Citation
Repeated nucleation of precipitates on dislocations in aluminum-copper

Material Information

Title:
Repeated nucleation of precipitates on dislocations in aluminum-copper
Creator:
Headley, Thomas Jeffrey, 1943-
Publication Date:
Language:
English
Physical Description:
viii, 208 leaves. : ill. ; 28 cm.

Subjects

Subjects / Keywords:
Alloys ( jstor )
Geometric planes ( jstor )
Mathematical vectors ( jstor )
Nucleation ( jstor )
Platelets ( jstor )
Precipitates ( jstor )
Precipitation ( jstor )
Quenching ( jstor )
Solutes ( jstor )
Stringers ( jstor )
Alloys -- Corrosion ( lcsh )
Aluminum-copper alloys ( lcsh )
Dislocations in metals ( lcsh )
Dissertations, Academic -- Materials Science and Engineering -- UF
Materials Science and Engineering thesis Ph. D
Nucleation ( lcsh )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Thesis:
Thesis--University of Florida.
Bibliography:
Bibliography: leaves 204-207.
Additional Physical Form:
Also available online.
General Note:
Typescript.
General Note:
Vita.
Statement of Responsibility:
Thomas Jeffrey Headley

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
0022775910 ( ALEPH )
14079047 ( OCLC )

Downloads

This item has the following downloads:


Full Text
















REPEATED NUCLEATION OF PRECIPITATES ON DISLOCATIONS IN ALUMINUM-COPPER







By



THOMAS JEFFREY HEADLEY










A DISSERTATION PRESENTED TO THE GRADUATE COUNCIL OF
THE UNIVERSITY OF FLORIDA
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE
DEGREE OF DOCTOR OF PHILOSOPHY



UNIVERSITY OF FLORIDA 1974
































Copyright by

Thomas Jeffrey Headley

1974






























Dedicated to my wife, Lynn















ACKNOW LED GMENTS



The author is deeply indebted to his Advisory Chairman, Dr. John J'. H-ren, who contributed unselfishly of his time,

and provided advice, encouragement, and stimulating discussion during the course of this research. He is indebted to his advisory committee for assistance, and to Dr. R. T. DeHoff

for many helpful discussions.

Thanks are due to Dr. R. W. Gould for providing alloy

materials; to the Sandia Corporation, Albuquerque, New Mexico,

for chemical analysis; to the Japan Electron Optics Laboratory, Boston, Massachusetts, for use of the JEOL 100B Electron Microscope; to Mr. E. J. Jenkins for assistance in the laboratory; to Mr. Paul Smith for assistance in the darkroom; and

to Mrs. Elizabeth Godey for typing this manuscript.

TLhe author's wife, Lynn, is acknowledged for her constant inspiration and support. His mother is acknowledged

for her lifelong encouragement.

Finally, the financial. support of the Atomic Energy Commission was deeply appreciated.















TABLE 0F CONTENTS


Page

ACKNOWLEDGMENTS........... .. .. .. .. .... iv

ABSTRACT.........................vii

CHAPTER

1INTRODUCTION ....................1

2 REVIEW OF THEORY AND PREVIOUS WORK .. .........5

2.1. Theory of Heterogenieous Nucleation
at Dislocations......... .......5
2.2. Precipitation in the Al-Cu System . .10 2.3. Dislocation Climb..............17
2.3.1. Quenched-in Vacancies and
the Chemical Climb Force......17 2.3.2. Theory of Dislocation Climb ...18 2.3-3. Dislocation Climb Sources ...22
2.4. Repeated Nucleation on Dislocations ...25
2.5. Pertinent Electron Microscopy Theory . .28
2.5.1. Two-Beam Diffraction Contrast
Theory *... .... ... 28
2.5.2. Defect Identification from
Invisibility Conditions .......35
2.5.3. Imaging Precipitates in the
Electron Microscope .........37

3 EXPERIMENTAL PROCEDURES AND MATERIALS .......41

S.1. Specimen Materials ..............41
5.2. Heat Treatments.......................42
3.3. Electron Microscope Specimen
Preparation..................4
5.4. Electron Microscopy.............48

4 EXPERIMENTAL RESULTS AND ANALYSES ..........5

4.1. Introduction ..............................5
4.2. Nature and Source of the Climbing
Dislocations .. .. .. .. .. .. .. ...3
4.2.1. Dislocation Climb Sources . .53
4.2.2. Glide Dislocations Which
Climb................70

v









TABLE OF C(INTENTS Continued


CH A f, TF R Page

4 (Continued)

4.3. Identification and Characterization
of the Precipi tate Phase .. ......... ...77
4.4. Further Geometric Analyses .. ........ .94
4.4.1. Distribution of Precipitates
in Colonies at Climb Sources . 94
4.4.2. Geometry of the Precipitate
Stringer';.' *.................. ..98
4.4.3. Determination of the Burgers Vectors of Small Loops Within Precipitate Colonies .. ....... ..106
4.4.4. "Secondary" Climb Sources ... 111 4.4.5. A Climb Source on (100) .....118
4.4.6. Nucleation of Preferred 0'
Orientations During Segmented Climb ..... .............. 122
4.4.7. Displacement Fringe Contrast
in a Precipitate Colony ..... 127
4.4.8. Precipitate Colonies Associated
with Subboundary Formation ... 132
4.5. Effects of Experimental Variables on
Microstructure ...... ............. ..135
4.5.1. Effect of Time at Constant
Aging Temperature .. ........ .135
4.5.2. Effect of Solution Treatment
Temperature ... ........... .140
4.5.3. Effect of Temperature to Which
Samples Are Direct-Quenched . 147 4.5.4. Effect of Quench Rate ... ...... 154
4.5.5. Effect of Copper Concentration 166
4.6. Summary ...... ................. .176

5 THE REPEATED NUCLEATION MECHANISM .. ....... ..178

5.1. Nucleation of 0' Near Edge Dislocations 178
5.2. Comparison with Previous Repeated
Nucleation Mechanisms ... .......... 182
5.3. The Mechanism in Al-Cu ... ......... .. 184
5.3.1. Local Solute Buildup .. ....... ..186
5.3.2. Precipitate Stringer Formation 190
5.4. Criteria for Repeated Nucleation in
Al-Cu and Application to Other Systems 196

6 CONCLUSIONS AND SUGGESTIONS FOR FUTURE WORK . 198

BIBLIOGRAPHY ......... ................... 204

BIOGRAPHICAL SKETCH ....... .................. .208
vi









Abstract of Dissertation Presented to the Graduate Council of the University of Florida Ln Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy


REPEATED NUCLEATION OF. PRECIPITATES
ON DISLOCATIONS !N ALUMINUM-COPPER


By

Thomas Je f rey Headley

August, 1974


Chai rman: John J. Hren
Major Department: Materials Science and. Engineering


Results are presented of an investigation of a newly

discovered -propagation mechanism for the ai-*e' transformation in Al-Cu: repeated nucleation on climbing dislocations. It was found that during the quench, dislocations are generated and climb by the annihilation of quenched-in vacancies. Densely populated colonies of e' precipitates nucleate in the stress fields of the climbing dislocations. In this way, the distribution of the entire volume fraction of e' is established during the quench.

The climbing dislocations were found to be a/2<110>

type, falling into three categories according to origin:

(1) pure-edge loops on {110) habits nucleated at dislocation climb sources, (2) glide dislocations initially on {111}, and (3) pure-edge loops on {llOl formed by the collapse of

vacancy clusters.

The effects of solution treatment temperature, aging

temperature, quench rate, and solute concentration on the vii










repeated nucleation process we re determined. It was found that repeated nucleation OCc-u-,s during quenching from all temperatures within the solid solution range1y, to all temperatures in the -range room temperature to 300'C. it occurs during slow and fast quenching as well, but does not occur in alloys with concentration Clwt.% Cu.

Mechanisms of repeated nucleation proposed earlier for other alloys are not applicable to Al-Cu. Dislocation climb and precipitation were found to be independently controlled

-processes. The relevant criteria for repeated nucleation in

this system are:

(1) a precipitate phase which nucleates easily on

dislocations,

(2) a source of dislocations during quenching,

(3) a driving force for dislocation climb which is

independent of the precipitation process, and

(4) a climb -rate slow enough to permit nucleation

but -rapid enough to avoid pinning.

It is suggested that pipe diffusion along the moving dislocation provides the necessary solute enhancement for successive nucleations.













1J.i i













CHAPTER 1

INTRODUCTION



Some of the most important strengthening mechanisms in alloys result from the precipitation of a second solid phase from a supersaturated solid solution. The age-hardening phenomenon in many aluminum alloys is a well-known example. A -precipitation reaction is a nucleation-and-growth transformation. Hence, the rate of the reaction is dependent -upon

(1) the nucleation rate of precipitates, and (2) the rate of their growth. If either or both of these rates is low the reaction rate will be low. Consequently, it is important to understand how and where precipitate reactions nucleate, apart from the problem of growth. Much is known about the kinetics of growth, but not about nucleation, especially heterogeneous nucleation.

Precipitate reactions nucleate either homogeneously or

heterogeneously within the matrix. If both the volume misfit and interfacial energy between precipitate and matrix are small, the reaction can nucleate homogeneously at random sites throughout the lattice. Homogeneous nucleation is known to, occur in only a few alloy systems, e.g., the precipitation of cobalt particles from dilute solutions of cobalt in copper (Servi and Turnbull, 1966). In most


1






2


precipitate reactions, either the volume misfit or interfacial energy, or both, is sufficiently large so that only heterogeneous nucleation occurs at preferred sites within the lattice. These sites are lattice defects such as grain boundaries,-dislocations, stacking faults, point defects, and other particles. Presumably, a portion of the energy

associated with the defect is supplied to help reduce the activation energy for formation of the critical nucleus, resulting in a nucleation event which is otherwise energetically unfavorable.

In the case of heterogeneous nucleation at dislocations, it is the dislocation strain energy in the matrix which helps overcome the barrier to nucleation. If the misfit strain caused by the precipitate is large, dislocation-nucleation may be the only method of decomposition of the supersaturated solid solution. Kelly and Nicholson (1963) and Nicholson (1970), have given excellent -reviews of the evidence for nucleation on dislocations in a number of alloy systems. A well-known example of heterogeneous nucleation on dislocations is that of the metastable 81 phase in Al-Cu alloys.

For a precipitate reaction which is dislocationnucleated, the following problem arises whenever the initial dislocation density is low, as is often the case following quenching. How can the reaction propagate once the available nucleation sites on dislocations have been saturated, i.e., what is the mechanism for propagation of the nucleation? Currently there are two known mechanisms whereby the reaction






3


may propagate. One mechanism is autocatalytic nucleation, first proposed by Lorimer (1968), and similar to the way in which martensite propagates. The initial precipitates nucleate on dislocations and grow into the matrix. In so doing, they generate stress fields in the matrix which aid in the nucleation of new precipitates. Thus, the reaction propagates in bands spreading out from the original dislocations to fill the lattice. Lorimer showed that the ca-*' reaction in Al-Cu could propagate by autocatalytic nucleation. Before

the present work, this was the only reported mechanism whereby the e' -reaction propagates from a low initial dislocation density.

Secondly, nucleation of the reaction can be propagated if the dislocation can somehow free itself from the initial

-precipitates and move away under a chemical or mechanical stress. It then presents fresh sites for the nucleation of

more precipitates. Nicholson (1970) was the first to use the term "repeated precipitation on dislocations" to describe this process. Repeated nucleation on climbing dislocations was first observed for carbide precipitation in austenitic stainless steel (Silcock and Tunstall, 1964). Since that time it has been reported for only a few other alloy systems. Very recently, Nes (1974) published a paper on the mechanism

for -repeated -precipitation on dislocations which he implied was universal with the statement that his model "can be applied to repeated precipitation (or colony growth) in any alloy system."






4


During experiments in which Al-Cu alloys were quenched directly to aging temperatures, this author observed that precipitation of the 01 phase occurred exclusively by repeated nucleation on climbing dislocations. Further examination revealed that the -repeated nucleation in this system could not be explained by the mechanism proposed by Nes (1974). Therefore, the primary purpose of this research was to establish the mechanism for repeated nucleation of 81 on climbing dislocations in Al-Cu, and in so doing, to determine if there

are aspects of the mechanism which might apply to precipitation in other alloy systems.














CHAPTER 2

REVIEW OF THEORY AND PREVIOUS WORK


2.1. Theory of Heterogeneous Nucleation
at Dislocations


Nucleation theory employs the concepts of a critical

nucleus and an activation energy for nucleation. An assumption of the theory is that random thermal fluctuations lead to the formation of small embryos of the transformed phase.

Embryos having a size and shape smaller than some critical size and shape will on the average disappear, and those with a larger size will grow and become stable. This critical size and shape is defined as the critical nucleus. The activation energy is the minimum energy barrier which must be overcome before nucleation can occur and turns out to be the free energy of the critical nucleus. This energy barrier is a maximum with respect to size and a minimum with respect to all other variables. The importance of the free energy of the critical nucleus can be recognized from its appearance in the equation for the steady state nucleation rate, which

is written in general form as

N s AG*
j = Z c exp 'RT (2.1)
s






6


where J =nucleation rate,

Z =the Zeldovich factor,

=frequency which atoms add to the nucleus,

N s= number of available nucleation sites,

cs= composition of the nucleus,

AG* = free energy of the critical nucleus,

k =the Boltzmann constant, and

T = temperature.

AG* has the form


AG* K surf) 2 (2.2)
(AGd-riv)T

where K =a shape factor,

surf particle/matrix interfacial energy, and
AG drive -driving force for the reaction.

For homogeneous nucleation, Ns is hig assuring reasonable nucleation -rates. For heterogeneous nucleation, N sis low

and the nucleation rate is usually dominated by the exponential dependence on the free energy of the critical nucleus.

For precipitation in solids, there is a chemical free

energy change per unit volume tending to drive the transformation to the new phase. If the atomic volumes in the matrix and particle are different, there is a misfit strain energy associated with formation of the new phase. Thus the free energy of the critical nucleus can be written


AG* = K( uf) 3
[AG(chem)+AG(strain)1






7


AG(chem) is negative whereas a surf and AG(strain) are positive. For a given reaction, AG(chem) can be calculated from thermodynamic parameters by the method of Aaronson et al. (1970). Attempts have been made to calculate surf from atomic bond models, but in general, the binding energies are unknown. In the absence of a proven model, surf is often taken as the measured bulk interfacial energy. The validity of this approximation is questionable. AG(strain) can be calculated by the method of Eshelby (1957,1961).

If a nucleus forms in the stress field of a dislocation, an additional energy term arises from the interaction of the displacement field of the precipitate with the stress field of the dislocation. We can then write AG* as


AG* = K(asurf)) 32
[AG(chem)+AG(strain)+AG(int)2"

AG(int) is negative and acts to reduce the positive AG(strain) term, so that it represents a major part of the advantage gained from nucleation at a dislocation. If AG(strain) is large (i.e., the precipitate misfit is large), nucleation at dislocations may be the only way the reaction can initiate. In addition to the AG(int) term, a second advantage for nucleation at a dislocation arises from the pre-exponential

term in Equation (2.1). This term is the frequency with

which atoms join the nucleus and depends on solute diffusion to the nucleus and across the interface. Solute pipediffusion along a dislocation core is always faster than






8


bulk diffusion, so that $ iic-reases for nucleation at a dislocation. Also, solute diffusion to dislocations themselves enhances the local concentration (e~g., Builough and Newman, 1959).

The task of calculating AG(int) is difficult, which is the main reason why the theory of heterogeneous nucleation at dislocations is less advanced than homogeneous nucleation theory. In fact, there have been only six published attempts to calculate AG* for nucleation at a dislocation. Cahn (19S7) made the first calculation. lie assumed an incoherent precipitate, an isotropic material, and completely neglected any interaction term. Despite these simplifications, his model

was able to predict qualitatively some experimental observations of nucleation at dislocations. Dollins (1970) calculated AG* for coherent, spherical and disk-shaped nuclei at a dislocation in an isotropic matrix. His work was reexamined by Barnett (1971). Lyubov and Solov'Yev (1965) have given the most complete treatment for calculating AG* for a coherent nucleus at a dislocation. Ram~rez and Pound (1973) attempted to include effects of the dislocation core energy on nucleation, effects that were omitted from the other models which use linear elasticity theory. An excellent recent -review of the present status of calculating AG* for nucleation at dislocations has been given by Larchd (1974). None of the above treatments, however, have included effects of elastic anisotropy, principally because the calculations involved are extremely difficult. It can be concluded that









the presence of the dislocation stress field aids in reducing the energy barrier to nucleation, but a rigorous calculation oF the el-fect is not yet available to provide an explicit expression for the rate of heterogeneous nucleation at dislocations.

It is instructive, however, to examine the order of magnitude or the terms in AG* to estimate the catalytic effect of the dislocation. Typical values of AG(chem) are in the range l-Sx109 ergs/cm3 (20-100 cal/cm 3). Values of AG(strain) are in the same range as AC(chem) for particles with appreciable misfit. Values of o obtained from bulk measurements are almost certainly too large since they relate to incoherent interfaces, whereas critical nuclei whose sizes
0
are of the order of 10's of Angstroms should have coherent interfaces. Estimates of coherent interfacial energies are
2
in the range 20-50 ergs/cm

If we take AG(chem) = 2x109 ergs/cm 3, surf = 20

ergs/cm2 and AG(strain) = 2x109 ergs/cm 3, then a spherical
0
nucleus with a diameter of 20A would have a chemical free
-13
energy change and strain energy = 8x10 ergs each, and a surface energy = 24x10-13 ergs. Due to the problems discussed above, no calculated values are available for AG(int),

but it is estimated that it can be of the same order of magnitude as AG(chem) and AG(strain). Then if one assumes that the surface energy is overestimated, as is likely (Barnett, 1973), the interaction energy due to the presence of the






10


dislocation can have an appreciable effect of lowering the total free energy of the critical nucleus, whenever AG(strain) is large.




2.2. Precipitation in the Al-Cu System


The aluminum-rich end of the Al-Cu phase diagram is a eutectic system between the aluminum solid solution and 8-CuAl (53 wt.% Cu). Figure 2.1 shows the portion of the diagram containing the a--solid solution region. Upon quenching from the solid solution region and aging, the equilibrium precipitate is the b.c.t. 0-phase. The existence of three metastable, transition phases, Guinier-Preston (G.P.) zones, 0" and 0', was established by the early x-ray work of Preston (1938a,b,c) on the 4 wt.% Cu alloy, and by Guinier (1938, 19391942,195Oi952) on the 4 and 5 wt.% Cu alloys. The solvus lines for these three transition phases are shown in Figure 2.1. The positions of the 0" and e, solvuses are due to Hornbogen (1967). That of the G.P. solvus is due to Beton and Rollason (1957).

Guinier (1938) and Preston (1938a) determined that the G.P. zones are coherent, copper-rich clusters of plate-like shape which form on {i00) planes of the matrix. The most reliable lattice parameters of the 0" and a' phases are given by Silcock et al. (1953). e" is complex tetragonal with a = 4.04X and c = 7.81. It is coherent with the matrix and








700


600
54 8'

500 +


U
0 400


+*
300 -+e


20- 6+ol"! SOLVUS
0+ C
E


SOLVUS
1 00

100 0+ G.P Zones

I

1 2 3 4 5 6 7
Al
Wt.0/o Cu


Figure 2.1. The aluminum-rich end of the Al-Cu phase
diagram, including the solvus lines for
G.P. zones, o", and e' precipitates.






12


forms as platelets on {1001 planes of the matrix. 0' is complex tetragonal with a = 4.04A and c = 5.8 It also forms as platelets parallel to {lOOI matrix planes, and is initially coherent on its broad faces and semi-coherent on its edge. As it grows its broad faces become semi-coherent. The orientation relationship for both 0" and 0' is {100} p.t {10 I lOmatrix and <100> ppt 11 <100> matrix' The tetragonal unit cell of e' is shown in Figure 2.2. There are 6 atoms/unit cell. The a-solid solution is f.c.c. with a = 4.0 45. and has 4 atoms/unit cell. When the atomic volumes are calculated for these two unit cells and compared, it is found that the o.-*0' transformation involves a 3.95% volume contraction. The resulting transformation strain can be partially compensated if vacancies are generated by the growing precipitates and supplied to the matrix.

The early x-ray work established the following precipitation sequence for quenching and aging below the G.P. solvus:

G.P. zones V" - el -+ e(CuAl 2).

However, as suggested by the x-ray work and later confirmed by many transmission electron microscope (TEM) studies, several of the reactions can proceed concurrently depending on the quenching and aging procedures. In addition, TEM investigations have clearly established the homogeneous or heterogeneous nature of the various reactions. Nicholson and Nutting (1958) -resolved G.P. zones and 0" platelets in the 4 wt.% Cu alloy and found them to be homogeneously distributed in the matrix. It is now clear, however, that 0"





13



(100)AI (OOl) e' /1!
/ 1,






5.8 A









O Aluminum Atoms Copper Atoms Figure 2.2. The tetragonal unit cell of 6' (after
Silcock, Heal, and Hardy, 1953).






14


must be nucleated on G.P. zones (Lorimer and Nicholson, 1969; Lorimer, 1970). If a sample is direct-quenched into the region below the e solvus but above the G.P. solvus (Figure 2.1), no 0" forms. However, if a sample is direct-quenched below the G.P. solvus and then up-quenched into this region and aged, 6" forms and its distribution is a function of the size distribution of G.P. ones present before the sample was up-quenched (Lorimer, 1970). Neither G.P. zones nor e" plays a role in the nucleation of 9'. Due to its misfit strain, 0' nucleates only heterogeneously in the presence of a stress field in the lattice. It nucleates either at dislocations (Nicholson and Nutting, 1958), in the stress fields of other 0' precipitates (Lorimer, 1968), or in the presence of a macroscopic stress applied to the sample during aging

(IHosford and Agrawal, 1974).

Numerous TEM investigations have confirmed the catalytic effect of dislocations for nucleating 6'. It was suggested early (Wilsdorf and Kuhlmarn-Wilsdorf, 1955; Thomas and Nutting, 1956), and later confirmed by TEM, that only certain 6' orientations will nucleate at a given dislocation. This is explained in terms of the misfit strain of the 8' platelet and the Burgers vector of the dislocation. In Figure 2.3, it is shown that the principal misfit around a 6' platelet is normal to the {100} plane of the platelet. In Figure 2.4, it is shown that a dislocation with Burgers vector a/2[110]

partially relieves the misfit strain around e' platelets on (100) and (010) whose misfits lie at 450 to the Burgers vector,











I



/ DIRECTION
0 OFMISFIT
I! !

PLATELET
a= b=4.04A c=5.8 aAl-Cu MATRIX
a: b: c: = 4.04A




Figure 2.3. Diagram showing that the distortion of (001)
planes around a 8' platelet is normal to the
platelet (not to scale).





LoTO] [100] COMPRESSION

MISFIT MISFIT [0
N*,/ MATRIX

G' on (010) f on (100) ORIENTATION


b=a/2[110]

TENSION



Figure 2.4. Diagram showing that a dislocation with
Burgers vector a/2[l10] relieves the misfit
around e' platelets lying on (100) and (010).
It does not relieve the misfit around a
platelet on (001).






16


but not for a 01 platelet on (001) whose misfit vector is perpendicular to the Burgers vector. Hence, the (001) orientation gains no advantage by nucleating in the stress field

of the dislocation. Likewise for a dislocation with F=a[1001, only the (100) orientation of el should nucleate in its stress field.

Many early TEM investigations of the 81 phase were conducted after long aging treatments at high temperatures in the a+e' field (Figure 2.1). The resulting microstructures contained a uniform distribution of large 81 platelets, and it was initially concluded that these 61 platelets were nucleated by a -random distribution of pre-existing 8". However, as it became clear that 81 nucleates at dislocations and not at e", the problem of how the -random distribution of e' could form by quenching and aging alone remained unresolved until the work of Lorimer (1968,1970). Lorimer showed that the a-*el reaction could propagate from an initially low dislocation density, introduced during quenching, by an autocatalytic nucleation mechanism. Early during the aging period, the initial dislocations become saturated with 0'. These platelets then grow into the matrix and produce their own stress fields which aid the nucleation of more precipitates. With long aging, the reaction propagates in bands spreading out from the dislocations to fill the structure with a uniform distribution of 0' platelets on all three {100} orientations. Until the present research, this was the only






17


-reported mechanism whereby the cx*e, -reaction was found to propagate from a low initial dislocation density.





2.3. Dislocation Climb

2.3.1. quenched-In Vacancies and the
Chemical Climb Force

It is now widely accepted that vacancies can exist in

crystals in thermal equilibrium with the lattice. The equilibrium concentration of vacancies increases exponentially with temperature according to the Arrhenius relation:


C 0 = A exp(-E f/kT)


where A =an entropy factor,

E f the activation energy for forming a vacancy,

k =the Boltzmann constant, and

T =temperature.

Large supersaturations of vacancies can be retained in the lattice by quenching rapidly from elevated temperatures. During and after the quench, the excess vacancies diffuse to

sinks such as surfaces, grain boundaries, and dislocations, or they may cluster and collapse into vacancy disks bounded by dislocation loops. The condensation of vacancies onto a dislocation causes it to undergo positive climb. The greater the supersaturation of vacancies, the greater is the driving force for climb. A simple picture of dislocation climb by vacancy annihilation is shown in Figure 2.5(a). Dislocations






18


can also climb by vacancy-emission (negative climb) and this is illustrated in Figure 2.5(b).

Christian (1965) has suggested that the vacancyannihilating climb of a/2 dislocations in f.c.c. lattices occurs easily only on the {111} and {l10} planes. A necessary condition for climb is that the Burgers vector has a component perpendicular to the plane of climb. Therefore there are two {lll} and five {l10} "planes of easy climb" for an a/2[l10] dislocation in f.c.c. Miekk-oja and Raty (1971) have considered the choice of climb planes in terms of the chemical climb force on each plane. This force arises from the supersaturation (or subsaturation) of vacancies (Bardeen and Herring, 1952). Its magnitude is proportional to (-xu), where Iis the Burgers vector and u is the dislocation line direction. Thus, according to Miekk-oja and Raty, a dislocation with F a/2[l10] is affected by the maximum climb force, F c,max'
on the {1101 plane perpendicular to F-. It is not affected at all on the {111} and {10} planes containing b. And it is affected by forces 0.82F and O.5F on the two {lll}
c, max c, max
and four {110} planes, respectively, which are inclined to E.


2.3.2. Theory of Dislocation Climb
In reality, dislocation climb is more complex than the

simple picture envisioned in Figure 2.5. The theory of climb

has been developed by Lothe (1960), Thomson and Balluffi (1962), Balluffi and Thomson (1962), Friedel (1964), Hirth





19






OOO00 O000000 000@000 OOO OO OO OO0OO OO~Oo oooooo0 1000000 0Q00O0 0000000 000000 000000 OOO OOO0 OOO OO OOO OOO0
OOO OO OO oOO OOO OOO
OOOOOO 000000 000000 000000 000000 000000
a




0000000 0000000 000@000 000@000 0000000 0000000

oooooo ooo ooo ooooOO ooo-qoo 00o0&00 0000 0 000000 OOOOOO 000000 000000 000000 000000
b



Figure 2.5. (a) Positive climb of an edge dislocation
by vacancy annihilation. (b) Negative climb by vacancy emission (after Reed-Hill, 1973).









and Lothe (1968) and a general review of the mi-echanisms has been given by Balluffi (196)).

Briefly, dislocation cLirib occurs by (i) the absorption of vacancies onto the dislocation core, (2) diffusion of the vacancies along the core to jogs, and (3) subsequent movement of the jogs by destruction of the vacancies. This sequence is illustrated in Figure 2.6 for climb of an undissociated edge dislocation (a similar model applies for climb by vacancy-emission). Then, according to Balluffi (1969), the dislocation climb velocity is

2 Dlb2 [c (R) -c]
V =

[kn(4I) A 7)

where D = vacancy diffusivity in the lattice,

b = magnitude of the Burgers vector,

c(R) = vacancy supersaturation at a large distance
R from the dislocation,

c = vacancy concentration maintained in the
lattice in equilibrium with the jogs,

z = mean migration distance of a vacancy along
the core before jumping off, and

= jog spacing.

Seidman and Balluffi (1968) surveyed the available experimental data on climb rates and concluded that, in the presence of moderate to large supersaturations, climb in aluminum appears to be highly efficient. In other words, jog production and motion is sufficiently fast that the climb rate is limited only by the diffusion of vacancies to the dislocation, and the dislocation acts as a perfect line sink.






21














VACANCIES

a






















C



Figure 2.6. Diagram of vacancy processes associated with
climb of the extra half-plane of an edge
dislocation. Vacancies absorb onto the core
(a) diffuse along the core (b) and annihilate
at jogs (c). Subsequent motion of the jog across the page moves the extra half-plane
up one atomic spacing.






22


2.3.3. Dislocation Climb Sources

The classical dislocation multiplication mechanism is

that proposed by Frank and Read (1950), whereby a dislocation, pinned at each end, expands in its slip plane by glide. Westmacott et al. (1959) observed dislocation sources in thin foils of Al-4 wt.% Cu by transmission electron microscopy, and interpreted them to be Frank-Read sources. Gulden and Nix (1968) have observed similar sources in Al-4 wt.% Cu3 wt.% Si. Analogous to the Frank-Read mechanism, a dislocation multiplication mechanism which operates by climb was

proposed by Bardeen and Herring (1952) to act as a continuous sink for excess vacancies. The Bardeen-Herring model for a dislocation climb source is shown in Figure 2.7. Initially, a straight dislocation between A and B has its slip plane normal to the plane of the paper. Hence, it can move in the plane of the paper only by climb. Condensation of vacancies onto this line would move the dislocation out through the sequence of positions shown. When the bottom segments of the loop meet, they annihilate and rejoin as shown by the dotted lines. The segment ABC is now free to repeat the process,

and there is left a vacancy loop outlined by the dislocation ring. As long as there remains a vacancy supersaturation in the region, this loop will expand, removing atoms from the lattice, and the operation can repeat removing an indefinite number of planes.

Dislocation climb source configurations were first observed by TEM in Al-Mg alloys (Westmacott et al. 1962;






23










































Figure 2.7. The Bardeen-Herring model of a dislocation
climb source. An edge dislocation, pinned
between AB, has its slip plane normal to the paper. It climbs in the plane of the paper, by vacancy annihilation, through the successive positions 1-4, rejoining at the bottom.
The portion ACB can then repeat the process
(after Bardeen and Herring, 1952).






24


Pmbury and Nicholson, 1963). Since then, dislocation climb sources have been identified in aluminum (Ldington and West, 1966) and a number of other aluminum alloys, including Al-Ag (Edington and West, 1966), Al-Ag ternary alloys (Passoja and Ansell, 1971), and Al-Cu (Boyd and Edington, 1971). They have also been observed in other systems, including silicon (Ravi, 1971) and NiAI (Marshall and Brittain, 1974). Often the dislocation climb sources observed by TEM have small particles at the center of the source loop. An example from the present work is shown in Figure 4.2. Although the particles are usually too snail to be identified, it is thought that they are insoluble particles existing at the solution treatment temperature. It is generally believed that vacancies diffusing to the particle interface activate the source which then operates to produce successive loops. The source itself is often thought to be a portion of a misfit dislocation at the particle/matrix interface.

The only reported observation of dislocation climb

sources in binary Al-Cu alloys is that of Boyd and Edington (1971). They observed source densities of about 3/grain. These sources generated pure-edge loops on {1101 habits with a/2 Burgers vectors.






25


2.4. Repeated Nucleation on Dislocations


The concept of repeated precipitation on climbing dislocations was first proposed by Silcock and Tunstall (1964) to explain the occurrence of planar colonies of NbC precipitates on stacking faults in austenitic stainless steels. In connection with the precipitate reaction, the stacking faults were found to grow by the climb of a/3<111> Frank partial dislocations bounding the fault. The transformation

to the NbC -phase involves a 23% volume expansion, so that the growing precipitates consume vacancies from the matrix in order to relieve the transformation strains. Silcock and Tunstall proposed that the Frank partial climbs by vacancyemission in order to feed vacancies to the transformation. Thus the driving force for the dislocation climb is this need to supply vacancies for the precipitate reaction. The principles of the Silcock-Tunstall model are outlined in Figure 2.8.

Repeated precipitation on climbing dislocations by this

mechanism has since been reported in a variety of systems, including different steels, an iron-vanadium alloy, a coppersilver alloy, superalloys, and semiconducting materials. The

phenomenon has been observed to occur on both partial and total dislocations. For the sake of brevity, the list of reports will not be given here, and the reader is referred to the complete list in the recent paper by Nes (1974). In every reported case to date, the precipitate phase has a larger atomic volume than the matrix, thereby consuming






26














XF


\\INbC~










a b C d e










Figure 2.8. The Silcock- Tuns tall model for repeated precipitation of NbC in austenitic stainless
steel. The precipitates nucleate on Frank
partials (a). Movement of jogs, J, provides
vacancies for the precipitates to grow (b)
with the consequent climb of the dislocation (c). The dislocation pinches off (d) and the
process repeats (e) (after Silcock and Tunstall,
1964).






27

vacancies during the transformation. In every case the dislocation climb has been reported to be vacancy-emitting in order to supply the necessary vacancies.

Nes (1974) has expanded the original model of Silcock and Tunstall (1964) into a more quantitative theory, which was intended to account for the various features of repeated precipitation in all the systems reported since 1964. The

fundamentals of the Nes theory are:

(1) Vacancies must be supplied to the transforming

particles in order to reduce the particle/matrix

mismatch.

(2) The subsequent particle growth causes vacancyemitting climb of the dislocation in order to

feed the transformation.

(3) The particle growth/dislocation climb sequence

between conservative nucleations is controlled

by balancing the rate at which vacancies must be supplied to the precipitates with the climb

rate of the dislocation.

(4) The particle is dragged some distance by the

dislocation before unpinning occurs.

(S) The rate controlling parameters in the kinetics

of colony growth are either (a) the atomic diffusion of the precipitating atoms, or (b) the core (interface) self-diffusion, depending on

which has the highest activation energy.






28


Nes implied that this mechanism is applicable to repeated precipitation on climbing dislocations in all systems, whereas in reality, it probably applies only when there is required a mass balance of vacancies between growing precipitates and

climbing dislocations.

There is one report of repeated precipitation on climbing dislocations which has not been attributed to the above mechanism. Embury (1963) observed that dislocations in Al-Mg alloys were drawn around M92 A 13 precipitate particles, pinched off leaving loops, and climbed away under the chemical force of a quenched-in vacancy supersaturation where the process repeated. However, this process is reported to occur only to a small extent.





2.S. Pertinent Electron Microscopy Theory


2.S.1. Two-Beam Diffraction Contrast Theory

In the transmission electron microscope, contrast arises from differences in intensity scattered out of the incident electron beam by Bragg diffraction from the crystal planes. The best contrast from defects occurs under so-called "twobeam" conditions. Owing to the large amount of tilt available in commercial goniometer stages, the crystal can be oriented so that the incident beam diffracts strongly only from one set of lattice planes. Then approximately 95% or more of the incident intensity is contained either in the






29


beam scattered in the forward direction (called the "transmitted" or "main" beam), or in the strong diffracted beam. The electron image is usually formed by placing an aperture around one of these beams and allowing it to pass through, while the other beam is blocked by the aperture holder (Figure 2.9). The unblocked beam is then magnified by successive lenses and projected onto the fluorescent screen. When the aperture is placed around the transmitted beam, the image formed is called a "bright-field" image. When it is placed around the diffracted beam, the image is called a "dayk-field" image. Contrast at defects arises in, say, the bright-field image because the strain in the lattice around the defect causes local distortions in the atomic planes which lead to local changes in the intensity scattered into the diffracted beam. This in turn leads to local variations in intensity in the bright-field image. This is illustrated for the case of a dislocation in Figure 2.10.

When a crystal of sufficient thickness is oriented very close to the Bragg condition for one set of planes, there can be a dynamic interchange of electrons between the two beams,

resulting from multiple scattering back and forth as the two beams pass through the crystal (Figure 2.11). In order to predict the intensities in the bright- or dark-field image, it is necessary to describe mathematically the physical processes which go on in this dynamic interchange. The two-beam dynamical theory of electron diffraction for a distorted crystal was developed by Howie and Whelan (1961). Their






30









INCIDENT BEAM










CRYSTAL



TRANSMITTED DIFFRACTE C

\D11
BEAM if BEAM


-OBJECTIVE
LENS PLANE





OBJECTIVE APERTURE

HOLDER












Figure 2.9. Method for forming a bright-field image under
two-beam conditions. The transmitted beam is
allowed to pass through the objective aperture
while the diffracted beam is blocked by the
aperture holder.






31





INCIDENT BEAM
















A All X






DIFFRACTED INTENSITY








TRANSMITTED INTENSITY




Figure 2.10. Diagram illustrating how contrast a-rises in
the transmitted and diffracted beams from
diffraction off the distorted planes around an edge dislocation. Planes to the left of the dislocation are tilted toward the Bragg angle. Planes to the right are tilted away
from the Bragg angle.







32











INCIDENT



BEAM












FOIL
THIC KNESS DIFFRACTING PLANES










DIFFRACTED
BEAM

TRANSMITTED BEAM








Figure 2.11. Diagram illustrating the dynamic interchange
of electron intensity between the two beams
resulting from multiple scattering events.






33


derivation is similar to the derivation of a two-beam theory for x-ray diffraction by Darwin (1914). The Howie-Whelan theory predicts the amplitudes T and S of the electron waves in the transmitted and scattered beams, respectively, at any point in the crystal. The formulation of the theory uses a column approximation, whereby the crystal is imagined to be divided up into parallel columns in the direction of the incident beam. Dynamic interchange between T and S is considered within a column, but not between neighboring columns. This is a valid approximation because the Bragg angles for high energy electron diffraction are small (-l/2).

The Howie-Whelan theory predicts the following coupled

pair of first-order differential equations for the variations in T and S with depth in the crystal:

dT = (Tri/ )T + (Ti/C )S exp(27risz+2Tig'*)


dS (ii/ o S + (Tri/ )T exp(-27isz-27ig.R)


where T = amplitude of the incident beam,

S = amplitude of the diffracted beam,

z = depth in the crystal in the direction
of the incident beam,

s = parameter measuring deviation from the
Bragg condition,

= the diffracting vector,

= the local displacement field at depth z,

o= parameter related to mean refractive index
of the crystal, and

= the extinction distance.
g






34


Each separate equation represents the variation in amplitude of the electron waves in that beam as it passes through the crystal. The first term in each equation represents the intensity scattered in the forward direction for that beam. The second term in each equation represents the intensity scattered into that beam from the other beam.

In order to account for experimentally observed effects

of absorption, it is necessary to replace the quantities l/ and l/Cg in the equations by the complex quantities (co+i/Eo') and (1/ +i/), respectively. One then obtains:

dT = ri(l/ o+i/C')T + Tri(l/ +i/Yg)S exp(2Tisz+27ig'7)
a T o g 9(~s+n~

dS= (l/ +i/E')S + rilWE +i/YT exp (-2risz -2'igR)


Multiplying the amplitudes T and S by their complex conjugates gives the relative intensities in the two beams at any point in the crystal. In particular, when the intensity is calculated at the bottom of all imaginary columns in the foil, it predicts the image projected onto the viewing screen, since no interaction occurs in vacuum once the beams exit the crystal.

In the absence of any displacement field (7=0), or in the presence of a fixed, rigid body displacement (7=constant), the equations can be solved analytically for T or S, and the solution predicts a uniform intensity over the bottom of the crystal. When the displacement field R7 varies with depth, as is the case around dislocations and other defects, the equations can no longer be solved analytically, and numerical

methods must be used to obtain T and S.






35


The validity of the Howie-Whelan equations in predicting intensities which correspond to two-beam images has been

overwhelmingly demonstrated by the success of computer simulation techniques for matching defect images (Head et al., 1973). (See for example Section 4.2.1.)


2.S.2. Defect Identification from
Invisibility Conditions

Although the solution of the two-beam equations is not straightforward for the case of defects with varying displacement fields, often the solution pfr se is not needed to identify the defect from its image. Instead, it is often possible to apply a simple criterion to identify defects in the electron microscope. This criterion is based on the fact that the term in the equations which gives rise to contrast is the product 9-7. The diffraction vector g is the reciprocal lattice vector normal to the diffracting planes,

so that the product -R samples the magnitude of the distortion created in the diffracting planes by the displacement field 7. If a defect happens to cause no distortion in the diffracting planes for a two-beam condition, then g-R=0 and the two-beam equations predict uniform intensity everywhere at the bottom of the foil. In other words, there is no contrast around the defect and it is said to be "invisible" for this diffraction condition. This criterion, applied to the identification of dislocations, can be described as follows. To a first approximation, the planes -parallel to the Burgers vector of a dislocation in an isotropic crystal are not






36


distorted. Then when the crystal is oriented so that one such set of planes is in the reflecting position, the dislocation will be "invisible" in the image. The diffracting vector is perpendicular to the diffracting planes, and therefore to the Burgers vector for this condition. Hence, the criterion for invisibility of a dislocation is the wellknown relation j.K-=O. To identify the Burgers vector of a dislocation, it is simply a matter of tilting the foil and selecting various two-beam conditions until two diffraction vectors, g, and 92,are found for which the dislocation is invisible in the bright-field image. The Burgers vector must be perpendicular to both g, and 92 so that it can be determined from their cross product, i.e., E-=( 9 x9. However, this technique is not capable of determining the Burgers vector unambiguously, i.e. whether it is +5 or -5b. Furthermore, the criterion B_=O for invisibility applies only to screw dislocations where, in the isotropic approximation, all sets of planes parallel to the Burgers vector are undistorted. This is not so for an edge dislocation. For a set of planes to remain undistorted by an edge dislocation, not only must 9-5=O, but in addition, g must be parallel to the dislocation line direction. Mathematically this is written 9-(ExY)=O, where a is the line direction. This is a very stringent condition which is seldom obtained in the microscope. Thus, edge dislocations, or dislocations with appreciable edge orientation, often exhibit strong "residual contrast" when -B_=O, due to the (Fxii) term. For






37


this reason, practical experience in recognizing "residual contrast" is necessary in order to identify dislocations from the invisibility criterion.

The criterion 9-57=0 for invisibility is valid only for total dislocations, where the product 9-F can be only zero or an integer (since it is the product of a reciprocal lattice vector and a real lattice vector). For partial dislocations, T-Fcan take on the non-integer values 1/3, 2/3, 4/3, etc., in cubic lattices. Howie and Whelan (1962) determined that partial dislocations are invisible when jT-b_=O or 1/3 and are visible for all other products. Silcock and Tunstall (1964) further determined that, for this to be strictly valid, the deviation from the Bragg condition cannot be too large.

The condition that a defect is "invisible" if its displacement field does not distort the reflecting planes can be applied to identify certain small precipitates. For example, in the case of el platelets in Al-Cu, the principal misfit in the lattice caused by the platelet is normal to the plane of the platelet (Section 2.2). If such platelets are too small to distinguish their shape, their orientation can still be determined since they will be invisible whenever g is perpendicular to the misfit vector.


2.5.3. Imaging Precipitates in the
Electron Microscope

Precipitates can be imaged by one or more of several mechanisms in the electron microscope. A good description






38


of these mechanisms is given by Hirsch et al. (1965, p. 336). Those pertinent to this research will be outlined below.

(1) Strain contrast in the matrix. All coherent and

semicoherent precipitates, and most incoherent precipitates, cause some strain in the matrix. These strain fields therefore give rise to diffraction contrast effects in the matrix.

This can be the only mechanism for imaging very small precipitates whose sizes are less than the resolution limit of the microscope, but whose long-range strain fields are

greater than this limit.

(2) Misfit dislocation imaging. Semicoherent precipitates have misfit dislocations over their semicoherent interfaces. The strain fields of these misfit dislocations can cause strain contrast just as for isolated dislocations in the matrix. Weatherly and Nicholson (1968) have investigated the conditions for imaging misfit dislocations. Often small platelets viewed normal to the platelets are imaged by the misfit-dislocation loops around their edges. This is referred to as "dislocation-ring" contrast.

(3) Structure factor contrast. According to Ashby and Brown (1963), this contrast arises whenever a coherent precipitate has a different structure factor from the matrix, and thus a different extinction distance. A particle of thickness Lt then increases the effective foil thickness in columns -passing through the particle, giving rise to an intensity change relative to columns in the matrix. Depending on the depth of the particles in the foil and the






39


relative values of the extinction distances in the particle and matrix, the particles can appear either lighter or barker than the surrounding matrix. Structure factor contrast arises only within the limits of the particle boundary.

(4) Orientation contrast. This contrast mechanism

arises whenever a foil is oriented such that a certain set of lattice -planes in the precipitate is diffracting strongly, whereas the matrix is diffracting weakly, or vice versa. The contrast is of a uniform light and dark nature, typically dark precipitates in a light matrix. Orientation contrast can arise only when there is appreciable difference in crystal structure between the precipitate and matrix, i.e., when the precipitates are semicoherent or incoherent. For example, when the electron beam is parallel to the thin dimension of large precipitate platelets, often certain lattice planes in the precipitate will also be parallel to the beam. In this case, the precipitate diffracts strongly. If the matrix is not oriented for strong Bragg diffraction, the bright-field image will show dark precipitates in a light matrix.

(5) Displacement fringe contrast. Displacement fringe

contrast arises when there is an abrupt change in the phase of the transmitted and diffracted waves as they encounter a thin sheet of precipitate which displaces the matrix planes in opposite directions on either side of it. This displacement IT around a typical semicoherent precipitate platelet is normal to the plane of the platelet, and its magnitude is given by






40


1 R A I nIF n I


where At = thickness of the platelet,

6 = precipitate misfit,

n = number of misfit dislocations at the
periphery of the platelet, and

= Burgers vector of the misfit dislocations.

When this displacement is substituted into the equations of the dynamical theory, the intensity of the transmitted beam is found to oscillate with thickness (Whelan and Hirsch, 1957). Thus, when the precipitate platelet is inclined to the electron beam, a fringe effect is observed. The socalled stacking fault fringes are the limiting case of displacement fringe contrast.














CHAPTER 3

EXPERIMENTAL PROCEDURES AND MATERIALS



3.1. Specimen Materials


The four Al-Cu alloys used in this work were obtained

as rolled sheets from a previous research project. They were prepared from 99.99% aluminum and 99.99% copper by double melting in an induction furnace using a graphite mold. After solidification, the billets were alternately cold-rolled and

annealed to reduce them to sheet form.

The target compositions were the 4, 2) 1, and 1/2 wt.% Cu alloys. The nominal copper concentrations of the four alloys were 3.85, 1.96, 0.99, and 0.5 wt.%, based on starting weights before melting. The impurity content in the 3.85 wt.% Cu alloy was determined by x-ray spectrographic analysis by the Sandia Corporation, Albuquerque, New Mexico. The impurity levels are given as ranges in Table 3.1. The barium level is suspect as it was determined from only one line. The impurity levels in the other three alloys were not determined. However, since all four alloys were prepared from the same starting materials, the other three probably had the same impurity levels as the 3.8S wt.% Cu alloy.

The 1 wt.% and 1/2 wt.% Cu alloys were obtained as

rolled sheet, 0.038 inch and 0.034 inch, respectively. They 41






42


Table 3.1

Impurity Levels in the A1-3.85 wt.% Copper Alloy


I purity Weight ppm

Fe 5-25

Pb <10

Si 1-10

Mo <10

Mg 5-20

Ca 1-5

Ga <10

Ba 10-40




were then cold-rolled to 0.005 inch for heat treatment. The

3.85 wt.% and 1.96 wt.% Cu'alloys were obtained as rolled sheet, 0.004 inch thick. They were not reduced further before heat treatment. Samples for heat treatment were cut from the rolled foils to the approximate dimensions 1/8 x 1/2 x 0.004 inch. This was found to be a suitable size for preparing electron microscope specimens after heat treatment.





3.2. Heat Treatments


All samples were solution treated for one hour at a

temperature in the a-solid solution range (Figure 2.1). Next, they were either (1) direct-quenched to an aging temperature






43


above the G.P. solves, or (2) quenched to a low temperature. The samples given direct-quenches were aged for various times from approximately one second to 24 hours, and then quenched into room-temperature water. Samples quenched to low temperatures were either prepared for electron microscopy without further treatment,, or they were up-quenched to a temperature above the G.P. solves and aged for various times. They were then quenched into room-temperature water.

The solution treatments were conducted in a vertical furnace in air. The temperature in the heat zone was controlled to within 2'C. One end of the sample was clamped in a stainless steel alligator clip attached to the bottom of a one-half inch diameter stainless steel tube, and this was inserted into the heat zone of the furnace. Before each treatment, the temperature in the heat zone was determined by inserting a thermocouple into a dummy stainless steel tube

suspended in the heat zone.

Quenching was achieved by dropping the specimen-stainless steel tube assembly out of the bottom of the furnace into the quench bath. For direct-quenches to the aging temperature, the specimen was dropped into a Lauda Constant Temperature Oil Bath, maintained at the aging temperature, to within 0.2'C. A schematic diagram of the apparatus used for solution treatment and direct-quenching is shown in Figure 3.1. For quenching to low temperatures, the specimen was dropped into one of several low temperature baths in place of the oil bath. Following the quench to low temperatures, some






44





CLAMP -STAINLESS STEEL TUBE




INSULATION- VERTICAL
FURNACE




ALUM IN UM -SPECIMEN
BLOCKALLIGATOR CLIP









DROP 18INCHES
QUENCH MOTOR





CONSTANT
OIL TEMPERATURE
dub OIL BATH
HEATER STIRRER






Figure 3.1. Diagram of the apparatus used for solution
treatment and direct-quenching. The specimen,
attached to the stainless steel tube for weight,
is dropped from the solution treatment furnace
into the constant temperature, aging bath.






45


sampDles were immediately up-quenched into the constant temperature oil bath maintained at the desired aging temperature. Due to the wide variety of solution treatment temperatures, aging temperatures, and aging times employed in this research, no table of heat treatments will be given here. Instead, the specific heat treatment information will be given either in the text or in the figure captions.





3.3. Electron Microscope Specimen Preparation


Electron microscope specimens were prepared from the

heat treated strips by electropolishing in a solution of 5% perchloric acid in methyl alcohol. A polishing potential of 18 volts d.c. was used with a stainless steel cathode. The electropolishing setup is shown in Figure 3.2. The beaker containing the polishing solution was immersed in a bath of dry ice and acetone to slow down the polishing reaction. The solution was circulated at a slow speed with a magnetic stirrer to keep it cold. Under these conditions, the polishing bath was maintained at -45'C.

The specimen strip was held with locking tweezers and polished by dipping the bottom end (approximately 1/8-3/16 inch) into the solution at a dipping rate of about 1/second. Dipping was found to reduce edge attack and to give a relatively uniform polish. The voltage dropped to about 12 volts during immersion. Total polishing time to obtain a suitable







46










18B VOLTS d.c. TWEEZERS











50/. PE RC HLO RIC SEIE
ACID IN METHANOL
!STAINLESS STEEL CATHODE (2 HALVES)














MAGNETIC
STIRRER









Figure 3.2. The electropolishing setup for preparing thin
foils for electron microscopy. Polishing is
accomplished by dipping the bottom end of the
specimen into the solution.






47

thin area was about 1S minutes. When it was determined (by experience) that the specimen was nearly thin enough, the stirring was stopped to avoid damaging the thin area. The

last 20-30 seconds of polishing was done by immersion and agitation instead of dipping, since dipping to the last was found to sometimes etch the thin foil edge. Polishing was continued until the bottom edge appeared very ragged or until small holes had broken through. Then the power was switched off and the specimen was rapidly removed and plunged into a beaker of cold acetone (99.8% pure) immersed in the dry iceacetone cooling bath. It was agitated for about 10 seconds and then immediately placed under a stream of room-temperature

acetone from a wash bottle. After washing for about 30 seconds, it was allowed to dry in air. The initial wash in cold acetone was necessary to remove most of the electrolyte which rapidly etched the polished surface if allowed to warm to room temperature.

The thinned, bottom portion of the sample strip was cut

off with an X-acto razor knife and mounted in a 3 mm, 7S mesh, locking, double copper grid for viewing in the electron microscope. Several more specimens could then be polished from the sane sample strip, if desired. However, the remaining bottom part of the strip was already polished quite thin. To avoid etching this polished surface, the specimen was agitated for about one minute in the small beaker of cold acetone prior to repolishing.






48

3.4. Electron Microscopy


The thin foils were examined in a Phillips EM 200 electron microscope operated at 100 Kv potential. A goniometer stage with 45' and 300 tilt on two orthogonal axes was used.

Two-beam diffraction conditions were established for taking all micrographs. To obtain two-beam conditions, the foils were oriented close to one of the low index poles shown on the Kikuchi line map for an f.c.c. crystal in Figure 3.3. Use of this Kikuchi line map during specimen tilting, as described by Head et al. (1973), enabled diffraction vectors

to be determined unambiguously in every case.






49













10 0 D200 --0












13T








Figure 3.3. Kikuchi line map over two adjacent stereographic triangles for a face-centered cubic
crystal (after Head et al., 1973).














cJii'PTR 4

EXPERIMENTAL RESULTS AND ANALYSES



4.1. Tvatroduction


Evidence for repeated nucleation of the 6' phase on dislocations was first observed in this research when the Al-3.85 wt.% Cu alloy was quenched directly to aging temperatures above the 6"' solvus. Figure 4.1 shows a typical inicrostructure resulting from direct-quenching and aging.

A brief description of the features and evolution of this microstructure (with the facts to be established in this chapter) is as follows. Dispersed throughout the foil are densely populated colonies of small e' precipitates. The colonies are bounded either totally or partially by dislocations, some of which are out of contrast in this image. The dislocations were generated and climbed during the quench From the solution treatment temperature. As they climbed, they nucleated and dispersed the e, colonies in their paths. All dislocations climbed during quenching and all nucleated precipitate colonies. The precipitate colonies may (1) be planar, (2) lie on smoothly curved surfaces, or (3) lie on corrugated-shaped surfaces, depending on the climb paths of the dislocations.



50






Si







01

16


























1A4










Figure 4.1. Typical microstructure resulting from quenching the Al-4 wt.% Cu alloy directly to the

of mal 8'precipitates, bounded by dislocation. (eattreatment: S.T. 1 hour, quench
to 200. ae Sminutes.)






?


The primary goal of this research was to determine the

mechanism by which repeated nucleation of 0' occurs in the Al-Cu system. However, since this work is the first reported observation of the phenomenon in Al-Cu alloys, a secondary goal was to characterize thoroughly the various features of the microstructures observed. The geometrical analyses are reported in detail here.

The material in this chapter is developed much in the

way in which the experimental analysis was performed. First, in Section 4.2, the nature and sources of the climbing dislocations are established. Next, in Section 4.3 the precipitate is identified as 0' and characterized as to distribution of orientations in the colonies. Section 4.4 contains descriptions of the various geometries and some of the diffraction effects. Finally, Section 4.5 describes results of experiments designed to determine the effects of different parameters on the repeated nucleation process. Most of the analyses for identification purposes were conducted on the

3.85 wt.'O Cu alloy. Accordingly, all micrographs in this chapter are from this alloy, except those in Section 4.5.5. In addition, most micrographs in this chapter are from samDles quenched directly to aging temperatures. For the sake

of brevity, the copper concentration of the alloys is listed in the figure captions as either 4, 2, 1, or 1/2 wt.%.






S3


4.2. Nature and Source of the Climbing Dislocations


The dislocations which climbed during quenching can be classified mainly into one of two categories according to origin: (1) those generated at dislocation climb sources, and (2) glide dislocations which subsequently climb (a third category found in alloys quenched into oil or water at room temperature will be discussed in Section 4.S.4).


4.2.1. Dislocation Climb Sources

Figure 4.2 shows micrographs from fOilS direct-quenched to 220'C and aged for short times before quenching to room temperature. Present in the microstructures are configurations consisting of concentric dislocation loops. When viewed edge-on, the loops are seen to be coplanar Since their traces are Straight lines, as at points A in Figure

4.2(c). Tilting the foil confirms that these straight lines are traces of coplanar loops. Concentric loops sectioned by the thin foil leave straight-line traces with the foil surfaces, e.g., at B-B in Figure 4.2(b) and (c). Often small particles were observed at the center of the loops as in Figure 4.2(b). The operation of dislocation sources has been

discussed in Section 2.3.3 and will not be repeated here. It will now be established that these are climb sources, and the Burgers vectors and habit planes of the loops will be identified.

A typical source is shown in Figure 4.3. Several loop habits have been generated at the source. We are concerned






54



















a b

















J A4


C



Figure 4.2. Dislocation sources in Al-4 wt.% Cu directquenched from 550'C to 220'C and aged for
8 seconds in (a) and (b) and one minute in (c).






55

















a b

~ 41 9 ii i:'i ,,i! i









C d












e



Figure 4.3. Series of micrographs for determining the
geometry and Burgers vectors of the source
loops. The beam direction is close to [101]
in (a), (b), and (c), to [112] in (d), and
to [001] in (e). (Heat treatment: S.T.
1 hour 550'C, quench to 220'C, age 8 seconds.)





56


in this analysis with the outermost ioop and the one inner loop which lie totally within the foil. The plane of the

foil was analyzed to be very close to (101) so that this loop habit must lie on or very close to (101). Consider first the three images (a), (b), and (c) taken about the [101] beam direction. In each image the source loops on (101) exhibit weak, residual contrast typical of "invisible"~ images of edge dislocations for which 9-BY=0, but 9-(bxT) O (Hirsch et al., 1965, p. 261). Those loop segments lying approximately parallel to the 9 vectors, where g. (ExiT>0-, are invisible. The loops are everywhere visible for the two reflections, g=131 and 220. From this analysis, the source loops are identified as pure edge-dislocation ioops lying on (101) with F-=a/2[101]. Since the Burgers vector is normal to the plane of the loops, the loops must expand in this

plane by the process of climb. The sources are therefore dislocation climb sources of the type observed by Boyd and Edington (1971) in Al-265 wt.% Cu.

The ioop habits of climb sources in these alloys were observed always to be {110} (with one exception to be discussed in Section 4.4). The typical source produced loops on more than one {1101 habit. Loops lying on as many as five of the six possible 1110} habits were observed at one source. The typical source also generated more than one loop on each habit. As many as five or six loops on one habit were commonly observed, although the average number varied with the heat treatment.






S7


When a given foil was first examined, a technique was

used for rapidly determining if the dislocation sources were indeed climb sources with {1101 habits, or if some or all of them might be dislocation glide sources which are known to operation the {1111 slip planes in AI-Cu alloys (Westmacott et al., 19S9). This technique was to tilt the foil to {0011, {1111, and fl0l) orientations and, in each orientation) to determine the number of different source habits viewed edgeon together with the angles between these habits. For example, when a foil was tilted to the (001) orientation, two edge-on habits at 90' apart were observed, Figure 4.4(a). Since the [001] pole is parallel to two {110} planes at 90' to each other., and not to any {111} planes, those sources are identified immediately as {110} climb sources. Likewise, when viewed in the (111) orientation, three edge-on habits at 600 apart were seen, Figure 4.4(b). Again, since the [111] pole is parallel to three {110} planes at 60' to each other and not to any 11111 planes, the sources are identified as climb sources. However., neither of these cases rules out the possibility that other sources seen inclined to the beam in these orientations might be glide sources lying on 1111 planes. Therefore, it was necessary to tilt to a {1011 orientation. The [101] pole is parallel to two {1111 planes at 70.S' and to only one fllOj plane. In this orientation, only one habit was ever seen edge-on, as shown in Figure 4.1. There was no evidence that any of the sources found in these

foils were glide sources.






58





120

220













a
)MI




















b

Figure 4.4. (a) Two edge-on habits of climb sources at 900
to each other in an (001)-oriented foil.
(b) Three edge-on habits at 600 to each other
in a (111)-oriented foil. (Heat treatments:
S.T. 1 hour 5500C, quenched to 2200C, aged
5 minutes.)






59


Occasionally, when viewing edge-on habits in the (001)

orientation, it was observed that two different habits did not lie exactly at 90' to one another, although the habits of other sources in the same field of view appeared to be perpendicular, Figure 4.5. It is concluded that climb of the loops is not necessarily confined strictly to the {1101 planes. This angular measurement between two adjacent habits is more accurate for determining if the loops lie exactly on {1101 planes than are measurements made from the rotation calibration between directions in the diffraction pattern and those in the image.

It has now been established that the dislocation sources in these foils are climb sources. However., the climb of pure-edge dislocations can be either vacancy-annihilating or vacancy-emitting. The former case removes lattice planes from the crystal whereas the latter case adds interstitial planes. It remains to be shown whether the source loops climb by vacancy annihilation or emission, although intuitive arguments favor vacancy-annihilating climb. For instance, it is known that quenching produces large vacancy supersaturations, but negligible concentrations of interstitial. As the temperature drops during quenching, the need for the excess vacancies to diffuse to sinks would promote the growth of vacancy loops and tend to annihilate any interstitial loops. Therefore, one would expect that the large climb sources operate by vacancy annihilation during quenching, but this is not a sufficient proof. In fact, in the past






60






















Figure 4.S. Two edge-on habits of climb sources lying
slightly off 900 from each other in an (001)oriented foil. (Heat treatment: S.T. 1 hour
5500C, quench to 220'C, aged 5 minutes.)























Figure 4.6. Climb source inclined through the foil. Two
loops are sectioned leaving dislocation arcs.
The arc at A was selected for computer matching to determine its Burgers vector. (Heat
treatment: S.T. 1 hour S500C, quenched to
2200C, aged 4 seconds.)






61


only intuitive reasoning has been used to show that climb sources in aluminum alloys operate by vacancy-annihilating climb.

In the present work, the technique of computer matching of dislocation images (Head et al., 1973) was employed to establish that these loops climb by vacancy annihilation, thereby removing planes locally from the lattice. This technique is capable of determining unambiguously the Burgers vector of a dislocation line segment, i.e., whether the

Burgers vector is +B- or -F.

Figure 4.6 shows a climb source in a sample directquenched to 220'C and held only four seconds. This source has generated two loops on a {1101 habit inclined through the

foil, so that each loop is sectioned and leaves two arcs of dislocation. The segment of the outermost loop at A is reasonably straight and was selected for the computer matching experiment (the oscillations in the image are contrast effects arising from the inclination of the dislocation through the foil). From invisibility conditions, the Burgers vectors of these loops were determined to be either a/2[0111 or a/2[0111. The loops are pure-edge and lie on (011). By stereographic analysis, the line direction of segment A was determined to be very close to [100] in (011), and the foil

normal was determined to be [3131.

Six experimental images of segment A are shown in Figure

4.7 along with the corresponding computed images for 5-=a/2[011] and 5-=a/2[011]. These six images represent






62
























d,





e




S..... . ...




b=a/2E [TT] b a/2EO11]


Figure 4.7. Six experimental and computed images of dislocation A in Figure 4.6. The line direction is [100], the foil normal is [313], and the beam
direction is close to [101] in (a) and (b), to [112] in (c) and (d), and to [111] in (e)
and (f).






63


reflections from three non-coplanar beam directions, a necessary condition for uniquely identifying a dislocation by computer matching (Head, 1969). From the rotation calibration of the electron microscope, the exact orientation of the g-vector was marked on each experimental image. Also, from the known geometry of the computer program, the orientation of the g-vector was marked on each computed image. Thus the direction of the j-vector serves as a basis for comparison when matching the features of the computed images with those in the experimental images. Now for a given diffraction vector 'g, the image of a dislocation with Burgers vector +U7 is identical to that of a dislocation with Burgers vector -F5 after a rotation of 1800 (Head et al. 1973, p. 382). Clearly it can be seen from Figure 4.7 that the Burgers vector of Segment A of the loop is a/2[011] and not a/2 [011].

By convention, the positive direction of the dislocation line in the computer program is always taken to be acute to the foil normal. Thus for the foil normal [313] the positive direction of Segment A is [100] and not [100]. Also, the computer program employs the finish-to-start, right-hand (FS/RH-) convention for establishing the direction of the Burgers vector with respect to the positive sense of the dislocation line, Figure 4.8. Thus, from the (PS/RH) convention, and the absolute Burgers vector and positive line direction of Segment A, the geometry of the two 1oops in Figure 4.6 can be established, and this is illustrated






64







o -* 0 0 0 e
o 0o0o 0 0 0 00 0

0 000 0 0 0 000 0

0 0 0 0 0 0 00 ()0
O F O 0-S 0
0S0 0000 F6 000

a b

Figure 4.8. Schematic representation of Burgers circuits
taken in a cubic lattice around an edge dislocation (a), and in perfect crystal (b), illustrating the FS/RH definition of the
Burgers vector. The positive sense of the
dislocation line is out of the paper afterr
Head et al., 1973).









(011) PLANE
SOURCES

5a/,- [OTT]I




Figure 4.9. The geometry of dislocation climb source loops
in Al-Cu as indicated by the absolute sense of
the Burgers vector determined from computer
matching. The pure-edge loops expand in their
habit plane by vacancy-annihilating climb.






65


schematically in Figure 4.9. Clearly the loops are pure-edge, vacancy loops which climb in their habit plane by vacancy annihilation. It is concluded that these climb sources operate during the quench to act as sinks for the excess, quenchedin vacancies.

Now that it is established that the source loops climb by vacancy-condensation onto the loops, one further experimental observation must be explained. The vacancy-annihilating climb of a/2<110> dislocations in f.c.c. lattices is generally believed to occur easily only on {1111 and {110} planes, removing one and two atom planes, respectively (Christian, 1965, p. 363). Stacking faults were never observed within these climb loops, even though the same loops were examined on many different reflections, e.g., Figure 4.3. The stacking of {110} planes in f.c.c. is ABAB (Figure 4.10). Removal of a single {1101 plane by vacancy condensation behind a climbing a/2<110> dislocation would create a stacking fault. Two adjacent fllO} planes must be eliminated to avoid a stacking fau t. It appears, therefore, that the source loops climb by condensation of vacancies onto two adjacent f1101 planes. Since the stacking fault energy in dilute aluminum alloys is high, it appears to be energetically favorable for the loops to expand in this manner. A schematic cross-section through a climb source is shown in Figure 4.11.

The nature of the source particles is undetermined.

Occasionally, sources were observed that nucleated at very






66











Lool]






















thr pln dow liA i A poiios Reoa







Figurof 41.Daga sgleg plae ofacBiatoms crates aitak


ing fault, A on A.





67












(110) PLANES
000000 A OO OOQOOOA
000;0 O00000O0O 000000 B
0OOOO 0000000000 OOOOOA
5'a b/2[110 SOURCE@ A F.Q
OOO@ ..OO OOOOO OOOO-,Buu
0000000000050000 H00000
OOOOO 000 OOOOOOA
00 00 00000 B







Figure 4.11. Schematic diagram of the cross-section through a climb source loop on (110). The pure-edge
loops with b=a/2[110] climb by vacancy condensation onto two adjacent (110) planes, thereby
avoiding creation of a stacking fault.





68


large sphei-oidized particles ( l/1O l/211 diameter), Figure

4.12. These particles were large enough to be analyzed on a JEOL l00B Analytical Electron Microscope using a finefocused electron beam (approximately l,OOOA diameter) and a non-dispersive detection system for fluorescent analysis of the emitted x-rays. The analysis of these large spheroids

identified them as pure lead. It is not known how lead entered the sample material. However, such particles were observed in only a few foils and nucleated only a small fraction (1<]%) of the climb sources present. The typical. source particle was so small as to be barely visible or not visible at all, Figure 4.2(b). Such particles were too small for the x-ray analysis, but they are most probably not lead. The chemical analysis of the 3.85 wt,% Cu alloy (Section 3.1) showed no appreciable concentration of any single impurity which would suggest a guess at the particle nature.

Although the chemical. composition of the source particles is unknown, some observations were made about their distribution. rIh1e climb sources were dispersed randomly in most portions of the foils. Occasionally, local high densities of sources were observed. In a few instances, sources were observed evenly spaced in a straight line (Figure 4.13), suggesting that the source particles were part of an impurity stringer produced when the original cast alloy was rolled down.

The density of active climb sources in these foils varied with heat treatment (Section 4.5) In the only other reported







69




















Figure 4.12. Climb sources generated at a large, spherical
lead particle. (Heat treatment: S.T. 1 hour
550C, quench to 220C, aged 1 minute.)





20
1)-0















Figure 4.13. Climb sources aligned in a row. (Heat treatment: S.T. 1 hour 550'C, quench to 2000C,
aged 5 minutes.)





70


observation of climb sources in Al-Cu alloys, Boyd and Edington (1971) observed a source density of about three per grain in Al-2.5 wt.% Cu (although it is not stated, most probably measured in the volume of a grain sectioned by the foil; grain size not reported).- Source densities many orders of magnitude higher than this were observed in the present foils. The maximum density was produced in a sample quenched from 550'C to 180'C,, A micrograph of this foil is shown in Figure 4.49(e) From this micrograph and the average grain size (approximately 2S5% diameter) the active source density was estimated to be approximately 6x10 sources per grain.


4.2.2. Glide Dislocations Which Climb

In addition to loops generated at climb sources, other dislocations were observed which had climbed during quenching and nucleated precipitate colonies. Examples are shown in Figure 4.14. Gene-rally, these dislocations were long and either smoothly curved or irregular-shaped, depending on their climb paths. The micrographs in. this section were taken from foils aged long enough so that the precipitate colonies are -readily visible, thereby delineating the climb paths of the dislocations. F7or the present, it is assumed that the precipitate colonies were nucleated by the climbing dislocations (this will be proven in Section 4.3). In Figure

4.14(a), the dislocation exits the foil surfaces at A and C, and the trai Iing precipitate colony intersects one foil






71



























414


B"




Fiue4.4 rciiaecoois ulaeda ogcib



in isoaios Tedslctini ()i









Figureou of14 coniptast along ABc.ete (Hat treatment:b

S.T. 1 hour 550'C, quenched to 220'C, aged
S minutes.)~





72


surface along the trace ABC. In Figure 4.14(b), the dislocation lies along ABC at the upper edge of the precipitate colony, where it is "invisible." Some residual contrast can be seen, however. The source of these dislocations is unknown. They could be (1) grown-in dislocations, (2) glide dislocations which existed at the solution treatment temperature prior to quenching, or (3) glide dislocations which were generated at some source, probably grain boundaries, at the onset of quenching. It is thought that most, if not all, fall into categories (2) and (3), i.e., they were a/2<110> glide dislocations on Ml} planes prior to climbing. Figure 4.1S, for example, shows two images of a precipitate colony nucleated by one such long dislocation. In (a) the precipitate colony is inclined through the foil. The curved dislocation has been sectioned twice by the foil, leaving two arcs, AB and CD, at the ends of the precipitate colony. The Burgers vector of this dislocation was determined to be a/2[101]. The micrograph in (b) was taken after the foil was tilted to the (111) orientation. Here the precipitate colony is viewed edge-on and appears as a curved, dark line. This indicates that the dislocation climbed on an irrational, smoothly curved surface normal to the (111) plane. Since (111) is a glide plane, it is reasonable to assume that it was a curved, glide dislocation on (111) prior to climbing.

Such long glide dislocations were observed often to

have cllmbod on smoothly cut'ved suTfacos. This was easily recognized by the curved. intersections which the asSoci-Ited






73





















a




















b
Figure 4.15. Precipitate colony nucleated by climb of a
glide dislocation, initially on (111), with
FIa/2[101]. The dislocation has been sectioned
twice by the foil, leaving arcs AB and CD in
(a). The beam direction is close to [011] in
(a) and to [111] in (b). (Heat treatment: S.T. 1 hour 550'C, quenched to 220'C, aged
5 minutes.)





74


precipitate colonies made with the foil surfaces, Figure 4.16. Similarly, Miekk-oja and R~.ty (1971) observed repeated nucleation of silver-rich precipitates from solid solutions of silver in copper behind dislocations which were shown to be a/2<110> glide dislocations on 11111 planes before climbing. They foune. that these dislocations subsequently climbed in one of two different ways: (1) off the slip plane on smoothly curved surfaces, similar to -that described above, or (2) into a crooked shape so that different segments of the dislocations climbed on different low-index -planes intersecting the original slip plane. They further showed that these low index planes were of the typos {1101 and'11111, i.e., the planes of "easy climb" (Section 2.3.1) on which the chemical climb force, from a subsaturation of vacancies, was the greatest.

In the present research, the shapes of precipitate

colonies behind certain glide dislocations suggested that

different segments of these Jislocations had climbed on separate crystallographic planes also. The term "segmented

climb" shall be used here to refer to this mode of climb. Micrographs of precipitate colonies apparently resulting from segmented climb are shown in Figures 4.17 and 4.14(b). In Figure 4.17, the dislocation between AB has climbed through the lattice From left to right. The precipitate colony nucleated by this dislocation is separated into bands of precipitates. This effect is thought to be associated with

the climb of adj acent dislocation segments on separate





75











'*:o
-4%






t* 0"0




Figure 4.16. precipitate colony exhibiting Curved traces
of intersection with the foil surfaces, indicating a curved climb path of the nucleating dislocation. (Heat treatment: S.T. 1 hour

550C, quenched to 220'C, aged 5 minutes.)
0.C'



















Figure 4.17. Banded precipitate colony nucleated by "segmented climb" of the dislocation AB from left
to right through the foil. (Heat treatment

S.T. 1 hour 550C, quenched to 220'C, aged
5 minutes.)






76


crystallographic planes. The resultant precipitate colony has a corrugated shape. No precipitation Occurred in areas between the bands, e.g., at C. Such precipitate-free areas can also be seen in the climb path of the dislocation in Figure 4.14(b). For some reason the dislocation is ineffective in nucleating precipitates in these regions of its climb path. A possible explanation is that precipitation Occurs readily on those segments of the dislocation which climb on the separate crystallographic planes, but not on those curved arcs of the dislocation which bridge the gaps between these planes. This will be discussed further in

Section 4.4.6.

The segmented climb of initial glide dislocations was observed only in the AI-3.8S wt.% Cu alloy. An attempt was made to determine the planes upon which segmented climb occurred by trace analysis of the intersections of the precipitate colony with the foil surface, but this proved to be impractical for two reasons. First, as pointed out by Miekk-oja and My (1971), the possible planes of easy climb can be numerous, i.e., six {1101 and four {1111 planes, so that the trace analysis is best accomplished by using single crystals cut to special orientations. Such crystals were not available in this research. Secondly, the traces of the intersections of the bands of precipitates with the foil surfaces were never well defined, a condition which leads to poor accuracy in the stereographic analysis.





77


There is evidence that the tendency for glide dislocations to climb either on smoothly curved surfaces or to segment and climb on different )lanes depends upon the line direction of the dislocation before climb began. For example, in Figure 4.18 the curved dislocation along ABC has nucleated a smoothly curved precipitate colony between A and B, and a corrugated colony between B and C. Presumably, the initial line direction of the dislocation segment between B and C was favorable for rapid climb onto the different planes of easy climb.





4.3. Identification and Characterization
of the Precinitate Pfihse


Figure 4.19(a) is a difFraction pattern in the exact (001) matrix orientation taken from the group of precipitates shown in Figure 4.19(b). The image quality in (b) is poor since the exact (001) orientation is a many-beam condition. The geometry in Figure 4.19(b) is as follows. The foil has sectioned three ll] 0 habits of dislocation cli-mb sources, numbered 1, 2 and 3. These habits are viewed edgeon in this orientation. Habits 1 and 2 lie on (110) while habit 3 lies on (i0) Small precipitate platelets are dispersed over the three habits. Two orientations of precipitates are present lying parallel to the (100) and (010) matrix )h1anes. The p:recipitate reflections in the diffraction pa":tern are streaked in the <100> directions owing to






78


























CN



V44




Figure 4.18. A glide dislocation which climbed on a
smoothly-curved surface between A and B, and underwent segmented climb between B
and C. (Heat treatment: S.T. 1 hour
5150C, quenched to 2200C, aged 5 minutes.)






79


































a"










b


Figure 4.19. (a) (001) diffraction pattern showing precipitate -reflections, taken from the area of the
foil shown in Nb. (Heat treatment: S.T.
1 hour SS00C, quenched to 220'C, aged 30
minutes.)






80


the relaxation of the Laue condition along the thin dimension of the platelets. The diffraction pattern can be analyzed

on the basis of three superimposed patterns shown in Figure

4.20(a)-(c), where we consider only the lower right-hand quadrant of the pattern. The patte-rn in (a) is the (001) matrix pattern. The pattern in (b) is indexed on the basis of two el orientations parallel to (100) and (010) matrix planes, using the lattice parameters of 4.04A and SAX for el (Section 2.2). The remaining reflections in (c) are due to double diffraction from the matrix {2001 and {2201 beams. Double diffraction is a common occurrence in foils containing

precipitates with dimensions much smaller than the foil thickness. The composite pattern, shown in (d), matches the experimental pattern. Thus the precipitates are positively identified as the of phase, in agreement with the known fact that a' is the only metastable phase which nucleates on dislocations in Al-Cu.

In the present research, conditions were chosen to insure that the e' phase was the only precipitate phase present after heat treatment. Its distribution was always associated with the climbing dislocations.

The appearance of the precipitate colonies at high

magnifications is shown in Figure 4.21. These are typical colonies nucleated on dislocation climb sources. The colony in (a) was nucleated on the (101) source habit in the plane of the foil. In (b), five different {1101 habits were generated at the same source particle and have nucleated






81





O00M 020M O00M 002
00-one. (0)
I OP' 20%(

1 1~

o2 00 0
e 0
200M 220M 103G
a b







oooM O00M

S- 0(D



I I





c d


Figure 4.20. Analysis of the lower, right-hand quadrant of
the diffraction pattern in Figure 4.14(a).
(a) (001) matrix pattern; (b) precipitate
reflections from e' platelets lying parallel
to (010) matrix planes (0j), and to (100) matrix
planes (01); (c) double diffraction from matrix
200 and 220 reflections; (d) combination of
(a), (b), and (c). Compare with Figure 4.14(a).






82











A04











a




~b

cooie en ated on. th -10 hbts
b s.a S
~B










S..1hu 55,qece t 2,ae
mi e


020



C


b

Figure 4.21. (a) and (b) Appearance of typical precipitate
colonies generated on the f11O1 habits of
climb sources. (c) and (d) Schematic diagrams
illustrating the geometry of the colonies.
See text for description. (Heat treatment: S.T. 1 hour 550'C, quenched to 220'C, aged
5 minutes.)






83






VIEWING DIRECTION., NORMAL TO (101)





5=2/2[1011











c


VIEWING
DIRECTION, NORMAL TO (101)














(f of)


SECTION OF THIN FOIL

d


Figure 4.21. Continued.






84'


precipitate colonies. One habit lies in the plane of the foil at A, one habit is viewed edge-on at B, and three other habits at C, D and E are inclined through the foil. The geometries of these sources are shown schematically in Figure

4. 2 1 (c) and (d) .
Often the best condition for imaging the precipitates in a colony was also a condition for "invisibility" of the dislocation loop bounding the colony, e.g., Figure 4.21(a). This was particularly true when examining colonies on dislocation climb sources where the best projected view of a colony was obtained with the beam oriented approximately normal to the colony, and hence to the bounding source loop. However, since the source loops are pure-edge with their Burgers vectors normal to their habit planes, any two-beam reflection selected to view the precipitate colony in this orientation has its g-vector perpendicular to the Burgers vectors of the loops. Thus the loop will be "invisible" when viewed normal to its habit plane. This is the case in Figure

4.21(a) where 9-F=0 for the loop bounding the precipitate colony and only residual contrast due to the pure edge-nature

of the loop is detected.

Likewise in Figure 4.21(b), the dislocations bounding

the colonies at A, C and D are invisible, whereas the one at E is visible. A consistent interpretation of the geometry of this source array is as follows. The beam direction is close to [101] and g=lli for this image. Habit A lies on the (101) plane of the foil (see Figure 4.21(d)). The






85


Burgers vector of its source loop, being pure edge, is a/2[101]. Thus the loop is invisible for g=lll. The source loops bounding the habits C and D are also invisible. These dislocations must have the other two a/2<110> Burgers vectors which cause invisibility for the lli reflection, namely, a/2[011] and a/2[liO]. Habit B lies on the (101) plane parallel to the beam and is viewed edge-on. The Burgers vector of its source loops must be a/2[iOl]. The source loop bounding habit E must have one of the two remaining a/2<110> Burgers vectors, namely, a/2[110] or a/2[Oli], both of which would be visible for the lli reflection. The dislocation is visible at E.

It is difficult to determine whether the smallest precipitates in these colonies are imaged by strain contrast in the matrix or by structure factor contrast (Ashby and Brown, 1963). The larger precipitates in a colony are imaged by the dislocation loops bounding the periphery of the platelets

(Section 4.4).

In Section 2.2) it was pointed out that because of dislocation strain effects, only two el orientations will nucleon any given a/2<110> dislocation. The missing orientation has its principal misfit (normal to the plane of the platelet) perpendicular to the Burgers vector of the dislocation so that its strain field is not relieved by the stress field of the dislocation. Careful examination of a number of precipitate colonies at dislocation climb sources revealed that only two 61 orientations were present in any given colony.





86


The missing orientation was always that (100} orientation whose misfit would be perpendicular to the Burgers vector of the source loop bounding the colony. This is illustrated in

Figures 4.21(b) 4.22 and 4.23.

First, Figure 4.22 shows bright and dark field images of several precipitate colonies on climb sources in a foil whose normal was close to [1011. The foil was oriented with the electron beam close to [101]. The dark field image was taken from a precipitate reflection from the (010) 01 orientation parallel to the beam. Climb source A (and its precipitate colony) lies on (101).' and its source loops, being pure edge, have Burgers vectors a/2[101]. At B, two other source habits lie on f1101 planes inclined to the foil. The Burgers Vectors of their source loops were not determined but they cannot be a/2[1011. Now, if all three {1001 orientations of el were present in the colony at A, the orientation imaged in the dark field would be observed throughout source A as in the sources at B. The misfit of this missing (010) orientation is perpendicular to the a/2[1011 Burgers vector of the source loops at A (which are "invisible" in this image). The few precipitates on (010) in the middle of source A in the dark field image were found to lie within small source loops lying on other T1101 habits, when this source was

examined in another orientation.

Next, in Figure 4.21(b), the source habit at B lies on the (101) plane and is viewed edge-on. The Burgers vectors of its source loops, being pure edge, must be a/2[101].





87































Figure 4.22. Bright-field and dark-field images of precipitate colonies on dislocation climb sources.
The colony at A lies in the (101) plane of the
foil. The colonies at B are inclined through
the foil. The dark-field was taken from a
precipitate reflection from the (010) 0' orientation lying parallel to the [101] beam direction. (Heat treatment: S.T. 1 hour 5500C,
quenched to 2200C, aged 5 minutes.)






88


The (010) 01 orientation, which is parallel to the beam direction and also viewed edge-on, is not present in habit B, although it is easily detected in habits C and E. Again, this is the 01 orientation whose misfit is perpendicular to the a/2[101] Burgers vector of the source loops bounding

habit B.

The fact that el platelets lying parallel to the beam

can indeed be seen if present in precip itate colonies viewed edge-on is shown in Figure 4.23. This micrograph was taken with the beam oriented near [001].. The two edge-on habits at A and B lie on (110) planes so that the bounding dislocation loops have Burge-rs vectors a/2[1101. Both the (010) and (100) orientations of 6' platelets can be clearly seen dispersed along the habits. Note that these are the two e' orientations whose misfits are not perpendicular to the Burgers vector of the bounding dislocation loops and are therefore favored to be nucleated by the loops.

All three 81 orientations were never observed in a given precipitate colony. As pointed out above, the missing orientation was always that whose nucleation is not aided by the stress field of the dislocations bounding the colony. This was true for precipitate colonies generated by both climb sources and glide dislocations which climbed. This evidence leads to the important conclusion that the precipitates must have nucleated in the stress field near the dislocations as they climbed through the lattice, and not at some later time when the influence of the dislocation was no longer present,






89































Z2.44






Figure 4.23. 0' precipitate colonies on climb source habits
A and B which lie parallel to the beam. Two
edge-on orientations of e' platelets, (100)
and (010) are clearly visible in colonies A and B. The beam direction is close to [001].
(Heat treatment: S.T. 1 hour 550'C, quenched
to 2200C, aged 30 minutes.)





90


e.g., during aging. Since the dislocation climb sources are known to have operated during the high-temperature part of the quench, the precipitates must have nucleated during quenching.

Further evidence to support this conclusion was obtained by in situ aging experiments in the electron microscope. The results of these experiments are shown in the micrographs of Figure 4.24. These are images of a foil from a sample solution treated for one hour at SSO'C, quenched into oil at 220'C and held only four seconds at 220'C, then water-quenched to room temperature. After electropolishing, the foil was placed in the heating stage of the microscope. The micrographs in Figure 4.24(a) and (b) Were taken prior to heating. In (a), a dislocation climb source is viewed normal to its

(011) habit of source loops which are "invisible" in this orientation and imaged by -residual contrast. Since this is a climb source, we know that these dislocations climbed during quenching. A-long, crooked dislocation, which was most probably a glide dislocation prior to quenching, is shown in (b). 'Its crooked shape is the only indication that it may have climbed during quenching. Now, if nucleation of the 01 precipitates does occur as the dislocations climb through the lattice during quenching, then the precipitates must already be present in the foil in (a) and (b). However, the four-second aging time at 220*C was insufficient to cause the precipitates to grow to visible sizes. The foil was then heated to 230'C in the microscope. After nine minutes at






91




















oil

(a)





















(b)

Figure 4.24. (a) A dislocation climb source, imaged by
-residual contrast, and (b) a long glide dislocation in a sample quenched from 550*C to 220*C and held only 4 seconds before quenching to room temperature.






92







++ %,-B,+


00








.1+ A,' I
44




~ 4
(c)








I -zr4 q
,+ p++- : ++.s ...
4A,





lip B





(d)

Figure 4.24. Continued. (c) and (d) Micrographs of the
same dislocations in (a) and (b) after aging
9 minutes at 230'C in the electron microscope.
The random precipitation at B has occurred at the foil surfaces. The precipitate colonies
associated with the dislocations are now
clearly visible at A.




Full Text
196
5.4. Criteria for Repeated Nucleation in Al-Cu
and Application to Other Systems
Based on the results reported herein, the following
criteria may be established for repeated nucleation in this
system:
(1) a phase which nucleates easily on dislocations,
namely the 0 phase;
(2) a source or sources of dislocations during the
quench, namely dislocation climb sources and
glide dislocations generated during quenching;
(3) a driving force for dislocation climb which is
independent of the precipitate reaction, namely
the annihilation of quenched-in vacancies; and
(4) a climb rate which is sufficiently slow to.
permit nucleation, but sufficiently fast so
that the precipitates do not grow so large as
to effectively pin the dislocations.
These criteria are far simpler than those required by the
mechanism of Nes (1974), and might well apply to other sys
tems in which dislocation-nucleated transformations occur.
(2) and (3) above apply to most aluminum alloys. In par
ticular, Al-Mg and certain Al-Mg ternary alloys have mestable
phases which nucleate on dislocations, and appreciable dis
location climb occurs when these alloys are quenched. Unlike
copper in Al-Cu, the magnesium atom is larger than the aluni-
num atom and it is unclear what effect this may have. The
Al?CuMg phase nucleates easily as laths on dislocations in


189
Figure 5.3. Micrograph from which a measurement was made
of the number density of precipitates in a
colony. (Heat treatment: S.T. 1 hour 550C,
quenched to 220C, aged 5 minutes.)


22
2.3.3. Dislocation Climb Sources
The classical dislocation multiplication mechanism is
that proposed by Frank and Read (1950) whereby a dislocation,
pinned at each end, expands in its slip plane by glide.
Westmacott et_ al. (1959) observed dislocation sources in thin
foils of Al-4 wt.l Cu by transmission electron microscopy,
and interpreted them to be Frank-Read sources. Gulden and
Nix (1968) have observed similar sources in Al-4 wt.% Cu-
3 wt.% Si. Analogous to the Frank-Read mechanism, a dislo
cation multiplication mechanism which operates by climb was
proposed by Bardeen and Herring (1952) to act as a continuous
sink for excess vacancies. The Bardeen-Herring model for a
dislocation climb source is shown in Figure 2.7. Initially,
a straight dislocation between A and B has its slip plane
normal to the plane of the paper. Hence, it can move in the
plane of the paper only by climb. Condensation of vacancies
onto this line would move the dislocation out through the
sequence of positions shown. When the bottom segments of the
loop meet, they annihilate and rejoin as shown by the dotted
lines. The segment ABC is now free to repeat the process,
and there is left a vacancy loop outlined by the dislocation
ring. As long as there remains a vacancy supersaturation
in the region, this loop will expand, removing atoms from
the lattice, and the operation can repeat removing an indefi
nite number of planes.
Dislocation climb source configurations were first
observed by TEM in Al-Mg alloys (Westmacott et al. 1962;


150
Figure 4.49
Continued


19
oooooo
oooooo
oooooo
ooc- oo
OOO'OOO
oooooo
oooooo
oooooo
oooooo
OOO-QOO
ooo ooo
oooooo
oooooo
oooooo
000*000
OOOOQOO
OOO OOO
OOOOOO
a
000*000
000*000
ooosooo
000800
oooaooo
oooooo
b
000*000
oooooo
oooooo
ooo ooo
ooo ooo
oooooo
ooo@ooo
000*000
000*000
000*0 o
ooo ooo
oooooo
Figure 2.5. (a) Positive climb of an edge dislocation
by vacancy annihilation. (b) Negative climb
by vacancy emission (after Reed-Hill, 1973).


13
Figure 2.2.
The tetragonal unit cell of 0' (after
Silcock, Heal, and Hardy, 1953).


77
There is evidence that the tendency for glide disloca
tions to climb either on smoothly curved surfaces or to seg
ment and climb on different planes depends upon the line
direction of the dislocation before climb began. For example,
in Figure 4.18 the curved dislocation along ABC has nucleated
a smoothly curved precipitate colony between A and B, and a
corrugated colony between B and C. Presumably, the initial
line direction of the dislocation segment between B and C
was favorable for rapid climb onto the different planes of
easy climb.
4.3. Identification and Characterization
of thePrecipitate Phase
Figure 4.19(a) is a diffraction pattern in the exact
(001) matrix orientation taken from the group of precipi
tates shown in Figure 4.19(b). The image quality in (b) is
poor since the exact (001) orientation is a many-beam condi
tion. The geometry in Figure 4.19(b) is as follows. The
foil has sectioned three {110} habits of dislocation climb
sources, numbered 1, 2 and 3. These habits are viewed edge-
on in this orientation. Habits 1 and 2 lie on (110) while
habit 3 lies on (110). Small precipitate platelets are dis
persed over the three habits. Two orientations of precipi
tates are present lying parallel to the (100) and (010)
matrix planes. The precipitate reflections in the diffrac
tion pattern are streaked in the <10Q> directions owing to


65
schematically in Figure 4.9. Clearly the loops are pure-edge,
vacancy loops which climb in their habit plane by vacancy
annihilation. It is concluded that these climb sources oper
ate during the quench to act as sinks for the excess, quenched-
in vacancies.
Now that it is established that the source loops climb
by vacancy-condensation onto the loops, one further experi
mental observation must be explained. The vacancy-annihilat
ing climb of a/2<110> dislocations in f.c.c. lattices is
generally believed to occur easily only on {111} and {110}
planes, removing one and two atom planes, respectively
(Christian, 1965, p. 363). Stacking faults were never
observed within these climb loops, even though the same loops
were examined on many different reflections, e.g., Figure 4.3.
The stacking of {110} planes in f.c.c. is ABAB (Figure 4.10).
Removal of a single {110} plane by vacancy condensation
behind a climbing a/2<110> dislocation would create a stack
ing fault. Two adjacent {110} planes must be eliminated to
avoid a stacking fau]^t. It appears, therefore, that the
source loops climb by condensation of vacancies onto two
adjacent {110} planes. Since the stacking fault energy in
dilute aluminum alloys is high, it appears to be energetic
ally favorable for the loops to expand in this manner. A
schematic cross-section through a climb source is shown in
Figure 4.11.
The nature of the source particles is undetermined.
Occasionally, sources were observed that nucleated at very


113
colonies. Some of these (110) habits are viewed edge-on (as
at B), some lie in the plane of the foil (as at C), and some
are inclined through the foil (as at D). A similar configu
ration was shown in Figure 4.14(b).
It is proposed that these configurations evolved in the
following manner. The long glide dislocation began to climb
shortly after the onset of quenching. When the sample tem
perature passed below the 0' solvus, this dislocation began
to nucleate 0 precipitates in its climb path. Some of the
earliest-nucleated platelets acted as source particles and
generated the climb source loops on {110} planes, and these
then nucleated their own precipitate colonies.
The term "secondary climb sources" shall be used to dis
tinguish sources nucleated at 0' platelets in this manner
from "primary" climb sources nucleated on insoluble particle
existing at the solution treatment temperature. This seems
an appropriate designation since the secondary sources nucle
ate only if the 0 precipitation reaction occurs, whereas
primary sources operate independent of the precipitation.
From the observation that secondary climb sources were
always located at the base of the precipitate colonies gener
ated by the long climbing dislocations, it can be deduced
that their nucleation occurs within some limited, time-
temperature range just below the 0 solvus when the earliest
0' precipitates formed. The diameters of the largest secon
dary source loops were never as large as the diameters of
the largest primary source loops in the same foil, in


REPEATED NUCLEATION OF PRECIPITATES
ON DISLOCATIONS IN ALUMINUM-COPPER
By
THOMAS JEFFREY HEADLEY
A DISSERTATION PRESENTED TO THE GRADUATE COUNCIL OF
THE UNIVERSITY OF FLORIDA
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE
DEGREE OF DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
1974

Copyright by
Thomas Jeffrey Headley
1974

Dedicated to my wife, Lynn

ACKNOWLEDGMENTS
The author is deeply indebted to his Advisory Chairman,
Dr. John J. Hren, who contributed unselfishly of his time,
and provided advice, encouragement, and stimulating discussion
during the course of this research. He is indebted to his
advisory committee for assistance, and to Dr. R. T. DeHoff
for many helpful discussions.
Thanks are due to Dr. R. W. Gould for providing alloy
materials; to the Sandia Corporation, Albuquerque, New Mexico,
for chemical analysis; to the Japan Electron Optics Laboratory,
Boston, Massachusetts, for use of the JEOL 100B Electron
Microscope; to Mr. E. J. Jenkins for assistance in the labo
ratory; to Mr. Paul Smith for assistance in the darkroom; and
to Mrs. Elizabeth Godey for typing this manuscript.
The author*s wife, Lynn, is acknowledged for her con
stant inspiration and support. His mother is acknowledged
for her lifelong encouragement.
Finally, the financial support of the Atomic Energy Com
mission was deeply appreciated.
iv

TABLE OF CONTENTS
Page
ACKNOWLEDGMENTS iv
ABSTRACT ....... vii
CHAPTER
1 INTRODUCTION 1
2 REVIEW OF THEORY AND PREVIOUS WORK 5
2.1. Theory of Heterogeneous Nucleation
at Dislocations 5
2.2. Precipitation in the Al-Cu System .... 10
2.3. Dislocation Climb 17
2.3.1. Quenched-In Vacancies and
the Chemical Climb Force 17
2.3.2. Theory of Dislocation Climb ... 18
2.3.3. Dislocation Climb Sources .... 22
2.4. Repeated Nucleation on Dislocations ... 25
2.5. Pertinent Electron Microscopy Theory ... 28
2.5.1. Two-Beam Diffraction Contrast
Theory ..... 28
2.5.2. Defect Identification from
Invisibility Conditions 35
2.5.3. Imaging Precipitates in the
Electron Microscope ....... 37
3 EXPERIMENTAL PROCEDURES AND MATERIALS 41
3.1. Specimen Materials ..... 41
3.2. Heat Treatments 42
3.3. Electron Microscope Specimen
Preparation 45
3.4. Electron Microscopy 48
4 EXPERIMENTAL RESULTS AND ANALYSES 50
4.1. Introduction 50
4.2. Nature and Source of the Climbing
Dislocations 53
4.2.1. Dislocation Climb Sources .... 53
4.2.2. Glide Dislocations Which
Climb 70
v

TABLE OF CONTENTS Continued
CHAPTER Page
4 (Continued)
4.3. Identification and Characterization
of the Precipitate Phase 77
4.4. Further Geometric Analyses 94
4.4.1. Distribution of Precipitates
in Colonies at Climb Sources ... 94
4.4.2. Geometry of the Precipitate
Stringers 9 8
4.4.3. Determination of the Burgers
Vectors of Small Loops Within
Precipitate Colonies 106
4.4.4. "Secondary" Climb Sources .... Ill
4.4.5. A Climb Source on (100) ..... 118
4.4.6. Nucleation of Preferred 6'
Orientations During Segmented
Climb 122
4.4.7. Displacement Fringe Contrast
in a Precipitate Colony 127
4.4.8. Precipitate Colonies Associated
with Subboundary Formation .... 132
4.5. Effects of Experimental Variables on
Microstructure 135
4.5.1. Effect of Time at Constant
Aging Temperature ........ 135
4.5.2. Effect of Solution Treatment
Temperature 140
4.5.3. Effect of Temperature to Which
Samples Are Direct-Quenched . 147
4.5.4. Effect of Quench Rate 154
4.5.5. Effect of Copper Concentration . 166
4.6. Summary 176
5 THE REPEATED NUCLEATION MECHANISM ....... 178
5.1. Nucleation of 0' Near Edge Dislocations 178
5.2. Comparison with Previous Repeated
Nucleation Mechanisms 182
5.3. The Mechanism in Al-Cu 184
5.3.1. Local Solute Buildup 186
5.3.2. Precipitate Stringer Formation . 190
5.4. Criteria for Repeated Nucleation in
Al-Cu and Application to Other Systems . 196
6 CONCLUSIONS AND SUGGESTIONS FOR FUTURE WORK . 198
BIBLIOGRAPHY 204
BIOGRAPHICAL SKETCH .................. 208
vi

Abstract of Dissertation Presented to the Graduate Council
of the University of Florida In Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
REPEATED NUCLEATI ON OF PRECIPITATES
ON DISLOCATIONS IN ALUMINUM-COPPER
By
Thomas Jeffrey Headley
August, 1974
Chairman: John J. Hren
Major Department: Materials Science and Engineering
Results are presented of an investigation of a newly
discovered propagation mechanism for the a->0 transformation
in Al-Cu: repeated nucleation on climbing dislocations. It
was found that during the quench, dislocations are generated
and climb by the annihilation of quenched-in vacancies.
Densely populated colonies of 6' precipitates nucleate in the
stress fields of the climbing dislocations. In this way, the
distribution of the entire volume fraction of 0' is estab
lished during the quench.
The climbing dislocations were found to be a/2
type, falling into three categories according to origin:
(1) pure-edge loops on {110} habits nucleated at dislocation
climb sources, (2) glide dislocations initially on {111},
and (3) pure-edge loops on {110} formed by the collapse of
vacancy clusters.
The effects of solution treatment temperature, aging
temperature, quench rate, and solute concentration on the
vii

repeated nucleation process wore determined. It was found
that repeated nucleation occurs during quenching from all
temperatures within the solid solution range, to all temper
atures in the range room temperature to 300C. It occurs
during slow and fast quenching as well, but does not occur
in alloys with concentration <1 wt.% Cu.
Mechanisms of repeated nucleation proposed earlier for
other alloys are not applicable to Al-Cu. Dislocation climb
and precipitation were found to be independently controlled
processes. The relevant criteria for repeated nucleation in
this system are:
(1) a precipitate phase which nucleates easily on
dislocations,
(2) a source of dislocations during quenching,
(3) a driving force for dislocation climb which is
independent of the precipitation process, and
(4) a climb rate slow enough to permit nucleation
but rapid enough to avoid pinning.
It is suggested that pipe diffusion along the moving dislo
cation provides the necessary solute enhancement for succes
sive nucleations.

CHAPTER 1
INTRODUCTION
Some of the most important strengthening mechanisms in
alloys result from the precipitation of a second solid phase
from a supersaturated solid solution. The age-hardening
phenomenon in many aluminum alloys is a well-known example.
A precipitation reaction is a nucleation-and-growth trans
formation. Hence, the rate of the reaction is dependent upon
(1) the nucleation rate of precipitates, and (2) the rate of
their growth. If either or both of these rates is low the
reaction rate will be low. Consequently, it is important to
understand how and where precipitate reactions nucleate,
apart from the problem of growth. Much is known about the
kinetics of growth, but not about nucleation, especially
heterogeneous nucleation.
Precipitate reactions nucleate either homogeneously or
heterogeneously within the matrix. If both the volume misfit
and interfacial energy between precipitate and matrix are
small, the reaction can nucleate homogeneously at random
sites throughout the lattice. Homogeneous nucleation is
known to occur in only a few alloy systems, e.g., the pre
cipitation of cobalt particles from dilute solutions of
cobalt in copper (Servi and Turnbull, 1966). In most
1

2
precipitate reactions, either the volume misfit or inter
facial energy, or both, is sufficiently large so that only
heterogeneous nucleation occurs at preferred sites within the
lattice. These sites are lattice defects such as grain
boundaries, dislocations, stacking faults, point defects,
and other particles. Presumably, a portion of the energy
associated with the defect is supplied to help reduce the
activation energy for formation of the critical nucleus,
resulting in a nucleation event which is otherwise energetic
ally unfavorable.
In the case of heterogeneous nucleation at dislocations,
it is the dislocation strain energy in the matrix which helps
overcome the barrier to nucleation. If the misfit strain
caused by the precipitate is large, dislocation-nucleation
may be the only method of decomposition of the supersaturated
solid solution. Kelly and Nicholson (1963) and Nicholson
(1970), have given excellent reviews of the evidence for
nucleation on dislocations in a number of alloy systems. A
well-known example of heterogeneous nucleation on disloca
tions is that of the metastable 0' phase in Al-Cu alloys.
For a precipitate reaction which is dislocation-
nucleated, the following problem arises whenever the initial
dislocation density is low, as is often the case following
quenching. How can the reaction propagate once the available
nucleation sites on dislocations have been saturated, i.e.,
what is the mechanism for propagation of the nucleation?
Currently there are two known mechanisms whereby the reaction

3
may propagate. One mechanism is autocatalytic nucleation,
first proposed by Lorimer (1968) and similar to the way in
which martensite propagates. The initial precipitates nucle
ate on dislocations and grow into the matrix. In so doing,
they generate stress fields in the matrix which aid in the
nucleation of new precipitates. Thus, the reaction propa
gates in bands spreading out from the original dislocations
to fill the lattice. Lorimer showed that the a+0' reaction
in Al-Cu could propagate by autocatalytic nucleation. Before
the present work, this was the only reported mechanism
whereby the 0' reaction propagates from a low initial dis
location density.
Secondly, nucleation of the reaction can be propagated
if the dislocation can somehow free itself from the initial
precipitates and move away under a chemical or mechanical
stress. It then presents fresh sites for the nucleation of
more precipitates. Nicholson (1970) was the first to use
the term "repeated precipitation on dislocations" to describe
this process. Repeated nucleation on climbing dislocations
was first observed for carbide precipitation in austenitic
stainless steel (Silcock and Tunstall, 1964). Since that
time it has been reported for only a few other alloy systems.
Very recently, Nes (1974) published a paper on the mechanism
for repeated precipitation on dislocations which he implied
was universal with the statement that his model "can be
applied to repeated precipitation (or colony growth) in any
alloy system."

4
During experiments in which Al-Cu alloys were quenched
directly to aging temperatures, this author observed that
precipitation of the 0' phase occurred exclusively by repeated
nucleation on climbing dislocations. Further examination
revealed that the repeated nucleation in this system could
not be explained by the mechanism proposed by Nes (1974).
Therefore, the primary purpose of this research was to estab
lish the mechanism for repeated nucleation of 0' on climbing
dislocations in Al-Cu, and in so doing, to determine if there
are aspects of the mechanism which might apply to precipita
tion in other alloy systems.

CHAPTER 2
REVIEW OF THEORY AND PREVIOUS WORK
2.1. Theory of Heterogeneous Nucleation
at Dislocations
Nucleation theory employs the concepts of a critical
nucleus and an activation energy for nucleation. An assump
tion of the theory is that random thermal fluctuations lead
to the formation of small embryos of the transformed phase.
Embryos having a size and shape smaller than some critical
size and shape will on the average disappear, and those with
a larger size will grow and become stable. This critical
size and shape is defined as the critical nucleus. The acti
vation energy is the minimum energy barrier which must be
overcome before nucleation can occur and turns out to be the
free energy of the critical nucleus. This energy barrier is
a maximum with respect to size and a minimum with respect to
all other variables. The importance of the free energy of
the critical nucleus can be recognized from its appearance
in the equation for the steady state nucleation rate, which
is written in general form as
J
ZB
/ exp
s
[
-AG*,
~TTJ
(2.1)
5

6
where J
Z
3
N
s
AG* =
k
T
AG* has the
nucleation rate,
the Zeldovich factor,
frequency which atoms add to the nucleus,
number of available nucleation sites,
composition of the nucleus,
free energy of the critical nucleus,
the Boltzmann constant, and
temperature.
form
AG*
K (a J3
^ surf7
(AGdrlye)Z
(2.2)
where K = a shape factor,
surf = Particle/matri-:x interfacial energy, and
n T* T VP
AG = driving force for the reaction.
For homogeneous nucleation, is high assuring reasonable
nucleation rates. For heterogeneous nucleation, Ng is low
and the nucleation rate is usually dominated by the exponen
tial dependence on the free energy of the critical nucleus.
For precipitation in solids, there is a chemical free
energy change per unit volume tending to drive the transfor
mation to the new phase. If the atomic volumes in the matrix
and particle are different, there is a misfit strain energy
associated with formation of the new phase. Thus the free
energy of the critical nucleus can be written
AG*
K (a r)
v surf^
1
[ AG(chem)+AG(strain)]

7
AG(chem) is negative whereas asur£ and AG(strain) are posi
tive. For a given reaction, AG(chem) can be calculated from
thermodynamic parameters by the method of Aarons on et al.
(1970) Attempts have been made to calculate crsur£ from
atomic bond models, but in general, the binding energies are
unknown. In the absence of a proven model, taken as the measured bulk interfacial energy. The validity
of this approximation is questionable. AG(strain) can be
calculated by the method of Eshelby (1957,1961).
If a nucleus forms in the stress field of a dislocation,
an additional energy term arises from the interaction of the
displacement field of the precipitate with the stress field
of the dislocation. We can then write AG* as
AG* =
K(a
surf)
[AG(chem)+AG(strain)+AG(int)]
AG(int) is negative and acts to reduce the positive AG(strain)
term, so that it represents a major part of the advantage
gained from nucleation at a dislocation. If AG(strain) is
large (i.e., the precipitate misfit is large), nucleation
at dislocations may be the only way the reaction can initiate.
In addition to the AG(int) term, a second advantage for
nucleation at a dislocation arises from the pre-exponential
6 term in Equation (2.1). This term is the frequency with
which atoms join the nucleus and depends on solute diffusion
to the nucleus and across the interface. Solute pipe-
diffusion along a dislocation core is always faster than

8
bulk diffusion, so that £ increases for nucleation at a dis
location. Also, solute diffusion to dislocations themselves
enhances the local concentration (e.g., Bullough and Newman,
1959) .
The task of calculating AG(int) is difficult, which is
the main reason why the theory of heterogeneous nucleation
at dislocations is less advanced than homogeneous nucleation
theory. In fact, there have been only six published attempts
to calculate AG* for nucleation at a dislocation. Cahn (1957)
made the first calculation. He assumed an incoherent pre
cipitate, an isotropic material, and completely neglected any
interaction term. Despite these simplifications, his model
was able to predict qualitatively some experimental observa
tions of nucleation at dislocations. Dollins (1970) calcu
lated AG* for coherent, spherical and disk-shaped nuclei at
a dislocation in an isotropic matrix. His work was reexamined
by Barnett (1971). Lyubov and Solov'Yev (1965) have given
the most complete treatment for calculating AG* for a coher
ent nucleus at a dislocation. Ramirez and Pound (1973)
attempted to include effects of the dislocation core energy
on nucleation, effects that were omitted from the other
models which use linear elasticity theory. An excellent
recent review of the present status of calculating AG* for
nucleation at dislocations has been given by Larch (1974).
None of the above treatments, however, have included effects
of elastic anisotropy, principally because the calculations
involved are extremely difficult. It can be concluded that

9
the presence of the dislocation stress field aids in reducing
the energy barrier to nucleation, but a rigorous calculation
of the effect is not yet available to provide an explicit
expression for the rate of heterogeneous nucleation at dis
locations .
It is instructive, however, to examine the order of mag
nitude of the terms in AG* to estimate the catalytic effect
of the dislocation. Typical values of AG(chem) are in the
range 1-5x10^ ergs/aA (20-100 cal/cm3). Values of AG(strain)
are in the same range as AG(chem) for particles with appre
ciable misfit. Values of o r obtained from bulk measure-
surf
ments are almost certainly too large since they relate to
incoherent interfaces, whereas critical nuclei whose sizes
o
are of the order of 10 !s of Angstroms should have coherent
interfaces. Estimates of coherent interfacial energies are
2
in the range 20-50 ergs/cm .
If we take AG(chem) = 2x10^ ergs/cm^, asur£ = 20
2 9 3
ergs/cm and AG(strain) = 2x10 ergs/cm then a spherical
nucleus with a diameter of 20A would have a chemical
-13
energy change and strain energy = 8x10 ergs each,
-13
surface energy = 24x10 ergs. Due to the problems
cussed above, no calculated values are available for
but it is estimated that it can be of the same order
free
and a
dis -
AG(int) ,
of mag
nitude as AG(chem) and AG(strain). Then if one assumes that
the surface energy is overestimated, as is likely (Barnett,
1973), the interaction energy due to the presence of the

dislocation can have an appreciable effect of lowering the
total free energy of the critical nucleus, whenever AG(strain)
is large.
2.2. Precipitation in the Al-Cu System
The aluminum-rich end of the Al-Cu phase diagram is a
eutectic system between the aluminum solid solution and
0-CuA19 (-53 wt.% Cu). Figure 2.1 shows the portion of the
diagram containing the a-solid solution region. Upon quench
ing from the solid solution region and aging, the equilibrium
precipitate is the b.c.t. 0-phase. The existence of three
metastable, transition phases, Guinier-Preston (G.P.) zones,
0" and 0, was established by the early x-ray work of Pres
ton (1938a,b,c) on the 4 wt.% Cu alloy, and by Guinier (1938,
1939,1942,1950,1952) on the 4 and 5 wt.% Cu alloys. The
solvus lines for these three transition phases are shown in
Figure 2.1. The positions of the 0" and 0 solvuses are due
to Hornbogen (1967). That of the G.P. solvus is due to Betn
and Rollason (1957) .
Guinier (19 38) and Preston (1938a) determined that the
G.P. zones are coherent, copper-rich clusters of plate-like
shape which form on {100} planes of the matrix. The most
reliable lattice parameters of the 0" and 0 phases are given
by Silcock ert al. (1953). 0" is complex tetragonal with
a = 4.04A and c = 7.8. It is coherent with the matrix and

Temperature
ii
Figure 2.1. The aluminum-rich end of the Al-Cu phase
diagram, including the solvus lines for
G.P. zones, 0", and 0' precipitates.

12
forms as platelets on {100} planes of the matrix. 0' is
complex tetragonal with a = 4.04A and c = 5. 8$.. It also forms
as platelets parallel to {100} matrix planes, and is initi
ally coherent on its broad faces and semi-coherent on its
edge. As it grows its broad faces become semi-coherent.
The orientation relationship for both 0" and 0' is {100}
PP^
|| {100} . and <100> || <100> .
11 matrix ppt 11 matrix
The tetragonal unit cell of 0 is shown in Figure 2.2.
There are 6 atoms/unit cell. The a-solid solution is f.c.c.
with a = 4.045, and has 4 atoms/unit cell. When the atomic
volumes are calculated for these two unit cells and compared,
it is found that the a+0' transformation involves a 3.95%
volume contraction. The resulting transformation strain can
be partially compensated if vacancies are generated by the
growing precipitates and supplied to the matrix.
The early x-ray work established the following precipi
tation sequence for quenching and aging below the G.P. solvus:
G.P. zones + 0" -* 0' + 0(CuA12).
However, as suggested by the x-ray work and later confirmed
by many transmission electron microscope (TEM) studies,
several of the reactions can proceed concurrently depending
on the quenching and aging procedures. In addition, TEM
investigations have clearly established the homogeneous or
heterogeneous nature of the various reactions. Nicholson
and Nutting (1958) resolved G.P. zones and 0" platelets in
the 4 wt.% Cu alloy and found them to be homogeneously dis
tributed in the matrix. It is now clear, however, that 0"

13
Figure 2.2.
The tetragonal unit cell of 0' (after
Silcock, Heal, and Hardy, 1953).

must be nucleated on G.P. zones (Lorimer and Nicholson, 1969;
Lorimer, 1970). If a sample is direct-quenched into the
region below the 0" solvus but above the G.P. solvus (Figure
2.1), no 6" forms. Howevei, if a sample is direct-quenched
below the G.P. solvus and then up-quenched into this region
and aged, 0" forms and its distribution is a function of the
size distribution of G.P. zones present before the sample
was up-quenched (Lorimer, 1970). Neither G.P. zones nor 0"
plays a role in the nucleation of 0'. Due to its misfit
strain, O' nucleates only heterogeneously in the presence of
a stress field in the lattice. It nucleates either at dislo
cations (Nicholson and Nutting, 1958), in the stress fields
of other 0' precipitates (Lorimer, 3968), or in the presence
of a macroscopic stress applied to the sample during aging
(Mosford and Agrawal, 19 74).
Numerous TEM investigations have confirmed the catalytic
effect of dislocations for nucleating 0'. It was suggested
early (Wilsdorf and Kuhlmarn-Wilsdorf, 1955; Thomas and
Nutting, 1956), and later confirmed by TEM, that only certain
0' orientations will nucleate at a given dislocation. This
is explained in terms of the misfit strain of the 0' platelet
and the Burgers vector of the dislocation. In Figure 2.3, it
is shown that the principal misfit around a 0' platelet is
normal to the {100} plane of the platelet. In Figure 2.4, it
is shown that a dislocation with Burgers vector a/2[110]
partially relieves the misfit strain around 0' platelets on
(100) and (010) whose misfits lie at 45 to the Burgers vector,

15
DIRECTION
OF MISFIT
9' PLATELET
a = b=4.04 c = 5.8 a
Ai-Cu MATRIX
a- b=c = 4.04
Figure 2,3. Diagram showing that the distortion of (001)
planes around a 9' platelet is normal to the
platelet (not to scale).
COMPRESSION
b= a/2 [110]
TENSION
[010]
[100]
if
[001]
MATRIX
ORIENTATION
Figure 2.4. Diagram showing that a dislocation with
Burgers vector a/2[110] relieves the misfit
around 0 platelets lying on (100) and (010).
It does not relieve the misfit around a
platelet on (001) .

16
but not for a 0' platelet on (001) whose misfit vector is
perpendicular to the Burgers vector. Hence, the (001) orien
tation gains no advantage by nucleating in the stress field
of the dislocation. Likewise for a dislocation with F=a[100],
only the (100) orientation of 0' should nucleate in its stress
field.
Many early TEM investigations of the 0' phase were con
ducted after long aging treatments at high temperatures in
the a+0' field (Figure 2.1). The resulting microstructures
contained a uniform distribution of large 0* platelets, and
it was initially concluded that these 0* platelets were
nucleated by a random distribution of pre-existing 0". How
ever, as it became clear that 0' nucleates at dislocations
and not at 0", the problem of how the random distribution of
0' could form by quenching and aging alone remained unresolved
until the work of Lorimer (1968,1970). Lorimer showed that
the a+0' reaction could propagate from an initially low dis
location density, introduced during quenching, by an auto-
catalytic nucleation mechanism. Early during the aging
period, the initial dislocations become saturated with 0'.
These platelets then grow into the matrix and produce their
own stress fields which aid the nucleation of more precipi
tates. With long aging, the reaction propagates in bands
spreading out from the dislocations to fill the structure with
a uniform distribution of 0' platelets on all three {100}
orientations. Until the present research, this was the only

17
reported mechanism whereby the a+0* reaction was found to
propagate from a low initial dislocation density.
2.3. Dislocation Climb
2.3.1. Quenched-In Vacancies and the
Chemical Climb Force
It is now widely accepted that vacancies can exist in
crystals in thermal equilibrium with the lattice. The equi
librium concentration of vacancies increases exponentially
with temperature according to the Arrhenius relation:
Cq = A exp(-E^/kT)
where A = an entropy factor,
E^ = the activation energy for forming a vacancy,
k = the Boltzmann constant, and
T = temperature.
Large supersaturations of vacancies can be retained in the
lattice by quenching rapidly from elevated temperatures.
During and after the quench, the excess vacancies diffuse to
sinks such as surfaces, grain boundaries, and dislocations
or they may cluster and collapse into vacancy disks bounded
by dislocation loops. The condensation of vacancies onto a
dislocation causes it to undergo positive climb. The greater
the supersaturation of vacancies, the greater is the driving
force for climb. A simple picture of dislocation climb by
vacancy annihilation is shown in Figure 2.5(a). Dislocations

18
can also climb by vacancy-emission (negative climb) and this
is illustrated in Figure 2.5(b).
Christian (1965) has suggested that the vacancy-
annihilating climb of a/2<110> dislocations in f.c.c. lattices
occurs easily only on the (111) and (110} planes. A necessary
condition for climb is that the Burgers vector has a component
perpendicular to the plane of climb. Therefore there are two
(111} and five {110} "planes of easy climb" for an a/2[110]
dislocation in f.c.c. Miekk-oja and Raty (1971) have con
sidered the choice of climb planes in terms of the chemical
climb force on each plane. This force arises from the super
saturation (or subsaturation) of vacancies (Bardeen and Her
ring, 1952). Its magnitude is proportional to (5~xu) where 5"
is the Burgers vector and u is the dislocation line direction.
Thus, according to Miekk-oja and Raty, a dislocation with
F=a/2[110] is affected by the maximum climb force, F ,
on the {110} plane perpendicular to F. It is not affected at
all on the {ill} and {110} planes containing b. And it is
affected by forces 0.82F and 0.5F on the two {111}
c y max c y max
and four {110} planes, respectively, which are inclined to F.
2.3.2. Theory of Dislocation Climb
In reality, dislocation climb is more complex than the
simple picture envisioned in Figure 2.5. The theory of climb
has been developed by Lothe (1960), Thomson and Balluffi
(1962), Balluffi and Thomson (1962), Friedel (1964), Hirth

19
oooooo
oooooo
oooooo
ooc- oo
OOO'OOO
oooooo
oooooo
oooooo
oooooo
OOO-QOO
ooo ooo
oooooo
oooooo
oooooo
000*000
OOOOQOO
OOO OOO
OOOOOO
a
000*000
000*000
ooosooo
000800
oooaooo
oooooo
b
000*000
oooooo
oooooo
ooo ooo
ooo ooo
oooooo
ooo@ooo
000*000
000*000
000*0 o
ooo ooo
oooooo
Figure 2.5. (a) Positive climb of an edge dislocation
by vacancy annihilation. (b) Negative climb
by vacancy emission (after Reed-Hill, 1973).

20
and Lothe (1968) and a general review of the mechanisms has
been given by Balluffi (1960).
Briefly, dislocation cLimb occurs by (1) the absorption
of vacancies onto the dislocation core, (2) diffusion of the
vacancies along the core to jogs, and (3) subsequent move
ment of the jogs by destruction of the vacancies. This
sequence is illustrated in Figure 2.6 for climb of an undis
sociated edge dislocation (a similar model applies for climb
by vacancy-emission). Then, according to Balluffi (1969),
the dislocation climb velocity is
v =
2ttD-, b^ [c (R) cv) j
r r2Zv L 2z ,-R-v -i
* ln(d]
where D.
vacancy diffusivity in the lattice,
magnitude of the Burgers vector,
c(R) = vacancy supersaturation at a large distance
R from the dislocation,
.0
vacancy concentration maintained in the
lattice in equilibrium with the jogs,
mean migration distance of a vacancy along
the core before jumping off, and
A
= jog spacing.
Seidman and Balluffi (1968) surveyed the available experi
mental data on climb rates and concluded that, in the pres
ence of moderate to large supersaturations, climb in aluminum
appears to be highly efficient. In other words, jog produc
tion and motion is sufficiently fast that the climb rate is
limited only by the diffusion of vacancies to the dislocation,
and the dislocation acts as a perfect line sink.

21
b
c
Figure 2.6. Diagram of vacancy processes associated with
climb of the extra half-plane of an edge
dislocation. Vacancies absorb onto the core
(a) diffuse along the core (b) and annihilate
at jogs (c). Subsequent motion of the jog
across the page moves the extra half-plane
up one atomic spacing.

22
2.3.3. Dislocation Climb Sources
The classical dislocation multiplication mechanism is
that proposed by Frank and Read (1950) whereby a dislocation,
pinned at each end, expands in its slip plane by glide.
Westmacott et_ al. (1959) observed dislocation sources in thin
foils of Al-4 wt.l Cu by transmission electron microscopy,
and interpreted them to be Frank-Read sources. Gulden and
Nix (1968) have observed similar sources in Al-4 wt.% Cu-
3 wt.% Si. Analogous to the Frank-Read mechanism, a dislo
cation multiplication mechanism which operates by climb was
proposed by Bardeen and Herring (1952) to act as a continuous
sink for excess vacancies. The Bardeen-Herring model for a
dislocation climb source is shown in Figure 2.7. Initially,
a straight dislocation between A and B has its slip plane
normal to the plane of the paper. Hence, it can move in the
plane of the paper only by climb. Condensation of vacancies
onto this line would move the dislocation out through the
sequence of positions shown. When the bottom segments of the
loop meet, they annihilate and rejoin as shown by the dotted
lines. The segment ABC is now free to repeat the process,
and there is left a vacancy loop outlined by the dislocation
ring. As long as there remains a vacancy supersaturation
in the region, this loop will expand, removing atoms from
the lattice, and the operation can repeat removing an indefi
nite number of planes.
Dislocation climb source configurations were first
observed by TEM in Al-Mg alloys (Westmacott et al. 1962;

23
4
Figure 2,7. The Bardeen-Herring model of a dislocation
climb source. An edge dislocation, pinned
between AB, has its slip plane normal to the
paper. It climbs in the plane of the paper,
by vacancy annihilation, through the succes
sive positions 1-4, rejoining at the bottom.
The portion ACB can then repeat the process
(after Bardeen and Herring, 1952).

Embury and Nicholson, 1963). Since then, dislocation climb
sources have been identified In aluminum (Edington and West,
1966) and a number of other aluminum alloys, including Al-Ag
(Edington and West, 1966), Al-Ag ternary alloys (Passoja and
Ansell, 1971), and Al-Cu (Boyd and Edington, 1971). They
have also been observed in other systems, including silicon
(Ravi, 1971) and NiAl (Marshall and Brittain, 1974). Often
the dislocation climb sources observed by TEM have small
particles at the center of the source loop. An example from
the present work is shown in Figure 4.2. Although the par
ticles are usually too small to be identified, it is thought
that they are insoluble particles existing at the solution
treatment temperature. It is generally believed that vacan
cies diffusing to the particle interface activate the source
which then operates to produce successive loops. The source
itself is often thought to be a portion of a misfit disloca
tion at the particle/matrix interface.
The only reported observation of dislocation climb
sources in binary Al-Cu alloys is that of Boyd and Edington
(1971). They observed source densities of about 3/grain.
These sources generated pure-edge loops on {110} habits with
a/2<110> Burgers vectors.

25
2.4. Repeated Nucleation on Dislocations
The concept of repeated precipitation on climbing dis
locations was first proposed by Silcock and Tunstall (1964)
to explain the occurrence of planar colonies of NbC precipi
tates on stacking faults in austenitic stainless steels.
In connection with the precipitate reaction, the stacking
faults were found to grow by the climb of a/3 Frank
partial dislocations bounding the fault. The transformation
to the NbC phase involves a 23% volume expansion, so that
the growing precipitates consume vacancies from the matrix
in order to relieve the transformation strains. Silcock and
Tunstall proposed that the Frank partial climbs by vacancy-
emission in order to feed vacancies to the transformation.
Thus the driving force for the dislocation climb is this need
to supply vacancies for the precipitate reaction. The prin
ciples of the Silcock-Tunstall model are outlined in Figure 2.8.
Repeated precipitation on climbing dislocations by this
mechanism has since been reported in a variety of systems,
including different steels, an iron-vanadium alloy, a copper-
silver alloy, superalloys, and semiconducting materials. The
phenomenon has been observed to occur on both partial and
total dislocations. For the sake of brevity, the list of
reports will not be given here, and the reader is referred to
the complete list in the recent paper by Nes (1974). In
every reported case to date, the precipitate phase has a
larger atomic volume than the matrix, thereby consuming

26
Figure 2.8. The Silcock-Tunstall model for repeated pre
cipitation of NbC in austenitic stainless
steel. The precipitates nucleate on Frank
partials (a). Movement of jogs, J, provides
vacancies for the precipitates to grow (b)
with the consequent climb of the dislocation
(c) The dislocation pinches off (d) and the
process repeats (e) (after Silcock and Tunstall,
1964).

27
vacancies during the transformation. In every case the dis
location climb has been reported to be vacancy-emitting in
order to supply the necessary vacancies.
Nes (1974) has expanded the original model of Silcock
and Tunstall (1964) into a more quantitative theory, which
was intended to account for the various features of repeated
precipitation in all the systems reported since 1964. The
fundamentals of the Nes theory are:
(1) Vacancies must be supplied to the transforming
particles in order to reduce the particle/matrix
mismatch.
(2) The subsequent particle growth causes vacancy-
emitting climb of the dislocation in order to
. feed the transformation.
(3) The particle growth/dislocation climb sequence
between conservative nucleations is controlled
by balancing the rate at which vacancies must
be supplied to the precipitates with the climb
rate of the dislocation.
(4) The particle is dragged some distance by the
dislocation before unpinning occurs.
(5) The rate controlling parameters in the kinetics
of colony growth are either (a) the atomic dif
fusion of the precipitating atoms, or (b) the
core (interface) self-diffusion, depending on
which has the highest activation energy.

28
Nes implied that this mechanism is applicable to repeated
precipitation on climbing dislocations in all systems, where
as in reality, it probably applies only when there is required
a mass balance of vacancies between growing precipitates and
climbing dislocations.
There is one report of repeated precipitation on climb
ing dislocations which has not been attributed to the above
mechanism. Embury (1963) observed that dislocations in Al-Mg
alloys were drawn around l^Al^ precipitate particles, pinched
off leaving loops, and climbed away under the chemical force
of a quenched-in vacancy supersaturation where the process
repeated. However, this process is reported to occur only to
a small extent.
2.5. Pertinent Electron Microscopy Theory
2.5.1. Two-Beam Diffraction Contrast Theory
In the transmission electron microscope, contrast arises
from differences in intensity scattered out of the incident
electron beam by Bragg diffraction from the crystal planes.
The best contrast from defects occurs under so-called "two-
beam" conditions. Owing to the large amount of tilt avail
able in commercial goniometer stages, the crystal can be
oriented so that the incident beam diffracts strongly only
from one set of lattice planes. Then approximately 95% or
more of the incident intensity is contained either in the

29
beam scattered in the forward direction (called the "trans
mitted" or "main" beam), or in the strong diffracted beam.
The electron image is usually formed by placing an aperture
around one of these beams and allowing it to pass through,
while the other beam is blocked by the aperture holder
(Figure 2.9). The unblocked beam is then magnified by
successive lenses and projected onto the fluorescent screen.
When the aperture is placed around the transmitted beam, the
image formed is called a "bright-field" image. When it is
placed around the diffracted beam, the image is called a
"dark-field" image. Contrast at defects arises in, say, the
bright-field image because the strain in the lattice around
the defect causes local distortions in the atomic planes
which lead to local changes in the intensity scattered into
the diffracted beam. This in turn leads to local variations
in intensity in the bright-field image. This is illustrated
for the case of a dislocation in Figure 2.10.
When a crystal of sufficient thickness is oriented very
close to the Bragg condition for one set of planes, there can
be a dynamic interchange of electrons between the two beams,
resulting from multiple scattering back and forth as the two
beams pass through the crystal (Figure 2.11). In order to
predict the intensities in the bright- or dark-field image,
it is necessary to describe mathematically the physical pro
cesses which go on in this dynamic interchange. The two-beam
dynamical theory of electron diffraction for a distorted
crystal was developed by Howie and Whelan (1961). Their

30
INCIDENT BEAM
Figure 2.9. Method for forming a bright-field image under
two-beam conditions. The transmitted beam is
allowed to pass through the objective aperture
while the diffracted beam is blocked by the
aperture holder.

31
INCIDENT BEAM
Figure 2.10. Diagram illustrating how contrast arises in
the transmitted and diffracted beams from
diffraction off the distorted planes around
an edge dislocation. Planes to the left of
the dislocation are tilted toward the Bragg
angle. Planes to the right are tilted away
from the Bragg angle.

32
INCIDENT
BEAM
TRANSMITTED
BEAM
Figure 2.11.
Diagram illustrating the dynamic interchange
of electron intensity between the two beams
resulting from multiple scattering events.

33
derivation is similar to the derivation of a two-beam theory
for x-ray diffraction by Darwin (1914). The Howie-Whelan
theory predicts the amplitudes T and S of the electron waves
in the transmitted and scattered beams, respectively, at
any point in the crystal. The formulation of the theory
uses a column approximation, whereby the crystal is imagined
to be divided up into parallel columns in the direction of
the incident beam. Dynamic interchange between T and S is
considered within a column, but not between neighboring
columns. This is a valid approximation because the Bragg
angles for high energy electron diffraction are small (^1/2).
The Howie-Whelan theory predicts the following coupled
pair of first-order differential equations for the variations
in T and S with depth in the crystal:
37 = (iri/£0)T + C tt i / fg) S exp(2risz + 2rig-R)
37 = ( tt i / EQ) S + (iri/ )T exp (-27risz-2TTig* R)
where
T = -amplitude of the incident beam,
S = amplitude of the diffracted beam,
z = depth in the crystal in the direction
of the incident beam,
s = parameter measuring deviation from the
Bragg condition,
g = the diffracting vector,
R = the local displacement field at depth z,
E = parameter related to mean refractive index
of the crystal, and
E = the extinction distance,
g

34
Each separate equation represents the variation in amplitude
of the electron waves in that beam as it passes through the
crystal. The first term in each equation represents the
intensity scattered in the forward direction for that beam.
The second term in each equation represents the intensity
scattered into that beam from the other beam.
In order to account for experimentally observed effects
of absorption, it is necessary to replace the quantities
1/£q and l/£ in the equations by the complex quantities
(1/Cq+/5q) and (l/£ +i/£g), respectively. One then obtains:
^ = Tri (l/C0 + i/C)T + ui (l/£ +i/ps exp(2TTsz + 27Tg* R)
Jj! = iri(l/£0+i/5)S + 1T (l/Cg+i/SpT exp (-27risz-2Trig-R)
Multiplying the amplitudes T and S by their complex conjugates
gives the relative intensities in the two beams at any point
in the crystal. In particular, when the intensity is calcu
lated at the bottom of all imaginary columns in the foil, it
predicts the image projected onto the viewing screen, since
no interaction occurs in vacuum once the beams exit the crystal
In the absence of any displacement field (R=0), or in the
presence of a fixed, rigid body displacement (R=constant), the
equations can be solved analytically for T or S, and the solu
tion predicts a uniform intensity over the bottom of the
crystal. When the displacement field R varies with depth, as
is the case around dislocations and other defects, the equa
tions can no longer be solved analytically, and numerical
methods must be used to obtain T and S.

35
The validity of the Howie-Whelan equations in predict
ing intensities which correspond to two-beam images has been
overwhelmingly demonstrated by the success of computer simu
lation techniques for matching defect images (Head et ad.,
1973). (See for example Section 4.2.1.)
2.5.2. Defect Identification from
Invisibility Conditions
Although the solution of the two-beam equations is not
straightforward for the case of defects with varying dis
placement fields, often the solution per se is not needed to
identify the defect from its image. Instead, it is often
possible to apply a simple criterion to identify defects in
the electron microscope. This criterion is based on the
fact that the term in the equations which gives rise to con
trast is the product gR. The diffraction vector g is the
reciprocal lattice vector normal to the diffracting planes,
so that the product g*K samples the magnitude of the distor
tion created in the diffracting planes by the displacement
field R. If a defect happens to cause no distortion in the
diffracting planes for a two-beam condition, then g-R=0 and
the two-beam equations predict uniform intensity everywhere
at the bottom of the foil. In other words, there is no con
trast around the defect and it is said to be "invisible" for
this diffraction condition. This criterion, applied to the
identification of dislocations, can be described as follows.
To a first approximation, the planes parallel to the Burgers
vector of a dislocation in an isotropic crystal are not

36
distorted. Then when the crystal is oriented so that one
such set of planes is in the reflecting position, the dislo
cation will be "invisible" in the image. The diffracting
vector is perpendicular to the diffracting planes, and there
fore to the Burgers vector for this condition. Hence, the
criterion for invisibility of a dislocation is the well-
known relation g*F=0. To identify the Burgers vector of a
dislocation, it is simply a matter of tilting the foil and
selecting various two-beam conditions until two diffraction
vectors g and g^, are found for which the dislocation is
invisible in the bright-field image. The Burgers vector
must be perpendicular to both g^ and so that it can be
determined from their cross product, i.e., 5"= (g^ ). How
ever, this technique is not capable of determining the
Burgers vector unambiguously, i.e., whether it is +F or -F.
Furthermore, the criterion g-F=0 for invisibility applies
only to screw dislocations where, in the isotropic approxi
mation, all sets of planes parallel to the Burgers vector
are undistorted. This is not so for an edge dislocation.
For a set of planes to remain undistorted by an edge dislo
cation, not only must g*F=0, but in addition, g must be
parallel to the dislocation line direction. Mathematically
this is written g-(Fxu)=0, where u is the line direction.
This is a very stringent condition which is seldom obtained
in the microscope. Thus, edge dislocations, or dislocations
with appreciable edge orientation, often exhibit strong
"residual contrast" when g*F=0, due to the (Fxu) term. For

37
this reason, practical experience in recognizing "residual
contrast" is necessary in order to identify dislocations from
the invisibility criterion.
The criterion g*F=0 for invisibility is valid only for
total dislocations, where the product g*F can be only zero
or an integer (since it is the product of a reciprocal lat
tice vector and a real lattice vector). For partial dislo
cations, g*F can take on the non-integer values 1/3, 2/3,
4/3, etc., in cubic lattices. Howie and Whelan (1962) deter
mined that partial dislocations are invisible when g-F=0 or
1/3 and are visible for all other products. Silcock and
Tunstall (1964) further determined that, for this to be
strictly valid, the deviation from the Bragg condition cannot
be too large.
The condition that a defect is "invisible" if its dis
placement field does not distort the reflecting planes can
be applied to identify certain small precipitates. For
example, in the case of 6' platelets in Al-Cu, the principal
misfit in the lattice caused by the platelet is normal to
the plane of the platelet (.Section 2.2). If such platelets
are too small to distinguish their shape, their orientation
can still be determined since they will be invisible when
ever g is perpendicular to the misfit vector.
2.5.3. Imaging Precipitates in the
Electron Microscope
Precipitates can be imaged by one or more of several
mechanisms in the electron microscope. A good description

38
of these mechanisms is given by Hirsch et ad. (1965, p. 336).
Those pertinent to this research will be outlined below.
(1) Strain contrast in the matrix. All coherent and
semicoherent precipitates, and most incoherent precipitates,
cause some strain in the matrix. These strain fields there
fore give rise to diffraction contrast effects in the matrix.
This can be the only mechanism for imaging very small pre
cipitates whose sizes are less than the resolution limit of
the microscope, but whose long-range strain fields are
greater than this limit.
(2) Misfit dislocation imaging. Semicoherent precipi
tates have misfit dislocations over their semicoherent inter
faces. The strain fields of these misfit dislocations can
cause strain contrast just as for isolated dislocations in
the matrix. Weatherly and Nicholson (1968) have investigated
the conditions for imaging misfit dislocations. Often small
platelets viewed normal to the platelets are imaged by the
misfit-dislocation loops around their edges. This is
referred to as "dislocation-ring" contrast.
(3) Structure factor contrast. According to Ashby and
Brown (1963), this contrast arises whenever a coherent pre
cipitate has a different structure factor from the matrix,
and thus a different extinction distance. A particle of
thickness At then increases the effective foil thickness in
columns passing through the particle, giving rise to an
intensity change relative to columns in the matrix. Depend
ing on the depth of the particles in the foil and the

39
relative values of the extinction distances in the particle
and matrix, the particles can appear either lighter or
darker than the surrounding matrix. Structure factor con
trast arises only within the limits of the particle boundary.
(4) Orientation contrast. This contrast mechanism
arises whenever a foil is oriented such that a certain set
of lattice planes in the precipitate is diffracting strongly,
whereas the matrix is diffracting weakly, or vice versa.
The contrast is of a uniform light and dark nature, typically
dark precipitates in a light matrix. Orientation contrast
can arise only when there is appreciable difference in crys
tal structure between the precipitate and matrix, i.e., when
the precipitates are semicoherent or incoherent. For example,
when the electron beam is parallel to the thin dimension of
large precipitate platelets, often certain lattice planes in
the precipitate will also be parallel to the beam. In this
case, the precipitate diffracts strongly. If the matrix is
not oriented for strong Bragg diffraction, the bright-field
image will show dark precipitates in a light matrix.
(5) Displacement fringe contrast. Displacement fringe
contrast arises when there is an abrupt change in the phase
of the transmitted and diffracted waves as they encounter a
thin sheet of precipitate which displaces the matrix planes
in opposite directions on either side of it. This displace
ment R around a typical semicoherent precipitate platelet is
normal to the plane of the platelet, and its magnitude is
given by

40
| R | = At<$ n | F |
1 n1 1 n1
where At = thickness of the platelet,
6 = precipitate misfit,
n = number of misfit dislocations at the
periphery of the platelet, and
E~ = Burgers vector of the misfit dislocations.
When this displacement is substituted into the equations of
the dynamical theory, the intensity of the transmitted beam
is found to oscillate with thickness (Whelan and Hirsch,
1957). Thus, when the precipitate platelet is inclined to
the electron beam, a fringe effect is observed. The so-
called stacking fault fringes are the limiting case of dis
placement fringe contrast.

CHAPTER 3
EXPERIMENTAL PROCEDURES AND MATERIALS
3.1. Specimen Materials
The four Al-Cu alloys used in this work were obtained
as rolled sheets from a previous research project. They were
prepared from 99.99% aluminum and 99.99% copper by double
melting in an induction furnace using a graphite mold. After
solidification, the billets were alternately cold-rolled and
annealed to reduce them to sheet form.
The target compositions were the 4, 2, 1, and 1/2 wt.%
Cu alloys. The nominal copper concentrations of the four
alloys were 3.85, 1.96, 0.99, and 0.5 wt.%, based on starting
weights before melting. The impurity content in the 3.85 wt.%
Cu alloy was determined by x-ray spectrographic analysis by
the Sandia Corporation, Albuquerque, New Mexico. The impurity
levels are given as ranges in Table 3.1. The barium level is
suspect as it was determined from only one line. The impurity
levels in the other three alloys were not determined. How
ever, since all four alloys were prepared from the same
starting materials, the other three probably had the same
impurity levels as the 3.85 wt.% Cu alloy.
The 1 wt.% and 1/2 wt.% Cu alloys were obtained as
rolled sheet, 0.038 inch and 0.034 inch, respectively. They
41

42
Table 3.1
Impurity Levels in the Al-3.85 wt.l Copper Alloy
Impurity
Weight ppm
Fe
5-25
Pb
<10
Si
1-10
Mo
<10
Mg
5-20
Ca
1-5
Ga
<10
Ba
10-40
were then cold-rolled to 0.005 inch for heat treatment. The
3.85 wtA and 1.96 wt.% Cu alloys were obtained as rolled
sheet, 0.004 inch thick. They were not reduced further
before heat treatment. Samples for heat treatment were cut
from the rolled foils to the approximate dimensions 1/8 x
1/2 x 0.004 inch. This was found to be a suitable size for
preparing electron microscope specimens after heat treatment.
3.2. Heat Treatments
All samples were solution treated for one hour at a
temperature in the a-solid solution range (Figure 2.1). Next,
they were either (1) direct-quenched to an aging temperature

43
above the G.P. solvus, or (2) quenched to a low temperature.
The samples given direct-quenches were aged for various times
from approximately one second to 24 hours, and then quenched
into room-temperature water. Samples quenched to low temper
atures were either prepared for electron microscopy without
further treatment, or they were up-quenched to a temperature
above the G.P. solvus and aged for various times. They were
then quenched into room-temperature water.
The solution treatments were conducted in a vertical
furnace in air. The temperature in the heat zone was con
trolled to within 2C. One end of the sample was clamped
in a stainless steel alligator clip attached to the bottom of
a one-half inch diameter stainless steel tube, and this was
inserted into the heat zone of the furnace. Before each
treatment, the temperature in the heat zone was determined by
inserting a thermocouple into a dummy stainless steel tube
suspended in the heat zone.
Quenching was achieved by dropping the specimen-stainless
steel tube assembly out of the bottom of the furnace into the
quench bath. For direct-quenches to the aging temperature,
the specimen was dropped into a Lauda Constant Temperature
Oil Bath, maintained at the aging temperature, to within
0.2C. A schematic diagram of the apparatus used for solu
tion treatment and direct-quenching is shown in Figure 3.1.
For quenching to low temperatures, the specimen was dropped
into one of several low temperature baths in place of the
oil bath. Following the quench to low temperatures, some

44
Figure
INSULATION-
ALUMINUM
BLOCK
CLAMP
2?
I
STAINLESS STEEL TUBE
1
$
1
VERTICAL
FURNACE
-SPECIMEN
ALLIGATOR CLIP
DOOR
T
3.1. Diagram of the apparatus used for solution
treatment and direct-quenching. The specimen,
attached to the stainless steel tube for weight,
is dropped from the solution treatment furnace
into the constant temperature, aging bath.

45
samples were immediately up-quenched into the constant tem
perature oil bath maintained at the desired aging temperature
Due to the wide variety of solution treatment temperatures,
aging temperatures, and aging times employed in this research
no table of heat treatments will be given here. Instead, the
specific heat treatment information will be given either in
the text or in the figure captions.
3.3. Electron Microscope Specimen Preparation
Electron microscope specimens were prepared from the
heat treated strips by electropolishing in a solution of 5%
perchloric acid in methyl alcohol. A polishing potential of
18 volts d.c. was used with a stainless steel cathode. The
electropolishing setup is shown in Figure 3.2. The beaker
containing the polishing solution was immersed in a bath of
dry ice and acetone to slow down the polishing reaction. The
solution was circulated at a slow speed with a magnetic
stirrer to keep it cold. Under these conditions, the polish
ing bath was maintained at -45C.
The specimen strip was held with locking tweezers and
polished by dipping the bottom end (approximately 1/8-3/16
inch) into the solution at a dipping rate of about 1/second.
Dipping was found to reduce edge attack and to give a rela
tively uniform polish. The voltage dropped to about 12 volts
during immersion. Total polishing time to obtain a suitable

46
5 l o
ACID
(+)
The electropolishing setup for preparing thin
foils for electron microscopy. Polishing is
accomplished by dipping the bottom end of the
specimen into the solution.
Figure 3.2.

47
thin area was about 15 minutes. When it was determined (by-
experience) that the specimen was nearly thin enough, the
stirring was stopped to avoid damaging the thin area. The
last 20-30 seconds of polishing was done by immersion and
agitation instead of dipping, since dipping to the last was
found to sometimes etch the thin foil edge. Polishing was
continued until the bottom edge appeared very ragged or until
small holes had broken through. Then the power was switched
off and the specimen was rapidly removed and plunged into a
beaker of cold acetone (99.81 pure) immersed in the dry ice-
acetone cooling bath. It was agitated for about 10 seconds
and then immediately placed under a stream of room-temperature
acetone from a wash bottle. After washing for about 30
seconds, it was allowed to dry in air. The initial wash in
cold acetone was necessary to remove most of the electrolyte
which rapidly etched the polished surface if allowed to warm
to room temperature.
The thinned, bottom portion of the sample strip was cut
off with an X-acto razor knife and mounted in a 3 mm, 75 mesh,
locking, double copper grid for viewing in the electron
microscope. Several more specimens could then be polished
from the same sample strip, if desired. However, the remain
ing bottom part of the strip was already polished quite thin.
To avoid etching this polished surface, the specimen was agi
tated for about one minute in the small beaker of cold ace
tone prior to repolishing.

48
3.4. Electron Microscopy
The thin foils were examined in a Phillips EM 200 elec
tron microscope operated at 100 Kv potential. A goniometer
stage with 45 and 30 tilt on two orthogonal axes was used.
Two-beam diffraction conditions were established for taking
all micrographs. To obtain two-beam conditions, the foils
were oriented close to one of the low index poles shown on
the Kikuchi line map for an f.c.c. crystal in Figure 3.3.
Use of this Kikuchi line map during specimen tilting, as
described by Head et al. (1973), enabled diffraction vectors
to be determined unambiguously in every case.

49
Figure 3.3.
Kikuchi line map over two adjacent stereo
graphic triangles for a face centered cubic
crystal (after Head e_t al_. 1973).

CHAPTER 4
EXPERIMENTAL RESULTS AND ANALYSES
4.1. I introduction
Evidence for repeated nucleation of the 0 phase on dis
locations was first observed in this research when the
Al-3.85 wt.% Cu alloy was quenched directly to aging temper
atures above the 6" solvus. Figure 4.1 shows a typical
microstructure resulting from direct-quenching and aging.
A brief description of the features and evolution of this
microstructure (with the facts to be established in this
chapter) is as follows. Dispersed throughout the foil are
densely populated colonies of small 6' precipitates. The
colonies are bounded either totally or partially by disloca
tions, some of which are out of contrast in this image. The
dislocations were generated and climbed during the quench
from the solution treatment temperature. As they climbed,
they nucleated and dispersed the 0' colonies in their paths.
All dislocations climbed during quenching and all nucleated
precipitate colonies. The precipitate colonies may (1) be
planar, (2) lie on smoothly curved surfaces, or (3) lie on
corrugated-shaped surfaces, depending on the climb paths of
the dislocations.
50

51
Figure 4.1. Typical microstructure resulting from quench
ing the Al-4 wt.% Cu alloy directly to the
aging temperature. The foil contains colonies
of small 0' precipitates, bounded by disloca
tions. (Heat treatment: S.T. 1 hour, quench
to 220C, age 5 minutes.)

5 2
The primary goal of this research was to determine the
mechanism by which repeated nucleation of O occurs in the
Al-Cu system. However, since this work is the first reported
observation of the phenomenon in Al-Cu alloys, a secondary
goal was to characterize thoroughly the various features of
the microstructures observed. The geometrical analyses are
reported in detail here.
The material in this chapter is developed much in the
way in which the experimental analysis was performed. First,
in Section 4.2, the nature and sources of the climbing dis
locations are established. Next, in Section 4.3 the precipi
tate is identified as 0' and characterized as to distribution
of orientations in the colonies. Section 4.4 contains
descriptions of the various geometries and some of the dif
fraction effects. Finally, Section 4.5 describes results of
experiments designed to determine the effects of different
parameters on the repeated nucleation process. Most of the
analyses for identification purposes were conducted on the
3.85 wt.% Cu alloy. Accordingly, all micrographs in this
chapter are from this alloy, except those in Section 4.5.5.
In addition, most micrographs in this chapter are from sam
ples quenched directly to aging temperatures. For the sake
of brevity, the copper concentration of the alloys is listed
In the figure captions as either 4, 2, 1, or 1/2 wt.%.

53
4.2. Nature and Source of the Climbing Dislocations
The dislocations which climbed during quenching can be
classified mainly into one of two categories according to
origin: (1) those generated at dislocation climb sources,
and (2) glide dislocations which subsequently climb (a third
category found in alloys quenched into oil or water at room
temperature will be discussed in Section 4.5.4).
4.2.1. Dislocation Climb Sources
Figure 4.2 shows micrographs from foils direct-quenched
to 220C and aged for short times before quenching to room
temperature. Present in the microstructures are configura
tions consisting of concentric dislocation loops. When
viewed edge-on, the loops are seen to be coplanar since
their traces are straight lines, as at points A in Figure
4.2(c). Tilting the foil confirms that these straight lines
are traces of coplanar loops. Concentric loops sectioned by
the thin foil leave straight-line traces with the foil sur
faces, e.g., at B-B in Figure 4.2(b) and (c). Often small
particles were observed at the center of the loops as in
Figure 4.2(b). The operation of dislocation sources has been
discussed in Section 2.3.3 and will not be repeated here.
It will now be established that these are climb sources, and
the Burgers vectors and habit planes of the loops will be
identified.
A typical source is shown in Figure 4.3. Several loop
habits have been generated at the source. We are concerned

54
c
Figure 4.2. Dislocation sources in Al-4 wt.% Cu direct-
quenched from 550C to 220C and aged for
8 seconds in (a) and (b) and one minute in (c) .

55
e
Figure 4.3. Series of micrographs for determining the
geometry and Burgers vectors of the source
loops. The beam direction is close to [101]
in (a), (b) and (c) to [112] in (d) and
to [001] in (e). (Heat treatment: S.T.
1 hour 550C, quench to 220C, age 8 seconds.)

56
in this analysis with the outermost loop and the one inner
loop which lie totally within the foil. The plane of the
foil was analyzed to be very close to (101) so that this loop
habit must lie on or very close to (101). Consider first the
three images (a), (b), and (c) taken about the [101] beam
direction. In each image the source loops on (101) exhibit
weak, residual contrast typical of "invisible images of
edge dislocations for which g*F=0, but g-(Fxu)^0 (Hirsch
ejt al_. 1965 p. 261). Those loop segments lying approxi
mately parallel to the g vectors, where g-(Fxu)~0, are in
visible. The loops are everywhere visible for the two
reflections, g=T3T and 220. From this analysis, the source
loops are identified as pure edge-dislocation loops lying on
(101) with F=a/2[101]. Since the Burgers vector is normal
to the plane of the loops, the loops must expand in this
plane by the process of climb. The sources are therefore
dislocation climb sources of the type observed by Boyd and
Edington (1971) in Al-2.5 wt.% Cu.
The loop habits of climb sources in these alloys were
observed always to be {110} (with one exception to be dis
cussed in Section 4.4). The typical source produced loops
on more than one {110} habit. Loops lying on as many as
five of the six possible {110} habits were observed at one
source. The typical source also generated more than one
loop on each habit. As many as five or six loops on one
habit were commonly observed, although the average number
varied with the heat treatment.

57
When a given foil was first examined, a technique was
used for rapidly determining if the dislocation sources were
indeed climb sources with {110} habits, or if some or all of
them might be dislocation glide sources which are known to
operate on the {111} slip planes in Al-Cu alloys (Westmacott
et al. 1959). This technique was to tilt the foil to {001},
{111}, and {101} orientations and, in each orientation, to
determine the number of different source habits viewed edge-
on together with the angles between these habits. For exam
ple, \\rhen a foil was tilted to the (001) orientation, two
edge-on habits at 90 apart were observed, Figure 4.4(a).
Since the [001] pole is parallel to two {110} planes at 90
to each other, and not to any {111} planes, those sources
are identified immediately as {110} climb sources. Likewise,
when viewed in the (111) orientation, three edge-on habits
at 60 apart were seen, Figure 4.4(b). Again, since the
[111] pole is parallel to three {110} planes at 60 to each
other and not to any {111} planes, the sources are identified
as climb sources. However, neither of these cases rules out
the possibility that other sources seen inclined to the beam
in these orientations might be glide sources lying on {111}
planes. Therefore, it was necessary to tilt to a {101} ori
entation. The [101] pole is parallel to two {111} planes at
70.5 and to only one {110} plane. In this orientation, only
one habit was ever seen edge-on, as shown in Figure 4.1.
There was no evidence that any of the sources found in these
foils were glide sources.

58
/
Figure 4.4. (a) Two edge-on habits of climb sources at 90
to each other in an (001)-oriented foil.
(b) Three edge-on habits at 60 to each other
in a (111)-oriented foil. (Heat treatments:
S.T. 1 hour 550C, quenched to 220C, aged
5 minutes.)

59
Occasionally, when viewing edge-on habits in the (001)
orientation, it was observed that two different habits did
not lie exactly at 90 to one another, although the habits
of other sources in the same field of view appeared to be
perpendicular, Figure 4.5. It is concluded that climb of
the loops is not necessarily confined strictly to the {110}
planes. This angular measurement between two adjacent habits
is more accurate for determining if the loops lie exactly on
(110) planes than are measurements made from the rotation
calibration between directions in the diffraction pattern
and those in the image.
It has now been established that the dislocation sources
in these foils are climb sources. However, the climb of
pure-edge dislocations can be either vacancy-annihilating or
vacancy-emitting. The former case removes lattice planes
from the crystal whereas the latter case adds interstitial
planes. It remains to be shown whether the source loops
climb by vacancy annihilation or emission, although intuitive
arguments favor vacancy-annihilating climb. For instance, it
is known that quenching produces large vacancy supersatura
tions, but negligible concentrations of interstitials. As
the temperature drops during quenching, the need for the
excess vacancies to diffuse to sinks would promote the growth
of vacancy loops and tend to annihilate any interstitial
loops. Therefore, one would expect that the large climb
sources operate by vacancy annihilation during quenching,
but this is not a sufficient proof. In fact, in the past

60
Figure 4.5. Two edge-on habits of climb sources lying
slightly off 90 from each other in an (001)-
oriented foil. (Heat treatment: S.T. 1 hour
550C, quench to 220C, aged 5 minutes.)
Figure 4.6. Climb source inclined through the foil. Two
loops are sectioned leaving dislocation arcs.
The arc at A was selected for computer match
ing to determine its Burgers vector. (Heat
treatment: S.T. 1 hour 550C, quenched to
220C, aged 4 seconds.)

61
only intuitive reasoning has been used to show that climb
sources in aluminum alloys operate by vacancy-annihilating
climb.
In the present work, the technique of computer matching
of dislocation images (Head et_ al. 1973) was employed to
establish that these loops climb by vacancy annihilation,
thereby removing planes locally from the lattice. This tech
nique is capable of determining unambiguously the Burgers
vector of a dislocation line segment, i.e., whether the
Burgers vector is +F or -F.
Figure 4.6 shows a climb source in a sample direct-
quenched to 220C and held only four seconds. This source
has generated two loops on a (110} habit inclined through the
foil, so that each loop is sectioned and leaves two arcs of
dislocation. The segment of the outermost loop at A is
reasonably straight and was selected for the computer match
ing experiment (the oscillations in the image are contrast
effects arising from the inclination of the dislocation
through the foil). From invisibility conditions, the Burgers
vectors of these loops were determined to be either a/2[Oil]
or a/2[Ol]. The loops are pure-edge and lie on (Oil). By
stereographic analysis, the line direction of segment A was
determined to be very close to [100] in (Oil) and the foil
normal was determined to be [313].
Six experimental images of segment A are shown in Figure
4.7 along with the corresponding computed images for
F=a/2[011] and F=a/2[Oil]. These six images represent

62
b=a/2[01f]
b=a/2[01l]
Figure 4.7. Six experimental and computed images of dislo
cation A in Figure 4.6. The line direction is
[100], the foil normal is [313], and the beam
direction is close to [101] in (a) and (b),
to [112] in (c) and (d) and to [111] in (e)
and (f) .

63
reflections from three non-coplanar beam directions, a neces
sary condition for uniquely identifying a dislocation by
computer matching (Head, 1969). From the rotation calibra
tion of the electron microscope, the exact orientation of
the g-vector was marked on each experimental image. Also,
from the known geometry of the computer program, the orienta
tion of the g-vector was marked on each computed image.
Thus the direction of the g-vector serves as a basis for
comparison when matching the features of the computed images
with those in the experimental images. Now for a given dif
fraction vector g, the image of a dislocation with Burgers
vector +F is identical to that of a dislocation with Burgers
vector -F after a rotation of 180 (Head e_t al_. 1973, p.
382). Clearly it can be seen from Figure 4.7 that the
Burgers vector of Segment A of the loop is a/2[Oil] and not
a/2[Oil].
By convention, the positive direction of the dislocation
line in the computer program is always taken to be acute to
the foil normal. Thus for the foil normal [313] the posi
tive direction of Segment A is [100] and not [100]. Also,
the computer program employs the finish-to-start, right-hand
(FS/RH) convention for establishing the direction of the
Burgers vector with respect to the positive sense of the
dislocation line, Figure 4.8. Thus, from the (FS/RH) conven
tion, and the absolute Burgers vector and positive line
direction of Segment A, the geometry of the two loops in
Figure 4.6 can be established, and this is illustrated

64
o
o
o
o
o
o
o
o
o
o
O O O O O O
a
oro_p. oe-o
o ob o o o o
b
O
O
O
o
o
o
Figure 4.8. Schematic representation of Burgers circuits
taken in a cubic lattice around an edge dis
location (a) and in perfect crystal (b) ?
illustrating the FS/RH definition of the
Burgers vector. The positive sense of the
dislocation line is out of the paper (after
Head et al. 19 7 3).
(011) PLANE
I
^5=a/4[0TT]
Figure 4.9. The geometry of dislocation climb source loops
in Al-Cu as indicated by the absolute sense of
the Burgers vector determined from computer
matching. The pure-edge loops expand in their
habit plane by vacancy-annihilating climb.

65
schematically in Figure 4.9. Clearly the loops are pure-edge,
vacancy loops which climb in their habit plane by vacancy
annihilation. It is concluded that these climb sources oper
ate during the quench to act as sinks for the excess, quenched-
in vacancies.
Now that it is established that the source loops climb
by vacancy-condensation onto the loops, one further experi
mental observation must be explained. The vacancy-annihilat
ing climb of a/2<110> dislocations in f.c.c. lattices is
generally believed to occur easily only on {111} and {110}
planes, removing one and two atom planes, respectively
(Christian, 1965, p. 363). Stacking faults were never
observed within these climb loops, even though the same loops
were examined on many different reflections, e.g., Figure 4.3.
The stacking of {110} planes in f.c.c. is ABAB (Figure 4.10).
Removal of a single {110} plane by vacancy condensation
behind a climbing a/2<110> dislocation would create a stack
ing fault. Two adjacent {110} planes must be eliminated to
avoid a stacking fau]^t. It appears, therefore, that the
source loops climb by condensation of vacancies onto two
adjacent {110} planes. Since the stacking fault energy in
dilute aluminum alloys is high, it appears to be energetic
ally favorable for the loops to expand in this manner. A
schematic cross-section through a climb source is shown in
Figure 4.11.
The nature of the source particles is undetermined.
Occasionally, sources were observed that nucleated at very

66
[ooi]
Figure 4.10. Diagram showing the stacking of atoms in
(110) planes in a face centered cubic crystal
(viewed normal to the planes). Atoms in the
third plane down lie in A positions. Removal
of a single plane of B atoms creates a stack
ing fault, A on A.

67
(110) PLANES
OOOOOo A ooOOOOOA
OOOQon 0o000000 qQOOOO b
oooo uooooooooo OOOOOA
OOOO1 oOOOOOOBOOon rOOOO =
00000o Ooo000000n00000
0000 Oo OOOO B
Figure 4.11. Schematic diagram of the cross-section through
a climb source loop on (110). The pure-edge
loops with b=a/2[l0] climb by vacancy conden
sation onto two adjacent (110) planes, thereby
avoiding creation of a stacking fault.

large spheroidized particles (-1/10 l/2p diameter), Figure
4.12. These particles were large enough to be analyzed on
a JEOL 100B Analytical Electron Microscope using a fine-
O
focused electron beam (approximately 1,000A diameter), and
a non-dispersive detection system for fluorescent analysis
of the emitted x-rays. The analysis of these large spheroids
identified them as pure lead. It is not known how lead
entered the sample material. However, such particles were
observed in only a few foils and nucleated only a small
fraction (<<1%) of the climb sources present. The typical
source particle was so small as to be barely visible or not
visible at all, Figure 4.2(b). Such particles were too
small for the x-ray analysis, but they are most probably not
lead. The chemical analysis of the 3.85 wt.% Cu alloy
(Section 3.1) showed no appreciable concentration of any
single impurity which would suggest a guess at the particle
nature.
Although the chemical composition of the source particles
is unknown, some observations were made about their dis
tribution. The climb sources were dispersed randomly in
most portions of the foils. Occasionally, local high densi
ties of sources were observed. In a few instances, sources
were observed evenly spaced in a straight line (Figure 4.13),
suggesting that the source particles were part of an impurity
stringer produced when the original cast alloy was rolled down.
The density of active climb sources in these foils varied
with heat treatment (Section 4.5). In the only other reported

69
Figure 4.12. Climb sources generated at a large, spherical
lead particle. (Heat treatment: S.T. 1 hour
550C, quench to 220C, aged 1 minute.)
Figure 4.13. Climb sources aligned in a row. (Heat treat
ment: S.T. 1 hour 550C, quench to 200C,
aged 5 minutes.)

70
observation of climb sources in Al-Cu alloys, Boyd and Eding-
ton (1971) observed a source density of about three per
grain in Al-2.5 wt.% Cu (although it is not stated, most
probably measured in the volume of a grain sectioned by the
foil; grain size not reported). Source densities many orders
of magnitude higher than this were observed in the present
foils. The maximum density was produced in a sample
quenched from 550C to 180C.. A micrograph of this foil is
shown in Figure 4.49(e). From this micrograph and the aver
age grain size (approximately 250y diameter), the active
source density was estimated to be approximately 6x10^ sources
per grain.
4.2.2. Glide Dislocations Which Climb
In addition to loops generated at climb sources, other
dislocations were observed which had climbed during quench
ing and nucleated precipitate colonies. Examples are shown
in Figure 4.14. Generally, these dislocations were long and
either smoothly curved or irregular-shaped, depending on
their climb paths. The micrographs in this section were
taken from foils aged long enough so that the precipitate
colonies are readily visible, thereby delineating the climb
paths of the dislocations. For the present, it is assumed
that the precipitate colonies were nucleated by the climbing
dislocations (this will be proven in Section 4.3). In Figure
4.14(a), the dislocation exits the foil surfaces at A and C,
and the trailing precipitate colony intersects one foil

71
Figure 4.14. Precipitate colonies nucleated at long, climb
ing dislocations. The dislocation in (b) is
out of contrast along ABC. (Heat treatment:
S.T. 1 hour 550C, quenched to 220C, aged
5 minutes.)

72
surface along the trace ABC. In Figure 4.14(b), the dislo
cation lies along ABC at the upper edge of the precipitate
colony, where it is "invisible." Some residual contrast can
be seen, however. The source of these dislocations is un
known. They could be (1) grown-in dislocations, (2) glide
dislocations which existed at the solution treatment tempera
ture prior to quenching, or (3) glide dislocations which were
generated at some source, probably grain boundaries, at the
onset of quenching. It is thought that most, if not all,
fall into categories (2) and (3), i.e., they were a/2<110>
glide dislocations on (ill) planes prior to climbing.
Figure 4.15, for example, shows two images of a precipitate
colony nucleated by one such long dislocation. In (a) the
precipitate colony is inclined through the foil. The curved
dislocation has been sectioned twice by the foil, leaving
two arcs, AB and CD, at the ends of the precipitate colony.
The Burgers vector of this dislocation was determined to be
a/2 [101]. The micrograph in (b) was taken after the foil
was tilted to the (111) orientation. Here the precipitate
colony is viewed edge-on and appears as a curved, dark line.
This indicates that the dislocation climbed on an irrational,
smoothly curved surface normal to the (111) plane. Since
(111) is a glide plane, it is reasonable to assume that it
was a curved, glide dislocation on (111) prior to climbing.
Such long glide dislocations were observed often to
have climbed on smoothly curved surfaces. This was easily
recognized by the curved intersections which the associated

73
Figure 4.15. Precipitate colony nucleated by climb of a
glide dislocation, initially on (111), with
D=a/2[10]. The dislocation has been sectioned
twice by the foil, leaving arcs AB and CD in
(a). The beam direction is close to [Oil] in
(a) and to [111] in (b) (Heat treatment:
S.T. 1 hour 550C, quenched to 220C, aged
5 minutes.)

74
precipitate colonies made with the foil surfaces, Figure 4.16.
Similarly, Miekk-oja and Raty (1971) observed repeated nucle-
ation of silver-rich precipitates from solid solutions of
silver in copper behind dislocations which were shown to be
a/2<110> glide dislocations on {111} planes before climbing.
They found that these dislocations subsequently climbed in
one of two different ways: (1) off the slip plane on smoothly
curved surfaces, similar to that described above, or (2) into
a crooked shape so that different segments of the disloca
tions climbed on different low-index planes intersecting the
original slip plane. They further showed that these low
index planes were of the types {110} and {ill}, i.e., the
planes of "easy climb" (Section 2.3.1) on which the chemical
climb force, from a subsaturation of vacancies, was the
greatest.
In the present research, the shapes of precipitate
colonies behind certain glide dislocations suggested that
different segments of these dislocations had climbed on
separate crystallographic planes also. The term "segmented
climb" shall be used here to refer to this mode of climb.
Micrographs of precipitate colonies apparently resulting from
segmented climb are shown in Figures 4.17 and 4.14(b). In
Figure 4.17, the dislocation between AB has climbed through
the lattice from left to right. The precipitate colony
nucleated by this dislocation is separated into bands of
precipitates. This effect is thought to be associated with
the climb of adjacent dislocation segments on separate

75
Figure 4.16. Precipitate colony exhibiting curved traces
of intersection with the foil surfaces, indi
cating a curved climb path of the nucleating
dislocation. (Heat treatment: S.T. 1 hour
550C, quenched to 220C, aged 5 minutes.)
Figure 4.17. Banded precipitate colony nucleated by "seg
mented climb" of the dislocation AB from left
to right through the foil. (Heat treatment:
S.T. 1 hour 550C, quenched to 220C, aged
5 minutes.)

76
crystallographic planes. The resultant precipitate colony
has a corrugated shape. No precipitation occurred in areas
between the bands, e.g., at C. Such precipitate-free areas
can also be seen in the climb path of the dislocation in
Figure 4.14(b). For some reason the dislocation is ineffec
tive in nucleating precipitates in these regions of its
climb path. A possible explanation is that precipitation
occurs readily on those segments of the dislocation which
climb on the separate crystallographic planes, but not on
those curved arcs of the dislocation which bridge the gaps
between these planes. This will be discussed further in
Section 4.4.6.
The segmented climb of initial glide dislocations was
observed only in the Al-3.85 wt.% Cu alloy. An attempt was
made to determine the planes upon which segmented climb
occurred by trace analysis of the intersections of the pre
cipitate colony with the foil surface, but this proved to be
impractical for two reasons. First, as pointed out by
Miekk-oja and Raty (1971), the possible planes of easy climb
can be numerous, i.e., six {110} and four {111} planes, so
that the trace analysis is best accomplished by using single
crystals cut to special orientations. Such crystals were
not available in this research. Secondly, the traces of the
intersections of the bands of precipitates with the foil
surfaces were never well defined, a condition which leads to
poor accuracy in the stereographic analysis.

77
There is evidence that the tendency for glide disloca
tions to climb either on smoothly curved surfaces or to seg
ment and climb on different planes depends upon the line
direction of the dislocation before climb began. For example,
in Figure 4.18 the curved dislocation along ABC has nucleated
a smoothly curved precipitate colony between A and B, and a
corrugated colony between B and C. Presumably, the initial
line direction of the dislocation segment between B and C
was favorable for rapid climb onto the different planes of
easy climb.
4.3. Identification and Characterization
of thePrecipitate Phase
Figure 4.19(a) is a diffraction pattern in the exact
(001) matrix orientation taken from the group of precipi
tates shown in Figure 4.19(b). The image quality in (b) is
poor since the exact (001) orientation is a many-beam condi
tion. The geometry in Figure 4.19(b) is as follows. The
foil has sectioned three {110} habits of dislocation climb
sources, numbered 1, 2 and 3. These habits are viewed edge-
on in this orientation. Habits 1 and 2 lie on (110) while
habit 3 lies on (110). Small precipitate platelets are dis
persed over the three habits. Two orientations of precipi
tates are present lying parallel to the (100) and (010)
matrix planes. The precipitate reflections in the diffrac
tion pattern are streaked in the <10Q> directions owing to

78
Figure 4.18. A glide dislocation which climbed on a
smoothly-curved surface between A and B,
and underwent segmented climb between B
and C. (Heat treatment: S.T. 1 hour
515C, quenched to 220C, aged 5 minutes.)

79
Figure 4,19.
(a) (001) diffraction pattern showing
tate reflections, taken from the area
foil shown in (b). (Heat treatment:
1 hour 550C, quenched to 220C, aged
minutes.)
precipi-
of the
S.T.
30

80
the relaxation of the Laue condition along the thin dimension
of the platelets. The diffraction pattern can be analyzed
on the basis of three superimposed patterns shown in Figure
4.20 (a)-(c), where we consider only the lower right-hand
quadrant of the pattern. The pattern in (a) is the (001)
matrix pattern. The pattern in (b) is indexed on the basis
of two 9' orientations parallel to (100) and (010) matrix
planes, using the lattice parameters of 4.04A and 5.8 for
0' (Section 2.2). The remaining reflections in (c) are due
to double diffraction from the matrix {200} and {220} beams.
Double diffraction is a common occurrence in foils containing
precipitates with dimensions much smaller than the foil
thickness. The composite pattern, shown in (d), matches the
experimental pattern. Thus the precipitates are positively
identified as the 0' phase, in agreement with the known fact
that 0' is the only metastable phase which nucleates on dis
locations in Al-Cu.
In the present research, conditions were chosen to in
sure that the 0' phase was the only precipitate phase present
after heat treatment. Its distribution was always associated
with the climbing dislocations.
The appearance of the precipitate colonies at high
magnifications is shown in Figure 4.21. These are typical
colonies nucleated on dislocation climb sources. The colony
in (a) was nucleated on the (101) source habit in the plane
of the foil. In (b) five different {110} habits were gener
ated at the same source particle and have nucleated

81
oooM
0
O
200m
oooM
0 -
I
o -
Figure 1
020,
O
M
0
220M
oooM
0
002!|
101c
I
'2
002^ 101^
0
200^
103^
I
20CW
e91 i 2ei o
103^
o
i
o
000M
0
I
I

- I -
0
!
I
o
.20. Analysis of the lower, right-hand quadrant of
the diffraction pattern in Figure 4.14(a).
(a) (001) matrix pattern; (b) precipitate
reflections from 0' platelets lying parallel
to (010) matrix planes (0^), and to (100) matrix
planes (6?) ; (c) double diffraction from matrix
200 and 220 reflections; (d) combination of
(a), (b) and (c) Compare with Figure 4.14(a).

Q'jA
82
Figure 4.21. (a) and (b) Appearance of typical precipitate
colonies generated on the {110} habits of
climb sources. (c) and (d) Schematic diagrams
illustrating the geometry of the colonies.
See text for description. (Heat treatment:
S.T. 1 hour 550C, quenched to 220C, aged
5 minutes.)

83
d
Figure 4.21. Continued.

84
precipitate colonies. One habit lies in the plane of the
foil at A, one habit is viewed edge-on at B, and three other
habits at C, D and E are inclined through the foil. The
geometries of these sources are shown schematically in Figure
4.21(c) and (d) .
Often the best condition for imaging the precipitates
in a colony was also a condition for "invisibility" of the
dislocation loop bounding the colony, e.g., Figure 4.21(a).
This was particularly true when examining colonies on dis
location climb sources where the best projected view of a
colony was obtained with the beam oriented approximately
normal to the colony, and hence to the bounding source loop.
However, since the source loops are pure-edge with their
Burgers vectors normal to their habit planes, any two-beam
reflection selected to view the precipitate colony in this
orientation has its g-vector perpendicular to the Burgers
vectors of the loops. Thus the loop will be "invisible" when
viewed normal to its habit plane. This is the case in Figure
4.21(a) where g*£T=0 for the loop bounding the precipitate
colony and only residual contrast due to the pure edge-nature
of the loop is detected.
Likewise in Figure 4.21(b), the dislocations bounding
the colonies at A, C and D are invisible, whereas the one at
E is visible. A consistent interpretation of the geometry
of this source array is as follows. The beam direction is
close to [101] and g=lll for this image. Habit A lies on
the (101) plane of the foil (see Figure 4.21(d)). The

85
Burgers vector of its source loop, being pure edge, is
a/2[101], Thus the loop is invisible for g=ll. The source
loops bounding the habits C and D are also invisible. These
dislocations must have the other two a/2<110> Burgers vectors
which cause invisibility for the 111 reflection, namely,
a/2[Oil] and a/2[110]. Habit B lies on the (101) plane
parallel to the beam and is viewed edge-on. The Burgers
vector of its source loops must be a/2[01]. The source loop
bounding habit E must have one of the two remaining a/2<110>
Burgers vectors, namely, a/2[110] or a/2[Oil], both of which
would be visible for the 111 reflection. The dislocation is
visible at E.
It is difficult to determine whether the smallest pre
cipitates in these colonies are imaged by strain contrast in
the matrix or by structure factor contrast (Ashby and Brown,
1963). The larger precipitates in a colony are imaged by
the dislocation loops bounding the periphery of the platelets
(Section 4.4).
In Section 2.2, it was pointed out that because of dis
location strain effects, only two 0' orientations will nucle
ate on any given a/2<110> dislocation. The missing orienta
tion has its principal misfit (normal to the plane of the
platelet) perpendicular to the Burgers vector of the dislo
cation so that its strain field is not relieved by the stress
field of the dislocation. Careful examination of a number
of precipitate colonies at dislocation climb sources revealed
that only two 0' orientations were present in any given colony.

86
The missing orientation was always that {100) orientation
whose misfit would be perpendicular to the Burgers vector of
the source loop bounding the colony. This is illustrated in
Figures 4.21(b), 4.22 and 4.23.
First, Figure 4.22 shows bright and dark field images of
several precipitate colonies on climb sources in a foil whose
normal was close to [101]. The foil was oriented with the
electron beam close to [101]. The dark field image was taken
from a precipitate reflection from the (010) 8* orientation
parallel to the beam. Climb source A (and its precipitate
colony) lies on (101), and its source loops, being pure edge,
have Burgers vectors a/2[101]. At B, two other source
habits lie on (110) planes inclined to the foil. The Burgers
vectors of their source loops were not determined but they
cannot be a/2[101]. Now, if all three {100} orientations of
0' were present in the colony at A, the orientation imaged
in the dark field would be observed throughout source A as
in the sources at B. The misfit of this missing (010) orien
tation is perpendicular to the a/2[101] Burgers vector of
the source loops at A (which are "invisible" in this image).
The few precipitates on (010) in the middle of source A in
the dark field image were found to lie within small source
loops lying on other {110} habits, when this source was
examined in another orientation.
Next, in Figure 4.21(b), the source habit at B lies on
the (101) plane and is viewed edge-on. The Burgers vectors
of its source loops, being pure edge, must be a/2[01].

87
i
Figure 4.22. Bright-field and dark-field images of precipi
tate colonies on dislocation climb sources.
The colony at A lies in the (101) plane of the
foil. The colonies at B are inclined through
the foil. The dark-field was taken from a
precipitate reflection from the (010) 0' orien
tation lying parallel to the [101] beam direc
tion. (Heat treatment: S.T. 1 hour 550C,
quenched to 220C, aged 5 minutes.)

88
The (010) 0' orientation, which is parallel to the beam
direction and also viewed edge-on, is not present in habit B,
although it is easily detected in habits C and E. Again,
this is the 0' orientation whose misfit is perpendicular to
the a/2[101] Burgers vector of the source loops bounding
h ab i t B.
The fact that 0' platelets lying parallel to the beam
can indeed be seen if present in precipitate colonies viewed
edge-on is shown in Figure 4.23. This micrograph was taken
with the beam oriented near [001]. The two edge-on habits
at A and B lie on (110) planes so that the bounding disloca
tion loops have Burgers vectors a/2[110]. Both the (010)
and (100) orientations of 0' platelets can be clearly seen
dispersed along the habits. Note that these are the two 0'
orientations whose misfits are not perpendicular to the
Burgers vector of the bounding dislocation loops and are
therefore favored to be nucleated by the loops.
All three 0' orientations were never observed in a given
precipitate colony. As pointed out above, the missing orien
tation was always that whose nucleation is not aided by the
stress field of the dislocations bounding the colony. This
was true for precipitate colonies generated by both climb
sources and glide dislocations which climbed. This evidence
leads to the important conclusion that the precipitates must
have nucleated in the stress field near the dislocations as
they climbed through the lattice, and not at some later time
when the influence of the dislocation was no longer present,

89
Figure 4.23. 0' precipitate colonies on climb source habits
A and B which lie parallel to the beam. Two
edge-on orientations of 0' platelets, (100)
and (010) are clearly visible in colonies A
and B. The beam direction is close to [001].
(Heat treatment: S.T. 1 hour 550C, quenched
to 220C, aged 30 minutes.)

90
e.g., during aging. Since the dislocation climb sources are
known to have operated during the high-temperature part of
the quench, the precipitates must have nucleated during
quenching.
Further evidence to support this conclusion was obtained
by in situ aging experiments in the electron microscope. The
results of these experiments are shown in the micrographs of
Figure 4.24. These are images of a foil from a sample solu
tion treated for one hour at 550C, quenched into oil at
220C and held only four seconds at 220C, then water-quenched
to room temperature. After electropolishing, the foil was
placed in the heating stage of the microscope. The micro
graphs in Figure 4.24(a) and (b) were taken prior to heating.
In (a), a dislocation climb source is viewed normal to its
(Oil) habit of source loops which are "invisible" in this
orientation and imaged by residual contrast. Since this is
a climb source, we know that these dislocations climbed
during quenching. A long, crooked dislocation, which was
most probably a glide dislocation prior to quenching, is
shown in (b). Its crooked shape is the only indication that
it may have climbed during quenching. Now, if nucleation
of the 0' precipitates does occur as the dislocations climb
through the lattice during quenching, then the precipitates
must already be present in the foil in (a) and (b). However,
the four-second aging time at 220C was insufficient to cause
the precipitates to grow to visible sizes. The foil was then
heated to 230C in the microscope. After nine minutes at

91
(a)
(b)
Figure 4.24. (a) A dislocation climb source, imaged by
residual contrast, and (b) a long glide dis
location in a sample quenched from 550C to
220C and held only 4 seconds before quench
ing to room temperature.

92
(d)
Figure 4.24. Continued. (c) and (d) Micrographs o£ the
same dislocations in (a) and (b) after aging
9 minutes at 230C in the electron microscope.
The random precipitation at B has occurred at
the foil surfaces. The precipitate colonies
associated with the dislocations are now
clearly visible at A.

93
230C, these dislocations were rephotographed and are shown
in (c) and (d). Clearly the precipitate colonies existed in
the climb paths of the dislocations after quenching, and the
nine-minute aging treatment has caused growth to visible
sizes, points A in the micrographs. Careful observation is
required to see the edge-on 0' platelets at A in the precipi
tate colony in (d). It is now apparent that this dislocation
underwent segmented climb (Section 4.2.2). The random pre
cipitation at points B throughout the micrographs in (c) and
(d) results from 0' precipitation on the foil surfaces, an
effect which is known to happen when thin foils of Al-Cu
alloys are heated in the electron microscope (Thomas and
Whelan, 1961).
From the pictures in Figure 4.24 alone, the argument
could be made that solute atoms may have segregated to the
dislocations and were then left behind when the dislocations
climbed away, creating supersaturated layers of copper in
the climb paths. Then aging at 230C caused nucleation and
growth of the precipitates from the supersaturated layers.
However, if this were the case, all three orientations of 0'
should nucleate in a given precipitate colony, and this is
inconsistent with the observations discussed above.

94
4.4. Further Geometric Analyses
The previous two sections established information about
the dislocation climb and precipitation which is basic to
understanding the details of this section. Included here are
(1) descriptions of the various geometrical aspects of the
precipitate colonies, (2) explanation of some diffraction
effects, and (3) other unusual features of the microstruc
tures, in addition to repeated precipitation, which have not
been previously reported. All these descriptions would prove
useful to someone examining these types of microstructures
for the first time.
4.4.1. Distribution of Precipitates in
Colonies at Climb Sources^
We consider the distribution of precipitates in a colony
on a single {110} habit of a climb source. A typical colony
is shown in Figure 4.25. The distribution of precipitates
is twofold. First, there is a densely-populated region of
small precipitates which covers approximately the central
three fourths of the colony. Secondly, there are two regions
marked A on either side of the colony which are comprised of
rows of somewhat larger precipitates extending from the
outer dislocation loop into the colony. These rows of pre
cipitates shall be referred to as "precipitate stringers."
The spacing between stringers is quite uniform. The dense
precipitation is not present in the stringer regions. The
stringers are always aligned in the <100> direction which is

95
Figure 4.25. A typical precipitate colony on a {110} habit
of a climb source exhibiting a central region
of dense precipitation, and two regions at A
of precipitate stringers aligned along the
[100] direction in the plane of the colony.

96
contained in the {110} plane of the colony. The dislocation
loop bounding the colony bows out locally between the stringers
and is generally smoothly curved in those regions away from
the stringers. The nature of the stringers will be discussed
in detail in the next section.
There were five basic shapes of the precipitate colonies
on climb source habits observed in these foils. These five
shapes are shown schematically, with corresponding micro
graphs, in Figure 4.26. Shapes I and II were the most com
monly observed. The typical colony with shape I was either
circular or slightly elliptical in the [001] direction, and
the region of dense precipitation was essentially continuous
over the center of the colony. Shape II is similar to Shape
I, except that there is a figure-eight-shaped region in the
center which is void of precipitates. The experimental
observations suggest that as the diameter of a colony with
Shape I increases, it will tend toward Shape II. At present,
there is no satisfactory explanation for the figure-eight -
shaped region void of precipitates. Shape III is very elon
gated in the [001] direction, and the regions of stringers
extend out beyond the projected sides of the ellipse. Such
colonies were observed only occasionally. In Shape IV, the
dislocation loop bounding the colony bows out along [110] at
two points. Narrow precipitate-free zones extend in this
direction from the center of the colony to these points.
The discussion in Section 4.1.5 may account for the tendency
to bow out to points, but there is no satisfactory

97
Figure 4.26. Five basic shapes of 0' precipitate colonies
nucleated on {110} habits of climb sources in
Al~Cu. All are oriented with reference to the
crystal directions at left.

98
explanation for the precipitate-free zones in these direc
tions. In Shape V, there is a well-defined boundary where
precipitation begins behind the climbing source loop. Inside
this boundary the region is void of precipitates (except
where another interior loop has climbed). A possible expla
nation for this shape is as follows. Different climb sources
become active in nucleating loops at different times during
quenching. Any climb of the loops which occurred before the
temperature passed below the 0' solvus temperature does not
nucleate precipitates. Further climb below the 0' solvus
temperature generates ring-shaped, precipitate colonies
behind the loops.
For the case where several concentric source loops are
generated on a given {110} habit, each successive, interior
loop nucleates the same basic colony shape on a smaller
scale, e.g., Shape IV in Figure 4.26.
4.4.2. Geometry of the Precipitate Stringers
One of the early problems to be solved concerned the
identification of the defects in the stringers. The stringers
were already visible in samples quenched to the aging temper
ature and held for very short times (Figure 4.27) although
the general precipitate colony was not yet visible. The
stringers appear as rows of small dots, so that the contrast
cannot be interpreted further at this stage. In samples aged
for longer times, the stringers are imaged as rows of small
dislocation loops, Figures 4.21 and 4.25.

99
Figure 4.27. Stringers of small dots along [010] in a
sample quenched from 550C to 220C and
aged only 8 seconds.

100
Three possible mechanisms were considered for the gener
ation of stringers of small loops behind the climbing source
loops. These were:
(1) Small loops were pinched off in some regular
manner from the climbing source loops.
(2) Vacancy debris was left behind the climbing
source loops from some regular arrangement of
superjogs or kinks. This debris then coalesced
into small loops.
(3) 0' precipitates were nucleated repeatedly in
regular rows. The small loops would be misfit
loops around the periphery of the platelets.
It is shown in the next section that the Burgers vectors of
the small loops in stringers are not the same as those of the
climbing source loops. Therefore, they are not pinched-off
loops and we can omit possibility (1). For the moment, we
disregard possibility (2) and show that the stringers are
definitely composed of rows of 0' precipitates.
When foils were viewed along beam directions parallel to
(l00) planes, 0' platelets were definitely imaged edge-on in
some stringers, as at A and B in Figure 4.28(a). This is
confirmed in the dark field image, taken from a 0' reflection,
in Figure 4.28(b). Other stringers at C contain 0' platelets
on another {100} orientation which is not parallel to the
beam. In this case, the precipitates are large enough to
exhibit displacement fringe contrast (see Section 2.5.3).
Another dark field image at high magnification, Figure 4.29(b),

101
Figure 4.28. (a) Bright-field and (b) dark-field images of
a foil oriented so that the stringers at A and
B are clearly imaged as 0' platelets viewed
edge-on. The precipitates at C are inclined
to the beam and exhibit fringes. (Heat treat
ment: S.T. 1 hour 550C, quenched to 220C,
aged 5 minutes.)

102
shows that the stringers contain separate, but closely-
spaced, precipitates oriented parallel to the beam. Hence,
the stringers are definitely composed of 0' platelets, and
the dislocation loops are assumed to be misfit loops at the
platelet peripheries (see Section 4.4.3).
Consider now mechanism number (2) above, requiring
vacancy debris to be left behind the climbing source loops.
Boyd and Edington (1971) observed small dislocation loops
lying just inside large, climb source loops in Al-2.5 wt.% Cu.
These were analyzed to be prismatic edge-loops with all pos
sible a/2<110> Burgers vectors. They proposed that the small
loops were present from condensation of vacancy debris gener
ated by the motion of edge jogs and screw kinks on the climb
ing source loops. In the present case, such small loops
could act as nucleation sites for the precipitate stringers.
If this were so, it would be expected that all three 0' orien
tations could be nucleated since the small loops had all
possible a/2<110> Burgers vectors. It was shown earlier
that only two of the three possible 0' orientations nucleated
in any given precipitate colony. The vacancy debris mecha
nism therefore does not explain the experimental results.
The origin of the precipitate stringers will be discussed in
Chapter 5.
As mentioned previously, the stringers in precipitate
colonies on climb sources always lie along the <100> direc
tion in the plane of the colony, Figure 4.30. Typically, the
boundaries of the two stringer regions are fan-shaped. That

103
b
Figure 4.29. (a) Bright-field and (b) dark-field images
at high magnification showing that the pre
cipitate stringers are actually composed of
separate but closely-spaced 0* platelets.
(Heat treatment: S.T. 1 hour 550C, quenched
to 220C, aged 5 minutes.)

104
is, the stringers at the center of these regions extend
farthest into the middle of the colony, points A in Figure
4.30. The stringers become progressively shorter in going
to the ends of the stringer regions, points B.
As aging time at temperature is increased, the precipi
tates in a stringer grow and probably coalesce, so that
stringer geometry evolves through the sequence of shapes
shown in Figure 4.31. The outermost precipitate tends to
grow into a Y-shape along the bowed-out, climb source loop.
A given source loop always nucleated stringers of the
two 9 orientations compatible with its Burgers vector.
However, platelets of only one orientation were nucleated in
any given stringer, Figure 4.29. Thus, there is some geo
metrical restriction about the origin of a stringer at or
near the dislocation loop which favors repeated nucleation
of only one 0 orientation.
Measurements were made of the average spacing between
stringers, and the average spacing between precipitates in a
stringer. The former were made in precipitate colonies
viewed normal to their {110} habits so as to obtain the true
projected spacing. The latter were made on the dark field
image in Figure 4.29(b). This sample was aged for a short
enough time that coalescence of precipitates had not yet
begun. The average spacing between stringers, measured normal
to the <100> direction was found to be 0.096y (960). The
average spacing between precipitates in a stringer was found

105
Figure 4.30. Typical precipitate colony on a climb source
illustrating the fan-shaped boundaries of the
precipitate stringer regions. (Heat treatment
S.T. 1 hour 550C, quenched to 220C, aged
5 minutes.)
Figure 4.31. Evolution of the geometry of precipitate
stringers at climb sources with increasing
time at the aging temperature.

106
to be 0.0 37y (370) or about one-third of the spacing between
stringers.
Precipitate stringers were also observed in 0* colonies
nucleated by glide dislocations which climbed, Figure 4.14(a).
The average spacing between these stringers was in good
agreement with the spacing of stringers at climb sources.
Stringers were not present at glide dislocations that under
went segmented climb, Figure 4.14(b).
A further feature of Figure 4.14(a) is that a well-
defined boundary exists between stringers and dense precipi
tation. Since the spacing between stringers is larger than
the spacing between precipitates in the region of dense
nucleation, it is unlikely that the latter nucleated at the
same preferred sites as the stringers. In other words, there
has been a change in the mode of nucleation at some point
near the end of the quench.
4.4.3. Determination of the Burgers Vectors of
Small Loops Within Precipitate Colonies
It was stated previously that the small loops visible
in a precipitate colony are loops at the periphery of 6'
platelets. It will now be shown that the diffraction con
trast at these loops is consistent with this hypothesis, and
that these are not loops formed by other possible mechanisms.
First, it is known that the misfit between matrix and pre
cipitate planes at the peripheral edge of 6' platelets is
accommodated by the presence of a<100> type edge-dislocation

107
loops around the platelets (Weatherly and Nicholson, 1968;
Laird and Aaronson, 1968).
Figure 4.32 shows a precipitate colony generated on a
climb source in the (101) plane of the foil. The source loop,
being pure-edge, has Burgers vector a/2[101], so that it is
invisible for this reflection. Consistent with this Burgers
vector, the precipitate colony contains 0' platelets lying on
(100) and (001) planes. If some of the platelets are large
enough, we should expect to see small loops with Burgers
vectors a[100] or a[001] at their peripheries. In Figure 4.32,
four stringers of loops are located with arrows and marked
either 1 or 2. The Burgers vectors of these loops were deter
mined by invisibility conditions from a number of two-beam
images taken for different foil orientations. Selected images
are shown in Figure 4.33, and the visibility data are summar
ized in Tab le 4.1.
Table 4.1
Summary of Visibility Data for the
Images of Figure 4.33
g=
111
020
ill
220
202
022
Source Loop
I
I
I
V
I
V
a/2[101]
Loops in
Stringers 1
V
I
V
I
V
V
a[001]
Loops in
Stringers 2
V
I
V
V
V
I
a[100]
Beam Direction
101
101
101
111
111
111
Note: V = Visible; I = Invisible.

108
L
Figure
.32. Micrograph of a precipitate colony on a climb
source showing the location (arrows) of four
rows of small dislocation loops whose Burgers
vectors were to be determined. (Heat treatment:
S.T. 1 hour 550C, quenched to 220C, aged 5
minutes.)

109
Figure 4.33.
Series o£ micrographs for determining the
Burgers vectors of the small dislocation loops
in the arrowed rows from invisibility condi
tions. The invisibility data are summarized
in Table 4.1.

110
The Burgers vectors of the loops in stringers 1 are iden
tified as a[001] and those in stringers 2 as a[100]. These
are consistent with the Burgers vectors of the peripheral
loops of the two orientations of 0' platelets expected in
this colony.
Now it will be shown how the data of Table 4.1 rule out
other possible origins for the loops:
(1) It is conceivable that kinks in the climbing source
loop could cause portions to trail behind and pinch off,
leaving the small loops. However, the small loops would then
have the same a/2[101] Burgers vector as the source loop.
Table 4.1 shows that the small loops are visible on several
reflections where the source loop is invisible. Therefore
they cannot be pinched-off loops.
(2) If the 6' platelets form in front of the advancing
source loop, the source loop could pinch off around the plate
lets, reacting with the a<100> peripheral loops around the
platelets. Two reactions are possible, depending on the
platelet orientation:
a/2 [101] + a[100] * a/2[301] for 0 on (100),
a/2 [101] + a [001] -* a/2[103] for 0' on (001).
However, both resultant a/2<103> Burgers vectors can be ruled
out since the loops in stringers 1 are invisible for the 220
reflection.
(3) Frank sessile loops, having a/3 Burgers vec
tors, can form by the collapse of vacancy clusters on {111}
planes. Such partial dislocations are invisible for g b=0

Ill
and 1/3 and visible for all other products (Silcock and
Tunstall, 1964; Hirsch et_ al. 1965). All four possible
a/3 Burgers vectors can be eliminated by comparing the
data in Table 4.1 with these visibility criteria, so that the
loops are not Frank loops.
(4) All six possible a/2<110> Burgers vectors can be
eliminated by the combinations of visible and invisible
images in Table 4.1. Therefore the loops cannot be prismatic
loops on {111} with F=a/2<110>, formed from Frank sessile
loops by the passage of a/6<112> Shockley partials over the
loops.
(5) The loops cannot be pure-edge, vacancy-condensation
loops with b=a/2<110> on (110} planes (of the type observed
by Boyd and Edington (1971) lying just within source loops),
since all possible a/2<110> Burgers vectors can be ruled out.
Thus, it is concluded that the small loops visible in
precipitate colonies have Burgers vectors of the type a<100>
and lie at the periphery of the 9' platelets.
4.4.4. "Secondary" Climb Sources
A characteristic grouping of precipitate colonies
observed often is shown in Figure 4.34(a). Here a long glide
dislocation climbed through the lattice to its final position
A-A, where it is out of contrast. In climbing, this disloca
tion nucleated the large precipitate colony P in its path.
At the base of this colony, a number of {110} habits of climb
sources have been activated and nucleated their own precipitate

112
Figure 4.34. (a) Secondary climb sources B, C, and D, gener
ated in the precipitate colony P that was
nucleated by climb of the long glide disloca
tion which is out of contrast between AA.
(b) Secondary climb sources A and B generated
in the precipitate colonies at primary climb
sources (viewed edge-on in this orientation).

113
colonies. Some of these (110) habits are viewed edge-on (as
at B), some lie in the plane of the foil (as at C), and some
are inclined through the foil (as at D). A similar configu
ration was shown in Figure 4.14(b).
It is proposed that these configurations evolved in the
following manner. The long glide dislocation began to climb
shortly after the onset of quenching. When the sample tem
perature passed below the 0' solvus, this dislocation began
to nucleate 0 precipitates in its climb path. Some of the
earliest-nucleated platelets acted as source particles and
generated the climb source loops on {110} planes, and these
then nucleated their own precipitate colonies.
The term "secondary climb sources" shall be used to dis
tinguish sources nucleated at 0' platelets in this manner
from "primary" climb sources nucleated on insoluble particle
existing at the solution treatment temperature. This seems
an appropriate designation since the secondary sources nucle
ate only if the 0 precipitation reaction occurs, whereas
primary sources operate independent of the precipitation.
From the observation that secondary climb sources were
always located at the base of the precipitate colonies gener
ated by the long climbing dislocations, it can be deduced
that their nucleation occurs within some limited, time-
temperature range just below the 0 solvus when the earliest
0' precipitates formed. The diameters of the largest secon
dary source loops were never as large as the diameters of
the largest primary source loops in the same foil, in

114
agreement with the conclusion that the latter are nucleated
earlier in the quench and consume more vacancies.
Now, if secondary climb sources are indeed nucleated by
0' platelets, they could nucleate in precipitate colonies on
primary climb sources as well. Such configurations were also
observed, as shown in Figure 4.34(b). In this micrograph,
the beam is approximately parallel to [001] so that climb
source habits on (110) and (110) planes are viewed edge-on
as at A and B. The smaller source habits at A and B were
obviously nucleated at different sites along the two larger
habits. Therefore, they all could not have been nucleated at
the original source particle. As they appear to have nucle
ated in the planes of the larger habits, they were most
probably nucleated at 0* precipitates in these planes.
A model will now be presented to illustrate how a 0'
platelet could act as a source particle for the nucleation of
climb source loops on {110} planes. As discussed in the pre
vious section, the misfit between matrix and precipitate
planes at the peripheral edge of 0' platelets is accommodated
by the presence of a<001> edge-dislocation loops around the
platelets. The extra half-plane must be contained in the
precipitate. Furthermore, Laird and Aaronson (1968) have
shown that 0' platelets of appreciable size are often octag
onal-shaped with their edges lying along the <100> and <110>
directions within the plane of the platelet. Consider then
such an octagonal-shaped platelet lying parallel to the (001)
matrix planes with its c-axis parallel to [001] matrix,

115
Figure 4.35(a). Two of its eight sides lie parallel to the
[100], [010], [110] and [110] directions, respectively. For
convenience, we assume the thickness of the platelet in its
c-direction is such that it has one a[001] dislocation loop
at its edge. Now let the two segments of this a[001] loop
along [010] between AB, and along [100] between CD dissociate
according to the reactions:
a[001] -* a/2 [101] + a/2 [101] between AB ,
a[001] -> a/2 [Oil] + a/2 [Oil] between CD,
which occur without energy change. The dissociated configu
ration is shown in Figure 4.35(b). The resultant dislocations
are pure-edge segments between AB and CD. Assume further that
these segments are free to climb as pure-edge dislocations
and that the corners A, B, C and D pin them at these points.
These segments of pure-edge dislocations can now operate by
climb to produce successive loops on the appropriate {110}
planes, as in Figure 4.35(c), in the same manner as the
original model of a climb source proposed by Bardeen and
Herring (1952). In this way, a given 6' platelet could pro
duce edge-loops on four possible {110} planes.
Since the extra half plane of the original a[100] loop
was contained within the precipitate, the initial climb of
the dissociated dislocation segments must proceed through a
small volume of precipitate before entering the matrix. As
the growth of 0' platelets from the matrix generates vacan
cies within the precipitate (Section 2.2), dislocation climb
is aided by the precipitation.

116
a
b
Figure 4.35. Model for the operation of a climb source at
a 9' platelet. (a) 9' platelet on (001) with
a[100] misfit dislocation loop around its
edge. (b) Dissociation of the a[100] loop
into total edge dislocations along the
portions AB and CD.

117
Figure 4.35. Continued. (c) The a/2[Oil] dislocation
segment, pinned between C and D, operates
as a Bardeen-Herring source by climb on
the (Oil) plane (compare with Figure 2.7).

118
4.4.5. A Climb Source on (100)
On one occasion, when a foil was being examined in the
(001) orientation, a planar, precipitate colony lying on (100)
was observed. This colony is shown in Figure 4.36(a) at A.
It is different from all other planar colonies observed which
were nucleated by dislocation climb sources operating on
{110} planes, e.g., at B in Figure 4.36(a). When the foil
was tilted away from the (001) orientation, the colony was
found to be bounded by an arc of a dislocation loop, Figure
4.36(b). Thus, it had the appearance of a climb source loop
and associated precipitate colony, but lying on a (100) plane.
This configuration was photographed on different two-beam
reflections in several orientations. Selected images are
shown in Figure 4.37. Invisibility conditions were used to
determine the Burgers vector of the dislocation arc and orien
tation of the precipitates in the colony. This information
is summarized in Table 4.2. The dislocation arc is invisible
for g=02Q and 022. This eliminates all a<001> and a/2<110>
Burgers vectors except a[100] and a/2[Oil]. Further, it is
visible for jf=IIl and 311. This eliminates a/2 [Oil] but not
a[100]. All partial dislocations with Burgers vector a/3
or a/6<112> are eliminated by comparison of the data in Table
4.2 with the visible and invisible criteria for partial dis
locations, namely, invisible for g*5 = 0 or 1/3 and visible
for all other products (Silcock and Tunstall, 1964; Hirsch
et al. 1965). It is concluded that the Burgers vector of
the dislocation is a[100]. Since it lies on (100), it must

119
Figure 4
36.
Two views of a precipitate colony A which is
seen to lie on (100) from its edge-on orien
tation in (a). The beam direction is close
to [001] in (a) and to [101] in (b).
(Heat treatment: S.T. 1 hour 550C, quenched
to 220C, aged 5 minutes.)

120
Figure 4.37. Series o£ micrographs for determining the
Burgers vector of the dislocation arc bound
ing the precipitate colony A in Figure 4.36,
and the orientation of 0' precipitates in
the colony. The invisibility data are summar
ized in Table 4.2.

121
Table 4.2
Summary of Visibility Data for the
Images of Figure 4.37
g=
200
0 20
III
020
220
022
Ill
311
Burgers
Vector
Ppt.
Orient.
Dislocation
Arc
V
I
V
I
V
I
V
V
a[10 0 ]
Precipitate
Colony
V
I
V
I
V
I
V
V
(100)
Beam
Direction
001
001
101
101
111
111
112
103
Note: V = Visible; I = Invisible.
be an arc of a pure-edge loop. Hence, it can expand in the
(100) plane only by climb. The presence of the precipitate
colony on (100) indicates that it did indeed climb in this
plane. As in the case of climb sources on {110} planes, it
is assumed that the precipitate colony was nucleated by the
a[100] dislocation as it climbed on (100). Now a dislocation
with b=a[100] favors nucleation of the (100) 9' orientation
only. From Table 4.2 we see that the precipitates in the
colony are invisible for g=020 and 022 and visible for all
reflections for which h in (hkl) is non-zero. Since the
precipitates are out of contrast only for g-vectors normal
to their misfit, and therefore to their habit plane, the
colony must consist of precipitates having only the (100)
orientation.

122
In summary, the configuration has all the characteristics
of a climb source operating on (100) with b=a[100], and a pre
cipitate colony of (100) 6* orientation nucleated by the
source loop.
As in the case of climb sources on (110), no stacking
fault was observed within this loop. Since the stacking of
planes in f.c.c. in the [100] direction is ABAB..., the loop
must climb by the condensation of vacancies onto two adjacent
(100) planes.
The small partial loop at C in Figure 4.36(a) was also
analyzed and found to have F=a/2[10l]. This must be a
secondary climb source of the (1101-type that was nucleated
at a 9' platelet lying in the precipitate colony on (100).
This observation is the only reported case of a climb
source in aluminum alloys operating to produce pure-edge
loops on a cube plane. This is not surprising, however,
since it was the only (100) source recognized as such among
thousands of sources scanned in all these foils.
4.4.6. Nucleation of Preferred 9* Orientations
During Segmented ClimE~
Figures 4.14(b) and 4.17 show corrugated-shaped, pre
cipitate colonies which were nucleated by the segmented climb
of glide dislocations on different crystallographic planes.
Close examination of such colonies revealed that they con
sisted of adjacent bands of precipitates, each band contain
ing only one of the two possible 9' orientations favored to
be nucleated by the climbing dislocation. This "preferred

1 2 3
nucleation" results in different fringe effects in the bands
marked 1 and 2 in Figure 4.14(b). The fringe appearance is
a diffraction effect, to be discussed in the next section.
This preferred nucleation of only one 0' orientation per
band is shown clearly in Figure 4.38, which is a magnified
portion of Figure 4.17 on a different reflection. Only the
edge-on orientation of 0' is present in the bands marked A.
Bands B therefore must contain the other possible 0' orien
tation favored to be nucleated, which is not parallel to the
beam in this orientation. Between some bands are regions
where no precipitates were nucleated, e.g., at points C.
Depending on the diffracting vector, the misfit of the single
9' orientation in a band can cause the band to go completely
out of contrast, as is the case in Figure 4.18 between C and D,
To account for the nucleation of only one 0' orientation
in a given band, it is now suggested that a dislocation under
going segmented climb tries to assume certain line directions
on the different crystallographic planes. This is probably
due largely to differences in line tension and ease of climb
on certain planes. If the local line direction is such that
the dislocation is pure edge, nucleation of two orientations
is favored. If it is pure screw, no nucleation is favored
since there are no tensile or compressive stresses around a
screw dislocation. If it is mixed, the resulting stress
field could favor the nucleation of one or the other 0' orien
tation, depending on the line direction and plane on which
the dislocation lies. As stated in Section 4.2.2, it was not

124
Figure 4.38. Micrograph of a precipitate colony at high
magnification illustrating that only one 9*
orientation is present in different bands
A and B which result from segmented climb of
a glide dislocation. No precipitates were
nucleated in the regions C between bands.
(Heat treatment: S.T. 1 hour 550C, quenched
to 220C, aged 5 minutes.)

125
possible to determine the different planes on which segmented
climb occurred in these foils. However, the vacancy-
annihilating climb of a/2<110> dislocation is easiest on
{ill) and illO) planes (Christian, 1965). Furthermore,
Miekk-oja and Raty (19 71) were able to show that segmented
climb occurred on the (111} and (110) planes in Cu-Ag alloys.
Their model for segmented climb on these planes will be used
here to suggest how this preferred nucleation could occur
and to explain the absence of precipitates between the bands.
Their model is shown in Figure 4.39 for segmented climb
of a dislocation with Burgers vector a/2[Oil] which was
initially a glide dislocation in its edge position on (111).
It quickly deviates from its edge position along XY, so that
some of its segments climb on the (Oil) plane and others
climb on the (111) and (111) planes, owing to the large
chemical climb forces on these planes. After a short time,
the dislocation assumes the multiply curved shape along RS
due to line tension effects which are required to mate up the
different segments between the planes. The exact curvatures
depend on the widths of the segments, the ease of climb on
the different planes, and the line tension, all of which are
unknown. If it is assumed that this model is valid for seg
mented climb in Al-Cu, it is possible that the resulting
curvatures favor nucleation of one of the two 0' orientations
on a certain plane and the other orientation on another plane.
1!" the curvature of the dislocation becomes so great near the
plane junctions as to be appreciably screw in orientation,

126
Ab=a/2[011]
G | (Oil) C
R s' XjITI) 5
X (011) A I (011) Y
(111)
Figure 4.39. Model for segmented climb on the planes of
easy climb. The dislocation, initially in
its edge position XY on the (111) plane,
climbs onto the planes of easy climb and
assumes the curved shape RS (after Miekk-oja
and Raty, 1971).

127
then no 0' would nucleate, accounting for the precipitate-
free regions between bands in Figure 4.38.
It is interesting to note that Miekk-oja and Raty (1971)
observed precipitation during segmented climb but did not
report any preferred orientations of the precipitates asso
ciated with the climb. In Al-Cu, segmented climb always
resulted in this preferred nucleation.
4.4.7. Displacement Fringe Contrast
in a Precipitate Colony
Often, fringe effects were observed in the precipitate
colonies, Figures 4.38 and 4.40(a). The intensity of the
fringes varied with the deviation from the Bragg condition,
becoming stronger as the deviation tended toward zero. Some
times the fringe intensity varied with position over a given
colony, appearing to be a function of the local precipitate
density. The intensity of fringes increased in areas where
the local precipitate density increased. The colony in
Figure 4.40(a) has a very dense, but uniform distribution of
precipitates so that the fringe intensity is large but uni
form.
Similarly, Miekk-oja and Raty (1971) reported fringe
effects in precipitate colonies nucleated behind climbing
dislocations in Cu-Ag alloys. However, they did not explain
the contrast mechanism, but described the effect simply as
image contrast variations with foil depth depending strongly
on diffraction conditions. It will now be shown that the
fringe effect can be explained simply on the basis of

128
Figure 4.40. Displacement fringe contrast in precipitate
colonies in samples quenched from 550C to
220C and aged for 5 minutes in (a) and for
1 minute in (b). The dislocation which
nucleated the colony in (a) is out of con
trast at the bottom of the colony.

129
displacement fringe contrast (Hirsch et_ al. 1965 p. 341).
In Section 2.5.3, it was shown that displacement fringe con
trast arises from abrupt phase changes in the incident and
diffracted beams encountering an inclined sheet of precipitate
which has displaced the matrix by an amount K in opposite
directions across the sheet. Figure 4.41 illustrates that a
planar, precipitate colony, composed of two orientations of
6' platelets of uniform size, has the resultant effect of a
sheet of displacement field which displaces the matrix normal
to the colony by a resultant displacement R^. Thus, whenever
such a precipitate colony is inclined to the electron beam,
the displacement fringe effect will be observed. Of course,
a colony consisting of only one orientation of 0', whose
misfit does not lie in the plane of the colony, would cause
a displacement fringe effect also, as seen in Figure 4.38.
Similarly, Warren (1974), using the method of Humble (1968),
has computed electron micrographs of a plane of dilatation
and obtained the displacement fringe effect. An example is
shown in Figure 4.42. In addition, Clarebrough (1973) has
shown that passage of a unit dislocation through an ordered
lattice creates a planar displacement field at the slip plane
which causes similar fringe effects in electron micrographs.
Now the intensity of displacement fringes is a function
of the magnitude of the normal displacement R. It can be
visualized from Figure 4.41 that, if the density of precipi
tates varies over the colony, the local resultant displace
ment, R-^, varies accordingly. Thus, the fringes will vary

130
a
e~
V
/£>
/\'y
-Ri /X'V
'X
R,
<$>
%s
Figure 4.41. Diagram illustrating the origin of fringes in
a 0' precipitate colony. (a) A planar colony
containing two orientations of 0' platelets of
uniform size is inclined through the foil.
(b) The combined displacement fields of all
the platelets act as an inclined sheet of dis
placement, giving rise to the conditions
for displacement fringe contrast.

li
Figure 4.42. Computer-simulated electron micrographs of a
plane of dilatation inclined through the foil,
illustrating the displacement fringe effect.
The fringes lie at about 45 to the horizontal,
and there is a horizontal dislocation image at
the top of the fringes. The strength of the
dilatation increases from the top image to the
bottom so that the intensity of the fringes
increases (after Warren, 1974).

132
in intensity in agreement with experimental observation.
When the sheet of displacement is not flat, the fringes will
not be straight and parallel. Accordingly, the precipitate
colony in Figure 4.40(a) has some curvature. Colonies were
observed in which the fringes curved as much as 90, indicat
ing large curvature in the precipitate colony and, therefore,
in the climb path of the dislocation.
Figure 4.40(b) shows fringes behind a dislocation in a
foil which was quenched to 220C and aged for only one minute
at 220C. Here the individual precipitates in the colony are
not large enough to be visible, but their presence is indi
cated by the displacement fringes. The intensity of these
fringes is less than in those of Figure 4.40(a). This could
be due to differences in the deviation from the Bragg condi
tion or by differences in the density of precipitates in the
two colonies. A more likely explanation is that the average
precipitate thickness is less for the shorter aging time,
and therefore the resulting ^ is smaller. Such fringes in
the absence of visible precipitates were observed often in
samples aged for short times. They further support the con
clusion that the 0' colonies are nucleated entirely during
the quench.
4.4.8. Precipitate Colonies Associated
with Subboundary Formation
Dislocation subboundaries were observed in all foils
examined in this research, regardless of heat treatment. An
example of a subboundary network is shown in Figure 4.43(a).

133
The dislocations were generated probably at grain boundaries
during the quench. Normally, subboundaries are observed in
cold-worked metals which have been given recovery anneals.
In the present samples, boundary formation was essentially
completed during the quench, since well-defined boundaries
were observed in foils quenched into liquid nitrogen with no
subsequent aging treatment. Both tilt boundaries and twist
boundaries were observed, although tilt boundaries were more
prevalent.
Figure 4.43(b) shows a junction of three tilt boundaries
in a sample direct-quenched to 220C and aged five minutes.
Clearly, the dislocations have nucleated precipitate colonies
in the process of climb. The climb paths of all dislocations
in a given boundary are in the same direction as indicated
by the positions of the precipitate colonies. It cannot be
determined from such micrographs if the precipitates were
nucleated while the boundaries were forming, or if the bound
aries, once formed, climbed in a cooperative manner and
nucleated the precipitate colonies. In either case, boundary
formation was completed during the quench. Since edge dislo
cations are potential nucleation sites for 0 precipitates
whereas screw dislocations are not, precipitate colonies were
always observed to be associated with tilt boundaries, but
they were not present at twist boundaries.

134
Figure 4,43. (a) Dislocation subboundaries in a sample
direct-quenched from 550C to 220C and held
only 8 seconds. (b) Precipitate colonies
associated with subboundaries in a sample
direct-quenched from 550C to 220C and
aged 5 minutes.

135
4.5. Effects of Experimental Variables
on Microstructure
Up to this point, evidence has been presented to charac
terize the nature of repeated precipitation at climbing dis
locations. What remains is to determine the mechanisms by
which the nucleation events take place. Information towards
this objective was obtained by varying independently the
following experimental parameters:
(1) time at constant aging temperature after
direct-quenching,
(2) solution treatment temperature,
(3) temperature to which samples are direct-
quenched,
(4) quench rate, and
(5) solute concentration.
In addition, a clearer picture of the operation of climb
sources in Al-Cu alloys during quenching was obtained as well.
4.5.1. Effect of Time at Constant Aging Temperature
In this experiment, samples were solution-treated for
one hour at 550C, direct-quenched in oil at 220C, and aged
for various times. The resulting microstructures after aging
for 8 seconds, 1 minute, 5 minutes, 30 minutes, and 2 hours,
respectively, are shown in Figure 4.44 at low magnification.
After 8 seconds at 220C, the only visible precipitates are
those in the stringers along the [OlO] direction. After
aging one minute, other precipitates are just visible in the

136
b
Figure 4.44. Sequence of micrographs showing the effect
of time at constant aging temperature on
colony growth. Samples were direct-quenched
from 550C and aged for (a) 8 seconds,
(b) 1 minute, (c) 5 minutes, (d) 30 minutes,
and (e) 2 hours.

137
Figure
4,44. Continued

138
Figure 444. Continued.

139
colonies. After 5 minutes aging, the interiors of the
colonies are seen to be densely precipitated. All precipi
tates in the colonies were present at the start of the aging
at 220C (Section 4.3). After 30 minutes at 220C, the pre
cipitates are large enough to individually exhibit displace
ment fringe contrast, but precipitation is still localized
within the original colonies. After two hours at 220C, some
scattered precipitates are observed outside the colonies,
but the vast majority are still associated with the original
colonies.
These results indicate that the density of precipitates
generated by repeated nucleation is sufficiently large that
aging for long times results essentially only in growth.
Since there was no evidence for bands of precipitates spread
ing out from the colonies, an autocatalytic nucleation mecha
nism (Section 2.2) can be ruled out. In the original work
on autocatalytic nucleation of 0' in Al-Cu, Lorimer (1970)
found a uniform distribution of precipitates throughout the
foil after aging 35 minutes at 240C. In the presence of a
large precipitate density, generated by repeated nucleation
on climbing dislocations, evidently the driving force for
autocatalytic nucleation is small, and solute depletion of
the supersaturated lattice can be accomplished by growth of
existing precipitates.

140
4.5.2. Effect of Solution Treatment Temperature
Five samples were solution treated for one hour at vari
ous temperatures within the solid solution range, then direct-
quenched to 220C and aged for five minutes. The five tem
peratures employed were 570, 550, 530, 515 and 504C,
covering the range from just below the solidus temperature to
just above the a+0 solvus temperature for the 3.85 wt.% Cu
alloy (Figure 4.45). The major difference between samples
was the quenched-in vacancy supersaturation, which increases
exponentially with quenching temperature. However, to a
lesser extent, the treatments also differed in quench rate.
The resulting microstructures are shown in Figure 4.46 at
low magnification.
First, repeated nucleation occurred during quenching in
all samples as indicated by the presence of precipitate
colonies associated with all dislocations. Therefore repeated
nucleation does not appear to depend on the vacancy super
saturation, at least for the range of supersaturations in
these direct-quenches. Secondly, both climb sources and
glide dislocations which subsequently climbed were present
in all samples, although their relative densities varied.
It is not possible to illustrate all the features of
these microstructures in one micrograph for each sample in
Figure 4.46. Accordingly, descriptions are given here based
on observations recorded during examination in the electron
micros cope.

141
Figure 4.45.
Diagram showing the five solution treatment
temperatures from which samples were direct-
quenched to 220C and aged.

142
Figure 4.46. Sequence of micrographs showing the effect on
microstructure of the solution treatment
temperature from which samples were direct-
quenched. Samples were solution treated for
1 hour at 570C in (a), 550C in (b), 530C
in (c), 515C in (d), and 504C in (e), then
direct-quenched to 220C and aged 5 minutes.

143

144
Figure 4.47. Large dislocation climb source in a sample
quenched from 570C to 220C and aged 5
minutes.

145
Two observations were made about the effect of quenching
temperatures on the operation of climb sources:
(1) The average size of the source loops increased with
increasing solution treatment temperature. In the samples
quenched from the lowest temperatures, the climb sources were
small, with few exceptions. In the sample quenched from
570C, both small and very large sources (~5-6p diameter)
were observed, the average size being the largest for all
I
samples. Figure 4.47 shows a very large source in this
sample.
(2) The number density of active climb sources was very
low for quenching from 504C, increased with quenching tem
perature to a maximum for 550C, then decreased to an inter
mediate value at 570C. Insufficient micrographs were
obtained to make reliable quantitative measurements of these
densities, so that the trend is described only qualitatively
here. An important influence on microstructure is that vari
ations in size and density of climb sources are reflected in
the density of precipitates which nucleate during quenching.
The above observations can be explained as follows. The
vacancy supersaturation increases exponentially with quench
ing temperature so that the average distance a source loop
will climb increases. Secondly, if nucleating sites for
climb sources are indeed particles, the particle solubility
will tend to increase with temperature. Thus, as the solu
tion treatment temperature was increased, a trade-off between
increasing vacancy supersaturation and a decreasing density

146
of undissolved source particles could lead to the observed
maximum density of sources in the sample quenched from 550C.
That is, at 570C, there would be a minimum number of undis
solved source particles, and a maximum vacancy concentration.
Upon quenching fewer sources would generate much larger source
loops compared to the structure quenched from 550C. At
504C, the undissolved particle density would be a maximum,
but the vacancy concentration is a minimum. Upon quenching
only a minimum operation of climb sources is required to pro
vide sinks for vacancy annihilation.
In the sample quenched from 570C, repeated nucleation
of 9' occurred within a limited band, approximately one
micron wide, immediately behind the final position of the
dislocations (Figures 4.46(a) and 4.47). As this sample was
quenched from the highest temperature and contained the
largest vacancy supersaturation, it is believed that appreci
able dislocation climb occurred above the 0' solvus tempera
ture, where no 0 could nucleate. The well-defined boundary
where precipitation begins indicates that the concept of a
0' solvus temperature is a very effective and sensitive bar
rier to nucleation during quenching. In addition, the slope
of the 0' solvus at the composition 3.85 wt.% Cu is small,
causing a large driving force for nucleation immediately
below it.

147
4.5.3. Effect of Temperature to Which
Samples Are Direct-Quenched
In this experiment, five samples were solution treated
for one hour at 550C, direct-quenched into oil at various
temperatures above the G.P. zone solvus, and aged (Figure
4.48). Therefore, the basic difference between samples was
the quenching rate. The temperatures were 180C, just above
the G.P. zone solvus; 200 and 220C, just below and above
the 0" solvus; 250 and 300C, increasing temperatures in the
a+9' region. The oil bath was limited to 300C. The tem
perature to which each sample was quenched shall be referred
to as the aging temperature, Ta. With decreasing Ta, the
samples were aged for longer times so that the 0' precipitates
would be visible. The resulting microstructures are shown at
low magnification in Figure 4.49.
In general, 0 nucleated repeatedly on climbing disloca
tions during quenching in all samples, although only to
small degree in the sample quenched to 300C (Figure 4.49(a)).
Furthermore, both glide dislocations which climbed and climb
sources were observed in all samples. Whereas the density of
glide dislocations varied little from sample to sample, the
density of climb sources varied greatly.
The sequence of micrographs in Figure 4.49 shows that
decreasing Ta increases the number of active climb sources
and decreases the average source loop diameter. The maximum
source density occurred for Ta=180C and was about 6x10^ per
grain (Section 4.2.1).

Temperature,
148
Al
Wt.% Cu
Figure 4.48. Diagram showing the five temperatures above
the G.P. solvus to which samples were direct-
quenched from 550C and aged.

149
Figure 4.49. Sequence of micrographs showing the effect on
microstructure of the temperature to which
samples were direct-quenched from 550C and
aged. The samples were quenched to (a) 300C
and aged 15 seconds, (b) 250C and aged 1
minute, (c) 220C and aged 5 minutes, (d) 200C
and aged 30 minutes, and (e) 180C and aged
1 hour.

150
Figure 4.49
Continued

Ibl
Figure 4.49. Continued

152
These observations can be explained if two factors are
accounted for. First, Figure 4.50 shows typical cooling
curves for identical samples quenched at two different rates.
Clearly, the sample quenched at the slower rate, takes
longer to reach a given temperature, T-^, than a sample
quenched at a faster rate, The former also spends longer
within any given temperature increment, AT. Secondly, the
variation in active source density can be treated as a problem
analogous to the nucleation of precipitates over a distribu
tion of favorable sites at different aging temperatures. It
is assumed only that some particles will be more favorable
sites for nucleating loops than others.
First consider the sample with the slowest quench rate
(Ta=300C). Shortly after the onset of quenching, source
loops nucleate only at the most favorable particles. Since
this sample spends a maximum time at high temperatures, the
diffusion distance for vacancies is large. The first loops
to nucleate can grow to be large, depleting the matrix of
vacancies, and thereby suppressing the nucleation of other
loops in nearby regions. As a result, this sample should
have the lowest active source density, but the largest aver
age loop size, in agreement with experiment (Figure 4.49(a)).
As the quench rate increases (i.e., as Ta decreases), the
time spent at high temperatures decreases and the diffusion
distance for vacancies decreases. The first loops to nucle
ate can no longer grow as large before other loops nucleate
on the next most favored particles (Figure 4.49(b)). Carried

153
Time to Reach Temperature
Figure 4.50. Diagram showing the differences in the time
to reach a given temperature and the time
spent in a given temperature range AT, for
two different quench rates.

154
still further, the sample with the fastest quench rate (Ta=
180C) should contain the maximum density of active sources
with the smallest average loop size, again in agreement with
experimental observations (Figure 4.49(e)).
An additional feature of the sample quenched to 300C
was that repeated nucleation occurred for only short distances
behind the climbing dislocations and resulted in a low density
of precipitates, Figure 4.49(e). This effect can be explained
also using Figure 4.50. Given the slowest quench rate, this
sample stayed longest at temperatures above the 0* solvus,
thus causing maximum dislocation climb and the maximum deple
tion of vacancies before any precipitates nucleate. Also,
the sample remained longest at high temperatures just below
the 0' solvus. Therefore only a few precipitates nucleate
and these grow rapidly (Figure 4.49(a)). This sample was aged
only 15 seconds at 300C so that the precipitate size is still
small.
In the two samples quenched to 200 and 180C (tempera
tures below the 0" solvus, Figure 4.48), 0" did not form as
predicted from the metastable phase diagram. This is in agree
ment with observations that 0" nucleates only on previously-
formed G.P. zones (Lorimer and Nicholson, 1969; Lorimer, 1970)
4.5.4. Effect of Quench Rate
The previous section described results of changing the
quench rate by direct-quenching to various aging temperatures.
However, direct-quenching is, in general, a slow quench. The

155
purpose of the following experiments was to determine if
repeated nucleation could be eliminated, or suppressed, by
faster quenching. The quench rate was changed by varying
the quenching medium.
Samples were solution treated for one hour and quenched
int o:
(1) air,
(2) liquid nitrogen,
(3) oil at room temperature, and
(4) water at room temperature.
The quench rates were not measured, but it is believed they
increased in the order given above. After quenching, the
samples were up-quenched into oil at 220C and aged for five
minutes in order to grow any 0' precipitates present to
visible sizes. The air-cooled sample was held for approxi
mately six seconds to insure complete cooling before up-
quenching to 220C. All other samples were up-quenched with
in about two seconds after down-quenching.
It was found that repeated nucleation of 0' had occurred
on climbing dislocations during all quenches listed above.
However, dislocation configurations varied widely as a result
of the different quench rates so that each treatment will be
discussed separately.
Air quench
Air quenching generated approximately the same density
of glide dislocations as did direct-quenching into oil. All

156
had climbed during quenching and had nucleated precipitate
colonies, Figure 4.51(a). Climb sources were found only
occasionally, but they had nucleated precipitate colonies
with the same general appearance as those in direct-quenched
specimens. A small group of sources is shown in Figure 4.51(b).
Liquid nitrogen quench
The structure of this sample was similar to that of the
air-cooled sample. The density of glide dislocations was
approximately twice that in the air-cooled sample. Both small
and large climb sources were present and all had nucleated
precipitate colonies, Figure 4.52(b).
Room-temperature oil quench
The dislocation structure changed abruptly in going to
a room-temperature oil quench. The resulting microstructure
is shown in Figure 4.53 at low magnification. The foil is
full of small dislocation loops. In addition, there were
some glide dislocations present which had climbed during
quenching and had nucleated precipitate colonies (Figure 4.53
at A and B). Note that the density of small loops is reduced
in the vicinity of the glide dislocations. Some climb sources
were observed, but their density was much lower than in the
samples direct-quenched into oil at the aging temperature.
Figure 4.54 shows an area where a number of climb sources are
grouped together. The density of small loops was low in the
regions adjacent to climb sources.

157

b
Figure 4.51. Precipitate colonies nucleated on (a) an
initial glide dislocation, and (b) climb
sources in a sample air-quenched from
550C, then up-quenched in oil to 220G
and aged 5 minutes.

158
b
Figure 4.52. Precipitate colonies nucleated on (a) initial
glide dislocations, and (b) climb sources in
a sample quenched from 550C into liquid
nitrogen, then up-quenched to 220C and aged
5 minutes.

159
Figure 4.53. Microstructure of a sample quenched from 550C
into room temperature oil, then up-quenched to
220C and aged 5 minutes, showing a high
density of small dislocation loops and several
glide dislocations which climbed (at points A).
Figure 4.54. Dislocation climb sources and associated pre
cipitate colonies in the sample quenched into
room-temperature oil, then up-quenched to
220C and aged 5 minutes.

160
The small loops are shown at high magnification in Figure
4.55. The interior of the loops are precipitated in much the
same manner as the interior of the climb source loops shown
in previous micrographs. For example, the loops and associ
ated precipitate colonies at points A have the same appearance,
but on a smaller scale, as climb source loops and their asso
ciated precipitate colonies viewed normal to their {110}
habits. The small loops encircling the precipitates at A
are invisible in this image, and in place of precipitate
stringers in the <10Q> directions, they have single larger
precipitates at the interior edges of the loops.
The small loops were found to lie on {110} planes as
illustrated by the edge-on habits at B in Figure 4.55. Their
Burgers vectors were determined from invisibility conditions
to be a/2<110> types normal to the loop planes. Thus, they
are prismatic edge-loops on {110} with E=a/2<110>. It is
assumed that they formed by collapse of vacancy clusters onto
{110} planes. Similar prismatic loops on {110} have been
observed in quenched Al-2.5 wt.% Cu (Boyd and Edington, 1971)
and in quenched Al-Mg alloys (Embury and Nicholson, 1963).
Once the loops form, they grow by climb from further vacancy
condensation. In so doing, they nucleate the small precipi
tate colonies. In fact, apart from their origin and size,
there is probably no difference in the mechanism of repeated
nucleation during the growth of these small loops and during
the growth of climb source loops.

161
Figure 4.55. Higher magnification of the structure in
Figure 4.53 showing that the interiors of the
small dislocation loops contain precipitate
colonies. Loops at B are viewed edge-on,
while loops at A are out of contrast around
their precipitate colonies.

162
The high density of small loops in this structure can
be explained as follows. The room temperature oil quench is
a moderately fast quench achieving a much larger supersatu
ration of vacancies than direct-quenching to the aging tem
perature. Most vacancies do not have time to diffuse to
active climb sources or to glide dislocations which are climb
ing. They therefore cluster rapidly and collapse into loops
which grow by further vacancy condensation.
If the origin of these small loops is neglected, this
microstructure can be thought of as an extension of the struc
tures described in the previous section, where direct-quench
ing into oil at progressively lower temperatures increased
the density of active climb sources and decreased their
average diameter.
From Figure 4.53, the density of small loops was esti-
13 3
mated to be 2.2x10 /cm Their average diameter was measured
from Figure 4.55 and found to be approximately 0.25y. From
these estimates, and assuming that each loop climbs by remov
ing two adjacent {110} planes to avoid a stacking fault, the
- 4
quenched-in vacancy concentration was estimated as 3x10
This value is a slight overestimate since it neglects any
vacancy contribution from the growing 0" platelets, which
have a smaller atomic volume than the matrix.
However, this estimate is in good agreement with a value
of approximately 2.5x10 ^ for the equilibrium concentration of
vacancies in pure aluminum at the 550 solution treatment tem
perature (Simmons and Balluffi, 1960 ; Guerard et al. 1974).

163
Room-temperature water quench
The structure of this sample was similar to that of the
sample quenched into room-temperature oil. Figure 4.56(a)
shows the microstructure to be full of small loops. Occa
sional glide dislocations were observed which had climbed and
nucleated 0' colonies (Figure 4.56(a) at point A). The edge-
on orientations of the small loops indicated that they lie
on {110} planes. As in the oil quenched sample, their Burgers
vectors were found to be a/2<110> normal to the loop planes.
These are prismatic edge-loops and, as before, it is assumed
that they formed by collapse of vacancy clusters onto {110}.
When viewed at high magnification (Figure 4.56(b)), the
interiors of the loops appear to contain small precipitate
colonies. It is assumed that the loops nucleated 0' in the
process of growth by climb.
Two features were observed in the microstructure of this
sample that were not present in the sample quenched into room-
temperature oil. First, in isolated areas, helical disloca
tions were observed which had partially broken up into loops,
e.g., Figure 4.57. This is in agreement with the work of
Thomas (1959) who showed that helical dislocations were
present in Al-4 wt.l Cu quenched from low temperatures in the
solid solution range. In the present case, the solution
treatment at 550C was sufficiently high that the structure
consists mainly of vacancy-condensation loops with occasional
helical dislocations.

164
Figure 4.56. (a) Low and (b) high magnification of micro
structure of sample quenched from 550C into
room temperature water, then up-quenched to
221C and aged 5 minutes. The foil contains
a high density of irregular-shaped loops which
are internally precipitated.

165
Figure 4.57. Helical dislocations in the sample quenched
from 550C into room temperature water, then
up-quenched to 220C and aged 5 minutes.
Figure 4.58. Loops which have partially moved off their
habit planes in the sample quenched from 550C
into room temperature water, then up-quenched
to 220C and aged 5 minutes.

166
Secondly, it was found that portions of many small loops
had moved off their {110} habits. Examples are shown in
Figure 4,58. Since the original {110} habit is the most
favorable climb plane for these edge loops (Miekk-oja and
Raty, 1971), it is unlikely that they would move off by climb.
They have most probably slipped onto intersecting {111} planes
under the quenching stresses.
4.5.5. Effect of Copper Concentration
Samples with decreasing copper content were given direct-
quenching treatments to determine if solute concentration
played an important role in the repeated nucleation mechanism.
The copper concentrations of the samples were nominally
1.96 wtJ, 1.0 wt.%, and 0.5 wt.%, respectively (evidence for
repeated nucleation in the 3.85 wt.I Cu alloy has already
been discussed in great detail). The samples were solution
treated for one hour at 545C, direct-quenched into oil at
210C, and aged at that temperature. These temperatures are
shown in relation to the 0' and 0" solvus lines in Figure
4.59. Samples of the 1.96 wt.% Cu alloy were aged for either
30 minutes or one hour at 210C. Samples of the 1 wt.% Cu
alloy were aged for either three seconds, one hour, or 24
hours at 210C. Samples of the 0.5 wt.% Cu alloy were aged
for either three seconds, one hour, or 7-1/2 hours at 210C.
It was found that repeated nucleation of 0' occurred
extensively in the 1.96 wt.% alloy, but did not occur at all
in the 1.0 and 0.5 wt.% Cu alloys. Appreciable dislocation

Temperature,
16 7
Figure 4.59. Diagram showing the solution treatment and
aging temperatures used for direct-quenching
the 2, 1, and 1/2 wt.% copper alloys.

168
climb occurred during quenching in all three alloys. Thus,
copper concentration is an important variable in the repeated
nucleation process, tending to suppress the mechanism alto
gether below some critical concentration between 1.96 and
1 wt.% Cu. The microstructures in these samples are dis
cussed separately below.
1.96 wt.% copper
Many glide dislocations were observed which had climbed
and nucleated precipitate colonies. An example is shown in
Figure 4.60(a), where the dislocation is invisible along AB.
All the precipitate stringers have coalesced into long plate
lets due to the one-hour aging treatment at 210C. Many
climb sources were observed which had nucleated precipitate
colonies during quenching, Figure 4.60(b).
1.0 wt.% copper
Again both glide dislocations and dislocation climb
sources were present in this alloy, Figure 4.61. However,
no evidence for repeated nucleation of precipitates was
observed and only a few precipitates were found near dislo
cations after aging up to 24 hours. All climb source loops
had the Class IV shape of Figure 4.26, e.g., the source sec
tioned by the foil in Figure 4.61(b).
0.5 wt.% copper
As before, both climb sources and glide dislocations
were present after direct-quenching this alloy. In addition,

169
Figure 4.60. Precipitate colonies nucleated on (a) a long
glide dislocation which is out of contrast
between AB, and (b) dislocation climb sources
in the Al-2 wt.% Cu alloy. (Heat treatment:
S.T. 1 hour 545C, quenched to 210C, aged
1 hour.)

170
Figure 4.61. Micrographs of Al-1 wt.% Cu alloy quenched
from 5 45C to 210C and aged 1 hour, shox^ing
that no repeated nucleation of 0' occurred
during quenching. (a) Glide dislocations;
(b) climb sources.

171
bands of small dislocation loops were present throughout the
microstructure of the sample aged only three seconds at 210C,
Figure 4.62. These small loops exhibit stacking fault con
trast and are assumed to be Frank loops formed by the collapse
of vacancy clusters. After aging one hour at 210C, no such
loops were observed and it is assumed that they annealed out.
No evidence for repeated precipitation was observed in this
alloy, even after aging for 7-1/2 hours at 210C. In fact,
no precipitates at all were detected in this alloy, although
it was aged in the two-phase a+0' region (Figure 4.59).
The shape of climb source loops changed appreciably in
going to the 0.5 wtA Cu alloy. After direct-quenching and
aging for three seconds at 210C, some climb source loops
had the Class IV shape of Figure 4.26, but most had the shape
of nearly perfect rhombuses. After aging for one hour at
210C, all source loops, without exception, had the shape of
nearly perfect rhombuses (Figure 4.63). The long axis of the
rhombus is close to the <100> direction in the {110} habit of
the loops. The loop sides lie along <112> directions to
within about 5%. These <112> directions are the lines of
intersection of the two {111} planes perpendicular to the
{110} plane of the loops, i.e., these are the two slip planes
containing the a/2<110> Burgers vectors of the loops. This
is the same geometry of climb source loops observed in Al-Mg
alloys by Embury and Nicholson (1963).
It is now apparent that, as the copper content of the
alloy decreases from 3.85 wtJ to 0.5 wt.%, the typical climb

rr
Figure 4.62. Band o£ small prismatic dislocation loops in
Al-0.5 wt.% Cu quenched from 545C to 210C
and aged only 3 seconds.
Figure 4.,6 3. Rhombus-shaped climb sources in Al-0.5 wt.% Cu
quenched from 550C to 210C and aged 1 hour.
Local segments at A, B, C, and D appear to have
slipped out of the climb plane.

173
source loop changes in shape from slightly elliptical in the
[100] direction to a rhombus elongated in the [100] direction,
as shown schematically in Figure 4.64.
Close examination of the sides of the rhombus loops in
the sample aged for one hour at 210C showed that they con
tained a very regular spacing of kinks or jogs, Figure 4.65.
It is suggested that these are kinks caused by slip in the
{.111) planes normal to the loops. In fact, it was observed
often that short segments of the loops had undergone extensive
slip out of the plane of the loops, e.g., at A, B, C, and D
in Figure 4.65. The kinks were not resolvable (if present)
in climb source loops in the sample aged for only three
seconds.
The average spacing of the kinks was measured normal to
the <100> direction in the loop plane and found to be about
O.lp. This is almost exactly the same as the measured spacing
(0.096y) between precipitate stringers in the colonies on
climb sources in the 3.85 wt.% Cu alloy (Section 4.4.2).
Therefore, it is suggested that the origin of the precipitate
stringers in the alloys of higher copper concentration is
associated with a regular spacing of kinks out of the climb
plane of the source loops.
Examination of the lines of intersection of concentric
rhombus loops at a climb source with the foil surfaces
revealed that these intersections were not parallel, Figure
4.63. This indicates that the planes of successive loops are
rotated with respect to each other. In addition, it was

174
4wt./oCu
0.5wt./oCu
Figure 4.64. Diagram showing the change in shape of the
typical climb source in Al-Cu with decreasing
copper concentration.

175
Figure 4.65. High magnification of climb source loops in
Al-0.5 wt.% Cu quenched from 545C to 210C
and aged 1 hour, revealing a regular spacing
of kinks along the dislocations.

176
often observed that, where a large outer loop intersected
both foil surfaces, the two lines of intersection were not
parallel, indicating that the loops themselves are not
strictly planar. These observations could be explained only
if the loops are allowed to slip or climb locally out of
their habit planes in varying step heights.
4.6. Summary
The results in this chaoter may be summarized as follows:
(1) The 0' phase nucleated repeatedly on climbing dis
locations during quenching, creating precipitate colonies
along the climb paths. The precipitate density thus generated
was sufficiently large to suppress autocatalytic nucleation
during long aging.
(2) The dislocations cLimb during the quench by annihi
lation of quenched-in vacancies. These dislocations fall into
three categories according t a origin:
(a) pure-edge loous on {110} planes, gener
ated at dislocation climb sources,
(b) glide dislocations on {111} planes, and
(c) prismatic edge-loops on {110} formed by
collapse of vacancy clusters.
The dislocation density in each category varied with heat
treatment. Categories (a) aid (b) were found in all samples.
Category (c) was found only In samples quenched into oil or
water at room temperature.

177
(3) Glide dislocations climb in one of two ways:
(a) on smoothly curved surfaces, nucleating
precipitate colonies containing two 0'
orientations, or
(b) on corrugated-shaped surfaces, nucleating
bands containing only one 0' orientation
in each band.
(4) Only the two 0' orientations compatible with the
Burgers vector of the climbing dislocation are nucleated in a
given colony. This suggests that the precipitates were nucle
ated during quenching in the wake of the climbing dislocations
In situ aging treatments in the TEM support this conclusion.
(5) Precipitate colonies generated on climb sources
contain a region of dense precipitation and regions of pre
cipitate stringers in <1Q0> directions. The stringers are
associated with a regular spacing of kinks on the climbing
source loops. Only one 0' orientation is nucleated in a
given stringer.
(6) The number density and average size of active climb
sources are functions of the quench rate and vacancy super
saturation for direct-quenches. The typical shape of a climb
source loop changes from an ellipse to a rhombus, with
decreasing copper concentration.
(7) Repeated nucleation does not occur in alloys with
copper concentration below some critical value between 1.96
and 1 wt.% copper.

CHAPTER 5
THE REPEATED NUCLEATION MECHANISM
In the previous chapter, it was convenient to include
some discussion and analysis in each section. Hence, the
discussion in this chapter will be limited to aspects of the
mechanism of repeated nucleation of 0' in Al-Cu.
5.1. Nucleation of 9' Near Edge Dislocations
The a^0' transformation occurs with a 3.95% volume con
traction in the lattice (Section 2.2). The resulting misfit
strain energy is sufficient to suppress nucleation except at
dislocations whose stress fields reduce this strain energy.
The observation that just two 0' orientations nucleate in
each colony (Section 4.3) provides insight as to where the
nucleation events occur about the dislocation. One 0' orien
tation does not nucleate because its misfit is perpendicular
to the Burgers vector. Hence, its strain field is not
relieved by the stress field of the dislocation (Section 2.2)
The other two orientations nucleate because their misfits
lie at 45 to the Burgers vector, so that their strain fields
are partially relieved by the stress field of the dislocation
Since the a-*9' transformation is a volume contraction, the
178

179
resulting strain puts the surrounding matrix in tension.
Thus, this strain can be relieved only if the 0' platelet
forms on the compressive side of the dislocation. In order
to determine how close to the dislocation it forms, one must
solve the elastic interaction problem discussed in Section
2.1. In the case of the pure-edge loops generated at climb
sources in these alloys, the compressive region lies in front
of the expanding loops, Figure 5.1(a). Hence, nucleation
occurs in advance of the climbing dislocations, Figure 5.1(b)
Since essentially all the visible precipitates in a colony on
a climb source lie within the loops (e.g., Figure 4.30), the
loops must climb through or by the particles, or pinch off
around them in order to continue expanding. Now 0' forms
parallel to {10 0 > with a<100> peripheral misfit loops, so
that the climbing dislocation with a/2<110> Burgers vector
cannot comprise a part of the platelet edge. And as dis
cussed in Section 4.3.3, there was no evidence that the climb
ing dislocations pinched off around the platelets. If a
nucleus forms above or below the climb plane, the climbing
dislocation can pass by more easily than if the nucleus forms
along the climb plane (or grows to intersect it). In the
latter case, the dislocation must either cut through the pre
cipitate or, if the stresses involved are too high, glide
locally around it.
Occasionally a large precipitate was observed outside a
source loop which bent locally around the precipitate, Figure
5.2. It is assumed that such precipitates were the last to

SOURCE
CLIMB
PLANE
/
i
"1
a
o
b
\
/
n/->/x|n//
\/\ \|7\\
I*
/
c
Figure 5.1. (a) Schematic cross-section through a climb
source particle and one loop showing regions
of compression (dashed) in advance of the
loop. (b) Nucleation in the compressive
regions. (c) After further climb and gener
ation of another source loop.

181
a
b
Figure 5.2. Two images of a precipitate colony on a climb
source showing that the precipitates 1, 2, and
3 lie outside the outermost loop. The source
loop is out of contrast in (a), and the three
precipitates are out of contrast in (b). The
beam direction is close to [Oil] in (a), and to
[001] in (b).

182
nucleate near the end of the quench when the dislocation
climb rate had dropped to a low value. Thus, these precipi
tates had time to grow large enough to retard the passage of
the climbing dislocation.
5.2. Comparison with Previous Repeated
Nucleation Mechanisms
I i
|
i
The fundamentals of the repeated nucleation mechanism
proposed by Nes (1974) were reviewed in Section 2.4. This
mechanism reportedly accounts for the phenomenon in all
systems in which it has been observed to date, but in all
these cases the transformed phase has had a larger atomic
volume than the matrix. The Nes model will now be compared
with repeated nucleation of 0' in Al-Cu. The fundamentals
of the Nes mechanism are:
(1) Vacancies must be supplied to the transforming
particle in order to reduce the particle/matrix
mismatch.
(2) The subsequent particle growth provides the
driving force for the vacancy-emitting climb
of the dislocation.
(3) The sequence between repeated nucleations is
controlled by balancing the rate of vacancy
consumption by the precipitates with the rate
of vacancy emission by the climbing dislocation.

183
(4) The particle is dragged some distance by the
dislocation before unpinning occurs.
(5) The parameters controlling colony growth are
either (a) the atomic diffusion of solute, or
(b) the core self-diffusion, whichever has the
highest activation energy.
We now compare each of the above with the evidence in
Al-Cu. ,
(1) The a+0' transformation involves a 3.95%
volume contraction.
Thus, if anything, the precipitates generate a few vacancies
instead of consuming them.
(2) The dislocation climb is essentially indepen
dent of the precipitation reaction.
Evidence for this is as follows. First, it was shown in
Section 4.2.1 that the dislocation climb is definitely vacancy-
annihilating instead of emitting. As climb occurs during
quenching, it is concluded that the driving force is the
annihilation of the quenched-in vacancy supersaturation. If
any vacancies produced by the transformation contribute to
climb, the experimental evidence suggests that this contribu
tion is small. For example, depending on quench rate and
vacancy supersaturation, the dislocations often climbed over
large distances before precipitation began, e.g., Figures
4.47 and 4.52(b). And as evidenced by Figures 4.25 and
4.44(a-d), the dislocations undergo little, if any, additional
climb during aging, while the precipitates grow orders of

184
magnitude in size.
(3) There is no vacancy balance required between
precipitate and dislocation in Al-Cu.
This conclusion is based on (1) and (2) above.
(4) Particle dragging was not observed in Al-Cu.
There was no visual evidence that the particles had been
dragged by the dislocations during 0' colony growth. If in
fact it did occur, the dragging distances involved must be
smaller than the resolution limit of a small particle near
a dislocation image. This is estimated to be less than 100.
(5) Colony growth in Al-Cu is controlled by the
dislocation climb rate.
It is suggested that the rate of colony growth is primarily
controlled by the rate at which the dislocation climbs through
the lattice after the sample temperature passes below the 0'
solvus. This in turn is controlled by the rate at which
excess vacancies reach the dislocation and depends, therefore,
on the self-diffusion coefficient. In addition, it is influ
enced by the degree of vacancy supersaturation below the 0'
solvus which depends on the solution treatment temperature
and the quench rate (Sections 4.5.2-4.5.4).
5.3. The Mechanism in Al-Cu
In light of the discussion above, it is clear that
repeated nucleation of 0' in Al-Cu cannot be explained by the

185
mechanism of Nes (1974). It does not involve a vacancy flux
problem between growing precipitate and climbing dislocation.
Rather it appears to be a basic heterogeneous nucleation
problem, requiring that the dislocation stress field be pres
ent long enough for nucleation to occur. The mechanism of
repeated nucleation in Al-Cu apparently involves three essen
tial factors:
(1) Nucleation occurs because the 1 dislocation
!
stress field is present to help overcome the
energy barrier.
(2) Repeated nucleation is possible because the
dislocation climbs under a driving force
independent of the precipitation process,
namely the quenched-in vacancy supersaturation.
This assures that the dislocation advances to
act as a catalyst at successive positions
along the climb surface.
(3) The climb rate is apparently slow enough to
allow nucleation, but rapid enough to move
the dislocation past the newly-formed precipi
tates before they grow large enough to pin it.
Factor (3) could be a self-regulating effect in that the
initial precipitate size may be limited by the strain field
of the dislocation. Growth after nucleation relies princi
pally on long-range bulk-diffusion of solute and may not be
fast enough to pin the climbing dislocation.

186
5.3.1. Local Solute Buildup
The matrix is a random distribution of 2-4 wt.% Cu (>1-2
atomic!) in the aluminum lattice. For nucleation to occur
repeatedly, it is necessary that concentration buildups occur
at the moving dislocation which can provide for the high
copper content of the 0! nuclei (v33 atomic!). It is sug
gested that the necessary concentration changes probably do
not come about by long-range bulk-diffusion of copper. Prior
to nucleation there is no concentration gradient to promote
long-range copper diffusion to the dislocation. There is,
though, a drift force on the copper to diffuse to the dislo
cation and lower the associated strain energy. However,
since nucleation occurs rapidly during quenching, it is un
likely that bulk-diffusion due to the drift force is rapid
enough to cause the necessary copper buildups.
In the absence of a sufficient contribution from long-
range bulk-diffusion, we consider two possibilities whereby
the copper concentration can be enhanced locally at the
moving dislocation.
(1) There is indirect evidence from resistivity measure
ments that copper atoms cluster during quenching of Al-Cu
alloys (Perry, 1966). Clusters located at or very near the
climb plane would provide copper-rich regions that might
transform to 0' as the dislocation stress field passes. It
is not valid to assume that, a cluster will automatically
transform to 0' if it contains more copper atoms than arc
required for the critical nucleus size in the presence of the

187
dislocation. The nucleation event still involves a statis
tical probability that the correct combination of copper and
aluminum atoms rearrange into the correct structure with
sufficient size and shape (i.e., the critical nucleus size
and shape). It is valid, however, to assume that the proba
bility that this event will occur is higher in a copper-rich
cluster than elsewhere in the matrix. In addition, it is
generally believed that clusters can trap vacancies with a
binding energy characteristic of the alloy (Federighi and
Thomas, 1961). Such trapped vacancies could aid in the local
atomic rearrangements involved in forming the critical nucleus
structure. However, without available information on cluster
sizes and densities, it is not possible to estimate if clus
tering alone could provide enough solute buildup for nucle
ation in these samples.
(2) If it is assumed that the dislocation acts as a
highly efficient sink for the copper atoms it encounters
along the climb surface, then rapid pipe diffusion along the
core to a nearby growing nucleus can provide sufficient copper
locally for 6' precipitation. Assuming that thermal fluctu
ations create growing embryos of the new phase at various
locations along the dislocation (a basic assumption of nucle
ation theory), then adjacent portions of the climbing dislo
cation could collect and transport solute to these growing
embryos. Since dislocation pipe diffusion is much faster
than bulk diffusion, the frequency of atoms joining the criti
cal nucleus (the term 8 in Equation 2.1) is increased over

188
that for bulk diffusion. This provides an advantage for
nucleation at the dislocation in addition to that gained from
the strain field. To test if this model could provide suffi
cient solute for the observed repeated nucleation, a calcula
tion was made to determine if there was enough copper over
the climb surface to account for the density of precipitates
generated in a colony. The density of precipitates per unit
area, measured from Figure 5.3, was found to be approximately
10 2 '
6.4x10 /(cm) It is assumed that the density of visible
precipitates at this stage is the same as the density of
nuclei which form initially. As discussed in Section 4.2.1,
an a/2[110] dislocation climbing by vacancy annihilation on a
{110} plane in f.c.c. removes two adjacent planes of atoms.
Initially, it was assumed that the dislocation could collect
easily all the solute on these two planes and each plane on
either side of it, or all copper atoms on a total of four
adjacent {110} planes. Assuming 2 atomic! copper (4 wt.! Cu)
on these {110} planes, the dislocation collects 5.8x10^
copper atoms in climbing over a square micron area. If by
pipe diffusion, the dislocation can distribute this amount
of solute to the 640 embryos per square micron, then on the
average there would be 900 copper atoms available to each
embryo. There are two copper atoms per unit cell of 0' so
this amounts to enough solute for 450 unit cells/embryo. An
estimate of the critical nucleus size must now be made to
determine if this is enough solute to form stable nuclei.
This was done on the basis of a plate-like nucleus on (100),

189
Figure 5.3. Micrograph from which a measurement was made
of the number density of precipitates in a
colony. (Heat treatment: S.T. 1 hour 550C,
quenched to 220C, aged 5 minutes.)

190
a thickness sufficient to nucleate one a[100] misfit dislo
cation around its edge (-2 unit cells thick) and a diameter
slightly less than the resolution limit of the electron
microscope (-30 maximum). Such a critical nucleus would
contain about 90 unit cells. Hence, this model can supply
more than enough solute to nucleate the precipitate density
observed, so that it is not required for the dislocation to
collect all the solute in its climb path.
It is suggested that the pipe diffusion model provides
the primary means for locally enhancing the solute concentra
tion for repeated nucleation. However, it is likely that
there is at least some contribution, however small, from
bulk-diffusion to the surface of the growing embryos. The
importance of clustering cannot be accurately assessed, but
it is not a necessary condition. It would be helpful, of
course, in attaining local concentration enhancement.
5.3.2. Precipitate Stringer Formation
All the above considerations would predict a random dis
tribution of 0' nucleated by a passing dislocation. This
could account for the regions of uniform precipitation
observed behind the climbing dislocations, but not for the
regions of straight and uniformly-spaced precipitate stringers
which always form along <100> directions on climb sources,
Figure 4.30. We now consider a possible mechanism for this
stringer formation, and we limit the discussion to stringers
produced in precipitate colonies on climb sources.

191
It was shown in Section 4.5.5 that the spacing between
stringers (about O.ly) is the same as the spacing of visible
superkinks on climb source loops. It is assumed that the
origin of the precipitate stringers is related to the move
ment of such superkinks. According to Balluffi (1969) the
climb of a dislocation which is rotated out of its pure edge
orientation introduces kinks in the dislocation line. The
spacing of kinks depends on the magnitude of the rotation
]
and is equal to F/tan(90-a), where F is the Burgers vector
and represents the kink height, and a is the angle between
the Burgers vector and the dislocation line direction. It is
now assumed that the climb source loops are rotated slightly
off their {110} habit about an axis parallel to the <001>
direction in the habit plane. Such rotations could result
from local slip out of the climb plane due to the stress
fields of the precipitates forming in advance of the climbing
loops. A kink spacing of O.ly implies a rotation of only
0.13 for kinks of one Burgers vector in height. Such kinks
would not be resolved in the electron microscope, as are
those in Figure 4.65. Rather, it is assumed that the rota
tion is on the order of several degrees, creating numerous
kinks. It is further assumed that at the high temperatures
where climb occurs, the kinks are mobile and coalesce into
visible superkinks with an equilibrium spacing of O.ly. Then
each time a source loop expands and increases its line length
by an amount O.ly (projected normal to <001>) a new super
kink becomes stable. It is assumed that the precipitate

192
nucleation process associated with the kink then becomes
active and contimies as the loop expands. This process of
forming successive superkinks and trailing precipitates
behind them as the loop expands is illustrated in Figure 5.4.
For clarity, the precipitate stringers are shown to be con
tinuous. The shape in Figure 5.4 compares favorably with the
shape of precipitate colonies observed on climb sources,
e.g., Figure 4.50.
Several further aspects of the above model are now con
sidered. First, the fact that the superkink configuration
is apparently a favored nucleation site is probably a func
tion of the stress field set up around it. In addition, pipe
diffusion of solute along the dislocation is most probably
slowed at these superkinks. Thus, solute builds up at the
kinks providing both a solute-rich environment for nucleation
and a continuing supply of solute for growth. Next, we con
sider why superkinks and their trailing precipitate stringers
are confined in their motion to the <1Q0> direction in the
loop habit. It may be that only one nucleation event occurs
near the kink in advance of the climbing loop. Then, as the
dislocation climbs by (or through), pipe diffusion to the kink
continues to supply solute so that a long, thin ribbon of 0'
trails behind the advancing kink. It can be seen from Figure
5.5 that the two possible {100} orientations of 0* which
nucleate on a given loop could be extended from such super
kinks only along the <100> direction contained in the loop
habit, provided the loop continues to expand macroscopically

193
Figure 5.4. Diagram illustrating the shape of precipitate
stringer regions which would result from
successive nucleation of superkinks of spacing
a, as a circular climb source loop expands.
The precipitate stringers are shown continuous
for clarity.

194
Figure 5.5. Diagram of an expanding climb source loop on
(110) with Burgers vector a/2[110]. The two
0' orientations, (100) and (010), which
nucleate on this dislocation, can be extended
in continuous ribbons from moving superkinks
only along the [100] direction.

195
in its habit plane. This is in agreement with the fact that
the stringers are always aligned along <100>.
This model predicts a thin, continuous ribbon of pre
cipitate, but it was shown in Figure 4.27 that the stringers
are composed of separate, small particles when they are first
visible after quenching. It is probable that immediately
after a thin ribbon forms, surface energy effects cause it to
pinch off into small platelets, in a manner similar to the
initial rapid spheroidization of eutectoid platelets upon
annealing. Further aging would then cause these platelets
to grow and coalesce as shown in Figure 4.31. This continu
ous ribbon model is consistent with the observation that
platelets of only one 0' orientation exist in any given
stringer (Section 4.4.2). In fact, it is difficult to imagine
a more logical interpretation.
It is suggested that initial rapid growth of the pre
cipitates in these stringers, enhanced by pipe diffusion
along the dislocation, creates a sufficient concentration
gradient in the matrix such that this is the only mechanism
which continues to operate once it begins. This could account
for the fact that uniform nucleation of precipitates does not
occur in the regions where the stringer mechanism operates
(Figure 4.30) .

196
5.4. Criteria for Repeated Nucleation in Al-Cu
and Application to Other Systems
Based on the results reported herein, the following
criteria may be established for repeated nucleation in this
system:
(1) a phase which nucleates easily on dislocations,
namely the 0 phase;
(2) a source or sources of dislocations during the
quench, namely dislocation climb sources and
glide dislocations generated during quenching;
(3) a driving force for dislocation climb which is
independent of the precipitate reaction, namely
the annihilation of quenched-in vacancies; and
(4) a climb rate which is sufficiently slow to.
permit nucleation, but sufficiently fast so
that the precipitates do not grow so large as
to effectively pin the dislocations.
These criteria are far simpler than those required by the
mechanism of Nes (1974), and might well apply to other sys
tems in which dislocation-nucleated transformations occur.
(2) and (3) above apply to most aluminum alloys. In par
ticular, Al-Mg and certain Al-Mg ternary alloys have mestable
phases which nucleate on dislocations, and appreciable dis
location climb occurs when these alloys are quenched. Unlike
copper in Al-Cu, the magnesium atom is larger than the aluni-
num atom and it is unclear what effect this may have. The
Al?CuMg phase nucleates easily as laths on dislocations in

the Al-Cu-Mg system and might be a candidate for repeated
nucleation. In steels, the Ni^Ti phase nucleates easily on
dislocations in austenitic stainless steel. This transforma
tion, like 0' in Al-Cu, involves a net volume contraction an
could be a candidate for repeated nucleation.

CHAPTER 6
CONCLUSIONS AND SUGGESTIONS FOR FUTURE WORK
This research has shown that the a+0' transformation in
Al-Cu can initiate and propagate during quenching by the
mechanism of repeated nucleation on climbing dislocations.
The dislocations are generated and climb during quenching by
annihilation of the quenched-in vacancies. Densely populated
colonies of 0' precipitates nucleate in the passing stress
fields of the climbing dislocations. The distribution of
the entire volume fraction of 6' is thus determined during
the quenching step (£10 1 seconds). Therefore, autocatalytic
nucleation is not required to propagate the reaction during
aging.
The goals of this research were twofold: (1) to charac
terize the microstructures resulting from repeated nucleation,
and (2) to establish the repeated nucleation mechanism in
this system. Pertaining to microstructure, the following
conclusions were drawn:
(1) The dislocations which climb during quenching fall
into three categories according to origin:
a. pure-edge loops on {110} planes with
a/2<110> Burgers vectors, generated at
dislocation climb sources.
198

3 99
b. a/2<110> glide dislocations on {111} planes.
c. prismatic edge-loops on {110} with
b=a/2<110>, formed by the collapse of
vacancy clusters.
The dislocation density in each category varied with heat
treatment. Categories (a) and (b) were found in all samples
regardless of quench procedure. Category (c) was found only
in samples quenched into oil or water at room temperature.
(2) Glide dislocations climb either (a) on smoothly
curved surfaces, nucleating precipitate colonies containing
two 0' orientations, or (b) on corrugated-shaped surfaces,
nucleating precipitate bands containing only one 0' orienta
tion per band.
(3) Only the two 0' orientations compatible with the
Burgers vector of the climbing dislocation are nucleated in
a given colony (with the exception of 2(b) above).
(4) Precipitate colonies generated on climb sources
exhibit a central region of uniform precipitation, and two
regions of precipitate stringers aligned in <100> directions.
The origin of the stringers is probably associated with the
movement of regularly spaced superkinks on the climbing
source loops. Only one 0' orientation is nucleated in a
given stringer.
(5) 0' platelets nucleated by repeated precipitation
can serve as source particles for generation of "secondary"
climb sources which nucleate more precipitates. In a sense,
this is a form of autocatalytic propagation of the reaction.

200
(6) The number density and the average size of active
climb sources are functions of the quench rate and vacancy
supersaturation for direct-quenches The typical shape of a
source loop changes from an ellipse to a rhombus with decreas
ing copper concentration.
(7) Repeated nucleation of 0' does not occur in alloys
with copper concentration below some value between 1.96 and
1.0 wt.%,
The morphologies resulting from repeated nucleation con
tained many fascinating features which had not been previously
reported, and it is certain that further investigation of
these microstructures would reveal even more details not dis
closed by this initial work.
The mechanism of repeated nucleation of 0' in Al-Cu is
different from other previously reported mechanisms. The
vacancy-annihilating dislocation climb appears to be inde
pendent of the precipitation process. Repeated nucleation
is possible simply because the climb rate is slow enough to
permit nucleation in advance of the moving dislocation, but
rapid enough to permit the dislocation to pass before parti
cle growth can result in effective pinning. The fact that
the phenomenon occurs during quenching assures that there is
a continuous supply of non-equilibrium vacancies to move the
dislocation along and present fresh sites for further nucle
ation. The criteria for repeated nucleation by this mechanism
are :
(1) a phase which nucleates easily on dislocations,

201
(2) a source of dislocations during quenching,
(3) a driving force for dislocation climb which
is independent of the precipitation process,
(4) a climb rate slow enough to permit nucleation
but rapid enough to move the dislocation past
the stable precipitates.
This research has provided a basic understanding of the
phenomenon in Al-Cu, but it is by no means complete. Several
topics for further investigation are suggested:
(a) If a dislocation can climb through the lattice at
elevated temperatures and provide the catalyst for repeated
nucleation, then perhaps a gliding dislocation can also. It
would be instructive to examine samples deformed either at
temperatures slightly below the 0 solvus or during quenching
from the solution treatment temperature for evidence of
repeated nucleation during glide.
(b) It is suggested that a small-angle x-ray investiga
tion at temperatures near the 0' solvus would provide direct
evidence for copper clustering if it exists. If so, a
determination of the density and size distribution of clus
ters would help resolve whether or not clustering could be
involved in the repeated nucleation mechanism.
(c) Trace element additions are known to alter the
kinetics of the a->0' transformation, presumably by segregat
ing to the particle interface where they can change both the
interfacial energy and, to some extent, the misfit strain.
It would be of interest to determine if trace additions of

202
such elements as Cd, In, or Sn could either enhance or elimi
nate repeated nucleation during quenching.
(d) An item, examined to some extent here, which needs
further investigation is the nature of the segmented climb
of glide dislocations in this system. In particular it would
be of interest to understand more completely why this mode
of climb nucleates precipitate bands containing only one e'
orientation instead of two.
(e) As a by-product, this investigation has established
a good understanding of the operation of dislocation climb
sources in Al-Cu alloys, but some questions remain to be
answered. First, the nature of the source particles is still
unknown. The capability of the new breed of scanning trans
mission electron microscopes (STEMs) to focus the electron
O
beam to diameters on the order of 15, and to thereby obtain
microdiffraction patterns and/or x-ray chemical analysis from
very small particles, might be employed to solve this problem.
Secondly, further work is required to understand the depen
dence of the shape of source loops on solute concentration.
(f) It is suggested that a calculation of the inter
action energy between a 0' nucleus and an edge dislocation
would reveal the location of the nucleation events around the
dislocation. The method outlined by Larch (1974) for a
coherent nucleus at an edge dislocation should be modified
for the known geometry of a 0' platelet on {100} at an edge
dislocation on (110). The position which maximizes the inter
action energy (more negative) should be sought.

203
(g) Finally, the criteria for repeated nucleation in
A'l-Cu are relatively simple and probably exist in a number
of other systems. it. may be that the effect has not been
observed in some alloys that have phases which nucleate on
dislocations, simply because slow quenches are not normally
employed in laboratory investigations. Commercial quenching
practice is often a different situation though, where thick-
section parts almost certainly receive relatively slow
quenches. It Is suggested that other aluminum alloys, which
meet the criteria above, should be investigated to determine
if the mechanism lias broader application than the Al-Cu
system. Al-Mg or Al-Mg ternary alloys should receive first
consideration.

BIBLIOGRAPHY
Aaronson, H. I., Kinsman, K. R., and Russell, K. C. (1970),
Scripta Met. 4_, 101.
Ashby, M. F. and Brown, L. M. (1963) Phil. Mag. 8_, 1649.
Balluffi, R. W. (1969), Phys. Stat. Sol., 31, 443.
Balluffi, R. W., and Thomson, R. M. (1962), J. Appl. Phys.,
33, 817.
Bardeen, J., and Herring, C. (1952), Imperfections in Nearly
Perfect Crystals, Wiley and Sons p. 2 61.
Barnett, D. M. (1971), Scripta Met. 5_, 261.
Barnett, D. M. (1973), private communication.
Betn, R. H., and Rollason, E. C. (1957-58), J. Inst. Metals
86, 77.
Boyd, J. D. and Edington, J. W. (1971), Phil. Mag., 23, 633
Bullough, R., and Newman, R. C. (1959), Proc. Royal Soc.
London, A249, 427.
Cahn, J. W. (1957), Acta Met. 5_, 169.
Christian, J. W. (1965) The Theory of Phase Transformations
in Metals and Alloys^ Pergamon Press, p. 363.
Clarebrough, L. M. (1973), Phys. Stat. Sol, (a), 18 427.
Darwin, C. G. (1914), Phil. Mag., 27, 315, 675.
Dollins, C. C. (1970), Acta Met., 18, 1209.
Edington, J. W., and West, D. R. (1966), J. Appl. Phys., 37,
3904.
Embury, J. D. (1963), Ph.D. Dissertation, University of
Cambridge.
Embury, J. D., and Nicholson, R. B. (1963), Acta Met., 11,
347.
204

Eshelby, J. D. (1957), Ptoc. Royal Soc, London, A241 376.
Eshelby, J. D. (1961), Progress in Solid Mechanics, Sneddon
and Hill, eds. North -Holland, p. ¡T9T
Federighi, T. and Thomas, G. (1961), Phil. Mag. 6_, 127.
Frank, F. C., and Read, W. T. (1950), Phys. Review, 79, 722.
Friedel, J. (1964), Dislocations, Addison-Wesley.
Guerard, B. von, Peisl, H., and Zitzmann, R. (1974), Appl.
Phys., 3, 37.
Guinier, A. (1938), Nature, 142, 569.
Guinier, A. (1939), Ann. Physique, 12, 161.
Guinier, A. (1942), J. Phys. Radium, _3, 129.
Guinier, A. (1950), Compt. Rend., 231 655.
Guinier, A. (1952) Acta Cryst. 5_, 121.
Gulden, M. E. and Nix, W. D. (1968) Phil. Mag. 18_, 217.
Head, A. K. (1969), Aust. J. Phys., 'n, 43.
Head, A. K., Humble, P., Clarebrough, L. M., Morton, A. J.,
and Forwood, C. T. (1973), Computed Electron Micrographs
and Defect Identifica!ion, Amelinckx, S'. Gevers R7,
and Nihoul, J., eds., North-Holland.
Hirsch, P. B., Howie, A., Nicholson, R. B., Pashley, D. W.,
and Whelan, M. J. (1965), Electron Microscopy of Thin
Crystals, Butterworths.
Hirth, J. P., and Lothe, J. (1968), Theory of Dislocations,
McGraw-Hill.
Hornbogen, G. (1967), Aluminum, 3, 163.
Hosford, W. F., and Agrawal, S. P. (1974), paper presented
at the 6th Annual Spring Meeting of A.I.M.E., Pittsburgh,
Pa.
|, A. ,
and
Whelan, M.
J. (1961), Proc.
Royal
Soc. London,
A2 6 3 ,
217.
A. ,
and
Whelan, M.
J. (1962), Proc.
Roval
Soc. London,
A26 7 ,
206.
. e P .
(19 £
>8) Aust.
J. Phys., 21, 32 5

206
Kelly, A., and Nicholson, R. B. (1963), Progress in Materials
Science, 3, Chalmers, B. ed. Pergaiiion Press, p. 149.
Laird, C. and Aaronson, II. I. (1968), Trans. AIME, 242 1393.
Larch, F. C. (1974), chapter to be published in book by
Nabarro, F.R.N.
Lorimer, G. W. (1968) paper presented at the Fourth European
Regional Conference on Electron Microscopy, Rome, Italy.
Lorimer, G. W. (1970), Fizika, 2_, Suppl. 2, p. 33.1.
Lorimer, G. W., and Nicholson, R. B. (1969), The Mechanism
of Phase Transformations in Crystalline Solids^ Inst.
of Metals, p. $6.
Lothe, J. (1960), J. Appl. Phys., 31, 1077.
Lyubov, B. Y., and Solov'Yev, V. A. (1965), Fiz. Metal.
Metalloved. 19_, 333.
Marshall, G. W., and Brittain, J. 0. (1974), paper presented
at the 6th Annual Spring Meeting of A.I.M.E., Pittsburgh,
Pa.
Miekk-oja, H. M., and Raty, R. (1971), Phil. Mag., 24, 1197.
Nes, E. (1974), Acta Met., 22, 81.
Nicholson, R. B. (1970), Phase Transformations, American
Society for Metals, p. 269".
Nicholson, R. B., and Nutting, J. (1958), Phil. Mag. 3_, 531.
Passoja, D. E., and Ansell, G. S. (1971), Acta Met., 19, 1253.
Perry, A. J. (1966), Acta Met., 14, 305.
Preston, G. I). (1938a), Nature 142 570 .
Preston, G. D. (1938b), Proc. Royal Soc. London, A16 7, 526.
Preston, G. D. (1938c), Phil. Mag., 26 855.
Ramrez, R. G., and Pound, G. M. (1973), Met. Trans. 1563.
Reed-Hill, R. E. (1973), Physical Metallurgy Principles,
Van Nostrand, p. 170.
Ravi, K. V. (1971), Met. Trans., 3, 1311.

207
Seidman, D. N. and Balluffi, R. W. (1968), Lattice Defects
and Their Interactions, Hasiguti, R. R., ed. Gordon
and Breach, p. $13.
Servi, I. S., and Turnbull, D. (1966), Acta Met., 14, 161.
Silcock, J. M. Heal, T. J., and Hardy, H. K. (1953-54),
J. Inst. Metals, 82, 239.
Silcock,
361.
J. M. ,
and
Tunstall, W. J.
(1964) Phil.
Mag., 10 ,
Simmons,
52,
R. 0. ,
117.
and
Balluffi, R. W.
(1960) Phys.
Review,
Thomas, G. (1959), Phil, Mag., 4, 1213.
Thomas, G. and Nutting, J. (1956) The Mechanism of Phase
Transformations in Metals, Institute of Metals, p. ST.
Thomas, G. and Whelan, M. J. (1961), Phil. Mag. 6_, 1103.
Thomson, R. M., and Balluffi, R. W. (1962), J. Appl. Phys.,
33, 803.
Warren, J. B. (1974), Ph.D. research in progress.
Weatherly, G. C., and Nicholson, R. B. (1968), Phil. Mag.,
17, 801.
Westmacott, K. H., Hull, D., and Barnes, R. S. (1959),
Phil. Mag. 4_, 10 89.
Westmacott, K. H., Barnes, R. S., and Smallman, R. E. (1962)
Phil. Mag., 7, 1585.
Whelan, M. J. and Hirsch, P. B. (1957), Phil. Mag. Z_, 1121
Wilsdorf, H. and Kuhlmann-Wilsdorf, D. (1955) Defects in
Crystalline Solids, Physical Society, p. 175.

BIOGRAPHICAL SKETCH
Thomas Jeffrey Headley was born June 22, 1943, at
Sheffield, Alabama. In June, 1961, he was graduated from
Coffee High School, Florence, Alabama. In June, 1965, he
was graduated from Virginia Polytechnic Institute with the
degree of Bachelor of Science in Metallurgical Engineering.
In June, 1967, he was graduated from Virginia Polytechnic
Institute with the degree of Master of Science in Metallur
gical Engineering. He worked, first as an Engineer, and
then as an Associate Scientist for the Lockheed-Georgia
Company, Marietta, Georgia, from July, 1967, to April, 1970.
He entered graduate school at the University of Florida in
April, 1970. Since that date, he has worked as a graduate
research assistant in the Materials Science and Engineering
Department while pursuing the degree of Doctor of Philosophy.
Thomas Jeffrey Headley is married to the former Lynn
Lancaster Moore and is the father of two children. He is a
member of the American Institute of Mining, Metallurgical,
and Petroleum Engineers, the American Society for Metals,
Alpha Sigma Mu, and the Society of Sigma Xi.
20 8

1 certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
John'J. Hren, Chairman
Processor of Materials Science
and Engineering
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
^ v.
P. N. Rhines
Professor of Materials Science
and Engineering
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
Vl
g.l. Q*l-UtP
Rl IP Reed-Hill
Professor of Materials Science
and Engineering
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
/
T.' "tfeHoff
Professor
and Engineering
Science

I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
Associate Professor of Materials
Science and Engineering
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
y. A. Eisenberg *
Associate Professor of Engineering
Science and Mechanics and
Aerospace Engineering
This dissertation was submitted to the Graduate Faculty of
the College of Engineering and to the Graduate Council, and
was accepted as partial fulfillment of the requirements for
the degree of Doctor of Philosophy.
August, 1974
Dean, Graduate School



118
4.4.5. A Climb Source on (100)
On one occasion, when a foil was being examined in the
(001) orientation, a planar, precipitate colony lying on (100)
was observed. This colony is shown in Figure 4.36(a) at A.
It is different from all other planar colonies observed which
were nucleated by dislocation climb sources operating on
{110} planes, e.g., at B in Figure 4.36(a). When the foil
was tilted away from the (001) orientation, the colony was
found to be bounded by an arc of a dislocation loop, Figure
4.36(b). Thus, it had the appearance of a climb source loop
and associated precipitate colony, but lying on a (100) plane.
This configuration was photographed on different two-beam
reflections in several orientations. Selected images are
shown in Figure 4.37. Invisibility conditions were used to
determine the Burgers vector of the dislocation arc and orien
tation of the precipitates in the colony. This information
is summarized in Table 4.2. The dislocation arc is invisible
for g=02Q and 022. This eliminates all a<001> and a/2<110>
Burgers vectors except a[100] and a/2[Oil]. Further, it is
visible for jf=IIl and 311. This eliminates a/2 [Oil] but not
a[100]. All partial dislocations with Burgers vector a/3
or a/6<112> are eliminated by comparison of the data in Table
4.2 with the visible and invisible criteria for partial dis
locations, namely, invisible for g*5 = 0 or 1/3 and visible
for all other products (Silcock and Tunstall, 1964; Hirsch
et al. 1965). It is concluded that the Burgers vector of
the dislocation is a[100]. Since it lies on (100), it must


73
Figure 4.15. Precipitate colony nucleated by climb of a
glide dislocation, initially on (111), with
D=a/2[10]. The dislocation has been sectioned
twice by the foil, leaving arcs AB and CD in
(a). The beam direction is close to [Oil] in
(a) and to [111] in (b) (Heat treatment:
S.T. 1 hour 550C, quenched to 220C, aged
5 minutes.)


Ibl
Figure 4.49. Continued


Copyright by
Thomas Jeffrey Headley
1974


7
AG(chem) is negative whereas asur£ and AG(strain) are posi
tive. For a given reaction, AG(chem) can be calculated from
thermodynamic parameters by the method of Aarons on et al.
(1970) Attempts have been made to calculate crsur£ from
atomic bond models, but in general, the binding energies are
unknown. In the absence of a proven model, taken as the measured bulk interfacial energy. The validity
of this approximation is questionable. AG(strain) can be
calculated by the method of Eshelby (1957,1961).
If a nucleus forms in the stress field of a dislocation,
an additional energy term arises from the interaction of the
displacement field of the precipitate with the stress field
of the dislocation. We can then write AG* as
AG* =
K(a
surf)
[AG(chem)+AG(strain)+AG(int)]
AG(int) is negative and acts to reduce the positive AG(strain)
term, so that it represents a major part of the advantage
gained from nucleation at a dislocation. If AG(strain) is
large (i.e., the precipitate misfit is large), nucleation
at dislocations may be the only way the reaction can initiate.
In addition to the AG(int) term, a second advantage for
nucleation at a dislocation arises from the pre-exponential
6 term in Equation (2.1). This term is the frequency with
which atoms join the nucleus and depends on solute diffusion
to the nucleus and across the interface. Solute pipe-
diffusion along a dislocation core is always faster than


Abstract of Dissertation Presented to the Graduate Council
of the University of Florida In Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
REPEATED NUCLEATI ON OF PRECIPITATES
ON DISLOCATIONS IN ALUMINUM-COPPER
By
Thomas Jeffrey Headley
August, 1974
Chairman: John J. Hren
Major Department: Materials Science and Engineering
Results are presented of an investigation of a newly
discovered propagation mechanism for the a->0 transformation
in Al-Cu: repeated nucleation on climbing dislocations. It
was found that during the quench, dislocations are generated
and climb by the annihilation of quenched-in vacancies.
Densely populated colonies of 6' precipitates nucleate in the
stress fields of the climbing dislocations. In this way, the
distribution of the entire volume fraction of 0' is estab
lished during the quench.
The climbing dislocations were found to be a/2
type, falling into three categories according to origin:
(1) pure-edge loops on {110} habits nucleated at dislocation
climb sources, (2) glide dislocations initially on {111},
and (3) pure-edge loops on {110} formed by the collapse of
vacancy clusters.
The effects of solution treatment temperature, aging
temperature, quench rate, and solute concentration on the
vii


149
Figure 4.49. Sequence of micrographs showing the effect on
microstructure of the temperature to which
samples were direct-quenched from 550C and
aged. The samples were quenched to (a) 300C
and aged 15 seconds, (b) 250C and aged 1
minute, (c) 220C and aged 5 minutes, (d) 200C
and aged 30 minutes, and (e) 180C and aged
1 hour.


144
Figure 4.47. Large dislocation climb source in a sample
quenched from 570C to 220C and aged 5
minutes.


83
d
Figure 4.21. Continued.


147
4.5.3. Effect of Temperature to Which
Samples Are Direct-Quenched
In this experiment, five samples were solution treated
for one hour at 550C, direct-quenched into oil at various
temperatures above the G.P. zone solvus, and aged (Figure
4.48). Therefore, the basic difference between samples was
the quenching rate. The temperatures were 180C, just above
the G.P. zone solvus; 200 and 220C, just below and above
the 0" solvus; 250 and 300C, increasing temperatures in the
a+9' region. The oil bath was limited to 300C. The tem
perature to which each sample was quenched shall be referred
to as the aging temperature, Ta. With decreasing Ta, the
samples were aged for longer times so that the 0' precipitates
would be visible. The resulting microstructures are shown at
low magnification in Figure 4.49.
In general, 0 nucleated repeatedly on climbing disloca
tions during quenching in all samples, although only to
small degree in the sample quenched to 300C (Figure 4.49(a)).
Furthermore, both glide dislocations which climbed and climb
sources were observed in all samples. Whereas the density of
glide dislocations varied little from sample to sample, the
density of climb sources varied greatly.
The sequence of micrographs in Figure 4.49 shows that
decreasing Ta increases the number of active climb sources
and decreases the average source loop diameter. The maximum
source density occurred for Ta=180C and was about 6x10^ per
grain (Section 4.2.1).


6
where J
Z
3
N
s
AG* =
k
T
AG* has the
nucleation rate,
the Zeldovich factor,
frequency which atoms add to the nucleus,
number of available nucleation sites,
composition of the nucleus,
free energy of the critical nucleus,
the Boltzmann constant, and
temperature.
form
AG*
K (a J3
^ surf7
(AGdrlye)Z
(2.2)
where K = a shape factor,
surf = Particle/matri-:x interfacial energy, and
n T* T VP
AG = driving force for the reaction.
For homogeneous nucleation, is high assuring reasonable
nucleation rates. For heterogeneous nucleation, Ng is low
and the nucleation rate is usually dominated by the exponen
tial dependence on the free energy of the critical nucleus.
For precipitation in solids, there is a chemical free
energy change per unit volume tending to drive the transfor
mation to the new phase. If the atomic volumes in the matrix
and particle are different, there is a misfit strain energy
associated with formation of the new phase. Thus the free
energy of the critical nucleus can be written
AG*
K (a r)
v surf^
1
[ AG(chem)+AG(strain)]


146
of undissolved source particles could lead to the observed
maximum density of sources in the sample quenched from 550C.
That is, at 570C, there would be a minimum number of undis
solved source particles, and a maximum vacancy concentration.
Upon quenching fewer sources would generate much larger source
loops compared to the structure quenched from 550C. At
504C, the undissolved particle density would be a maximum,
but the vacancy concentration is a minimum. Upon quenching
only a minimum operation of climb sources is required to pro
vide sinks for vacancy annihilation.
In the sample quenched from 570C, repeated nucleation
of 9' occurred within a limited band, approximately one
micron wide, immediately behind the final position of the
dislocations (Figures 4.46(a) and 4.47). As this sample was
quenched from the highest temperature and contained the
largest vacancy supersaturation, it is believed that appreci
able dislocation climb occurred above the 0' solvus tempera
ture, where no 0 could nucleate. The well-defined boundary
where precipitation begins indicates that the concept of a
0' solvus temperature is a very effective and sensitive bar
rier to nucleation during quenching. In addition, the slope
of the 0' solvus at the composition 3.85 wt.% Cu is small,
causing a large driving force for nucleation immediately
below it.


179
resulting strain puts the surrounding matrix in tension.
Thus, this strain can be relieved only if the 0' platelet
forms on the compressive side of the dislocation. In order
to determine how close to the dislocation it forms, one must
solve the elastic interaction problem discussed in Section
2.1. In the case of the pure-edge loops generated at climb
sources in these alloys, the compressive region lies in front
of the expanding loops, Figure 5.1(a). Hence, nucleation
occurs in advance of the climbing dislocations, Figure 5.1(b)
Since essentially all the visible precipitates in a colony on
a climb source lie within the loops (e.g., Figure 4.30), the
loops must climb through or by the particles, or pinch off
around them in order to continue expanding. Now 0' forms
parallel to {10 0 > with a<100> peripheral misfit loops, so
that the climbing dislocation with a/2<110> Burgers vector
cannot comprise a part of the platelet edge. And as dis
cussed in Section 4.3.3, there was no evidence that the climb
ing dislocations pinched off around the platelets. If a
nucleus forms above or below the climb plane, the climbing
dislocation can pass by more easily than if the nucleus forms
along the climb plane (or grows to intersect it). In the
latter case, the dislocation must either cut through the pre
cipitate or, if the stresses involved are too high, glide
locally around it.
Occasionally a large precipitate was observed outside a
source loop which bent locally around the precipitate, Figure
5.2. It is assumed that such precipitates were the last to


74
precipitate colonies made with the foil surfaces, Figure 4.16.
Similarly, Miekk-oja and Raty (1971) observed repeated nucle-
ation of silver-rich precipitates from solid solutions of
silver in copper behind dislocations which were shown to be
a/2<110> glide dislocations on {111} planes before climbing.
They found that these dislocations subsequently climbed in
one of two different ways: (1) off the slip plane on smoothly
curved surfaces, similar to that described above, or (2) into
a crooked shape so that different segments of the disloca
tions climbed on different low-index planes intersecting the
original slip plane. They further showed that these low
index planes were of the types {110} and {ill}, i.e., the
planes of "easy climb" (Section 2.3.1) on which the chemical
climb force, from a subsaturation of vacancies, was the
greatest.
In the present research, the shapes of precipitate
colonies behind certain glide dislocations suggested that
different segments of these dislocations had climbed on
separate crystallographic planes also. The term "segmented
climb" shall be used here to refer to this mode of climb.
Micrographs of precipitate colonies apparently resulting from
segmented climb are shown in Figures 4.17 and 4.14(b). In
Figure 4.17, the dislocation between AB has climbed through
the lattice from left to right. The precipitate colony
nucleated by this dislocation is separated into bands of
precipitates. This effect is thought to be associated with
the climb of adjacent dislocation segments on separate


20
and Lothe (1968) and a general review of the mechanisms has
been given by Balluffi (1960).
Briefly, dislocation cLimb occurs by (1) the absorption
of vacancies onto the dislocation core, (2) diffusion of the
vacancies along the core to jogs, and (3) subsequent move
ment of the jogs by destruction of the vacancies. This
sequence is illustrated in Figure 2.6 for climb of an undis
sociated edge dislocation (a similar model applies for climb
by vacancy-emission). Then, according to Balluffi (1969),
the dislocation climb velocity is
v =
2ttD-, b^ [c (R) cv) j
r r2Zv L 2z ,-R-v -i
* ln(d]
where D.
vacancy diffusivity in the lattice,
magnitude of the Burgers vector,
c(R) = vacancy supersaturation at a large distance
R from the dislocation,
.0
vacancy concentration maintained in the
lattice in equilibrium with the jogs,
mean migration distance of a vacancy along
the core before jumping off, and
A
= jog spacing.
Seidman and Balluffi (1968) surveyed the available experi
mental data on climb rates and concluded that, in the pres
ence of moderate to large supersaturations, climb in aluminum
appears to be highly efficient. In other words, jog produc
tion and motion is sufficiently fast that the climb rate is
limited only by the diffusion of vacancies to the dislocation,
and the dislocation acts as a perfect line sink.


Temperature
ii
Figure 2.1. The aluminum-rich end of the Al-Cu phase
diagram, including the solvus lines for
G.P. zones, 0", and 0' precipitates.


156
had climbed during quenching and had nucleated precipitate
colonies, Figure 4.51(a). Climb sources were found only
occasionally, but they had nucleated precipitate colonies
with the same general appearance as those in direct-quenched
specimens. A small group of sources is shown in Figure 4.51(b).
Liquid nitrogen quench
The structure of this sample was similar to that of the
air-cooled sample. The density of glide dislocations was
approximately twice that in the air-cooled sample. Both small
and large climb sources were present and all had nucleated
precipitate colonies, Figure 4.52(b).
Room-temperature oil quench
The dislocation structure changed abruptly in going to
a room-temperature oil quench. The resulting microstructure
is shown in Figure 4.53 at low magnification. The foil is
full of small dislocation loops. In addition, there were
some glide dislocations present which had climbed during
quenching and had nucleated precipitate colonies (Figure 4.53
at A and B). Note that the density of small loops is reduced
in the vicinity of the glide dislocations. Some climb sources
were observed, but their density was much lower than in the
samples direct-quenched into oil at the aging temperature.
Figure 4.54 shows an area where a number of climb sources are
grouped together. The density of small loops was low in the
regions adjacent to climb sources.


110
The Burgers vectors of the loops in stringers 1 are iden
tified as a[001] and those in stringers 2 as a[100]. These
are consistent with the Burgers vectors of the peripheral
loops of the two orientations of 0' platelets expected in
this colony.
Now it will be shown how the data of Table 4.1 rule out
other possible origins for the loops:
(1) It is conceivable that kinks in the climbing source
loop could cause portions to trail behind and pinch off,
leaving the small loops. However, the small loops would then
have the same a/2[101] Burgers vector as the source loop.
Table 4.1 shows that the small loops are visible on several
reflections where the source loop is invisible. Therefore
they cannot be pinched-off loops.
(2) If the 6' platelets form in front of the advancing
source loop, the source loop could pinch off around the plate
lets, reacting with the a<100> peripheral loops around the
platelets. Two reactions are possible, depending on the
platelet orientation:
a/2 [101] + a[100] * a/2[301] for 0 on (100),
a/2 [101] + a [001] -* a/2[103] for 0' on (001).
However, both resultant a/2<103> Burgers vectors can be ruled
out since the loops in stringers 1 are invisible for the 220
reflection.
(3) Frank sessile loops, having a/3 Burgers vec
tors, can form by the collapse of vacancy clusters on {111}
planes. Such partial dislocations are invisible for g b=0


93
230C, these dislocations were rephotographed and are shown
in (c) and (d). Clearly the precipitate colonies existed in
the climb paths of the dislocations after quenching, and the
nine-minute aging treatment has caused growth to visible
sizes, points A in the micrographs. Careful observation is
required to see the edge-on 0' platelets at A in the precipi
tate colony in (d). It is now apparent that this dislocation
underwent segmented climb (Section 4.2.2). The random pre
cipitation at points B throughout the micrographs in (c) and
(d) results from 0' precipitation on the foil surfaces, an
effect which is known to happen when thin foils of Al-Cu
alloys are heated in the electron microscope (Thomas and
Whelan, 1961).
From the pictures in Figure 4.24 alone, the argument
could be made that solute atoms may have segregated to the
dislocations and were then left behind when the dislocations
climbed away, creating supersaturated layers of copper in
the climb paths. Then aging at 230C caused nucleation and
growth of the precipitates from the supersaturated layers.
However, if this were the case, all three orientations of 0'
should nucleate in a given precipitate colony, and this is
inconsistent with the observations discussed above.


126
Ab=a/2[011]
G | (Oil) C
R s' XjITI) 5
X (011) A I (011) Y
(111)
Figure 4.39. Model for segmented climb on the planes of
easy climb. The dislocation, initially in
its edge position XY on the (111) plane,
climbs onto the planes of easy climb and
assumes the curved shape RS (after Miekk-oja
and Raty, 1971).


the Al-Cu-Mg system and might be a candidate for repeated
nucleation. In steels, the Ni^Ti phase nucleates easily on
dislocations in austenitic stainless steel. This transforma
tion, like 0' in Al-Cu, involves a net volume contraction an
could be a candidate for repeated nucleation.


16
but not for a 0' platelet on (001) whose misfit vector is
perpendicular to the Burgers vector. Hence, the (001) orien
tation gains no advantage by nucleating in the stress field
of the dislocation. Likewise for a dislocation with F=a[100],
only the (100) orientation of 0' should nucleate in its stress
field.
Many early TEM investigations of the 0' phase were con
ducted after long aging treatments at high temperatures in
the a+0' field (Figure 2.1). The resulting microstructures
contained a uniform distribution of large 0* platelets, and
it was initially concluded that these 0* platelets were
nucleated by a random distribution of pre-existing 0". How
ever, as it became clear that 0' nucleates at dislocations
and not at 0", the problem of how the random distribution of
0' could form by quenching and aging alone remained unresolved
until the work of Lorimer (1968,1970). Lorimer showed that
the a+0' reaction could propagate from an initially low dis
location density, introduced during quenching, by an auto-
catalytic nucleation mechanism. Early during the aging
period, the initial dislocations become saturated with 0'.
These platelets then grow into the matrix and produce their
own stress fields which aid the nucleation of more precipi
tates. With long aging, the reaction propagates in bands
spreading out from the dislocations to fill the structure with
a uniform distribution of 0' platelets on all three {100}
orientations. Until the present research, this was the only


192
nucleation process associated with the kink then becomes
active and contimies as the loop expands. This process of
forming successive superkinks and trailing precipitates
behind them as the loop expands is illustrated in Figure 5.4.
For clarity, the precipitate stringers are shown to be con
tinuous. The shape in Figure 5.4 compares favorably with the
shape of precipitate colonies observed on climb sources,
e.g., Figure 4.50.
Several further aspects of the above model are now con
sidered. First, the fact that the superkink configuration
is apparently a favored nucleation site is probably a func
tion of the stress field set up around it. In addition, pipe
diffusion of solute along the dislocation is most probably
slowed at these superkinks. Thus, solute builds up at the
kinks providing both a solute-rich environment for nucleation
and a continuing supply of solute for growth. Next, we con
sider why superkinks and their trailing precipitate stringers
are confined in their motion to the <1Q0> direction in the
loop habit. It may be that only one nucleation event occurs
near the kink in advance of the climbing loop. Then, as the
dislocation climbs by (or through), pipe diffusion to the kink
continues to supply solute so that a long, thin ribbon of 0'
trails behind the advancing kink. It can be seen from Figure
5.5 that the two possible {100} orientations of 0* which
nucleate on a given loop could be extended from such super
kinks only along the <100> direction contained in the loop
habit, provided the loop continues to expand macroscopically


183
(4) The particle is dragged some distance by the
dislocation before unpinning occurs.
(5) The parameters controlling colony growth are
either (a) the atomic diffusion of solute, or
(b) the core self-diffusion, whichever has the
highest activation energy.
We now compare each of the above with the evidence in
Al-Cu. ,
(1) The a+0' transformation involves a 3.95%
volume contraction.
Thus, if anything, the precipitates generate a few vacancies
instead of consuming them.
(2) The dislocation climb is essentially indepen
dent of the precipitation reaction.
Evidence for this is as follows. First, it was shown in
Section 4.2.1 that the dislocation climb is definitely vacancy-
annihilating instead of emitting. As climb occurs during
quenching, it is concluded that the driving force is the
annihilation of the quenched-in vacancy supersaturation. If
any vacancies produced by the transformation contribute to
climb, the experimental evidence suggests that this contribu
tion is small. For example, depending on quench rate and
vacancy supersaturation, the dislocations often climbed over
large distances before precipitation began, e.g., Figures
4.47 and 4.52(b). And as evidenced by Figures 4.25 and
4.44(a-d), the dislocations undergo little, if any, additional
climb during aging, while the precipitates grow orders of


101
Figure 4.28. (a) Bright-field and (b) dark-field images of
a foil oriented so that the stringers at A and
B are clearly imaged as 0' platelets viewed
edge-on. The precipitates at C are inclined
to the beam and exhibit fringes. (Heat treat
ment: S.T. 1 hour 550C, quenched to 220C,
aged 5 minutes.)


76
crystallographic planes. The resultant precipitate colony
has a corrugated shape. No precipitation occurred in areas
between the bands, e.g., at C. Such precipitate-free areas
can also be seen in the climb path of the dislocation in
Figure 4.14(b). For some reason the dislocation is ineffec
tive in nucleating precipitates in these regions of its
climb path. A possible explanation is that precipitation
occurs readily on those segments of the dislocation which
climb on the separate crystallographic planes, but not on
those curved arcs of the dislocation which bridge the gaps
between these planes. This will be discussed further in
Section 4.4.6.
The segmented climb of initial glide dislocations was
observed only in the Al-3.85 wt.% Cu alloy. An attempt was
made to determine the planes upon which segmented climb
occurred by trace analysis of the intersections of the pre
cipitate colony with the foil surface, but this proved to be
impractical for two reasons. First, as pointed out by
Miekk-oja and Raty (1971), the possible planes of easy climb
can be numerous, i.e., six {110} and four {111} planes, so
that the trace analysis is best accomplished by using single
crystals cut to special orientations. Such crystals were
not available in this research. Secondly, the traces of the
intersections of the bands of precipitates with the foil
surfaces were never well defined, a condition which leads to
poor accuracy in the stereographic analysis.


25
2.4. Repeated Nucleation on Dislocations
The concept of repeated precipitation on climbing dis
locations was first proposed by Silcock and Tunstall (1964)
to explain the occurrence of planar colonies of NbC precipi
tates on stacking faults in austenitic stainless steels.
In connection with the precipitate reaction, the stacking
faults were found to grow by the climb of a/3 Frank
partial dislocations bounding the fault. The transformation
to the NbC phase involves a 23% volume expansion, so that
the growing precipitates consume vacancies from the matrix
in order to relieve the transformation strains. Silcock and
Tunstall proposed that the Frank partial climbs by vacancy-
emission in order to feed vacancies to the transformation.
Thus the driving force for the dislocation climb is this need
to supply vacancies for the precipitate reaction. The prin
ciples of the Silcock-Tunstall model are outlined in Figure 2.8.
Repeated precipitation on climbing dislocations by this
mechanism has since been reported in a variety of systems,
including different steels, an iron-vanadium alloy, a copper-
silver alloy, superalloys, and semiconducting materials. The
phenomenon has been observed to occur on both partial and
total dislocations. For the sake of brevity, the list of
reports will not be given here, and the reader is referred to
the complete list in the recent paper by Nes (1974). In
every reported case to date, the precipitate phase has a
larger atomic volume than the matrix, thereby consuming


206
Kelly, A., and Nicholson, R. B. (1963), Progress in Materials
Science, 3, Chalmers, B. ed. Pergaiiion Press, p. 149.
Laird, C. and Aaronson, II. I. (1968), Trans. AIME, 242 1393.
Larch, F. C. (1974), chapter to be published in book by
Nabarro, F.R.N.
Lorimer, G. W. (1968) paper presented at the Fourth European
Regional Conference on Electron Microscopy, Rome, Italy.
Lorimer, G. W. (1970), Fizika, 2_, Suppl. 2, p. 33.1.
Lorimer, G. W., and Nicholson, R. B. (1969), The Mechanism
of Phase Transformations in Crystalline Solids^ Inst.
of Metals, p. $6.
Lothe, J. (1960), J. Appl. Phys., 31, 1077.
Lyubov, B. Y., and Solov'Yev, V. A. (1965), Fiz. Metal.
Metalloved. 19_, 333.
Marshall, G. W., and Brittain, J. 0. (1974), paper presented
at the 6th Annual Spring Meeting of A.I.M.E., Pittsburgh,
Pa.
Miekk-oja, H. M., and Raty, R. (1971), Phil. Mag., 24, 1197.
Nes, E. (1974), Acta Met., 22, 81.
Nicholson, R. B. (1970), Phase Transformations, American
Society for Metals, p. 269".
Nicholson, R. B., and Nutting, J. (1958), Phil. Mag. 3_, 531.
Passoja, D. E., and Ansell, G. S. (1971), Acta Met., 19, 1253.
Perry, A. J. (1966), Acta Met., 14, 305.
Preston, G. I). (1938a), Nature 142 570 .
Preston, G. D. (1938b), Proc. Royal Soc. London, A16 7, 526.
Preston, G. D. (1938c), Phil. Mag., 26 855.
Ramrez, R. G., and Pound, G. M. (1973), Met. Trans. 1563.
Reed-Hill, R. E. (1973), Physical Metallurgy Principles,
Van Nostrand, p. 170.
Ravi, K. V. (1971), Met. Trans., 3, 1311.


44
Figure
INSULATION-
ALUMINUM
BLOCK
CLAMP
2?
I
STAINLESS STEEL TUBE
1
$
1
VERTICAL
FURNACE
-SPECIMEN
ALLIGATOR CLIP
DOOR
T
3.1. Diagram of the apparatus used for solution
treatment and direct-quenching. The specimen,
attached to the stainless steel tube for weight,
is dropped from the solution treatment furnace
into the constant temperature, aging bath.


Embury and Nicholson, 1963). Since then, dislocation climb
sources have been identified In aluminum (Edington and West,
1966) and a number of other aluminum alloys, including Al-Ag
(Edington and West, 1966), Al-Ag ternary alloys (Passoja and
Ansell, 1971), and Al-Cu (Boyd and Edington, 1971). They
have also been observed in other systems, including silicon
(Ravi, 1971) and NiAl (Marshall and Brittain, 1974). Often
the dislocation climb sources observed by TEM have small
particles at the center of the source loop. An example from
the present work is shown in Figure 4.2. Although the par
ticles are usually too small to be identified, it is thought
that they are insoluble particles existing at the solution
treatment temperature. It is generally believed that vacan
cies diffusing to the particle interface activate the source
which then operates to produce successive loops. The source
itself is often thought to be a portion of a misfit disloca
tion at the particle/matrix interface.
The only reported observation of dislocation climb
sources in binary Al-Cu alloys is that of Boyd and Edington
(1971). They observed source densities of about 3/grain.
These sources generated pure-edge loops on {110} habits with
a/2<110> Burgers vectors.


115
Figure 4.35(a). Two of its eight sides lie parallel to the
[100], [010], [110] and [110] directions, respectively. For
convenience, we assume the thickness of the platelet in its
c-direction is such that it has one a[001] dislocation loop
at its edge. Now let the two segments of this a[001] loop
along [010] between AB, and along [100] between CD dissociate
according to the reactions:
a[001] -* a/2 [101] + a/2 [101] between AB ,
a[001] -> a/2 [Oil] + a/2 [Oil] between CD,
which occur without energy change. The dissociated configu
ration is shown in Figure 4.35(b). The resultant dislocations
are pure-edge segments between AB and CD. Assume further that
these segments are free to climb as pure-edge dislocations
and that the corners A, B, C and D pin them at these points.
These segments of pure-edge dislocations can now operate by
climb to produce successive loops on the appropriate {110}
planes, as in Figure 4.35(c), in the same manner as the
original model of a climb source proposed by Bardeen and
Herring (1952). In this way, a given 6' platelet could pro
duce edge-loops on four possible {110} planes.
Since the extra half plane of the original a[100] loop
was contained within the precipitate, the initial climb of
the dissociated dislocation segments must proceed through a
small volume of precipitate before entering the matrix. As
the growth of 0' platelets from the matrix generates vacan
cies within the precipitate (Section 2.2), dislocation climb
is aided by the precipitation.


ACKNOWLEDGMENTS
The author is deeply indebted to his Advisory Chairman,
Dr. John J. Hren, who contributed unselfishly of his time,
and provided advice, encouragement, and stimulating discussion
during the course of this research. He is indebted to his
advisory committee for assistance, and to Dr. R. T. DeHoff
for many helpful discussions.
Thanks are due to Dr. R. W. Gould for providing alloy
materials; to the Sandia Corporation, Albuquerque, New Mexico,
for chemical analysis; to the Japan Electron Optics Laboratory,
Boston, Massachusetts, for use of the JEOL 100B Electron
Microscope; to Mr. E. J. Jenkins for assistance in the labo
ratory; to Mr. Paul Smith for assistance in the darkroom; and
to Mrs. Elizabeth Godey for typing this manuscript.
The author*s wife, Lynn, is acknowledged for her con
stant inspiration and support. His mother is acknowledged
for her lifelong encouragement.
Finally, the financial support of the Atomic Energy Com
mission was deeply appreciated.
iv


54
c
Figure 4.2. Dislocation sources in Al-4 wt.% Cu direct-
quenched from 550C to 220C and aged for
8 seconds in (a) and (b) and one minute in (c) .


154
still further, the sample with the fastest quench rate (Ta=
180C) should contain the maximum density of active sources
with the smallest average loop size, again in agreement with
experimental observations (Figure 4.49(e)).
An additional feature of the sample quenched to 300C
was that repeated nucleation occurred for only short distances
behind the climbing dislocations and resulted in a low density
of precipitates, Figure 4.49(e). This effect can be explained
also using Figure 4.50. Given the slowest quench rate, this
sample stayed longest at temperatures above the 0* solvus,
thus causing maximum dislocation climb and the maximum deple
tion of vacancies before any precipitates nucleate. Also,
the sample remained longest at high temperatures just below
the 0' solvus. Therefore only a few precipitates nucleate
and these grow rapidly (Figure 4.49(a)). This sample was aged
only 15 seconds at 300C so that the precipitate size is still
small.
In the two samples quenched to 200 and 180C (tempera
tures below the 0" solvus, Figure 4.48), 0" did not form as
predicted from the metastable phase diagram. This is in agree
ment with observations that 0" nucleates only on previously-
formed G.P. zones (Lorimer and Nicholson, 1969; Lorimer, 1970)
4.5.4. Effect of Quench Rate
The previous section described results of changing the
quench rate by direct-quenching to various aging temperatures.
However, direct-quenching is, in general, a slow quench. The


191
It was shown in Section 4.5.5 that the spacing between
stringers (about O.ly) is the same as the spacing of visible
superkinks on climb source loops. It is assumed that the
origin of the precipitate stringers is related to the move
ment of such superkinks. According to Balluffi (1969) the
climb of a dislocation which is rotated out of its pure edge
orientation introduces kinks in the dislocation line. The
spacing of kinks depends on the magnitude of the rotation
]
and is equal to F/tan(90-a), where F is the Burgers vector
and represents the kink height, and a is the angle between
the Burgers vector and the dislocation line direction. It is
now assumed that the climb source loops are rotated slightly
off their {110} habit about an axis parallel to the <001>
direction in the habit plane. Such rotations could result
from local slip out of the climb plane due to the stress
fields of the precipitates forming in advance of the climbing
loops. A kink spacing of O.ly implies a rotation of only
0.13 for kinks of one Burgers vector in height. Such kinks
would not be resolved in the electron microscope, as are
those in Figure 4.65. Rather, it is assumed that the rota
tion is on the order of several degrees, creating numerous
kinks. It is further assumed that at the high temperatures
where climb occurs, the kinks are mobile and coalesce into
visible superkinks with an equilibrium spacing of O.ly. Then
each time a source loop expands and increases its line length
by an amount O.ly (projected normal to <001>) a new super
kink becomes stable. It is assumed that the precipitate


75
Figure 4.16. Precipitate colony exhibiting curved traces
of intersection with the foil surfaces, indi
cating a curved climb path of the nucleating
dislocation. (Heat treatment: S.T. 1 hour
550C, quenched to 220C, aged 5 minutes.)
Figure 4.17. Banded precipitate colony nucleated by "seg
mented climb" of the dislocation AB from left
to right through the foil. (Heat treatment:
S.T. 1 hour 550C, quenched to 220C, aged
5 minutes.)


193
Figure 5.4. Diagram illustrating the shape of precipitate
stringer regions which would result from
successive nucleation of superkinks of spacing
a, as a circular climb source loop expands.
The precipitate stringers are shown continuous
for clarity.


108
L
Figure
.32. Micrograph of a precipitate colony on a climb
source showing the location (arrows) of four
rows of small dislocation loops whose Burgers
vectors were to be determined. (Heat treatment:
S.T. 1 hour 550C, quenched to 220C, aged 5
minutes.)


60
Figure 4.5. Two edge-on habits of climb sources lying
slightly off 90 from each other in an (001)-
oriented foil. (Heat treatment: S.T. 1 hour
550C, quench to 220C, aged 5 minutes.)
Figure 4.6. Climb source inclined through the foil. Two
loops are sectioned leaving dislocation arcs.
The arc at A was selected for computer match
ing to determine its Burgers vector. (Heat
treatment: S.T. 1 hour 550C, quenched to
220C, aged 4 seconds.)


163
Room-temperature water quench
The structure of this sample was similar to that of the
sample quenched into room-temperature oil. Figure 4.56(a)
shows the microstructure to be full of small loops. Occa
sional glide dislocations were observed which had climbed and
nucleated 0' colonies (Figure 4.56(a) at point A). The edge-
on orientations of the small loops indicated that they lie
on {110} planes. As in the oil quenched sample, their Burgers
vectors were found to be a/2<110> normal to the loop planes.
These are prismatic edge-loops and, as before, it is assumed
that they formed by collapse of vacancy clusters onto {110}.
When viewed at high magnification (Figure 4.56(b)), the
interiors of the loops appear to contain small precipitate
colonies. It is assumed that the loops nucleated 0' in the
process of growth by climb.
Two features were observed in the microstructure of this
sample that were not present in the sample quenched into room-
temperature oil. First, in isolated areas, helical disloca
tions were observed which had partially broken up into loops,
e.g., Figure 4.57. This is in agreement with the work of
Thomas (1959) who showed that helical dislocations were
present in Al-4 wt.l Cu quenched from low temperatures in the
solid solution range. In the present case, the solution
treatment at 550C was sufficiently high that the structure
consists mainly of vacancy-condensation loops with occasional
helical dislocations.


102
shows that the stringers contain separate, but closely-
spaced, precipitates oriented parallel to the beam. Hence,
the stringers are definitely composed of 0' platelets, and
the dislocation loops are assumed to be misfit loops at the
platelet peripheries (see Section 4.4.3).
Consider now mechanism number (2) above, requiring
vacancy debris to be left behind the climbing source loops.
Boyd and Edington (1971) observed small dislocation loops
lying just inside large, climb source loops in Al-2.5 wt.% Cu.
These were analyzed to be prismatic edge-loops with all pos
sible a/2<110> Burgers vectors. They proposed that the small
loops were present from condensation of vacancy debris gener
ated by the motion of edge jogs and screw kinks on the climb
ing source loops. In the present case, such small loops
could act as nucleation sites for the precipitate stringers.
If this were so, it would be expected that all three 0' orien
tations could be nucleated since the small loops had all
possible a/2<110> Burgers vectors. It was shown earlier
that only two of the three possible 0' orientations nucleated
in any given precipitate colony. The vacancy debris mecha
nism therefore does not explain the experimental results.
The origin of the precipitate stringers will be discussed in
Chapter 5.
As mentioned previously, the stringers in precipitate
colonies on climb sources always lie along the <100> direc
tion in the plane of the colony, Figure 4.30. Typically, the
boundaries of the two stringer regions are fan-shaped. That


58
/
Figure 4.4. (a) Two edge-on habits of climb sources at 90
to each other in an (001)-oriented foil.
(b) Three edge-on habits at 60 to each other
in a (111)-oriented foil. (Heat treatments:
S.T. 1 hour 550C, quenched to 220C, aged
5 minutes.)


132
in intensity in agreement with experimental observation.
When the sheet of displacement is not flat, the fringes will
not be straight and parallel. Accordingly, the precipitate
colony in Figure 4.40(a) has some curvature. Colonies were
observed in which the fringes curved as much as 90, indicat
ing large curvature in the precipitate colony and, therefore,
in the climb path of the dislocation.
Figure 4.40(b) shows fringes behind a dislocation in a
foil which was quenched to 220C and aged for only one minute
at 220C. Here the individual precipitates in the colony are
not large enough to be visible, but their presence is indi
cated by the displacement fringes. The intensity of these
fringes is less than in those of Figure 4.40(a). This could
be due to differences in the deviation from the Bragg condi
tion or by differences in the density of precipitates in the
two colonies. A more likely explanation is that the average
precipitate thickness is less for the shorter aging time,
and therefore the resulting ^ is smaller. Such fringes in
the absence of visible precipitates were observed often in
samples aged for short times. They further support the con
clusion that the 0' colonies are nucleated entirely during
the quench.
4.4.8. Precipitate Colonies Associated
with Subboundary Formation
Dislocation subboundaries were observed in all foils
examined in this research, regardless of heat treatment. An
example of a subboundary network is shown in Figure 4.43(a).


70
observation of climb sources in Al-Cu alloys, Boyd and Eding-
ton (1971) observed a source density of about three per
grain in Al-2.5 wt.% Cu (although it is not stated, most
probably measured in the volume of a grain sectioned by the
foil; grain size not reported). Source densities many orders
of magnitude higher than this were observed in the present
foils. The maximum density was produced in a sample
quenched from 550C to 180C.. A micrograph of this foil is
shown in Figure 4.49(e). From this micrograph and the aver
age grain size (approximately 250y diameter), the active
source density was estimated to be approximately 6x10^ sources
per grain.
4.2.2. Glide Dislocations Which Climb
In addition to loops generated at climb sources, other
dislocations were observed which had climbed during quench
ing and nucleated precipitate colonies. Examples are shown
in Figure 4.14. Generally, these dislocations were long and
either smoothly curved or irregular-shaped, depending on
their climb paths. The micrographs in this section were
taken from foils aged long enough so that the precipitate
colonies are readily visible, thereby delineating the climb
paths of the dislocations. For the present, it is assumed
that the precipitate colonies were nucleated by the climbing
dislocations (this will be proven in Section 4.3). In Figure
4.14(a), the dislocation exits the foil surfaces at A and C,
and the trailing precipitate colony intersects one foil


59
Occasionally, when viewing edge-on habits in the (001)
orientation, it was observed that two different habits did
not lie exactly at 90 to one another, although the habits
of other sources in the same field of view appeared to be
perpendicular, Figure 4.5. It is concluded that climb of
the loops is not necessarily confined strictly to the {110}
planes. This angular measurement between two adjacent habits
is more accurate for determining if the loops lie exactly on
(110) planes than are measurements made from the rotation
calibration between directions in the diffraction pattern
and those in the image.
It has now been established that the dislocation sources
in these foils are climb sources. However, the climb of
pure-edge dislocations can be either vacancy-annihilating or
vacancy-emitting. The former case removes lattice planes
from the crystal whereas the latter case adds interstitial
planes. It remains to be shown whether the source loops
climb by vacancy annihilation or emission, although intuitive
arguments favor vacancy-annihilating climb. For instance, it
is known that quenching produces large vacancy supersatura
tions, but negligible concentrations of interstitials. As
the temperature drops during quenching, the need for the
excess vacancies to diffuse to sinks would promote the growth
of vacancy loops and tend to annihilate any interstitial
loops. Therefore, one would expect that the large climb
sources operate by vacancy annihilation during quenching,
but this is not a sufficient proof. In fact, in the past


106
to be 0.0 37y (370) or about one-third of the spacing between
stringers.
Precipitate stringers were also observed in 0* colonies
nucleated by glide dislocations which climbed, Figure 4.14(a).
The average spacing between these stringers was in good
agreement with the spacing of stringers at climb sources.
Stringers were not present at glide dislocations that under
went segmented climb, Figure 4.14(b).
A further feature of Figure 4.14(a) is that a well-
defined boundary exists between stringers and dense precipi
tation. Since the spacing between stringers is larger than
the spacing between precipitates in the region of dense
nucleation, it is unlikely that the latter nucleated at the
same preferred sites as the stringers. In other words, there
has been a change in the mode of nucleation at some point
near the end of the quench.
4.4.3. Determination of the Burgers Vectors of
Small Loops Within Precipitate Colonies
It was stated previously that the small loops visible
in a precipitate colony are loops at the periphery of 6'
platelets. It will now be shown that the diffraction con
trast at these loops is consistent with this hypothesis, and
that these are not loops formed by other possible mechanisms.
First, it is known that the misfit between matrix and pre
cipitate planes at the peripheral edge of 6' platelets is
accommodated by the presence of a<100> type edge-dislocation


138
Figure 444. Continued.


BIBLIOGRAPHY
Aaronson, H. I., Kinsman, K. R., and Russell, K. C. (1970),
Scripta Met. 4_, 101.
Ashby, M. F. and Brown, L. M. (1963) Phil. Mag. 8_, 1649.
Balluffi, R. W. (1969), Phys. Stat. Sol., 31, 443.
Balluffi, R. W., and Thomson, R. M. (1962), J. Appl. Phys.,
33, 817.
Bardeen, J., and Herring, C. (1952), Imperfections in Nearly
Perfect Crystals, Wiley and Sons p. 2 61.
Barnett, D. M. (1971), Scripta Met. 5_, 261.
Barnett, D. M. (1973), private communication.
Betn, R. H., and Rollason, E. C. (1957-58), J. Inst. Metals
86, 77.
Boyd, J. D. and Edington, J. W. (1971), Phil. Mag., 23, 633
Bullough, R., and Newman, R. C. (1959), Proc. Royal Soc.
London, A249, 427.
Cahn, J. W. (1957), Acta Met. 5_, 169.
Christian, J. W. (1965) The Theory of Phase Transformations
in Metals and Alloys^ Pergamon Press, p. 363.
Clarebrough, L. M. (1973), Phys. Stat. Sol, (a), 18 427.
Darwin, C. G. (1914), Phil. Mag., 27, 315, 675.
Dollins, C. C. (1970), Acta Met., 18, 1209.
Edington, J. W., and West, D. R. (1966), J. Appl. Phys., 37,
3904.
Embury, J. D. (1963), Ph.D. Dissertation, University of
Cambridge.
Embury, J. D., and Nicholson, R. B. (1963), Acta Met., 11,
347.
204


CHAPTER 3
EXPERIMENTAL PROCEDURES AND MATERIALS
3.1. Specimen Materials
The four Al-Cu alloys used in this work were obtained
as rolled sheets from a previous research project. They were
prepared from 99.99% aluminum and 99.99% copper by double
melting in an induction furnace using a graphite mold. After
solidification, the billets were alternately cold-rolled and
annealed to reduce them to sheet form.
The target compositions were the 4, 2, 1, and 1/2 wt.%
Cu alloys. The nominal copper concentrations of the four
alloys were 3.85, 1.96, 0.99, and 0.5 wt.%, based on starting
weights before melting. The impurity content in the 3.85 wt.%
Cu alloy was determined by x-ray spectrographic analysis by
the Sandia Corporation, Albuquerque, New Mexico. The impurity
levels are given as ranges in Table 3.1. The barium level is
suspect as it was determined from only one line. The impurity
levels in the other three alloys were not determined. How
ever, since all four alloys were prepared from the same
starting materials, the other three probably had the same
impurity levels as the 3.85 wt.% Cu alloy.
The 1 wt.% and 1/2 wt.% Cu alloys were obtained as
rolled sheet, 0.038 inch and 0.034 inch, respectively. They
41


Ill
and 1/3 and visible for all other products (Silcock and
Tunstall, 1964; Hirsch et_ al. 1965). All four possible
a/3 Burgers vectors can be eliminated by comparing the
data in Table 4.1 with these visibility criteria, so that the
loops are not Frank loops.
(4) All six possible a/2<110> Burgers vectors can be
eliminated by the combinations of visible and invisible
images in Table 4.1. Therefore the loops cannot be prismatic
loops on {111} with F=a/2<110>, formed from Frank sessile
loops by the passage of a/6<112> Shockley partials over the
loops.
(5) The loops cannot be pure-edge, vacancy-condensation
loops with b=a/2<110> on (110} planes (of the type observed
by Boyd and Edington (1971) lying just within source loops),
since all possible a/2<110> Burgers vectors can be ruled out.
Thus, it is concluded that the small loops visible in
precipitate colonies have Burgers vectors of the type a<100>
and lie at the periphery of the 9' platelets.
4.4.4. "Secondary" Climb Sources
A characteristic grouping of precipitate colonies
observed often is shown in Figure 4.34(a). Here a long glide
dislocation climbed through the lattice to its final position
A-A, where it is out of contrast. In climbing, this disloca
tion nucleated the large precipitate colony P in its path.
At the base of this colony, a number of {110} habits of climb
sources have been activated and nucleated their own precipitate


130
a
e~
V
/£>
/\'y
-Ri /X'V
'X
R,
<$>
%s
Figure 4.41. Diagram illustrating the origin of fringes in
a 0' precipitate colony. (a) A planar colony
containing two orientations of 0' platelets of
uniform size is inclined through the foil.
(b) The combined displacement fields of all
the platelets act as an inclined sheet of dis
placement, giving rise to the conditions
for displacement fringe contrast.


86
The missing orientation was always that {100) orientation
whose misfit would be perpendicular to the Burgers vector of
the source loop bounding the colony. This is illustrated in
Figures 4.21(b), 4.22 and 4.23.
First, Figure 4.22 shows bright and dark field images of
several precipitate colonies on climb sources in a foil whose
normal was close to [101]. The foil was oriented with the
electron beam close to [101]. The dark field image was taken
from a precipitate reflection from the (010) 8* orientation
parallel to the beam. Climb source A (and its precipitate
colony) lies on (101), and its source loops, being pure edge,
have Burgers vectors a/2[101]. At B, two other source
habits lie on (110) planes inclined to the foil. The Burgers
vectors of their source loops were not determined but they
cannot be a/2[101]. Now, if all three {100} orientations of
0' were present in the colony at A, the orientation imaged
in the dark field would be observed throughout source A as
in the sources at B. The misfit of this missing (010) orien
tation is perpendicular to the a/2[101] Burgers vector of
the source loops at A (which are "invisible" in this image).
The few precipitates on (010) in the middle of source A in
the dark field image were found to lie within small source
loops lying on other {110} habits, when this source was
examined in another orientation.
Next, in Figure 4.21(b), the source habit at B lies on
the (101) plane and is viewed edge-on. The Burgers vectors
of its source loops, being pure edge, must be a/2[01].


97
Figure 4.26. Five basic shapes of 0' precipitate colonies
nucleated on {110} habits of climb sources in
Al~Cu. All are oriented with reference to the
crystal directions at left.


62
b=a/2[01f]
b=a/2[01l]
Figure 4.7. Six experimental and computed images of dislo
cation A in Figure 4.6. The line direction is
[100], the foil normal is [313], and the beam
direction is close to [101] in (a) and (b),
to [112] in (c) and (d) and to [111] in (e)
and (f) .


45
samples were immediately up-quenched into the constant tem
perature oil bath maintained at the desired aging temperature
Due to the wide variety of solution treatment temperatures,
aging temperatures, and aging times employed in this research
no table of heat treatments will be given here. Instead, the
specific heat treatment information will be given either in
the text or in the figure captions.
3.3. Electron Microscope Specimen Preparation
Electron microscope specimens were prepared from the
heat treated strips by electropolishing in a solution of 5%
perchloric acid in methyl alcohol. A polishing potential of
18 volts d.c. was used with a stainless steel cathode. The
electropolishing setup is shown in Figure 3.2. The beaker
containing the polishing solution was immersed in a bath of
dry ice and acetone to slow down the polishing reaction. The
solution was circulated at a slow speed with a magnetic
stirrer to keep it cold. Under these conditions, the polish
ing bath was maintained at -45C.
The specimen strip was held with locking tweezers and
polished by dipping the bottom end (approximately 1/8-3/16
inch) into the solution at a dipping rate of about 1/second.
Dipping was found to reduce edge attack and to give a rela
tively uniform polish. The voltage dropped to about 12 volts
during immersion. Total polishing time to obtain a suitable


56
in this analysis with the outermost loop and the one inner
loop which lie totally within the foil. The plane of the
foil was analyzed to be very close to (101) so that this loop
habit must lie on or very close to (101). Consider first the
three images (a), (b), and (c) taken about the [101] beam
direction. In each image the source loops on (101) exhibit
weak, residual contrast typical of "invisible images of
edge dislocations for which g*F=0, but g-(Fxu)^0 (Hirsch
ejt al_. 1965 p. 261). Those loop segments lying approxi
mately parallel to the g vectors, where g-(Fxu)~0, are in
visible. The loops are everywhere visible for the two
reflections, g=T3T and 220. From this analysis, the source
loops are identified as pure edge-dislocation loops lying on
(101) with F=a/2[101]. Since the Burgers vector is normal
to the plane of the loops, the loops must expand in this
plane by the process of climb. The sources are therefore
dislocation climb sources of the type observed by Boyd and
Edington (1971) in Al-2.5 wt.% Cu.
The loop habits of climb sources in these alloys were
observed always to be {110} (with one exception to be dis
cussed in Section 4.4). The typical source produced loops
on more than one {110} habit. Loops lying on as many as
five of the six possible {110} habits were observed at one
source. The typical source also generated more than one
loop on each habit. As many as five or six loops on one
habit were commonly observed, although the average number
varied with the heat treatment.


3 99
b. a/2<110> glide dislocations on {111} planes.
c. prismatic edge-loops on {110} with
b=a/2<110>, formed by the collapse of
vacancy clusters.
The dislocation density in each category varied with heat
treatment. Categories (a) and (b) were found in all samples
regardless of quench procedure. Category (c) was found only
in samples quenched into oil or water at room temperature.
(2) Glide dislocations climb either (a) on smoothly
curved surfaces, nucleating precipitate colonies containing
two 0' orientations, or (b) on corrugated-shaped surfaces,
nucleating precipitate bands containing only one 0' orienta
tion per band.
(3) Only the two 0' orientations compatible with the
Burgers vector of the climbing dislocation are nucleated in
a given colony (with the exception of 2(b) above).
(4) Precipitate colonies generated on climb sources
exhibit a central region of uniform precipitation, and two
regions of precipitate stringers aligned in <100> directions.
The origin of the stringers is probably associated with the
movement of regularly spaced superkinks on the climbing
source loops. Only one 0' orientation is nucleated in a
given stringer.
(5) 0' platelets nucleated by repeated precipitation
can serve as source particles for generation of "secondary"
climb sources which nucleate more precipitates. In a sense,
this is a form of autocatalytic propagation of the reaction.


109
Figure 4.33.
Series o£ micrographs for determining the
Burgers vectors of the small dislocation loops
in the arrowed rows from invisibility condi
tions. The invisibility data are summarized
in Table 4.1.


TABLE OF CONTENTS
Page
ACKNOWLEDGMENTS iv
ABSTRACT ....... vii
CHAPTER
1 INTRODUCTION 1
2 REVIEW OF THEORY AND PREVIOUS WORK 5
2.1. Theory of Heterogeneous Nucleation
at Dislocations 5
2.2. Precipitation in the Al-Cu System .... 10
2.3. Dislocation Climb 17
2.3.1. Quenched-In Vacancies and
the Chemical Climb Force 17
2.3.2. Theory of Dislocation Climb ... 18
2.3.3. Dislocation Climb Sources .... 22
2.4. Repeated Nucleation on Dislocations ... 25
2.5. Pertinent Electron Microscopy Theory ... 28
2.5.1. Two-Beam Diffraction Contrast
Theory ..... 28
2.5.2. Defect Identification from
Invisibility Conditions 35
2.5.3. Imaging Precipitates in the
Electron Microscope ....... 37
3 EXPERIMENTAL PROCEDURES AND MATERIALS 41
3.1. Specimen Materials ..... 41
3.2. Heat Treatments 42
3.3. Electron Microscope Specimen
Preparation 45
3.4. Electron Microscopy 48
4 EXPERIMENTAL RESULTS AND ANALYSES 50
4.1. Introduction 50
4.2. Nature and Source of the Climbing
Dislocations 53
4.2.1. Dislocation Climb Sources .... 53
4.2.2. Glide Dislocations Which
Climb 70
v


47
thin area was about 15 minutes. When it was determined (by-
experience) that the specimen was nearly thin enough, the
stirring was stopped to avoid damaging the thin area. The
last 20-30 seconds of polishing was done by immersion and
agitation instead of dipping, since dipping to the last was
found to sometimes etch the thin foil edge. Polishing was
continued until the bottom edge appeared very ragged or until
small holes had broken through. Then the power was switched
off and the specimen was rapidly removed and plunged into a
beaker of cold acetone (99.81 pure) immersed in the dry ice-
acetone cooling bath. It was agitated for about 10 seconds
and then immediately placed under a stream of room-temperature
acetone from a wash bottle. After washing for about 30
seconds, it was allowed to dry in air. The initial wash in
cold acetone was necessary to remove most of the electrolyte
which rapidly etched the polished surface if allowed to warm
to room temperature.
The thinned, bottom portion of the sample strip was cut
off with an X-acto razor knife and mounted in a 3 mm, 75 mesh,
locking, double copper grid for viewing in the electron
microscope. Several more specimens could then be polished
from the same sample strip, if desired. However, the remain
ing bottom part of the strip was already polished quite thin.
To avoid etching this polished surface, the specimen was agi
tated for about one minute in the small beaker of cold ace
tone prior to repolishing.


Dedicated to my wife, Lynn


39
relative values of the extinction distances in the particle
and matrix, the particles can appear either lighter or
darker than the surrounding matrix. Structure factor con
trast arises only within the limits of the particle boundary.
(4) Orientation contrast. This contrast mechanism
arises whenever a foil is oriented such that a certain set
of lattice planes in the precipitate is diffracting strongly,
whereas the matrix is diffracting weakly, or vice versa.
The contrast is of a uniform light and dark nature, typically
dark precipitates in a light matrix. Orientation contrast
can arise only when there is appreciable difference in crys
tal structure between the precipitate and matrix, i.e., when
the precipitates are semicoherent or incoherent. For example,
when the electron beam is parallel to the thin dimension of
large precipitate platelets, often certain lattice planes in
the precipitate will also be parallel to the beam. In this
case, the precipitate diffracts strongly. If the matrix is
not oriented for strong Bragg diffraction, the bright-field
image will show dark precipitates in a light matrix.
(5) Displacement fringe contrast. Displacement fringe
contrast arises when there is an abrupt change in the phase
of the transmitted and diffracted waves as they encounter a
thin sheet of precipitate which displaces the matrix planes
in opposite directions on either side of it. This displace
ment R around a typical semicoherent precipitate platelet is
normal to the plane of the platelet, and its magnitude is
given by


184
magnitude in size.
(3) There is no vacancy balance required between
precipitate and dislocation in Al-Cu.
This conclusion is based on (1) and (2) above.
(4) Particle dragging was not observed in Al-Cu.
There was no visual evidence that the particles had been
dragged by the dislocations during 0' colony growth. If in
fact it did occur, the dragging distances involved must be
smaller than the resolution limit of a small particle near
a dislocation image. This is estimated to be less than 100.
(5) Colony growth in Al-Cu is controlled by the
dislocation climb rate.
It is suggested that the rate of colony growth is primarily
controlled by the rate at which the dislocation climbs through
the lattice after the sample temperature passes below the 0'
solvus. This in turn is controlled by the rate at which
excess vacancies reach the dislocation and depends, therefore,
on the self-diffusion coefficient. In addition, it is influ
enced by the degree of vacancy supersaturation below the 0'
solvus which depends on the solution treatment temperature
and the quench rate (Sections 4.5.2-4.5.4).
5.3. The Mechanism in Al-Cu
In light of the discussion above, it is clear that
repeated nucleation of 0' in Al-Cu cannot be explained by the


71
Figure 4.14. Precipitate colonies nucleated at long, climb
ing dislocations. The dislocation in (b) is
out of contrast along ABC. (Heat treatment:
S.T. 1 hour 550C, quenched to 220C, aged
5 minutes.)


3
may propagate. One mechanism is autocatalytic nucleation,
first proposed by Lorimer (1968) and similar to the way in
which martensite propagates. The initial precipitates nucle
ate on dislocations and grow into the matrix. In so doing,
they generate stress fields in the matrix which aid in the
nucleation of new precipitates. Thus, the reaction propa
gates in bands spreading out from the original dislocations
to fill the lattice. Lorimer showed that the a+0' reaction
in Al-Cu could propagate by autocatalytic nucleation. Before
the present work, this was the only reported mechanism
whereby the 0' reaction propagates from a low initial dis
location density.
Secondly, nucleation of the reaction can be propagated
if the dislocation can somehow free itself from the initial
precipitates and move away under a chemical or mechanical
stress. It then presents fresh sites for the nucleation of
more precipitates. Nicholson (1970) was the first to use
the term "repeated precipitation on dislocations" to describe
this process. Repeated nucleation on climbing dislocations
was first observed for carbide precipitation in austenitic
stainless steel (Silcock and Tunstall, 1964). Since that
time it has been reported for only a few other alloy systems.
Very recently, Nes (1974) published a paper on the mechanism
for repeated precipitation on dislocations which he implied
was universal with the statement that his model "can be
applied to repeated precipitation (or colony growth) in any
alloy system."


15
DIRECTION
OF MISFIT
9' PLATELET
a = b=4.04 c = 5.8 a
Ai-Cu MATRIX
a- b=c = 4.04
Figure 2,3. Diagram showing that the distortion of (001)
planes around a 9' platelet is normal to the
platelet (not to scale).
COMPRESSION
b= a/2 [110]
TENSION
[010]
[100]
if
[001]
MATRIX
ORIENTATION
Figure 2.4. Diagram showing that a dislocation with
Burgers vector a/2[110] relieves the misfit
around 0 platelets lying on (100) and (010).
It does not relieve the misfit around a
platelet on (001) .


125
possible to determine the different planes on which segmented
climb occurred in these foils. However, the vacancy-
annihilating climb of a/2<110> dislocation is easiest on
{ill) and illO) planes (Christian, 1965). Furthermore,
Miekk-oja and Raty (19 71) were able to show that segmented
climb occurred on the (111} and (110) planes in Cu-Ag alloys.
Their model for segmented climb on these planes will be used
here to suggest how this preferred nucleation could occur
and to explain the absence of precipitates between the bands.
Their model is shown in Figure 4.39 for segmented climb
of a dislocation with Burgers vector a/2[Oil] which was
initially a glide dislocation in its edge position on (111).
It quickly deviates from its edge position along XY, so that
some of its segments climb on the (Oil) plane and others
climb on the (111) and (111) planes, owing to the large
chemical climb forces on these planes. After a short time,
the dislocation assumes the multiply curved shape along RS
due to line tension effects which are required to mate up the
different segments between the planes. The exact curvatures
depend on the widths of the segments, the ease of climb on
the different planes, and the line tension, all of which are
unknown. If it is assumed that this model is valid for seg
mented climb in Al-Cu, it is possible that the resulting
curvatures favor nucleation of one of the two 0' orientations
on a certain plane and the other orientation on another plane.
1!" the curvature of the dislocation becomes so great near the
plane junctions as to be appreciably screw in orientation,


1 2 3
nucleation" results in different fringe effects in the bands
marked 1 and 2 in Figure 4.14(b). The fringe appearance is
a diffraction effect, to be discussed in the next section.
This preferred nucleation of only one 0' orientation per
band is shown clearly in Figure 4.38, which is a magnified
portion of Figure 4.17 on a different reflection. Only the
edge-on orientation of 0' is present in the bands marked A.
Bands B therefore must contain the other possible 0' orien
tation favored to be nucleated, which is not parallel to the
beam in this orientation. Between some bands are regions
where no precipitates were nucleated, e.g., at points C.
Depending on the diffracting vector, the misfit of the single
9' orientation in a band can cause the band to go completely
out of contrast, as is the case in Figure 4.18 between C and D,
To account for the nucleation of only one 0' orientation
in a given band, it is now suggested that a dislocation under
going segmented climb tries to assume certain line directions
on the different crystallographic planes. This is probably
due largely to differences in line tension and ease of climb
on certain planes. If the local line direction is such that
the dislocation is pure edge, nucleation of two orientations
is favored. If it is pure screw, no nucleation is favored
since there are no tensile or compressive stresses around a
screw dislocation. If it is mixed, the resulting stress
field could favor the nucleation of one or the other 0' orien
tation, depending on the line direction and plane on which
the dislocation lies. As stated in Section 4.2.2, it was not


168
climb occurred during quenching in all three alloys. Thus,
copper concentration is an important variable in the repeated
nucleation process, tending to suppress the mechanism alto
gether below some critical concentration between 1.96 and
1 wt.% Cu. The microstructures in these samples are dis
cussed separately below.
1.96 wt.% copper
Many glide dislocations were observed which had climbed
and nucleated precipitate colonies. An example is shown in
Figure 4.60(a), where the dislocation is invisible along AB.
All the precipitate stringers have coalesced into long plate
lets due to the one-hour aging treatment at 210C. Many
climb sources were observed which had nucleated precipitate
colonies during quenching, Figure 4.60(b).
1.0 wt.% copper
Again both glide dislocations and dislocation climb
sources were present in this alloy, Figure 4.61. However,
no evidence for repeated nucleation of precipitates was
observed and only a few precipitates were found near dislo
cations after aging up to 24 hours. All climb source loops
had the Class IV shape of Figure 4.26, e.g., the source sec
tioned by the foil in Figure 4.61(b).
0.5 wt.% copper
As before, both climb sources and glide dislocations
were present after direct-quenching this alloy. In addition,


4
During experiments in which Al-Cu alloys were quenched
directly to aging temperatures, this author observed that
precipitation of the 0' phase occurred exclusively by repeated
nucleation on climbing dislocations. Further examination
revealed that the repeated nucleation in this system could
not be explained by the mechanism proposed by Nes (1974).
Therefore, the primary purpose of this research was to estab
lish the mechanism for repeated nucleation of 0' on climbing
dislocations in Al-Cu, and in so doing, to determine if there
are aspects of the mechanism which might apply to precipita
tion in other alloy systems.


81
oooM
0
O
200m
oooM
0 -
I
o -
Figure 1
020,
O
M
0
220M
oooM
0
002!|
101c
I
'2
002^ 101^
0
200^
103^
I
20CW
e91 i 2ei o
103^
o
i
o
000M
0
I
I

- I -
0
!
I
o
.20. Analysis of the lower, right-hand quadrant of
the diffraction pattern in Figure 4.14(a).
(a) (001) matrix pattern; (b) precipitate
reflections from 0' platelets lying parallel
to (010) matrix planes (0^), and to (100) matrix
planes (6?) ; (c) double diffraction from matrix
200 and 220 reflections; (d) combination of
(a), (b) and (c) Compare with Figure 4.14(a).


8
bulk diffusion, so that £ increases for nucleation at a dis
location. Also, solute diffusion to dislocations themselves
enhances the local concentration (e.g., Bullough and Newman,
1959) .
The task of calculating AG(int) is difficult, which is
the main reason why the theory of heterogeneous nucleation
at dislocations is less advanced than homogeneous nucleation
theory. In fact, there have been only six published attempts
to calculate AG* for nucleation at a dislocation. Cahn (1957)
made the first calculation. He assumed an incoherent pre
cipitate, an isotropic material, and completely neglected any
interaction term. Despite these simplifications, his model
was able to predict qualitatively some experimental observa
tions of nucleation at dislocations. Dollins (1970) calcu
lated AG* for coherent, spherical and disk-shaped nuclei at
a dislocation in an isotropic matrix. His work was reexamined
by Barnett (1971). Lyubov and Solov'Yev (1965) have given
the most complete treatment for calculating AG* for a coher
ent nucleus at a dislocation. Ramirez and Pound (1973)
attempted to include effects of the dislocation core energy
on nucleation, effects that were omitted from the other
models which use linear elasticity theory. An excellent
recent review of the present status of calculating AG* for
nucleation at dislocations has been given by Larch (1974).
None of the above treatments, however, have included effects
of elastic anisotropy, principally because the calculations
involved are extremely difficult. It can be concluded that


160
The small loops are shown at high magnification in Figure
4.55. The interior of the loops are precipitated in much the
same manner as the interior of the climb source loops shown
in previous micrographs. For example, the loops and associ
ated precipitate colonies at points A have the same appearance,
but on a smaller scale, as climb source loops and their asso
ciated precipitate colonies viewed normal to their {110}
habits. The small loops encircling the precipitates at A
are invisible in this image, and in place of precipitate
stringers in the <10Q> directions, they have single larger
precipitates at the interior edges of the loops.
The small loops were found to lie on {110} planes as
illustrated by the edge-on habits at B in Figure 4.55. Their
Burgers vectors were determined from invisibility conditions
to be a/2<110> types normal to the loop planes. Thus, they
are prismatic edge-loops on {110} with E=a/2<110>. It is
assumed that they formed by collapse of vacancy clusters onto
{110} planes. Similar prismatic loops on {110} have been
observed in quenched Al-2.5 wt.% Cu (Boyd and Edington, 1971)
and in quenched Al-Mg alloys (Embury and Nicholson, 1963).
Once the loops form, they grow by climb from further vacancy
condensation. In so doing, they nucleate the small precipi
tate colonies. In fact, apart from their origin and size,
there is probably no difference in the mechanism of repeated
nucleation during the growth of these small loops and during
the growth of climb source loops.


Temperature,
148
Al
Wt.% Cu
Figure 4.48. Diagram showing the five temperatures above
the G.P. solvus to which samples were direct-
quenched from 550C and aged.


165
Figure 4.57. Helical dislocations in the sample quenched
from 550C into room temperature water, then
up-quenched to 220C and aged 5 minutes.
Figure 4.58. Loops which have partially moved off their
habit planes in the sample quenched from 550C
into room temperature water, then up-quenched
to 220C and aged 5 minutes.


dislocation can have an appreciable effect of lowering the
total free energy of the critical nucleus, whenever AG(strain)
is large.
2.2. Precipitation in the Al-Cu System
The aluminum-rich end of the Al-Cu phase diagram is a
eutectic system between the aluminum solid solution and
0-CuA19 (-53 wt.% Cu). Figure 2.1 shows the portion of the
diagram containing the a-solid solution region. Upon quench
ing from the solid solution region and aging, the equilibrium
precipitate is the b.c.t. 0-phase. The existence of three
metastable, transition phases, Guinier-Preston (G.P.) zones,
0" and 0, was established by the early x-ray work of Pres
ton (1938a,b,c) on the 4 wt.% Cu alloy, and by Guinier (1938,
1939,1942,1950,1952) on the 4 and 5 wt.% Cu alloys. The
solvus lines for these three transition phases are shown in
Figure 2.1. The positions of the 0" and 0 solvuses are due
to Hornbogen (1967). That of the G.P. solvus is due to Betn
and Rollason (1957) .
Guinier (19 38) and Preston (1938a) determined that the
G.P. zones are coherent, copper-rich clusters of plate-like
shape which form on {100} planes of the matrix. The most
reliable lattice parameters of the 0" and 0 phases are given
by Silcock ert al. (1953). 0" is complex tetragonal with
a = 4.04A and c = 7.8. It is coherent with the matrix and


162
The high density of small loops in this structure can
be explained as follows. The room temperature oil quench is
a moderately fast quench achieving a much larger supersatu
ration of vacancies than direct-quenching to the aging tem
perature. Most vacancies do not have time to diffuse to
active climb sources or to glide dislocations which are climb
ing. They therefore cluster rapidly and collapse into loops
which grow by further vacancy condensation.
If the origin of these small loops is neglected, this
microstructure can be thought of as an extension of the struc
tures described in the previous section, where direct-quench
ing into oil at progressively lower temperatures increased
the density of active climb sources and decreased their
average diameter.
From Figure 4.53, the density of small loops was esti-
13 3
mated to be 2.2x10 /cm Their average diameter was measured
from Figure 4.55 and found to be approximately 0.25y. From
these estimates, and assuming that each loop climbs by remov
ing two adjacent {110} planes to avoid a stacking fault, the
- 4
quenched-in vacancy concentration was estimated as 3x10
This value is a slight overestimate since it neglects any
vacancy contribution from the growing 0" platelets, which
have a smaller atomic volume than the matrix.
However, this estimate is in good agreement with a value
of approximately 2.5x10 ^ for the equilibrium concentration of
vacancies in pure aluminum at the 550 solution treatment tem
perature (Simmons and Balluffi, 1960 ; Guerard et al. 1974).


94
4.4. Further Geometric Analyses
The previous two sections established information about
the dislocation climb and precipitation which is basic to
understanding the details of this section. Included here are
(1) descriptions of the various geometrical aspects of the
precipitate colonies, (2) explanation of some diffraction
effects, and (3) other unusual features of the microstruc
tures, in addition to repeated precipitation, which have not
been previously reported. All these descriptions would prove
useful to someone examining these types of microstructures
for the first time.
4.4.1. Distribution of Precipitates in
Colonies at Climb Sources^
We consider the distribution of precipitates in a colony
on a single {110} habit of a climb source. A typical colony
is shown in Figure 4.25. The distribution of precipitates
is twofold. First, there is a densely-populated region of
small precipitates which covers approximately the central
three fourths of the colony. Secondly, there are two regions
marked A on either side of the colony which are comprised of
rows of somewhat larger precipitates extending from the
outer dislocation loop into the colony. These rows of pre
cipitates shall be referred to as "precipitate stringers."
The spacing between stringers is quite uniform. The dense
precipitation is not present in the stringer regions. The
stringers are always aligned in the <100> direction which is


40
| R | = At<$ n | F |
1 n1 1 n1
where At = thickness of the platelet,
6 = precipitate misfit,
n = number of misfit dislocations at the
periphery of the platelet, and
E~ = Burgers vector of the misfit dislocations.
When this displacement is substituted into the equations of
the dynamical theory, the intensity of the transmitted beam
is found to oscillate with thickness (Whelan and Hirsch,
1957). Thus, when the precipitate platelet is inclined to
the electron beam, a fringe effect is observed. The so-
called stacking fault fringes are the limiting case of dis
placement fringe contrast.


177
(3) Glide dislocations climb in one of two ways:
(a) on smoothly curved surfaces, nucleating
precipitate colonies containing two 0'
orientations, or
(b) on corrugated-shaped surfaces, nucleating
bands containing only one 0' orientation
in each band.
(4) Only the two 0' orientations compatible with the
Burgers vector of the climbing dislocation are nucleated in a
given colony. This suggests that the precipitates were nucle
ated during quenching in the wake of the climbing dislocations
In situ aging treatments in the TEM support this conclusion.
(5) Precipitate colonies generated on climb sources
contain a region of dense precipitation and regions of pre
cipitate stringers in <1Q0> directions. The stringers are
associated with a regular spacing of kinks on the climbing
source loops. Only one 0' orientation is nucleated in a
given stringer.
(6) The number density and average size of active climb
sources are functions of the quench rate and vacancy super
saturation for direct-quenches. The typical shape of a climb
source loop changes from an ellipse to a rhombus, with
decreasing copper concentration.
(7) Repeated nucleation does not occur in alloys with
copper concentration below some critical value between 1.96
and 1 wt.% copper.


95
Figure 4.25. A typical precipitate colony on a {110} habit
of a climb source exhibiting a central region
of dense precipitation, and two regions at A
of precipitate stringers aligned along the
[100] direction in the plane of the colony.


26
Figure 2.8. The Silcock-Tunstall model for repeated pre
cipitation of NbC in austenitic stainless
steel. The precipitates nucleate on Frank
partials (a). Movement of jogs, J, provides
vacancies for the precipitates to grow (b)
with the consequent climb of the dislocation
(c) The dislocation pinches off (d) and the
process repeats (e) (after Silcock and Tunstall,
1964).


116
a
b
Figure 4.35. Model for the operation of a climb source at
a 9' platelet. (a) 9' platelet on (001) with
a[100] misfit dislocation loop around its
edge. (b) Dissociation of the a[100] loop
into total edge dislocations along the
portions AB and CD.


107
loops around the platelets (Weatherly and Nicholson, 1968;
Laird and Aaronson, 1968).
Figure 4.32 shows a precipitate colony generated on a
climb source in the (101) plane of the foil. The source loop,
being pure-edge, has Burgers vector a/2[101], so that it is
invisible for this reflection. Consistent with this Burgers
vector, the precipitate colony contains 0' platelets lying on
(100) and (001) planes. If some of the platelets are large
enough, we should expect to see small loops with Burgers
vectors a[100] or a[001] at their peripheries. In Figure 4.32,
four stringers of loops are located with arrows and marked
either 1 or 2. The Burgers vectors of these loops were deter
mined by invisibility conditions from a number of two-beam
images taken for different foil orientations. Selected images
are shown in Figure 4.33, and the visibility data are summar
ized in Tab le 4.1.
Table 4.1
Summary of Visibility Data for the
Images of Figure 4.33
g=
111
020
ill
220
202
022
Source Loop
I
I
I
V
I
V
a/2[101]
Loops in
Stringers 1
V
I
V
I
V
V
a[001]
Loops in
Stringers 2
V
I
V
V
V
I
a[100]
Beam Direction
101
101
101
111
111
111
Note: V = Visible; I = Invisible.


117
Figure 4.35. Continued. (c) The a/2[Oil] dislocation
segment, pinned between C and D, operates
as a Bardeen-Herring source by climb on
the (Oil) plane (compare with Figure 2.7).


91
(a)
(b)
Figure 4.24. (a) A dislocation climb source, imaged by
residual contrast, and (b) a long glide dis
location in a sample quenched from 550C to
220C and held only 4 seconds before quench
ing to room temperature.


173
source loop changes in shape from slightly elliptical in the
[100] direction to a rhombus elongated in the [100] direction,
as shown schematically in Figure 4.64.
Close examination of the sides of the rhombus loops in
the sample aged for one hour at 210C showed that they con
tained a very regular spacing of kinks or jogs, Figure 4.65.
It is suggested that these are kinks caused by slip in the
{.111) planes normal to the loops. In fact, it was observed
often that short segments of the loops had undergone extensive
slip out of the plane of the loops, e.g., at A, B, C, and D
in Figure 4.65. The kinks were not resolvable (if present)
in climb source loops in the sample aged for only three
seconds.
The average spacing of the kinks was measured normal to
the <100> direction in the loop plane and found to be about
O.lp. This is almost exactly the same as the measured spacing
(0.096y) between precipitate stringers in the colonies on
climb sources in the 3.85 wt.% Cu alloy (Section 4.4.2).
Therefore, it is suggested that the origin of the precipitate
stringers in the alloys of higher copper concentration is
associated with a regular spacing of kinks out of the climb
plane of the source loops.
Examination of the lines of intersection of concentric
rhombus loops at a climb source with the foil surfaces
revealed that these intersections were not parallel, Figure
4.63. This indicates that the planes of successive loops are
rotated with respect to each other. In addition, it was


157

b
Figure 4.51. Precipitate colonies nucleated on (a) an
initial glide dislocation, and (b) climb
sources in a sample air-quenched from
550C, then up-quenched in oil to 220G
and aged 5 minutes.


100
Three possible mechanisms were considered for the gener
ation of stringers of small loops behind the climbing source
loops. These were:
(1) Small loops were pinched off in some regular
manner from the climbing source loops.
(2) Vacancy debris was left behind the climbing
source loops from some regular arrangement of
superjogs or kinks. This debris then coalesced
into small loops.
(3) 0' precipitates were nucleated repeatedly in
regular rows. The small loops would be misfit
loops around the periphery of the platelets.
It is shown in the next section that the Burgers vectors of
the small loops in stringers are not the same as those of the
climbing source loops. Therefore, they are not pinched-off
loops and we can omit possibility (1). For the moment, we
disregard possibility (2) and show that the stringers are
definitely composed of rows of 0' precipitates.
When foils were viewed along beam directions parallel to
(l00) planes, 0' platelets were definitely imaged edge-on in
some stringers, as at A and B in Figure 4.28(a). This is
confirmed in the dark field image, taken from a 0' reflection,
in Figure 4.28(b). Other stringers at C contain 0' platelets
on another {100} orientation which is not parallel to the
beam. In this case, the precipitates are large enough to
exhibit displacement fringe contrast (see Section 2.5.3).
Another dark field image at high magnification, Figure 4.29(b),


89
Figure 4.23. 0' precipitate colonies on climb source habits
A and B which lie parallel to the beam. Two
edge-on orientations of 0' platelets, (100)
and (010) are clearly visible in colonies A
and B. The beam direction is close to [001].
(Heat treatment: S.T. 1 hour 550C, quenched
to 220C, aged 30 minutes.)


TABLE OF CONTENTS Continued
CHAPTER Page
4 (Continued)
4.3. Identification and Characterization
of the Precipitate Phase 77
4.4. Further Geometric Analyses 94
4.4.1. Distribution of Precipitates
in Colonies at Climb Sources ... 94
4.4.2. Geometry of the Precipitate
Stringers 9 8
4.4.3. Determination of the Burgers
Vectors of Small Loops Within
Precipitate Colonies 106
4.4.4. "Secondary" Climb Sources .... Ill
4.4.5. A Climb Source on (100) ..... 118
4.4.6. Nucleation of Preferred 6'
Orientations During Segmented
Climb 122
4.4.7. Displacement Fringe Contrast
in a Precipitate Colony 127
4.4.8. Precipitate Colonies Associated
with Subboundary Formation .... 132
4.5. Effects of Experimental Variables on
Microstructure 135
4.5.1. Effect of Time at Constant
Aging Temperature ........ 135
4.5.2. Effect of Solution Treatment
Temperature 140
4.5.3. Effect of Temperature to Which
Samples Are Direct-Quenched . 147
4.5.4. Effect of Quench Rate 154
4.5.5. Effect of Copper Concentration . 166
4.6. Summary 176
5 THE REPEATED NUCLEATION MECHANISM ....... 178
5.1. Nucleation of 0' Near Edge Dislocations 178
5.2. Comparison with Previous Repeated
Nucleation Mechanisms 182
5.3. The Mechanism in Al-Cu 184
5.3.1. Local Solute Buildup 186
5.3.2. Precipitate Stringer Formation . 190
5.4. Criteria for Repeated Nucleation in
Al-Cu and Application to Other Systems . 196
6 CONCLUSIONS AND SUGGESTIONS FOR FUTURE WORK . 198
BIBLIOGRAPHY 204
BIOGRAPHICAL SKETCH .................. 208
vi


Temperature,
16 7
Figure 4.59. Diagram showing the solution treatment and
aging temperatures used for direct-quenching
the 2, 1, and 1/2 wt.% copper alloys.


155
purpose of the following experiments was to determine if
repeated nucleation could be eliminated, or suppressed, by
faster quenching. The quench rate was changed by varying
the quenching medium.
Samples were solution treated for one hour and quenched
int o:
(1) air,
(2) liquid nitrogen,
(3) oil at room temperature, and
(4) water at room temperature.
The quench rates were not measured, but it is believed they
increased in the order given above. After quenching, the
samples were up-quenched into oil at 220C and aged for five
minutes in order to grow any 0' precipitates present to
visible sizes. The air-cooled sample was held for approxi
mately six seconds to insure complete cooling before up-
quenching to 220C. All other samples were up-quenched with
in about two seconds after down-quenching.
It was found that repeated nucleation of 0' had occurred
on climbing dislocations during all quenches listed above.
However, dislocation configurations varied widely as a result
of the different quench rates so that each treatment will be
discussed separately.
Air quench
Air quenching generated approximately the same density
of glide dislocations as did direct-quenching into oil. All


96
contained in the {110} plane of the colony. The dislocation
loop bounding the colony bows out locally between the stringers
and is generally smoothly curved in those regions away from
the stringers. The nature of the stringers will be discussed
in detail in the next section.
There were five basic shapes of the precipitate colonies
on climb source habits observed in these foils. These five
shapes are shown schematically, with corresponding micro
graphs, in Figure 4.26. Shapes I and II were the most com
monly observed. The typical colony with shape I was either
circular or slightly elliptical in the [001] direction, and
the region of dense precipitation was essentially continuous
over the center of the colony. Shape II is similar to Shape
I, except that there is a figure-eight-shaped region in the
center which is void of precipitates. The experimental
observations suggest that as the diameter of a colony with
Shape I increases, it will tend toward Shape II. At present,
there is no satisfactory explanation for the figure-eight -
shaped region void of precipitates. Shape III is very elon
gated in the [001] direction, and the regions of stringers
extend out beyond the projected sides of the ellipse. Such
colonies were observed only occasionally. In Shape IV, the
dislocation loop bounding the colony bows out along [110] at
two points. Narrow precipitate-free zones extend in this
direction from the center of the colony to these points.
The discussion in Section 4.1.5 may account for the tendency
to bow out to points, but there is no satisfactory


200
(6) The number density and the average size of active
climb sources are functions of the quench rate and vacancy
supersaturation for direct-quenches The typical shape of a
source loop changes from an ellipse to a rhombus with decreas
ing copper concentration.
(7) Repeated nucleation of 0' does not occur in alloys
with copper concentration below some value between 1.96 and
1.0 wt.%,
The morphologies resulting from repeated nucleation con
tained many fascinating features which had not been previously
reported, and it is certain that further investigation of
these microstructures would reveal even more details not dis
closed by this initial work.
The mechanism of repeated nucleation of 0' in Al-Cu is
different from other previously reported mechanisms. The
vacancy-annihilating dislocation climb appears to be inde
pendent of the precipitation process. Repeated nucleation
is possible simply because the climb rate is slow enough to
permit nucleation in advance of the moving dislocation, but
rapid enough to permit the dislocation to pass before parti
cle growth can result in effective pinning. The fact that
the phenomenon occurs during quenching assures that there is
a continuous supply of non-equilibrium vacancies to move the
dislocation along and present fresh sites for further nucle
ation. The criteria for repeated nucleation by this mechanism
are :
(1) a phase which nucleates easily on dislocations,


CHAPTER 4
EXPERIMENTAL RESULTS AND ANALYSES
4.1. I introduction
Evidence for repeated nucleation of the 0 phase on dis
locations was first observed in this research when the
Al-3.85 wt.% Cu alloy was quenched directly to aging temper
atures above the 6" solvus. Figure 4.1 shows a typical
microstructure resulting from direct-quenching and aging.
A brief description of the features and evolution of this
microstructure (with the facts to be established in this
chapter) is as follows. Dispersed throughout the foil are
densely populated colonies of small 6' precipitates. The
colonies are bounded either totally or partially by disloca
tions, some of which are out of contrast in this image. The
dislocations were generated and climbed during the quench
from the solution treatment temperature. As they climbed,
they nucleated and dispersed the 0' colonies in their paths.
All dislocations climbed during quenching and all nucleated
precipitate colonies. The precipitate colonies may (1) be
planar, (2) lie on smoothly curved surfaces, or (3) lie on
corrugated-shaped surfaces, depending on the climb paths of
the dislocations.
50


53
4.2. Nature and Source of the Climbing Dislocations
The dislocations which climbed during quenching can be
classified mainly into one of two categories according to
origin: (1) those generated at dislocation climb sources,
and (2) glide dislocations which subsequently climb (a third
category found in alloys quenched into oil or water at room
temperature will be discussed in Section 4.5.4).
4.2.1. Dislocation Climb Sources
Figure 4.2 shows micrographs from foils direct-quenched
to 220C and aged for short times before quenching to room
temperature. Present in the microstructures are configura
tions consisting of concentric dislocation loops. When
viewed edge-on, the loops are seen to be coplanar since
their traces are straight lines, as at points A in Figure
4.2(c). Tilting the foil confirms that these straight lines
are traces of coplanar loops. Concentric loops sectioned by
the thin foil leave straight-line traces with the foil sur
faces, e.g., at B-B in Figure 4.2(b) and (c). Often small
particles were observed at the center of the loops as in
Figure 4.2(b). The operation of dislocation sources has been
discussed in Section 2.3.3 and will not be repeated here.
It will now be established that these are climb sources, and
the Burgers vectors and habit planes of the loops will be
identified.
A typical source is shown in Figure 4.3. Several loop
habits have been generated at the source. We are concerned


120
Figure 4.37. Series o£ micrographs for determining the
Burgers vector of the dislocation arc bound
ing the precipitate colony A in Figure 4.36,
and the orientation of 0' precipitates in
the colony. The invisibility data are summar
ized in Table 4.2.


Q'jA
82
Figure 4.21. (a) and (b) Appearance of typical precipitate
colonies generated on the {110} habits of
climb sources. (c) and (d) Schematic diagrams
illustrating the geometry of the colonies.
See text for description. (Heat treatment:
S.T. 1 hour 550C, quenched to 220C, aged
5 minutes.)


135
4.5. Effects of Experimental Variables
on Microstructure
Up to this point, evidence has been presented to charac
terize the nature of repeated precipitation at climbing dis
locations. What remains is to determine the mechanisms by
which the nucleation events take place. Information towards
this objective was obtained by varying independently the
following experimental parameters:
(1) time at constant aging temperature after
direct-quenching,
(2) solution treatment temperature,
(3) temperature to which samples are direct-
quenched,
(4) quench rate, and
(5) solute concentration.
In addition, a clearer picture of the operation of climb
sources in Al-Cu alloys during quenching was obtained as well.
4.5.1. Effect of Time at Constant Aging Temperature
In this experiment, samples were solution-treated for
one hour at 550C, direct-quenched in oil at 220C, and aged
for various times. The resulting microstructures after aging
for 8 seconds, 1 minute, 5 minutes, 30 minutes, and 2 hours,
respectively, are shown in Figure 4.44 at low magnification.
After 8 seconds at 220C, the only visible precipitates are
those in the stringers along the [OlO] direction. After
aging one minute, other precipitates are just visible in the


31
INCIDENT BEAM
Figure 2.10. Diagram illustrating how contrast arises in
the transmitted and diffracted beams from
diffraction off the distorted planes around
an edge dislocation. Planes to the left of
the dislocation are tilted toward the Bragg
angle. Planes to the right are tilted away
from the Bragg angle.


90
e.g., during aging. Since the dislocation climb sources are
known to have operated during the high-temperature part of
the quench, the precipitates must have nucleated during
quenching.
Further evidence to support this conclusion was obtained
by in situ aging experiments in the electron microscope. The
results of these experiments are shown in the micrographs of
Figure 4.24. These are images of a foil from a sample solu
tion treated for one hour at 550C, quenched into oil at
220C and held only four seconds at 220C, then water-quenched
to room temperature. After electropolishing, the foil was
placed in the heating stage of the microscope. The micro
graphs in Figure 4.24(a) and (b) were taken prior to heating.
In (a), a dislocation climb source is viewed normal to its
(Oil) habit of source loops which are "invisible" in this
orientation and imaged by residual contrast. Since this is
a climb source, we know that these dislocations climbed
during quenching. A long, crooked dislocation, which was
most probably a glide dislocation prior to quenching, is
shown in (b). Its crooked shape is the only indication that
it may have climbed during quenching. Now, if nucleation
of the 0' precipitates does occur as the dislocations climb
through the lattice during quenching, then the precipitates
must already be present in the foil in (a) and (b). However,
the four-second aging time at 220C was insufficient to cause
the precipitates to grow to visible sizes. The foil was then
heated to 230C in the microscope. After nine minutes at


112
Figure 4.34. (a) Secondary climb sources B, C, and D, gener
ated in the precipitate colony P that was
nucleated by climb of the long glide disloca
tion which is out of contrast between AA.
(b) Secondary climb sources A and B generated
in the precipitate colonies at primary climb
sources (viewed edge-on in this orientation).


33
derivation is similar to the derivation of a two-beam theory
for x-ray diffraction by Darwin (1914). The Howie-Whelan
theory predicts the amplitudes T and S of the electron waves
in the transmitted and scattered beams, respectively, at
any point in the crystal. The formulation of the theory
uses a column approximation, whereby the crystal is imagined
to be divided up into parallel columns in the direction of
the incident beam. Dynamic interchange between T and S is
considered within a column, but not between neighboring
columns. This is a valid approximation because the Bragg
angles for high energy electron diffraction are small (^1/2).
The Howie-Whelan theory predicts the following coupled
pair of first-order differential equations for the variations
in T and S with depth in the crystal:
37 = (iri/£0)T + C tt i / fg) S exp(2risz + 2rig-R)
37 = ( tt i / EQ) S + (iri/ )T exp (-27risz-2TTig* R)
where
T = -amplitude of the incident beam,
S = amplitude of the diffracted beam,
z = depth in the crystal in the direction
of the incident beam,
s = parameter measuring deviation from the
Bragg condition,
g = the diffracting vector,
R = the local displacement field at depth z,
E = parameter related to mean refractive index
of the crystal, and
E = the extinction distance,
g


124
Figure 4.38. Micrograph of a precipitate colony at high
magnification illustrating that only one 9*
orientation is present in different bands
A and B which result from segmented climb of
a glide dislocation. No precipitates were
nucleated in the regions C between bands.
(Heat treatment: S.T. 1 hour 550C, quenched
to 220C, aged 5 minutes.)


188
that for bulk diffusion. This provides an advantage for
nucleation at the dislocation in addition to that gained from
the strain field. To test if this model could provide suffi
cient solute for the observed repeated nucleation, a calcula
tion was made to determine if there was enough copper over
the climb surface to account for the density of precipitates
generated in a colony. The density of precipitates per unit
area, measured from Figure 5.3, was found to be approximately
10 2 '
6.4x10 /(cm) It is assumed that the density of visible
precipitates at this stage is the same as the density of
nuclei which form initially. As discussed in Section 4.2.1,
an a/2[110] dislocation climbing by vacancy annihilation on a
{110} plane in f.c.c. removes two adjacent planes of atoms.
Initially, it was assumed that the dislocation could collect
easily all the solute on these two planes and each plane on
either side of it, or all copper atoms on a total of four
adjacent {110} planes. Assuming 2 atomic! copper (4 wt.! Cu)
on these {110} planes, the dislocation collects 5.8x10^
copper atoms in climbing over a square micron area. If by
pipe diffusion, the dislocation can distribute this amount
of solute to the 640 embryos per square micron, then on the
average there would be 900 copper atoms available to each
embryo. There are two copper atoms per unit cell of 0' so
this amounts to enough solute for 450 unit cells/embryo. An
estimate of the critical nucleus size must now be made to
determine if this is enough solute to form stable nuclei.
This was done on the basis of a plate-like nucleus on (100),


SOURCE
CLIMB
PLANE
/
i
"1
a
o
b
\
/
n/->/x|n//
\/\ \|7\\
I*
/
c
Figure 5.1. (a) Schematic cross-section through a climb
source particle and one loop showing regions
of compression (dashed) in advance of the
loop. (b) Nucleation in the compressive
regions. (c) After further climb and gener
ation of another source loop.


CHAPTER 6
CONCLUSIONS AND SUGGESTIONS FOR FUTURE WORK
This research has shown that the a+0' transformation in
Al-Cu can initiate and propagate during quenching by the
mechanism of repeated nucleation on climbing dislocations.
The dislocations are generated and climb during quenching by
annihilation of the quenched-in vacancies. Densely populated
colonies of 0' precipitates nucleate in the passing stress
fields of the climbing dislocations. The distribution of
the entire volume fraction of 6' is thus determined during
the quenching step (£10 1 seconds). Therefore, autocatalytic
nucleation is not required to propagate the reaction during
aging.
The goals of this research were twofold: (1) to charac
terize the microstructures resulting from repeated nucleation,
and (2) to establish the repeated nucleation mechanism in
this system. Pertaining to microstructure, the following
conclusions were drawn:
(1) The dislocations which climb during quenching fall
into three categories according to origin:
a. pure-edge loops on {110} planes with
a/2<110> Burgers vectors, generated at
dislocation climb sources.
198


203
(g) Finally, the criteria for repeated nucleation in
A'l-Cu are relatively simple and probably exist in a number
of other systems. it. may be that the effect has not been
observed in some alloys that have phases which nucleate on
dislocations, simply because slow quenches are not normally
employed in laboratory investigations. Commercial quenching
practice is often a different situation though, where thick-
section parts almost certainly receive relatively slow
quenches. It Is suggested that other aluminum alloys, which
meet the criteria above, should be investigated to determine
if the mechanism lias broader application than the Al-Cu
system. Al-Mg or Al-Mg ternary alloys should receive first
consideration.


must be nucleated on G.P. zones (Lorimer and Nicholson, 1969;
Lorimer, 1970). If a sample is direct-quenched into the
region below the 0" solvus but above the G.P. solvus (Figure
2.1), no 6" forms. Howevei, if a sample is direct-quenched
below the G.P. solvus and then up-quenched into this region
and aged, 0" forms and its distribution is a function of the
size distribution of G.P. zones present before the sample
was up-quenched (Lorimer, 1970). Neither G.P. zones nor 0"
plays a role in the nucleation of 0'. Due to its misfit
strain, O' nucleates only heterogeneously in the presence of
a stress field in the lattice. It nucleates either at dislo
cations (Nicholson and Nutting, 1958), in the stress fields
of other 0' precipitates (Lorimer, 3968), or in the presence
of a macroscopic stress applied to the sample during aging
(Mosford and Agrawal, 19 74).
Numerous TEM investigations have confirmed the catalytic
effect of dislocations for nucleating 0'. It was suggested
early (Wilsdorf and Kuhlmarn-Wilsdorf, 1955; Thomas and
Nutting, 1956), and later confirmed by TEM, that only certain
0' orientations will nucleate at a given dislocation. This
is explained in terms of the misfit strain of the 0' platelet
and the Burgers vector of the dislocation. In Figure 2.3, it
is shown that the principal misfit around a 0' platelet is
normal to the {100} plane of the platelet. In Figure 2.4, it
is shown that a dislocation with Burgers vector a/2[110]
partially relieves the misfit strain around 0' platelets on
(100) and (010) whose misfits lie at 45 to the Burgers vector,


207
Seidman, D. N. and Balluffi, R. W. (1968), Lattice Defects
and Their Interactions, Hasiguti, R. R., ed. Gordon
and Breach, p. $13.
Servi, I. S., and Turnbull, D. (1966), Acta Met., 14, 161.
Silcock, J. M. Heal, T. J., and Hardy, H. K. (1953-54),
J. Inst. Metals, 82, 239.
Silcock,
361.
J. M. ,
and
Tunstall, W. J.
(1964) Phil.
Mag., 10 ,
Simmons,
52,
R. 0. ,
117.
and
Balluffi, R. W.
(1960) Phys.
Review,
Thomas, G. (1959), Phil, Mag., 4, 1213.
Thomas, G. and Nutting, J. (1956) The Mechanism of Phase
Transformations in Metals, Institute of Metals, p. ST.
Thomas, G. and Whelan, M. J. (1961), Phil. Mag. 6_, 1103.
Thomson, R. M., and Balluffi, R. W. (1962), J. Appl. Phys.,
33, 803.
Warren, J. B. (1974), Ph.D. research in progress.
Weatherly, G. C., and Nicholson, R. B. (1968), Phil. Mag.,
17, 801.
Westmacott, K. H., Hull, D., and Barnes, R. S. (1959),
Phil. Mag. 4_, 10 89.
Westmacott, K. H., Barnes, R. S., and Smallman, R. E. (1962)
Phil. Mag., 7, 1585.
Whelan, M. J. and Hirsch, P. B. (1957), Phil. Mag. Z_, 1121
Wilsdorf, H. and Kuhlmann-Wilsdorf, D. (1955) Defects in
Crystalline Solids, Physical Society, p. 175.


28
Nes implied that this mechanism is applicable to repeated
precipitation on climbing dislocations in all systems, where
as in reality, it probably applies only when there is required
a mass balance of vacancies between growing precipitates and
climbing dislocations.
There is one report of repeated precipitation on climb
ing dislocations which has not been attributed to the above
mechanism. Embury (1963) observed that dislocations in Al-Mg
alloys were drawn around l^Al^ precipitate particles, pinched
off leaving loops, and climbed away under the chemical force
of a quenched-in vacancy supersaturation where the process
repeated. However, this process is reported to occur only to
a small extent.
2.5. Pertinent Electron Microscopy Theory
2.5.1. Two-Beam Diffraction Contrast Theory
In the transmission electron microscope, contrast arises
from differences in intensity scattered out of the incident
electron beam by Bragg diffraction from the crystal planes.
The best contrast from defects occurs under so-called "two-
beam" conditions. Owing to the large amount of tilt avail
able in commercial goniometer stages, the crystal can be
oriented so that the incident beam diffracts strongly only
from one set of lattice planes. Then approximately 95% or
more of the incident intensity is contained either in the


175
Figure 4.65. High magnification of climb source loops in
Al-0.5 wt.% Cu quenched from 545C to 210C
and aged 1 hour, revealing a regular spacing
of kinks along the dislocations.


114
agreement with the conclusion that the latter are nucleated
earlier in the quench and consume more vacancies.
Now, if secondary climb sources are indeed nucleated by
0' platelets, they could nucleate in precipitate colonies on
primary climb sources as well. Such configurations were also
observed, as shown in Figure 4.34(b). In this micrograph,
the beam is approximately parallel to [001] so that climb
source habits on (110) and (110) planes are viewed edge-on
as at A and B. The smaller source habits at A and B were
obviously nucleated at different sites along the two larger
habits. Therefore, they all could not have been nucleated at
the original source particle. As they appear to have nucle
ated in the planes of the larger habits, they were most
probably nucleated at 0* precipitates in these planes.
A model will now be presented to illustrate how a 0'
platelet could act as a source particle for the nucleation of
climb source loops on {110} planes. As discussed in the pre
vious section, the misfit between matrix and precipitate
planes at the peripheral edge of 0' platelets is accommodated
by the presence of a<001> edge-dislocation loops around the
platelets. The extra half-plane must be contained in the
precipitate. Furthermore, Laird and Aaronson (1968) have
shown that 0' platelets of appreciable size are often octag
onal-shaped with their edges lying along the <100> and <110>
directions within the plane of the platelet. Consider then
such an octagonal-shaped platelet lying parallel to the (001)
matrix planes with its c-axis parallel to [001] matrix,


36
distorted. Then when the crystal is oriented so that one
such set of planes is in the reflecting position, the dislo
cation will be "invisible" in the image. The diffracting
vector is perpendicular to the diffracting planes, and there
fore to the Burgers vector for this condition. Hence, the
criterion for invisibility of a dislocation is the well-
known relation g*F=0. To identify the Burgers vector of a
dislocation, it is simply a matter of tilting the foil and
selecting various two-beam conditions until two diffraction
vectors g and g^, are found for which the dislocation is
invisible in the bright-field image. The Burgers vector
must be perpendicular to both g^ and so that it can be
determined from their cross product, i.e., 5"= (g^ ). How
ever, this technique is not capable of determining the
Burgers vector unambiguously, i.e., whether it is +F or -F.
Furthermore, the criterion g-F=0 for invisibility applies
only to screw dislocations where, in the isotropic approxi
mation, all sets of planes parallel to the Burgers vector
are undistorted. This is not so for an edge dislocation.
For a set of planes to remain undistorted by an edge dislo
cation, not only must g*F=0, but in addition, g must be
parallel to the dislocation line direction. Mathematically
this is written g-(Fxu)=0, where u is the line direction.
This is a very stringent condition which is seldom obtained
in the microscope. Thus, edge dislocations, or dislocations
with appreciable edge orientation, often exhibit strong
"residual contrast" when g*F=0, due to the (Fxu) term. For


27
vacancies during the transformation. In every case the dis
location climb has been reported to be vacancy-emitting in
order to supply the necessary vacancies.
Nes (1974) has expanded the original model of Silcock
and Tunstall (1964) into a more quantitative theory, which
was intended to account for the various features of repeated
precipitation in all the systems reported since 1964. The
fundamentals of the Nes theory are:
(1) Vacancies must be supplied to the transforming
particles in order to reduce the particle/matrix
mismatch.
(2) The subsequent particle growth causes vacancy-
emitting climb of the dislocation in order to
. feed the transformation.
(3) The particle growth/dislocation climb sequence
between conservative nucleations is controlled
by balancing the rate at which vacancies must
be supplied to the precipitates with the climb
rate of the dislocation.
(4) The particle is dragged some distance by the
dislocation before unpinning occurs.
(5) The rate controlling parameters in the kinetics
of colony growth are either (a) the atomic dif
fusion of the precipitating atoms, or (b) the
core (interface) self-diffusion, depending on
which has the highest activation energy.


51
Figure 4.1. Typical microstructure resulting from quench
ing the Al-4 wt.% Cu alloy directly to the
aging temperature. The foil contains colonies
of small 0' precipitates, bounded by disloca
tions. (Heat treatment: S.T. 1 hour, quench
to 220C, age 5 minutes.)


49
Figure 3.3.
Kikuchi line map over two adjacent stereo
graphic triangles for a face centered cubic
crystal (after Head e_t al_. 1973).


72
surface along the trace ABC. In Figure 4.14(b), the dislo
cation lies along ABC at the upper edge of the precipitate
colony, where it is "invisible." Some residual contrast can
be seen, however. The source of these dislocations is un
known. They could be (1) grown-in dislocations, (2) glide
dislocations which existed at the solution treatment tempera
ture prior to quenching, or (3) glide dislocations which were
generated at some source, probably grain boundaries, at the
onset of quenching. It is thought that most, if not all,
fall into categories (2) and (3), i.e., they were a/2<110>
glide dislocations on (ill) planes prior to climbing.
Figure 4.15, for example, shows two images of a precipitate
colony nucleated by one such long dislocation. In (a) the
precipitate colony is inclined through the foil. The curved
dislocation has been sectioned twice by the foil, leaving
two arcs, AB and CD, at the ends of the precipitate colony.
The Burgers vector of this dislocation was determined to be
a/2 [101]. The micrograph in (b) was taken after the foil
was tilted to the (111) orientation. Here the precipitate
colony is viewed edge-on and appears as a curved, dark line.
This indicates that the dislocation climbed on an irrational,
smoothly curved surface normal to the (111) plane. Since
(111) is a glide plane, it is reasonable to assume that it
was a curved, glide dislocation on (111) prior to climbing.
Such long glide dislocations were observed often to
have climbed on smoothly curved surfaces. This was easily
recognized by the curved intersections which the associated


119
Figure 4
36.
Two views of a precipitate colony A which is
seen to lie on (100) from its edge-on orien
tation in (a). The beam direction is close
to [001] in (a) and to [101] in (b).
(Heat treatment: S.T. 1 hour 550C, quenched
to 220C, aged 5 minutes.)


127
then no 0' would nucleate, accounting for the precipitate-
free regions between bands in Figure 4.38.
It is interesting to note that Miekk-oja and Raty (1971)
observed precipitation during segmented climb but did not
report any preferred orientations of the precipitates asso
ciated with the climb. In Al-Cu, segmented climb always
resulted in this preferred nucleation.
4.4.7. Displacement Fringe Contrast
in a Precipitate Colony
Often, fringe effects were observed in the precipitate
colonies, Figures 4.38 and 4.40(a). The intensity of the
fringes varied with the deviation from the Bragg condition,
becoming stronger as the deviation tended toward zero. Some
times the fringe intensity varied with position over a given
colony, appearing to be a function of the local precipitate
density. The intensity of fringes increased in areas where
the local precipitate density increased. The colony in
Figure 4.40(a) has a very dense, but uniform distribution of
precipitates so that the fringe intensity is large but uni
form.
Similarly, Miekk-oja and Raty (1971) reported fringe
effects in precipitate colonies nucleated behind climbing
dislocations in Cu-Ag alloys. However, they did not explain
the contrast mechanism, but described the effect simply as
image contrast variations with foil depth depending strongly
on diffraction conditions. It will now be shown that the
fringe effect can be explained simply on the basis of


34
Each separate equation represents the variation in amplitude
of the electron waves in that beam as it passes through the
crystal. The first term in each equation represents the
intensity scattered in the forward direction for that beam.
The second term in each equation represents the intensity
scattered into that beam from the other beam.
In order to account for experimentally observed effects
of absorption, it is necessary to replace the quantities
1/£q and l/£ in the equations by the complex quantities
(1/Cq+/5q) and (l/£ +i/£g), respectively. One then obtains:
^ = Tri (l/C0 + i/C)T + ui (l/£ +i/ps exp(2TTsz + 27Tg* R)
Jj! = iri(l/£0+i/5)S + 1T (l/Cg+i/SpT exp (-27risz-2Trig-R)
Multiplying the amplitudes T and S by their complex conjugates
gives the relative intensities in the two beams at any point
in the crystal. In particular, when the intensity is calcu
lated at the bottom of all imaginary columns in the foil, it
predicts the image projected onto the viewing screen, since
no interaction occurs in vacuum once the beams exit the crystal
In the absence of any displacement field (R=0), or in the
presence of a fixed, rigid body displacement (R=constant), the
equations can be solved analytically for T or S, and the solu
tion predicts a uniform intensity over the bottom of the
crystal. When the displacement field R varies with depth, as
is the case around dislocations and other defects, the equa
tions can no longer be solved analytically, and numerical
methods must be used to obtain T and S.


174
4wt./oCu
0.5wt./oCu
Figure 4.64. Diagram showing the change in shape of the
typical climb source in Al-Cu with decreasing
copper concentration.


46
5 l o
ACID
(+)
The electropolishing setup for preparing thin
foils for electron microscopy. Polishing is
accomplished by dipping the bottom end of the
specimen into the solution.
Figure 3.2.


CHAPTER 2
REVIEW OF THEORY AND PREVIOUS WORK
2.1. Theory of Heterogeneous Nucleation
at Dislocations
Nucleation theory employs the concepts of a critical
nucleus and an activation energy for nucleation. An assump
tion of the theory is that random thermal fluctuations lead
to the formation of small embryos of the transformed phase.
Embryos having a size and shape smaller than some critical
size and shape will on the average disappear, and those with
a larger size will grow and become stable. This critical
size and shape is defined as the critical nucleus. The acti
vation energy is the minimum energy barrier which must be
overcome before nucleation can occur and turns out to be the
free energy of the critical nucleus. This energy barrier is
a maximum with respect to size and a minimum with respect to
all other variables. The importance of the free energy of
the critical nucleus can be recognized from its appearance
in the equation for the steady state nucleation rate, which
is written in general form as
J
ZB
/ exp
s
[
-AG*,
~TTJ
(2.1)
5


78
Figure 4.18. A glide dislocation which climbed on a
smoothly-curved surface between A and B,
and underwent segmented climb between B
and C. (Heat treatment: S.T. 1 hour
515C, quenched to 220C, aged 5 minutes.)


185
mechanism of Nes (1974). It does not involve a vacancy flux
problem between growing precipitate and climbing dislocation.
Rather it appears to be a basic heterogeneous nucleation
problem, requiring that the dislocation stress field be pres
ent long enough for nucleation to occur. The mechanism of
repeated nucleation in Al-Cu apparently involves three essen
tial factors:
(1) Nucleation occurs because the 1 dislocation
!
stress field is present to help overcome the
energy barrier.
(2) Repeated nucleation is possible because the
dislocation climbs under a driving force
independent of the precipitation process,
namely the quenched-in vacancy supersaturation.
This assures that the dislocation advances to
act as a catalyst at successive positions
along the climb surface.
(3) The climb rate is apparently slow enough to
allow nucleation, but rapid enough to move
the dislocation past the newly-formed precipi
tates before they grow large enough to pin it.
Factor (3) could be a self-regulating effect in that the
initial precipitate size may be limited by the strain field
of the dislocation. Growth after nucleation relies princi
pally on long-range bulk-diffusion of solute and may not be
fast enough to pin the climbing dislocation.


5 2
The primary goal of this research was to determine the
mechanism by which repeated nucleation of O occurs in the
Al-Cu system. However, since this work is the first reported
observation of the phenomenon in Al-Cu alloys, a secondary
goal was to characterize thoroughly the various features of
the microstructures observed. The geometrical analyses are
reported in detail here.
The material in this chapter is developed much in the
way in which the experimental analysis was performed. First,
in Section 4.2, the nature and sources of the climbing dis
locations are established. Next, in Section 4.3 the precipi
tate is identified as 0' and characterized as to distribution
of orientations in the colonies. Section 4.4 contains
descriptions of the various geometries and some of the dif
fraction effects. Finally, Section 4.5 describes results of
experiments designed to determine the effects of different
parameters on the repeated nucleation process. Most of the
analyses for identification purposes were conducted on the
3.85 wt.% Cu alloy. Accordingly, all micrographs in this
chapter are from this alloy, except those in Section 4.5.5.
In addition, most micrographs in this chapter are from sam
ples quenched directly to aging temperatures. For the sake
of brevity, the copper concentration of the alloys is listed
In the figure captions as either 4, 2, 1, or 1/2 wt.%.


137
Figure
4,44. Continued


129
displacement fringe contrast (Hirsch et_ al. 1965 p. 341).
In Section 2.5.3, it was shown that displacement fringe con
trast arises from abrupt phase changes in the incident and
diffracted beams encountering an inclined sheet of precipitate
which has displaced the matrix by an amount K in opposite
directions across the sheet. Figure 4.41 illustrates that a
planar, precipitate colony, composed of two orientations of
6' platelets of uniform size, has the resultant effect of a
sheet of displacement field which displaces the matrix normal
to the colony by a resultant displacement R^. Thus, whenever
such a precipitate colony is inclined to the electron beam,
the displacement fringe effect will be observed. Of course,
a colony consisting of only one orientation of 0', whose
misfit does not lie in the plane of the colony, would cause
a displacement fringe effect also, as seen in Figure 4.38.
Similarly, Warren (1974), using the method of Humble (1968),
has computed electron micrographs of a plane of dilatation
and obtained the displacement fringe effect. An example is
shown in Figure 4.42. In addition, Clarebrough (1973) has
shown that passage of a unit dislocation through an ordered
lattice creates a planar displacement field at the slip plane
which causes similar fringe effects in electron micrographs.
Now the intensity of displacement fringes is a function
of the magnitude of the normal displacement R. It can be
visualized from Figure 4.41 that, if the density of precipi
tates varies over the colony, the local resultant displace
ment, R-^, varies accordingly. Thus, the fringes will vary


large spheroidized particles (-1/10 l/2p diameter), Figure
4.12. These particles were large enough to be analyzed on
a JEOL 100B Analytical Electron Microscope using a fine-
O
focused electron beam (approximately 1,000A diameter), and
a non-dispersive detection system for fluorescent analysis
of the emitted x-rays. The analysis of these large spheroids
identified them as pure lead. It is not known how lead
entered the sample material. However, such particles were
observed in only a few foils and nucleated only a small
fraction (<<1%) of the climb sources present. The typical
source particle was so small as to be barely visible or not
visible at all, Figure 4.2(b). Such particles were too
small for the x-ray analysis, but they are most probably not
lead. The chemical analysis of the 3.85 wt.% Cu alloy
(Section 3.1) showed no appreciable concentration of any
single impurity which would suggest a guess at the particle
nature.
Although the chemical composition of the source particles
is unknown, some observations were made about their dis
tribution. The climb sources were dispersed randomly in
most portions of the foils. Occasionally, local high densi
ties of sources were observed. In a few instances, sources
were observed evenly spaced in a straight line (Figure 4.13),
suggesting that the source particles were part of an impurity
stringer produced when the original cast alloy was rolled down.
The density of active climb sources in these foils varied
with heat treatment (Section 4.5). In the only other reported


REPEATED NUCLEATION OF PRECIPITATES
ON DISLOCATIONS IN ALUMINUM-COPPER
By
THOMAS JEFFREY HEADLEY
A DISSERTATION PRESENTED TO THE GRADUATE COUNCIL OF
THE UNIVERSITY OF FLORIDA
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE
DEGREE OF DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
1974


128
Figure 4.40. Displacement fringe contrast in precipitate
colonies in samples quenched from 550C to
220C and aged for 5 minutes in (a) and for
1 minute in (b). The dislocation which
nucleated the colony in (a) is out of con
trast at the bottom of the colony.


63
reflections from three non-coplanar beam directions, a neces
sary condition for uniquely identifying a dislocation by
computer matching (Head, 1969). From the rotation calibra
tion of the electron microscope, the exact orientation of
the g-vector was marked on each experimental image. Also,
from the known geometry of the computer program, the orienta
tion of the g-vector was marked on each computed image.
Thus the direction of the g-vector serves as a basis for
comparison when matching the features of the computed images
with those in the experimental images. Now for a given dif
fraction vector g, the image of a dislocation with Burgers
vector +F is identical to that of a dislocation with Burgers
vector -F after a rotation of 180 (Head e_t al_. 1973, p.
382). Clearly it can be seen from Figure 4.7 that the
Burgers vector of Segment A of the loop is a/2[Oil] and not
a/2[Oil].
By convention, the positive direction of the dislocation
line in the computer program is always taken to be acute to
the foil normal. Thus for the foil normal [313] the posi
tive direction of Segment A is [100] and not [100]. Also,
the computer program employs the finish-to-start, right-hand
(FS/RH) convention for establishing the direction of the
Burgers vector with respect to the positive sense of the
dislocation line, Figure 4.8. Thus, from the (FS/RH) conven
tion, and the absolute Burgers vector and positive line
direction of Segment A, the geometry of the two loops in
Figure 4.6 can be established, and this is illustrated


1 certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
John'J. Hren, Chairman
Processor of Materials Science
and Engineering
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
^ v.
P. N. Rhines
Professor of Materials Science
and Engineering
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
Vl
g.l. Q*l-UtP
Rl IP Reed-Hill
Professor of Materials Science
and Engineering
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
/
T.' "tfeHoff
Professor
and Engineering
Science


161
Figure 4.55. Higher magnification of the structure in
Figure 4.53 showing that the interiors of the
small dislocation loops contain precipitate
colonies. Loops at B are viewed edge-on,
while loops at A are out of contrast around
their precipitate colonies.


38
of these mechanisms is given by Hirsch et ad. (1965, p. 336).
Those pertinent to this research will be outlined below.
(1) Strain contrast in the matrix. All coherent and
semicoherent precipitates, and most incoherent precipitates,
cause some strain in the matrix. These strain fields there
fore give rise to diffraction contrast effects in the matrix.
This can be the only mechanism for imaging very small pre
cipitates whose sizes are less than the resolution limit of
the microscope, but whose long-range strain fields are
greater than this limit.
(2) Misfit dislocation imaging. Semicoherent precipi
tates have misfit dislocations over their semicoherent inter
faces. The strain fields of these misfit dislocations can
cause strain contrast just as for isolated dislocations in
the matrix. Weatherly and Nicholson (1968) have investigated
the conditions for imaging misfit dislocations. Often small
platelets viewed normal to the platelets are imaged by the
misfit-dislocation loops around their edges. This is
referred to as "dislocation-ring" contrast.
(3) Structure factor contrast. According to Ashby and
Brown (1963), this contrast arises whenever a coherent pre
cipitate has a different structure factor from the matrix,
and thus a different extinction distance. A particle of
thickness At then increases the effective foil thickness in
columns passing through the particle, giving rise to an
intensity change relative to columns in the matrix. Depend
ing on the depth of the particles in the foil and the


201
(2) a source of dislocations during quenching,
(3) a driving force for dislocation climb which
is independent of the precipitation process,
(4) a climb rate slow enough to permit nucleation
but rapid enough to move the dislocation past
the stable precipitates.
This research has provided a basic understanding of the
phenomenon in Al-Cu, but it is by no means complete. Several
topics for further investigation are suggested:
(a) If a dislocation can climb through the lattice at
elevated temperatures and provide the catalyst for repeated
nucleation, then perhaps a gliding dislocation can also. It
would be instructive to examine samples deformed either at
temperatures slightly below the 0 solvus or during quenching
from the solution treatment temperature for evidence of
repeated nucleation during glide.
(b) It is suggested that a small-angle x-ray investiga
tion at temperatures near the 0' solvus would provide direct
evidence for copper clustering if it exists. If so, a
determination of the density and size distribution of clus
ters would help resolve whether or not clustering could be
involved in the repeated nucleation mechanism.
(c) Trace element additions are known to alter the
kinetics of the a->0' transformation, presumably by segregat
ing to the particle interface where they can change both the
interfacial energy and, to some extent, the misfit strain.
It would be of interest to determine if trace additions of


Eshelby, J. D. (1957), Ptoc. Royal Soc, London, A241 376.
Eshelby, J. D. (1961), Progress in Solid Mechanics, Sneddon
and Hill, eds. North -Holland, p. ¡T9T
Federighi, T. and Thomas, G. (1961), Phil. Mag. 6_, 127.
Frank, F. C., and Read, W. T. (1950), Phys. Review, 79, 722.
Friedel, J. (1964), Dislocations, Addison-Wesley.
Guerard, B. von, Peisl, H., and Zitzmann, R. (1974), Appl.
Phys., 3, 37.
Guinier, A. (1938), Nature, 142, 569.
Guinier, A. (1939), Ann. Physique, 12, 161.
Guinier, A. (1942), J. Phys. Radium, _3, 129.
Guinier, A. (1950), Compt. Rend., 231 655.
Guinier, A. (1952) Acta Cryst. 5_, 121.
Gulden, M. E. and Nix, W. D. (1968) Phil. Mag. 18_, 217.
Head, A. K. (1969), Aust. J. Phys., 'n, 43.
Head, A. K., Humble, P., Clarebrough, L. M., Morton, A. J.,
and Forwood, C. T. (1973), Computed Electron Micrographs
and Defect Identifica!ion, Amelinckx, S'. Gevers R7,
and Nihoul, J., eds., North-Holland.
Hirsch, P. B., Howie, A., Nicholson, R. B., Pashley, D. W.,
and Whelan, M. J. (1965), Electron Microscopy of Thin
Crystals, Butterworths.
Hirth, J. P., and Lothe, J. (1968), Theory of Dislocations,
McGraw-Hill.
Hornbogen, G. (1967), Aluminum, 3, 163.
Hosford, W. F., and Agrawal, S. P. (1974), paper presented
at the 6th Annual Spring Meeting of A.I.M.E., Pittsburgh,
Pa.
|, A. ,
and
Whelan, M.
J. (1961), Proc.
Royal
Soc. London,
A2 6 3 ,
217.
A. ,
and
Whelan, M.
J. (1962), Proc.
Roval
Soc. London,
A26 7 ,
206.
. e P .
(19 £
>8) Aust.
J. Phys., 21, 32 5


166
Secondly, it was found that portions of many small loops
had moved off their {110} habits. Examples are shown in
Figure 4,58. Since the original {110} habit is the most
favorable climb plane for these edge loops (Miekk-oja and
Raty, 1971), it is unlikely that they would move off by climb.
They have most probably slipped onto intersecting {111} planes
under the quenching stresses.
4.5.5. Effect of Copper Concentration
Samples with decreasing copper content were given direct-
quenching treatments to determine if solute concentration
played an important role in the repeated nucleation mechanism.
The copper concentrations of the samples were nominally
1.96 wtJ, 1.0 wt.%, and 0.5 wt.%, respectively (evidence for
repeated nucleation in the 3.85 wt.I Cu alloy has already
been discussed in great detail). The samples were solution
treated for one hour at 545C, direct-quenched into oil at
210C, and aged at that temperature. These temperatures are
shown in relation to the 0' and 0" solvus lines in Figure
4.59. Samples of the 1.96 wt.% Cu alloy were aged for either
30 minutes or one hour at 210C. Samples of the 1 wt.% Cu
alloy were aged for either three seconds, one hour, or 24
hours at 210C. Samples of the 0.5 wt.% Cu alloy were aged
for either three seconds, one hour, or 7-1/2 hours at 210C.
It was found that repeated nucleation of 0' occurred
extensively in the 1.96 wt.% alloy, but did not occur at all
in the 1.0 and 0.5 wt.% Cu alloys. Appreciable dislocation


99
Figure 4.27. Stringers of small dots along [010] in a
sample quenched from 550C to 220C and
aged only 8 seconds.


88
The (010) 0' orientation, which is parallel to the beam
direction and also viewed edge-on, is not present in habit B,
although it is easily detected in habits C and E. Again,
this is the 0' orientation whose misfit is perpendicular to
the a/2[101] Burgers vector of the source loops bounding
h ab i t B.
The fact that 0' platelets lying parallel to the beam
can indeed be seen if present in precipitate colonies viewed
edge-on is shown in Figure 4.23. This micrograph was taken
with the beam oriented near [001]. The two edge-on habits
at A and B lie on (110) planes so that the bounding disloca
tion loops have Burgers vectors a/2[110]. Both the (010)
and (100) orientations of 0' platelets can be clearly seen
dispersed along the habits. Note that these are the two 0'
orientations whose misfits are not perpendicular to the
Burgers vector of the bounding dislocation loops and are
therefore favored to be nucleated by the loops.
All three 0' orientations were never observed in a given
precipitate colony. As pointed out above, the missing orien
tation was always that whose nucleation is not aided by the
stress field of the dislocations bounding the colony. This
was true for precipitate colonies generated by both climb
sources and glide dislocations which climbed. This evidence
leads to the important conclusion that the precipitates must
have nucleated in the stress field near the dislocations as
they climbed through the lattice, and not at some later time
when the influence of the dislocation was no longer present,


181
a
b
Figure 5.2. Two images of a precipitate colony on a climb
source showing that the precipitates 1, 2, and
3 lie outside the outermost loop. The source
loop is out of contrast in (a), and the three
precipitates are out of contrast in (b). The
beam direction is close to [Oil] in (a), and to
[001] in (b).


43
above the G.P. solvus, or (2) quenched to a low temperature.
The samples given direct-quenches were aged for various times
from approximately one second to 24 hours, and then quenched
into room-temperature water. Samples quenched to low temper
atures were either prepared for electron microscopy without
further treatment, or they were up-quenched to a temperature
above the G.P. solvus and aged for various times. They were
then quenched into room-temperature water.
The solution treatments were conducted in a vertical
furnace in air. The temperature in the heat zone was con
trolled to within 2C. One end of the sample was clamped
in a stainless steel alligator clip attached to the bottom of
a one-half inch diameter stainless steel tube, and this was
inserted into the heat zone of the furnace. Before each
treatment, the temperature in the heat zone was determined by
inserting a thermocouple into a dummy stainless steel tube
suspended in the heat zone.
Quenching was achieved by dropping the specimen-stainless
steel tube assembly out of the bottom of the furnace into the
quench bath. For direct-quenches to the aging temperature,
the specimen was dropped into a Lauda Constant Temperature
Oil Bath, maintained at the aging temperature, to within
0.2C. A schematic diagram of the apparatus used for solu
tion treatment and direct-quenching is shown in Figure 3.1.
For quenching to low temperatures, the specimen was dropped
into one of several low temperature baths in place of the
oil bath. Following the quench to low temperatures, some


139
colonies. After 5 minutes aging, the interiors of the
colonies are seen to be densely precipitated. All precipi
tates in the colonies were present at the start of the aging
at 220C (Section 4.3). After 30 minutes at 220C, the pre
cipitates are large enough to individually exhibit displace
ment fringe contrast, but precipitation is still localized
within the original colonies. After two hours at 220C, some
scattered precipitates are observed outside the colonies,
but the vast majority are still associated with the original
colonies.
These results indicate that the density of precipitates
generated by repeated nucleation is sufficiently large that
aging for long times results essentially only in growth.
Since there was no evidence for bands of precipitates spread
ing out from the colonies, an autocatalytic nucleation mecha
nism (Section 2.2) can be ruled out. In the original work
on autocatalytic nucleation of 0' in Al-Cu, Lorimer (1970)
found a uniform distribution of precipitates throughout the
foil after aging 35 minutes at 240C. In the presence of a
large precipitate density, generated by repeated nucleation
on climbing dislocations, evidently the driving force for
autocatalytic nucleation is small, and solute depletion of
the supersaturated lattice can be accomplished by growth of
existing precipitates.


69
Figure 4.12. Climb sources generated at a large, spherical
lead particle. (Heat treatment: S.T. 1 hour
550C, quench to 220C, aged 1 minute.)
Figure 4.13. Climb sources aligned in a row. (Heat treat
ment: S.T. 1 hour 550C, quench to 200C,
aged 5 minutes.)


30
INCIDENT BEAM
Figure 2.9. Method for forming a bright-field image under
two-beam conditions. The transmitted beam is
allowed to pass through the objective aperture
while the diffracted beam is blocked by the
aperture holder.


169
Figure 4.60. Precipitate colonies nucleated on (a) a long
glide dislocation which is out of contrast
between AB, and (b) dislocation climb sources
in the Al-2 wt.% Cu alloy. (Heat treatment:
S.T. 1 hour 545C, quenched to 210C, aged
1 hour.)


48
3.4. Electron Microscopy
The thin foils were examined in a Phillips EM 200 elec
tron microscope operated at 100 Kv potential. A goniometer
stage with 45 and 30 tilt on two orthogonal axes was used.
Two-beam diffraction conditions were established for taking
all micrographs. To obtain two-beam conditions, the foils
were oriented close to one of the low index poles shown on
the Kikuchi line map for an f.c.c. crystal in Figure 3.3.
Use of this Kikuchi line map during specimen tilting, as
described by Head et al. (1973), enabled diffraction vectors
to be determined unambiguously in every case.


145
Two observations were made about the effect of quenching
temperatures on the operation of climb sources:
(1) The average size of the source loops increased with
increasing solution treatment temperature. In the samples
quenched from the lowest temperatures, the climb sources were
small, with few exceptions. In the sample quenched from
570C, both small and very large sources (~5-6p diameter)
were observed, the average size being the largest for all
I
samples. Figure 4.47 shows a very large source in this
sample.
(2) The number density of active climb sources was very
low for quenching from 504C, increased with quenching tem
perature to a maximum for 550C, then decreased to an inter
mediate value at 570C. Insufficient micrographs were
obtained to make reliable quantitative measurements of these
densities, so that the trend is described only qualitatively
here. An important influence on microstructure is that vari
ations in size and density of climb sources are reflected in
the density of precipitates which nucleate during quenching.
The above observations can be explained as follows. The
vacancy supersaturation increases exponentially with quench
ing temperature so that the average distance a source loop
will climb increases. Secondly, if nucleating sites for
climb sources are indeed particles, the particle solubility
will tend to increase with temperature. Thus, as the solu
tion treatment temperature was increased, a trade-off between
increasing vacancy supersaturation and a decreasing density


159
Figure 4.53. Microstructure of a sample quenched from 550C
into room temperature oil, then up-quenched to
220C and aged 5 minutes, showing a high
density of small dislocation loops and several
glide dislocations which climbed (at points A).
Figure 4.54. Dislocation climb sources and associated pre
cipitate colonies in the sample quenched into
room-temperature oil, then up-quenched to
220C and aged 5 minutes.


18
can also climb by vacancy-emission (negative climb) and this
is illustrated in Figure 2.5(b).
Christian (1965) has suggested that the vacancy-
annihilating climb of a/2<110> dislocations in f.c.c. lattices
occurs easily only on the (111) and (110} planes. A necessary
condition for climb is that the Burgers vector has a component
perpendicular to the plane of climb. Therefore there are two
(111} and five {110} "planes of easy climb" for an a/2[110]
dislocation in f.c.c. Miekk-oja and Raty (1971) have con
sidered the choice of climb planes in terms of the chemical
climb force on each plane. This force arises from the super
saturation (or subsaturation) of vacancies (Bardeen and Her
ring, 1952). Its magnitude is proportional to (5~xu) where 5"
is the Burgers vector and u is the dislocation line direction.
Thus, according to Miekk-oja and Raty, a dislocation with
F=a/2[110] is affected by the maximum climb force, F ,
on the {110} plane perpendicular to F. It is not affected at
all on the {ill} and {110} planes containing b. And it is
affected by forces 0.82F and 0.5F on the two {111}
c y max c y max
and four {110} planes, respectively, which are inclined to F.
2.3.2. Theory of Dislocation Climb
In reality, dislocation climb is more complex than the
simple picture envisioned in Figure 2.5. The theory of climb
has been developed by Lothe (1960), Thomson and Balluffi
(1962), Balluffi and Thomson (1962), Friedel (1964), Hirth


CHAPTER 5
THE REPEATED NUCLEATION MECHANISM
In the previous chapter, it was convenient to include
some discussion and analysis in each section. Hence, the
discussion in this chapter will be limited to aspects of the
mechanism of repeated nucleation of 0' in Al-Cu.
5.1. Nucleation of 9' Near Edge Dislocations
The a^0' transformation occurs with a 3.95% volume con
traction in the lattice (Section 2.2). The resulting misfit
strain energy is sufficient to suppress nucleation except at
dislocations whose stress fields reduce this strain energy.
The observation that just two 0' orientations nucleate in
each colony (Section 4.3) provides insight as to where the
nucleation events occur about the dislocation. One 0' orien
tation does not nucleate because its misfit is perpendicular
to the Burgers vector. Hence, its strain field is not
relieved by the stress field of the dislocation (Section 2.2)
The other two orientations nucleate because their misfits
lie at 45 to the Burgers vector, so that their strain fields
are partially relieved by the stress field of the dislocation
Since the a-*9' transformation is a volume contraction, the
178


23
4
Figure 2,7. The Bardeen-Herring model of a dislocation
climb source. An edge dislocation, pinned
between AB, has its slip plane normal to the
paper. It climbs in the plane of the paper,
by vacancy annihilation, through the succes
sive positions 1-4, rejoining at the bottom.
The portion ACB can then repeat the process
(after Bardeen and Herring, 1952).


55
e
Figure 4.3. Series of micrographs for determining the
geometry and Burgers vectors of the source
loops. The beam direction is close to [101]
in (a), (b) and (c) to [112] in (d) and
to [001] in (e). (Heat treatment: S.T.
1 hour 550C, quench to 220C, age 8 seconds.)


35
The validity of the Howie-Whelan equations in predict
ing intensities which correspond to two-beam images has been
overwhelmingly demonstrated by the success of computer simu
lation techniques for matching defect images (Head et ad.,
1973). (See for example Section 4.2.1.)
2.5.2. Defect Identification from
Invisibility Conditions
Although the solution of the two-beam equations is not
straightforward for the case of defects with varying dis
placement fields, often the solution per se is not needed to
identify the defect from its image. Instead, it is often
possible to apply a simple criterion to identify defects in
the electron microscope. This criterion is based on the
fact that the term in the equations which gives rise to con
trast is the product gR. The diffraction vector g is the
reciprocal lattice vector normal to the diffracting planes,
so that the product g*K samples the magnitude of the distor
tion created in the diffracting planes by the displacement
field R. If a defect happens to cause no distortion in the
diffracting planes for a two-beam condition, then g-R=0 and
the two-beam equations predict uniform intensity everywhere
at the bottom of the foil. In other words, there is no con
trast around the defect and it is said to be "invisible" for
this diffraction condition. This criterion, applied to the
identification of dislocations, can be described as follows.
To a first approximation, the planes parallel to the Burgers
vector of a dislocation in an isotropic crystal are not


142
Figure 4.46. Sequence of micrographs showing the effect on
microstructure of the solution treatment
temperature from which samples were direct-
quenched. Samples were solution treated for
1 hour at 570C in (a), 550C in (b), 530C
in (c), 515C in (d), and 504C in (e), then
direct-quenched to 220C and aged 5 minutes.


CHAPTER 1
INTRODUCTION
Some of the most important strengthening mechanisms in
alloys result from the precipitation of a second solid phase
from a supersaturated solid solution. The age-hardening
phenomenon in many aluminum alloys is a well-known example.
A precipitation reaction is a nucleation-and-growth trans
formation. Hence, the rate of the reaction is dependent upon
(1) the nucleation rate of precipitates, and (2) the rate of
their growth. If either or both of these rates is low the
reaction rate will be low. Consequently, it is important to
understand how and where precipitate reactions nucleate,
apart from the problem of growth. Much is known about the
kinetics of growth, but not about nucleation, especially
heterogeneous nucleation.
Precipitate reactions nucleate either homogeneously or
heterogeneously within the matrix. If both the volume misfit
and interfacial energy between precipitate and matrix are
small, the reaction can nucleate homogeneously at random
sites throughout the lattice. Homogeneous nucleation is
known to occur in only a few alloy systems, e.g., the pre
cipitation of cobalt particles from dilute solutions of
cobalt in copper (Servi and Turnbull, 1966). In most
1


187
dislocation. The nucleation event still involves a statis
tical probability that the correct combination of copper and
aluminum atoms rearrange into the correct structure with
sufficient size and shape (i.e., the critical nucleus size
and shape). It is valid, however, to assume that the proba
bility that this event will occur is higher in a copper-rich
cluster than elsewhere in the matrix. In addition, it is
generally believed that clusters can trap vacancies with a
binding energy characteristic of the alloy (Federighi and
Thomas, 1961). Such trapped vacancies could aid in the local
atomic rearrangements involved in forming the critical nucleus
structure. However, without available information on cluster
sizes and densities, it is not possible to estimate if clus
tering alone could provide enough solute buildup for nucle
ation in these samples.
(2) If it is assumed that the dislocation acts as a
highly efficient sink for the copper atoms it encounters
along the climb surface, then rapid pipe diffusion along the
core to a nearby growing nucleus can provide sufficient copper
locally for 6' precipitation. Assuming that thermal fluctu
ations create growing embryos of the new phase at various
locations along the dislocation (a basic assumption of nucle
ation theory), then adjacent portions of the climbing dislo
cation could collect and transport solute to these growing
embryos. Since dislocation pipe diffusion is much faster
than bulk diffusion, the frequency of atoms joining the criti
cal nucleus (the term 8 in Equation 2.1) is increased over


98
explanation for the precipitate-free zones in these direc
tions. In Shape V, there is a well-defined boundary where
precipitation begins behind the climbing source loop. Inside
this boundary the region is void of precipitates (except
where another interior loop has climbed). A possible expla
nation for this shape is as follows. Different climb sources
become active in nucleating loops at different times during
quenching. Any climb of the loops which occurred before the
temperature passed below the 0' solvus temperature does not
nucleate precipitates. Further climb below the 0' solvus
temperature generates ring-shaped, precipitate colonies
behind the loops.
For the case where several concentric source loops are
generated on a given {110} habit, each successive, interior
loop nucleates the same basic colony shape on a smaller
scale, e.g., Shape IV in Figure 4.26.
4.4.2. Geometry of the Precipitate Stringers
One of the early problems to be solved concerned the
identification of the defects in the stringers. The stringers
were already visible in samples quenched to the aging temper
ature and held for very short times (Figure 4.27) although
the general precipitate colony was not yet visible. The
stringers appear as rows of small dots, so that the contrast
cannot be interpreted further at this stage. In samples aged
for longer times, the stringers are imaged as rows of small
dislocation loops, Figures 4.21 and 4.25.


140
4.5.2. Effect of Solution Treatment Temperature
Five samples were solution treated for one hour at vari
ous temperatures within the solid solution range, then direct-
quenched to 220C and aged for five minutes. The five tem
peratures employed were 570, 550, 530, 515 and 504C,
covering the range from just below the solidus temperature to
just above the a+0 solvus temperature for the 3.85 wt.% Cu
alloy (Figure 4.45). The major difference between samples
was the quenched-in vacancy supersaturation, which increases
exponentially with quenching temperature. However, to a
lesser extent, the treatments also differed in quench rate.
The resulting microstructures are shown in Figure 4.46 at
low magnification.
First, repeated nucleation occurred during quenching in
all samples as indicated by the presence of precipitate
colonies associated with all dislocations. Therefore repeated
nucleation does not appear to depend on the vacancy super
saturation, at least for the range of supersaturations in
these direct-quenches. Secondly, both climb sources and
glide dislocations which subsequently climbed were present
in all samples, although their relative densities varied.
It is not possible to illustrate all the features of
these microstructures in one micrograph for each sample in
Figure 4.46. Accordingly, descriptions are given here based
on observations recorded during examination in the electron
micros cope.


29
beam scattered in the forward direction (called the "trans
mitted" or "main" beam), or in the strong diffracted beam.
The electron image is usually formed by placing an aperture
around one of these beams and allowing it to pass through,
while the other beam is blocked by the aperture holder
(Figure 2.9). The unblocked beam is then magnified by
successive lenses and projected onto the fluorescent screen.
When the aperture is placed around the transmitted beam, the
image formed is called a "bright-field" image. When it is
placed around the diffracted beam, the image is called a
"dark-field" image. Contrast at defects arises in, say, the
bright-field image because the strain in the lattice around
the defect causes local distortions in the atomic planes
which lead to local changes in the intensity scattered into
the diffracted beam. This in turn leads to local variations
in intensity in the bright-field image. This is illustrated
for the case of a dislocation in Figure 2.10.
When a crystal of sufficient thickness is oriented very
close to the Bragg condition for one set of planes, there can
be a dynamic interchange of electrons between the two beams,
resulting from multiple scattering back and forth as the two
beams pass through the crystal (Figure 2.11). In order to
predict the intensities in the bright- or dark-field image,
it is necessary to describe mathematically the physical pro
cesses which go on in this dynamic interchange. The two-beam
dynamical theory of electron diffraction for a distorted
crystal was developed by Howie and Whelan (1961). Their


105
Figure 4.30. Typical precipitate colony on a climb source
illustrating the fan-shaped boundaries of the
precipitate stringer regions. (Heat treatment
S.T. 1 hour 550C, quenched to 220C, aged
5 minutes.)
Figure 4.31. Evolution of the geometry of precipitate
stringers at climb sources with increasing
time at the aging temperature.


9
the presence of the dislocation stress field aids in reducing
the energy barrier to nucleation, but a rigorous calculation
of the effect is not yet available to provide an explicit
expression for the rate of heterogeneous nucleation at dis
locations .
It is instructive, however, to examine the order of mag
nitude of the terms in AG* to estimate the catalytic effect
of the dislocation. Typical values of AG(chem) are in the
range 1-5x10^ ergs/aA (20-100 cal/cm3). Values of AG(strain)
are in the same range as AG(chem) for particles with appre
ciable misfit. Values of o r obtained from bulk measure-
surf
ments are almost certainly too large since they relate to
incoherent interfaces, whereas critical nuclei whose sizes
o
are of the order of 10 !s of Angstroms should have coherent
interfaces. Estimates of coherent interfacial energies are
2
in the range 20-50 ergs/cm .
If we take AG(chem) = 2x10^ ergs/cm^, asur£ = 20
2 9 3
ergs/cm and AG(strain) = 2x10 ergs/cm then a spherical
nucleus with a diameter of 20A would have a chemical
-13
energy change and strain energy = 8x10 ergs each,
-13
surface energy = 24x10 ergs. Due to the problems
cussed above, no calculated values are available for
but it is estimated that it can be of the same order
free
and a
dis -
AG(int) ,
of mag
nitude as AG(chem) and AG(strain). Then if one assumes that
the surface energy is overestimated, as is likely (Barnett,
1973), the interaction energy due to the presence of the


57
When a given foil was first examined, a technique was
used for rapidly determining if the dislocation sources were
indeed climb sources with {110} habits, or if some or all of
them might be dislocation glide sources which are known to
operate on the {111} slip planes in Al-Cu alloys (Westmacott
et al. 1959). This technique was to tilt the foil to {001},
{111}, and {101} orientations and, in each orientation, to
determine the number of different source habits viewed edge-
on together with the angles between these habits. For exam
ple, \\rhen a foil was tilted to the (001) orientation, two
edge-on habits at 90 apart were observed, Figure 4.4(a).
Since the [001] pole is parallel to two {110} planes at 90
to each other, and not to any {111} planes, those sources
are identified immediately as {110} climb sources. Likewise,
when viewed in the (111) orientation, three edge-on habits
at 60 apart were seen, Figure 4.4(b). Again, since the
[111] pole is parallel to three {110} planes at 60 to each
other and not to any {111} planes, the sources are identified
as climb sources. However, neither of these cases rules out
the possibility that other sources seen inclined to the beam
in these orientations might be glide sources lying on {111}
planes. Therefore, it was necessary to tilt to a {101} ori
entation. The [101] pole is parallel to two {111} planes at
70.5 and to only one {110} plane. In this orientation, only
one habit was ever seen edge-on, as shown in Figure 4.1.
There was no evidence that any of the sources found in these
foils were glide sources.


195
in its habit plane. This is in agreement with the fact that
the stringers are always aligned along <100>.
This model predicts a thin, continuous ribbon of pre
cipitate, but it was shown in Figure 4.27 that the stringers
are composed of separate, small particles when they are first
visible after quenching. It is probable that immediately
after a thin ribbon forms, surface energy effects cause it to
pinch off into small platelets, in a manner similar to the
initial rapid spheroidization of eutectoid platelets upon
annealing. Further aging would then cause these platelets
to grow and coalesce as shown in Figure 4.31. This continu
ous ribbon model is consistent with the observation that
platelets of only one 0' orientation exist in any given
stringer (Section 4.4.2). In fact, it is difficult to imagine
a more logical interpretation.
It is suggested that initial rapid growth of the pre
cipitates in these stringers, enhanced by pipe diffusion
along the dislocation, creates a sufficient concentration
gradient in the matrix such that this is the only mechanism
which continues to operate once it begins. This could account
for the fact that uniform nucleation of precipitates does not
occur in the regions where the stringer mechanism operates
(Figure 4.30) .


103
b
Figure 4.29. (a) Bright-field and (b) dark-field images
at high magnification showing that the pre
cipitate stringers are actually composed of
separate but closely-spaced 0* platelets.
(Heat treatment: S.T. 1 hour 550C, quenched
to 220C, aged 5 minutes.)


143


190
a thickness sufficient to nucleate one a[100] misfit dislo
cation around its edge (-2 unit cells thick) and a diameter
slightly less than the resolution limit of the electron
microscope (-30 maximum). Such a critical nucleus would
contain about 90 unit cells. Hence, this model can supply
more than enough solute to nucleate the precipitate density
observed, so that it is not required for the dislocation to
collect all the solute in its climb path.
It is suggested that the pipe diffusion model provides
the primary means for locally enhancing the solute concentra
tion for repeated nucleation. However, it is likely that
there is at least some contribution, however small, from
bulk-diffusion to the surface of the growing embryos. The
importance of clustering cannot be accurately assessed, but
it is not a necessary condition. It would be helpful, of
course, in attaining local concentration enhancement.
5.3.2. Precipitate Stringer Formation
All the above considerations would predict a random dis
tribution of 0' nucleated by a passing dislocation. This
could account for the regions of uniform precipitation
observed behind the climbing dislocations, but not for the
regions of straight and uniformly-spaced precipitate stringers
which always form along <100> directions on climb sources,
Figure 4.30. We now consider a possible mechanism for this
stringer formation, and we limit the discussion to stringers
produced in precipitate colonies on climb sources.


12
forms as platelets on {100} planes of the matrix. 0' is
complex tetragonal with a = 4.04A and c = 5. 8$.. It also forms
as platelets parallel to {100} matrix planes, and is initi
ally coherent on its broad faces and semi-coherent on its
edge. As it grows its broad faces become semi-coherent.
The orientation relationship for both 0" and 0' is {100}
PP^
|| {100} . and <100> || <100> .
11 matrix ppt 11 matrix
The tetragonal unit cell of 0 is shown in Figure 2.2.
There are 6 atoms/unit cell. The a-solid solution is f.c.c.
with a = 4.045, and has 4 atoms/unit cell. When the atomic
volumes are calculated for these two unit cells and compared,
it is found that the a+0' transformation involves a 3.95%
volume contraction. The resulting transformation strain can
be partially compensated if vacancies are generated by the
growing precipitates and supplied to the matrix.
The early x-ray work established the following precipi
tation sequence for quenching and aging below the G.P. solvus:
G.P. zones + 0" -* 0' + 0(CuA12).
However, as suggested by the x-ray work and later confirmed
by many transmission electron microscope (TEM) studies,
several of the reactions can proceed concurrently depending
on the quenching and aging procedures. In addition, TEM
investigations have clearly established the homogeneous or
heterogeneous nature of the various reactions. Nicholson
and Nutting (1958) resolved G.P. zones and 0" platelets in
the 4 wt.% Cu alloy and found them to be homogeneously dis
tributed in the matrix. It is now clear, however, that 0"


176
often observed that, where a large outer loop intersected
both foil surfaces, the two lines of intersection were not
parallel, indicating that the loops themselves are not
strictly planar. These observations could be explained only
if the loops are allowed to slip or climb locally out of
their habit planes in varying step heights.
4.6. Summary
The results in this chaoter may be summarized as follows:
(1) The 0' phase nucleated repeatedly on climbing dis
locations during quenching, creating precipitate colonies
along the climb paths. The precipitate density thus generated
was sufficiently large to suppress autocatalytic nucleation
during long aging.
(2) The dislocations cLimb during the quench by annihi
lation of quenched-in vacancies. These dislocations fall into
three categories according t a origin:
(a) pure-edge loous on {110} planes, gener
ated at dislocation climb sources,
(b) glide dislocations on {111} planes, and
(c) prismatic edge-loops on {110} formed by
collapse of vacancy clusters.
The dislocation density in each category varied with heat
treatment. Categories (a) aid (b) were found in all samples.
Category (c) was found only In samples quenched into oil or
water at room temperature.


134
Figure 4,43. (a) Dislocation subboundaries in a sample
direct-quenched from 550C to 220C and held
only 8 seconds. (b) Precipitate colonies
associated with subboundaries in a sample
direct-quenched from 550C to 220C and
aged 5 minutes.


84
precipitate colonies. One habit lies in the plane of the
foil at A, one habit is viewed edge-on at B, and three other
habits at C, D and E are inclined through the foil. The
geometries of these sources are shown schematically in Figure
4.21(c) and (d) .
Often the best condition for imaging the precipitates
in a colony was also a condition for "invisibility" of the
dislocation loop bounding the colony, e.g., Figure 4.21(a).
This was particularly true when examining colonies on dis
location climb sources where the best projected view of a
colony was obtained with the beam oriented approximately
normal to the colony, and hence to the bounding source loop.
However, since the source loops are pure-edge with their
Burgers vectors normal to their habit planes, any two-beam
reflection selected to view the precipitate colony in this
orientation has its g-vector perpendicular to the Burgers
vectors of the loops. Thus the loop will be "invisible" when
viewed normal to its habit plane. This is the case in Figure
4.21(a) where g*£T=0 for the loop bounding the precipitate
colony and only residual contrast due to the pure edge-nature
of the loop is detected.
Likewise in Figure 4.21(b), the dislocations bounding
the colonies at A, C and D are invisible, whereas the one at
E is visible. A consistent interpretation of the geometry
of this source array is as follows. The beam direction is
close to [101] and g=lll for this image. Habit A lies on
the (101) plane of the foil (see Figure 4.21(d)). The


186
5.3.1. Local Solute Buildup
The matrix is a random distribution of 2-4 wt.% Cu (>1-2
atomic!) in the aluminum lattice. For nucleation to occur
repeatedly, it is necessary that concentration buildups occur
at the moving dislocation which can provide for the high
copper content of the 0! nuclei (v33 atomic!). It is sug
gested that the necessary concentration changes probably do
not come about by long-range bulk-diffusion of copper. Prior
to nucleation there is no concentration gradient to promote
long-range copper diffusion to the dislocation. There is,
though, a drift force on the copper to diffuse to the dislo
cation and lower the associated strain energy. However,
since nucleation occurs rapidly during quenching, it is un
likely that bulk-diffusion due to the drift force is rapid
enough to cause the necessary copper buildups.
In the absence of a sufficient contribution from long-
range bulk-diffusion, we consider two possibilities whereby
the copper concentration can be enhanced locally at the
moving dislocation.
(1) There is indirect evidence from resistivity measure
ments that copper atoms cluster during quenching of Al-Cu
alloys (Perry, 1966). Clusters located at or very near the
climb plane would provide copper-rich regions that might
transform to 0' as the dislocation stress field passes. It
is not valid to assume that, a cluster will automatically
transform to 0' if it contains more copper atoms than arc
required for the critical nucleus size in the presence of the


85
Burgers vector of its source loop, being pure edge, is
a/2[101], Thus the loop is invisible for g=ll. The source
loops bounding the habits C and D are also invisible. These
dislocations must have the other two a/2<110> Burgers vectors
which cause invisibility for the 111 reflection, namely,
a/2[Oil] and a/2[110]. Habit B lies on the (101) plane
parallel to the beam and is viewed edge-on. The Burgers
vector of its source loops must be a/2[01]. The source loop
bounding habit E must have one of the two remaining a/2<110>
Burgers vectors, namely, a/2[110] or a/2[Oil], both of which
would be visible for the 111 reflection. The dislocation is
visible at E.
It is difficult to determine whether the smallest pre
cipitates in these colonies are imaged by strain contrast in
the matrix or by structure factor contrast (Ashby and Brown,
1963). The larger precipitates in a colony are imaged by
the dislocation loops bounding the periphery of the platelets
(Section 4.4).
In Section 2.2, it was pointed out that because of dis
location strain effects, only two 0' orientations will nucle
ate on any given a/2<110> dislocation. The missing orienta
tion has its principal misfit (normal to the plane of the
platelet) perpendicular to the Burgers vector of the dislo
cation so that its strain field is not relieved by the stress
field of the dislocation. Careful examination of a number
of precipitate colonies at dislocation climb sources revealed
that only two 0' orientations were present in any given colony.


158
b
Figure 4.52. Precipitate colonies nucleated on (a) initial
glide dislocations, and (b) climb sources in
a sample quenched from 550C into liquid
nitrogen, then up-quenched to 220C and aged
5 minutes.


152
These observations can be explained if two factors are
accounted for. First, Figure 4.50 shows typical cooling
curves for identical samples quenched at two different rates.
Clearly, the sample quenched at the slower rate, takes
longer to reach a given temperature, T-^, than a sample
quenched at a faster rate, The former also spends longer
within any given temperature increment, AT. Secondly, the
variation in active source density can be treated as a problem
analogous to the nucleation of precipitates over a distribu
tion of favorable sites at different aging temperatures. It
is assumed only that some particles will be more favorable
sites for nucleating loops than others.
First consider the sample with the slowest quench rate
(Ta=300C). Shortly after the onset of quenching, source
loops nucleate only at the most favorable particles. Since
this sample spends a maximum time at high temperatures, the
diffusion distance for vacancies is large. The first loops
to nucleate can grow to be large, depleting the matrix of
vacancies, and thereby suppressing the nucleation of other
loops in nearby regions. As a result, this sample should
have the lowest active source density, but the largest aver
age loop size, in agreement with experiment (Figure 4.49(a)).
As the quench rate increases (i.e., as Ta decreases), the
time spent at high temperatures decreases and the diffusion
distance for vacancies decreases. The first loops to nucle
ate can no longer grow as large before other loops nucleate
on the next most favored particles (Figure 4.49(b)). Carried


BIOGRAPHICAL SKETCH
Thomas Jeffrey Headley was born June 22, 1943, at
Sheffield, Alabama. In June, 1961, he was graduated from
Coffee High School, Florence, Alabama. In June, 1965, he
was graduated from Virginia Polytechnic Institute with the
degree of Bachelor of Science in Metallurgical Engineering.
In June, 1967, he was graduated from Virginia Polytechnic
Institute with the degree of Master of Science in Metallur
gical Engineering. He worked, first as an Engineer, and
then as an Associate Scientist for the Lockheed-Georgia
Company, Marietta, Georgia, from July, 1967, to April, 1970.
He entered graduate school at the University of Florida in
April, 1970. Since that date, he has worked as a graduate
research assistant in the Materials Science and Engineering
Department while pursuing the degree of Doctor of Philosophy.
Thomas Jeffrey Headley is married to the former Lynn
Lancaster Moore and is the father of two children. He is a
member of the American Institute of Mining, Metallurgical,
and Petroleum Engineers, the American Society for Metals,
Alpha Sigma Mu, and the Society of Sigma Xi.
20 8


194
Figure 5.5. Diagram of an expanding climb source loop on
(110) with Burgers vector a/2[110]. The two
0' orientations, (100) and (010), which
nucleate on this dislocation, can be extended
in continuous ribbons from moving superkinks
only along the [100] direction.


92
(d)
Figure 4.24. Continued. (c) and (d) Micrographs o£ the
same dislocations in (a) and (b) after aging
9 minutes at 230C in the electron microscope.
The random precipitation at B has occurred at
the foil surfaces. The precipitate colonies
associated with the dislocations are now
clearly visible at A.


202
such elements as Cd, In, or Sn could either enhance or elimi
nate repeated nucleation during quenching.
(d) An item, examined to some extent here, which needs
further investigation is the nature of the segmented climb
of glide dislocations in this system. In particular it would
be of interest to understand more completely why this mode
of climb nucleates precipitate bands containing only one e'
orientation instead of two.
(e) As a by-product, this investigation has established
a good understanding of the operation of dislocation climb
sources in Al-Cu alloys, but some questions remain to be
answered. First, the nature of the source particles is still
unknown. The capability of the new breed of scanning trans
mission electron microscopes (STEMs) to focus the electron
O
beam to diameters on the order of 15, and to thereby obtain
microdiffraction patterns and/or x-ray chemical analysis from
very small particles, might be employed to solve this problem.
Secondly, further work is required to understand the depen
dence of the shape of source loops on solute concentration.
(f) It is suggested that a calculation of the inter
action energy between a 0' nucleus and an edge dislocation
would reveal the location of the nucleation events around the
dislocation. The method outlined by Larch (1974) for a
coherent nucleus at an edge dislocation should be modified
for the known geometry of a 0' platelet on {100} at an edge
dislocation on (110). The position which maximizes the inter
action energy (more negative) should be sought.


79
Figure 4,19.
(a) (001) diffraction pattern showing
tate reflections, taken from the area
foil shown in (b). (Heat treatment:
1 hour 550C, quenched to 220C, aged
minutes.)
precipi-
of the
S.T.
30


64
o
o
o
o
o
o
o
o
o
o
O O O O O O
a
oro_p. oe-o
o ob o o o o
b
O
O
O
o
o
o
Figure 4.8. Schematic representation of Burgers circuits
taken in a cubic lattice around an edge dis
location (a) and in perfect crystal (b) ?
illustrating the FS/RH definition of the
Burgers vector. The positive sense of the
dislocation line is out of the paper (after
Head et al. 19 7 3).
(011) PLANE
I
^5=a/4[0TT]
Figure 4.9. The geometry of dislocation climb source loops
in Al-Cu as indicated by the absolute sense of
the Burgers vector determined from computer
matching. The pure-edge loops expand in their
habit plane by vacancy-annihilating climb.


136
b
Figure 4.44. Sequence of micrographs showing the effect
of time at constant aging temperature on
colony growth. Samples were direct-quenched
from 550C and aged for (a) 8 seconds,
(b) 1 minute, (c) 5 minutes, (d) 30 minutes,
and (e) 2 hours.


141
Figure 4.45.
Diagram showing the five solution treatment
temperatures from which samples were direct-
quenched to 220C and aged.


171
bands of small dislocation loops were present throughout the
microstructure of the sample aged only three seconds at 210C,
Figure 4.62. These small loops exhibit stacking fault con
trast and are assumed to be Frank loops formed by the collapse
of vacancy clusters. After aging one hour at 210C, no such
loops were observed and it is assumed that they annealed out.
No evidence for repeated precipitation was observed in this
alloy, even after aging for 7-1/2 hours at 210C. In fact,
no precipitates at all were detected in this alloy, although
it was aged in the two-phase a+0' region (Figure 4.59).
The shape of climb source loops changed appreciably in
going to the 0.5 wtA Cu alloy. After direct-quenching and
aging for three seconds at 210C, some climb source loops
had the Class IV shape of Figure 4.26, but most had the shape
of nearly perfect rhombuses. After aging for one hour at
210C, all source loops, without exception, had the shape of
nearly perfect rhombuses (Figure 4.63). The long axis of the
rhombus is close to the <100> direction in the {110} habit of
the loops. The loop sides lie along <112> directions to
within about 5%. These <112> directions are the lines of
intersection of the two {111} planes perpendicular to the
{110} plane of the loops, i.e., these are the two slip planes
containing the a/2<110> Burgers vectors of the loops. This
is the same geometry of climb source loops observed in Al-Mg
alloys by Embury and Nicholson (1963).
It is now apparent that, as the copper content of the
alloy decreases from 3.85 wtJ to 0.5 wt.%, the typical climb


repeated nucleation process wore determined. It was found
that repeated nucleation occurs during quenching from all
temperatures within the solid solution range, to all temper
atures in the range room temperature to 300C. It occurs
during slow and fast quenching as well, but does not occur
in alloys with concentration <1 wt.% Cu.
Mechanisms of repeated nucleation proposed earlier for
other alloys are not applicable to Al-Cu. Dislocation climb
and precipitation were found to be independently controlled
processes. The relevant criteria for repeated nucleation in
this system are:
(1) a precipitate phase which nucleates easily on
dislocations,
(2) a source of dislocations during quenching,
(3) a driving force for dislocation climb which is
independent of the precipitation process, and
(4) a climb rate slow enough to permit nucleation
but rapid enough to avoid pinning.
It is suggested that pipe diffusion along the moving dislo
cation provides the necessary solute enhancement for succes
sive nucleations.


104
is, the stringers at the center of these regions extend
farthest into the middle of the colony, points A in Figure
4.30. The stringers become progressively shorter in going
to the ends of the stringer regions, points B.
As aging time at temperature is increased, the precipi
tates in a stringer grow and probably coalesce, so that
stringer geometry evolves through the sequence of shapes
shown in Figure 4.31. The outermost precipitate tends to
grow into a Y-shape along the bowed-out, climb source loop.
A given source loop always nucleated stringers of the
two 9 orientations compatible with its Burgers vector.
However, platelets of only one orientation were nucleated in
any given stringer, Figure 4.29. Thus, there is some geo
metrical restriction about the origin of a stringer at or
near the dislocation loop which favors repeated nucleation
of only one 0 orientation.
Measurements were made of the average spacing between
stringers, and the average spacing between precipitates in a
stringer. The former were made in precipitate colonies
viewed normal to their {110} habits so as to obtain the true
projected spacing. The latter were made on the dark field
image in Figure 4.29(b). This sample was aged for a short
enough time that coalescence of precipitates had not yet
begun. The average spacing between stringers, measured normal
to the <100> direction was found to be 0.096y (960). The
average spacing between precipitates in a stringer was found


87
i
Figure 4.22. Bright-field and dark-field images of precipi
tate colonies on dislocation climb sources.
The colony at A lies in the (101) plane of the
foil. The colonies at B are inclined through
the foil. The dark-field was taken from a
precipitate reflection from the (010) 0' orien
tation lying parallel to the [101] beam direc
tion. (Heat treatment: S.T. 1 hour 550C,
quenched to 220C, aged 5 minutes.)


21
b
c
Figure 2.6. Diagram of vacancy processes associated with
climb of the extra half-plane of an edge
dislocation. Vacancies absorb onto the core
(a) diffuse along the core (b) and annihilate
at jogs (c). Subsequent motion of the jog
across the page moves the extra half-plane
up one atomic spacing.


17
reported mechanism whereby the a+0* reaction was found to
propagate from a low initial dislocation density.
2.3. Dislocation Climb
2.3.1. Quenched-In Vacancies and the
Chemical Climb Force
It is now widely accepted that vacancies can exist in
crystals in thermal equilibrium with the lattice. The equi
librium concentration of vacancies increases exponentially
with temperature according to the Arrhenius relation:
Cq = A exp(-E^/kT)
where A = an entropy factor,
E^ = the activation energy for forming a vacancy,
k = the Boltzmann constant, and
T = temperature.
Large supersaturations of vacancies can be retained in the
lattice by quenching rapidly from elevated temperatures.
During and after the quench, the excess vacancies diffuse to
sinks such as surfaces, grain boundaries, and dislocations
or they may cluster and collapse into vacancy disks bounded
by dislocation loops. The condensation of vacancies onto a
dislocation causes it to undergo positive climb. The greater
the supersaturation of vacancies, the greater is the driving
force for climb. A simple picture of dislocation climb by
vacancy annihilation is shown in Figure 2.5(a). Dislocations


32
INCIDENT
BEAM
TRANSMITTED
BEAM
Figure 2.11.
Diagram illustrating the dynamic interchange
of electron intensity between the two beams
resulting from multiple scattering events.


rr
Figure 4.62. Band o£ small prismatic dislocation loops in
Al-0.5 wt.% Cu quenched from 545C to 210C
and aged only 3 seconds.
Figure 4.,6 3. Rhombus-shaped climb sources in Al-0.5 wt.% Cu
quenched from 550C to 210C and aged 1 hour.
Local segments at A, B, C, and D appear to have
slipped out of the climb plane.


170
Figure 4.61. Micrographs of Al-1 wt.% Cu alloy quenched
from 5 45C to 210C and aged 1 hour, shox^ing
that no repeated nucleation of 0' occurred
during quenching. (a) Glide dislocations;
(b) climb sources.


42
Table 3.1
Impurity Levels in the Al-3.85 wt.l Copper Alloy
Impurity
Weight ppm
Fe
5-25
Pb
<10
Si
1-10
Mo
<10
Mg
5-20
Ca
1-5
Ga
<10
Ba
10-40
were then cold-rolled to 0.005 inch for heat treatment. The
3.85 wtA and 1.96 wt.% Cu alloys were obtained as rolled
sheet, 0.004 inch thick. They were not reduced further
before heat treatment. Samples for heat treatment were cut
from the rolled foils to the approximate dimensions 1/8 x
1/2 x 0.004 inch. This was found to be a suitable size for
preparing electron microscope specimens after heat treatment.
3.2. Heat Treatments
All samples were solution treated for one hour at a
temperature in the a-solid solution range (Figure 2.1). Next,
they were either (1) direct-quenched to an aging temperature


I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
Associate Professor of Materials
Science and Engineering
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
y. A. Eisenberg *
Associate Professor of Engineering
Science and Mechanics and
Aerospace Engineering
This dissertation was submitted to the Graduate Faculty of
the College of Engineering and to the Graduate Council, and
was accepted as partial fulfillment of the requirements for
the degree of Doctor of Philosophy.
August, 1974
Dean, Graduate School


66
[ooi]
Figure 4.10. Diagram showing the stacking of atoms in
(110) planes in a face centered cubic crystal
(viewed normal to the planes). Atoms in the
third plane down lie in A positions. Removal
of a single plane of B atoms creates a stack
ing fault, A on A.


153
Time to Reach Temperature
Figure 4.50. Diagram showing the differences in the time
to reach a given temperature and the time
spent in a given temperature range AT, for
two different quench rates.


li
Figure 4.42. Computer-simulated electron micrographs of a
plane of dilatation inclined through the foil,
illustrating the displacement fringe effect.
The fringes lie at about 45 to the horizontal,
and there is a horizontal dislocation image at
the top of the fringes. The strength of the
dilatation increases from the top image to the
bottom so that the intensity of the fringes
increases (after Warren, 1974).


37
this reason, practical experience in recognizing "residual
contrast" is necessary in order to identify dislocations from
the invisibility criterion.
The criterion g*F=0 for invisibility is valid only for
total dislocations, where the product g*F can be only zero
or an integer (since it is the product of a reciprocal lat
tice vector and a real lattice vector). For partial dislo
cations, g*F can take on the non-integer values 1/3, 2/3,
4/3, etc., in cubic lattices. Howie and Whelan (1962) deter
mined that partial dislocations are invisible when g-F=0 or
1/3 and are visible for all other products. Silcock and
Tunstall (1964) further determined that, for this to be
strictly valid, the deviation from the Bragg condition cannot
be too large.
The condition that a defect is "invisible" if its dis
placement field does not distort the reflecting planes can
be applied to identify certain small precipitates. For
example, in the case of 6' platelets in Al-Cu, the principal
misfit in the lattice caused by the platelet is normal to
the plane of the platelet (.Section 2.2). If such platelets
are too small to distinguish their shape, their orientation
can still be determined since they will be invisible when
ever g is perpendicular to the misfit vector.
2.5.3. Imaging Precipitates in the
Electron Microscope
Precipitates can be imaged by one or more of several
mechanisms in the electron microscope. A good description


122
In summary, the configuration has all the characteristics
of a climb source operating on (100) with b=a[100], and a pre
cipitate colony of (100) 6* orientation nucleated by the
source loop.
As in the case of climb sources on (110), no stacking
fault was observed within this loop. Since the stacking of
planes in f.c.c. in the [100] direction is ABAB..., the loop
must climb by the condensation of vacancies onto two adjacent
(100) planes.
The small partial loop at C in Figure 4.36(a) was also
analyzed and found to have F=a/2[10l]. This must be a
secondary climb source of the (1101-type that was nucleated
at a 9' platelet lying in the precipitate colony on (100).
This observation is the only reported case of a climb
source in aluminum alloys operating to produce pure-edge
loops on a cube plane. This is not surprising, however,
since it was the only (100) source recognized as such among
thousands of sources scanned in all these foils.
4.4.6. Nucleation of Preferred 9* Orientations
During Segmented ClimE~
Figures 4.14(b) and 4.17 show corrugated-shaped, pre
cipitate colonies which were nucleated by the segmented climb
of glide dislocations on different crystallographic planes.
Close examination of such colonies revealed that they con
sisted of adjacent bands of precipitates, each band contain
ing only one of the two possible 9' orientations favored to
be nucleated by the climbing dislocation. This "preferred


182
nucleate near the end of the quench when the dislocation
climb rate had dropped to a low value. Thus, these precipi
tates had time to grow large enough to retard the passage of
the climbing dislocation.
5.2. Comparison with Previous Repeated
Nucleation Mechanisms
I i
|
i
The fundamentals of the repeated nucleation mechanism
proposed by Nes (1974) were reviewed in Section 2.4. This
mechanism reportedly accounts for the phenomenon in all
systems in which it has been observed to date, but in all
these cases the transformed phase has had a larger atomic
volume than the matrix. The Nes model will now be compared
with repeated nucleation of 0' in Al-Cu. The fundamentals
of the Nes mechanism are:
(1) Vacancies must be supplied to the transforming
particle in order to reduce the particle/matrix
mismatch.
(2) The subsequent particle growth provides the
driving force for the vacancy-emitting climb
of the dislocation.
(3) The sequence between repeated nucleations is
controlled by balancing the rate of vacancy
consumption by the precipitates with the rate
of vacancy emission by the climbing dislocation.


2
precipitate reactions, either the volume misfit or inter
facial energy, or both, is sufficiently large so that only
heterogeneous nucleation occurs at preferred sites within the
lattice. These sites are lattice defects such as grain
boundaries, dislocations, stacking faults, point defects,
and other particles. Presumably, a portion of the energy
associated with the defect is supplied to help reduce the
activation energy for formation of the critical nucleus,
resulting in a nucleation event which is otherwise energetic
ally unfavorable.
In the case of heterogeneous nucleation at dislocations,
it is the dislocation strain energy in the matrix which helps
overcome the barrier to nucleation. If the misfit strain
caused by the precipitate is large, dislocation-nucleation
may be the only method of decomposition of the supersaturated
solid solution. Kelly and Nicholson (1963) and Nicholson
(1970), have given excellent reviews of the evidence for
nucleation on dislocations in a number of alloy systems. A
well-known example of heterogeneous nucleation on disloca
tions is that of the metastable 0' phase in Al-Cu alloys.
For a precipitate reaction which is dislocation-
nucleated, the following problem arises whenever the initial
dislocation density is low, as is often the case following
quenching. How can the reaction propagate once the available
nucleation sites on dislocations have been saturated, i.e.,
what is the mechanism for propagation of the nucleation?
Currently there are two known mechanisms whereby the reaction


133
The dislocations were generated probably at grain boundaries
during the quench. Normally, subboundaries are observed in
cold-worked metals which have been given recovery anneals.
In the present samples, boundary formation was essentially
completed during the quench, since well-defined boundaries
were observed in foils quenched into liquid nitrogen with no
subsequent aging treatment. Both tilt boundaries and twist
boundaries were observed, although tilt boundaries were more
prevalent.
Figure 4.43(b) shows a junction of three tilt boundaries
in a sample direct-quenched to 220C and aged five minutes.
Clearly, the dislocations have nucleated precipitate colonies
in the process of climb. The climb paths of all dislocations
in a given boundary are in the same direction as indicated
by the positions of the precipitate colonies. It cannot be
determined from such micrographs if the precipitates were
nucleated while the boundaries were forming, or if the bound
aries, once formed, climbed in a cooperative manner and
nucleated the precipitate colonies. In either case, boundary
formation was completed during the quench. Since edge dislo
cations are potential nucleation sites for 0 precipitates
whereas screw dislocations are not, precipitate colonies were
always observed to be associated with tilt boundaries, but
they were not present at twist boundaries.


121
Table 4.2
Summary of Visibility Data for the
Images of Figure 4.37
g=
200
0 20
III
020
220
022
Ill
311
Burgers
Vector
Ppt.
Orient.
Dislocation
Arc
V
I
V
I
V
I
V
V
a[10 0 ]
Precipitate
Colony
V
I
V
I
V
I
V
V
(100)
Beam
Direction
001
001
101
101
111
111
112
103
Note: V = Visible; I = Invisible.
be an arc of a pure-edge loop. Hence, it can expand in the
(100) plane only by climb. The presence of the precipitate
colony on (100) indicates that it did indeed climb in this
plane. As in the case of climb sources on {110} planes, it
is assumed that the precipitate colony was nucleated by the
a[100] dislocation as it climbed on (100). Now a dislocation
with b=a[100] favors nucleation of the (100) 9' orientation
only. From Table 4.2 we see that the precipitates in the
colony are invisible for g=020 and 022 and visible for all
reflections for which h in (hkl) is non-zero. Since the
precipitates are out of contrast only for g-vectors normal
to their misfit, and therefore to their habit plane, the
colony must consist of precipitates having only the (100)
orientation.


164
Figure 4.56. (a) Low and (b) high magnification of micro
structure of sample quenched from 550C into
room temperature water, then up-quenched to
221C and aged 5 minutes. The foil contains
a high density of irregular-shaped loops which
are internally precipitated.


80
the relaxation of the Laue condition along the thin dimension
of the platelets. The diffraction pattern can be analyzed
on the basis of three superimposed patterns shown in Figure
4.20 (a)-(c), where we consider only the lower right-hand
quadrant of the pattern. The pattern in (a) is the (001)
matrix pattern. The pattern in (b) is indexed on the basis
of two 9' orientations parallel to (100) and (010) matrix
planes, using the lattice parameters of 4.04A and 5.8 for
0' (Section 2.2). The remaining reflections in (c) are due
to double diffraction from the matrix {200} and {220} beams.
Double diffraction is a common occurrence in foils containing
precipitates with dimensions much smaller than the foil
thickness. The composite pattern, shown in (d), matches the
experimental pattern. Thus the precipitates are positively
identified as the 0' phase, in agreement with the known fact
that 0' is the only metastable phase which nucleates on dis
locations in Al-Cu.
In the present research, conditions were chosen to in
sure that the 0' phase was the only precipitate phase present
after heat treatment. Its distribution was always associated
with the climbing dislocations.
The appearance of the precipitate colonies at high
magnifications is shown in Figure 4.21. These are typical
colonies nucleated on dislocation climb sources. The colony
in (a) was nucleated on the (101) source habit in the plane
of the foil. In (b) five different {110} habits were gener
ated at the same source particle and have nucleated


61
only intuitive reasoning has been used to show that climb
sources in aluminum alloys operate by vacancy-annihilating
climb.
In the present work, the technique of computer matching
of dislocation images (Head et_ al. 1973) was employed to
establish that these loops climb by vacancy annihilation,
thereby removing planes locally from the lattice. This tech
nique is capable of determining unambiguously the Burgers
vector of a dislocation line segment, i.e., whether the
Burgers vector is +F or -F.
Figure 4.6 shows a climb source in a sample direct-
quenched to 220C and held only four seconds. This source
has generated two loops on a (110} habit inclined through the
foil, so that each loop is sectioned and leaves two arcs of
dislocation. The segment of the outermost loop at A is
reasonably straight and was selected for the computer match
ing experiment (the oscillations in the image are contrast
effects arising from the inclination of the dislocation
through the foil). From invisibility conditions, the Burgers
vectors of these loops were determined to be either a/2[Oil]
or a/2[Ol]. The loops are pure-edge and lie on (Oil). By
stereographic analysis, the line direction of segment A was
determined to be very close to [100] in (Oil) and the foil
normal was determined to be [313].
Six experimental images of segment A are shown in Figure
4.7 along with the corresponding computed images for
F=a/2[011] and F=a/2[Oil]. These six images represent


67
(110) PLANES
OOOOOo A ooOOOOOA
OOOQon 0o000000 qQOOOO b
oooo uooooooooo OOOOOA
OOOO1 oOOOOOOBOOon rOOOO =
00000o Ooo000000n00000
0000 Oo OOOO B
Figure 4.11. Schematic diagram of the cross-section through
a climb source loop on (110). The pure-edge
loops with b=a/2[l0] climb by vacancy conden
sation onto two adjacent (110) planes, thereby
avoiding creation of a stacking fault.