Precipitation and dislocation nucleation in quench-aged Al-Mg alloys

PRECIPITATION

AND DISLOCATION

NUCLEATION

A. EIKUMt$

IN QUENCH-AGED

Al-Mg

ALLOYS*

and G. THOMAS?

Smallprecipitates have been observed in thin foils prepared from bulk Al-5 % Mg and Al-6.5% Mg alloys after quenching followed by aging at 106°C. The precipitates, concluded to be ~XIintermediate form of Mg,Al,, will apparentiy nucleate only when a critical supersaturation of both Mg atoms and vacancies exists in the specimen. The vacancy denuded zone between colonies of prismatic dislocation loops and grain boundaries is one good example where the critical supersaturation oan occur. A high stress is associated with the precipitates, which can be relaxed by the punching out of rows of prismatic dislocation loops. This has been observed directly in the electron microscope. A comparison of the experimental results with the theory of Bullough and Newman enables the critical shear stress for prismatic glide to be estimated. The value obtained is 3.4 x lo-* C. Some general observations on the nature of the precipitation and the generation of dislocation loops are discussed. PRECIPITATION

ET GENERATION DE DISLOCATIONS DARS LES ALLIAGES Al-Mg VIEILLIS APRES TREMPE Les auteurs ont observe de petits precipites dans des pelliculos minces prepari?asQ partir d’alliages Al-5 % Mg et Al-6,5 % Mg trempes et vieillis 8. 100°C. Les precipites, que les auteurs estiment etre une forme intermediaire de Mg,Al,, semblent ne se former que quand il existe dans l’echantillon une snrsaturation critique Ztla fois on atomes Mg et en Iacunes. Laeone sihueeentre les colonies de boueles de dislocation prismatiques et les joints de grains est un bon exemple dune region oit la sursaturatio~ critique peut se prod&e. Des contraintes elevees sont associees anx procipites, contraintes qui peuvent btre relaxees par la generation de ran&es de boucles de dislocation prismatiques. Ce phenomene a 6th observe directement au microscope Blectronique. En comparant les resultats experimentaux et la theorie de Bullough et Newman, il est possible d’obtenir une estimation de la contrainte tangentielle oritique necessaire an glissement prismatique. La valeur obtenu est 3,4 x 10e4 G. Certaines observations d’ordre general sur la nature de la precipitation et la generation des boucles de dislocat*ionsont discut&es. AUSSCHEIDUNG

UND

VERSETZUNGSKEIMBILDUNG IN ABGESCH~~CKTEN UND ANGELASSENEN Al-Mg-LEGIERUNGEN Kleine Ausscheidungen wurden in diinnen Folien beobachtet, welche aus Legierungen der Zusammensetzung Al--5 ‘A Mg und Al-6.5 oAMg nach Abschrecken und Anlassen bei 100°C hergestellt wurden. Die Ausscheidungen, vermutlich eine Zwischenform von Mg,Al,, werden offenbar nur gebildet, wenn eine kritische Ubersattigung an Magnesiumatomon und Leerstellen in der Probe vorhanden ist. Die Zone ohne loops zwischen Gruppen prismatischer Versetzungsringe und Korngrenzen ist ein gutes Beispiel fiir Gebiete, in denen die kritische ~er~ttig~g auftreten kann. Starke innere Spannnngen sind mit den Ausscheidungen verkniipft, die duroh das AusstoBen von prismatischen Versetzungsringen abgebaut werden konnen. Dieser Vorgang wurde im Elektronenmikroskop direkt beobachtet. Ein Vergleich der experimentellen Ergebnisse mit der Theorie von Bullough und Newman gestattet eine Abschbtzung der kritischen Schubspannung fur prismatisches Gleiten. Das Ergebnis ist 3.4 x 10m4G. Einige allgemeine Beobachtungen iiber die Art der Ausscheidung und die Erzeugung von Versetzungsringen werden diskutiert. 1. I~TROD.UCTIO~

Precipitation of the /3-phase (MgsAl,) in binary Al-Mg alloys is energetically difficult because of the relatively high strain energy associated with the formation of MgsAls. (See e.g. Refs. l-3.) The structure of the j? phase is very complex, as there are ~1200 atoms per unit cell. c4) Although a G.P. zone stage has not yet been positively detected during the aging of A1-Mg alloys, an intermediate ,8’ phase has been observed by transmission electron mioroscopy.(ly2) However from the kinetics of the early stages of aging it has been suggested that clustering of ma~esium atoms occurs prior to precipitiation (Federighi, private communication). In view of the high binding energy between Mg atoms and vacancies(5) the * Received September 9, 1963; revised October 21, 1963. t Department of Mineral Technology and Inorganic Materials Research Division, Lawrence Radiation Labor&tory, Un+iv;osy ZJ California, Berkeley. Department of Metallurgy, Birmingham U&ersity, Eigland. ACTA METALLURGICA,

VOL. 12, MAY 1964

clustering probably involves magnesium atomvacancy complexes.(6) It is expected therefore that precipitation will occur when a critical Mg-vacancy ratio is attained in the complex. Such a ratio will depend upon the heat treatment given to a particular AI-Mg alloy. Preliminary investig~tions(?’ involving quench-age treatments showed that rapid quenching and aging at 100°C induced homogeneous precipitation in localized regions of an Al-5 % Mg alloy. Furthermore, punching of dislocation loops from the precipitates was observed to occur whilst under examination in the electron microscope, indicating that the precipitates generate large stresses at the precipitate-matrix interface. In other cases dislocation climb sources are formed. The present investigation was undertaken to study these effects in more detail.

537

2. EXPERIMENTAL Specimens of the Al-5 % Mg and Al-6.5 % Mg alloys,

0.005 in. thick, were annealed no less than 12 hr above

538

ACTA

METALLURGICA,

500°C, raised to and quenched from 55O’C into brine solution at approximately -10°C. The specimens were subsequently aged for 100 hr at 100°C. Thin foils were then prepared from the aged bulk specimens by electro-polishing in a 20% perchloric acid-ethyl alcohol solution using the standard window method. All the foils were examined in a Siemens Elmiskop lb electron microscope operated at 100 kV. 3. RESULTS

(1) The quench-aged substructure The dislocation substructure in quenched and aged Al-6.5 wt. % Mg is similar to that previously observed in Al-5 wt. % Mg.(5f7*8f The dislocation substructure within a single grain can vary appreciably over short distances. This variation is greater than that observed over the entire 2-in. specimen length. Figure l(a) shows a typical region containing large helices and loops. The density of very small defects seen in the background, e.g. at A, is less near the larger, more stable dislocations, as one would expect if these were clusters of vacancies, small loops or precipitates. However, in Fig. l(b), it can be seen that prismatic punching has occurred from these small

VOL.

12, 1904

defects which shows that they are precipitates and not dislocation loops. These precipitates were not present in specimens which had been quenched and aged at room temperature. This indicates that they are not impurities undissolved after the homogenization treatment as had been observed by Embury and Nicholson.(9J The precipitates are thus assumed TV be essentially Mg,A.& (or an intermediate phase). They gave rise to no anomalous effects in the electron diffraction patterns (e.g. streaks or extra reflections) so that no structural determinations were possible. The contrast from the small precipitates (e.g. at S, Kg. 2) is similar to that observed from cobalt particles in Cu-Co aIloys(loJ1) and is sensitive to the diffracting conditions. The contrast is thus typical of that due to coherency strains. From the nature of the contrast and from the symmetry with which (110) loops can be generated from a given precipitate, it is ~oneluded that the precipitates are small spheres. The following point also indicates that the small defects are not dislocation loops. In order to maintain a loop of radius r = 100 d at IOO’C, the vacancy supersaturation c/c0 must be -10. One would not

(b) FIG. 1. Struoture typical of Al-6.5 wt. % Mg quenched from 55O’C into brine and aged 100 hr at 100°C; (a) helical dislocations and loops witahsmall precipitates, e.g. at A and, (b) precipitates which have generated dislocations by prismatic punching.

EIKUM

AND

THOMAS:

PRECIPITATION

AND

DISLOCATION

NUCLEATION

539

(W

(cl

FIG. 2. A punching sequence observed directly in the microscope showing; (a) precipitate at S exhibiting coherency strain contrast, (b) a single loop punched out at S, (c) a second larger loop and another row at S and a row of five loops at T. Foil prepared from Al-6.5 wt% Mg after quenching from 550°C into iced brine, and aging 100 hr at 100°C.

ACTA

540

METALLURGICA,

expect this value of excess quenched-in an anneal

vacancies after

of 100 hr at lOO”C, unless the precipitate

is supplying formation

vacancies,

for example,

during a trans-

in appearance to those observed for climb sources.(12g13) A double loop climb source, with the inner attached loop is shown in Fig. 4(a), taken from a bulk specimen temperature.

Climb sources

punching

of loops,

to be of the interstitial

which

are expected

type in view of the Mg,Al,

was observed

foils in the microscope.

during

examination

A typical

sequence

of the is shown

the innermost

attached

The stability

of attached

loop is then usually missing.

be discussed

later.

loops in climb sources will

The defect

which has generated

climb sources in Fig. 4 is similar in appearance

in Fig. 2. The loops are always generated along (110)

larger than those which generate

glide cylinders

thin foils.

and are only visible when the foil is

oriented to give diffraction glide

along

succeeding

the cylinder

contrast.

loops, they must be of the a/2(110)

S along

emitted

in Fig. 2(c).

[lOi]

and a second In addition

has been generated

along [Oil].

the

of measurement,

same

emitted

size

from

kind.

out from the one has been

a new row of loops A uniform

row has also been created at T (Fig. 2(c)). accuracy

5 loop

Within the

all five loops in row T are

(r E 40 A),

whereas

the

first loop

S was smaller than the second.

This

might indicate that some growth of the precipitate or change

in strain field occurs

during

some oases the first loop punched subsequent seen

loops,

In

loops are produced

other

punching.

In

of which

imporbant,

cases uniformly

of the variations with

loops indicates

they nucleate

and grow.

respect

to

those

loops

As pointed

prismatic

out by Bullough

and Newman(14) the final configuration

of a row of

loops is attained when the shear stress acting on each loop due to its neighbors

is equal to the critical shear

stress required to move a loop. derived

the following

Bullough and Newman

expression

critical shear stress assuming

for calculating

161Gv Tc = B&7-(1 The dimensionless spacing

parameter to loop

Y)

v is related to the ratio

diameter,

diameter

in a row, Y is Poisson’s

modulus

and 6 the Burgers

D is the average ratio, G the shear

vector.

Experimental

A decrease in radius with distance traveled

Fig. 5 together

the loops

is usually

are interstitial

and absorb vacancies, an effect which would be more pronounced with small loop radii. This cannot explain ssrved

the random

variation

in this investigation.

clude per se that vacancy loops along

always the

accounts

loop

row,

in sizes of loops Thus one cannot

absorption

for a decrease although

all

expected to be of the interstitial investigated here. The complex,

structure

adjacent

Fig. 3(b),

to

as many

by interstitial in loop the

loops

with the theoretical

radii are

kind in the alloy

in

curve for v =

0.05(14) which gave the best fit to our data.

Taking

v = 0.05 and r = 50 A, then the critical shear stress is found to be 3.4 x 1O-4 G.

obcon-

this

the loops to have con-

stant radii

of loop

in loop radii are

generate

values of z/D, where z = loop spacing, are plotted

to occur because

in

that high stresses are set

which decrease in size with distance from the source. assumed

but

up around the particles due to a change in volume as

sized

but there appears to be no tend-

especially

The fact that the precipitates dislocation

can be

ency for any one kind of effect to predominate. The observations

loops

(3) The critical shear stress for prismatic glide

out is larger than

some examples

in Fig. 3(a).

punched

Since the loops

due to the stress field of

In Fig. 2(b) a loop has been punched particle

are also

found in specimens which have been rapidly quenched and aged above room temperature, Fig. 4(b), but

(2) Prismatic punching

structure,

12, 1964

aged at room

from /3’ to p.

Prismatic

VOL.

4. DISCUSSION (1) Nucleation

of precipitates

The influence of dislocation

substructure

on the dis-

tribution of precipitates in A1-Mg alloys appears to be more complicated than is found in other aluminum alloys.

One would expect the presence of dislocations

precipitates

becomes

to result in the efficient operation

of heterogeneous

dislocations

are gen-

nucleation

the

erated. The contrast is probably due dislocations e.g. at the precipitate-matrix

to tangled interface.

Another observation of interest is the formation of a single dislocation loop that apparently remains attached to the particle, Fig. 3 at U. Only the precipitate is visible in Fig. 3(a), but in Fig. 3(b) a loop is clearly present. The single attached loop is similar

such

as

is found

in

Al-Cu

and

Al-Ag systems.(1-3) In concentrated Al-Mg (9.3% Mg) heterogeneous nucleation has been observud(13) in the form of rod-shaped precipitates on prismatic dislocation loops. However, the precipitates were relatively small and the loops large. This indicates that one must be very careful to choose an aging treatment suitable for the formation of observable

EIKUM

AND THOMAS:

PRECIPITATION

AND

DISLOCATION

NUCLEATION

541

(b)

FIG. 3. A punching sequence where, (a) faint precipitate contrast is visible at U and, (b) a single attached dislocation has been formed at U. M-6.5 wt. % Mg quenched from 550% into iced brine, sged 100 hr at 100°C.

542

ACTA

METALLURGICA,

VOL.

12,

1964

(b) FIG. 4. Climb sources in Al-Mg alloys. (a) TWO loop source containing the inner attached loop. Al-5 wt. % Mg, quenched from 550°C into iced brine and aged ~1 mo. at 20°C. (b) Showing precipitate contrast andmultiple sources in Al-6.5 wt. % Mg, quenched from 550°C into iced brine, aged 100 hr at 100°C.

precipitate while not allowing the dislocations to climb appreciably. From previous climb experiments(6) it is now known that a 2 hr-1OO’C anneal in foils about 2000 A thick is not adequate for perceptible dislocation climb. The grain boundary precipitates shown in Fig. 6 were observed after aging Al-5 ‘A Mg for 50 hr at 150°C. Most of the dislocations had annealed out and precipitation was observed only in the grain boundaries. The formation of the precipitate also gives rise to high local stresses as evidenced by the generation of dislocations, and

-

Is2

rb

P,

P*

Ps

v = 0.05

P&J P,

(Ref.14)

Pa

P,

Pa

Fro. 5. Variation in the ratio of loop spacing (z) to loop diameter (D) as a function of position (pi). The squares and triangle represent data from rows T and V, Figs. 2(c), 3(b), respectively.

the precipitates themselves are not readily seen but rather the contrast due to their strain fields. A possible case of homogeneous nucleation is shown in Fig. 7 where the relationship between dislocations, small angle boundaries and small precipitates is readily observed. The life time for free vacancies is ~10-~ set at 20°C assuming lo8 jumps per vacancy. Thus, after quenching from 550°C into iced brine, the prismatic loops probably nucleated at room temperature and grew until the unassociated excess vacancies were out of solution. This micrograph shows that the free vacancy supersaturation increases with distance away from the boundary (notice how the loop size and density changes from X to Y) and that precipitates form in highest densities about half way along the denuded zone. This is just where one would expect the concentration of vacancy-magnesium complexes to be a maximum. As shown in an earlier paper,c5) the total concentration of vacancies in supersaturated Al-Mg alloys is the sum of free vacancies (which go to form loops) and bound vacancies. In the case of Al-5 % Mg alloy the binding energy was estimated to be No.2 eV so that a large fraction of vacancies are trapped by magnesium

EIKUM

AND

THOMAS:

PRECIPITATION

AND

DISLOCATION

t-l = lo-l5

cm2/sec

is calculated,

than the diffusivity the literature.

543

NUCLEATION

expected

However,

which

this value

when one considers the high vacancy which

is simultaneously

is higher

from data available

in

is reasonable supersaturation

eliminated,

and

suggests

that some kind of G.P. zone stage may occur prior to precipitation.(6) from

There is some evidence

resistivity

experiments

for this

(Federighi,

private

communication). (2) Nucleation of dislocations The nucleation of dislocations is expected

from Mg2Al, particles

to occur in view of the large stresses set

up due to the volume cipitation.

The

changes

criteria

single attached

loop

accompanying

for precipitates

(climb source)

a

or to generate

a

row of prismatic loops are not obvious. ing of interstitial the formation annihilation

Some punch-

loops may occur concurrently

of a climb

small concentration density

pre-

to form

loop,

with

and only

a

of vacancies would be lost during

of the loop.

associated

source

For example,

with foils showing

the particle climb-sources

(see Fig. 4(b)) is of the order of 1Or3 cm-3, and if each particle punched FIQ. 6. Grain boundary precipitate which has punched out loops. Al-5 wt.% Mg quenched from 520°C into water at 78°C and aged ~50 hr at 150%.

atoms.

In this case it is then

precipitation

plex necessary

for precipitation

This is not unexpected ture of MgsAl,.

in a Mg-vacancy

Precipitation

is thus expected

to the Mg atom-vacancy

the quenching

temperature

of permanent

sinks.

The latter

observe

the loop density

why

this type of precipitation but all heat treated

in the same way and conversely these precipitates

explains

in one alloy

treatments.

The

are not observed

fact

after that

in areas where

is high suggests that in these areas

atoms

prefer

to

be

distributed

along

dislocation lines rather than precipitate, in spite of the high supersaturation. This idea is also supported

of ~10~~

could

annihilate all loops, i.e. approximately per interstitial

loop.

supersaturation

Therefore,

completely

1000 vacancies

a sufficient vacancy

would still be available for generation

of the climb source.

The vacancy

concentration

cal-

culated from the sources in Fig. 4(b) is approximately 10-b.

Had

at D in Fig. 3(b)

the dislocation

formed

when

the

matrix

vacancy

supersaturation,

still

contained

a climb

source

been

a large may have

operated. (3) Vacancy loop stability A brief discussion

cipitation occurs. From the width of the precipitation free zone in Fig. 8 (a, N 0.4 p) a value for the diffusivity D w $Apz

of the vacancy

supersaturation

required to operate a climb source is prompted following

observation.

source is sometimes slowly quenched

The attached

The vacancy tain positive

present when a sample has been

with no additional

curvature

aging, but never

necessary

of a dislocation

and aged to main-

with radius of

r is.d5) ln

5= CO

Taking

supersaturation

climb

by the

inner loop of a

after a specimen has been rapidly quenched at 100°C or more.

by the result that no heterogeneous precipitation has been observed under conditions where matrix pre-

7

out 10 loops, each 40 A in radius, a

concentration

to be

ratio, i.e. to

is also dependent This

in alloys of different compositions quench-aging

struc-

as well as to the density

the Mg concentration.(s)

we do not always

com-

of the small spheres.

in view of the complex

very sensitive

magnesium

of large

,u). There thus appears to be some critical of vacancies

different

for

the large size of the loop free

concentration

upon

favorable

to occur rather than nucleation

loops as this explains zone (“3

more

vacancy

Gb4 pql

_

y)47Tr

IInr/b+ 2.51.

G = 2.6 x 10n dyn/cm2,

b = 2.86 A

and

Y = 0.33, the supersaturation required to maintain loops with several values of r at 100°C and 160°C

ACTA

644

METALLURGICA,

VOL.

12, 1964

?&a. 7. As Fig. 2 showing small angle boundaries, prismatic loops and pr~ipit~tes. Note precipitates in the loop-free zone and the variation in loop size between X and Y.

were calculated, and the results are shown in Table 1. The inner loop radius in Fig. 4(a) is ~300 A which, from Table 1, would require c/co N 2, and as previously mentioned a small amount of su~r~turation TABLE

-

1. Vacancy supersaturation to maintain climb

rr A

T,“C

Cl%

50

100 160 100 160 100 160 100 160 100 160

58 10 7 2.4 2.1 1.4 1.3 1.2 1.2

100 300 1000 2000

should be present after only room temperature aging, As more loops are generated by the source, the vacancy ooncentration within the small volume of matrix surround~g the preeipita~ should steadily diminish. Thus, during extended aging at 100°C the innermost unattached loop (r N 2000 A) becomes much more stable than the attached loop (r N 300 A or even less), and a vacancy flux from inside to outside should

be set up. Under these conditions the inside loop should shrink and rejoin the particle-matrix interface. 5. CONCLUSIONS

(1) Homogeneous

precipitation in Al-Mg alloys containing 5 and 6g wt. % Mg appears possible when a critical supersaturation of both Mg atoms and vacancies is obtained, e.g. in the zone between colonies of prismatic dislocation loops and grain boundaries. (2) Precipitation is inhibited when the prismatic loop density is high (high free vacancy supersaturation) when the Mg atoms apparently redistribute along the dislocations probably forming atmospheres. (3) Stresses associated with the formation of precipitates can be relaxed by prismatic punching of dislocation loops or by the operation of a climb source. The former predominates when the vaoancy supersaturation is ineu&ient to ann~ilate the loops, and the climb sources operate when the vacancy supersaturation is relatively high. (4) The critical shear stress, T, for glide of loops produced by prismatic punching was estimated

EIKUM

AND

THOMAS:

PRECIPITATION

from the Bullough and Newman theory, and wa8 found to be 3.4 x 1O-4 G. REFERENCES 1. G. THOMAS, Electron Microscopy and Strew#h of ClrJstals, edited by G. THOMAS and J. WASHBURN, p. 793. Interscience, New York (1963). 2. R. B. NICHOLSON, ibid. p. 861. 3. A. KELLY and R. B. NICHOLSON, Precipitation Hardening, Progrees in Materiab Science 10, p. 149. Macmillan, New York (1963). 4. M. HANSEN, Constitution of Binary Alloys, 2nd edition. McGraw-Hill, New York (1958). 5. A. EI~UM and G. THOMAS, Proc. Int. Conf. on Cry&. Lattice Defects, Kyoto, Japan (1962); J. Phys. Sot. Japan 18, Suppl. III, 98 (1963).

AND

DISLOCATION

NUCLEATION

545

t. T. FEDERI~EII and G. THOMAS, PhiE. Msg. 7, 127 (1962).

8. 9. 10. 11. 12. 13. 14. 15.

A. EIKUM, M.S. Thesis, University of California at Berkeley (i962), Lawrence Radiation” Laboratory Report UCRL-10879. G. THOMAS;P~~Z. Mug. 4, 1213 (1959). J. D. EMBURY and R. B. NICHOLSON, J. Inst. Met. 91. 119 (1962). V. A. PHILIPS and J. D. LIVINQSTON, Phil. Mug. 7, 969 (1962). M. F. ASHBY and L. M. BROWN, Phil. Mag. 8,1083 (1963). K. H. WESTMACOTT,R. S. BARNES and R. E. SMALLMAN, Phil. Nag. 7, 1585 (1962). J. D. EMBURY and R. B. NICHOLSON, Acta 1Met. 11, 347 (19631. k. B&LO~C+H and R. C. NEWMAN, Phil. Mag. 5,921(1960). A. EIRUM and G. THOMAS, J. Appl. Phya. 54, 3363 __ (1963).