A study of the interdiffusion of aluminum and zirconium

JOURNAL

OF NUCLEAR

A STUDY

MATERZALS

1, 12 (1964)

61-69,

0

OF THE INTERDIFFUSION G. V. KIDSON

NORTH-HOLLAND

OF ALUMINUM

PUBLISHING

CO., AMSTERDAM

AND ZIRCONIUM

and G. D. MILLER

Chalk River Nuclear Laboratories, Atomic Energy of Caru&a Limited, Chalk River, Ontario, Canaa’.u Received

A study conium

of the interdiffusion has been made

temperature.

Only

ZrAls, was detected according

one

of aluminum

as a function

and zir-

of time

intermediate

1 November

alloy

W(T, t) <

2.22

suivant

and

in the diffusion zone, and grew

x 100 exp -

The diffusion mechanism

46 000 RT

la relation empirique

W(T, t) <

phase,

to the empirical relation

1963

cm/h+.

in ZrAls is discussed

2.22 x 108 exp -

On discute du mecanisme en fonction

1t*

[

Es

wurde

suivante 46 000 RT

in

1t*

cm/h*.

de diffusion dans ZrAla

de sa structure. eine

Untersuchung

iiber

die

diffusion von Al and Zr als Funktion

terms of its structure.

:

Temperatur

unternommen.

Legierungszwischenphase,

Gebildet

Zwischen-

der Zeit und

wurde nur eine

ZrAls, welche in der Diffu-

sionszone auftritt und nach der empirischen Beziehung w&hat Une

Etude de l’h&&odiffusion

r&C&e

en for&ion

du temps

de Al

et Zr a BtB

:

W(T, t) <

2.22 x 109 exp -

et de la tempdrature. Der Diffusionsmechanismus

11 se forme seulement une phase intermddiaire ZrAls,

46 000 RT

1t*

cm/h*.

in ZrAls wird an Hand

dans la zone de diffusion, cette phase se d&eloppant

seiner Struktur

1.

For heavy water moderated, organic-cooled reactors there is a requirement for pressure

metal I), it would appear to be particularly suitable for this purpose. Consideration must be given, however, to the long term compatibility

tubes that have a low neutron capture cross section, good corrosion resistance and acceptable mechanical properties at operating temper-

of the two metals when in intimate contact at elevated temperatures. Under these conditions the two metals will interdiffuse, one into the

atures of about 400’ C. A zirconium-rich alloy, zircaloy-2, fulfils the first two requirements well and the third with reservations. While the high temperature properties are initially satisfactory, they subsequently deteriorate due to embrittlement caused by an uptake of hydrogen from the coolant. It has been suggested that one means of overcoming this difficulty would be to block the passage of hydrogen from the coolant to the Zircaloy by providing a protective coating. Since aluminum has a low neutron capture cross section, is cheap, easily fabricated and has the lowest hydrogen permeability of any

other and by so doing, produce in the diffusion zone thin layers of zirconium-aluminum alloys. According to the published equilibrium diagram 2) (see fig. I), a total of nine intermediate phases occur in the Al-Zr system. If the diffusion rates were very high, it is possible that all of the aluminum would be transformed into one or more of these phases, with the resultant loss of protection from hydrogen permeation to the zircaloy. Moreover the mechanical properties of such alloy phases may be such that repeated thermal cycling would produce cracks, again reducing the degree of protection afforded. The purpose of the present investigation was

Introduction

61

diskutiert.

Q.

62 to

gather

the

necessary

V.

data

KIDSON

on the

AND

inter-

G.

D.

MILLER

to certain aspects

of their phase diagrams

in

diffusion behaviour of aluminum and pure zirconium in order to provide a means of extra-

section 2 below. In section 3 the experimental methods used and results observed are pre-

polation to times and temperatures

sented, and the latter are discussed in section 4.

operating

conditions,

of practical

as well as to note

the

The problem

mechanical properties of the alloy layers formed. It was hoped, also, that such a study would have some intrinsic value in the more fundamental aspects of diffusion in systems form a series of intermediate phases. By adopting

a simple diffusion

general features

of the growth

by the interdiffusion

2.

which

model,

of the fabrication

Diffusion

Model

p

of alloy layers

molar concentration

of component Al -3

WEIGHT 1900--1

’

1,’

6

6

IO

‘,

’

’ ,

1652O h \\ -I \ \ \ \ \

1900

15 8

PER

CENT

20 1

ALUMINUM 30

40 1 I

I

I

I

hlil i

w-IN*) i

50 I

60 II,,

70

I

N

60 90

I

tcl d

;:

~~zid<

50

59’ ,OO

,I’

\ \,’

I4800

’ \ ,I E?l

‘\

00 . \ \

\ \

73.5

49

\ \

1400

\

-

\ \

Y w

\

1300-

-

P

\

!50~

\

f

\ 1200

-

I

I \ I

w c II00

I

I I

1000

97 lO(3~~

9400 1113.5)

900 66p

a

000

700

1.2 (0.35) 660.5’ 6600 i

(ALb

ZR

Fig. 1.

ATOMIC

in this

Consider the interface between two phases and y, separated by a miscibility gap SC,,= CBv-Crs (see fig. 2). Here C refers to the

the

of two metals are related

I234

of composite

pressure tubes has not been considered paper.

PER

CENT

L 70 ALUMINUM

I

J 90

The equilibrium diagram for the aluminum-zirconium

90

100 AL

system.

A. Let the

A

STUDY

THE

OF

INTERDIFFUSION

ANNEALING

OF

TEMPERATURE

ALUMINUM

AND

63

ZIRCONIUM

of time and temperature

is

-A,,(T)1 t*

W&T, t) = V,(T)

= B, t*.

(3)

This type of description to any number

can be generalized

of intermediate

phases

in a

simple way 3). In general it is assumed that an observed parabolic relationship between the width -OlSTANCE-

Fig.

2.

diagram, diffusion

-TEMPERATURE-

The

between

relationship

the diffusion zone

annealing

interface phase

the

equilibrium

temperature

concentration

for

and the a

three

system.

diffusion coefficient in the B phase be D, and that in the y phase be D,. In a time dt, the amount of component A arriving at the interface EB,,from the /3 phase is JB = - D,(i%‘/b~)~~~. Similarly, the amount leaving the interface in the y phase is J,,= - D,(~C/~)X)~,,~, where, for example, (bC/b~)+~ refers to the concentration gradient evaluated on the /l phase side of the interface. If JB =_Jy there will be an accumulation of component A in excess of the equilibrium concentration in the y phase, and new t!3phase material will be formed. If dt, is the increment in p phase during dt, then clearly

[Q(WW,,,,

- D,(WWQ,,I.

(1)

Some developments of this relation have been discussed more fully previously 3). In particular it can be shown that the time dependence of the interface position is just &7#) = A&V)

tt

zone and the time of

Experimental order

to

Procedure assess

the

compatibility

of

aluminum and zircaloy-2, it was important to ensure that the experimentally measured rate of growth of the diffusion layers would represent

That is d&,/dt = WC’,,

3.

In

F&y-C,1 dt,= [JrJyl dt

of the diffusion

annealing is evidence that the process is governed by volume diffusion. Conversely, a departure from the parabolic behaviour indicates the operation of some effect not taken into account in the simple description outlined above (for example, the presence of inhibiting oxide films, the rate of nucleation of the new phases, or failure to maintain a constant area of surface contact etc.). Moreover, while the equilibrium diagram suggests that all phases will be formed in the diffusion zone, the kinetics of nucleation and growth of some phases may be so slow as to preclude their formation to a detectable degree during normal times of observation. As it turns out, the Al-Zr system provides an example in which effects neglected in the simple model appear to be operative. These will be discussed in section 4.

(2)

where A,(T) is a rate constant independent of time, but involving the miscibility gap, the solubility range and the diffusion coefficients concerned. An analogous consideration at the a@ interface gives an expression for &.,(t) and hence the width of the /3 phase as a function

the maximum

to be encountered

in practice.

A primary aim, therefore, was to obtain intimate contact between clean faces of the metals to promote good bonding and unimpeded diffusion across the interface. Preliminary studies using both pure zirconium and zircaloy-2 indicated no significant differences in the results. In subsequent studies only pure zirconium was used since the results would have a wider range of usefulness, and the system could be analysed more simply. The aluminum was cut in the form of & in. diameter discs from 99.999 yO pure & in. thick strip. Zirconium discs, f in. diameter and 4 in. thick were machined from

64

c.

v.

KIDSON

AND

U.

D.

TABLE 1

a reactor grade rod. The surface preparation finally adopted for the aluminum was to electropolish

the discs in a TO parts ethyl

30 parts perchloric density

of

acid solution

1.5 amps/cm2

at

alcohol,

at a current

16-20 V for

10

seconds. The discs were rinsed in distilled water, ethyl

alcohol

mechanically

and dried. polished,

standard zirconium

The

zirconium

followed

was

by a dip in a

chemical polishing solution

of 45 parts HNOs, 50 parts HsO and 5 parts HP, rinsed in water and alcohol and dried. were Composite couples of Al-Zr-Al-Zr-Al stacked in a clamp as shown in fig. 3. The clamp N. STL.

HEMISPHERE

ANTALUM

THERMOCOUPLE

Fig. 3.

FOIL

WELL

Annealing assembly for zirconium-aluminum diffusion couples.

was designed so that the couples could be compressed to a reproducible degree in a metallurgical press before tightening the restraining nut. A chromel-alumel thermocouple was inserted into a hole drilled in the bottom of the clamp, and the whole assembly placed in a vacuum furnace tube. The system was evacuated to about 5 x 10-s mm of mercury, and the clamp then pushed into the pre-heated hot zone of the furnace tube by means of the attached thermocouple tube. Temperatures were measured frequently during the course of an anneal, and were controlled to i 2’ C by a HoneywellBrow-n Pyrovane controller. A series of anneals at varying times were made at a number of temperatures to evaluate the rate constants in eq. (3). Times and temperatures used are listed in table 1. Following the heating period the specimens were withdrawn, cooled, removed from the clamp and mounted in a cold setting plastic. They were sectioned, ground and polished on a plane parallel to the diffusion direction.

MILLER

Annealing

times and temperatures

“C

Hours 1

640

2

629

1t (2)

2 (3)

16

17fr (3)

600

2 (3)

4 (5)

30

70

578

3

7

24

47

553

24

30

72

144

Note

:

Numbers

in parentheses

88

denote

number

of

runs made at those times and temperatures.

A cross section of a composite couple annealed at 553” C for 30 hours is shown in fig. 4 at a magnification of about 30 x . As can be seen, these couples provided four interfaces under identical conditions of time and temperature. In general there was considerable variation in the widths of the diffusion layers from one interface to another and even along the length of a given interface. Consequently a policy was adopted of recording the maximum width observed and of repolishing a specimen a number of times and checking the diffusion zone. In some cases as many as four or five couples were annealed before acceptable diffusion layers were obtained. The widths of the diffusion zones were measured to an accuracy of 1O-3 cm on a Vickers’ projection microscope. Some detailed features of the diffusion bands are shown in fig.

5 on

a couple

annealed

at

629” C for

179 hours.

3.1.

OBSERVATIONS

The most striking and significant observation was that only one of the nine intermediate alloy phases in the Zr-Al system grew to a detecable width in the diffusion zone, and it grew very rapidly. The phase was unambiguously identified as ZrAla both by Debye Scherrer X-ray powder patterns and also by electron probe microanalyser scans +. As a further check on the t

The latter were provided

through the courtesy

of Associated Electrical Industries Type SEM, 2 Micro Analyser.

Ltd.,

using

a

A

STUDY

OF

THE

INTERDIFFUSION

OF

ALUMINUM

AND

ZIRCONIUM

65

66

G.

V.

KIDSON

possibility

of other phases, the ZrAla-Zr

face

examined

was

by

electron

AND

G.

by

inter-

D.

weighing

during

microscopy,

MILLER

the results of

the

best

the initial stages of growth,

a parabolic

dependence,

couples assuming

and extrapolating

the

using a two-stage carbon replica and a magnification of 5000 x . No evidence of additional

log10 (width)

layers was seen. The second observation

t = 1 h. Values of B(T) were thus determined for five temperatures ranging from 550” C to

was that in spite of

the considerable care taken, the growth of the diffusion zones did not, in general, follow a simple parabolic

time dependence.

In fig. 6 a

plot of log10 (width) versus loglo (time) is given for couples annealed at 629 f 3” C. It can be seen that while the initial growth rate is as expected, it subsequently falls off and in general the results become quite erratic for the longer annealing times. Referring again to fig. 5, which shows most of the macroscopic features of the diffusion zones, both a string of lenticular voids at the ZrAls-Al interface and a series of small round voids in the interior of the band near the Al side can be seen. The lenticular voids led to a very weak bond at the ZrAla-Al interface and care had to be exercised during the handling of the couples to avoid breaking them. The ZrAls-Zr bond, however, was very strong. The large transverse crack across the alloy phase in fig. 5 was also typical. It is believed to be due to thermal stresses generated during the cooling of the specimen. 3.2

RESULTS

640” C. While the analytical form of the temperature dependence

for each annealing

of the rate constants

B(T) is

known to be rather complex, it is nevertheless frequently found that the observed results can be fitted reasonably well by a simple Arrhenius plot of log B(T) versus l/T “K (see fig. 7). Treating the present results in this way it is found that the time, temperature dependence of the diffusion zone width can be adequately represented by

WV’, t)< 2.22 x 109 exp [ - 46 OOO/RT]t*cm/h* I

-

20

0010

An upper limit for the rate constant was obtained

versus log10 (time) plot back to

(4)

TPC)

-

9-

B(T)

8-

temperature

76-

‘:0 ; 100 B ;z 5 70 ; 60 G 50 5-

5-

40 Y z

30

% i3 ,’

20

2_

k -e 4 10 1

2

3 LOG,,

Fig. 6.

4

5678910

(ANNEALING

TIME I

20 IN

HOURS

30

40

x,

-

Logarithmic plot of growth of diffusion band for specimens annealed at 629” C.

Fig.

7.

Log10 (rate constant) versus _. anneahng temperature.

reciprocal

of

A

4.

STUDY

OF

THE

INTERDIFFUSION

OF

would

Discussion The expression

layer

for the width of the diffusion

as a function

of time and temperature

can be used to extrapolate

the experimental

ALUMINUM

AND

never approach

experiments,

67

ZIRCONIUM

those formed

in these

a very thin layer may not suffer

in this way. Clearly there is a need for tests of the Al-ZrAls

bond under thermal cycling

con-

data to the anticipated operating time of about fifteen years and temperature of about 400” C.

ditions. This, could be done by first annealing at an elevated temperature to produce the

Inserting

equivalent diffusion layer thickness, slow cooling

these

values

in

eq.

(4)

W(400°C, 15 yrs) = 1.02 x 1O-3 cm=0.4

we

find

x lo-ain.

Thus if an aluminum coating of about 2 x 10-a in. thick was put on the zircaloy pressure tube, less than one quarter of it would be consumed by the diffusion layer. It must be borne in mind, moreover, that during the course of these measurements, considerable difficulty was experienced in achieving conditions for the growth of any diffusion layer at all. The present results, therefore, are conservative in so far as they represent the maximum growth. In this regard it should be noted that one of the early attempts to produce diffusion couples consisted of hot extruding aluminum on to the surface of clean zirconium +. The aluminumzirconium bonds thus produced were very strong as shown by standard “stud tests” but on subsequent annealing the growth of a diffusion zone was very slow. If the surface of the zirconium (or zircaloy-2) was anodized before extruding the aluminum, the diffusion rates were even slower, although the bonding was again very strong. Finally, of course, it must be recalled that the widths of the diffusion layers included penetration into the zirconium as well as the aluminum. It seems safe to conclude,

therefore,

that the

use of a thin layer of aluminum as a protective coating on a zircaloy pressure tube is reasonable for the assumed operating conditions provided one takes the criterion of acceptability entirely in terms of the rate of consumption of aluminum. The point of concern that remains is the fragility of the Al-ZrAla bond. It is possible that the aluminum could flake off under the action of thermal cycling once a diffusion layer had been formed. Since, however, the width of the zone t

This work was done with

R. D. Watson, Branch, AECL.

Applied

the cooperation

Engineering

of

Development

to 400” C and then cycling. The remainder of this discussion will be concerned with an interpretation of the observations in terms of the properties of the Al-Zr system. The specific points of interest are: (i)

the departure from parabolic growth rates

(ii)

the formation of voids in the diffusion zone

(iii)

the unique occurrence rapid growth

of ZrAls

and its

(iv)

some aspects of the microanalyser

results.

It seems probable that the observed departure from parabolic behaviour arose from the formation of voids in the ZrAls phase, particularly from the lenticular voids at the alloy-Al interface. If, as will be discussed later, the voids were formed as a result of the intrinsic diffusion mechanism, then they would increase in size and number as diffusion proceeded, eventually reducing the area and bond at the ZrAls-Al interface and inhibiting further growth. It is well known of course that voids can arise as a result of the Kirkendall effect in diffusion couples 4). Reasons

as to

why

it should

be

particularly large in ZrAla will be suggested later in terms of the structure of this alloy phase. The sequence of events leading to the formation of an intermediate alloy phase in the diffusion zone involves initially the saturation of the primary solid solutions with solute atoms, the subsequent nucleation of the intermediate phase and finally, the growth of that phase. Since the solubility limit of Zr in Al is only about 0.07 at %, whereas that of Al in Zr is 1.2 at “/” at around 600” C, one would expect the Al primary solid solution to be saturated first. The ease of nucleation of the ZrAls phase next to the Al can be understood in terms of the structures of the two lattices.

68

G. v. KIDSON

AND

The aluminum, of course, has the simple FCC structure, with a near neighbor distance

G.

D.

MILLER

that the diffusion coefficient of the atoms must

of 2.86 8. ZrAls has an ordered body centered

have been of the order of 10-s cms/sec to produce a zone of that size +. On the other hand

tetragonal

an empirical

type Doss space lattice 5). It is not

correlation

between melting point

difficult to show, however, that the arrangement

and diffusion coefficient would suggest that the

of the atoms

latter should be around lo-15 cms/sec at 629” C

can be considered

as a slightly

distorted modification of a simple FCC structure, as indicated in fig. 8. The interatomic spacings

in systems

see, therefore,

that the diffusion rates in ZrAla

of the Al-Al

are unusually

high ; so high in fact, that the

atoms are 2.80 A and 2.85 A, close

to that of the pure aluminum phase. Evidently, once the Zr atoms enter the Al matrix and occupy an ordered arrangement of sites, the subsequent formation of ZrAls will not result in any significant misfit energy.

0 Aluminum &= 2.808, 0 Zirconiumdr = 2.85k, Fig. 8.

Structure of ZrAla lattice showing its similarity

to the simple FCC aluminum Note:

The complete

lattice.

cell contains the layer sequence

ABAB’AB’ABA.

The growth of ZrAls to the apparent exclusion of all others must be due primarily to relatively large diffusion coefiioients of Zr and Al in that phase, This can be seen from the fact that while the rate constant for each phase is a function of the diffusion coefficients, the miscibility gaps and the solutility ranges of all the phases present, the latter two parameters do not differ greatly from one phase to another. A rough measure of just how large the diffusion coefficients in ZrAls are can be obtained as follows. As indicated in fig. 6, the diffusion layer grew to a width of 5 x 10-s cm after annealing 10 hours at 629” C. It is not difficult to show

which melt at about

1550’ C. We

rapid transport of excess Al to the Zr side of the phase effectively swamps the formation and growth of any of the expected intervening phases. This rapidity of diffusion is made even more surprising by the very small concentration gradient available as a driving force across the ZrAls phase, which is reported as having an ordered stoichiometric composition 5). These considerations, and others to be discussed suggest that the concentration gradient exists through the formation of a defect lattice. In particular, let us assume that in moving from the Al rich side towards the Zr rich side of the ZrAls, there is a progressive increase in the number of unoccupied Al sites in the Al sublattice. Since the diffusion coefficient is proportional to the product of the vacancy concentration C, and the average vacancy jump frequency TV, it is apparent that a defect lattice would account for the unusually large diffusion rates observed. An examination of fig. 8 shows that each Al atom

has eight

Al and four

Zr sites in its

coordination shell, whereas each Zr atom is completely surrounded by Al sites. As a consequence, a vacant Al site common to both Al and Zr atoms is more likely to exchange with the Al atom than the Zr atom, since the latter would have to overcome the ordering force of the lattice in addition to the normal barrier energies in occupying an Al site. It follows that rym i Yv’,A’, whence Dzr< DAM. That this may be a rather large effect is evident from the fact that the structure remains ordered up to the melting point and hence must have an ordering t

If we assume the mean square displacement

to be about i$W, then D=x2/2t 3.6 x 104 w lo-* cm2/sec.

N [5x

lo-2/2]2/2

_T%? x

A STUDY OF THE INTERDIFFUSION

OF ALUMINUM

69

AND ZIRCONIUM

CONCLUSIONS

4.1.

The use of a thin layer of aluminum to prevent the uptake of hydrogen by a zircaloy-2 pressure tube is not jeopardized by the rate of consumption of the aluminum through the formation of diffusion layers. On the basis of the present study a maximum

of 0.4 x 10-S in.

of ZrA13 would grow in 15 years at 400” C, and

DFFUSION

I-

ZONEI

DIFFUSION

I

1

(4 Fig.

9.

Electron

ZONE1

(b) beam micro-analyser

scans across

diffusion zone. a) Zirconium.

b) Aluminum.

energy of about 0.6 eV. A significant difference in flow rates for the two species would result in a large Kirkendall effect, and hence account for the formation of voids on the Al rich side of the ZrAls phase. The assumption of a defect lattice is thus consistent with the observed behaviour of the system. It is also consistent with the results of the electron probe microanalyser scan shown in fig. 9. It will be noted that within the limits of accuracy of the data, there is no concentration gradient of Zr atoms/unit volume across the phase, whereas there appears to be a distinct decrease in Al atoms towards the Zr rich side. While the results of such scans are not in themselves conclusive evidence of a defect structure they do, in the present case, substantiate the intermetation niven above.

this could probably be substantially reduced by anodizing the zircaloy surface prior to extruding the aluminum. Further work should be done to test the mechanical properties of the aluminum-ZrAls interface for very thin diffusion layers and under thermal cycling conditions. The formation and growth of ZrAl3 in the absence of all of the other intermediate phases can be understood in terms of the structure of ZrAl3. The occurrence of voids in the aluminum rich region of the diffusion zone is probably a manifestation of a Kirkendall effect, and is an intrinsic result of the diffusion mechanism. The assumption of a defect Al sublattice in ZrAl3 is consistent with all of the observations. Acknowledgement The authors are indebted to Mr. Jacques I’Hereux for performing a considerable amount of careful preliminary work in this study. References It has been suggested that the thin oxide film on the

aluminum

is the

real

barrier

permeation.

See A.

Sawetzky

Chalk River

Report

No.

also C. J. Smithells, Sot.

(London)

to

and M.

AECL

J.

Rees:

1252, 1960, and

C. E. Ramsley,

A 152 (1935)

hydrogen

Proc. Roy.

706

M. Hansen, Constitution of Binary Alloy (McGrawHill,

1958) p. 153

G. Kidson, See

for

Phys.

J. Nucl.

example:

gemon

3 (1961)

D.

21

LeCls,ire,

Prog.

Met.

1

See for example: Lattice

Mst. A.

Spacings Press,

W.

B.

Pearson,

and Structures

1958) p. 391

Handbook

of Metals

of

(Per-