Aging Effects in Copper-Based Shape Memory Alloys N.F. KENNON, D.P. DUNNE, and L. MIDDLETON Aging of three copper-based shape memory alloys was studied by measuring the time dependence of hardness, martensitic transformation temperatures, lattice parameters, and shape memory capability at temperatures in the range 200 to 450 ~ The ultimate loss of the shape memory effect in each alloy was preceded by changes in the other properties which resulted from thermally activated processes having activation energies in the range 60 to 80 kJ mol -~. At temperatures above about 300 ~ the aging process involved the eventual formation of a and 3/2 phases. Although the activation energy appears to be insensitive to temperature and alloy composition, at lower temperatures other thermally activated processes, such as change in the type or degree of order, may, at least in the initial stages, be significant aging phenomena.

I.

Table I.

INTRODUCTION

THE shape memory effect occurs in alloys having a metastable structure of martensite, retained parent phase, or a mixture of the two. In each case the metastable phase or phases will transform by diffusional processes at a temperature dependent rate to a more stable structure with concomitant degradation of the shape memory capacity. For applications of shape memory alloys that involve thermal cycling, it is important that transformation during thermal excursions does not reduce the capacity of a device to function satisfactorily during service. Alternatively, the gradual cumulative degradation which occurs inevitably during cycling is an important factor in determining the life expectancy of a device. This kind of limitation on the exploitation of the shape memory effect seems to have escaped detailed scrutiny, although some work on aging effects in copper-based shape memory and similar alloys has been reported. Some of these studies ~-7 detailed structural changes that occurred during thermal treatment, while others 8-13 described some accompanying changes in properties. Of the papers concerned specifically with shape memory alloys, 2'6'8'1~ the changes in properties related to the degradation of the shape memory effect have received very little attention. Consequently, the purpose of the present paper is to provide an account of the influence of aging on the behavior and shape memory capacity of three copperbased alloys. A companion paper describing the detailed microstructural changes resulting from the aging treatments will be published later.

II.

EXPERIMENTAL PROCEDURES AND RESULTS

Three copper-based alloys with nominal compositions given in Table I were prepared by induction melting under a protective cover of graphite powder. Small coarse grained specimens of the alloys were produced by homogenizing in the/3 phase field at 900 ~ and water quenching. The specimens were then aged for times

N.F. KENNON, D.P. DUNNE, and L. MIDDLETON are all with The University of Wollongong. Department of Metallurgy, P. O. Box 1144, Wollongong, N. S. W, 2500, Australia. Manuscript submitted June 5, 1981. METALLURGICAL TRANSACTIONS A

9

Alloy 1 2 3

Cu 81.8 82.9 72.8

Compositions of Alloys (Wt Pct)

Zn

21.2

AI 15.1 14.2 6.0

Ni 3.1 2.9

M, -11 ~ 146 ~ 62.5 ~

Martensite y', ( 2 H ) /31 (18R) 131( 9 R )

up to approximately 10 6 seconds at temperatures between 200 ~ and 450 ~ Changes (at ambient temperature) in the ordered bcc/31 phase were studied using alloy 1 ( M , = - l l ~ while changes in the faulted/31 martensite were studied in alloys 2 and 3 (Ms, Mr=ambient). The following measurements were made as a function of aging time at the various aging temperatures: (i) Vickers hardness number (20 kg load) of the 131 phase (alloy 1, Figure l(a)) and the /3'1 phase (alloys 2, 3; Figures 2(a), 3(a)), (ii) shape recovery capacity by direct observation of thin specimens aged, then bent 2 3 0 deg before heating above the AI temperature (Figures l(b), 2(b), and 3(b)), (iii) martensitic transformation temperatures by direct observation of relief effects and color changes ~4on the surface of specimens polished after aging and either cooled in liquid nitrogen (alloy 1, Figure l(c)) or heated in a simple microscope heating stage (alloy 3, Figure 3(c)); this measurement was not made for alloy 2, (iv) lattice parameter of the/31 phase in alloy 1 (Figure 1(d)) determined from X-ray diffraction powder patterns obtained from needle shaped specimens in a 114.6 mm diameter camera with CuK~ radiation and corrected for shrinkage and for absorption using the Nelson-Riley function. 15 The aging times required at each temperature to initiate change in the four properties conformed approximately with an Arrhenius relationship, and the activation energies obtained from the relationship are presented in Table II.

III.

DISCUSSION

Deformation at low strains can occur in shape memory alloys in a number of ways. For alloys having a ~ structure (Ms
ISSN 0360-2133/82/0311-0551500.75/0 AMERICAN SOCIETY FOR METALS AND THE METALLURGICAL SOCIETY OF AIME

VOLUME 13A, APRIL 1982--551

0+[

o"

+

i

.

,

ud

+ +

+[

..+,..

200 "

~

1

2

3

t,

LOG

5

4

9

6

I

I

I

I

I

1

2

3

4

5

sec

LOG

(a)

t,

sec

(a)

uu >

100

<

: _~.--.--~..-.-~

100 I

+ I

I

I !

l

0

< 9 00"r

300"r

I 2

t,

LOG

4so~

I

I

I

I

3

4

5

I

sec

I 2 LOG

(b)

I '

0

Z

200"C

I I

l

t.

400"r

300"c

I

I

I

3

4

5

sec

(b) Fig. 2 - - Diagrams showing the influence of aging time and temperature on the properties of Alloy 2. (a) Vickers hardness number, (b) shape recovery.

(Ms>ambient) the various kinds of (semi-) coherent interfaces present may become rearranged (as in alloys 2 and 3) in the manner described by Wayman and Shimizu.16 The shape memory effect occurs in both cases by reversal of those interface movements during subsequent heating. It is to be expected that microstructural features that interfere with the creation of, or the glissile nature of, the interfaces will reduce their capacity to move under the influence of stress and to revert to an initial configuration during heating. An adverse influence on the shape memory effect is therefore to be expected. Such microstructural features can arise by diffusional transformation of the metastable phases involved in the shape memory effect to a more stable structure during an aging treatment. In the present work, aging resulted in both the bcc fl, phase in alloy 1 and the 13~ martensite in alloys 2 and 3 being replaced ultimately with the finely dispersed structures shown in Figures 4 and 5. These structures were identified by X-ray diffraction of severely overaged* speci:

oU

A~

.-...o-.+t "~ _ _ o "..~.~

+, , . . , , . . . . ~

"%.

-<- - 6 0 u.i

II

M+

_2

_/. je

9

o.---i~,,,l,.e,~--o

--

,oo.~

300"r i00%

-120 I

I

I

I

I

1

2

3

4

5

t,

LOG

ser

(c) E

293

Q.

<

292 40~'C

291

aSO+C

300"C

2~0C

~OC

I

I

I

I

I

1

2

3

4

5

LOG t, s e c

(d) Fig. 1 - - Diagrams showing the influence of aging time and temperature on the properties of Alloy 1. (a) Vickers hardness number, (b) shape recovery, (c) Ms, Ms, and A+ temperatures, and (d) lattice parameter of the fl, phase. 552--VOLUME 13A, APRIL 1982

*It was necessary to age for prolonged periods to obtain diffraction patterns from which parameter measurements could be made; for short aging periods the patterns contained only very diffuse lines.

mens as mixtures of the fcc a phase and cubic 72 phase as indicated in Table III. It should be noted that the aging temperature of 400 ~ was above the peritectoid temperature of 363 ~ proposed for Cu-Al alloys by Jewett and Mack, 17 and consequently the possibility of the ordered fcc a2 phase occurring in the Cu-A1-Ni alloys as a result of the peritectoid transformation was precluded. METALLURGICAL TRANSACTIONS A

Table II. Activation Energies Obtained from Property Measurements

Property Increase in hardness Loss of shape memory Change of Ms, Mz, As, .4: Decrease in lattice parameter

Activation Energy (kJ mol-i) with 95 Pct Confidence Limits Alloy 1 Alloy 2 Alloy 3 78 ---8 58 ---4 72 ___1 60 +6 62 ---12 75 +--8 68 ---1 73 - 10 78 +3

(a)

(b)

Fig. 4--Photomicrograph showing the structure of Alloy 1 aged for 24 h at 300 ~ etched in potassium dichromate, magnification 750 times.

(c) Fig. 3 - Diagrams showing the influence of aging time and temperature on the properties of Alloy 3. (a) Vickers hardness number, (b) shape recovery, and (c) M,, As, and A: temperatures.

As the microstructure of the /3~ or the 13'1 changed progressively during aging, time and temperature dependent changes in properties occurred, as shown in Figures l to 3. In alloy 1, the times required at each aging temperature to initiate the increase in hardness (Figure l(a)) and to initiate the decrease in lattice parameter (Figure l(d)) were similar, with similar activation energies (Table II). These observations strongly suggested that the common origin of these property changes was precipitation of 3/2 phase (Table III), or a transition phase that ultimately became the "Y2phase, which hardened the structure and simultaneously depleted the maxtrix of A1 and Ni solute atoms, thereby decreasing the lattice parameter of the 131 phase. This conclusion is consistent with the appearance of precipitates in the microMETALLURGICALTRANSACTIONS A

Fig. 5--Photomicrograph showing the structure of Alloy 3 aged for 24h at 300 ~ etched in one pct aqueous chromic acid; magnification 750 times.

structure after aging times similar to those required to initiate the increase in hardness. The progressive decrease in lattice parameter, without an initial increase as found in C u - S n , 9 indicates that the /3~ phase remained ordered at all aging temperatures, consistent with a D O 3 ordering temperature of 540 ~ reported by Kurdjumow TM for a Cu-15 pct A1 alloy. Although changes in the transformation temperatures (Figure l(c)) are curious, the times required to initiate them were comparable with the times required to initiate changes in other properties, suggesting that the onset of precipitation VOLUME 13A, APRIL 1982--553

Table IIl. X-Ray Diffraction Analyses of Overaged Specimens

Alloy

Aging Treatment Temperature Time(s)

Phases Detected

Lattice Parameter (pm)

a Tz

365.4 871.3

1

400 ~

3.4 • 103

2

400 ~

8.4 x 104

ot

365.4

8.4 x l04

3'2 o~ 3'2

872.7 367.4 871.0

3

400 ~

is also the dominant influence in this case. A change in transformation temperature may originate from such effects as the following. (i) Formation of solute rich precipitates which deplete the matrix and raise Ms, MI, As, At. (ii) Formation of solute lean precipitates which enrich the matrix and lower Ms, MI, A~, As. (iii) Formation of segregates or precipitates which interfere with either the nucleation (e. g., by embryo exhaustion 19':~ or the growth (e. g., by partitioning) of the martensite and which lower Ms, Ms, As, Ay. (iv) Large decrease in grain size, changing the distribution of embryos in some grains with consequent decrease in Ms, Mj, As, AI:'. (v) Change in the degree or type of order in the high temperature phase, changing the thermodynamical properties of the system and so changing Ms, M/, As, Af. 13.22 At 400 ~ Ms, etc. decreased during aging, as reported previously for Cu-Su 9 and also as observed for alloy 3 (Figure 3(c)). This result is consistent with embryo exhaustion or some similar process, since it is unlikely that the precipitates were solute lean. At 200 ~ Ms, etc. increased, as would be expected with the formation of solute rich precipitates that do not interfere significantly with the nucleation or growth of the martensite. Stobbs et al .z3 also observed Ms to increase during aging of a similar alloy at 250 ~ and ascribed the change to the formation of "to-like" precipitates. Thus, it appears for this alloy that there is a temperature dependent change in the type of precipitation process that occurs during aging with consequent changed influence on Ms, My, As, and Ay. In alloys 2 and 3 the formation of the equilibrium a and 7: phases during prolonged heat treatment may occur by either: (i) aging of/3, formed from the/31 martensite on heating; or (ii) tempering of /3'~ martensite remaining at the aging temperature. In process (i), precipitation from the/3~ may occur as 7,particles or a transition phase related to 72. Alternatively, however, precipitates may be in the form of heavily faulted plates similar to those termed "bainite" reported by Hornbogen and Warlimont 6 in fl-CuZn as a transition phase which became the a phase as the faults decayed during aging. In alloy 3, hot stage and other metallographic observations z4 suggested that both processes (i) and (ii) occurred, for a significant volume fraction of the /3~ martensite ap554--VOLUME 13A. APRIL 1982

peared to be stabilized and remained untransformed during aging treatments at temperatures below about 300 ~ Stabilization of some of the fll martensite in alloy 3 left only a dispersed volume fraction of material on which to make observations of the relief effects to determine the transformation temperatures on heating and cooling. Consequently, the results shown in Figure 3(c) have limited accuracy, as it was difficult to obtain reproducible results. Nevertheless, the decreased Ms, etc. in alloy 3 is compatible with the results obtained by Cook and Brown 13 for a similar Cu-Zn-AI alloy aged at 50 to 120 ~ and ascribed to changed degree of DO3 ordering of the/31 phase. Although such an aging mechanism is feasible, Melton and Mercier25 have reported that the martensite formed in quenched Cu-Zn-A1 alloys of similar composition to the one studied by Cook and Brown ~3has the 9R structure derived from parent phase with B2 order. On this basis an alternative explanation for the aging mechanism is that, because the DO3 ordering temperature is lower than that for B2 ordering, ~2a transition occurs from B2 to DO3 order on aging at low temperatures. The times required to initiate the decrease in Ms, etc. and the increase in hardness for alloy 3 are similar (Figures 3(a) and 3(c)), indicating that both changes probably have a common origin. While at low aging temperatures this may be changed DO3 order 13or a B2 ~ DO3 transition, ~2 at higher temperatures, e.g., 300 ~ the changes were associated with the appearance in the microstructurc of precipitates which were morphologically similar to the "bainite" reported by Hombogen and Warlimont. 6 There is no reason to suppose that the increase in hardness in alloy 2 (Figure 2(a)) cannot also be associated with the onset of precipitation of some kind, probably involving the 72 phase. It would be expected that the microstructural changes that caused the changes in hardness and transformation characteristics would also reduce the shape memory capacity of the alloys. Figures l(b), 2(h), and 3(b) show this to be the case. The similarity of the activation energies derived from loss of the shape memory capability and from changes in other properties (Table II) indicated that similar rate controlling processes were operative for each property. It is likely that at the higher aging temperatures the measured activation energies related to processes that led to the formation of structures consisting of the a and 72 phases. However, at the lower temperatures it is possible that different processes took place; for example, the peritectoid transformation suggested by Jewett and Mack ~7may have occurred in alloys 1 and 2, and changes in the degree or type of order ~2''3 may have occurred in alloy 3. Consequently, it is likely that the measured activation energies do not relate to a single atomic process, even though the results appear to conform with an Arrhenius type relationship. Although it follows that values of the activation energies have doubtful physical significance, nevertheless they are consistent with the value of 65 • kJ mol -~ obtained by Cook and Brown ~3 for a Cu-26 pet Zn-4 pct AI alloy. As these values are much lower than the activation energy for diffusion in similar alloys ( - 1 5 0 to 200 kJ mol-~), 26 it is strongly suggested that the processes which occurred during the aging treatments were diffusional, but much enhanced by a superconcentration of quenched-in vacancies, as proposed by Clark and Brown 27 and supported METALLURGICALTRANSACTIONS A

Table IV. Aging Times Required to Lose Shape Memory Capability Temperature 100 ~ 50 ~ 20 ~

Alloy 1 12 days 240 days 6.5 years

Alloy 2 3.0 years 69 years 744 years

Alloy 3 42.5 days 5 years 82 years

by Cook and Brown. 13 An alternative but a less likely possibility, at least for alloys 2 and 3, is enhancement by the structural defects inevitably generated by transformation of the/3; martensite to/31 during heating. The results clearly show that the microstructural changes can produce quite large changes in hardness and transformation temperatures before the shape memory capacity is affected. This result is most pronounced in alloy 2, for which the time required to degrade the shape memory effect was approximately three orders of magnitude greater than the time required to initiate change in other properties. Although the physical significance of the activation energies derived from measurements of shape memory is more obscure than for those derived from other property changes, the values can be used with the Arrhenius equation to predict the life expectancy of shape memory devices operating at temperatures lower than those examined. These predictions, Table IV, are based on the assumption that the particular Arrhenius relationships extrapolate to the lower temperatures. Table IV indicates that alloys 1 and 3 should have considerably lower stabilities than alloy 2 on aging at temperatures of 100 ~ or less. Alloy 1 is more highly supersaturated than alloy 2, and this could account for its lower stability on aging. The rapid aging of the Zn-bearing alloy 3 could also be associated with the susceptibility to decomposition to the "bainite" product, which appears to be absent in the Cu-Ni-A1 alloys. It is evident from the recent survey by Wayman z8 that Cu-Zn-A1 alloys, similar to alloy 3, are preferred materials for commercial exploitation of Cu-based shape memory alloys. However, the results shown in Table IV clearly indicate that such alloys have a relatively short life expectancy at temperatures as low as 100 ~ and certainly for any device operating under conditions that involve thermal excursions above 50 ~ the Cu-A1-Ni alloys would have significantly greater thermal stability.

CONCLUSION Aging of/31, Cu-A1-Ni./3; Cu-A1-Ni. and/3; Cu-Zn-A1 shape memory alloys at temperatures in the range 200 to 450 ~ resulted ultimately in loss of the shape mem, ory capacity. In all cases the loss was preceded by an increase in hardness which at higher temperatures, at least, was due to precipitation leading to the formation of the stable a and Y2 phases, identified by X-ray diffraction. In addition, there was an associated change in the martensitic transformation

METALLURGICALTRANSACTIONS A

temperatures which depended on alloy composition and aging temperature. The time required to initiate change in these properties was related to aging temperature by an Arrhenius equation from which activation energies in the range 60 to 80 kJ mol -~ were obtained. The relationship was used to predict the life expectancy of shape memory devices operating at temperatures lower than 200 ~ These predictions indicate that the/3 ( Cu-A1-Ni alloy is the most resistant to degradation.

ACKNOWLEDGMENTS The authors gratefully acknowledge the financial support for this work provided by The University of Wollongong and the Australian Research Grants Committee, and are indebted to R. Doherty for some of the experimentation.

REFERENCES 1. R.G. Cope: J. Inst. Metals, 1958-59, vol. 87, p. 330. 2. R. Rapacioli and M. Chandrasekaran: Proc. Internat. Conf, on Mart. Trans., M.I.T., Cambridge, MA, 1979, p. 596. 3. A.A. Hussein, L. I. E1-Menanati, and H. J. Klaar: Metall. Trans. A, 1978, vol. 9A, p. 1783. 4. N. Kuwauo, I. Otaga, and Y. Tomkioy: Trans. Jap. Inst. Met., 1977, vol. 18, p. 195. 5. H. Goldenstein and I. G. S. Falleiros: Proc. Internat. Conf. on Mart. Transf., M.I.T., Cambridge, MA, 1979, p. 566. 6. E. Hombogen and H. Warlimont: Acta, Met., 1967, vol. 15, p. 943. 7. L. Guzman and M. Ruhlie: Proc. lnternat. Conf. on Mart. Trans., M.I.T., Cambridge, MA, 1979, p. 590. 8. R.K. Govila: Acta. Met., 1964, vol. 12, p. 273. 9. N.F. Kennon: Met. Sci. J., 1972, vol. 6, p. 64. 10. R. Rapacioli, M. Chandrasekaran, and L. Delaey: Shape Memory Effects in Alloys, J. Perkins, ed., Plenum Press, New York, NY, 1975, p. 365. 11. L. Chandrasekaran and A. E Miodownik: Proc. Internat. Conf. on Mart. Trans., M. I. T., Cambridge, MA, 1979, p. 584. 12. L. Delaey, A. Deruyttere, E. Aemoudt, and J. R. Roos: Report 78RI1978 Shape Memory Effect, Department Metaalkunde, Katholicke Universiteit, Leuven, Belgium. 13. J.M. Cook and L. M. Brown: Scripta Met., 1978, vol. 12, p. 949. 14. K. Otsuka and K. Shimizu: Scripta Met., 1970, vol. 4, p. 469. 15. I.B. Nelson and D. P. Riley: Proc. Phys. Soc., 1945, vol. 57, p. 160. 16. C.M. Wayman and K. Shimizu: Met. Sci. J., 1972, vol. 6, p. 175. 17. R.P. Jewett and D. J. Mack: J. Inst. Metals, 1963-64, vol. 92, p. 59. 18. G.V. Kurdjumow: Byull. AN USSR, Ser. Khim., 1936, vol. 2, p. 271. 19. M. Cohen. E.S. Machlin, and V.G. Paranjpe: Thermodynamics in Physical Metallurgy, ASM, Cleveland, OH, 1950, p. 242. 20. M. Cohen: Trans. ASM, 1949, vol. 41, p. 35. 21. K.E. Easterling and G.C. Weatherly: Acta Met., 1969, vol. 17, p. 845. 22. D.P. Dunne and C. M. Wayman: Metall. Trans., 1973, vol.4, p. 147. 23. W.M. Stobbs. R. J. Henderson, and A. M. Crossley: Proc. Internat. Conf. on Mart. Trans., M.I.T., Cambridge, MA, 1979, p. 578. 24. D.P. Dunne. N.F. Kennon, and L. Middleton: University of Wollongong. Australia. unpublished research, 1982. 25. K.N. Melton and O. Mercier: Metall. Trans. A , 1979, vol. 10A, p. 875. 26. C.I. Smithells: Metals Reference Book, (5th ed.), Butterworths, London. 1976. p. 860. 27. J.S. Clark and N. Brown: J. Phys. Chem. Solids, 1961, vol. 19, p. 291. 28. C.M Wayman: J. Metals, 1980, vol. 32, no. 6, p. 129.

VOLUME 13A, APRIL 1982

555