Stress-Induced Martensitic Transformation Cycling and Two-Way Shape Memory Training in Cu-Zn-Al Alloys

Stress-Induced Martensitic Transformation Cycling and Two-Way Shape Memory Training in Cu-Zn-AI Alloys

JEFF PERKINS and R.O. SPONHOLZ

The character and mechanism of two-way shape memory in Cu-Zn-A1 alloys is investigated by means of closely controlled thermomechanical cycling and careful measurement of the progressive effect of the particular "training" routine, as well as by correlary studies of submicrostructural evolution as training proceeds. The results establish the quantitative relationship between the cyclic training routine and the ability of the sample to exhibit two-way shape memory. The variation of numerous training parameters with cycling is presented and interpreted. Microscopic studies indicate that as two-way shape memory training proceeds, specific physical features develop in the parent phase submicrostructure, particularly dislocation tangles and "vestigial" martensite markings; these assist in the nucleation and growth of a preferred martensite plate arrangement during cooling.

I. INTRODUCTION

IN recent years, considerable attention has been given to alloys which exhibit so-called "shape memory." These alloys can undergo surprisingly large amounts of strain and then, upon temperature increase or unloading, revert to their original shape. Shape memory has now been demonstrated in quite a few alloys, the most prominent of these being NiTi, CuA1Ni, and CuZnA1. The last system is the subject of the present investigation.

The literature is rife with special terminology, and a brief summary of terms in in order: (1) "Shape Memory" (SM): this is a general term which refers to various thermoelastic behaviors where there is a reversal of apparent plastic strain; (2) "Shape Memory Effect" (SME): a specimen which has been deformed maintains that deformation when the applied stress is removed, but recovers its original shape when heated; (3)"Stress Induced Martensite" (SIM): a specimen is deformed while in the parent phase (at a temperature above the Ms temperature), producing martensite; (4) "Pseudoelastic Effect"; the specimen recovers a pro- portion of the induced strain immediately on unloading; (5) "Two-Way Shape Memory" (TWSM): a specimen changes shape in both directions as a function of temperature (heating and cooling); in effect, the alloy remem- bers both high and low temperature shapes. TWSM can be achieved by subjecting a specimen to a number of SME cycles, SIM cycles, or, as in this work, a series of combined SIM-SME cycles. (See Figure 1.)

The essence of shape memory depends upon the alloy's ability to undergo a thermoelastic martensitic transformation. A thermoelastic martensitic transformation is realized if martensite forms and grows continuously as the temperature is lowered and/or as stress is increased, and shrinks and vanishes continuously along the same path as temperature is raised and/or stress is decreased. Cooling or the application of stress promotes the martensitic transformation, while heating or unstressing reverses it. Combining these two aspects has led to the development of certain

JEFF PERKINS is Associate Professor of Materials Science, Department of Mechanical Engineering, Naval Postgraduate School, Code 69Ps, Monterey, CA 93940. R.O. SPONHOLZ, LCDR, United States Navy and Graduate Student, is with Naval Postgraduate School, Monterey, CA 93940.

Manuscript submitted February 3, 1983.

so-called training routines for two-way shape memory. It is implicit in the concept of thermoelastic martensitic transformation that there is no irreversible behavior involved in the P-M transformations, i .e. , no plastic deformation. However, there is an increasing amount of indirect 1'2 and direct 3'4'5 evidence that this is not the case, at least on the microscopic scale, and that local transformation strengthen- ing processes take place.

The purpose of this work was to investigate the effects of TWSM training routines. Of particular interest was how the experimental data would compare with the idealized pat- terns depicted schematically in Figure 1. That is, what values of stress and displacement are exhibited as the specimen is subjected to a single training cycle and in what manner

SME Training for TWSM

TWSM / Ms ~ Behavior

SIM Training for TWSM

.A

�9 SIM Training for TWSM: ~ I I ~ ] ABCDA repeated D

�9 SMETroining for TWSM~ AEFGHIJA repeated D INDUCED STRAIN, c i

�9 Combined Training for TWSM = ABCGHIJ repeated ~.

Q TWSM Behavior= O(~(~@ ~>

Fig. l - Schematic illustration of training routines used to obtain 2-way shape memory (TWSM) behavior; SIM training for TWSM involves repeating the pattern ABCDA; SME training for TWSM involves repeating the pattern AEFGHIJA; "combined" training for TWSM involves repeating the pattern ABCGHIJA. After training the alloy will spontaneously follow the TWSM behavior pattern (1)(2)(3)(4) on cooling and heating.

METALLURGICAL TRANSACTIONS A VOLUME 15A, FEBRUARY 1984--313

do these values change as the number of training cycles increases. Therefore, this work was concerned with the important quality of reproducibility in thermomechanical behavior, with the ability of specimens to improve their two-way shape memory with training, and with points of diminishing returns in terms of such training. Further- more, it was an aim of this work to investigate possible submicrostructural changes associated with TWSM training in order to understand the mechanisms of two-way shape memory.

II. EXPERIMENTAL

The Cu-Zn-AI alloy material studied was provided by Delta Materials Research Limited, Ipswick, Suffolk, England, with a nominal composition of 69.25 wt pct Cu, 26.75 wt pct Zn, and 4.0 wt pct A1. This alloy was selected on the basis of its Ms temperature, which would place it in the parent phase at room temperature. The as-received material was machined into tensile test specimens 2.250 inches (57.15 mm) in length and 0.1378 inch (3.5 mm) uniform diameter; these were sealed in evacuated quartz tubes, solution treated at 900 ~ for 15 minutes and quenched in ice water. Microscopic examination showed the samples to con- sist of homogeneous parent phase with a large grain structure, about 0.3 mm in diameter. Each tensile test specimen was hand polished followed by electropolishing to a mirror finish in 10 pct KCN solution (6 volts AC for 30 seconds). Disks 3.0 mm in diameter and 0.030 mm thick were cut out for use in determining martensite and parent start and finish temperatures using a Perkin-Elmer Corporation Model DSC-2 Differential Scanning Calorimeter (DSC). The temperatures recorded for alloy D are as follows: Ps = 264 K, Pi = 279 K, Ms = 268 K, and M I = 253 K.

Training routines were conducted on an Instron Testing Machine, Model 1102 TM-S-L, at a constant strain rate of 0.008 inch per minute. A 1.0 inch (25.4 mm) specimen gauge length was used. The grips were drilled and tapped to provide a serrated gripping surface and then slit to per- mit constriction of the specimen when collar screws were tightened. Preliminary tests were conducted to ensure non- slippage of the grips while providing uniform strain along the length of the tensile test specimens. Surrounding the grip assembly and test specimen was an insulated bucket which could be raised or lowered on the grip and into which liquids at various temperatures could flow via a simple siphon system. In order to cycle samples between the martensite and parent phases, a mixture of methyl alcohol and liquid nitro- gen was used to achieve the low temperature phase while room temperature alcohol was used for the high temperature phase. Temperature was determined by means of a Newport Laboratories pyrometer, Model 267B-KC1, coupled to a chromel/alumel thermocouple. Displacement was measured at the grips by means of a noncontact Electro Corporation Electro-Micrometer Model PAl1547, which was mounted on extension arms to remove it from the liquid in the bucket. The displacement output signal was fed to a Honeywell Electronik 194 strip chart recorder.

The general training routine for a given specimen consisted of the following sequence: (1) starting with the alloy in the parent phase (T > PI); (2) strain 3 pct; (3) while maintaining strain, cool specimen below MI; (4) unload by moving Instron crosshead to achieve zero stress; (5) unfix

314--VOLUME 15A, FEBRUARY 1984

one end by removing pin from upper grip; (6)warm to T > Pi- This routine corresponds to the path ABCGHIJ in Figure 1 and was conducted for various number of cycles.

Evaluation of TWSM after a given number of training cycles consisted of the following sequence: (1) starting with the trained alloy in parent phase (T > PI) and one end unfixed; (2) cool to T < M I, record displacement; (3) warm to T > Pi, record displacement.

Disks were cut from trained tensile specimens and pre- pared for transmission electron microscopy by jet polishing in a 3 pct perchloric/methanol solution using a Struers Tenupol 2 apparatus. The foils were examined in a Philips EM 201 transmission electron microscope. Foils were examined prior to and after various number of training cycles in order to observe changes in microstructure that would help characterize the training mechanism.

III. RESULTS AND DISCUSSION

A. Characterization of the Training Routine

The experimental data confirm the general shape of the training pattern ABCGHIJ represented in Figure 1. In the course of characterizing the training routine, and changes in it during multiple training cycles, a number of particular parameters were monitored. These are described with the as- sistance of Figures 2(a) and 2(b). Figure 2(a) represents the stress recorded by the Instron machine load cell plotted against the crosshead movement, while Figure 2(b) depicts the corresponding electronic displacement data plotted by the Honeywell strip chart recorder.

First of all, a distinction can be seen between the first training cycle and subsequent cycles. During the first training cycle, the peak stress level (point B in Figure 2(a)), which corresponds to the first-time formation of stress- induced martensite from virgin parent phase, is greater than in a typical follow-on training cycle (point F in Figure 2(a)) and, as will be discussed later, this stress decreases continuously with training cycles.

Once the desired initial induced displacement m i w a s

reached, the Instron machine was stopped. While maintaining the grips in this fixed position, chilled methanol (T < Ms) was siphoned into the bucket surrounding the specimen. When this was done, the stress level dramatically decreased, (B ~ C in Figure 2(a)), while the elongation simultaneously increased, (B ~ C in Figure 2(b)). This elongation, due to formation of thermoelastic martensite under stress, is designated A,,. After a short period of time, this thermoelastic martensitic transformation is complete, at point C. In essence, the specimen, at point C, has under- gone a stress induced martensitic transformation with a thermoelastic martensitic transformation superimposed upon it. It is this low temperature, low stress, maximum elongation point that the specimen is being trained to remember. From point C the crosshead was moved to achieve a zero stress level. With this accomplished, the specimen, unfixed at one end and heated to T > PI, returned to the vicinity of its starting shape.

Of particular note, and another distinction of the very first training cycle, was the imparting of some amount of irreversible plastic deformation, Ap, to the specimens, as sum- marized in Table I. The cause of this deformation may be stress accommodation at parent phase grain boundaries.

METALLURGICAL TRANSACTIONS A

]st CYCLE

f LATER CYCLE

/

w F- ff)

E D,

CROSSHEAD MOVEMENT

W

I--

-Md>T>Pf I

I # I

,,,,, J J

T<Mf

(a)

L

,L..__._._ _ H

D C

~Ap LDISPLACEMENTAi I Am J T, T ?

(b) Fig. 2--(a) Stress-strain profile on initial and later training cycles; (b) corresponding time-displacement profile, defining the parameters Ap, A, Am, and A+.

That is, the strain limit for completely reversible shape memory (eL) may be exceeded in local areas of the polycrystalline material. In any event, it is clear that some points within the specimen experience irreversible plastic deformation. Nagasawa 6 has suggested that a necessary condition of TWSM is severe deformation below Ms, while Otsuka and Shimizu 7 note that by inducing plastic strain one assists the stress-induced martensite transformation. Otsuka and Shimizu explain that dislocations introduced by the plastic strain assist nucleation of a particular variant of stress induced martensite. Perkins 8 notes that the most pronounced TWSM effects are observed in cases where an alloy is deformed severely and nonuniformly, such that some fibers exceed the strain limit, (eL).

In any case, the starting point for each subsequent training cycle is the final elongation of the previous cycle. In Figure 2(b) point E is the point of elongation after com- pletion of the first training cycle. Point E then becomes the starting point for the second training cycle.

The net displacement induced in the specimen by crosshead movement (Ai) as a function of cycling is tabulated in Table II and presented graphically in Figure 3. A first cycle induced strain of 3 pct was selected. The Table II entries for the first cycle subtract the amount of first cycle plastic deformation listed in Table I. That is, the Table I plus the Table II (first cycle) entries total 30 mils displacement or 3 pct strain. As can be seen from Table II, the amount of induced displacement decreases with cycling and approaches some constant level. The absolute value of this decrease in induced displacement was approximately 4 to 5 mils (0.4 to 0.5 pct strain) in the highly trained specimens. There are a number of possible sources of this downward trend. First, an increment of transformation strain hardening may occur during each cycle, eventually achieving a saturation point.

Table II. Induced Displacement (Ai) as a Function of Training (Ai in Mils; 1 Mil = 0.001 Inch = 0.0025 cm)

Cycle D-10 D-5 D-3 D-4 D-9 D-6 D-7

1 23.4 21.6 23.4 19.2 27.0 27.6 24.0 2 23.4 21.6 24.0 27.0 27.0 28.2 24.6 3 23.4 20.4 22.2 27.0 25.8 27.6 24.0 4 20.4 22.8 26.4 25.2 27.0 24.0 5 19.8 22.8 24.6 24.6 26.4 24.6 6 21.6 25.8 24.6 26.4 23.6 7 22.3 24.6 24.6 25.2 24.0 8 21.0 25.2 22.8 25.2 24.0 9 21.6 24.6 24.0 24.0 24.0

10 22.2 23.4 24.0 24.0 22.8 11 22.2 23.4 23.7 22.8 12 22.8 23.4 22.8 23.4 13 22.2 23.4 22.8 24.4 14 22.2 23.4 22.8 23.1 15 22.2 24.4 22.8 24.4 16 24.0 22.8 22.2 17 23.4 21.0 23.7 18 23.4 21.6 23.7 19 23.4 21.6 22.2 20 23.4 22.2 22.2 21 21.6 22 22.2 23 22.2 24 22.2 25 22.2 26 21.0 27 21.0 28 21.9 29 21.6 30 22.8

Table I. Amount of Plastic Deformation (A,) Retained in the Specimen after the First Training Cycle (in Mils)*

Specimen Designation D- 10 D-5 D-3 D-4 D-9 D-6 D-7 Eventual Training Cycles (3x) (5x) (10• (15x) (20x) (20x) (30x)

Ap 6.6 8.4 6.6 13.6 3.0 2.4 6.0

*Since the specimen's original gauge length was 1 inch or 1000 mils, Ap, first cycle, can be converted to pct plastic strain by dividing table entries by 10. For example, first cycle pct plastic strain for specimen D-10 would be 0.66 pct.


28.5 28

27.5 ~ 27 E 26.5 ~ 26 z 25.5 ~ 25 ~ 24.5 ~. 24 m 23.5 ~ 23 o w 22.5 ~ 22 z 21.5

21 20,5

0 2 4 6 8 I0 12 14 16 18 NUMBER OF TRAINING CYCLES

Fig. 3--1nduced displacement, Ai, v s number of training cycles.

! ~ '~ Specimen D - 6 - I ~ ~ O Specimen D-9 e o ~ �9 [] Specimen D - 7

a o ~ o n n

o o a . . ~ ' ~ ' ~ . - . ~ [] [] I1

I I I I I I I I I I 20

Second, it is possible that the alloy begins to "remember," via microstructural change, its strained low temperature position and thus does not fully recover its initial position upon heating to T > PI. The microstructural mechanism for this unwillingness of the specimen to recover its original position might be the retention of small, but (with cycling) increasing amounts of martensite on heating. This latter explanation is supported by experiments where it was shown that it is possible to drive the alloy back nearly to its previous cycle displacement starting position by increasing the temperature of the bath. Thus the retention of some amount of metastable retained martensite is likely. Delaey, et al.,9 have reported retained martensite on both thermal and stress cycling. Perkins and Muesing 5 have made similar findings in the case of thermal cycling.

The elongation due to thermoelastic martensitic transformation under stress, Am, is presented in Table III. This elongation occurs to the greatest extent in the first or second cycle and eventually achieves a fairly constant value. It should be remembered that at this point in the training cycle the Instron machine crosshead is fixed. In order for the specimen to elongate at all, its displacement must be accom- modated by the Instron/grip assembly, either by thermal contraction or machine compliance. Since the observed stress decreases dramatically during the thermoelastic martensitic transformation, rig thermal contraction alone is ruled out. It is likely that slight deflection of the load cell accommodates the small displacements associated with Am. Since the stress level does not reach zero, additional A m

would be possible if the specimen was inclined toward greater displacement.

The total elongation due to training, designated At, is the sum of the induced elongation Ai and the thermoelastic elongation Am. It is this (At) displacement which will be used later as a measure of the trainability of specimens.

The peak stress on the specimen during a given training cycle is represented by points such as B and F in Fig- ure 2(a). Figure 4 plots peak stress as a function of cycles at a fixed temperature of 10 ~ for a number of specimens. The peak stress, in general, decreases as the number of training cycles increases, with the magnitude of this decrease being of the order of 1000 to 2000 psi. It seems that the alloy, by microstructurally adjusting to the training routine, is able to achieve its maximum elongation at lower stress as training proceeds. Schroeder and Wayman 1~ have

Table IlI. Thermoelastic Martensitic Displacement (Am) as a Function of Training (Am in Mils;

1 Mil = 0.001 Inch = 0.0025 cm)

Cycle D- 10 D-5 D-3 D-4 D-9 D-6 D-7

1 6.6 9.6 7.2 7.2 7.2 6.6 7.2 2 5.1 7.5 4.2 3.6 7.2 7.8 5.4 3 5.1 7.2 4.8 4.5 7.2 6.6 5.4 4 7.5 4.8 4.8 5.4 7.2 4.8 5 7.2 4.8 4.8 6.0 7.2 5.4 6 5.0 4.2 6.0 6.9 4.8 7 4.2 6.0 6.0 6.9 5.4 8 3.6 6.0 4.2 6.0 5.4 9 3.6 5.4 5.4 5.4 6.6

10 4.2 6.6 4.8 5.4 4.8 11 6.0 4.8 6.6 4.8 12 6.6 4.8 6.6 4.8 13 4.8 4.8 5.4 5.4 14 5.1 4.2 6.0 4.8 15 4.2 4.8 4.8 5.4 16 5.4 4.2 5.4 17 4.8 4.2 6.0 18 4.8 4.2 6.0 19 5.4 4.2 7.2 20 5.1 5.4 4.8 21 6.0 22 6.0 23 5.4 24 6.0 25 6.0 26 5.4 27 4.2 28 5.1 29 3.6 30 5.4

16000

15500 -- ~.,~

15000- .~ ~ Specimen 0-9

=ml_n..w 1400014500 _-- ~ 1 PSI= 6.B9x10 -3 N / r a m 2

03

15500

15000 ~ ~ . , , ~ ~

12500 / I I I I I I I I I I 0 2 4 6 8 10 12 14 16 18 20

NUMBER OF TRAINING CYCLES ~.- Fig. 4 - - P e a k stress v s number of training cycles.

concluded that TWSM results from a preferred variant of martensite, namely, that variant which permits the lowest stress to be obtained. Otsuka and Shimizu 7 have found that after permanent strain was introduced, the stress for forward transformation was lowered; that is, the introduction of strain assists the transformation.

A summary of training cycle parameters is presented in Table IV. Peak stress has been previously defined. Cooling stress (o'coo,) corresponds to the stress level after the specimen has transformed from parent to martensite (points such

316--VOLUME 15A, FEBRUARY 1984 METALLURGICAL TRANSACTIONS A

Table IV. Training Cycle Parameters--Typical Beginning and Ending Values*

Alloy Cycle O'max (psi) ~rcoot (psi) Start Temp. (~ Cool Temp. (~ Ai (Mils) A,, (Mils)

D-3 2 12,800 2,335 10 -33 24.0 4.2 10 11,266 1,000 10 -31 22.2 4.2

D-4 3 20,000 8,660 12 - 34 27.0 4.5 13 15,530 5,330 12 -34 22.2 4.8

D-6 1 15,825 2,550 10 -30 27.6 6.6 2 16,760 2,350 12 -40 28.2 7.8

20 14,080 805 12 - 40 22.2 5.4

D-9 1 18,775 4,025 12 -40 27.0 7.2 4 17,770 5,630 11 -40 25.2 6.0

20 15,220 3,350 11 -40 23.4 5.1

D-7 1 16,600 2,666 10 -38 24.0 7.2 2 13,670 1,000 9 �9 -43 24.6 5.4

29 11,670 500 10 -43 21.6 3.6

*1 psi = 6.89 x 10 -3 N/mm2; 1 mil = 0.0025 cm

as C or G, Figure 2(a)). Start temperature is the temperature that the specimen experienced during the formation of stress induced martensite (points A ~ B, Figure 2(a)). Cooling temperature is the temperature at which formation of thermoelastic martensite took place (points B ~ C, Figure 2(a)). Samples are also compared for early and late cycles. In general, the parameters decrease in value as the number of cycles increases. It is thus apparent that the alloy is learning its training routine. The elongation due to the thermoelastic martensitic transformation decreases because, after the first few cycles, the alloy retains some amount of martensite and therefore less is available to transform.

B. Two-Way Shape Memory Trainability

The ability of specimens to demonstrate TWSM as a result of training is presented in Figure 5 and Table V. The percentage entries are obtained from the ratio of the displacement recorded during a cooling sequence, designated AxwsM (see Experimental) divided by the average induced displacement (At) in the training cycles (AvwsM/Ar X 100). The average induced displacement, in mils, is the figure in parentheses in Table V. Because of the way it is defined, this percentage trainability parameter is conservative, because

t-- r

I-- Z I,I

e. O3 E3

O3

l--

5 5 - -

50 - -

4O

3 5 - -

3 0 - -

2 5 - -

2 0

15 / /

10 /,,'

5 /

0 I I 0 2 4

/ t

I I I I I I I I I I I I I 6 8 I0 12 14 16 18 20 2 2 2 4 2 6 2 8 30 NUMBER OF TRAINING CYCLES .=

Fig. 5--Two-way shape memory trainability, as characterized by the parameter (AvwsM/Ar) • 100, plotted v s the number of training cycles.

the denominator, At, is an average which incorporates greater displacements from earlier cycles. Figure 5 is a composite graph fitting all TWSM data in Table V with a third order polynominal.

Table V. TWSM Performance

Number of Training Cycles

Alloy 3 5 10 13 15 18 20 25 28 30

D-10

D-5

D-3

D-4

D-9

D-6

D-7

10.4 pct (29.0)

27.3 pct (28.6)

45.6 pct (29.1)

49.5 pct (29.1)

34.4 pct (29.6)

33.9 pct 38.5 pct 45.7 pct (34.4) (32.8) (30.2) 20.2 pct 30.6 pct 40.1 pct 45.2 pct 43.3 pct 43.4 pct 47.9 pct 47.3 pct 48.6 pct

(29.8) (29.4) (29.2) (29.2) (29.1) (29.0) (28.8) (28.5) (28.4)


The data indicate that subjecting a sample to a larger number of training cycles allows it to achieve greater TWSM displacement during cooling. The trained alloys were able to repeat their trained displacement over many thermal cycles and to repeat their training after long waiting periods at room temperature (periods up to several weeks were examined).

C. Mechanisms of Two-Way Shape Memory

In their work on single crystal CuZn alloys, Shroeder and Wayman l~ indicate that TWSM results from the preferential formation and reversal of a "trained" variant of martensite; the trained variant results in a single orientation where initially there were four orientations. After only two SIM training cycles, they were able to achieve a strain upon cooling equal to that during superelastic loading. In other words, the specimen has transformed to a single crystal of the preferred orientation for maximum strain upon cooling or stressing. There is reason to believe that, with certain limitations, this same tendency is exhibited by polycrystalline material. Some years ago, Delaey and Thienel H presented very con- vincing evidence in the form of a film sequence which documents the changes occurring in the microstructure of a polycrystalline CuZnA1 sample as it is subjected to up to ten 'combined' training cycles (see Figure 1). A particular martensite plate arangement is observed during the induced displacement (SIM) phase of training, and after ten training cycles the microstructure induced by cooling only is virtu- ally identical. The conclusion again is that nucleation of a particular martensite plate variant is responsible for the spontaneous shape change associated with TWSM.

These observations do not allow us to understand how, on the submicroscopic scale, training can lead to the development of a preferred arrangement of martensite plates. Delaey and co-workers, in earlier work on polycrystalline alloys, 12 noted that stressing above Ms and subsequently cooling below 1t4: "freezes" in a particular microstructure. This will result in large internal stresses at grain boundaries and around tapered ends of martensite plates, and upon heating, the stored elastic energy which is released may assist the martensite-to-parent transformation. Delaey further notes that local plastic deformation will cause residual stresses which may influence the martensite microstructure obtained in cooling the sample during training manipu- lations. Similarly, Perkins 8 noted that elastic back stresses, arising from plastically deformed local regions or arrays of dislocations, may partly explain TWSM.

The obvious question is whether there are physical features in the microstructure which are associated with TWSM behavior. Is it possible, for example, that successive training cycles induce additional dislocation substructures which may act as nucleation points for the preferred martensite? Otsuka and Shimizu 7 have reported that dislocations assist the nucleation of particular martensite variants and result in lowering the critical stress for inducing martensite. It is clear that TWSM results from the development of a particular variant of martensite, and there is indirect evidence that dislocation tangles resulting from plastic deformation play a role in nucleating these preferential martensite variants by providing local residual stresses which assist the transformation of fl to martensite. Furthermore, once the particular martensite variant is formed by temperature change,

there are back stresses associated with the plate arrangement which will assist the martensite to fl transformation and associated shape change.

In order to gain a deeper understanding of the submicrostructural changes accompanying TWSM training, a program of transmission electron microscopic examination was carried out in the present work. The microstructure of the untrained alloy is shown in Figure 6. The parent phase, which at room temperature is about 20 K above P:, contains no residual martensite and no visible dislocation substructure and exhibits the typical mottling or tweed-like appearance often seen in these alloys. 13 The substructure developed after just a few training cycles is illustrated by Figure 7. It is apparent that significant changes in the substructure have occurred. An increased dislocation density is seen, which often takes the form of a set of dark streaks or bands. Initially it was thought that these must be areas of incomplete reversion of martensite plates, but closer analy- sis shows that these "vestiges" of the martensite plate structure are areas which are completely reverted to the crystal structure of the parent phase but with a residual substructure. A pattern seen quite often in the trained specimens consists of a set of these dark linear regions lying parallel to one another. It is obvious that martensite plates resided in such regions when the alloy was under stress or at low temperature, and upon reversion to the parent phase, vestiges of the plates in the form of dislocation tangles and local residual stress fields remain.

The microstructure after 15 training cycles, which is approximately at the saturation point noticed in the training data (Figures 3, 4, 5), is exhibited in Figure 8. Again we can see clearly where martensite plates formerly resided. We may see regions ripe for martensite plates to form, under stress or temperature change; that is, the martensite plates would grow into places now provided for them. We sometimes see, near plate areas, parallel markings running the length of the prior plate. In highly cycled specimens of polycrystalline CuA1Ni alloy, Ritter, et al., 14 found similar markings near receding martensite plates. These "vestigial marks" apparently provide a physical guide for the martensite plates to grow and shrink. In Figure 8, parallel arrays

Fig. 6--Microstructure of the parent phase in the untrained Cu-Zn-A1 alloy.


Fig. 7--Microstructure of the parent phase after several training cycles. (a) to (d) 3 cycles, (e) to (f) 5 cycles.

of dislocations are evident in the background, both between and through the plate areas.

This pattern becomes more clearly defined as training increases; the microstrncture after 20 cycles, which is beyond the saturation points in the training data, is seen in Figure 9. Again, closely spaced light and dark regions are evident; in these more highly trained specimens the bands are closely spaced and more prevalent. In addition, we now occasionally observe bona fide retained martensite plates. (It should be noted that the very fine "dashed" substructure evident in the background of these photomicrographs is considered to be part of the normal "tweed" structure of the parent phase.)

The microstructure after 30 training cycles, well beyond the saturation point in the training data (Figures 3, 4, 5), is

Fig. 8--Microstructure of the parent phase after 15 training cycles.

METALLURGICAL TRANSACTIONS A

Fig. 9--Microstructure of the parent phase after 20 training cycles.

VOLUME 15A, FEBRUARY 1984--319

Fig. lO--Microstructure of the parent phase after 30 training cycles. (See text for description.)

illustrated in Figure 10. Several views are provided because, although the microstructure has stabilized as well, i.e., is not changing with further cycling, it is not uniform. The most characteristic features are still the dark vestigial ridges, which represent the "ghosts" of some of the martensite plates. In Figure 10(a) a rather cluttered arrangement of these linear features is seen, while in other cases, such as Figure 10(b), a stack of aligned streaks is seen. These arrangements probably correspond in the fully developed martensitic microstructure to the typical self- accommodating groups of plates and alternating stacks of plates, respectively. In Figure 10(c) another characteristic plate arrangement is reflected in a V-shaped parallelpiped arrangement of vestigial streaks. At the relatively low mag- nifications represented by Figures 10(a), (b), (c), there is no obvious strain field or substructure detected around these linear features. However, it is notable that there are not enough of these features to correspond to all the martensite plates in the fully developed microstructure. Therefore, apparently only some of the plates in the stress-induced trained microstructure are leaving these features behind after reversion. Logically these may be those plates for which the net strain vector is least compatible with the applied strain.

If these vestigial ridges are examined at higher magnification, particularly in the vicinity of extinction contours, it is seen that they have a distinct substructure. In Figure 10(d) a striated structure is revealed in the contour, while in Figure 10(e) the substructure of the vestigial band is seen and its interaction with the contour is seen at the right. A higher magnification view of a portion of this band is shown in Figure 10(f). Again, it is clear that not every plate is

leaving debris of this sort. What we seem to have developed is a sort of microscopic "parking lot" for martensitic plates. At the point where we are observing it (above PI) most of the "cars" have gone home. Furthermore, this is a very poorly marked parking lot; not all the spaces are marked. But if even a few slots are designated, this will, as in a real parking lot, have an effect of the overall arrangement of cars (plates). That is, you don't have a vestigial marking for every plate, but those which you do have effect the subsequent positioning of the others.

A general pattern emerges from the TEM data. In the early stages of training, individual and quite distinct dislocations are generated. Additionally, there may be residual stresses remaining in the parent phase as the stress induced and thermoelastic martensite plates" recede on heating. As training proceeds, additional dislocations are generated, which tangle and are sometimes difficult to distinguish with- out the aid of increased magnification. Whereas one might expect these dislocations to inhibit the formation of stress induced martensite, it has already been established that the peak stress decreases as cyclic training proceeds. It is prob- able, therefore, that these dislocation tangles are either neu- tral toward, or assist the martensite transformation. Areas of vestigial markings and of bona fide retained martensite are developed in more highly trained alloys. In the case of the vestigial markings, it is clear that these define the pattern in which the martensitic microstrncture will deploy itself as stress is increased or temperature decreased below M~. That is, the submicrostructure contains distinct physical features which cause the microstructure to accommodate a particular variant of martensite on cooling, thus explaining TWSM.


The refinement, improvement, or increased definition of the substructure with cycling accounts for the ability of specimens to exhibit increased TWSM with increased training.

IV. SUMMARY AND CONCLUSIONS

This study has demonstrated that polycrystalline CuZnAI alloys are capable of improving their TWSM through "training." It is apparent from the data that one could expect to achieve displacements approaching 40 to 50 pct of the "trained" displacement during TWSM, provided the specimen was subjected to a sufficient number of prior training cycles. A sufficient number is considered to be approximately ten to fifteen based on the data generated in this study. A number of factors appear to contribute to the improvement of TWSM with training, as follows:

1. Plastic Deformation: Based on the decreasing amount of induced displacement (Table II) and the TEM evidence of increased dislocation density with cycling, it is apparent that light plastic deformation assists the training process and results in improved TWSM. It is concluded, therefore, the initial plastic deformation in the specimens studied in this work was a useful and necessary prerequi- site to TWSM.

2. Retained Martensfe: As training progresses, the specimens retain some amount of martensite. Dislocations generated during training may inhibit the reversion of martensite to parent/3 phase. During subsequent cycles, this retained martensite does not have to be induced by stress or temperature and thus the training process becomes easier, as evidenced by the decreasing trend in peak stress. While not verified directly by this study, it seems likely that the retained martensite serves as nucleation points for the thermoelastic martensitic transformation during TWSM.

3. Substructural Refinement: The combination of light plastic deformation and retained martensite leads to subtle substructural changes whose end result is the adjustment of the parent matrix in such a manner as to nucleate and grow a particular and preferential variant of martensite. This substructural adjustment continues and becomes more refined with cycling.

Therefore, based on this research, it would be expected that optimum training for TWSM displacements will be achieved when the alloy, initially in the parent phase, is

subjected to some amount of plastic deformation during the first cycle, followed by approximately fifteen training cycles. These training cycles should encompass displacements 100 pct greater than the desired TWSM cooling displacement. Of course, alloy selection should be based upon temperature requirements allowing training to be conducted between M I and Pl.

ACKNOWLEDGMENTS

This work was sponsored by the Division of Materials Research of the National Science Foundation through Grant DMR-81-08407. Particular thanks are due to Dr. J.J. Rayment for the insights gained from nearly three years of close collaboration.

REFERENCES 1. J. Perkins: Material Science and Engineering, 1981, vol. 51, p. 181. 2. J. Perkins, G.R. Edwards, C.R. Such, J.M. Johnson, and R.R.

Allen: Shape Memory Effects in Alloys, J. Perkins, ed., Plenum, New York, NY, 1975, p. 273.

3. J. Perkins: Metals Forum, The Journal of the Australasian Institute of Metals, 1981, vol. 4, p. 153.

4. S. Kajiwara and T. Kikuchi: Acta Metall., 1982, vol. 30, p. 589. 5. J. Perkins and W.E. Muesing: MetaU. Trans. A, 1983, vol. 14A,

p. 33. 6. A. Nagasawa, K. Enami, Y. lshino, and S. Nenno: Scripta Metall.,

1974, vol. 8, p. 1055. 7. K. Otsuka and K. Shimizu: Proceedings of an International Confer-

ence on Solid --* Solid Phase Transformations, H. I. Aaronson, D. E. Laughlin, R.E Sekerka, and C.M. Wayman, eds., TMS-AIME, Warrendale, PA, 1982, pp. 1267-86.

8. J. Perkins: Scripta Metall., 1974, vol. 8, p. 1469. 9. L. Delaey, J. Van Humbeeck, M. Chandrasekaran, J. Janssen,

M. Androde, and N. Mwanta: Metals Forum, the Journal of the Australasian Institute of Metals, 1981, vol. 4, p. 164.

10. T.A. Schroeder and C. M. Wayman: Scripta Metall., 1977, vol. 11, p. 225.

11. L. Delaey and J. Thienel: Shape Memory Effects in Alloys, J. Perkins, ed., Plenum, New York, NY, 1975, p. 341.

12. L. Delaey, M. Chandrasekaran, W. DeJonge, R. Rapacioli, and A. Deruyttere: Annual Report, INCRA Project No. 238: Shape Memory Effect and Internal Damping in Cu-Al and Cu-Zn-Based Alloys, International Copper Research Association, New York, NY, April 1975, p. 20.

13. L. Delaey, J. Perkins, and T.B. Massalski: J. Mater. Sci., 1972, vol. 7, p. 1197.

14. A. Ritter, N.Y.C. Yang, D. P. Pope, and C. Laird: Metall. Trans. A, 1979, vol. 10A, p. 667.

15. J. Perkins: Metall. Trans. A, 1982, vol. 13A, p. 1367.


Documents

Stress-Induced Martensitic Transformation Cycling and Two-Way Shape Memory Training in Cu-Zn-Al Alloys