A PIECE OF METAL THAT has been slightly deformed by a blow or a strain will return to its original shape when the pressure is released. Most metals will not do that, however, if the deformation is severe. An exception is found in certain unusual alloys that retain a "memory" of their original shape even after severe deformation. When an object made of a "shape memory" alloy is deformed and is later heated above a certain temperature, it returns to its original shape. Its returning motion may be strong enough to push or pull something else. In other words, the re-forming object can do work.

Can a useful engine be based on this principle? Several workers have tried to build thermal engines with working parts made of shape-memory alloys. A few engines of this type, based on simple designs, are now on the market. David Johnson of Berkeley, Calif., studies such engines and their alloys. Recently he sent me a manuscript describing how to construct two of the engines. He has also offered to supply sample wires made of the shape-memory alloy called Nitinol. (The name comes from the nickel and titanium of the alloy and from the U.S. Naval Ordnance Laboratory, where the alloy was first studied.)

A shape is implanted in Nitinol wire by annealing it: the wire is held in the wanted shape and brought to a high temperature. Suppose the wire is formed into a U shape and heated. After it cools it is twisted into a new shape, say an S. When the wire is then heated above a transition temperature (which depends on its composition), it returns to the U shape.

Nitinol can also develop a memory of its shape when it is below the transition temperature if it is put through a "training" procedure. The wire is repeatedly heated to a point above the transition temperature, allowed to regain its annealed shape (say a U) and then cooled below the transition temperature. Each time it is cooled it is forced into a particular second shape, say the S again. After several cycles of heating and cooling the wire begins to remember the S. Thereafter it will form into an S whenever it is cooled below the transition temperature. This memory changes somewhat the memory of the annealed shape; it is also weaker than the memory of the annealed shape.

One of Johnson's designs for a Nitinol thermoturbine is shown in Figure 1. Pulleys A and B are fixed to a shaft. A's radius exceeds B's by 30 percent. Two other pulleys, called idlers, are immersed in water baths, one hot and one cold. Running around the array of pulleys is a continuous length of Nitinol wire that has been coiled in a tight helical spring. Pulleys A and B are grooved to keep the spring from slipping.

When Johnson's apparatus is properly made, the spring pulls itself around the pulleys, forcing the shaft on which A and B are mounted to turn rapidly. The motion continues for as long as the water baths are kept hot and cold.

The motion results from the changes in the Nitinol as it passes through the water baths. In the hot water it heats above the transition temperature and contracts, pulling strongly on A and B. In the cold water it cools below the transition temperature and expands. Now its pull on A and B is weak.

Consider the forces on A and B as the spring contracts in the hot water. Note that the spring is wrapped around the pulleys in opposite directions. Hence the force on B is toward a counterclockwise rotation and the force on A is toward a clockwise one. The competition is not a stalemate because the rotation of an object depends on the torque, rather than just the force, on the object. Torque is the product of the force and a lever arm. The lever arm for the pulley is its radius. Since A's radius is larger than B's, the torque on A dominates the rotation. The shaft turns clockwise.

The spring in the cold-water bath also pulls on A and B. Again A has the larger torque because of its larger radius. The result should be a counterclockwise rotation of the pulleys and the shaft. Remember, however, that the forces from the expanded spring in the cold water are weaker than the forces from the contracted spring in the hot water. The net motion of the pulleys and the shaft is therefore clockwise.

Several precautions are necessary to ensure that the device functions properly. A and B must have steep grooves (at an angle of at least 60 degrees) to prevent the spring from slipping. The shafts of all the pulleys must be mounted on ball bearings to reduce friction. The idler pulleys must be far enough from A and B to stretch the sections of spring between them by about 50 percent each.

The Nitinol wire is.02 inch in diameter. Make the spring by wrapping the wire in a tight spiral around a metal rod 3/16 inch in diameter. Heat the wire and rod to a temperature of about 520 degrees Celsius, at which the array glows a dull red. This step anneals the wire, so that whenever it is heated above its transition temperature by the hot water, it tends to regain the shape it had on the rod. When the engine is run and the spring passes through the cold-water bath several times, it becomes trained to a second shape: a looser helix. Thereafter the spring has a memory of two shapes, the tight helix when it is above the transition temperature and the loose helix when it is below it.

The memory in Nitinol arises from the types of crystals that develop in it. They are sensitive to temperature and external stress. When the wire is above its transition temperature, it is in what is called the parent phase. If it is cooled below the transition temperature, large groups of atoms rearrange themselves to form a crystal structure sometimes called quench-induced martensite. (The name is misleading because it implies a need for rapid cooling.)

If the cooled wire is put under strain by stretching, some of the martensite is transformed into what is sometimes called stress-induced martensite. Greater stress leads to more transformation. If the wire is then heated above the transition temperature, the atoms change from the martensite formations to the parent phase. All these transformations are said to be diffusionless, meaning they involve a large-scale reordering of the atoms rather than a diffusion of individual atoms through the crystal structure.

The parent phase consists of cubic crystals and the martensite phase consists of needlelike crystals that collect in small domains. Within a domain the crystals are aligned. One explanation for the different types of martensite derives from the relative alignment of the domains. When Nitinol under no stress is cooled from the parent phase, the martensite domains are randomly oriented. This arrangement characterizes quench-induced martensite. If the wire is then stretched enough to distort its shape, some of the domains are reoriented so that they point along the direction of stress. The reorientation allows the wire to accommodate for the change in its length. Greater strain leads to more extensive alignment and thus to a greater amount of this type of martensite.

When the wire is heated above its transition temperature, the needlelike crystals of martensite change into the cubic crystals of the parent phase. Because the cubic crystals and the needlelike ones have different shapes, they do not fit into the same space. The new crystals form under strain and have strain energy stored in them. They move to change their positions and orientation in order to relieve the strain and release the stored energy. Where the martensite domains are randomly oriented this attempt to relieve strain does not result in any net movement of the wire, but where the domains are aligned the relief from the strain forces the wire to undo the effects of the previous stretching and deformation. By this action the wire regains its annealed shape.

One generates work by reheating the wire while it is under a moderate strain produced by external tension. As the wire passes the transition temperature, conversion to the parent phase again yields cubic crystals under strain. As before, they tend to relieve the strain by returning the wire to its annealed shape. This time the wire must push or pull against the external tension, thereby doing work. The energy for the work the wire does comes from the strain energy stored in the cubic crystals as they form.

At present no one fully understands why Nitinol can be trained to remember a second shape when it is below the transition temperature. One hypothesis is that the repetition of heating and cooling and the related crystal transformations introduce defects into the crystal structure. Each defect creates stress in the crystal. When a trained wire is cooled below its transition temperature, the stresses arising as a result of the defects force the wire into its trained shape.

Another explanation for training assumes that martensite cooled from the parent phase can exist in several different formations. Suppose there are only two such formations and one of them grows faster if the wire is under compression when it is cooled whereas the other grows faster if the wire is under tensile stress.

If the alloy is under no stress when it is cooled, the two types of crystal grow at the same rate. Equal amounts develop in the cooled wire. They each introduce strain into the wire, but since they are equal in amount, the wire is under no net strain that would alter the shape it had before it was cooled.

If the wire is under compression when it is cooled, one type of crystal grows faster than the other; the relation is reversed if the wire is under tensile stress when it is cooled. Suppose the wire is constrained to be bent when it is cooled. The compression type of martensite forms along the inner part of the bend and the elongation type forms along the outer part. When the wire is repeatedly heated and cooled while it is being held in a bent form, the two types begin to form automatically in their places in the bend. Eventually the process becomes so automatic that the constraints are no longer required. When the wire is cooled, one side develops the compression type of martensite and the other side develops the elongation type. The strain introduced by the crystals on each side of the wire bends the wire. It is then trained.

To follow the transformations of crystal structure in Johnson's engine I mentally labeled a small section of the wire spring as x. I also drew a phase diagram of stress v. temperature. As x travels around the engine, it also moves along the path shown in the phase diagram When it changes to or from the parent phase, it crosses the line separating the parent phase from the martensite formations. A complete cycle of x around the engine corresponds to a complete trip along the path in the phase diagram.

When x moves from the cold-water bath to pulley A, it is at a low temperature because it has just passed through the cold water. It is under only a moderate strain because the segment of the spring currently in the cold water is expanding. Hence x consists of a mixture of quench- and stress-induced martensite. This state corresponds to the part of the path at the lower left in the diagram.

When x passes over A, the tension on x increases because the segment of the spring then in the hot-water bath is contracting. The resulting increase in strain orients more of the martensite domains in x. This state corresponds to the part of the path at the upper left in the diagram.

On reaching the hot-water bath x is heated above the transition temperature. The martensite transforms to the parent phase. The cubic crystals form under a severe strain and store much strain energy. They relieve the strain and reduce the stored energy by making x contract. Because x is connected to the rest of the spring, it pulls on the segments of Nitinol wire between itself and the grooved pulleys. The unbalanced torques on the pulleys give rise to rotation.

The success of the engine depends on the ability of x and the other sections of the wire to do work when they are in the hot water. If just prior to the bath the tension on x is low, few of the martensite domains are oriented. When the needle crystals convert into cubic crystals, the attempt to contract x is small. Hence the force on the grooved pulleys is small. You might be tempted to move the idler pulley farther from A and B to increase the stretch of the spring and the stress on x. This step entails a disadvantage, which is that the tension in the spring is then so great that x cannot regain its annealed shape when it is in the hot water. It gradually begins to lose that memory.

When x passes around pulley B, the tension on x relaxes because the segment of the spring then in the cold-water bath is expanding. This state corresponds to the lower-right part of the phase diagram. (I am ignoring the fact that the wire is already cooling when it reaches B because of the surrounding air.) When x reaches the cold water, it is cooled and the parent phase is largely transformed to quench-induced martensite. Some of the parent phase is also transformed to stress-induced martensite because x is still under some tension even though it has expanded. This state corresponds to the lower-left part of the phase diagram. Thereafter x repeats the cycle.

Another engine employing Nitinol consists of a hub to which spokes of steel piano wire are attached [see Figure 3]. The hub can be a plastic wheel from a toy The sections of piano wire are about .03 inch in diameter and seven inches long. Attach them to the hub with epoxy glue. Helically wound sections of Nitinol wire run between the outer points of the spokes. Tie the Nitinol wire to the spokes with short lengths of copper wire. When the sections of the spring are in place, each one is stretched so that its coiled length is increased by about 20 percent. Attach quarter-ounce fishing weights near the outer ends of the spokes. Balance the wheel by adding dabs of solder to the weights. An unbalanced wheel will not turn freely on a horizontal shaft.

To set the engine in motion direct steam from a teapot onto a spring that is on each side of the wheel near the top. Once the spring has been heated above its transition temperature it contracts and pulls on the adjacent spokes. The motion shifts weight into the region being heated, thereby upsetting the balance of the wheel. As the heavy section falls, the wheel rotates. When the heated spring moves out of the steam, it cools and expands, restoring the shape of the spokes. Meanwhile another spring moves into the steam, contracts and pulls weight toward it The cycle continues in this way. The wheel goes on rotating as long as steam is directed onto the springs. The speed of rotation is determined by the cooling rate of the springs after they leave the steam.

A sample of Nitinol wire can be obtained from Johnson at TiNi Sales, P.O. Box 1 431, Lafayette, Calif. 94549-1431; send $3 and a self-addressed, stamped envelope.

Bibliography

ON THE THERMODYNAMICS OF THERMOELASTIC MARTENSITIC TRANSFORMATIONS. R. J. Salzbrenner and Morris Cohen in Acta Metallurgica, Vol. 27, No. 5, pages 739-748,- May, 1979.

SHAPE-MEMORY ALLOYS. L. McDonald Schetky in Scientific American, Vol. 241, No. 5, pages 74-82; November, 1979.

The Society for Amateur Scientists (SAS) is a nonprofit research and educational organization dedicated to helping people enrich their lives by following their passion to take part in scientific adventures of all kinds.