Keywords

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

1 Introduction

Shape memory materials have excellent mutifunctional properties, centered in shape recovery by heating after severe pseudo plastic deformation in the martensitic state [1]. Other functional characteristics that they display are pseudoelastic [2] and superplastic behavior, biocompatibility [3], etc. The performance and large energy density stored in such alloys are very interesting parameters for use in actuators, including MEMS [4]. The main features of these materials rely on the Martensitic Transformation, MT [5, 6], a first order, displacive, thermoelastic transformation from a high-temperature, high-symmetry phase (Austenite or parent phase) to a low-temperature, low-symmetry (Martensite) phase. Arising from the symmetry difference between parent and martensite phases, a single crystal of the parent phase spontaneously deforms giving rise to a set of different oriented martensite crystals called variants. In the case of no external forces applied during the transformation, the variants’ distribution reflects the effort of the martensite to accommodate the large strains associated with the crystal lattice change between austenite and martensite, which can reach values over 10 % in many cases. The variant distribution after the transformation is therefore dictated by the need of keeping the same external shape as in the austenite. Variants can be rearranged by external stress resulting in large pseudo plastic deformations in the martensite. During heating on the way back to austenite, however, all variants transform to a single austenite crystal, so that the initial shape is recovered.

The prototype of SMA is NiTi close to the equiatomic composition. It is known by its trademark name Nitinol® as developed in the Naval Ordnance Laboratory (USA) [7]. Other SMA are Ag-Cd, Au-Cd, Cu-Al-Ni, Cu-Sn, Cu-Zn-(X), etc.

Some of the shape memory alloys are ferromagnetically ordered, in which the interactions between lattice and magnetic subsystems give rise to new advanced properties as compared to the nonmagnetic counterparts. Those are called magnetic shape memory alloys (MSMA), and were studied for the first time in the late 1980s by V. A. Chernenko and V. V. Kokorin [8, 9]. In 1996, a deformation of 0.2 % in a Ni-Mn-Ga single crystal under a magnetic field of 1 T was reported by K. Ullakko, R. C. O’Handley et al. [10]. In this kind of materials, large strains can be triggered by a magnetic field at a constant temperature in the martensitic phase.

The prototype and most studied Magnetic Shape Memory Alloys (MSMA) are the Heusler compound Ni2MnGa and its off-stoichiometric analogs. Particularly, off stoichiometric alloys with martensitic and Curie temperatures above room temperature are currently used for applications. Nevertheless, many other FSMAs such as Fe-Pd [11], Fe-Pt [12], Ni-Mn-Al [13], Co-Ni-(Ga,Al) [14], and Fe-Ni-(Ga,Al) [15] have been subject of intense studies as well.

Thin films of SMAs and MSMAs have been widely studied regarding the preparation method [16, 17], substrate [18, 19], internal stress [20, 21], crystal structure [22, 23], and magnetic properties [24, 25]. These films are interesting for MEMS because of the easy integration into microelectronic circuits and the possibility of mass production with traditional microelectronic technology. An additional figure of merit of SMA is the large power density, over 10 W/g [26], that they can develop in actuators. This value is much larger than normal piezoelectric [27] and magnetostrictive [28] devices and therefore a large number of micro-device applications of SMAs are forecast in the near future.

The martensitic transformation in SMAs and MSMAs is easy to detect by calorimetric methods, through the associated latent heat [29]. DSC and DTA are currently used to detect the transformation, which expands over several degrees. In some cases, quite sharp peaks are obtained both during heating and cooling. In other cases, however, the transformation is broad and peaks are smeared and can be hardly identified as in the DSC scans shown in Fig. 23.1 that we measured for a Mn50Ni35Sn10Fe5 alloy.

Fig. 23.1
figure 1

DSC scans (endotherm up) for a MSMA alloy of composition Mn50Ni35Sn10Fe5. The broad peak on heating reveals the martensitic transformation, but on cooling it is not visible because of the much broad extension

Full size image

The heating and cooling rates in standard DSC measurements are close to those available in bulk actuators, but much lower that the heating rates expected for MEMS operation. In the later one expects a much rapid heating under a high-intensity electric current (Joule heating) or LASER pulses, to achieve fast actuation. Double beam cantilevers or bridges must be used for Joule heating, whereas single beam cantilevers (Fig. 23.2) are enough for LASER heating. Features of the martensitic transformation at high heating or cooling rates are essential to design microactuators as the mentioned MEMS. In this work we report the study of the martensitic transformation in a magnetic shape memory alloy thin film directly deposited onto the Flash DSC chip by means of Magnetron sputtering.

Fig. 23.2
figure 2

Over 10 μm displacement can be obtained in a Si cantilever of 150 × 20 × 5 μm, coated with 3 μm NiMnGa film, and heated through the martensitic transformation. Rapid heating by means of a LASER beam will allow the deformation to proceed in milliseconds

Full size image

2 Experimental

The alloys used in this study had the bulk composition Ni50Mn28Ga22 and were prepared from pure constituents by induction melting. To produce a sputtering target, casting was done in a special-shape mold, 2 in. in diameter 3 mm thick. The films were deposited, from such a target, in a Pfeiffer Vacuum Classic 500 sputtering system by High field DC magnetron sputtering. Details about the alloy preparation, deposition parameters, and procedure can be found in [30]. Deposition was done directly onto the Flash DSC chip. Special masks were used to avoid deposition in the reference site. The metallic mask depicted in Fig. 23.3 was first used in the back of the chip and also on the front, as discussed below. When used in the front, a small Copper shield was used to further protect the reference. The chip was attached to a substrate holder that can be heated up to 500 °C in order to promote the desired film structure.

Fig. 23.3
figure 3

Left: Schematic mask arrangement for sputtering deposition of the films onto the DSC chips. The Copper shield was used to further protect the reference site from film deposition. Right: Actual metallic mask made out of stainless steel

Full size image

Witness films were simultaneously deposited onto semiconductor grade Si with a native SiO2 surface layer. These witness films were used to characterize the composition, structure, and transformation properties of the films by traditional methods, including EDX, X-ray diffraction, and electric and magnetic measurements to determine the transformation temperatures. Compositions of the films are listed in Table 23.1. They differ from the target composition because the sputtering process favors some elements over others, depending on the deposition parameters. In most cases, a 10 nm thick Cr layer was deposited prior to the alloy, in order to increase the adherence. This buffer layer didn't introduce any perturbation in the film performance, as tested in the witness films.

Table 23.1 Characteristics of the samples and Flash DSC response
Full size table

The transformation was studied at ultrahigh speed in a Mettler-Toledo Flash DSC, at the Institute of Physics of Materials (University of the Basque Country-CSIC) in San Sebastian, Spain.

3 Results

Several films were deposited at different substrate temperatures and with the mask either on the back or the front of the DSC chip. They are summarized in Table 23.1 together with the adherence and the existence of the correct crystalline structure, denoted by the martensitic transformation in the case of MSMAs which was deduced from resistivity measurements. The last column indicates the performance in Flash DSC experiments. Witness films were tested in all cases to determine martensitic transformation and Curie temperatures. MSMA films not presenting the martensitic transformation were not run in the DSC.

In the first series of films the mask was placed on the back of the chip as close as possible to it. Some previous trial depositions were carried out at different temperatures and conditions. Bad adherence was found in all films deposited directly onto the DSC chip (sample #1) and the films were seen to peel out spontaneously. A thin layer (10 nm) of Cr circumvents this issue, and such a buffer layer was used in all subsequent depositions [31].

As seen in Table 23.1, no transformation was found in MSMA films deposited below 500 °C, even after in-situ annealing in the calorimeter. The absence of transformation for films deposited at room temperature was already known from previous experiments [30], but sometimes annealing is enough to recover the performance of the films. This was not the case here, since the maximum achievable temperature for the calorimeter was not enough to promote the atomic crystalline ordering. Such heat treatment would need conditions of no less than 800 °C and high vacuum [20], therefore the more feasible approach of high-temperature deposition was chosen.

The only transforming film prepared with the mask placed on the back of the chip is number 4 (see Table 23.1) deposited at 500 °C. The magnetic and martensitic transformations, as studied in the witness film, are represented in Fig. 23.4. The martensitic transformation is clear, but it extends over a very broad temperature range of almost 140 °C. This would probably prevent the detection of the transformation by DSC as the expected broad peak would be probably hardly discerned from the base line. However, the main problem consisted in the metallic deposit on the back of the DSC chip, or the heating to 500 °C, which produced a short circuit of the thermopile wires that made impossible to obtain any reliable signal from the Flash DSC apparatus.

Fig. 23.4
figure 4

Martensitic (upper part) and magnetic (lower part) transition in the film #4 as determined by resistivity and magnetic measurements, respectively. The austenite start (As) and finish (Af) temperatures, as well as the Curie temperature (Tc), are shown in the plot

Full size image

To avoid such issues, new films were deposited on the front of the DSC chip with the same mask. This time, however, the distance mask/substrate was large, increasing the risk of the deposited film to attain the reference site in the chip (see Fig. 23.3). This is true for sputtering, because the technique is not very directional but rather isotropic. To take care of this problem a Copper shield covering the reference site was incorporated (Fig. 23.5). A further test was carried out on the chip by heating up to 500 °C without any deposition to rule out heating effects on its performance. Such a test was successful.

Fig. 23.5
figure 5

The film deposited onto the Flash DSC chip, with the mask placed in the front. The sample site is fully covered with the metallic film, which extends somewhat into the reference site. The Copper shutter was not fully effective in protecting the reference, but the reference thermopile is clean

Full size image

The new series of samples were produced using different sputtering parameters, such as lower Ar pressure and sputtering power. This resulted in a change of the films composition increasing the Ni content and decreasing the Mn one (see Table 23.1). Film thickness was reduced to 1 μm. Influence of sputtering parameters on the film composition is discussed in [30, 32].

The new composition of the films (Ni54Mn24Ga22) results in a quite similar magnetic transformation temperature (Tc ≈ 83 °C) but the martensitic transformation is now much sharper. It extends from about 30 to 100 °C, as determined by resistivity measurements. It must be noticed that even a slight decrease of the substrate temperature to 450 °C destroys the transformation (sample 5 in Table 23.1).

The structure of the film at room temperature was studied by X-ray diffraction, revealing a polycrystalline tetragonal 10 M martensite with lattice parameters a = 5.90 Ǻ and c = 5.53 Ǻ. Figure 23.6 shows a SEM micrography of the film, where the austenitic crystal grains can be seen and their average diameter can be estimated as 100 nm. For the martensite, the grain size can be calculated using the Scherrer equation, which gives a value of 11 nm.

Fig. 23.6
figure 6

Scanning electron microscopy image in secondary electron mode of the NiMnGa thin film deposited on the DSC sensor. Crystal grains of about 100 nm diameter can be seen

Full size image

Film #6 was scanned by Flash DSC at 10,000 K/s and shows the transformation as a peak centered at about 50 °C (Fig. 23.7, upper). The peak is similar to that of bulk Mn50Ni35Sn10Fe5, depicted in Fig. 23.1. The only heating rate giving a good DSC scan was just 10,000 K/s, as higher rates produced too noisy curves and lower rates did not produce a high enough heat flow signal to detect the martensitic transformation. The latter effect can be appreciated from the cooling curve, measured at 4000 K/s not evidencing the martensitic transformation, although it does take place. The reversibility of the transformation was checked by repeating several DSC heating-cooling measurements in addition to the resistivity test, both revealing the reversible character of the transformation.

Fig. 23.7
figure 7

Martensitic transformation in film #6 (upper part) studied by Flash DSC at 10,000 K/s (endotherm up). A pure Ni film scan at the same heating rate is shown (lower part) for comparison. Nickel has no phase changes in the studied temperature range. Cooling rate was 4000 K/s in both cases. Three successive runs are superimposed in all measurements, giving confidence about the reversible and repetitive character of the transformation

Full size image

In order to rule out any artifact in the measurements, a test film of pure Ni (sample #7 in Table 23.1) was deposited with the same geometry at room temperature in another DSC chip for comparison. Nickel has no transformations in the temperature range studied, apart from its Curie temperature appearing at 360 °C. The Ni film scan is also plotted in Fig. 23.7, and does not show any features in the temperature range studied. This confirms the correspondence of the peak observed in the Flash DSC and the martensitic transformation of the alloy.

By subtracting the base line, a clean scan of the transformation is obtained (Fig. 23.8). The start and finish temperatures do agree with the witness film ones, and the central peak temperature too. Enthalpy of the film transformation was estimated by assuming the nominal thickness of 1 μm homogeneously extended all over the sample sensor (Fig. 23.5). The obtained value is 1.4 J/g, less than half the bulk value determined from the target. The latter is 3.8 J/g. Several factors can, however, explain the discrepancy. First, the composition of the target and film is not the same as shown in Experimental and Table 23.1, so the transformation enthalpy can be different too. Second, the alloy could contain some residual austenite as a consequence of the small grain size. This mass fraction does not undergo the martensitic transformation and therefore does not contribute to the enthalpy. Finally the thickness of the film deposited onto the DSC chip through the mask can be much lower than the witness film, since an important part of the incoming atom trajectories are not allowed by the aperture angle of the mask while all directions are possible for the witness. The thickness of the latter was determined by a profilometer with great accuracy, but the actual film onto the chip was not suitable for the thickness determination, so it could be thinner by a factor of two or so. Determination of the exact mass is always a problem in Flash DSC measurements. This problem is more difficult in thin film studies, because the corresponding reference material (so simple to obtain in polymers) is not available for “normal” DSC runs. In the present approach the sample mass has been calculated considering only the fraction of the film laying on the top of the sensor (rsensor = 2 × 10−4 m, Sample area = πrsensor 2) as the heat transfer between this part and the DSC sensor is faster than in any other because of its proximity. In the future a method for exact determination of the sample mass is to be found in order to have quantitative Flash DSC measurements.

Fig. 23.8
figure 8

Martensitic transformation in film #6, studied by Flash DSC, after base line subtraction (endotherm up). Start, peak, and end transformation temperatures are shown. The gray area was used to estimate the enthalpy change at the transformation, yielding ΔH ≈ 1.4 J/g

Full size image

4 Conclusions

NiMnGa Magnetic Shape Memory Alloy thin films have been deposited onto Flash DSC chips, preserving good functional properties. The martensitic transformation has been detected at high heating rates, similar to those intended in MEMS, in a Flash-DSC apparatus. More work is needed to achieve full Flash-DSC performance in metallic thin films with a low transformation enthalpy like the martensitic one. However, the technique is promising and will be extremely useful for the design of MEMS based on SMAs.