Phase Formation and Structural Sequence of Highly Oriented MBE Grown NiTiCu Shape Memory Thin Films

Full text

(1)Materials Transactions, Vol. 43, No. 5 (2002) pp. 933 to 938 Special Issue on Smart Materials-Fundamentals and Applications c 2002 The Japan Institute of Metals. Phase Formation and Structural Sequence of Highly-Oriented MBE-Grown NiTiCu Shape Memory Thin Films Ralf Hassdorf1 , Jürgen Feydt1 , René Pascal1 , Sigurd Thienhaus1 , Markus Boese2 , Tobias Sterzl1 , Bernhard Winzek1 and Michael Moske1 1 2. Center of Advanced European Studies and Research (caesar), D-53111 Bonn, Germany Institut für Anorganische Chemie, Universität Bonn, D-53117 Bonn, Germany. We present a study demonstrating the capability for controlled shape memory thin film growth using molecular beam epitaxy. Here, NiTiCu alloy films were grown which are known to exhibit the martensitic transformation well above room temperature. Remarkably, the microstructure of these films was found to be very different compared to conventionally sputtered polycrystalline films: here, the crystallites are highly oriented within ±3◦ along the film plane normal. Moreover, a splitting of the martensite orientation is detected indicating the selection of two specific sets of martensite variants. Mechanical stress measurements reveal a high ratio of recoverable stress even for films below 500 nm thickness. These results open up the possibility to specifically modify the microstructure and crystallographic orientation of shape memory thin films and thus suggest promising characteristics, especially in regard to their superelastic behavior. (Received March 11, 2002; Accepted April 15, 2002) Keywords: shape memory alloy, martensitic transformation, thin film, molecular beam epitaxy, phase formation. 1. Introduction. 2. Experimental Details. Shape memory alloys constitute a special materials category due to an inner structural transformation occurring under changing external parameters like temperature, strain, or mechanical forces which leads to a recovery of macroscopic deformation and to superelastic behavior. To date, bulk materials are employed mainly the classical NiTi-based alloys for applications in medical technology like tooth braces, cardiovascular stents, microgrippers, or endoscopes.1) The martensitic transformation occurring in these materials and its subsequent reversal are, however, of further practical interest, e.g., in miniaturized devices like electric or thermal switches.2) Therefore, thin films are currently under rapid development opening up the possibility for lithographic microstructuring.3) Molecular beam epitaxy (MBE) provides high variability regarding the elemental metallic components and compositions as well as controlled deposition on wafer-size substrates. In order to compare with previous results known for, e.g., sputtered NiTi-based films,4) here, NiTiCu alloy films were grown which are known to exhibit the martensitic transformation well above room temperature.5) After deposition, the films are amorphous and they crystallize upon post-growth annealing at around 450◦ C. On subsequent cooling, the transformation into the metastable distorted martensite structure occurs which has been studied in detail through reversible thermal processing in a cantilever beam stress apparatus and by X-ray diffraction in combination with a specially designed heating stage. From the structural point of view, the martensitic transformation in NiTiCu films is known to proceed in two steps:5) from parent austenite B2 phase (bcc) to the orthorhombic B19 intermediate structure and further to the monoclinic distorted B19 phase. Specially, in films with a Cu content near and above 10 at% the second-stage transformation appears to be sterically hindered by a strain-induced dislocation network6, 7) giving rise that the monoclinic structure is not observed.. 400 nm-thick-NiTiCu alloy films were grown onto thermally oxidized 4-inch single-crystal Si-(001) wafer substrates at 200◦ C in an ultrahigh vacuum system using MBE technique. The chamber pressure during deposition was below 1 × 10−6 Pa. The substrate was rotated with 30 rpm during the deposition process to maintain a constant film thickness along the overall wafer. Thickness variations as observed by a stylus profiler were less than 2% from center to wafer edge. Deposition rates were monitored by a quartz crystal microbalance and by using an atomic absorption spectroscopy unit.8) For the alloy composition on target, here Ti50 Ni40 Cu10 , the rates were 0.18 nm/s for Ti, 0.09 nm/s for Ni, and 0.02 nm/s for Cu. Specially, for depositing Ti, an effusion cell and alternatively an electron-beam evaporator were used. By means of energy-dispersive X-ray (EDX) analysis the element concentrations of the alloy films were examined and found to match the nominal ones by less than 2% deviation. Crystallization of the as-deposited amorphous films was carried out by insitu thermal treatment which involved a peak temperature of 650◦ C held for an interval of 1 min and constant ramp slopes of 20◦ C/min upon heating and cooling. As observed by simultaneous reflection high-energy electron diffraction crystallization occurs at around 450◦ C. The structure of the films was characterized by X-ray diffraction and cross-sectional transmission electron microscopy (CS TEM). The X-ray diffraction measurements were made in reflection geometry using Cu–Kα radiation. The instrumentation set-up comprised parallel beam primary optics (point focus), an Euler cradle, and a HiStar area detector allowing for a large coverage along the 2Θ and χ axis simultaneously. A specially designed heating stage capable to employ the area detector enabled us to study in detail the structural reorientations associated with the martensitic transformation. The stage features a high-temperature vacuum chamber mounted to the Euler cradle. It is capped with a.

(2) R. Hassdorf et al.. dome-shaped X-ray transparent Be window which provides access to almost the overall diffraction space of the film sample under investigation. To prevent the alloy films from oxidation upon heat exposure the chamber was evacuated and flushed by Ar gas at a dynamical pressure of 1 Pa. TEM cross-section specimen were prepared by means of standard mechanical thinning and low-angle ion-milling procedures. The cross-sections were examined in the diffraction and high-resolution imaging mode using 120 kV and 300 kV instruments, respectively, with EDX microprobe analysis. Surface analysis and depth profiling were carried out using in-situ X-ray photoemission spectroscopy (XPS) along with an Al-Kα source (1486.6 eV) and a hemispherical energy analyzer. The photoemission spectra were taken at an energy resolution of typically ∆E = 1 eV. The data were fitted according to a superposition of Gaussian and Lorentzian profile and corrected as for the electron mean free path and analyzer transmission function. A Shirley-type background was subtracted and the spectrum intensities were corrected by the elemental Scofield sensitivity factors.9) For the element concentrations, the numbers were rescaled so as to add up in total to 100%. The change in average film stress due to the martensitic transformation was monitored by ex-situ stress measurements employing a capacitance bending beam technique details of which are given in Ref. 10). By evaluating the sample curvature from three positions (capacitances) along the cantilever beam the stress was calculated using Stoney’s formula11) (here for df ds ): σ = E s ds2 /6R(1 − vs )df ,. (1). where ds and df are the thickness of substrate and film, respectively, E s denotes the in-plane Young’s modulus of the substrate, vs the Poisson ratio, and R the radius of sample curvature. The surface topography of the films was studied in situ using a variable-temperature scanning tunneling microscope (VT STM) with W tips in constant current mode. Prior to analysis the films were sputter-cleaned for 120 min at 2.0– 2.5 kV, 10 mA using high-purity Ar gas, and then annealed at 620◦ C under ultrahigh vacuum conditions. 3. Results and Discussion Stress measurements reveal a high ratio of recovery stress in the films under investigation. Figure 1 shows a typical stress-temperature curve of a Ti51 Ni38 Cu11 film on thermally oxidized Si after in-situ crystallization. The transformation from the high-temperature austenite to the low-temperature martensitic phase and vice versa is reflected by a significant change of the in-plane tensile stress of around 300 MPa. This value is by far smaller than the 400–600 MPa reported for sputtered films of similar composition.12) However, it clearly signals that a high fraction of the film volume is involved in the transformation, even for films far below 1 µm thickness. Moreover, the temperature range as well as the width of the transformation hysteresis fits very well the data found for sputtered films. Figure 2 shows the structural reorientation of a Ti51 Ni38 Cu11 film undergoing martensitic transformation as observed by. 300. Change in Tensile Stress (MPa). 934. MS = 61 C. 250 Af = 73 C. 200 150 100 50. Mf = 42 C. 0 AS = 54 C -50 20. 30. 40. 50. 60. 70. 80. 90. 100. Temperature ( C) Fig. 1 Stress-temperature curve of the transformation of a Ti51 Ni38 Cu11 film on thermally oxidized Si. Ms and Mf are the start and finish temperature, respectively, of the martensite formation. As and Af denote the start and finish temperature of the austenite formation.. X-ray diffraction. Remarkably, using MBE deposition technique, the microstructure of the film is very much different from conventionally sputtered films.5) Here, the crystallites are not randomly but highly oriented normal to the film plane with the orientation of the unit cell axes matching the coordinate system of the substrate, i.e., we do not detect the close-packed (111) lattice planes of the martensite or the corresponding (110) lattice planes of the austenite phase. As can be seen from the χ coverage the variation of the crystallographic orientations is limited to a small range within ±3◦ both in the austenite (left frame) and martensite phase (right frame). At 70◦ C only the austenite-(200) reflection is visible while upon cooling to room temperature the martensite-(200) and split (022) reflections become dominant. The Debye rings observed in the diffraction images are due to the Be dome used in connection with the sample heating stage. From the peak intensities of the reflections, the transformation hysteresis in terms of film volume affected by the transformation is derived (Fig. 3) and found to match the stress-temperature hysteresis (Fig. 1) in their overall behavior. Remarkably, the splitting of the martensite-(022) reflection along the χ direction as shown explicitly in Fig. 4 indicates the selection of two sets of martensite variants with opposite monoclinic distortion. It remains to be noted here that the underlying crystal structure is well ordered as can be seen from the occurrence of considerable first-order reflections ((100) and (011), not shown here). This point is of further relevance to XPS analysis described below. The microstructure of the films as observed by CS TEM reveals nearly columnar crystal growth parallel to the film normal extending in size almost over the entire film thickness (Fig. 5(a)). An intermediate layer of approximately 50 nm thickness is found formed at the interface to the oxidized Si substrate, identified by EDX microprobe analysis as Ti2 Ni, together with an additional thickness-dependent gradient in composition. The sharpness of the interface between substrate and intermediate layer indicates that obviously there is no interdiffusion of Ti and Si. High-resolution images and electron diffraction diagrams reveal the coexistence of tilted crystals showing a characteristic orientation order of 3◦ and 7◦ , respectively, relative to each other (Fig. 5(b)). This is con-.

(3) Phase Formation and Structural Sequence of Highly-Oriented MBE-Grown NiTiCu Shape Memory Thin Films. 935. Martensite. Austenite Be (200). (200). 2 (022). 70 C. Be. 50 C. RT. 1.0. 0.0. 0.8. 0.2. 0.6. 0.4. 0.4. 0.6. 0.2. 0.8. 0.0. 1.0. 20. 30. 40. 50. 60. Martensite Volume Fraction. Austenite Volume Fraction. Fig. 2 X-ray diffraction frame patterns of the martensitic transformation taken from a Ti51 Ni38 Cu11 film (2Θ range: 42◦ –77◦ , χ range: −69◦ –−111◦ ).. 70. Temperature ( C). Intensity (a.u.). Fig. 3 Transformation hysteresis of the martensitic transformation derived from integrating the peak intensities from the X-ray diffraction data shown in Fig. 2.. -105. -100. -95. -90. -85. -80. -75. (deg) Fig. 4 Martensite-(022) reflection. Intensities derived from peak integration along 2Θ.. sistent with STM results (see Fig. 6) exhibiting the contours of a distribution of well aligned crystals having a distribution of slightly tilted terrace orientations. Quantitatively, at the surface the average crystallite size is of about 120 nm in diameter with a corrugation length of 9.2 nm (0.8 nm for the smallest steps). These findings microscopically underline the overall characteristics observed by X-ray diffraction. However, no explicit martensite domain structure was found within the crystals. The depth profile as observed by XPS (Fig. 7) is at a first glance well consistent with the above EDX results. Initially,. however, the Ti signal shows a dramatic increase compared with the average concentration level resulting from a thin TiO2 layer at the surface which has been formed when exposing the sample to atmosphere. As for Cu, the initial concentration is significantly enriched by segregation following the crystallization process of the sample, a surfactant effect well known for Cu–Ni alloys.13) The photoemission signals taken from the 2 p3/2 and 2 p1/2 absorption edges, respectively, of the alloy constituents (Fig. 8) show explicitly that the segregation process proceeds via the Ni–Cu sublattice of the ordered bcc structure (see above). While the signal intensities for Ni and Cu behave reciprocally upon post-annealing of the film to 250◦ C the Ti signature remains unchanged (cf. Figs. 8(a), (b)). Regarding the Cu content, it must be noted that the concentration profile at the surface is expected to be much more complex than observed from the above XPS spectra. Normally, the enrichment in Cu due to segregation is restricted to the outermost layer.13) Here, however, following the restrictions in sensitivity typically, the information depth of XPS covers several atomic layers compositional variations cannot be resolved layer-by-layer. The phase sequence and the related compositional profile of the films as observed by EDX and XPS obviously are closely connected with the reaction scheme at the filmsubstrate interface. In particular, the strong chemical interaction between Ti atoms arriving from the vapor phase and oxygen from the Si-oxide barrier appears to control formation and growth of the thin intermediate layer identified here as Ti2 Ni phase. Remarkably, from the crystal structure and its lattice parameters this phase cannot be distinguished from oxygen-enriched Ti4 Ni2 O.14, 15) Even XPS does not provide further hints to this case since the energy shift of the Ti absorption edge arising from oxygen incorporation is estimated to be below the resolution limit of the energy analyzer. Nevertheless, the strong influence oxidized substrates regarding growth and microstructure of Ti thin films demands for discussion on the role that oxygen is possibly playing in the alloy films. Abermann and co-workers16, 17) investigated the formation of thin Ti-oxide films on alumina substrates. These films were then used as templates for deposition of pure Ti layers. From the stress evolution during the growth process and subsequent TEM analysis, they deduced two specific microstructural patterns: (i) formation of a polycrystalline Ti film through island growth on an as-deposited amorphous or nanocrystalline in-.

(4) 936. R. Hassdorf et al.. Normalized Atomic Concentration ( ). Fig. 5 (a) Inverted cross-sectional TEM dark field image of a nominally Ti50 Ni40 Cu10 film. Sequence of the phase compositions as determined from EDX microprobe analysis. (b) Diffraction pattern of the austenite phase. The arrow indicates an orientation perpendicular to the film plane normal.. 90 Ti Ni Cu. 80 70 60 50 40 30 20 10 0. 0. 100. 200. 300. 400. Sputter Depth (nm) Fig. 7. Fig. 6 STM topographic image of a Ti51 Ni38 Cu11 film at room temperature taken in constant current mode (Ugap = 0.1 V, I = 0.2 nA), scale 2×2 µm.. termediate Ti-oxide layer and (ii) epitaxial growth of a quasi single-crystalline Ti film on a highly-ordered TiO2 (rutile) layer formed after annealing at 400◦ C. A more detailed study showed that crystallization of the Ti-oxide layer starts at around 105◦ C which is significantly below the deposition temperature applied here (typically 200◦ C). Thus we can assume that, depending on the amount of oxygen from the underlying Si-oxide barrier reacting with the growing NiTiCu alloy film through interdiffusion, a thin crystalline well-oriented Ti-oxide layer is formed at the interface. Such behavior is confirmed by the observation of the (110) principal reflection of the TiO2 rutile structure in our films through X-ray diffraction patterns (not shown here). They also exhibit a strong confinement in the χ direction according to that found for the austenite and martensite phase. As for the Ti2 Ni phase formed on top of the TiO2 layer, the diffraction image reveals as well a sharp (333)/(511) prin-. Sputter-depth profile of a nominally Ti50 Ni40 Cu10 film.. cipal reflection spot (see Fig. 9), indicating here the specific orientation of the fcc-Ti2 Ni unit cell normal to the film plane additionally. Although the (333) and (511) orientation cannot be distinguished from each other due to their superposition along the 2Θ direction14) we assume that for pure symmetry reasons, here the quasi close-packed lattice plane, i.e. (333) orientation, is favored for the structural alignment. In view of the complex (Fd 3̄m)-type cubic crystal structure of Ti2 Ni with 96 atoms per unit cell14) it appears necessary to point out how the in-plane atomic configuration does look like with respect to the (333) orientation. It turns out that a number of lattice planes do exist with a normal aligned along the 111 direction which, however, due to the complexity of the crystal structure cannot be identified unequivocally. Notably, there is only one plane where the atomic sites are taken completely by Ti. This configuration is assumed to be the relevant one. It is illustrated in Figs. 10(a) and (b) based on a lattice parameter of 1.1438 nm for the Ti2 Ni-unit cell as determined from the electron diffraction data. (The slightly higher value than the one reported in Ref. 18) can be explained by the presence of Cu in the Ti2 Ni phase (see Fig. 5(a))). For the underlying.

(5) Phase Formation and Structural Sequence of Highly-Oriented MBE-Grown NiTiCu Shape Memory Thin Films Al K. 937. 3/2 3/2 1/2. Intensity (a.u.). 3/2 1/2. (b) Cu. Ni Ti (a) 420 440 460 480. 840 860 880 900 920 940. Binding Energy (eV) Fig. 8 Photoemission spectra of a nominally Ti50 Ni40 Cu10 film: (a) after in-situ crystallization and exposure to atmosphere (Ti 47.3 at%, Ni 42.3 at%, Cu 10.4 at%), (b) after post-annealing at 250◦ C (Ti 47.6 at%, Ni 27.7 at%, Cu 24.7 at%).. (440). (333) (511). Fig. 10 Atomic configuration for a quasi (333) plane of the Ti2 Ni crystal structure (space group Fd 3̄m). The atomic sites are completely occupied by Ti: (a) Normal projection view, (b) in-plane view along the [1̄, 1̄, 2] direction. Here, the atomic size scales with the interatomic distances, the metallic (Goldschmidt) radius of Ti taken as reference.19). (422). 2. Fig. 9 X-ray diffraction frame pattern of the Ti2 Ni phase taken from a Ti51 Ni38 Cu11 film (2Θ range: 22◦ –57◦ , χ range: −69◦ –−111◦ ).. structural base, the reader is referred to Ref. 15). Although the configuration is considerably corrugated (see Fig. 10(b)) the quasi hexagonal lattice structure observed is found to match the underlying TiO2 -(110) layer by less than 6.5% deviation in regard to its 111 direction. Similar holds for the NiTiCu alloy film on top of it, here with a mismatch of less than 3.5% as for the 110 direction of the austenite phase. Regarding the crystallization behavior of the phase sequence observed we refer to the data collected by Chen and Wu20) from which a crystallization scheme can be derived. It is found that for NiTi bulk alloys and thin films the crystallization temperature drops from around 550◦ C more or less monotonously by about 100◦ C over the composition range from Ti40 Ni60 to Ti70 Ni30 . Moreover, the crystallization temperature decreases further when Ni is substituted by Cu. Specifically, for the composition Ti50 Ni40 Cu10 , the crystallization temperature is reduced additionally by about 25◦ C compared with the non-Cu containing composition. This signals that due to the concentration profile along the growth direction in our films the crystallization starts from the Ti-rich. (Ti2 Ni) phase at the underlying TiO2 /Si-oxide barrier interface and proceeds gradually to the surface according to the compositional gradient. Thus the structural matches reported above are interlinked with the occurring phase sequence and the preferred orientation resulting therefrom. Preferred structural orientations have been achieved by DC magnetron sputtering as well. Typically, a multilayer sequence with layers from a pure Ti target (2 nm) and from an alloy target of the composition Ti50 Ni35 Cu15 (50 nm) has been deposited onto an oxidized Si substrate. After deposition, the multilayer sequence was dissolved by annealing at 650◦ C for 15 min which leads to a homogeneous monolayer as described by Chang and Grummon.21) After crystallization, again a preferred orientation was found, but in this case the 111 martensite direction is perpendicular to the substrate plane with a variation of the orientation of ±4◦ (Figs. 11(a), (b)). The differences and similarities to the MBE-grown films have not been fully elaborated yet and will be reported elsewhere. 4. Conclusion Using MBE deposition highly-oriented NiTiCu shape memory thin films were fabricated. Remarkably, the austenite-martensite transition in these films occurs well below 1 µm film thickness with comparable transition temperatures as known for thick sputter-deposited films (1–10 µm). Mechanical stress measurements reveal a stress recovery in the order of 300 MPa. Regarding the microstructure, the crystallites are oriented within ±3◦ along the film plane normal as observed by X-ray diffraction. Moreover, a splitting of the martensite orientation is detected indicating the selection of two sets of martensite variants with defined orientational re-.

(6) 938. R. Hassdorf et al.. (a) (b) Intensity (a.u.). (111). 2. -120. -110. -100. -90. -80. -70. -60. (deg) Fig. 11 (a) X-ray diffraction frame pattern taken from a sputter-deposited Ti50 Ni35 Cu15 film in the martensite state (2Θ range: 25◦ –55◦ , χ range: −69◦ –−111◦ ). The film structure originates from an as-grown NiTiCu/Ti multilayer stack dissolved upon annealing. (b) Integration along 2Θ.. lationship. This is consistent with high-resolution TEM and diffraction images revealing the coexistence of tilted crystals with specific orientation difference. For the surface topography, STM results exhibit contours of well aligned crystals with slightly tilted terrace orientations. These special orientation characteristics give rise to expect novel properties in such thin films, especially in regard to their superelastic behavior. Acknowledgments We thank A. Sehrbrock and R. Borowski for technical assistance, and P. Zerfass for image analysis of STM data. As for the development and construction of the X-ray heating stage used in this work kind cooperation with mri Physikalische Geräte GmbH, Karlsruhe is highly appreciated. REFERENCES 1) For a review, see, e.g., S. Miyazaki: Shape Memory Materials, ed. by K. Otsuka and C. M. Wayman, (Cambridge University Press, Cambridge, 1998) pp. 267–281. 2) See, e.g., Conf. Proc. ACTUATOR 2000, ed. by H. Borgmann, (Messe Bremen GmbH, Bremen, 2000). 3) See, e.g., M. Kohl, D. Dittmann, E. Quandt, B. Winzek, S. Miyazaki and D. M. Allen: Mater. Sci. Eng. A 273–275 (1999) 784–788. 4) S. Miyazaki and A. Ishida: Mater. Sci. Eng. A 273–275 (1999) 106–133. 5) S. Miyazaki, T. Hashinaga and A. Ishida: Thin Solid Films 281–282. (1996) 364–367. 6) L. Chang and D. S. Grummon: Philos. Mag. A 76 (1997) 163–189. 7) L. Chang and D. S. Grummon: Philos. Mag. A 76 (1997) 191–219. 8) See, e.g., C. Lu and Y. Guan: J. Vac. Sci. Technol. A 13 (1995) 1797– 1801. 9) J. H. Scofield: J. Elec. Spectr. Rel. Phenom. 8 (1976) 129–137. 10) M. Moske and K. Samwer: Mater. Res. Soc. Symp. Proc. 356, ed. by S. P. Baker, C. A. Ross, P. H. Townsend, C. A. Volkert and P. Børgesen, (Materials Research Society, Pittsburgh, 1995) pp. 27–32. 11) M. Ohring: The Materials Science of Thin Films, (Academic Press, London, 1992) pp. 416–418. 12) See, e.g., B. Winzek and E. Quandt: Z. Metallk. 90 (1999) 796–802. 13) H. H. Brongersma, P. A. J. Ackermans and A. D. van Langeveld: Phys. Rev. B 34 (1986) 5974–5976, and refs. therein. 14) M. H. Mueller and H. W. Knott: Trans. Metall. Soc. AIME 227 (1963) 674–678. 15) H. T. Takeshita, H. Tanaka, N. Kuriyama, T. Sakai, I. Uehara and M. Haruta: J. Alloys Comp. 311 (2000) 188–193. 16) P. Oberhauser and R. Abermann: Thin Solid Films 350 (1999) 59–66. 17) P. Oberhauser, M. Poppeller and R. Abermann: Mater. Res. Soc. Symp. Proc. 648, ed. by S. C. Moss, (Materials Research Society, Pittsburgh, 2001) pp. P3.19.1–6. 18) G. A. Yurko, J. W. Barton and J. G. Parr: Acta Cryst. 12 (1959) 909. 19) N. W. Alcock: Bonding and Structure, (Ellis Horwood, 1990) App. Tables of Radii A.3. 20) J. Z. Chen and S. K. Wu: J. Non-Cryst. Solids 288 (2001) 159–165, and refs. therein. 21) L. Chang and D. S. Grummon: Mater. Res. Soc. Symp. Proc. 311, ed. by M. Atzmon, A. L. Greer, J. M. E. Harper and M. R. Libera, (Materials Research Society, Pittsburgh, 1993) pp. 167–172..

(7)

No results found