Shape Memory Micro- and Nanowire Libraries for the High

Jul 31, 2017 - The scaling behavior of Ti–Ni–Cu shape memory thin-film micro- and ... start, Ms) and the thermal hysteresis from −3.5 to 20 K. I...
0 downloads 0 Views 3MB Size
Research Article pubs.acs.org/acscombsci

Shape Memory Micro- and Nanowire Libraries for the HighThroughput Investigation of Scaling Effects Tobias Oellers,† Dennis König,† Aleksander Kostka,‡ Shenqie Xie,§ Jürgen Brugger,§ and Alfred Ludwig*,† †

Institute for Materials, Faculty of Mechanical Engineering, Ruhr-Universität Bochum, 44801 Bochum, Germany ZGH, Ruhr-Universität Bochum, 44801 Bochum, Germany § Ecole Polytechnique Federale de Lausanne (EPFL), Laboratoire de Microsystemes, CH-1015 Lausanne, Switzerland ‡

S Supporting Information *

ABSTRACT: The scaling behavior of Ti−Ni−Cu shape memory thin-film micro- and nanowires of different geometry is investigated with respect to its influence on the martensitic transformation properties. Two processes for the highthroughput fabrication of Ti−Ni−Cu micro- to nanoscale thin film wire libraries and the subsequent investigation of the transformation properties are reported. The libraries are fabricated with compositional and geometrical (wire width) variations to investigate the influence of these parameters on the transformation properties. Interesting behaviors were observed: Phase transformation temperatures change in the range from 1 to 72 °C (austenite finish, (Af), 13 to 66 °C (martensite start, Ms) and the thermal hysteresis from −3.5 to 20 K. It is shown that a vanishing hysteresis can be achieved for special combinations of sample geometry and composition. KEYWORDS: combinatorial material science, shape memory alloy, phase transformation, zero hysteresis, scaling effects, thin film



INTRODUCTION The development of micro- and nanomechanical devices is closely related to the development of new or improved functional materials, which enable further integration of functions. To effectively apply new materials in micro- or nanosystems, it is important to investigate their properties on the applied length scale because the functional properties of a material can change significantly due to scaling effects, which are dependent on, for example, the surface to volume ratio, if the material is used on the bulk-, micro-, or nanoscale.1 For micro- and nanomechanical systems this dependence is important, as technological progress results in further downscaling of devices. Therefore, the general understanding of the influence of scaling effects has to be improved to be able to efficiently design functional structures at the micro- and nanoscale. In addition to the influence of structure geometry on materials properties, it is also necessary to investigate the dependence of functional materials properties on composition. Both parameters do strongly affect material performance and can show interdependent behavior.2,3 This can especially be observed toward the nanoscale where the composition of a material can be influenced, for example, by surface oxide formation, or the formation of film/substrate interface layers.4−6 So far, such approaches have not been explored extensively and are mainly performed for few, discrete compositions and geometries.7−9 The lack of such investigations is primarily due to the difficulty of systematically fabricating a statistically relevant number of samples with well-controlled geometry and composi© XXXX American Chemical Society

tional variations and furthermore a lack of techniques for a subsequent rapid characterization of the properties. This paper introduces a concept to overcome these issues which is based on a combinatorial approach to synthesize microand nanowire libraries, that is, a multitude of well-controlled wire samples fabricated in a parallel process on a wafer-scale substrate. The wire libraries comprise a geometrical variation from the micro- to the nanoscale and a compositional variation, and enable conducting a systematic analysis of the sample performance with regards to surface to volume ratio dependent scaling effects and composition, while sample fabrication and characterization can be conducted in a time-saving manner. Additionally, with the availability of a combinatorial test platform, it is not only possible to simply characterize the influence of different parameters on the material performance but also to do the reverse and purposefully tailor the parameters to achieve a certain performance. An example for this are transducer materials such as shape memory alloys (SMA), where through compositional and geometrical changes the material can be optimized for applications, which require either high response speed or high actuation strength. The general concept of influencing functional properties through scaling effects has been successfully demonstrated for diverse materials and applications10,11 such as the increase in solar absorption for solar cells,12 Received: April 13, 2017 Revised: July 25, 2017 Published: July 31, 2017 A

DOI: 10.1021/acscombsci.7b00065 ACS Comb. Sci. XXXX, XXX, XXX−XXX

Research Article

ACS Combinatorial Science

Figure 1. Arrangement of fabricated samples on a 100 mm diameter wafer. (a) Samples for EDX measurements are positioned on a rectangular grid in between the wire samples. (b) The compositional variation is indicated by the multilayer structure. The geometry variation is indicated in horizontal direction. Insets c and d show magnified SEM images of the fabricated structures for better clarity.



photoelectrochemical conversion efficiency,13 the detection range of biosensors,14 or tailoring of the hysteresis for SMA.15,16 For SMA, the investigations of nanoscale effects in literature are mainly limited to chemically elementary nanostructures,17−22 serial investigations of a small number of samples with limited variation,23,24 or investigations based on simulations.25−28 Hou et al. presented the fabrication of freestanding NiTi nanowires through skiving. They showed the capability of the technique to fabricate wires of up to several hundreds of micrometers in length and various nanoscale cross-section geometries. The presence of the transforming B2 crystal phase was confirmed, but the transformation properties were not further characterized.29,30 Phillips et al. showed the possibility for the fabrication of InTl shape memory nanowires in the range from 10 to 650 nm by using the mechanical pressure injection method on a porous template. The transformation was confirmed through X-ray diffraction and transmission electron microscopy experiments.31 These examples show that it is of high interest to characterize and understand scaling effects, especially on a large-scale approach coupled with significant sample variation. As a proof of concept, Ti−Ni−Cu SMA micro- and nanowire libraries were fabricated and investigated. Ti−Ni−Cu was selected, because it is promising for solid state actuator, transducer, and sensor applications,32 where it can achieve the integration on a small scale and potentially high-cycle applications.33 Ti−Ni−Cu SMAs are working on the principle of the martensitic phase transformation, which can be reliably investigated through temperature-dependent 4-point resistance measurements.34 Additionally, some information is available about thickness-dependent scaling effects for Ti50.2Ni49.8 thin films by Fu et al.,35 who investigated continuous films in a thickness range of approximately 4 μm to 50 nm. They reported, that in their experiments no transformation could be achieved for thicknesses 0 mm. It should be noted, that the thickness of the investigated measurement areas varied in the range of approximately 30−45 nm, which is at the lower limit for the EDX analysis. Therefore, the measurement error may be increased and could partially explain the inhomogeneous variation of the composition. But

Table 1. Grain Size of Wire Structures for Different Geometrical Widths wire width [nm]

grain size [nm]

1940 1060 560

13 ± 1 13 ± 1 14 ± 1

approximately 13−14 nm was observed for all investigated samples, with no significant variations along the wire width variation. Because of the almost constant grain size, it is concluded that variations of the transformation properties along the geometry gradient are not a result of changes in grain size. Phase Transformation Properties. For the samples fabricated by EBL, wire samples along the geometry variation, as well as along the compositional gradient, were measured. This was done to characterize the individual influence of both parameters on the transformation properties and to demonstrate the viability of the systematic investigation. In a first step, wire structures with a constant composition were measured along the geometry variation. Figure 7a shows R(T) measurements of wire structures with a composition of Ti48Ni40Cu12 and widths in a range from 560 to 1960 nm as well as the corresponding transformation temperatures (Figure 7b) and thermal hysteresis ΔT values (Figure 7c). The wire F

DOI: 10.1021/acscombsci.7b00065 ACS Comb. Sci. XXXX, XXX, XXX−XXX

Research Article

ACS Combinatorial Science

Figure 7. Transformation properties acquired through R(T) measurements of samples fabricated by e-beam lithography. The first set of graphs show the acquired R(T) curves (a) of wires with a composition of Ti48Ni40Cu12 with the corresponding values for the transformation temperatures (b) and thermal hysteresis (c). The second set of graphs show the acquired R(T) curves (d) of wires with a constant wire width of 750 nm with the corresponding values for the transformation temperatures (e) and thermal hysteresis (f). The third set of graphs show the acquired R(T) curves (g) of wires with a constant composition of Ti61Ni27Cu12 with the corresponding values for the transformation temperatures (h) and thermal hysteresis (i).

width range covers wire structures from the widest samples fabricated for this corresponding composition, down to the structure width, which no longer exhibited a stable transformation. The transformation is indicated by the nonlinearity of the R(T) curves. The linear behavior of the measurement which was

observed for elevated temperatures indicates a completed transformation of the Ti−Ni−Cu structures into the austenitic phase. The full transformation into the martensitic phase was not observed due to insufficient cooling. The nonlinear behavior was observed for samples in the range from 690 to 1960 nm. It could be concluded, that the reversible phase transformation was G

DOI: 10.1021/acscombsci.7b00065 ACS Comb. Sci. XXXX, XXX, XXX−XXX

Research Article

ACS Combinatorial Science

thermal hysteresis (Figure 7f). The R(T) curves show a nonlinear behavior for samples in the composition range from Ti48Ni40Cu12 to Ti63Ni25Cu12 with a low thermal hysteresis. For samples with a Ti concentration > 63 at. % a nonlinear behavior could still be observed, which is approaching a zero hysteresis behavior, but because of the increased noise level, it was not possible to determine exact Af and Ms values. Thus, the phase transformation properties could not be safely concluded for these samples. In the compositional range from Ti48Ni40Cu12 to Ti58Ni30Cu12 Af and Ms (Figure 7e) increased slightly from 28 and 30 °C to 31 and 33 °C. In the range from Ti58Ni30Cu12 to Ti61Ni27Cu12 a sharp decrease of Af and Ms from 28 and 30 °C down to 13 °C for both temperatures was observed. For an increasing Ti content to Ti63Ni25Cu12 both Af and Ms increased by 1 to 14 °C. The ΔT values (Figure 7f) increased from approximately −1.7 K for Ti48Ni40Cu12 to a “zero hysteresis” behavior for compositions in the range from Ti61Ni27Cu12 to Ti63Ni25Cu12. A similar behavior for transformation temperatures and the thermal hysteresis was observed by Zarnetta et al.45 who investigated the transformation properties of continuous Ti−Ni−Cu films for similar compositions and annealing conditions as in this work. There it was observed that the transformation temperatures exhibited a maximum and the thermal hysteresis a minimum for approximately Ti50Ni38Cu12 and showed a decrease (Af and Ms) or respectively increase (ΔT) for both increasing and decreasing Ti concentrations along the Ti88‑xNixCu12 gradient. The general behavior of the transformation properties for the continuous films was similar to the behavior of the wire samples with decreasing transformation temperatures and increasing thermal hysteresis for increasing Ti concentrations, but the exact values showed a significant difference. In the investigated compositional range, the transformation characteristics of the continuous films changed from Ms ≈ 66 °C, Af ≈ 72 °C and ΔT ≈ 7 K for a composition of Ti50Ni38Cu12 to Ms ≈ 45 °C, Af ≈ 55 °C and ΔT ≈ 12 K for increasing Ti and constant Cu content. The transformation characteristics of the investigated wire samples showed a change from Ms ≈ 49 °C, Af ≈ 46 °C and ΔT ≈ −2.2 K for a composition of Ti50Ni39Cu11 to Ms ≈ 13 °C, Af ≈ 13 °C, and ΔT ≈ 0 K for increasing Ti and constant Cu content. The similar behavior may indicate that the observed influence on the transformation parameters was caused by the compositional variation, because the only varied parameter for this specific set of wires samples, as well as the continuous films, was the composition while the geometry remained constant. From literature, the origin of the composition-dependent behavior was expected to be the formation of Ti-rich precipitate phases due to the increasing Ti content of the alloy, which were observed by Zarnetta et al. for the continuous films. The difference of the absolute property values in comparison to this work might be explained by the aforementioned constraining effects which differ for continuous films and wire samples. A comparison between the continuous films and wires is difficult due to the different constraints and without further investigations only allows for a qualitative comparison. It was mentioned before that the thermal hysteresis is influenced by the composition of the sample and that increasing lateral constraints cause a depletion of the Ti content for the “active” Ti−Ni−Cu material due to surface oxide formation. This depletion results in the vanishing of the hysteresis due to the reduced Ti content. To prevent this, it is assumed that the depletion mechanism could be counteracted by increasing the initial Ti concentration to stabilize the reversible transformation

occurring. For samples with a width < 690 nm, the curves only showed a slight nonlinearity with significant noise. Thus, no transformation properties could be obtained for these samples, and it is uncertain if a reversible phase transformation occurs at lower temperatures. For a composition of Ti48Ni40Cu12, the austenite finish (Af) and martensite start (Ms) transformation temperatures (Figure 7b) show a similar behavior for different wire widths. As the wire width gets smaller from 1960 to 1060 nm, the transformation temperatures increase slightly from Af = 31 °C and Ms = 35 °C to Af = 37 °C to Ms = 40 °C. In the range from 1060 down to 750 nm, a sharp decrease from Af = 37 °C and Ms = 40 °C to Af = 25 °C to Ms = 27 °C is observed. For a width of 690 nm transformation temperatures again show a slight increase to Af = 29 °C and Ms = 30 °C. The values for the thermal hysteresis (Figure 7c) increased from approximately −3.5 K to −0.7 K over the full width range from 1960 down to 690 nm. All obtained curves showed a negative thermal hysteresis. The behavior of the transformation temperatures was similar to the data obtained by König et al.15 for continuous films where the varied geometry parameter was the film thickness. The values for the thermal hysteresis on the other hand show a different behavior. The change in transformation temperatures might be explained by the influence of competing effects which are compositional changes due to the surface oxide formation and varying stress states due to the lateral constraints of the wire-like geometry. The stress state of the structures is not identical to the work presented by König et al. due to the different sample geometries. Additionally, in contrast to most prior studies which focused on the characterization of continuous films, this study focused on wire like structures. Because of the shape of the wire structures, a surface oxide layer not only forms on the top of the wire structure, but also on the sidewalls. This effect contributes to the compositional change of the Ti−Ni−Cu material by reducing the Ti content, since the surface oxide mainly consists of different oxidation states of TixOx as investigated by König et al.44 Because of the increased surface-area to volume ratio of the wires for decreasing widths, the influence scales proportional to the width variation. The stress state of the sample could also be influenced by the geometry of the wire. A continuous thin film is laterally constrained by the surrounding film as opposed to the wire structure which is only constrained by the Ti−Ni−Cu/surface oxide interface. Therefore, the actual stress state most probably differs from a continuous film, but the actual stress state could not be determined. The different influences of the film stress and compositional changes may explain the different behavior for the thermal hysteresis, but a definite conclusion cannot be given. The vanishing of the reversible phase transformation for widths < 690 nm may be explained by the influence of the wire thickness. Correlating the thickness data and transformation temperatures, the vanishing transformation coincides with a significant decrease of the wire thickness. The changes of the wire geometry in turn exhibit a significant influence on the competing scaling effects of composition and stress state, which may cause the suppression of the transformation. After the influence of the compositional variation on the transformation properties was investigated, samples along the Ti−Ni gradient with a constant geometry were measured to investigate the individual influence of the composition on the transformation characteristics. Figure 7d shows R(T) measurements of samples with a constant wire width of 750 nm and a compositional variation from Ti48Ni40Cu12 to Ti66Ni22Cu12 with the corresponding transformation temperatures (Figure 7e) and H

DOI: 10.1021/acscombsci.7b00065 ACS Comb. Sci. XXXX, XXX, XXX−XXX

Research Article

ACS Combinatorial Science

transformation for thickness values