Microfabrication Approaches on Magnetic Shape Memory Films

Magnetic shape memory alloys are emerging multifunctional materials that enable applications like high‐stroke actuation, solid‐state refrigeration, and energy harvesting of waste heat. Thin films of these alloys promise integration in microsystems to exploit their multifunctional properties at the microscale. However, the microfabrication process of these Heusler alloys is difficult. Herein, different etching techniques are investigated for the microfabrication of epitaxial Ni‐Mn‐Ga films, the encountered challenges are explained, and ways to overcome them are demonstrated. The results show that wet chemical etching is suitable for large patterned structures, while reactive ion etching of Ni‐Mn‐Ga films is unsuitable due to redeposition. For patterning structures below 10 μm with clean and sharp edges, the best results are obtained by ion beam etching with adjusted sample‐stage tilt. Finally, a microfabrication process using Si microtechnology to fabricate partially freestanding structures is demonstrated. These findings give guidelines for the fabrication and integration of these smart materials in Si‐based microsystems.


Introduction
Magnetic shape memory alloys are an emerging class of smart materials which combine ferromagnetism and ferroelasticity to produce promising capabilities [1] like magnetically induced strain [2] up to 12%, [3] high actuation frequency up to 100 kHz, [4] multicaloric effects, [5] and thermomagnetic harvesting of waste heat. [6]These alloys have been used to realize conventional actuation systems like grippers and valves [7] and are potential candidates for limbs of robots used in biomedical applications, [8][9][10][11] solidstate refrigeration, [12,13] and thermomagnetic powering of smart devices and sensors. [6]ost of the mentioned properties have been studied in the bulk form of Ni-Mn-based Heusler alloys and often optimum properties are only obtained in single crystals.To realize these multifunctional properties in microsystems, a large effort had been put on epitaxial film growth [14][15][16][17] and on microfabrication, summarized later.A unique advantage of these alloys is the "the material is the machine" [18] concept, which supports miniaturization for applications and devices at micro-and nanoscale.Thermal microsystems utilizing these alloy films also benefit from an accelerated heat exchange due to their large surface area to volume ratio compared to bulk, which increases frequency and thus power density. [6,19]tudies on patterned magnetic shape memory films have gained prominence over recent years.Kohl et al. [20] demonstrated microfabrication and integration technologies for polycrystalline Ni-Mn-Ga films in microvalves and microactuators.They performed film deposition on removable substrates and then transferred films to a carrier Si wafer for micromachining by optical lithography and wet chemical etching.The drawback of this approach is the additional transfer and bonding steps, which complicates batch fabrication.In another study, Kohl et al. [21] demonstrated reversible actuation based on thermal shape memory effect in nanoactuators fabricated from an epitaxial Ni-Mn-Ga film down to a feature dimension of 100 nm.The nanoactuators were fabricated by electron beam lithography and ion beam etching followed by wet chemical underetching of the Cr buffer layer to release the Ni-Mn-Ga actuator from the MgO(001) substrate.This release process employs wet etching and hence, requires careful transfer of sample between the etching step and the freeze drying or critical point drying step to avoid damage to structures from the capillary forces.In another study, the focused ion beam (FIB) process was used by Jenkins et al. [22] to fabricate partially freestanding Ni-Mn-Ga cantilevers.Although this approach does not require sample transfer, FIB, a serial patterning process, is too slow for mass production.Alvarez-Alonso et al. [23] developed a novel bottom-up fabrication process of Ni-Mn-Ga antidots using self-assembled polystyrene spheres acting as mask during film deposition.This masking approach (nanosphere lithography) was used in another study to fabricate Ni-Mn-Ga nanodiscs by Ar plasma etching and study their martensitic microstructure and actuation response. [24]Though this fabrication method allows convenient control of the pattern size, it is limited to only one pattern design.Taksha Ghahfarokhi et al. [25] also used Ar plasma etching for the microfabrication of epitaxial Ni-Mn-Ga films.They studied the influence of patterned structure dimension and orientation on the martensite microstructure, which revealed possibilities for microstructure engineering.
These limited studies so far on patterned magnetic shape memory films have focused on the martensitic microstructure and properties instead of practical aspects related to microfabrication.A systematic study on the suitability of existing microfabrication techniques and their particular challenges for these materials is missing.Such a study is necessary to guide the microfabrication of magnetic shape memory films and further the goal of integrating these emerging materials into microsystems.
Here, we investigate the suitability of conventional wet and dry etching techniques for patterning epitaxial Ni-Mn-Ga films.Epitaxial Ni-Mn-Ga films are selected because the best actuation properties have been observed in single crystals. [3,26]We choose a film thickness of at least 500 nm for our study as actuators with this thickness give a reasonable force and have a sufficient total heat capacity for most thermal applications. [20,27,28]We analyze the challenges encountered in the studied etching techniques and present ways to overcome them.Finally, we use the optimized etching process to demonstrate the fabrication of partially freestanding patterns from epitaxial films grown on Si-based substrates in a monolithic fashion.

Wet Chemical Etching of Ni-Mn-Ga Films
To test Ni-Mn-Ga microfabrication, we begin with wet chemical etching of 500 nm-thick sputter-deposited Ni-Mn-Ga films on MgO (001) substrates using a 37 vol% HNO 3 and 1.25 vol% HCl solution at room temperature.This solution was diluted with deionized water in a 1:10 ratio to obtain an etch rate of about 80 nm min À1 .Prior to the etching step, the film surface was masked with a photoresist by following the photolithography steps described in the Experimental section.Figure 1a shows an optical image of square-shaped patterns etched on Ni-Mn-Ga film with a 1.5 μm-thick AZ 5214E photoresist mask.At the outset, we notice large underetch regions of %5-6 μm width below the mask for each pattern.The underetching exhibits rounded corners under the mask, indicating the isotropic nature of Ni-Mn-Ga etching by the acid solution.Due to underetching, the patterns lose their intended shape and dimensions.This problem is amplified as the size of the pattern design is reduced.To investigate the effect of a hard mask instead of photoresist on the extent of underetching, we used a 225 nm-thick SiN x hard mask, deposited by chemical vapor deposition (CVD) and patterned by reactive ion etching with CF 4 plasma.Figure 1b shows exemplarily the corresponding optical image of structures obtained after the wet chemical etching.In this case, the underetched regions have reduced significantly to %1 μm in width (see Figure S1, Supporting Information, for a closer look at the underetched regions).Stripe with width of a few micrometers can also be fabricated, as shown in Figure 1c.However, wet chemical etching always results in rough edges of the patterned structures, appearing as "mouse bites." The isotropic nature of wet chemical etching leads to simultaneous lateral etching while etching through the film thickness.It gives rise to underetched regions below the mask, as illustrated in Figure 1d.In our case of wet etching with the photoresist mask (Figure 1a), the lateral underetching width (5-6 μm) is nearly ten times the film thickness (500 nm).This indicates poor adhesion of photoresist to film surface in the etching solution.This problem is solved using a hard mask, as exhibited by a reduced underetch width below the SiN x mask (see Figure S1, Supporting Information).The SiN x layer can be etched away selectively by CF 4 plasma, as described in the Experimental section.Nevertheless, the wet chemical etching technique could be challenging to achieve dimensional control of patterned structures with sharp edges due to its isotropic nature.
Previously, Eichhorn et al. [29] performed wet chemical etching of Ni-Mn-Ga films using a photoresist mask and dilute nitric acid (10 vol%) solution.The patterned structure dimensions, however, were in the order of hundreds of micrometers, which is significantly larger than the pattern dimensions fabricated in our study.Furthermore, electrochemical polishing based on nitric acid and ethanol solution had been used to fabricate magnetic shape memory foils from bulk single crystals. [30,31]his shows that wet chemical etching technique is easy to adapt for microfabricating thick films and foils.Based on these observations, wet chemical etching of Ni-Mn-Ga films is a suitable approach for fabricating structures with large lateral dimensions (several tens of micrometers and further) and for micrometerthick magnetic shape memory films.

Reactive Ion Etching of Ni-Mn-Ga Films
To fabricate sharp patterned structures by anisotropic etching, we next tested dry etching of 500 nm-thick Ni-Mn-Ga films.First, we investigated a reactive ion etching process based on Ar and SF 6 plasma.The inductively coupled plasma (ICP) power (700-1000 W) and radio frequency (RF) power (150-300 W) were varied during the etching experiments using a 10 μm-thick AZ 10XT photoresist as mask.A thick photoresist was selected to ensure the masked regions remain protected during the experiments.The sample stage was also cooled with a cryostat to 263 K during the etching experiments to prevent: 1) thermal softening of photoresist structures, which affects the transferred design to the film, and 2) baking of photoresist from heating, which makes photoresist removal after the etching step difficult.The etching process results in high etch rates, increasing monotonously from 55 nm min À1 (ICP power: 700 W, RF power: 150 W) to 107 nm min À1 (ICP power: 1000 W, RF power: 300 W) (see Figure S2, Supporting Information, for the variation of etch rate with ICP and RF power).However, after removing the photoresist, the patterned structure reveals a region of redeposited material with a width of %200 nm around the edges (Figure 2a).The sidewall "fences" of redeposited material are stable in some regions even after the sonication treatment in acetone and reach up to several micrometers in height.To study the influence of the photoresist profile on redeposition, we next used a 7 μm-thick AZ nLOF 2070 photoresist as mask.This photoresist has an undercut profile and is typically used for lift-off applications.Figure 2b shows the corresponding patterned structure after removing the photoresist.Unlike the previous case, the structure does not show sidewall fences.Instead, a dark area on the patterned structure is seen, which corresponds to the shadowed region under the V-shape photoresist.It has a granular morphology (see Figure S3a, Supporting Information), made of redeposited material of about 90 nm in height (see Figure S4, Supporting Information).Such granular morphology of redeposition is also observed on the top surface of the photoresist when investigating a partially etched sample (Figure 2c and S3b, Supporting Information).
The sidewall redeposition problem is described schematically in Figure 3.In an etching process dominated by physical sputtering, the etched material is generally nonvolatile and can redeposit on the sample, for instance, on the sidewalls of patterns.In case of a steep photoresist profile (case A), "fences" of redeposited material get created and remain around the structure after removing the photoresist mask (Figure 2a).On the contrary, the redeposition-induced fences are not observed after removing the photoresist with undercut profile (Figure 2b).In this case (Case B), redeposition occurs as usual on photoresist and pattern sidewalls and additionally on the shadow region under the V-shape photoresist profile.We propose that the undercut V-shape profile particularly prevents formation of a continuous redeposited "fence" on the sidewall, as shown in Figure 3 in the magnified region within a black circle.Hence, on removing the photoresist, we end up with patterned structures covered with redeposited material only on the shadowed regions.
On analyzing the existing literature, we found no reactive ion-etching studies involving fluorine-based plasma chemistry on Ni-Mn-Ga films.Nevertheless, Hong et al. [32,33] studied different halogen-based gas chemistries combined with Ar for reactive ion etching of Ni-Mn-Sb films.They found Ar þ BCl 3 , H 2 þ BCl 3 , and Ar þ SF 6 gas chemistries to be suitable for etching with high etch rates (0.5-1 μm min À1 ) and without redeposition.They reported redeposition and surface enrichment with Mn while using Ar þ NF 3 gas chemistry.Redeposition during etching is encountered primarily in Ar ion beam etching/ion milling technique due to the nonvolatile nature of etch products. [34,35]Hence, the occurrence of redeposition in our results indicates limited or no chemical reaction with volatile products using the fluorine plasma chemistry.Although redeposition process can be used to fabricate 3D structures, [36,37] it is critical to avoid this here because the redeposited material can act as a constraint and prevent actuation of the Ni-Mn-Ga structure in response to stimuli.The redeposition also affects the functioning of the reactive ion etching tool, as it can significantly coat the inner lining of the chamber when etching thick films.As a remedy, some microfabrication studies on magnetic materials used for magnetic tunnel junctions (MTJs) demonstrated that pulsed DC bias in reactive ion etching helped to reduce redeposition. [38,39]Nevertheless, our etching experiments suggest that Ni-Mn-Ga films are incompatible with the reactive ion etching process based on Ar and SF 6 plasma chemistry.

Ion Beam Etching of Ni-Mn-Ga Films
To solve the problem of redeposition, we performed a series of experiments with Ar ion beam etching on the Ni-Mn-Ga films.The ion beam etching tool is equipped with a tilting sample stage that allows optimization of the etching angle to reduce redeposition of the etched material.We performed etch tests on 2 μmthick Ni-Mn-Ga films patterned with 7 μm-thick AZ nLOF 2070 photoresist at tilt angles of 0°, 15°, 30°, and 45°for 10 min each.The angle is defined with respect to the standard configuration of sample stage during etching (90°between sample stage and ion beam).Contrary to the results of reactive ion etching (Figure 2c), no redeposition was observed on top of the photoresist.However, redeposition on the photoresist sidewalls is seen, as indicated by arrows (see Figure S5, Supporting Information).
Figure 4 shows the corresponding sample surfaces after removing the photoresist.The films are in martensitic state and exhibit a twinned microstructure consisting of Type X and Type Y mesoscopic twin boundaries. [40]No redeposition fences are observed here around the pattern.In case of no tilt (0°), the pattern has smooth edges with the etched region (Figure 4a).In Case A, the redeposition during etching occurs on sidewalls of photoresist and structure, which results in "fences" around the structure.In Case B, the redeposition additionally occurs on the top shadowed unmasked regions of structures.Furthermore, the redeposited material is discontinuous between sidewalls of photoresist and structure due to the undercut photoresist profile.Hence, redeposition on top of structures only remains instead of "fences" after removing photoresist.This is due to the undercut photoresist profile, which leads to partial etching in the shadowed regions.On tilting the sample during etching, this problem appears to be solved, as the pattern edges appear sharp in other cases (Figure 4b-d).To investigate the etch profile, we performed FIB cuts on the 15°and 45°tilt samples, as exemplarily indicated by white dotted lines.Figure 5a,b shows the corresponding FIB cross sections of the etched Ni-Mn-Ga films respectively.The characteristic twin boundary microstructure of martensitic Ni-Mn-Ga films is also observed in both of these cross sections. [40]However, depending on the etching angle, the films exhibit different surface etch profiles in the "etched area" marked in Figure 5a,b.The sample etched at 15°tilt shows an increase in etch rate from 16 nm min À1 at the edge of the masked area to 50 nm min À1 about 2.5 μm away from the edge, after which the etch rate remains stable.On the other hand, the 45°tilt etched sample shows a continuous increase in etch rate from 20 nm min À1 at the edge of masked area to 50 nm min À1 about 8 μm away from the edge.To further investigate the effect of the difference in etch rate, we etched a 500 nm-thick Ni-Mn-Ga film with a 3 μm-thick AZ nLOF 2020 photoresist at a tilt angle of 45°for 10 min.Figure 5c shows an optical image of the etched sample.The Ni-Mn-Ga film is completely etched after 10 min and the bare substrate (appearing green) can be seen in the unmasked areas.However, an unetched film region about 2 μm wide remains around the edges of the mask.
The variation in etch rate (when tilting the sample) originates from ion-beam shadowing of regions by the photoresist.This problem is illustrated in simplified form in Figure 5d.The photoresist has a finite thickness and creates an additional "shadow" region when the sample stage is tilted.Although the sample stage rotates during etching, the etch rate differences accumulate over the etching time and lead to unetched material remaining around the pattern.Using simple trigonometry, one can estimate the width of the shadow region based on the thickness of the photoresist.The width of shadow region is 1.9 μm (15°tilt) and 7 μm (45°tilt) for 7 μm-thick AZ nLOF 2070 photoresist.These values are close to the measured values (2.5 μm and 8 μm) in the etch profiles of Figure 5a,b, respectively.In case of Figure 5c (45°tilt of 3 μm-thick photoresist), the shadow region is about 3 μm wide and we ended up with about a 2 μm-wide unetched region.This problem associated with tilting has been encountered before in ion beam etching studies of multilayer stacks and referred to as pattern "foots" [41] or "tails." [42]he ion beam etching of Ni-Mn-Ga films up to now highlights three key points.1) Without tilting the sample stage, photoresist with undercut profile is not suitable for obtaining patterns with sharp edges (Figure 4a).2) Tilting the sample stage during etching results in a varying surface etch profile (Figure 5a,b).
3) Reducing the photoresist thickness reduces the width of shadow region during tilting (Figure 5c).
Hence, in the next step, we decided to use a thinner (1.5 μm thick) AZ 15nXT photoresist with a steep sidewall profile.This photoresist results in "fences" of reduced height around the pattern when etched at 3°tilt (see Figure S6, Supporting Information).To remove these "fences," we need an additional etching step.The adopted two-step etch process is shown schematically in Figure 6a.In the first etching step, the inclination of the sample stage is set to 3°for 15 min.This step completely etches the film and leads to structures of desired shape and size.However, it also causes redeposition on the sidewalls.To remove the redeposited material on the sidewalls, the sample stage is tilted to 60°and etched for 12 min in the second step.A pattern after the two-step etching process is shown exemplarily in Figure 6b.The walls of the photoresist and the structure below are free from redeposited material.After removing the photoresist, sharp edges and the twinned martensitic microstructure of epitaxial Ni-Mn-Ga film can be seen on the top surface (Figure 6c).This shows that the martensitic microstructure is retained after patterning.The surface inhomogeneities of the surrounding MgO substrate in Figure 6b,c are due to the topography variations of the martensitic microstructure, [43] which are imprinted on the substrate surface during etching.
Previous studies on dry etching of Ni-Mn-Ga films mainly used Ar plasma etching [24,25] and ion beam etching techniques [44] for microfabrication.No redeposition was reported in the plasma etching studies, possibly due to the lower film thicknesses (75 and 200 nm) compared to our study.However, Schmitt et al. [44] encountered redeposition during ion beam etching of 125-500 nm-thick Ni-Mn-Ga films with PMMA photoresist and used a two-step tilt etching process (30°and 70°) for microfabrication.Similarly, studies on ion beam etching of multilayer stacks have developed two-or multistep etching processes to counteract redeposition. [42,45]In summary, ion beam etching of Ni-Mn-Ga films can be used with a two-step tilt etching process to obtain sharp and clean patterned structures of dimensions below 10 μm.

Fabrication of Freestanding Patterns
As the next step toward commercialization and integration of magnetic shape memory films in actuation microsystems, it is crucial to transfer the epitaxial growth of magnetic shape memory alloys to Si-based substrates and fabricate partly freestanding structures.The MgO(001) substrate used so far in our experiments is incompatible with conventional microsystem technologies.In a recent study, we successfully demonstrated the epitaxial growth of Ni-Mn-Ga films on Si and silicon-on-insulator (SOI) substrates using a 4 nm-thick epitaxial SrTiO 3 buffer. [46]ence, we next selected a 500 nm-thick Ni-Mn-Ga film deposited on SOI substrate to fabricate partly freestanding structures.The developed process flow is shown schematically in Figure 7a.The optimized microfabrication process using ion beam etching can be directly transferred to the films grown on Si-based substrates.) with a 45°tilt.The characteristic twin boundary microstructure of Ni-Mn-Ga martensite is also observed here in both images.Additionally, the etch profiles show an increase in etch rate away from the pattern.This is due to the shadowing effect from sample tilting.The width of the shadow region is indicated by white dotted lines.The difference between etch profiles is discussed in the main text.The black line between Pt cover and film is carbon deposited to protect the surface during Pt deposition.The bright line between the film and the substrate is Cr buffer.c) Optical image of a completely etched sample with a tilt of 45°.Unetched film material of about 2 μm width around the mask remains due to the shadowing effect.d) Illustration of the shadowing effect using a simplified sketch.Tilting the sample during etching creates shadow regions (highlighted in light gray) beside the pattern.The etch rate is lower in these regions than in other parts.
Therefore, we first fabricated Ni-Mn-Ga patterned structures using the two-step ion-beam etching process.Subsequently, a 1.5 μm-thick AZ 5214E photoresist was patterned by photolithography to expose selected regions of the SOI substrate for Si underetching.The top Si layer (7 μm thick) of the SOI substrate was then isotropically etched with XeF 2 gas.Dry chemical etching with XeF 2 is highly selective to Si and the underlying SiO x acts as an effective etch-stop layer.In the final step, the photoresist is removed by O 2 plasma.Figure 7b shows an array of partly freestanding double-beam cantilevers with electrical contact pads.By optimizing the XeF 2 etch time, lateral etching of Si can be controlled to fabricate a high areal density of freestanding structures.The Ni-Mn-Ga double-beam cantilever can be deflected by Joule heating using the thermal shape memory effect. [21]reestanding magnetic shape memory films were fabricated previously by growing on soluble substrates such as NaCl [47,48] and polyvinyl alcohol [6,19] or by underetching the Cr buffer layer on MgO substrate. [24,29,43]However, these substrates are not suitable for existing Si-based microsystem technologies.Hence, additional bonding and integration steps had to be developed to fix the released films on a Si wafer before micromachining. [20]ing epitaxial films grown directly on Si-based substrates, additional steps in microfabrication and integration into microsystems can be avoided.Our process thus significantly simplifies the fabrication of actuation microsystems based on magnetic shape memory films.

Conclusion and Outlook
To summarize our work, we have tested wet chemical etching, reactive ion etching, and ion-beam etching techniques on epitaxial Ni-Mn-Ga films to understand their suitability for microfabrication, prevailing challenges, and solutions.Based on our findings of these etching techniques, we come to the following conclusions.1) Wet etching of Ni-Mn-Ga films is only suitable for large patterned structures of dimensions exceeding tens of micrometers.2) Reactive ion etching based on sulfur hexafluoride chemistry is not suitable for microfabrication of Ni-Mn-Ga.
3) A two-step ion-beam etching technique based on Ar plasma results in sharp and clean Ni-Mn-Ga patterned structures.This technique is suitable for patterned structures of dimensions below 10 μm. 4) Finally, we have used epitaxial Ni-Mn-Ga films Beyond the examined pattern dimensions, we expect that our optimized ion-beam etching process is adaptable to the submicrometer range along when using displacement Talbot lithography or electron beam lithography.However, optimization of mask material (photoresist or hard mask) and its thickness will be necessary to suit the film thickness to be etched.
As an outlook, the developed process flow allows integration of these multifunctional materials in microsystems with a potential to upscale the number density of structures.It also supports conducting extensive testing on design geometry optimization and reliability of the functional element on a single film sample.

Experimental Section
Film Deposition: As described in a previous work, epitaxial Ni-Mn-Ga films were grown on single-crystal MgO(001) substrates by DC magnetron sputtering with a 20 nm-thick Cr buffer layer. [17]Films of different thicknesses (500 and 2000 nm) were deposited from differently alloyed Ni-Mn-Ga targets (Ni 44 Mn 32 Ga 24 , Ni 48 Mn 22 Ga 30 , and Ni 48 Mn 32 Ga 20 ) during the study.For preparing freestanding patterns, Ni-Mn-Ga films were deposited on a commercially available SOI substrate (Lumiphase AG, Switzerland) with a 4 nm-thick SrTiO 3 (001) buffer and a 7 μm-thick Si (001) device layer instead of the MgO (001) substrate.The prepared samples included both austenitic and martensitic films at room temperature.
Microfabrication Process: Microfabrication studies on Ni-Mn-Ga films followed a standard photolithography step and tested different etching techniques.Different photoresists were tested as masks, which included 1.5 μm-thick AZ 5214E, 10 μm-thick AZ 10XT, 7 μm-thick AZ nLOF 2070, 3 μm-thick AZ nLOF 2020, and 1.5 μm-thick AZ 15nXT photoresists.All photoresists were exposed using an MLA 100 laser writer with a wavelength of 365 nm.The AZ 5214E, AZ nLOF 2070, AZ nLOF 2020, and AZ 15nXT photoresists were developed in AZ 726MIF solution and AZ 10XT in AZ 400 K (1:4 diluted with deionized water) solution and removed after etching using standard photoresist removers.For testing a hard mask, a 225 nm-thick SiN x was deposited on Ni-Mn-Ga film by CVD using a SENTECH SI500D plasma-enhanced CVD tool (deposition parameters: chamber pressure, 1.5 Pa; Ar flow, 140 sccm; N 2 flow, 7.8 sccm; SiH 4 flow, 250 sccm; power, 500 W; temperature, 363 K; duration, 588 s).The SiN x layer was patterned by reactive ion etching in CF 4 plasma using a SENTECH EtchLab 200 (parameters: power, 200 W; CF 4 gas flow, 30 sccm; pressure, 5 Pa; duration, 600 s).The reactive ion etching study on Ni-Mn-Ga films was carried out in an Oxford PlasmaLab 100 ICP etching tool using an Ar (10 sccm) þ SF 6 (5 sccm) gas flow, 0.01 mbar chamber pressure, and different combinations of ICP (700-1000 W) and RF (150-300 W) power.To prevent overheating and photoresist damage, the cryostat was set to 263 K.The etch rate after etching was measured using a Bruker DektakXT stylus surface profiler.
The ion-beam etching study of films was performed in a scia Mill 150 tool using an Ar gas flow rate of 12 sccm in an ion source, He substrate back-cooling flow rate of 5 sccm, a beam voltage of 700 V (Figure 6 and 7) or 800 V (Figure 5), sample stage rotation frequency of 10 rotations per minute, and different sample stage tilts.To prevent overheating and baking of photoresist during etching, the chiller temperature was set to 273 K.For preparing freestanding patterns, underetching of the Si layer was done Ni-Mn-Ga pattern is in austenitic phase at room temperature.The debris around pattern is likely to be contaminated material which is not attacked by XeF 2 gas.using a Xactix e2 XeF 2 gas etching system (parameters: chamber pressure, 400 Pa; duration, 60 s).
Microstructure Characterization: The optical images of the films were taken using an OLYMPUS BX53M optical microscope.Scanning electron microscope images of the films were taken in a Zeiss Sigma 300 scanning electron microscope.For studying the etch profile, the FIB cross sections of partially etched films were prepared and studied using an FEI Helios NanoLab 600i device.

Figure 1 .
Figure 1.Results of wet chemical etching of Ni-Mn-Ga films.a) Optical image of the patterns after wet chemical etching.The photoresist mask is still on the sample.Large underetched regions (5-6 μm wide) below the mask are observed.b) Optical image of the patterns after wet chemical etching with a SiN x hard mask (still on the pattern).The underetched regions are relatively less here (%1 μm wide).c) Scanning electron microscope (SEM) image showing rough edges of the patterns after wet chemical etching.d) Illustration of the underetch issue during wet chemical etching (indicated by arrows) leading to poor dimensional control of the patterns.

Figure 2 .
Figure 2. Results of the reactive ion etching of 500 nm-thick Ni-Mn-Ga films.a) SEM image showing the "fences" of redeposited material that remain after removing the AZ 10XT photoresist [ICP power: 900 W, RF power: 250 W]. b) SEM image showing the darker regions of redeposited material inside the pattern edges after removing the AZ nLOF 2070 photoresist [ICP power: 1000 W, RF power: 300 W].The undercut photoresist profile avoids "fences" but allows redeposition regions on the surface of the patterned structure.c) SEM image of a partially etched sample showing redeposition on the top of the AZ nLOF 2070 photoresist [ICP power: 1000 W, RF power: 300 W].

Figure 3 .
Figure 3. Sketch illustrating redeposition during reactive ion etching of Ni-Mn-Ga films with Case A: steep photoresist profile and Case B: undercut photoresist profile sequentially in steps 1-3.In Case A, the redeposition during etching occurs on sidewalls of photoresist and structure, which results in "fences" around the structure.In Case B, the redeposition additionally occurs on the top shadowed unmasked regions of structures.Furthermore, the redeposited material is discontinuous between sidewalls of photoresist and structure due to the undercut photoresist profile.Hence, redeposition on top of structures only remains instead of "fences" after removing photoresist.

Figure 4 .
Figure 4. SEM images showing the formation of patterns during ion beam etching of 2 μm-thick Ni-Mn-Ga films at a) 0°, b) 15°, c) 30°, and d) 45°tilt angles for 10 min using a 7 μm-thick AZ nLOF 2070 photoresist mask.The martensitic microstructure of the films can be clearly seen in the images, consisting of twin boundaries visible as parallel lines at 45°and 90°to the MgO[100] substrate edges.White dotted lines in (b) and (d) depict the orientation of FIB cross sections discussed later in the text.

Figure 5 .
Figure5.FIB cross-section SEM images of ion-beam-etched Ni-Mn-Ga films a) with a 15°tilt and b) with a 45°tilt.The characteristic twin boundary microstructure of Ni-Mn-Ga martensite is also observed here in both images.Additionally, the etch profiles show an increase in etch rate away from the pattern.This is due to the shadowing effect from sample tilting.The width of the shadow region is indicated by white dotted lines.The difference between etch profiles is discussed in the main text.The black line between Pt cover and film is carbon deposited to protect the surface during Pt deposition.The bright line between the film and the substrate is Cr buffer.c) Optical image of a completely etched sample with a tilt of 45°.Unetched film material of about 2 μm width around the mask remains due to the shadowing effect.d) Illustration of the shadowing effect using a simplified sketch.Tilting the sample during etching creates shadow regions (highlighted in light gray) beside the pattern.The etch rate is lower in these regions than in other parts.

Figure 6 .
Figure 6.Optimization of ion beam etching of Ni-Mn-Ga films.a) Illustration showing the two-step tilt etching process for Ni-Mn-Ga microfabrication.b) SEM image taken at 45°tilt showing a pattern with photoresist mask after the sidewall cleaning step.No "fences" are seen here around the pattern.c) SEM image of Ni-Mn-Ga pattern taken at 45°tilt after removing the photoresist.The martensite microstructure of the continuous film is also observed here after the patterning step.The edges are sharp and free from redeposited material.

Figure 7 .
Figure 7. Fabrication of partly freestanding Ni-Mn-Ga structures.a) Process flow for the fabrication of free-standing patterns on an SOI substrate.b) SEM images of freestanding Ni-Mn-Ga structures (left image) and a closer look at one structure (right image) after underetching the Si layer by XeF 2 gas.The Ni-Mn-Ga pattern is in austenitic phase at room temperature.The debris around pattern is likely to be contaminated material which is not attacked by XeF 2 gas.