f epitaxial NiMnGa / InGaAs nanomembranes grown on GaAs substrates into freestanding microtubes

We report the fabrication of Ni2.7Mn0.9Ga0.4/InGaAs bilayers on GaAs (001)/InGaAs substrates by molecular beam epitaxy. To form freestanding microtubes the bilayers have been released from the substrate by strain engineering. Microtubes with up to three windings have been successfully realized by tailoring the size and strain of the bilayer. The structure and magnetic properties of both, the initial films and the rolled-up microtubes, are investigated by electron microscopy, X-ray techniques and magnetization measurements. A tetragonal lattice with c/a 1⁄4 2.03 (film) and c/a 1⁄4 2.01 (tube) is identified for the Ni2.7Mn0.9Ga0.4 alloy. Furthermore, a significant influence of the cylindrical geometry and strain relaxation induced by roll-up on the magnetic properties of the tube is found.


Introduction
Ferromagnetic shape memory alloys (FSMA) are a class of smart materials which are in scope for research due to their multifunctional properties.2][3][4] The interactions among the ferromagnetic and ferroelastic order parameters, applied stress elds and applied magnetic elds are complex and may result in a variety of interesting phenomena.The most promising results to date have been achieved with the ferromagnetic Heusler alloy Ni 2 MnGa and in particular the family of Ni 2+x+y -Mn 1Àx Ga 1Ày .In these alloys magnetically controlled strains of up to 10% have been measured. 5The FSMA effect in these alloys is observed when the material undergoes the martensite transition.][11][12][13][14] Compared to the bulk material, thin lms have better mechanical strength 15,16 and show unique magnetic properties. 17,18Among various fabrication techniques like ion-beam sputtering, 9 pulsed laser deposition, 11 r.f.-sputtering 12 and laser-beam ablation 14 epitaxial growth has been considered as the most promising deposition technique to produce high quality lms with high macroscopic strain.Deposition of epitaxial Ni-Mn-Ga lms has been successfully realized on different substrates e.g.GaAs, 19 MgO, 20 SrTiO 3 , 21 Al 2 O 3 , 22 NaCl 23,24 and YSZ. 25 Frequently polycrystalline lms were obtained, where the FSMA effect was suppressed due to grain boundaries.Recently, free-standing epitaxial Ni-Mn-Ga lms with thickness in the micrometer-range have been fabricated and characterized.However, the structural and magnetic properties of Ni-Mn-Ga nanomembranes remain largely unexplored.Moreover, the control of composition and microstructure in thin lms are challenging issues.
In addition to epitaxial lm growth, strain engineering offers an elegant approach to rearrange nanomembranes into three-dimensional micro-and nanotubes.Generally, the fabrication process requires a selective etching of the underlaying sacricial layer to release the nanomembranes from their substrate and a certain amount of strain in the nanomembrane.Initially, molecular beam epitaxy (MBE) was used to fabricate strained bilayers based on semiconductors (i.e.InGaAs/GaAs [26][27][28] ).More recently, rst magnetic materials such as Fe 3 Si, have been introduced into epitaxial semiconductor/ magnetic radial superlattices. 29,302][33][34] However, the integral magnetic properties of rolled-up Ni-Mn-Ga alloys and their potential for applications have not been studied yet.In this work, the fabrication of In 0.2 Ga 0.8 As/Ni-Mn-Ga nanomembranes on GaAs (001) substrate and their rolled-up freestanding counterparts is described in detail.We investigate the structural and magnetic properties with respect to the sample geometry.

Experimental
Initially, a strained bilayer consisting of 20 nm AlAs and 20 nm In 0.2 Ga 0.8 As was fabricated by III-V molecular beam epitaxy (MBE) on GaAs (001) substrates.Aer the rst epitaxial growth the samples were transferred to a second custom designed ultrahigh-vacuum MBE multichamber equipped with Ni, Mn, and Ga effusion cells.Prior to the growth of Ni-Mn-Ga, the MBE-grown III-V surface was thermally deoxidized at 423 K for 30 min.The removal of the oxide layer was controlled by RHEED analysis.The growth experiments were then performed at 423 K by opening the shutters during 40 min, resulting in a 40 nm thick Ni-Mn-Ga layer.To prepare the microtubes periodic stripes of 100 mm width, separated by 50 mm wide trenches were dened by photolithography along the (100) direction.A solution of H 3 PO 4 (85%)-H 2 O 2 (31%)-H 2 O (2 : 10 : 500) was used to remove the Ni-Mn-Ga/InGaAs/AlAs layers of the trench.The high dilution of the etchants and the short etching time (<1 min) ensured the selective removal of the topmost layers.Aer li-off the Ni-Mn-Ga/InGaAs/AlAs layer stack remained only at the 100 mm stripes.Tube formation was initiated by wet chemical removal of the AlAs sacricial layer with 0.1% HF during 10 min.By releasing the strained bilayer from the substrate rolled-up InGaAs/Ni-Mn-Ga microtubes were prepared.
X-ray diffraction (XRD) q/2q measurements were performed in Brag-Brentano geometry using Cu-K a radiation.Atomic force microscopy (AFM) images were obtained using a 5600LS microscope from Agilent.Ex situ X-ray photoemission spectroscopy (XPS) studies were performed with a PHI 5600 X-ray Photoelectron Spectrometer.The non-monochromatic Mg-K a (hn ¼ 1254.6 eV) X-ray source provides a spectral resolution of 1.1 eV for the Ag 3d5/2 peak.Quantication calculations were done using standard single element sensitivity factors.
Magnetic measurements on the planar samples and microtubes were carried out in a vibrating sample magnetometer (Quantum Design Evercool II) with eld cooling (FC) and zero-eld cooling (ZFC) magnetization measurements with magnetic eld applied along the key directions to the sample.The morphology of the samples was studied by optical microscopy and scanning electron microscopy (SEM) using a FEI NanoSEM 200.To investigate the InGaAs/Ni-Mn-Ga microtube cross section a lamella was prepared by focused ion beam (FIB) cutting in a Zeiss NVision40.The TEM investigations have been done in a FEI Tecnai F30 operating at of 300 kV.The STEM mode was used for EDX analysis and imaging using the high angle annular dark eld detector (HAADF).The planar sample was prepared by mechanical polishing and Ar-ion etching at 3 keV.The cross section of the lm was analyzed with an energy-ltered Philips CM 20 FEG high resolution transmission electron microscope (HRTEM) equipped with Gatan GIF operated at 200 kV.

Surface morphology
The surface of the as prepared epitaxial nanomembranes was analysed by AFM.The topography measurements show a nanocrystalline texture (Fig. 1(a)) and a typical root mean square roughness (RMS) of $1 nm.Aer denition of the stripes by photolithography and selective etching of the trench cross-section analysis was performed by AFM.From the prole taken along the indicated direction in Fig. 1(b) a total height for the Ni-Mn-Ga/InGaAs/AlAs layer stack of 80 nm was obtained, resulting in an effective layer thickness for Ni-Mn-Ga of 40 nm.

XPS analysis
Fig. 2 shows the Ni 2p3/2, Mn 2p3/2 and Ga 2p3/2 peaks of the lm obtained from XPS aer sputtering with Ar ions for 2 min (removed $7 nm of the topmost layer).
The peaks appear at 852 eV, 639 eV and 1116 eV, respectively.The short Ar ion sputtering step removed surface contaminations, e.g.native oxide, and carbon species, and increased signicantly the intensity of the Ni 2p3/2, Mn 2p3/2 and Ga 2p3/ 2 peaks, enabling to determine the composition of the alloy to Ni 2.7 Mn 0.9 Ga 0.4 .It is known from earlier works 35 that the composition is strongly related to the sputtering time.To prove the composition, EDX-analysis was performed inside a SEM (not shown).From the measurements similar composition were obtained for the Ni-Mn-Ga alloy with deviations of <4 at% from the XPS data.

XRD and TEM analysis
The crystallographic structures of the substrate and the constrained Ni-Mn-Ga lm have been analyzed by XRD with q-2q scans.The intensity of the GaAs (004) Bragg peak was used to normalize the XRD patterns.In Fig. 3 there are two peaks from the Ni-Mn-Ga lm, which can be assigned to the (103) and (114) Bragg reections, indicating a tetragonal structure with c/a > 1.
To examine the crystal structure more in detail TEM experiments were performed on the lms.Fig. 4(a) shows the cross section of the lm.The GaAs, AlAs, InGaAs and Ni-Mn-Ga layers can be clearly identied by their different mass-thickness contrast.However, the interface between Ni-Mn-Ga and InGaAs is not clearly visible, allowing only an estimation of the layer thickness.The diffraction pattern taken from the Ni-Mn-Ga layer shows a combination of diffuse rings and spots, indicating a polycrystalline nature (Fig. 4(b)).Aer analyzing the position of the ( 112) and (200) diffraction rings a body centered tetragonal,  Aer releasing the strained InGaAs/Ni-Mn-Ga bilayers from the substrate rolled-up double microtubes orientated along the GaAs (100) axis with a typical radius of 2.5 mm, measured by scanning electron microscopy (SEM), and a length of several millimeters were obtained (Fig. 5(a) and (b)).
Fig. 6 summarizes the results from cross-sectional microscopic analysis.According to the tube radius and the stripe size microtubes with three periods are obtained (Fig. 6(a) and (b)).Dened FIB thinning of tubular structures was adversely affected by the cavity and loose rolling.During lamella preparation only one of the twin tubes shown in Fig. 5(b) survived.The lower layer in Fig. 6(b) is material redeposited during FIB milling (mainly GaAs).The layer thickness of InGaAs and for the Ni-Mn-Ga layer was measured to 20 nm and 40 nm, respectively, in agreement with the nominal deposition parameters and the AFM investigations.The Ni-Mn-Ga layer appears polycrystalline with grain sizes of about 20 nm.An EDX line scan was were carried out over the marked line of the layer stack (Fig. 7(a)).The element intensities allow to clearly distinguish between the three different materials NiMnGa, InGaAs, and GaAs redeposited during FIB cutting (Fig. 7(b)).The compositional analysis of the NiMnGa layer of the tube gave Ni 2.7 -MnGa 0.3 which is close to the XPS data.The oxygen signal is higher in the NiMnGa layer than in the GaAs without remarkable enrichment at the interfaces.This probably originates from oxygen incorporated during deposition.Similar results were observed in other rolled-up structures. 29he crystal structure of the rolled-up Ni-Mn-Ga was investigated by HRTEM (Fig. 6(c)).The diffraction pattern shows a combination of rings and spots, indicating a polycrystalline     That means aer roll-up the same lattice type as for the lm is observed.When comparing the results from the microtube with the lm, a slight mismatch between the lattice parameters of 2%, 3% and 1%, respectively for a, c and c/a can be found.
The difference can be seen as an indication for tensile strain in the NiMnGa lm.During roll-up of the strained bilayer relaxation can take place, accompanied by a decrease of the lattice parameters of NiMnGa.At the same time the compressed InGaAs layer (a ¼ 5.85 A for the lm) is released.

Magnetic properties
Magnetization measurements were carried out on the strained InGaAs/Ni-Mn-Ga bilayer and the corresponding rolled-up tubes along the main directions.The hysteresis curves measured at 10 K for the magnetic eld applied along the two [010] directions (0 and 90 to the tube axis, in-plane) and [001] direction (out-of-plane) are shown in Fig. 8.Typical ferromagnetic behavior and strong magnetic in-plane anisotropy is observed for the lm Fig. 8(a).Saturation magnetizations M s of 560 emu cm À3 at 4 T were obtained when taking into account the lm dimensions.Measurements along both in-plane directions exhibit atypical easy axis rectangular loop with a coercivity (H c ) of $25 mT.The  coercivity along [001] is signicant larger (100 mT) and the hysteresis curve is more tilted.
Aer roll-up the in-plane magnetization curves exhibit a smoother shape, indicating harder switching behavior (Fig. 8(b)).
Furthermore, the hysteresis curve of the in-plane direction (0 , tube long axis) coincides with the out-of-plane direction (tube short axis).The coercivity along [001] is signicant larger ($130 mT) and the hysteresis curve is more tilted.These results indicate that shape anisotropy of the tube plays a minor role.The polycrystalline character of the NiMnGa layer and the absence of a pronounced microstructure allow us to attribute this change of anisotropy due to magnetoelastic effect (stress).Then the change of stress induced by roll-up along the 90 direction can be estimated from the intersection of both hysteresis curves measured along 90 .This gives an average anisotropy eld of H A ¼ 2K/M s of 65 mT, which is related to an anisotropy constant of K ¼ 18.4 kJ m À3 .The difference of stress can be calculated from K ¼ 3/2l s s, where l s is the magnetostrictive constant and s is the stress.Using a value for NiMnGa of l s ¼ 1.3 Â 10 À4 (ref.38) one obtains a difference of stress between lm and rolled-up tube of 94 MPa, indicating a higher tensile stress along the 90 axis of the lm.The orientation of the easy axis of the lm seems to be supported by this stress, originating from the lattice mismatch between GaAs and InGaAs/Ni-Mn-Ga bilayer.Aer roll-up stress release occurs preferentially along the 90 axis (rolling axis) and therefore tilting of the hysteresis curve is observed.As expected from geometry all perpendicular directions to the long tube axis behave similar.Similar observations aer roll-up were made in. 29Roll-up of the strained bilayer induces compression of NiMnGa along the 90 axis.Probably, a tensile strength remains along the 0 direction due to incomplete underetching between neighbouring tubes, resulting in similar magnetization curves as observed in Fig. 8(b).Strain relaxation also can be seen as the reason for strong increase of coercivity along the 0 and 90 axis.Important to note, that the result from stress evolution found by magnetization measurements is in agreement with our data from TEM, showing a small drop of the lattice parameters aer roll-up.
In addition, the temperature dependence of the ZFC and FC magnetizations is shown in Fig. 9 measured in the range 10 # T # 400 K under a magnetic eld of 50 mT.Film (Fig. 9(a)) and tubes (Fig. 9(b)) show similar results.The ZFC magnetization initially increases with decreasing temperature and passes a maximum at T max ¼ 150 K, indicating a transition from superparamagnetic to ferromagnetic behavior.Above T max the samples become magnetically isotropic, which is also observed in the in-plane hysteresis loops measured at 300 K (Fig. 8(c)).Whereas below T max a splitting of the FC and ZFC curves is observed, implying the presence of magnetically blocked states.The ZFC-FC irreversibility found in this work is consistent with reports on polycrystalline Ni 50 Mn 35 Ga 15 samples with tetragonal structure. 39,40Similar to aforementioned mentioned works no martensitic transformations can be seen in the magnetization curves due to the nanocrystalline nature.

Conclusions
In conclusion, we have shown a fabrication route to realize freestanding FSMA based on Ni-Mn-Ga alloys grown by molecular beam epitaxy on GaAs.The evolution of structural and magnetic properties induced by roll-up is investigated in detail.Our results prove the inuence of crystallographic orientation and strain state on structural and magnetic properties of Ni-Mn-Ga nanomembranes and their tubular counterparts.These insights are fundamental in order to realize thin nanomembranes and freestanding three-dimensional FSMA structures with composition of choice for smart applications as compact actuators and microsensors.

Fig. 1
Fig. 1 (a) Atomic force microscopy topography image taken near the boundary of a Ni-Mn-Ga/InGaAs/AlAs stripe and GaAs.(b) Corresponding cross-sectional analysis of the height, with the measurement direction as indicated in (a).
L1 0 structure (I4/mmm) with lattice parameter a ¼ b ¼ 3.58 A, c ¼ 7.27 A, and c/a ¼ 2.03 was found.This result is comparable with diffraction data for Ni 2 Mn 1.2 Ga 0.8 of ref. 36, reporting a ¼ b ¼ 3.7 A, c ¼ 7.1 A. Shao Meng et al. 37 found a ¼ b ¼ 3.9 A, c ¼ 6.4 A for Ni 2.1 Mn 0.9 Ga.However, in the present work an alloy with significant higher amount of Ni was obtained.

Fig. 3
Fig.3XRD pattern of Ni 2.7 Mn 0.9 Ga 0.4 films and GaAs substrate measured at room temperature.

Fig. 4
Fig. 4 (a) HRTEM cross section of the Ni-Mn-Ga/InGaAs/AlAs/GaAs layer stack.(b) Corresponding selected area electron diffraction (SAED) of the Ni-Mn-Ga layer with electron beam orientated along the inplane.

Fig. 5
Fig. 5 Optical microscopy images: (a) taken from the tube array.The rolling direction is indicated with bright arrows.(b) Showing two rolled-up double tubes with $5 mm diameter.
layer.Based on the diffractions rings the crystalline structure can be indexed as body centered tetragonal, L1 0 structure (I4/ mmm; a ¼ b ¼ 3.51 A, c ¼ 7.06 A, c/a ¼ 2.01).

Fig. 6
Fig. 6 (a) SEM cross section of the rolled-up Ni-Mn-Ga/InGaAs bilayer.(b) Corresponding HRTEM image of the marked area in (a).(c) Corresponding selected area electron diffraction (SAED) of the rolledup Ni-Mn-Ga layer.

Fig. 7
Fig. 7 (a) EDX line scan and HAADF brightness inside the rolled up structure.(b) HAADF-STEM image with location of the EDX line scan.The image is located near the right edge of the marked area in Fig.6(a).The line scan spans from a partly thinned NiMnGa layer (upper) to GaAs redeposited in the hollow tube interior.

Fig. 8
Fig. 8 Magnetization curves measured at 10 K with the magnetic field applied along the two [010] directions (0 and 90 to the stripe/tube axis, in-plane) and [001] direction (out-of-plane): (a) of Ni 2.7 Mn 0.9 Ga 0.4 films and (b) after roll-up.(c) Magnetization curves for film and tube measured at 300 K.

Fig. 9
Fig. 9 ZFC and FC magnetization curves for a cooling field of 50 mT with the magnetic field applied along the 0 direction: (a) of Ni 53 -Mn 28 Ga 19 films and (b) after roll-up.