Crystallinity Control of Ferromagnetic Contacts in Stressed Nanowire

Jul 23, 2012 - We report the controlled growth of single-crystalline ferromagnetic contacts through solid state reaction at nanoscale. Single-crystal ...
0 downloads 0 Views 4MB Size
Download PDF
Letter pubs.acs.org/NanoLett

Crystallinity Control of Ferromagnetic Contacts in Stressed Nanowire Templates and the Magnetic Domain Anisotropy Yung-Chen Lin,† Yu Chen,† Renjie Chen,§ Kaushik Ghosh,§ Qihua Xiong,§,∥ and Yu Huang*,†,‡ †

Department of Materials Science and Engineering and ‡California Nano Systems Institute, University of California, Los Angeles, Los Angeles, California § Division of Physics and Applied Physics, School of Physical and Mathematical Sciences and ∥Division of Microelectronics, School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore S Supporting Information *

ABSTRACT: We report the controlled growth of singlecrystalline ferromagnetic contacts through solid state reaction at nanoscale. Single-crystal Mn5Si3 and Fe5Ge3 contacts were grown within stressed Si and Ge nanowire templates, where oxide-shells were used to exert compressive stress on the silicide or germanide. Compared to polycrystalline silicide and germanide structures observed within bare nanowires, the built-in high strain in the oxide-shelled nanostructures alters the nucleation behavior of the ferromagnetic materials, leading to single crystal growth in the transverse/radial direction. Interestingly, the compressive stress is also found to affect the magnetic anisotropy of the ferromagnetic contacts. In-plane and out-of-plane magnetization were observed in Fe5Ge3 for different crystal orientations, showing distinctly preferred domain orientations. These interesting results display the capability to control both the crystallinity and the magnetic anisotropy of ferromagnetic contacts in engineered nanostructures. KEYWORDS: Germanide, silicide, strain, ferromagnetic contacts, crystallinity, magnetic anisotropy intermetallic films have been epitaxially grown on Si or Ge because of their small lattice mismatch with the host materials,30−33 while it is desirable to develop robust growth methods for FM germanide or silicide that are compatible with current device processing and structure. Especially, the growth of nanoscale contacts has been regarded one of the most important aspects in future device applications.34,35 Here, we report the controlled growth of single crystal Mn5Si3 and Fe5Ge3 contacts in single-crystal Si or Ge nanowires (NWs) by solid-state reaction. Mn5Si3 is of the D88 type (space group P63/mmc) crystal structure with Mn atoms located in two different crystallographic sites. Mn5Si3 has been reported to be antiferromagnetic36 and can be converted to room-temperature ferromagnetic material by carbon implantation.37 The iron germanide (Fe5Ge3) adopts the InNi2 (B82) hexagonal structure (space group P63/mcm) with various composition ratios for Fe and Ge. Importantly, the Curie temperature of Fe5Ge3 was reported to be higher than 450 K,38,39 which is desirable for room-temperature spin operation. Si and Ge NWs were grown by the vapor−liquid−solid (VLS) method and were used as the growth template for silicide and germanide NWs. Mn or Fe metal source was

S

ilicides and germanides are crucial for silicon (Si) and germanium (Ge) based nanoelectronics, where they can serve as contacts or interconnects.1−8 It has been reported that the contacts formed through solid state reaction within nanostructures can improve device performance by reducing interface defects and contaminations, as well as by eliminating interface oxides.9−16 In ultrasmall or nanoscale devices, even a few grain boundaries (GBs) in silicide or germanide contacts may cause serious degradation to device reliability due to stress migration or electromigration.17−20 Especially for some small signal device applications, a few GBs may serve as free carrier traps, leading to a loss of carrier polarization, an increase of background noise, and a larger signal hysteresis. To gauge the above issues, the growth of single-crystal contact attracts considerable attention.21−25 For example, spintronics is one of the fields that require high quality magnetic contact and interface.26−28 Extensive research efforts have been dedicated to achieving high quality ferromagnetic (FM) contacts due to their crucial importance in the realization of spintronic devices with high signal-to-noise ratio. However, FM metals (e.g., Fe, Co) on Si or Ge usually form a nonmagnetic dead layer, such as nonmagnetic silicide, germanide, or alloys, during the high temperature processes, leading to a low spin-polarized injection.29 One solution is to directly grow high quality FM silicide or germanide as the contacts with good interfacial properties with Si or Ge. To date, only molecular beam expitaxy (MBE) Fe3Si and Mn5Ge3 thin © 2012 American Chemical Society

Received: June 4, 2012 Revised: July 3, 2012 Published: July 23, 2012 4341

dx.doi.org/10.1021/nl302113r | Nano Lett. 2012, 12, 4341−4348

Nano Letters

Letter

Figure 1. Mn5Si3 formation in the Si/oxide core−shell NWs with various oxide-shell thickness. (a−c) Mn5Si3 grown in bare silicon NWs with 20, 50, and 80 nm diameters, respectively. (d,e) Mn5Si3 grown in 10 nm oxide-shelled Si NWs with 20 and 50 nm core diameters, respectively. (f,g) Mn5Si3 grown in 30 nm oxide-shelled Si NWs with 50 and 80 nm core diameters, respectively. The scale bars are all 100 nm. Arrows indicate the grain boundaries and voids. To the left of the dashed line, single crystal Mn5Si3 is observed.

Figure 2. The growth transition of Mn5Si3 in the bare and oxide-shelled regions along an individual Si NW. (a) A schematic of a Si-AlOx core−shell NW patterned with a Mn pad, near which a NW region without the oxide-shell is presented. (b) A bright field image shows that Mn5Si3 grows in a NW structure depicted in (a). (c) A high-magnification image of the square area in (b), showing irregular grain boundaries. (d) A schematic of a Mn pad and silicon NW diffusion couple patterned with oxide-shelled and bare regions. (e) A bright-field image shows crystal grains in the oxide-shelled region and in the bare region. Arrows indicate the grain boundaries. The area indicated by the circle highlights the volume expansion prior to growth into the shelled NW. (f) A dark-field image of (e). Arrows indicate grain boundary regions. Double-sided arrows indicate single-crystal grains.

deposited onto the NWs. Aluminum oxide (AlOx) shells ∼10− 30 nm were grown by atomic layer deposition (ALD) around the Si or Ge NWs (Figure 1 schematic). (Please see the Supporting Information for details.) The bare and oxidewrapped NWs were then annealed at 650 (or 600) °C to form the single crystal silicide (or germanide) contacts within the NW templates. Si NWs with bare surface, with 10 nm oxide-

shell or with 30 nm oxide-shell are patterned with Mn pads, respectively (Figure 1 schematics). After rapid thermal processing (RTP) for 30 s at 650 °C, Mn5Si3 starts to grow into the Si NWs (Figure 1). In small bare Si NWs (∼20−30 nm), single-crystal Mn5Si3 is obtained (Figure 1a and Supporting Information Figure S1a,b). However, in bare Si NWs with ∼50 and ∼80 nm diameters, polycrystalline Mn5Si3 4342

dx.doi.org/10.1021/nl302113r | Nano Lett. 2012, 12, 4341−4348

Nano Letters

Letter

released, the nucleation rate of Mn5Si3 is still retarded by the limited Mn supply at the Mn5Si3/Si interface possibly due to the slower metal diffusion through the stressed NW region.1,12,41 Hence, these results suggest that we can suppress the growth of polycrystalline silicides in Si NW templates by retarding the nucleation rate of the silicides, which in turn can be controlled by both the external stress around the nucleation interface and the supply of metal atoms to the nucleation interface.1 To understand the phenomena more quantitatively, we take a look at the transformation free energy of the silicides. The nucleation of the silicides is driven by the free-energy change for transformation, ΔG, and will be opposed by the increased surface energy at the silicide/Si interface, σ. The formation of most metal-rich silicides from Si is associated with volume expansion. Within the oxide-shell, as the silicide is forced to adopt a similar cross-sectional area to the Si NW despite the volume expansion, it will experience strain. The strain energy per unit volume, ΔGs, needs to be taken into account of the free energy change, ΔG v. Adopting the homogeneous nucleation model with a circular shape nucleus (see details on Supporting Information on “The decrease of nucleation rate under stress”, Schematic 1),1,12 a nucleus of average radius r will have a free energy change per unit volume given by

(more than one single-crystal grain in the transverse/radial direction) is observed with grain boundaries and with voids formed near the silicide/Si interface (Figure 1b,c). We note that in the following discussions the number of crystal grains all refer to the grains in the transverse direction, unless otherwise noted. In Figures 1b,c,e,g, we observe that polycrystalline Mn5Si3 forms with irregular shapes and with volume expansion to about twice of that of the original Si NWs. For polycrystalline silicide, as the grains are randomly oriented, severe volume expansion is observed in the radial direction, leading to a rough NW morphology. When the compressive oxide-shell is applied, suppression on both void and polycrystalline formation is observed (Supporting Information Figures S2 and S3). For 50 nm Si NWs, 10 nm shell can effectively suppress the void formation near the silicide/Si interface, while the grain boundaries are still observed (Figure 1e and Supporting Information Figure S2). When the oxide-shell is increased to 30 nm, single-crystal Mn5Si3 forms in the 50 nm NWs with flat Mn5Si3/Si interface (no voids) and adopts the original shape and size of the Si NW template (Figure 1f and Supporting Information Figure S3a). In 80 nm Si NWs with 30 nm oxide-shell, polycrystalline Mn5Si3 is still observed, although the voids near the interface or silicide are mostly eliminated (Figure 1g and Supporting Information Figure S3b). We can infer from these observations that both the oxide-shell and the small size of a NW can promote single-crystal Mn5Si3 growth, most possibly through suppressing the number of nuclei in the cross-sectional area of the NWs. To verify the effect of the oxide-shells, experiments were performed to observe the growth transition of Mn5Si3 from bare to oxide-shelled region along a Si NW, as well as the growth of Mn5Si3 through an oxide notch structure (Figure 2). In Figure 2a, Mn5Si3 grows in the bare region first and then into the oxide-shelled region where compressive stress is applied by the shell.40 In the bare NW section that is close to the Mn pad, it is observed that Mn5Si3 grows with polycrystalline structure where randomly orientated grains and irregular (nonflat) grain boundaries are observed (Figure 2b,c; Supporting Information Figure S4). However, once Mn5Si3 grows into the oxide-shelled region, single crystalline Mn5Si3 forms. This polycrystalline to single crystalline structure transition induced by the oxide-shell corresponds well with our hypothesis that the polycrystalline formation is suppressed in the stressed NW region. The second transition structure is an oxide notch created in the middle of a Si NW, as shown in Figure 2d, where a Si NW is partially wrapped with oxide-shell at two ends. Thus, Mn5Si3 first grows into the oxide-shelled region experiencing the compressive stress, then through the notch region where the compression is released, and finally back to the compressed region. TEM analysis indicates that for Mn5Si3 grown in the compressed region, single crystal grain can grow up to more than 500 nm long in the axial direction (Figure 2e,f; Supporting Information Figures S5 and S6). Once Mn5Si3 grows out of the oxideshelled (compressed) region, the grain size in the axial direction is reduced to around 200−300 nm (Figure 2e,f). This indicates the oxide-shell can suppress the formation of silicide grain boundaries, leading to larger grain size in the NW axial direction. Interestingly, the multiple grain boundaries across the transverse direction as those observed in Figures 1c,e and 2b,c, are rarely observed in this notch region. The difference here (Figure 2d) is that Mn5Si3 grows initially as one grain on the transverse direction because of the compressive stress. When it later grows into the bare NW region, although the stress is

ΔGn = 2πraσ − aπr 2(ΔGv − ΔGs)

Where a, σ, ΔGv and ΔGs represents one-atom thickness, surface energy, bulk Gibbs free energy change, and strain energy change, respectively. With the increasing radius of the nucleus, the free energy of the nucleus increase first (due to the increase of surface energy) and then decreases (due to the decreasing free energy), passing a maximum at a critical size, r*. r* can be represented by the following equation σ r* = (ΔGv − ΔGs) And thus the nucleation activation energy, ΔG* = πσ2a/(ΔGv − ΔGs) Since the strain energy, ΔGs, is always positive, it lowers the driving force, ΔGv − ΔGs, to grow a nucleus. Thus, the critical radius, r*, and the free energy of critical nuclei, ΔG*, will increase due to the strain energy built-up in a stressed structure. The rate of nucleation, ρ*, is proportional to the concentration of the critical nuclei, and can be expressed as ⎛ ΔG* ⎞ ⎛ Q ⎞ ⎟ ⎟exp⎜ − ρ* = K exp⎜ − ⎝ kT ⎠ ⎝ kT ⎠

Where K is a constant and Q is related to the activation energy of diffusion. Under strain the nucleation rate will decrease due to the increasing nucleation barrier energy, ΔG*. Enhanced single-crystal growth of Mn5Si3 in oxide-shelled NWs can hence be attributed to the lower nucleation rate due to the compressive stress. If we assume Mn5Si3 is under 10% strain, which is close and at the same order of magnitude to reported values in NiSi/Si, PtSi/Si, and MnSi/Si NW structures,5,6,13 the strain energy is around 0.19 eV/atom (see details in Supporting Information, Strain energy calculation). For most stable silicide materials, the formation energy42 is around 0.5 eV/atom. Thus, under high strain, the strain energy at nanoscale can be elevated to the same order of magnitude with the formation energy. Here, due to small (negligible) entropy change (ΔS) in solidstate transformation, we can use the formation enthalpy (ΔH to 4343

dx.doi.org/10.1021/nl302113r | Nano Lett. 2012, 12, 4341−4348

Nano Letters

Letter

represent the free energy change (ΔG).42 Including the strain energy to the total Gibbs free energy, the nucleation rate under strain will dramatically decrease for manganese silicide (ρ*/ ρ*under strain = 8.8 × 1030; see Supporting Information on The decrease of nucleation rate under stress). On the other hand, the nucleation (or interfacial) controlled silicides such as NiSi2 and CoSi2 often have higher nucleation activation barrier (ΔG*), leading to a lower nucleation rate. This attributes to almost all NiSi2 formed in Si NW templates being singlecrystal,40,43 consistent with our explanation of nucleation ratecontrolled crystallinity in stressed NW templates. According to our experimental observations and theoretical estimation, strain energy could be an important term influencing material formation at nanoscale by elevating the nucleation activation barrier. Prior to this study, it is generally believed that as the formation energy (ΔH, driving force; ΔH ≅ ΔG) for most of compound materials such as silicide or germanide is large, the growth of these material systems is dominated by diffusion-control; and hence nucleation is regarded as a minor factor.1 However, at nanoscale high strain energy can build up in the confined structures, leading to interesting growth behaviors related to nucleation events around the interface. Thus, we believe that compressive oxide-shell assisted single-crystal growth can be extended to broad material systems at nanoscale. To prove our assumption, we implement our growth method to other functional materials such as single-crystal ferromagnetic contacts for spintronics. As a demonstration of a different phase transformation system at nanoscale, we have grown single-crystal FM germanide in the Ge/oxide (∼20−30 nm) core−shell NWs. Similar to silicide systems, with the aid of the oxide-shells single crystal Fe5Ge3 can grow into the Ge/oxide core−shell NWs from the Fe pads with atomically sharp interface as shown in Figure 3a,b (please find experimental details in the Supporting Information ). In the bare Ge NWs, polycrystalline germanides are obtained, and void formations are frequently observed near the interface (Supporting Information Figure S7a). For the oxide-shelled NWs in the region right next to the Fe pad (∼100−200 nm) polycrystalline Fe5Ge3 (no voids) is observed, which transforms into single crystal structure with further growth into the shelled region (Supporting Information Figures S7b,c), confirming the oxideshell’s effect on suppressing void formation and promoting single-crystal growth. To further examine the crystal structure, selective-area electron diffraction pattern (SAED) shows a single-crystal structure corresponding to the Fe5Ge3 lattice near the interface with grain size ∼1 μm long in the NW axial direction (Figure 3a inset). The energy dispersive X-ray spectrum (EDS) analysis (Figure 3c) verifies the iron-rich composition and gives the Fe/Ge ratio ∼1.5 ± 0.15 for most germanide NWs (>20 NWs). High-resolution TEM was used to analyze the crystal structures of Fe5Ge3 NWs. Two different diameters of Ge NWs, ∼50 and ∼30 nm, wrapped in 30 nm oxide-shell were used as the germanide growth templates. Surprisingly, two epitaxial relations, the Fe5Ge3 c-axis//Ge NW axial direction and the c-axis tilted from Ge NW axial axis, were observed in the ∼50 and ∼30 nm NWs, respectively. For NWs with ∼50 nm diameter, the epitaxial relation, Ge[11̅0]//Fe5Ge3[110] and Ge(111)//Fe5Ge3(001), was observed with a lattice mismatch ∼0.5% as shown in Figures 3d−f. The same epitaxial relation was often observed in the single-crystal Fe5Ge3 thin film grown on Ge(111) substrate.44,45 The schematic of the Fe5Ge3 crystal

Figure 3. The growth of Fe5Ge3 NW and Fe5Ge3/Ge heterostructure. (a) Bright-field image of Fe5Ge3/Ge NW heterostructure. A Ge NW is patterned with Fe pad and wrapped with AlOx shell. The structure is annealed at 600 °C to grow Fe5Ge3. Selective area electron diffraction (SAED) pattern (inset) indicates Fe5Ge3[110] zone axis. (b) An enlarged image around the interface. The dashed line indicates the thickness of the ALD AlOx shell. (c) The energy dispersion spectrum (EDS) of the Fe5Ge3 NW. The analyzed location is indicated by the asterisk. (d) HRTEM image of the Fe5Ge3/Ge epitaxial interface in a Ge NW of ∼50 nm diameter. (e,f) FFT of Ge[11̅0] zone axis and Fe5Ge3[110] zone axis, respectively. (g) HRTEM image of the Fe5Ge3/Ge epitaxial interface in a ∼27 nm diameter Ge NW. (h,i) FFT along Ge[1̅1̅2] zone axis and Fe5Ge3[5̅4̅2] zone axis, respectively. (j) The schematic shows the c-axis of Fe5Ge3 NW is parallel to the NW axial direction (white dashed arrow). It represents the Fe5Ge3 crystal orientation for the interface epitaxial relation in (d) for the ∼50 nm NW. (k) The schematic shows the c-axis of Fe5Ge3 crystal is tilted 70.18° in regard to the NW axis (white dash arrow). It represents the Fe5Ge3 crystal orientation in the epitaxial interface relation in (g) for the ∼27 nm NW.

orientation related to the NW is shown in Figure 3j. The c-axis of the hexagonal Fe5Ge3 is parallel to the axial direction of the 50 nm Ge NW. For the Ge NWs with ∼30 nm diameter, another epitaxial relation, Ge[2̅ 2 0]//Fe 5 Ge 3 [01̅ 2̅ ] and Ge(111)//Fe5Ge3(2̅21̅), with ∼1.8% lattice mismatch is consistently adopted (Figure 3g−i), where the c-axis of Fe5Ge3 crystal deviates from the axial direction by ∼70.18° (Figure 3k). We suggest that the latter epitaxial relation is a result of the crystal’s attempt to accommodate a lower energy configuration during the fast dynamic growth. It has been reported that different epitaxial relations or random crystal orientations may be observed owing to the different rates of incorporating atoms.46 In Fe5Ge3 thin film grown by MBE, 4344

dx.doi.org/10.1021/nl302113r | Nano Lett. 2012, 12, 4341−4348

Nano Letters

Letter

Figure 4. The magnetic properties of Fe5Ge3 NWs (a) SEM image of ∼30 nm Fe5Ge3/Ge NW heterostructure. Brighter contrast of the NW is the Fe5Ge3 portion. (b) The corresponding AFM image and (c) MFM phase image, which displays an out-of-plane magnetization. (d) The enlarged MFM image of (c), showing a single domain and out-of-plane magnetization close to the interface. The length of single domain is ∼1 μm. The arrow indicates out-of-plane magnetization. (e) The MFM image corresponding to (d) with reversed tip magnetization. (f) A typical AFM image of 50 nm Fe5Ge3/Ge heterstructure NW and (g) the corresponding MFM image. It shows an in-plane magnetization close to the Fe5Ge3/Ge interface with the magnetic polarization indicated by the arrows. (h,i) Schematics of the MFM measurement geometry that led to the (d) out-of-plane magnetization and (g) in-plane magnetization, respectively.

the magnetization of the MFM tip) of the Fe5Ge3 NW shows a reversed contrast image (Figure 4e) of similar shape and length to Figure 4d, confirming the single domain nature of the Fe5Ge3 contact. In contrast, the region close to the Fe pad shows a more random domain structure together with volume expansion (Figure 4d,f,g and Supporting Information Figure S10), which is a result of the polycrystal Fe5Ge3 formation in this area (Supporting Information Figure S7b,c). In the ∼50 nm Ge NW templates, the Fe5Ge3 contact close to the Fe5Ge3/ Ge interface shows an in-plane magnetization, the direction of which is parallel to the NW axis (Figure 4f and Supporting Information Figure S10) The schematics shown in Figure 4h (out-of-plane magnetization) and Figure 4i (in-plane magnetization) describe the magnetization polarization of the two different Fe5Ge3/Ge epitaxial relations in ∼30 nm NWs (Figure 4d) and ∼50 nm NWs (Figure 4g), respectively. The out-of-plane and in-plane magnetization directions shown for ∼30 nm and ∼50 nm Fe5Ge3 NWs indicate the existence of magnetic anisotropy for the Fe5Ge3 nanostructures. The magnetic anisotropy can be induced by three mechanisms: shape anisotropy, magnetocrystalline anisotropy and stress anisotropy. From our results, we infer that the shape anisotropy plays a minor role. If the shape anisotropy is dominant we should expect the magnetization of the smaller Fe5Ge3 NWs to lie in the NW axial direction due to the higher aspect ratio. However the opposite trend is observed in our studies: i.e. the magnetization of ∼50 nm NW lies in the axial direction, while that of the smaller ∼30 nm Fe5Ge3 NWs does not. We suggest that in these Fe5Ge3 nanostructures the mangetocrystalline and the stress anisotropy play more significant roles in magnetization than the shape anisotropy. To better verify the origin of the magnetic anisotropy in Fe5Ge3 NWs, crystal orientations are compared with the directions of the magnetic domains (Figure 3 and Figure 4). We note that the magnetization of 50 nm NWs is in-plane with c-axis lying in the NW axial direction; and the magnetization of 30 nm NWs is out-of-plane with tilted

limited metal sources are supplied leading to a slow growth rate. In such a case, the smallest lattice mismatch, Ge[11̅0]// Fe5Ge3[110] and Ge(111)//Fe5Ge3(001), is always obtained on Ge(111) substrate.45 According to our observations, Fe5Ge3 usually grows faster in the smaller diameter Ge NWs. With the faster growth rate in a smaller NW, it becomes more difficult for the Fe5Ge3/Ge interface to adopt a more close-packed structure (Ge(111)//Fe5Ge3(001)), which involves more atom jumping or bond breaking of Ge crystals. By adopting the tilted crystal orientation, the Fe5Ge3 crystal accommodates the less ideal epitaxial relation Ge(111)//Fe5Ge3(2̅21̅) with ∼1.8% lattice mismatch in order to avoid the formation of more defective grain boundaries (Supporting Information Figure S8).46 To investigate the magnetic properties of the single-crystal Fe5Ge3 contacts the magnetic force microscope (MFM) was utilized to explore the magnetic domains at room temperature. For a sample that was annealed at 600 °C for 20 s, a clear contrast between Fe5Ge3 and Ge in a ∼30 nm NW can be visualized with scanning electron microscope (SEM) (Figure 4a). The same sample was then scanned by atomic force microscope (AFM) shown in Figure 4b, showing no obvious volume change after the germanide formation. The diameter of the core Ge NW was ∼28 nm as verified by the height profile of ∼48 nm in AFM that includes ∼20 nm ALD oxide-shell thickness. MFM analysis shows a clear phase contrast between Fe5Ge3 and Ge. Only Fe5Ge3 shows a phase shift in the interleave mode, demonstrating that the Fe5Ge3 contact was ferromagnetic at room temperature. The length of Fe5Ge3 region mapped by MFM is also consistent with the observation by SEM (Figure 4a−c). The MFM analysis shows a bright contrast of ∼1 μm from Fe pad to the Fe5Ge3/Ge interface (Figure 4c), corresponding to a single-domain with out-ofplane magnetization (see the enlarged image is shown in Figure 4d, and see Supporting Information Figure S9 for more MFM phase characterizations). Furthermore, reverse MFM (reversing 4345

dx.doi.org/10.1021/nl302113r | Nano Lett. 2012, 12, 4341−4348

Nano Letters

Letter

c-axis ∼70.18° from the NW axis. According to previous studies,45,47 in Fe5Ge3 the c-axis is the magnetic hard axis and the a axis in the hexagonal crystal basal plane is the easy axis. As the magnetization aligned with c-axis will lead to higher energy state, the spontaneous magnetization will tend to lie in the basal plane, resulting in magnetocrystalline anisotropy. Here, in an ∼50 nm Fe5Ge3 NW the magnetic polarization lies in the NW axial direction that is the magnetization hard axis, deviating from the expectation when the magnetocrystalline anisotropy is dominant.45 We therefore need to take into account the stress anisotropy, as a thick oxide layer (∼30 nm) is used to apply considerable compressive stress (or residual stress) in the radial direction of the NW. The stress anisotropy is related to the spin−orbit coupling called the magnetostriction.47 Upon magnetization, a previously demagnetized crystal experiences a strain that can be measured as a function of applied magnetic field along the principal crystallographic axis. The inverse effect, the change of magnetization direction under the applied stress, will also occur. A uniaxial stress can produce a unique easy axis of magnetization if the stress induced field is sufficient to overcome all other anisotropic fields. Fe5Ge3 has been reported to have positive magnetostriction constant in the c-axis and a negative magnetostriction constant in the a-axis direction in the basal plane, suggesting that the compressive stress on the basal plane will render c-axis as the unique easy axis and that the tensile stress on the c-axial direction will favor a-axis as an easy axis. In the core−shell (Fe5Ge3/oxide) NWs, the oxide shells exert a compressive stress in the radial direction. For a 50 nm NW, the c-axis aligns with the axial direction. As the a-axis in the hexagonal basal plane is stressed compressively by the oxide-shell the stressinduced anisotropy may dominate over crystal orientation anisotropy, causing the switch of domain magnetization to form a-axis to c-axis. For a ∼30 nm Fe5Ge3 NW, the c-axis tilts ∼70.18° from NW axial direction (Figure 3k) and the basal plane tilts ∼70.18° from the radial direction. In this case, we expect the stress to be applied to both the basal plane and the caxis, where the observed out-of-plane magnetization is expected. Although further experiments are necessary to investigate the direction of the out-of-plane anisotropy in the stressed core−shell Fe5Ge3 NWs. Theoretically, the stress-induced anisotropy field is given by 3λσ/Ms,48 in which λ(001, 100) ∼ −20 × 10−6 is the magnetostriction coefficient, σ = Yε is the uniaxial stress and Ms is the saturation magnetization, the Young’s modulus, Y, is around 1.1 × 1012 dyn/cm2,49 and ε represents the strain value. The magnetocrystalline anisotropy field is given by 2K1/Ms with the magnetocrystalline anisotropy constant, K1 = 5.1 × 106 erg/cm3.47 The negative value of λ(001,100) indicates that a compressive strain measured along [100] (the a-axis) when the magnetization is parallel to [001] (the c-axis). If we assume the compression strain ε is ∼5−10% in the nanostructure,4−6,13 it gives a stress-induced anisotropy field ∼3.3 × 106/Ms to 6.6 × 106/Ms, which tends to align the magnetization along the c-axis. For the magnetocrystalline anisotropy, the induced anisotropy field which forces the magnetization toward a-axis is ∼10.2 × 106/Ms. Those two anisotropy fields are on the same order of magnitude for Fe5Ge3. However, the estimation above only considers the uniaxial stress (stress applied on one axis) in the material, while in the core−shell NWs the oxide-shell can provide multiple stress components that contribute to the stress-induced anisotropy field. For example, if we approximate

the stress from the oxide-shell to a biaxial stress, the stressinduced anisotropy filed can be doubled compared to the uniaxial stress. Therefore, the stress-induced anisotropy can well dominate over the magnetic anisotropy in the NW structures with the wrap-around oxide-shell. Conclusions. In conclusion, single-crystal Mn5Si3 and Fe5Ge3 contacts are grown within stressed nanowire structures with the oxide-shells. Single-crystal or polycrystalline Mn5Si3 growth has been systematically studied with various oxide thickness and Si NW diameters. It was found that the singlecrystal growth of Mn5Si3 in NW template can be achieved by applying a compressive stress from a thick oxide-shell (∼30 nm), which can significantly decrease the nucleation rate at the Mn5Si3/Si interface and leads to single grain growth in the transverse/radial direction. Similar approach has been applied to grow single-crystal FM Fe5Ge3 contact within the Ge/AlOx core−shell NWs. Two different epitaxial relations of the Fe5Ge3/Ge interfaces were found in different sized Ge NW templates, which in turn lead to differently preferred magnetic domain orientations. We further demonstrated that the magnetic polarization of the Fe5Ge3 contact is dominated by stress-induced magnetic anisotropy. In short, our studies open the pathway to control both the crystallinity and the direction of magnetization for ferromagnetic materials in engineered nanostructures, which is of great importance for future device applications including nanoscale spintronics.

■

ASSOCIATED CONTENT

S Supporting Information *

Experimental methods; additional TEM analysis and highmagnification views of Mn5Si3 and Fe5Ge3 structures; magnetic force microscope analysis of Fe5Ge3 structures; strain energy calculation for silicides in the core−shell NWs and the estimation of nucleation rate decrease under stress. This material is available free of charge via the Internet at http:// pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This project is supported by the Focus Center Research Program (FCRP)-Center on Functional Engineered Nano Architectonics (FENA). Y.H. acknowledges support from Henry Samueli School of Engineering and Applied Science Fellowship. Q.X. gratefully acknowledges Singapore National Research Foundation via a NRF fellowship grant (NRFRF2009-06) and start-up grant (M58110061) from Nanyang Technological University.

■

REFERENCES

(1) Gambino, J. P.; Colgan, E. G. Silicides and Ohmic Contacts. Mater. Chem. Phys. 1998, 52, 99−146. (2) Lavoie, C.; d’Heurle, F. M.; Detavernier, C.; Cabral, C. Towards Implementation of a Nickel Silicide Process for CMOS Technologies. Microelectron. Eng. 2003, 70, 144−157. (3) Zhu, S. Y.; Li, R.; Lee, S. J.; Li, M. F.; Du, A. Y.; Singh, J.; Zhu, C. X.; Chin, A.; Kwong, D. L. Germanium P-MOSFETs with SchottkyBarrier Germanide S/D, High-Kappa Gate Dielectric and Metal Gate. IEEE Electron Device Lett. 2005, 26, 81−83.

4346

dx.doi.org/10.1021/nl302113r | Nano Lett. 2012, 12, 4341−4348

Nano Letters

Letter

(4) Wu, Y.; Xiang, J.; Yang, C.; Lu, W.; Lieber, C. M. Single-Crystal Metallic Nanowires and Metal/Semiconductor Nanowire Heterostructures. Nature 2004, 430, 61−65. (5) Lin, Y. C.; Chen, Y.; Shaios, A.; Huang, Y. Detection of Spin Polarized Carrier in Silicon Nanowire with Single Crystal MnSi as Magnetic Contacts. Nano Lett. 2010, 10, 2281−2287. (6) Lin, Y. C.; Lu, K. C.; Wu, W. W.; Bai, J. W.; Chen, L. J.; Tu, K. N.; Huang, Y. Single Crystalline PtSi Nanowires, PtSi/Si/PtSi Nanowire Heterostructures, and Nanodevices. Nano Lett. 2008, 8, 913−918. (7) Tang, J. S.; Wang, C. Y.; Xiu, F. X.; Lang, M. R.; Chu, L. W.; Tsai, C. J.; Chueh, Y. L.; Chen, L. J.; Wang, K. L. Oxide-Confined Formation of Germanium Nanowire Heterostructures for HighPerformance Transistors. ACS Nano 2011, 5, 6008−6015. (8) Chen, Y.; Lin, Y. C.; Huang, C. W.; Wang, C. W.; Chen, L. J.; Wu, W. W; Huang, Y. Kinetic Competition Model and SizeDependent Phase Selection in 1-D Nanostructures. Nano Lett. 2012, 12, 3115−3120. (9) Kittl, J. A.; Lauwers, A.; Chamirian, O.; Van Dal, M.; Akheyar, A.; De Potter, M.; Lindsay, R.; Maex, K. Ni- and Co-Based Silicides for Advanced CMOS Applications. Microelectron. Eng. 2003, 70, 158−165. (10) Foggiato, J.; Yoo, W. S.; Ouaknine, M.; Murakami, T.; Fukada, T. Optimizing the Formation of Nickel Silicide. Mat. Sci. Eng., B 2004, 114, 56−60. (11) Woodruff, S. M.; Dellas, N. S.; Liu, B. Z.; Eichfeld, S. M.; Mayer, T. S.; Redwing, J. M.; Mohney, S. E. Nickel and Nickel Silicide Schottky Barrier Contacts to N-Type Silicon Nanowires. J. Vac. Sci. Technol., B 2008, 26, 1592−1596. (12) Chen, L. J.; Wu, W. W. In Situ Tem Investigation of Dynamical Changes of Nanostructures. Mat. Sci. Eng., R 2010, 70, 303−319. (13) Chou, Y. C.; Wu, W. W.; Chen, L. J.; Tu, K. N. Homogeneous Nucleation of Epitaxial CoSi2 and NiSi in Si Nanowires. Nano Lett. 2009, 9, 2337−2342. (14) Lu, K. C.; Wu, W. W.; Wu, H. W.; Tanner, C. M.; Chang, J. P.; Chen, L. J.; Tu, K. N. In Situ Control of Atomic-Scale Si Layer with Huge Strain in the Nanoheterostructure NiSi/Si/NiSi through Point Contact Reaction. Nano Lett. 2007, 7, 2389−2394. (15) Xiang, J.; Lu, W.; Hu, Y. J.; Wu, Y.; Yan, H.; Lieber, C. M. Ge/Si Nanowire Heterostructures as High-Performance Field-Effect Transistors. Nature 2006, 441, 489−493. (16) Appenzeller, J.; Knoch, J.; Bjork, M. I.; Riel, H.; Schmid, H.; Riess, W. Toward Nanowire Electronics. IEEE Trans. Electron Devices 2008, 55, 2827−2845. (17) Jarmar, T.; Seger, J.; Ericson, F.; Mangelinck, D.; Smith, U.; Zhang, S. L. Morphological and Phase Stability of NickelGermanosilicide on Si1‑xGex under Thermal Stress. J. Appl. Phys. 2002, 92, 7193−7199. (18) Koubuchi, Y.; Onuki, J.; Suwa, M.; Fukada, S.; Moribe, S.; Tanigaki, Y. Stress Migration Resistance and Contact Characterization of Al-Pd-Si Interconnects for Very Large-Scale Integrations. J. Vac. Sci. Technol. B 1990, 8, 1232−1238. (19) Tu, K. N. Recent Advances on Electromigration in Very-LargeScale-Integration of Interconnects. J. Appl. Phys. 2003, 94, 5451−5473. (20) Hu, C. K.; Rosenberg, R.; Lee, K. Y. Electromigration Path in Cu Thin-Film Lines. Appl. Phys. Lett. 1999, 74, 2945−2947. (21) Fleetwood, D. M.; Winokur, P. S.; Reber, R. A.; Meisenheimer, T. L.; Schwank, J. R.; Shaneyfelt, M. R.; Riewe, L. C. Effects of Oxide Traps, Interface Traps, and Border Traps on Metal-Oxide-Semiconductor Devices. J. Appl. Phys. 1993, 73, 5058−5074. (22) Evetts, J. E.; Blamire, M. G.; Mathur, N. D.; Isaac, S. P.; Teo, B. S.; Cohen, L. F.; Macmanus-Driscoll, J. L. Defect-Induced Spin Disorder and Magnetoresistance in Single-Crystal and Polycrystal Rare-Earth Manganite Thin Films. Philos. Trans. R. Soc. London, Ser. A 1998, 356, 1593−1613. (23) Gupta, A.; Gong, G. Q.; Xiao, G.; Duncombe, P. R.; Lecoeur, P.; Trouilloud, P.; Wang, Y. Y.; Dravid, V. P.; Sun, J. Z. Grain-Boundary Effects on the Magnetoresistance Properties of Perovskite Manganite Films. Phys. Rev. B 1996, 54, 15629−15632.

(24) DeGrave, J. P.; Schmitt, A. L.; Selinsky, R. S.; Higgins, J. M.; Keavney, D. J.; Jin, S. Spin Polarization Measurement of Homogeneously Doped Fe1‑xCoxSi Nanowires by Andreev Reflection Spectroscopy. Nano Lett. 2011, 11, 4431−4437. (25) Higgins, J. M.; Ding, R. H.; DeGrave, J. P.; Jin, S. Signature of Helimagnetic Ordering in Single-Crystal MnSi Nanowires. Nano Lett. 2010, 10, 1605−1610. (26) Schmehl, A.; Vaithyanathan, V.; Herrnberger, A.; Thiel, S.; Richter, C.; Liberati, M.; Heeg, T.; Rockerath, M.; Kourkoutis, L. F.; Muhlbauer, S.; Boni, P.; Muller, D. A.; Barash, Y.; Schubert, J.; Idzerda, Y.; Mannhart, J.; Schlom, D. G. Epitaxial Integration of the Highly Spin-Polarized Ferromagnetic Semiconductor EuO with Silicon and GaN. Nat. Mater. 2007, 6, 882−887. (27) Jonker, B. T. Progress toward Electrical Injection of SpinPolarized Electrons into Semiconductors. Proc. IEEE 2003, 91, 727− 740. (28) Anguelouch, A.; Gupta, A.; Xiao, G.; Abraham, D. W.; Ji, Y.; Ingvarsson, S.; Chien, C. L. Near-Complete Spin Polarization in Atomically-Smooth Chromium-Dioxide Epitaxial Films Prepared Using a CVD Liquid Precursor. Phys. Rev. B 2001, 64, 180408R. (29) Tong, L. N.; Pan, M. H.; Wu, J.; Wu, X. S.; Du, J.; Lu, M.; Feng, D.; Zhai, H. R.; Xia, H. Magnetic and Transport Properties of Sputtered Fe/Si Multilayers. Eur. Phys. J. B 1998, 5, 61−66. (30) Lin, F. T.; Jiang, D. M.; Ma, X. M.; Shi, W. Z. Structural Order and Magnetic Properties of Fe3Si/Si(100) Heterostructures Grown by Pulsed-Laser Deposition. Thin Solid Films 2007, 515, 5353−5356. (31) Sellai, A.; Mesli, A.; Petit, M.; Le Thanh, V.; Taylor, D.; Henini, M. Barrier Height and Interface Characteristics of Au/Mn5Ge3/ Ge(111) Schottky Contacts for Spin Injection. Semicond. Sci. Technol. 2012, 27. (32) Spiesser, A.; Dau, M. T.; Michez, L. A.; Petit, M.; Coudreau, C.; Glachant, A.; Thanh, V. L. Boosting the Curie Temperature in Carbon-Doped Mn5Ge3/Ge Heterostructures. Int. J. Nanotechnol. 2012, 9, 428−438. (33) Miyoshi, T.; Matsui, T.; Tsuda, H.; Mabuchi, H.; Morii, K. Magnetic and Electric Properties of Mn5Ge3/Ge Nanostructured Films. J. Appl. Phys. 1999, 85, 5372−5374. (34) Morimoto, T.; Ohguro, T.; Momose, H. S.; Iinuma, T.; Kunishima, I.; Suguro, K.; Katakabe, I.; Nakajima, H.; Tsuchiaki, M.; Ono, M.; Katsumata, Y.; Iwai, H. Self-Aligned Nickel-Mono-Silicide Technology for High-Speed Deep-Submicrometer Logic CMOS ULSI. IEEE Trans. Electron Devices 1995, 42, 915−922. (35) Iwai, H.; Ohguro, T.; Ohmi, S. NiSi Salicide Technology for Scaled CMOS. Microelectron. Eng. 2002, 60, 157−169. (36) Gopalakrishnan, B.; Surgers, C.; Montbrun, R.; Singh, A.; Uhlarz, M.; Von Lohneysen, H. Electronic Transport in Magnetically Ordered Mn5Si3Cx Films. Phys. Rev. B 2008, 77, 104414. (37) Surgers, C.; Potzger, K.; Strache, T.; Moller, W.; Fischer, G.; Joshi, N.; Von Lohneysen, H. Magnetic Order by C-Ion Implantation into Mn5Si3 and Mn5Ge3 and Its Lateral Modification. Appl. Phys. Lett. 2008, 93, 062503. (38) Stroppa, A.; Kresse, G.; Continenza, A. Spin Polarization Tuning in Mn5‑xFexGe3. Appl. Phys. Lett. 2008, 93, 092502. (39) Austin, A. E. Magnetic Properties of Fe5Ge3-Mn5Ge3 Solid Solutions. J. Appl. Phys. 1969, 40, 1381−1382. (40) Lin, Y. C.; Chen, Y.; Xu, D.; Huang, Y. Growth of Nickel Silicides in Si and Si/SiOx Core/Shell Nanowires. Nano Lett. 2010, 10, 4721−4726. (41) Chou, Y. C.; Lu, K. C.; Tu, K. N. Nucleation and Growth of Epitaxial Silicide in Silicon Nanowires. Mat. Sci. Eng., R 2010, 70, 112− 125. (42) Pretorius, R.; Marais, T. K.; Theron, C. C. Thin-Film Compound Phase-Formation Sequence -an Effective Heat of Formation Model. Mat. Sci. Eng., R 1993, 10, 1−83. (43) Dellas, N. S.; Liu, B. Z.; Eichfeld, S. M.; Eichfeld, C. M.; Mayer, T. S.; Mohney, S. E. Orientation Dependence of Nickel Silicide Formation in Contacts to Silicon Nanowires. J. Appl. Phys. 2009, 105. (44) Belmeguenai, M.; Berling, D.; Garreau, G.; Cherif, S. M.; Faurie, D.; Seiler, W.; Moch, P. Static and Dynamic Magnetic Properties of 4347

dx.doi.org/10.1021/nl302113r | Nano Lett. 2012, 12, 4341−4348

Nano Letters

Letter

Epitaxial Fe1.7Ge Thin Films Grown on Ge(111). J. Appl. Phys. 2012, 111, 07D502. (45) Jaafar, R.,Nehme, Y.,Berling, D.,Bubendorff, J. L.,Mehdaoui, A.,Pirri, C.,Garreau, G. Uhlaq-Bouillet, C. Room-Temperature Ferromagnetism in Single Crystal Fe1.7Ge Thin Films of High Thermal Stability Grown on Ge(111). Appl. Phys. Lett. 2008, 93. (46) Petroff, P. M.; Feldman, L. C.; Cho, A. Y.; Williams, R. S. Properties of Aluminum Epitaxial-Growth on GaAs. J. Appl. Phys. 1981, 52, 7317−7320. (47) Tawara, Y. Magnetocrystalline Anisotropy and Magnetostriction of Intermetallic Compound Fe5Ge3. J. Phys. Soc. Jpn. 1966, 21, 237− 243. (48) Thang, P. D.; Rijnders, G.; Blank, D. H. A. Stress-Induced Magnetic Anisotropy of CoFe2O4 Thin Films Using Pulsed Laser Deposition. J. Magn. Magn. Mater. 2007, 310, 2621−2623. (49) Spiesser, A.; Olive-Mendez, S. F.; Dau, M. T.; Michez, L. A.; Watanabe, A.; Le Thanh, V.; Glachant, A.; Derrien, J.; Barski, A.; Jamet, M. Effect of Thickness on Structural and Magnetic Properties of Mn5Ge3 Films Grown on Ge(111) by Solid Phase Epitaxy. Thin Solid Films 2010, 518, S113−S117.

4348

dx.doi.org/10.1021/nl302113r | Nano Lett. 2012, 12, 4341−4348