Probing the magnetic properties of cobalt–germanium nanocable arrays

Abstract

We report the synthesis of high density arrays of coaxial nanocables, consisting of germanium nanowires surrounded by cobalt nanotube sheaths, within anodic aluminium oxide membranes. The nanocable arrays were prepared using a supercritical fluid inclusion process, whereby the cobalt nanotubes were first deposited on the pore walls of the nanoporous membranes and subsequently filled with germanium to form coaxial nanocables. The composition and structure of the metal–semiconductor nanostructures was investigated by electron microscopy, energy dispersive X-ray mapping and X-ray diffraction at high angles. The magnetic properties of the co-axial nanocables were probed using a superconducting quantum interference device (SQUID). In essence, this paper describes a technique for merging semiconductor and magnetic technologies into well-defined building blocks which may ultimately lead to new multifunctional devices, such as spin-field effect transistors.

---

Introduction

Many current integrated microelectronic circuits consist of arrays of devices constructed from semiconductor heterostructures in two dimensions.^1,2 Further miniaturization of electronic circuits, to facilitate Moore's Law,³ may necessitate the construction of appropriate devices in single dimensions.^4–8 Consequently, the development of one-dimensional (1D) coaxial nanocables, which consist of a continuous crystalline nanowire encapsulated in a sheath of a different crystalline material, has been the focus of extensive research, as these structures are expected to play an important role in the next generation of electrical heterostructured devices.^2,9–18 In particular, the introduction of magnetic/nonmagnetic layer combinations into electronic devices could allow interaction, storage, processing and transfer of information via spin rather than electric charge.^19–21

Several research groups have been investigating ferromagnetic/semiconductor epitaxially deposited thin films and layered heterostructures.^22–25 In particular, researchers at Hewlett-Packard laboratories have described ultra fast spin injection devices based on coaxial nanocables consisting of alternating layers of ferromagnetic and semiconducting materials.²⁶ These researchers have postulated that heterostructured coaxial nanocables possess an ideal structure to maximize spin injection by reducing the Schottky layer barrier between magnetic and non-magnetic layers. The opportunity to fully exploit the potential of nanocables in applications such as spintronics therefore rests in the ability to understand, control and ultimately engineer the required materials. The nanowire geometry is especially beneficial for the construction of nanodevices as strong confinement can cause advantageous changes in general properties, e.g. leading to the enhancement of magnetic moments, but also at the same time allowing the ‘wiring problem’ in real devices to be addressed.

AAO membranes that contain a highly ordered uni-directional pore structure have been widely used as hosts for various materials synthesised from electrodeposition^27–29 and incipient wetness techniques.^30,31 These methods have yielded nanocrystals within the AAO pores but with certain materials pore plugging, which prevents precursor penetration into the cavities, is a significant problem.

In our laboratories, we have developed supercritical fluid (SCF) inclusion-phase methods for producing metal and metal oxide,³² as well as semiconductor,^33–36 nanowires and nanotubes within mesoporous matrices and anodized aluminium oxide (AAO) membranes. The high-diffusivity, high precursor solubility and reduced surface tension of a SCF³⁷ results in rapid nucleation and growth of the nanowires and nanotubes within the templates, without pore plugging, resulting in filling of the porous structures. Furthermore, we have recently shown that with certain metallic precursors (Co, Fe₃O₄) that nucleation and growth of the nanowires within mesoporous templates occurs through the initial binding and coating of the pore walls with the metal atoms to form tube-like structures.³²

In this paper we describe a highly reproducible method for synthesizing high density arrays of cobalt–germanium coaxial nanocables within the channels of porous AAO using a modified SCF method. The coaxial nanocables synthesized were characterized by electron microscopy and energy dispersive X-ray analysis (EDX) analysis. The magnetic properties of the systems were characterised on a superconducting quantum interference device (SQUID). The synthesis of the coaxial nanocables was undertaken in a two-step process. Firstly cobalt was deposited on the inner walls of the AAO nanopores to form nanotubes. These tubes were subsequently filled with germanium nanowires to form arrays of coaxial nanocables. We believe that the technique described leads to well-defined nanoscale building blocks which may ultimately lead to new multifunctional devices.

Experimental

Porous anodized aluminium oxide (AAO) membranes with an average pore diameter of 150 nm were purchased from Anodisc filters, SPI supplies, West Chester, PA, USA. Porous AAO with a pore diameter between 50 and 65 nm were synthesized in-house using a well known procedure described previously.^38,39 Cobalt nanotubes within the AAO porous membranes were synthesized in a 25 ml high-pressure cell. An appropriate amount of dicobalt octacarbonyl (Co₂(CO)₈) was placed adjacent to the membranes inside an open top quartz glass boat. The cell was attached via a three-way valve, to a stainless steel reservoir (∼48 ml). A high-pressure pump (Isco Instruments, PA) was used to pump CO₂ through the reservoir in to the reaction cell. The cell was placed in a furnace and heated to 873 K and pressurized to 37.5 MPa simultaneously using a platinum resistance thermometer and temperature controller. The reaction proceeded at these conditions for 30 min. After the synthesis of Co nanotubes the AAO membranes were cooled and removed from the reaction cell. Germanium nanowires were grown within the Co nanotubes incorporated within the AAO membranes within a clean cell, by the degradation of diphenylgermane (Ph₂GeH₂) in supercritical carbon dioxide (sc-CO₂). Warning: The high pressures and temperatures used in these experiments and the volatile nature of the chemicals could potentially lead to fire or explosion. Suitable safety precautions should be taken into consideration including the use of a blast screen.

X-Ray diffraction (XRD) analysis of the coaxial nanowires was carried out using a Philips X'Pert PW3207 diffractometer equipped with Cu Kα source and an accelerator detector. Scanning electron microscopy (SEM) was carried out on JEOL 5510. Transmission electron microscopy measurements (TEM) were conducted on a JEOL 2000 FX operating at a voltage of 200 kV and on a JEOL2010F field emission high resolution TEM (courtesy of Intel Ireland). The EDX mapping was carried out using a Phillips 2000 TEM equipped with a field emission gun. The spot size used for the EDX mapping had a diameter of 1–2 nm. Prior to EDX mapping the AAO membranes were polished using diamond spray. In a two-stage process the membranes were polished with a 10 µm diamond paste and then with a 1 µm diamond paste. Sample thinning was achieved using a Gatan Precision Ion Polishing System (PIPS). To obtain images of the longitudinal aspect of the nanocables, they first had to be liberated from their alumina matrix, by dissolving the alumina in dilute potassium hydroxide solution (3 M). The liberated nanocables were washed with distilled water and filtered on a 0.02 µm Anodisc filter, SPI supplies, West Chester, PA, USA. The filter was suspended in methanol and immersed in an ultrasonic bath for 20 min. A drop of the resultant liquor was dispersed on a carbon coated, 400 mesh, copper grid for TEM analysis.

Magnetization measurements were performed using a commercial SQUID magnetometer (MPMS XL, Quantum Design) at temperatures between 1.8 and 365 K and in fields up to 10 kOe. As the AAO matrix is very brittle, the samples were mounted onto Si substrates (3 × 3 mm² size) in order to perform SQUID measurements. Magnetic field was applied along the nanowire long axis, i.e. perpendicular to the sample surface. Correction for a diamagnetic contribution from the substrate was done on all of the data. Moreover, the low temperature data were additionally corrected for a paramagnetic contribution, which appeared due to the presence of a small amount of paramagnetic impurities in the Si substrate and the AAO matrix.

Results and discussion

As-synthesized porous AAO displaying a high density of pores with an average pore diameter of 50 nm was used as the template for cobalt nanotube synthesis. By controlling the amount of the cobalt precursor in the SCF cell, nanotubes with varying wall thicknesses could be produced within the AAO templates, up to a point where 50% of the pore volume was filled (Table 1). Fig. 1 displays TEM images from cross sections of individual Co nanotubes after liberation from the AAO matrices. Figs 1a–b show Co nanotubes with wall thicknesses of 7 and 5.5 nm, formed within 50 nm diameter AAO pores and Fig. 1c displays a cobalt nanotube with a wall thickness of 3.9 nm, formed within an AAO membrane with a mean pore diameter of 65 nm. These images demonstrate that the mean thickness of the nanotube walls can be precisely tuned by accurate control of the precursor concentration and that the thickness of the tubes is uniform throughout their length. The preferential formation of metallic nanotubes within our templates under SCF condidtions is not fully understood. There is obviously a favourable interaction between the cobalt precursor, dissolved in the fluid phase, and the pore walls of the AAO at high pressures⁴⁰ but further investigation is required before any conclusions can be drawn.

Fig. 1 TEM images of Co nanotubes formed within AAO membranes (mean pore diameter of 50 nm) showing nanotube wall thicknesses of (a) 7 nm, (b) 5.5 nm and (c) 3.8 nm respectively.

Table 1 Nanotube wall thickness, as a function of cobalt precursor concentration, for an AAO with 50 nm pores

Percentage of pores filled	Precursor mass used per AAO membrane/g	Nanotube wall thickness/nm
10	0.0041	1.5
20	0.0082	2.5
30	0.0123	4.0
40	0.0164	5.5
50	0.0205	7.5

---

Fig. 2a depicts a bright field TEM image of a polished AAO membrane, with a pore diameter of 150 nm, after deposition of Co nanotubes (30 nm) and subsequent filling of these tubes with Ge to form Co–Ge coaxial nanocables. The Co nanotubes appear as dark rings within the image while the lighter region inside the cobalt nanotubes consists of germanium nanowires. EDX mapping was used to determine the elemental composition of the nanocables within the AAO template, as shown in Fig. 2b–d. Fig. 2b shows the EDX spectrum originating from the surface of the AAO pore wall (position A). The spectrum consists of peaks due to Al (1.5 keV) and small Co (0.6, 6.8, 7.9 keV) and Ge (0.7, 9.8, 11.0 keV) peaks. A peak due to Cu (1.3, 8.0, 8.9 keV) is also observed which originates from the TEM grid. The EDX spectrum of the cobalt nanotubes within the AAO membranes (position B) can be seen in Fig. 2c. There is a clearly observed increase in the relative intensity of the cobalt peak (at 6.8 keV), with respect to the copper peaks (8.0, 8.9 keV), compared to position A. There is also an increase in the intensity of the germanium peak (9.8 keV) which is due to the increased proximity of the germanium or possibly due to spreading of the beam as it penetrates the sample. Moving to position C, which corresponds to the Ge core, the relative intensity of the cobalt peak decreases (6.8 keV), compared to position B, as shown in Fig. 2d, and the Ge peak increases relative to Co. The EDX data presented is evidence for the formation of heterostructured coaxial nanocables.

Fig. 2 (a) Bright field TEM image showing Co–Ge coaxial nanocables within an 150 nm AAO membrane (mean pore diameter of 150 nm) and EDX spectra of (b) the AAO pore wall (position A), (c) the cobalt sheath (position B) and (d) the germanium core (position C).

Fig. 3a depicts an SEM image of an assembly of coaxial nanocables where the alumina matrix has been partially dissolved. The nanocables can be seen as continuous white lines separated by darker lines. The image shows complete filling of the porous matrix and the coaxial nanocables have a continuous length of > 90 µm. Fig. 3b shows a TEM image from the section of an individual Co–Ge nanocable completely liberated from its 150 nm diameter AAO template and dispersed on a carbon-coated TEM grid. The core–sheath nature of the nanocables can clearly be seen. Fig. 3c shows this discrete cobalt–germanium interface at higher magnification. The lattice planes shown in the image are separated by a d-spacing of approximately 1.7 Å corresponding to the <200> face centred cubic (FCC) structure of crystalline cobalt.³⁸ The germanium core showed strong texture containing polycrystalline grains with well-defined grain boundaries as well as some amorphous regions.

Fig. 3 Images of cobalt–germanium coaxial nanocables. (a) SEM image of nanocables partially liberated from AAO matrix and (b) close-up of the cobalt–germanium interface and (c) high resolution TEM image of the cobalt–germanium coaxial interface.

The crystallinity of the nanocables was investigated using XRD at high angles as shown in Fig. 4. The peaks shown can be indexed to germanium (diamond structure, JCPDS card no. 03-0486) and cobalt (FCC structure, JCPDS card no. 10-0425) from the nanocables and peaks originating from the alumina matrix (JCPDS card no. 02-1420; 02-1421).⁴¹ There was only one cobalt peak visible, which corresponds to the (111) reflection at 46–47° 2θ, which is the most intense diffraction peak for this phase. The absence of other peaks, which should be intense enough to easily exceed the noise level of the data, suggests that cobalt is strongly orientated with respect to the sample. We note that γ-Al₂O₃ has a (111) d-spacing of 0.1977 nm (JCPDS card no. 10-0425) compared to the cobalt (111) d-spacing of 0.2045 nm (JCPDS card no. 01-1296); a close enough agreement to allow low strain epitaxial growth. The FCC structure of cobalt is consistent with the supercritical conditions of 837 K and 37.5 MPa, employed for the deposition of the metal.³² All of the expected reflections from the germanium diamond phase were observed, completely consistent with the polycrystalline structure observed in the TEM and SEM images. No trace of any Co–Ge alloys was found by XRD analysis. On the basis of these arguments we suggest that cobalt nucleates in an epitaxial form at the pore walls leading to strong orientation effects. However, with no simple epitaxial relationship for cobalt–germanium a random polycrystalline arrangement is favoured.

Fig. 4 High angle XRD pattern of Co–Ge nanocables within an AAO matrix (mean pore diameter of 150 nm).

Fig. 5 demonstrates the field dependence of the magnetic moment for an array of hollow Co nanotubes with an external diameter of 50 nm and wall thickness of about 3 nm, which corresponds to about 10% of the pore volume. The mvs. H dependencies were measured at temperatures of 1.8 and 300 K. In both cases the curves are nearly linear with a change of the slope at low fields. A very small hysteretic effects characterized by coercivity, H_C, in the order of 20–30 Oe were detected when the field decreased below 300 and 700 Oe as measured at 1.8 and 300 K, respectively. No tendency to saturation was observed up to the highest applied field. A linear paramagnetic contribution can be subtracted from the experimental data. Consequently, saturating hysteresis loops which barely depend on the temperature can be obtained. Such behaviour might be interpreted as a co-existence of both ferromagnetic and paramagnetic structural phases within the Co nanotubes over the whole temperature interval, where the later phase is strongly dominant. The temperature dependence of magnetization was measured during field cooling at H = 2 kOe. By subtracting the temperature independent magnetic moment of the ferromagnetic phase at saturation, m = 1.5 × 10⁻⁵ emu, the temperature dependence of a pure paramagnetic contribution can be obtained. The paramagnetic component of magnetization is inversely proportional to the temperature (see inset in Fig. 5) in good agreement with the Curie law, χ = C/T, for paramagnetic materials. Suppression of the ferromagnetic ordering, which would normally be observed in Co planar structures of the same thickness, is probably due to the formation of cobalt oxide (or other contaminated) layers both on the interface with AAO matrix and free cobalt surface.

Fig. 5 Magnetic moment versus field for hollow Co nanotubes (external diameter is 50 nm, wall thickness is 3 nm) at different temperatures as indicated in the figure. The inset shows the linear dependence of the magnetic moment on the inverse temperature at B = 2 kOe (○ experimental data for the whole system; Δ paramagnetic component of magnetization).

When the cobalt nanotubes were filled with germanium forming Co–Ge coaxial nanocables (total diameter is ≈50 nm, diameter of Ge core is ≈47 nm), their magnetic properties changed significantly. The M(H) curves clearly demonstrate the hysteretic effect (Fig. 6), which weakly depends on temperature. In their as-prepared state Co–Ge nanocables are characterized by coercivity, H_C = 45–62 Oe, and well-pronounced saturation, H_s ≈ 8–10 kOe. The magnetic moment is nearly temperature independent in the range between 1.8 to 260 K, as measured during field cooling at H = 2 kOe. As the temperature increases further, the moment was gradually reduced. In the measured temperature interval (up to 365 K) the total moment decreased by 12% compared with its maximum value at 1.8 K. Both field and temperature behaviour of the magnetic moment in the Co–Ge nanocables can therefore be attributed to a ferromagnetic nature of the effects at least at room temperature and below. High temperature measurements are necessary for exact determination of the Curie temperature, T_c. However, a lower limit of T_c (a point where the temperature starts to shift from its constant value) can be defined from our experiment as 260 K. Bulk cobalt yields a significantly higher T_c = 1388 K, therefore, the effects observed here cannot be attributed entirely to the Co part of the system. Although pure germanium does not possess a net magnetic moment, resulting in diamagnetism with a molar susceptibility of −11.6 × 10⁻⁶ cm³ mol⁻¹, the explanation for the observed effects is more complex than a simple superposition of ferromagnetic properties of Co and diamagnetic ones of Ge. Although the XRD analysis does not reveal any localized Co–Ge alloys in the investigated system and the TEM images (Fig. 3) demonstrate a sharp Co–Ge interface, the formation of a narrow (of the order of a few atomic layers) interdiffused layer at the interface can not be entirely excluded. Such a region forms a thin magnetically inactive layer (with thickness up to about six monolayers)⁴² which should unavoidably reduce the total moment of the system. Experimentally we observed that the saturated magnetic moment of the Co–Ge nanocables is indeed smaller than expected for pure cobalt. However, it is difficult to quantify the amount of Co–Ge alloy from the magnetization experiment. Other factors, such as the formation of a thin cobalt oxide layer on the Co–AAO interface, may as well lead to a reduction of the total moment. On the other hand, the fine effect of antiferromagnetic coupling observed earlier in Co–Ge superlattices⁴³ can probably be excluded as the Ge core is relatively thick and no exchange biasing effect was observed in the system even at low temperatures.

Fig. 6 The normalized magnetization versus magnetic field for as prepared and annealed Co–Ge nanocables at T = 1.8 K. The top inset is a low-field magnification of the hysteresis curves. The bottom inset shows the temperature dependence of coercivity for hollow Co nanotubes (▲), as prepared (○) and annealed (♦) Co–Ge nanocables.

Generally, ultrathin magnetic films of materials such as Co and Fe are characterized by an alteration of magnetic anisotropy, exchange coupling and electronic band structure, as compared with corresponding bulk values due to the broken perpendicular symmetry. In particular, the significant enhancement of the orbital and total magnetic moments, as well as moment reorientation, has been observed in ultra-thin Co films and multilayers and is strongly dependent on the overlayer material.^42,44–46 The latter fact is usually explained in terms of surface/interface magnetic anisotropy, K_s. For cobalt, the surface anisotropy normally changes the sign, when a free cobalt layer is covered by another material,⁴⁷ which causes reorientation of the moments. Thus, embedding of the Ge core into the Co nanotube has a double effect. Firstly, it serves as a protective layer to prevent a surface oxidation or contamination and, secondly, it changes the surface anisotropy and hence, the total anisotropy balance in the system. It may also lead to enhancement of the total Co moment, as it was shown for Co–Ge planar structures.³⁹ However, a detailed knowledge of Co and Ge electronic band structures is required. On the other hand, migration of individual Co atoms into the Ge core during the nanocable growth is an alternative possible origin of ferromagnetism leading to the local magnetization of the germanium matrix and effects similar to those in dilute magnetic semiconductors (DMS) will appear.

In order to clarify the source of ferromagnetism in the Ge–Co nanocables, the samples were further annealed for 4 h at 750 °C in an N₂ environment, which should facilitate the interdiffusion of cobalt and germanium atoms. Fig. 6 shows the influence of annealing on the demagnetisation process in Co–Ge nanocables as measured at 1.8 K. Similar effects were observed at room temperature as well. The main results of annealing can be seen in an increase of the magnetization saturation value (about 30%), general broadening of the hysteresis loops (see the top inset in Fig. 6) and reducing of the saturation field, H_s = 5–7 kOe. Annealing leads to a relative remanence increase: ΔM_r/M_s = 80 and 100% as measured at 300 and 1.8 K respectively. The inset at the bottom of Fig. 6 summarizes the temperature behaviour of coercivity for all investigated samples. The hollow Co nanotubes possess the smallest H_C. The coercivity is on the whole larger in as prepared Co–Ge nanocables, H_C = 45–62 Oe, and weakly decreases as temperature increases. Annealing of the Co–Ge sample leads to a significant increase of its coercivity, H_C = 100–155 Oe. The thermal treatment affects also the temperature dependence of the magnetic moment. While the total drop of magnetization after annealing remains the same, ≈12%, in the whole investigated temperature interval, the shape of Mvs. T curve is changed in the intermediate range, T = 150–275 K. The fact demonstrates a decrease of T_c and further broadening of ferro-/paramagnetic phase transition. Thus, the annealing experiment possibly demonstrates further penetration of individual Co atoms from the interdiffusion layer into the Ge core, which leads to a more homogeneous distribution of ferromagnetic atoms within the nanocables and, therefore, may cause an additional magnetization of the semiconductor core. However, currently we have no further evidence to support this conclusion. Hence, in the annealed sample two different mechanisms of ferromagnetism may coexist—original ferromagnetism of the transition metal (in the cobalt sheath) and the diluted magnetic semiconductor type of ferromagnetism due to the migration of the cobalt atoms into the germanium core. The shift of the Curie temperature further away of T_c(Co) is also consistent with this model, as diluted magnetic semiconductors have generally a lower T_c than pure transition metals.

Conclusions

We have shown that porous AAO templates can be used for the synthesis of tunable cobalt nanotubes and the subsequent production of high-density arrays of nanotubes of Co–Ge coaxial nanocables up to 90 µm in length. The coaxial nanocables consist of two discrete layers of a semiconducting core surrounded by a ferromagnetic sheath. From TEM and PXRD evidence it appears that the cobalt layer exists as highly orientated single crystals, while the germanium core favours a random polycrystalline arrangement. From our magnetic analysis we have shown that the nanocables are ferromagnetic with near room temperature T_c. The ferromagnetism does not result from the cobalt layer alone but is intrinsic to the Co–Ge heterostructure.

Acknowledgements

We would like to acknowledge the financial support from Intel (Ireland), the Irish Research Council for Science and Engineering (IRCSET), Science Foundation Ireland, the Latvian Council of Science and NMS Quantum Metrology Program, project QMP04.3.4. We would also like to thank Prof. Melihov from Tallin Technical University for access to SEM equipment and Dr Juan Perez Camacho for the HRTEM image.

References

1. C. Weisbuch and B. Vinter, Quantum Semiconductor Structures, Academic Press, Boston, 1991.
2. Q. Li and C. R. Wang, J. Am. Chem. Soc., 2003, 125, 9892.
3. G. E. Moore, Electronics, 1965, 38, 84.
4. J. T. Hu, T. W. Odom and C. M. Lieber, Acc. Chem. Res., 1999, 32, 435.
5. S. W. Chung, J. Y. Yu and J. R. Heath, Appl. Phys. Lett., 2000, 76, 2068.
6. J. Kong, H. T. Soh, A. M. Cassell, C. F. Quate and H. J. Dai, Nature, 1998, 395, 878.
7. Y. Feldman, E. Wasserman, D. J. Srolovitz and R. Tenne, Science, 1995, 267, 222.
8. S. Iijima, Nature, 1991, 354, 56.
9. J. Q. Hu, Y. Bando, Z. W. Liu, T. Sekiguchi, D. Golberg and J. H. Zhan, J. Am. Chem. Soc., 2003, 125, 11306.
10. X. M. Meng, J. Q. Hu, Y. Jiang, C. S. Lee and S. T. Lee, Appl. Phys. Lett., 2003, 83, 2241.
11. A. Kolmakov, Y. X. Zhang and M. Moskovits, Nano Lett., 2003, 3, 1125.
12. D. F. Liu, S. S. Xie, X. Q. Yan, L. J. Ci, F. Shen, J. X. Wang, Z. P. Zhou, H. J. Yuan, Y. Gao, L. Song, L. F. Liu, W. Y. Zhou and G. Wang, Chem. Phys. Lett., 2003, 375, 269.
13. Y. C. Zhu, Y. Bando and Y. Uemura, Chem. Commun., 2003, 836.
14. Q. Li and C. R. Wang, Appl. Phys. Lett., 2003, 82, 1398.
15. J. Q. Hu, X. M. Meng, Y. Jiang, C. S. Lee and S. T. Lee, Adv. Mater., 2003, 15, 70.
16. J. Q. Hu, Q. Li, X. M. Meng, C. S. Lee and S. T. Lee, Chem. Mater., 2003, 15, 305.
17. Y. D. Yin, Y. Lu, Y. G. Sun and Y. N. Xia, Nano Lett., 2002, 2, 427.
18. X. C. Wu, W. H. Song, B. Zhao, W. D. Huang, M. H. Pu, Y. P. Sun and J. J. Du, Solid State Commun., 2000, 115, 683.
19. M. N. Baibich, J. M. Broto, A. Fert, F. N. Vandau, F. Petroff, P. Eitenne, G. Creuzet, A. Friederich and J. Chazelas, Phys. Rev. Lett., 1988, 61, 2472.
20. J. Barnas, A. Fuss, R. E. Camley, P. Grunberg and W. Zinn, Phys. Rev. B, 1990, 42, 8110.
21. S. S. P. Parkin and D. Mauri, Phys. Rev. B, 1991, 44, 7131.
22. Y. Ohno, D. K. Young, B. Beschoten, F. Matsukura, H. Ohno and D. D. Awschalom, Nature, 1999, 402, 790.
23. J. De Boeck, W. Van Roy, V. Motsnyi, Z. Liu, K. Dessein and G. Borghs, Thin Solid Films, 2002, 412, 3.
24. F. Mireles and G. Kirczenow, Phys. Rev. B, 2002, 66, 214415.
25. J. P. Hong, S. B. Lee, Y. W. Jung, J. H. Lee, K. S. Yoon, K. W. Kim, C. O. Kim, C. H. Lee and M. H. Jung, Appl. Phys. Lett., 2003, 83, 1590.
26. V. V. Osipov and A. M. Bratkovsky, J. Phys. Condens. Matter, 2004, 2, 307656.
27. X. Y. Zhang, L. D. Zhang, W. Chen, G. W. Meng, M. J. Zheng and L. X. Zhao, Chem. Mater., 2001, 13, 2511.
28. C. Yoon and J. S. Suh, Bull. Korean Chem. Soc., 2002, 23, 1519.
29. D. Xu, X. Shi, G. Guo, L. Gui and Y. Tang, J. Phys. Chem. B., 2000, 104, 5061.
30. Y. Li, D. Xu, Q. Zhang, D. Chen, F. Huang, Y. Xu, G. Guo and Z. Gu, Chem. Mater., 1999, 11, 3433.
31. T. Peng, H. Yang, K. Dai, X. Pu and K. Hirao, Chem. Phys. Lett., 2003, 379, 432.
32. T. A. Crowley, K. J. Ziegler, D. M. Lyons, D. Erts, H. Olin, M. A. Morris and J. D. Holmes, Chem. Mater., 2003, 15, 3518.
33. K. M. Ryan, D. Erts, H. Olin, M. A. Morris and J. D. Holmes, J. Am. Chem. Soc., 2003, 125, 6284.
34. D. M. Lyons, K. M. Ryan, M. A. Morris and J. D. Holmes, Nano Lett., 2002, 2, 811.
35. N. R. B. Coleman, M. A. Morris, T. R. Spalding and J. D. Holmes, J. Am. Chem. Soc., 2001, 123, 187.
36. N. R. B. Coleman, M. A. Morris, T. R. Spalding and J. D. Holmes, J. Am. Chem. Soc., 2001, 123, 7010.
37. T. Clifford, Fundamentals of Supercritical Fluids, Oxford University Press, Oxford, 1999.
38. H. Masuda and K. Fukuda, Science, 1995, 268, 1466.
39. H. Masuda, H. Yamada, M. Satoh, H. Asoh and M. Nakao, Appl. Phys. Lett., 1997, 71, 2770.
40. S. Sarrade, G. M. Rios and M. Carles, J. Membr. Sci., 1996, 114, 81.
41. Peaks indexed using X'Pert Highscore software (PW3209), 2002, referenced to PCPDFWIN database JCPDS-ICDD, 2001.
42. P. Ryan, R. P. Winarski, D. J. Keavney, J. W. Freeland, R. A. Rosenberg, S. Park and C. M. Falco, Phys. Rev. B., 2004, 69, 54416.
43. Y. Endo, N. Kikuchi, O. Kitakami and Y. Shimada, J. Phys. Condens. Matter, 1999, 11, L133.
44. O. Kazakova, M. Hanson and A. Yu, J. Appl. Phys., 2004, 96, 9612.
45. S. Park, S. Lee and C. M. Falco, J. Appl. Phys., 2002, 91, 8141.
46. P. Beauvillian, A. Bounouh, C. Chappert, R. Mégy, S. Ould-Mahfoud, J. P. Renard, P. Veillet, D. Weller and J. Corno, J. Appl. Phys., 1994, 76, 6078.
47. F. C. Chen, Y. E. Wu, C. W. Su and C. S. Shern, Phys. Rev. B., 2002, 66, 184417.

---