ARTICLE IN PRESS Journal of Magnetism and Magnetic Materials 320 (2008) 2682– 2687
Contents lists available at ScienceDirect
Journal of Magnetism and Magnetic Materials journal homepage: www.elsevier.com/locate/jmmm
Magnetic anisotropy study of triangular-shaped Co nanostructures P. Imperia a,, W. Kandulski b, A. Kosiorek b, H. G!aczyn´ska b,c, H. Maletta a, M. Giersig b a b c
Hahn Meitner Institut, Glienicker Strasse 100, 14109 Berlin, Germany CESAR Research Center, Ludwig-Erhard-Allee 2, 53175 Bonn, Germany ´ , Poland Department of Macromolecular Physics, Adam Mickiewicz University, Umultowska 85, Poznan
a r t i c l e in f o
a b s t r a c t
Article history: Received 26 March 2008 Received in revised form 16 May 2008 Available online 4 June 2008
Atomic force microscopy (AFM), X-ray magnetic circular dichroism (XMCD), magnetic force microscopy (MFM) and vibrating sample magnetometry (VSM) have been used to measure the magnetic and geometrical characteristics of triangular-shaped Co structures of lateral size 730 nm and thickness 32 nm, prepared by nanosphere lithography (NSL). Evidence of in-plane six-fold magnetic anisotropy induced by the symmetry of the structure has been found. By means of XMCD measurements, performed at remanence after applying a pulsed field, a structure rotation angle-dependent oscillation of about 15% with a periodicity of 601 has been observed for both the orbital and spin moments. Furthermore, the system exhibits the angular hysteresis effect. The magnetic measurements performed by MFM show a reduction of the magnetic configurations to only two states, one quasi-single domain Y state and second, a combination of vortex and Y state. & 2008 Elsevier B.V. All rights reserved.
PACS: 75.75.+a 75.30.GW Keywords: Magnetic anisotropy XMCD Nanosphere lithography
1. Introduction The quest for new materials and technical solutions, able to match the requirements of increasing performances for data storage devices, produced in the last few years a large number of studies about the magnetic properties of nanostructured materials. The control of the magnetic properties by means of the shape could be decisive in improving the performances of newly designed devices [1–3]. For example, the associated shape anisotropy could be of primary importance to achieve a better control in view of their application in spin electronic devices and it could lead in the future to their full technological exploitation [4,5]. The ways devised to induce shape anisotropy have especially tampered with adjustments of the substrate topology [6,7] or have deployed the power of electron beam lithography [8,9]. The experimental results have also driven a wealth of theoretical studies aimed to achieve a better understanding of the nature of the magnetic anisotropy at the nanoscale under different geometrical conditions [10–12]. The study of pure ferromagnetic elements on a well-defined surface is important to better understand the magnetic properties of such complex systems considering the easy reproducibility of the preparation conditions at different laboratories and the well-known magnetic properties
Corresponding
author. Present address: ANSTO, Institute of Materials Engineering, New Illawarra Road, Lucas Heights, Sydney, NSW 2234, Australia. E-mail address:
[email protected] (P. Imperia). 0304-8853/$ - see front matter & 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.jmmm.2008.05.044
of pure ferromagnetic materials. Furthermore, for the study of shape anisotropy, the orientation of the substrate crystal with respect of the shaped particles must be well known and defined. Reports of studies of arrays of shaped particles cover several kinds of ferromagnetic pristine materials (Co, Fe and Ni) and alloys (permalloy) and few different shapes like triangles, rectangles and ellipses arranged on the surface with a simple geometry. The reports, in general, deal with the total magnetization angular dependence only. Here, we report evidence of in-plane magnetic anisotropy measured for both spin and magnetic moment found in triangularly nanostructured Co elements hexagonally arranged in a regular pattern. The simultaneous variation of the orbital and spin moment of the same quantity, relatively to their respective order of magnitude, confirms that the observed anisotropy is connected with the shape and arrangement of magnetic elements and not with their crystal structure. The measured samples have been prepared by nanosphere lithography [13,14], a relatively young technique that allows the simple and economical preparation of thin films of magnetic metals laterally structured in a number of different topologies [15].
2. Sample preparation Polystyrene (PS) latex particles of 1710 nm diameter were deposited from a water surface on the polished side of a Si wafer chemically cleaned by ultrasonication in acetone and in ultra pure (18 MO cm) water for 2 min each, with a methodology related to
ARTICLE IN PRESS P. Imperia et al. / Journal of Magnetism and Magnetic Materials 320 (2008) 2682–2687
the preparation of Langmuir–Blodget films [15]. A Co thin layer of 32 nm thickness was deposited by electron beam evaporation through the PS mask in high vacuum (HV) conditions with a base pressure during evaporation better than 2 106 mbar. After chemical removal of the nanosphere mask by ultrasonication in tetrahydrofurane (THF), the final result is a matrix of polycrystalline triangular elements arranged in a long-rangeordered hexagonal symmetry. The samples were studied by atomic force microscopy (AFM) in order to investigate their homogeneity and to determine the orientation of the evaporated triangular nanostructures with respect to the substrate (see Fig. 1 left panel). All the samples were well ordered and with a low density of defects over a large area in the range of about 100 mm2. The simplicity and effectiveness of the sample preparation constitutes the main advantage of this technique. However, the necessity to chemically remove the mask after the metal deposition ex-situ produces Co structures covered with a natural thin layer of CoO even after the shortest exposition to air. The X-ray absorption spectra (XAS) of the samples measured as introduced in the ultra high vacuum chamber having a base pressure better than 4 1010 mbar, show the typical splitting of the Co L3 edge expected when CoO is present on the surface of a Co layer. A suitable way designed to remove the CoO layer from the sample surface is H+ ion sputtering. This method has been already successfully used, for example, to remove the oxidized shell of chemically synthesized exposed Co nanoparticles [16]. The samples were therefore etched for 210 min at a relatively low sputtering energy (700 eV) with an H+ pressure of 2.4 105 mbar. The sputtering parameters were carefully adjusted to allow an effective removal of the CoO layer without destruction of the sample regular pattern. Experience tells that longer sputtering time at lower energy as well as shorter sputtering time at higher energies can partially or completely destroy the regularity of the patterns. The sputtering energy and the time necessary to remove the CoO thin layer depend on the sample thickness and the diameter of the PS latex spheres. For each combination of Co thickness and lateral dimension of triangles, the correct sputtering time must be found with a time-consuming trial-and-error procedure.
2683
The AFM pictures of Fig. 1 show a typical sample. On the left panel, there is the AFM scan just after the Co deposition. The triangular structures are well defined with a very low density of defects. The shape and dimension of the Co pattern are defined by the curvature of the PS latex spheres. The AFM scan in the right panel of Fig. 1 shows that the samples retain their well-ordered geometry after the correct H+ sputtering procedure. Small debris, having a diameter between 2 and 10 nm, are present on the Si substrate in between the metal triangular structures. The debris are the result of the cleaning action of the H+ ions on the CoO surface as well as on the exposed large area of the Si substrate. After sputtering, the thickness of the Co structures is slightly reduced by about 0.5 nm. In this case, the sample thickness measurements are affected by the simultaneous reduction of the CoO and of the SiO surface, eventually leading to an incorrect evaluation of the final Co total thickness. The native oxide layer has normally a thickness between about 1.5 and 2 nm and it stabilizes after about 16 days. H+ sputtering etches both oxides, CoO and SiO2, but with slightly different rates. The thickness estimate after sputtering is strictly valid only in case of the same etching rate for both oxides. The etching rates are very difficult to be estimated depending on the sputtering energy, gas pressure, amount of ions hitting the sample, sputtering angle and therefore depending on the specific setup used for the sputtering procedure.
3. Experimental 3.1. XMCD The magnetic properties of the samples were studied by means of X-ray magnetic circular dichroism (XMCD) at beam line UE 46 PGM at the synchrotron radiation facility BESSY, Berlin. XMCD has the advantage with respect to other techniques that it allows to extract the values of the orbital and of the spin moments separately in a relatively plain way applying the sum rules [17,18]. After sputtering in the separated preparation chamber, the samples were introduced in the measuring chamber without breaking the vacuum. The XAS measurements were taken with 90% circularly polarized light at remanent magnetization and at
Fig. 1. AFM pictures of an array of triangular nanostructured crystals Co thickness 30 nm, radius of the PS latex mask 1710 nm. Left panel: before sputtering. Right panel: after sputtering for 210 min with H+ ions (700 eV). In the inset: indication of the azimuth angle with respect to the pattern orientation.
ARTICLE IN PRESS 2684
P. Imperia et al. / Journal of Magnetism and Magnetic Materials 320 (2008) 2682–2687
room temperature. A pulsed field of 800 Oe with a pulse width of 20 ms was applied alternating its direction planar to the sample surface at each energy scan point. After each scan, the sample was rotated around its axis to perform a new energy scan at a different azimuthal angle. The arrows, imposed on the AFM picture of Fig. 1, show the geometry of the experiment. The angle j defines the direction along the triangle pattern; j ¼ 01 is defined as the direction along the spherical voids, while j ¼ 301 defines the direction along the triangle edges. The geometry of the experiment has been carefully planned to assure that the position of the center of the samples was unaltered after changing the azimuthal angle and that always the same area remains under the synchrotron light. The correct position of the sample with respect to the incident beam was carefully checked at each scan. The XAS spectra were recorded in total electron yield collecting the sample drain current. The curves were normalized toward the incident incoming light by means of the last mirror current. The measurements were performed with an incidence angle of the sample surface with respect to the incoming circularly polarized light of y ¼ 201. In the data analysis, no self-absorption effects were taken into account. The upper panel of Fig. 2 shows the typical line shape of the absorption spectra after sputtering. Three lines are drawn: the sum of the positively and negatively magnetized spectra (dash line), the step function representing the L2,3 absorption edge jump (dot line), and the spectrum resulting from the subtraction of the previous two (full line), the so-called isotropic spectrum. According to the sum rules, the number of 3d holes can be estimated from this last spectrum. During the data analysis, the orbital and spin moments, ml and ms, respectively, were calculated assuming the 3d electron occupation number for bulk hcp Co n3d ¼ 7.5 [19].
The good quality of the polycrystalline Co patterns has been confirmed by the presence of the EXAFS oscillations [20]. The small shoulders at the lower and higher energies of the main peak at the Co L3 edge, as they can be seen in Fig. 2, reveal that despite the H+ etching a small amount of oxide is still present on the sample surface. The separation of the metallic and nonmetallic contributions to the magnetic properties of the Co crystal cannot be easily achieved. The presence of a small amount of CoO on the surface of the samples has an impact on the orbital and spin moment evaluation obtained by the surface sensitive XAS technique in the soft X-ray regime with the total electron yield. It leads to a reduced remanent magnetization of the samples. The calculated values of the orbital and spin moment are lowered by an unknown factor with respect to a perfectly oxide-free surface. The lower panel of Fig. 2 shows the dichroic signal as a result of the subtraction of the positively and negatively magnetized XAS spectra. It also shows, together with the XMCD signal, its integral value. From this value and from the value of the integrated area under the isotropic spectrum, the orbital and spin moments can be separately evaluated. The spin moment ms and the orbital moment ml vs. the azimuth angle j, plotted in Figs. 3 and 4, respectively, clearly show that the patterned samples have an angular dependency with a period j ¼ 601. Both graphs show two set of data: the full black line represents the first run from j ¼ –301 to 601 and the dotted line from j ¼ –601 to 601. The ms and ml maxima were found at the angles of j ¼ –301 and 301 while the minima at j ¼ 601, 01 and 601, respectively. According to the geometry of the experimental setup, this means that the easy axis of magnetization lies along the 301 direction indicated in
norm. absorption (arb. u.)
6 5 4 3 2 1 0 1.0
0.4
XMCD (arb. u.)
0.0
0.0 -0.2
-0.5
-0.4
-1.0
Integral (arb. u.)
0.5
0.2
-0.6 -1.5
-0.8 -1.0 770
780 790 Photon Energy (eV)
800
-2.0 810
Fig. 2. Upper panel: dash line, isotropic spectrum; dot line, step function; full line, isotropic spectrum minus step function. Lower panel: dichroism spectrum of an array of triangular nanostructured crystals (30 nm) at Co L3,2 edges after 210 min H+ etching (700 eV).
Fig. 3. Spin moment vs. azimuth angle of an array of triangular nanostructured crystals (30 nm). Full line: from 301 to 601. Dot line and open circle symbols: from 601 to 601. The lines are just a guide for the eyes.
ARTICLE IN PRESS P. Imperia et al. / Journal of Magnetism and Magnetic Materials 320 (2008) 2682–2687
2685
3.2. MFM
Fig. 4. Orbital moment vs. azimuth angle of an array of triangular nanostructured crystals (30 nm). Full line: from 301 to 601. Dot line and open circle symbols: from 601 to 601. The lines are just a guide for the eyes.
the right-hand side of Fig. 1 (301+np/3, n ¼ 0, 1, 2,y), while the hard magnetization axis lies along the sphere to sphere center direction (01+np/3, n ¼ 0, 1, 2y) in the same image. The orbital and spin moment variation in both cases is approximately the same, about 15% of the signal. The maximum value calculated for the spin moment is ms ¼ 0.68 mB/atom at j ¼ 301 while the minimum value is ms ¼ 0.43 mB/atom at j ¼ 601. The maximum and minimum value for the orbital moment are ml ¼ 0.051 mB/ atom and ml ¼ 0.031 mB/atom at j ¼ 301 and 601, respectively. The orbital to spin moment ratio remains constant at all angles inside the error bars, at about ml/ms ¼ 0.080. The ml and ms absolute values as well as the ml/ms ratio calculated by the sum rules are lower with respect to calculated or measured values of orbital and spin moments expected for hcp bulk and thin films of Co [18]. Several factors play a role in such a result. It is possible that the applied pulsed field of 800 Oe leads to an incomplete magnetization of the samples measured in remanence. Static magnetization loops show that the saturation for non-H+ sputtered samples is achieved at about 2 kOe and the cycle becomes reversible at about 800 Oe. For samples having their oxidized surface removed the magnetic hysteresis cycles should be different, however, 800 Oe could be still not enough to fully saturate them. Despite the H+ sputtering procedure, the oxide could not be completely removed. The signature of a residual amount of CoO remains on the surface, as can be seen in Fig. 2, even after the best sputtering procedure. This, of course, affects the absolute calculated values of the orbital and spin moments, but it is irrelevant in the contest of this paper.
The direct magnetic force microcopy (MFM) observation and the micromagnetic calculations for the non-sputtered honeycomb pattern of triangularly shaped submicron Co particles are presented in Fig. 5a. Before measurement the sample was saturated at 3 kOe along the in-plane direction. From MFM images, one can distinguish three kinds of contrast presented in Fig. 5b. Each of those contrasts corresponds to a characteristic magnetic configuration. Their rough sketch is based on computations carried out by the object oriented micromagnetic framework (OOMMF) [21] (Fig. 5c). The first MFM blowup, Fig. 5bi, presents a quasi-single domain Y state. The strong contrast visible as black and white spots at the triangle corners results from the magnetization splitting from one towards two remaining vertexes, diagram 5ci [22]. The 5bii magnification reveals features characteristic for two states. The particular combination of dark and bright colors located in the center is typical for the vortex (V) state, while strongly contrasted regions in the corners arise from magnetization parallel to the edges like in the Y state. Therefore, this configuration is introduced as VY state. The 5biii image is also VY state but with the magnetization in all corners directed outwards the triangle. Statistically about 90% of the particles measured in three different areas of the sample exhibit VY state. The calculations of remanent state of Co triangle like single particle were performed using the free available OOMMF code. Both the geometry of the particles and the field of 3 kOe like in MFM experiment were mapped in the simulations. The cobalt material parameters were the following: magnetic saturation Ms ¼ 14 105 A/m; exchange stiffness A ¼ 30 1012 J/m; crystalline anisotropy constant K1 ¼ 2 103 J/m3; damping coefficient dc ¼ 0.05. The elementary mesh size was set to 3 nm, which corresponds to some polycrystalline properties of the real structure. The results of calculated magnetic configurations from initial random distribution of magnetic moments are presented in Fig. 6. The simulated contrast of Fig. 6a reveals VY magnetic configuration that very well corresponds to the measured MFM contrast. However, this alignment was formed in the field range from 0 to 3 kOe. After decreasing the field to zero, the remanent magnetic configuration as can be seen in Fig. 6b is a C state. This quasi-single domain configuration minimizes the exchange energy, while closed magnetization states like vortex (in which there is no external magnetic field) minimizes the magnetostatic energy. The problems with simulation of the remanent VY state are a consequence of limitations of the OOMMF code calculating the polycrystalline structure (the free available OOMMF was meant for single-crystalline materials). Moreover, the program calculates precisely only 2D spin arrays and the details of spin configuration in the vertical plane of the triangle are roughly approximated. Another factor playing a significant role is the edge roughness, which is responsible for the creation of nucleation sites allowing the presence of vortex at lower fields. With OOMMF, the roughness is automatically simulated by the inability of the cubic mesh to accurately represent triangular geometries. Thus, 3 nm mesh size due to polycrystalline structure could be not enough to simulate behavior like the corresponding experimental structure isolated particles. Furthermore, we cannot exclude that the remanent state is influenced by inter-particle interactions. The footmark of interparticle interactions seems to be visible at the MFM contrast of some triangles, where the neighboring particles have opposite colors at the corners. However, the validity of the above argument should be still verified and it could be done by performing multi-scale simulations [22]. The inter-particle
ARTICLE IN PRESS 2686
P. Imperia et al. / Journal of Magnetism and Magnetic Materials 320 (2008) 2682–2687
Fig. 5. (a) The zero field MFM image (10 10) mm of 34 nm thick cobalt particles evaporated through 1710 nm PS mask, obtained after sample saturation at 3 kOe. (b) Focus of the three types of MFM contrast: (i) the (quasi) single domain Y state, (ii) VYI state, mixture of vortex and Y state, where the magnetic flux come in by one corner and flow out by the two others, (iii) VYII state, in this case the magnetic flux come out (or come in) from all corners. (c) Schematic representations of magnetic configuration of Y state, VYI state and VYII state corresponding to the focused MFM contrasts and drew up on the basis of OOMMF simulations.
Fig. 6. OOMMF simulations of 32 nm thick triangle like particle: (a) VY configuration with corresponding contrast formed in the field range 0–3 kOe and (b) remanent state with C-state configuration.
interactions could be also confirmed by angular magnetization hysteresis behavior as proposed in Ref. [23].
4. Discussion The VY states combination observed with the MFM measurements could originate from two competing factors. On one hand, the exchange length of cobalt, lex ¼ 7 nm [24] is significantly smaller than the characteristic size of particles, thus privileging the vortex state, on the other hand the presence of elements characteristic for the Y state is favored by the triangular shape of the particles, exactly by their very tiny and sharp corners. Therefore, the occurrence of VY state is reasonable. The presence of VY states can explain the observed angular magnetic anisotropy. The calculated magnetic flux lines of the VY states are compatible with the easy axis of magnetization parallel and the hard axis perpendicular to the triangle corners. The observed anisotropy affects both the spin and the orbital moment in the same degree, this points to an effect connected with the
geometry of the structured samples and not to the crystal structure. The inter-particle interaction could also explain the differences in the curves measured while turning the azimuthal angle in one direction and in the opposite. This angular hysteresis behavior, dotted line one direction and continuous line the way back can be clearly observed in Figs. 3 and 4. Turning the angle j counter-clockwise, the minima/maxima gets slightly shifted, it could be that the magnetic configuration at remanent state plays a role that must be taken into account also discussing the observed orbital and spin moments angular hysteresis. The variation of the magnetic moment as a function of the azimuthal angle obtained by XMCD has been also confirmed independently. In this case, an angular dependence of the coercive field (Hc) has been obtained from magnetic hysteresis loops measured using vibrating sample magnetometry (VSM). The VSM measurements have been realized, using non-H+ sputtered samples with a Co layer of 34 nm. As shown in the polar plot of Hc in Fig. 7, a clear period of 601 is recognizable, confirming the results obtained by XMCD. The VSM data show no angular dependence in remanence considering the detection limits. The
ARTICLE IN PRESS P. Imperia et al. / Journal of Magnetism and Magnetic Materials 320 (2008) 2682–2687
2687
Note. The MFM observations, micromagnetic calculations and interpretations were done by H. Glaczynska. The VSM measurements have been done by U. Ebels.
Acknowledgments We thank S. Rudorff for the technical support during measurements at BESSY and D. Schmitz for useful scientific discussions and support. We also would like to thank M. Gmitra for the help during the micromagnetic simulations interpretation. W. K. would like to acknowledge the financial support from EUproject, Grant no. HPRN-CT-19999-00150 and from Deutsche Forschungs Gemeinschaft (DFG) Nr. Gi 298/5-1. References
Fig. 7. Polar plot of the coercive field of an array of triangular nanostructured crystals (44 nm thickness) measured by vibrating sample magnetometry. The sixfold symmetry is evident.
six-fold magnetic anisotropy as well as the angular hysteresis shown by Co triangular particles probably arises from the interplay between the pattern arrangement and particle particular shape.
5. Summary A six-fold symmetry have been found by means of XMCD and confirmed by VSM measurements in both the orbital and the spin moments of a patterned Co surface of interacting triangular elements prepared by nanosphere lithography. This kind of anisotropy, driven by the special pattern of the Co structures, can be attributed to configurational anisotropy effects arising from differences in energy appearing when the magnetization direction is varied with respect to the pattern orientation. MFM images supported by micromagnetic simulations show a mixture of Y and VY states for the magnetized elements of the sample in a ratio 9–1.
[1] C. Chappert, H. Bernas, J. Ferre´, V. Kottler, J.-P. Jamet, Y. Chen, E. Cambril, T. Devolder, F. Rousseaux, V. Mathet, H. Launois, Science 280 (1998) 1919. [2] R.P. Cowburn, J. Phys. D: Appl. Phys. 33 (2000) R1. [3] A. Ney, C. Pampuch, R. Koch, K.H. Ploog, Nature 425 (2003) 485. [4] R.P. Cowburn, M.E. Welland, Science 287 (2000) 1466. [5] G.A. Prinz, Science 282 (1998) 1466. [6] D.S. Chuang, C.A. Ballentine, R.C. O’Handley, Phys. Rev. B 49 (1994) 15084. [7] F. Nguyen-Van-Dau, M. Sussiau, A. Schuhl, P. Galtier, J. Appl. Phys. 81 (1997) 4482. [8] R.P. Cowburn, D.K. Koltsov, A.O. Adeyeye, M.E. Welland, D.M. Tricker, Phys. Rev. Lett. 83 (1999) 1042. [9] D.K. Koltsov, R.P. Cowburn, M.E. Welland, J. Appl. Phys. 88 (2000) 5315. [10] K. Yu. Guslienko, V. Novosad, Y. Otani, H. Shima, K. Fukamichi, Phys. Rev. B 65 (2001) 024414. [11] K. Yu. Guslienko, W. Scholz, R.W. Chantrell, V. Novosad, Phys. Rev. B 71 (2005) 144407. ˜ oz-Sandoval, J.J. Torres-Heredia, F. Lo´pez-Urı´as, J. Appl. Phys. 97 (2005) [12] E. Mun 10E318. [13] J. Rybczynski, U. Ebels, M. Giersig, Colloid. Surf. A: Physicochem. Eng. Aspects 219 (2003) 1. [14] J.C. Hulteen, R.P. Van Duyne, J. Vac. Sci. Technol. A 13 (1995) 1553. [15] A. Kosiorek, W. Kandulski, P. Chudzinski, K. Kempa, M. Giersig, Nano Letters 4 (2004) 1359. [16] P. Imperia, D. Schmitz, H. Maletta, N.S. Sobal, M. Giersig, Phys. Rev. B 72 (2005) 014448. [17] B.T. Thole, P. Carra, F. Sette, G. van der Laan, Phys. Rev. Lett. 68 (1992) 1943. [18] P. Carra, B.T. Thole, M. Altarelli, Xindong Wang, Phys. Rev. Lett. 70 (1993) 694. [19] C.T. Chen, Y.U. Idzerda, H.-J. Lin, N.V. Smith, G. Meigs, E. Chaban, G.H. Ho, E. Pellegrin, F. Sette, Phys. Rev. Lett. 75 (1995) 152. [20] P. Imperia, H. Rossner, unpublished results. [21] Object oriented micromagnetic framework (OOMMF) project at ITL/NIST. [22] D. Horvath, M. Gmitra, J. Magn. Magn. Mater. 256 (2003) 195. [23] D. Horvath, M. Gmitra, I. Vavra, J. Magn. Magn. Mater. 231 (2001) 273. [24] D.K. Koltsov, R.P. Cowburn, M.E. Welland, J. Appl. Phys. 88 (2000) 5315.
