Magnetization reversal of CrO 2 nanomagnet arrays

Transcription

1 JOURNAL OF APPLIED PHYSICS VOLUME 96, NUMBER DECEMBER 2004 Magnetization reversal of CrO 2 nanomagnet arrays Qiang Zhang, Y. Li, and A. V. Nurmikko Division of Engineering and Department of Physics, Brown University, Providence, Rhode Island G. X. Miao and G. Xiao Department of Physics, Brown University, Providence, Rhode Island A. Gupta IBM TJ Watson Research Center, Yorktown Heights, New York (Received 4 June 2004; accepted 16 September 2004) We report on fabrication techniques and the study of magnetic behavior of submicron-sized ferromagnetic CrO 2 arrays. Periodic rectangular arrays composed of circular dots with diameters ranging from 100 nm to 2 m were fabricated by electron-beam lithography and reactive ion etching of CrO 2 epitaxial thin films. The magnetization reversal of the nanomagnet arrays was measured by the magneto-optical Kerr effect, with hysteresis characteristics found to be sensitively dependent on array details, film thickness, and its built-in strain. A rich variety of behaviors such as single domain, probable multidomain, as well as magnetization reorientation transition were observed, with magnetostatic interaction between the neighboring dots affecting the collective behavior of the nanomagnet arrays American Institute of Physics. [DOI: / ] I. INTRODUCTION The half metallic ferromagnet CrO 2 is of considerable interest because of its high spin polarization 1 that may find possible applications in spin-based devices. High quality single crystal CrO 2 thin films can now be prepared on single crystal (rutile) TiO 2 substrates using chemical vapor deposition, 2 which demonstrate coherent magnetization rotation by single domain behavior in macroscopic samples, both in static 3 and ultrafast pulsed optical experiments. 4 Given both fundamental questions as well as applied potential about magnetic materials on a submicron scale (referred here as nanomagnets ), we have undertaken an initial study of the magnetic characteristics about patterned nanomagnet arrays of CrO 2. We note that recent work, especially on submicron patterned high-density arrays of permalloy films, has divulged interesting magnetization reversal pathways, including those via vortex states. 5 In general, nanostructured magnetic materials may exhibit quite different properties compared to their bulk or film counterparts, due to the small length scale, lower dimensionality and interacting features. Here, we demonstrate a particular approach to fabrication of CrO 2 nanomagnet arrays, with circular dots arranged in rectangular arrays of varying pitch with diameters ranging from 100 nm to 2 m, etched out from CrO 2 films of thicknesses ranging from 25 to 150 nm. The hysteresis loops of the arrays were measured by a high sensitivity transverse magnetooptic Kerr (TMOKE) technique. In these initial studies we have found that the details of the collective magnetization reversal behavior of the nanomagnet arrays were crucially dependent on several factors: the individual particle diameter, the CrO 2 film thickness, the details of the film preparation method, as well as the magnetostatic (dipole-dipole) intercoupling between the individual nanomagnets. The TMOKE measurements suggest a wide range of magnetization reversal pathways, ranging from participation by singledomain-like state, by probable multidomain states, and by magnetization reorientation transition. The main goal of this paper is to provide a phenomenological picture of how magnetic response of arrays of CrO 2 nanomagnets can show many facets of behavior with only moderate changes in their geometry and different parameters of size. The fabrication of magnetic nanostructures from single crystal CrO 2 films by high resolution lithography presents specific challenges in the development of a suitable process strategy. One issue is the metastable nature of the thin film material associated with the high strain originating from the finite (3.91% along b axis and 1.44% along c axis) lattice mismatch with the (100) TiO 2 substrate. 6 Another requirement is the need for a suitable low-damage chemical process in conjunction with etching. We have developed a particular chemical ion dry etching technique to enable the fabrication of patterned CrO 2 structures, with individual feature size resolution below 100 nm. II. SAMPLE FABRICATION The single crystal CrO 2 films were epitaxially grown on (100) single crystal TiO 2 substrate by chemical vapor deposition method. 2 The desired array patterns, typically m 2 in total area, were created by high resolution electron-beam lithographic technique, with the process steps summarized schematically in Figs. 1(a) 1(e). First, a double layer of positive polymethylmethacrylate (PMMA), with molecular weights of 495 and 950 kd, respectively, were spin coated onto the sample. After a bake at 180 C for 90 s for each layer, the sample was exposed in a 30 kv electronbeam lithography system based on a LEO 1530VP scanning electron microscope. The sample was developed in a 1:3 solution of methyl isobutyl ketone/isopropyl alcohol for 65 s and then evaporated with a thin layer of aluminum film of 30 nm, followed by liftoff process. After liftoff, the residue /2004/96(12)/7527/5/$ American Institute of Physics

2 7528 J. Appl. Phys., Vol. 96, No. 12, 15 December 2004 Zhang et al. FIG. 1. Upper trace: Schematic of fabrication processes (a) spin-coating the sample with double layer of PMMA, (b) expose the sample with e-beam lithography, after developing, a thin layer of Al was evaporated onto the sample, (c) liftoff, (d) reactive ion etching of CrO 2 film using Al as the etch mask, (e) remove the Al mask by wet etching. Lower trace: SEM images of fabricated CrO 2 nanomagnet array, (f) tilted view of dot array with d =500 nm and h=200 nm, and (g) top view of square lattice dot array with d=250 nm, h=150 nm, and s=30 nm. aluminum pattern over the exposed area was used as the etch mask for reactive ion etching (RIE) of the underneath CrO 2 film. A recipe of mixture of bromotrifluoromethane CBrF 3 and oxygen with a ratio of 1:1 was used for RIE etching. The introduction of bromium into the reactive species was important so as to stabilize the etch process: it greatly reduced unwanted lateral chemical attack to material under the etch mask because Br is known for its ability to passivate the CrO 2 surface. 7 Finally, the aluminum mask was removed by wet etching in phosphoric acid. Prior to this work, the typical approaches of patterning CrO 2 film have employed wet etching, selective growth on prepatterned substrate, 8 and ionbeam etching on a micrometer scale or beyond. Compared to these methods, our RIE etching is compatible with the goal of fabricating nanoscale feature sizes, while producing minimal damage to the film (typically by thermal and ionbombardment effects in a dry etch situation). Examples of the fabricated CrO 2 nanomagnet arrays are shown in the lower panel of Fig. 1, as scanning electron microscope (SEM) images. Figure 1(f) displays a portion of an array with individual nanomagnet diameter d = 500 nm, etched from a 200 nm thick film. As can be seen, both etched and unetched surfaces remain clean and smooth after etching. The etched-wall tilting angle is typically 20 off the film normal direction. An example of a close-packed square array is shown in Fig. 1(g), with a periodicity of 280 nm and composed of individual 150 nm tall nanomagnets of diameter d =250 nm. The SEM image shows that the dots remain distinct and well separated from each other, even though their edge-to-edge spacing is only about 30 nm. The vertical wall and edge of each dot is quite well defined. III. CHARACTERIZATION AND DISCUSSION The magnetic characteristics of the fabricated arrays were measured at room temperature by acquiring hysteresis loops via the magneto-optic Kerr effect in the transverse geometry. A semiconductor laser diode operating at 850 nm was used as the light source. The laser beam was focused to FIG. 2. Easy-axis hysteresis characteristics of noninteracting nanomagnet array on h=50 nm as-grown film measured by transverse magneto-optical Kerr effect at room temperature, the diameters of the dots are (a) d =1000 nm, (b) d=400 nm, (c) d=200 nm, (d) d=100 nm. a spot size of 30 m on the sample, i.e., well within the total array area. Only the reflected laser beam was collected and detected. A balanced dual-photodiode detector scheme was employed to minimize the influence of noise from the laser itself. Instead of using a conventional lock-in technique by modulating the light intensity, the strength of the magnetic field was periodically varied (typically between ±800 Oe at the frequency of a few hertz) and carefully monitored by a field sensor. The magnetic field sensor signal, together with the differential output from the photodetector, was inputted into a dual-channel digital oscilloscope for signal averaging. The consequence is that the source laser noise and scattered light noises induced by mechanical vibration, even the Fourier components in the very low frequency range, can be greatly suppressed. This approach has earlier proved to be very useful for magneto-optic measurement of the patterned magnetic samples 9 and was of particular use here in those instances where the CrO 2 material coverage was low on the wafer after etching. We next show selected examples of data to emphasize the widely varying magnetic behavior of the fabricated arrays. The hysteresis loops of the nanomagnet arrays fabricated out of an h=50 nm thick as-grown CrO 2 film are shown in Fig. 2, for various individual element diameters d (h/d ratios). The lattice constant l in this case was large enough (l=2 m when d=1 m; l=1 m when d =400 nm; l=3d when d=100 nm or 200 nm) so that the interparticle dipolar coupling between elements is negligible. Given the spot size of the optical probe, these measured hysteresis loops can thus be regarded as an ensemble average of individual dot behavior. The dot diameters shown here are d=1 m, 400, 200, and 100 nm, respectively. The axes of symmetry for the array were directed along the easy and hard axis of magnetization of the as-grown film. The loops were measured with B field along the easy axis of the sample. For d larger than 100 nm, the loop shape suggests that magnetization reversal by formation of a multidomain structure is being observed. Due to the strong uniaxial magnetocrystalline anisotropy of CrO 2 film [strain dependent, but in the approximate range of K eff erg/cm 3 (Ref. 10)], it would seem that the alternative possibility of

3 J. Appl. Phys., Vol. 96, No. 12, 15 December 2004 Zhang et al FIG. 4. Etch-down film vs as-grown film: easy-axis hysteresis characteristics of noninteracting nanomagnet array fabricated on etch-down film with film thickness h=40 nm, the diameters of the dot are d=100, 400, 800 nm, and 2 m. FIG. 3. Influence of film thickness on easy-axis hysteresis characteristics: noninteracting nanomagnet array with film thickness h=25, 36, and 50 nm, the diameters of the dots are d=100, 200, 400, and 800 nm. formation of vortex states, in analogy to earlier measurements on permalloy arrays, 5 is not favored in general for CrO 2. As the particle diameter is reduced (increasing the h/d ratio), both the coercive and saturation field increased notably, by nearly a factor of 3 compared to the case of d =1 m. The dependence of saturation field on the dot sizes can be understood from the magnetostatic energy aspects. At small dot diameters, the magnetostatic energy would predominate over the magnetocrystalline anisotropic energy so that the saturation field will increase with demagnetizing field at saturation. However, when d was reduced to 100 nm, the shape of the hysteresis loop changed dramatically. The coercivity for the nanomagnets seems virtually disappeared whereas the saturation field remains unchanged. The approximately linear response of M-H curve then suggests that the easy axis of magnetization has switched to the film normal direction, or a magnetization reorientation transition (MRT) has happened. We also investigated the influence of film thickness on the magnetic behavior of the nanomagnets, and a representative result is shown in the Fig. 3. The hysteresis loops of arrays with dots with d=100, 200, 400, and 800 nm and film thickness of 50, 36, and 25 nm are plotted. As observed, the film thickness has a significant impact on the shape of the hysteresis loop, especially in case of the small diameter dots. Thinner film, with reduced height/ diameter ratio, appeared to suppress the tendency of magnetization reorientation transition and the squareness of the loop was accordingly enhanced. Theoretical nanomagnetic calculation 11 has predicted that the critical condition for such MRT to happen occurs at a particle height-to-diameter ratio h/ d = This prediction was obtained for the cylindrical dot case where the magnetocrystalline anisotropy was neglected. For our d =100 nm dot shown above, the critical value of h/d ratio was 0.5. To consider this discrepancy, we first note that the shape of our nanomagnet is not a perfect cylinder, but rather more like a truncated cone. Secondly, it is known the strain originating from the lattice mismatch between CrO 2 and TiO 2 (Ref. 10) has a strong influence on the magnetoanisotropy of CrO 2 films thinner than 100 nm. To examine the influence of the (inhomogeneous) strain distribution in the context of our nanomagnets, we also fabricated patterned arrays on CrO 2 films whose thickness was adjusted to the target thickness by RIE etching, prior to lithography and patterning. Here we started with relatively thick as-grown films, typically about 200 nm. In Fig. 4, the hysteresis loops of nanomagnets with various dot diameters are shown, fabricated from a 40 nm etch-down film. These nanomagnets distinguished themselves from the previously shown patterned structures on as-grown films (Figs. 2 and 3), by showing a magnetization reversal much closer to single domain behavior, even though the sharpness of the switching might be compromised due to the fluctuation of the dot sizes and shapes in our ensemble averaging TMOKE measurement. Other than the shape of the loop, the coercivity dependence on the dot size also gave indication of single domain state. The coercivity remained almost constant throughout the dot size range, which is not consistent with the typical characteristics of nucleation/propagation type of reversal mechanism. 12 We surmise that the contrast is likely due to the strain-releasing effect during the etching process. In fact, it has been recently demonstrated that the strain in CrO 2 thin film induced by the lattice mismatch between CrO 2 and TiO 2 can be at least partially released in a wet etching process, 13 resulting in an enhancement of the in-plane uniaxial magnetocrystalline anisotropy. Finally, we have also investigated the collective aspect and behavior of high density, interacting nanomagnet dot arrays, in the proximity regime where dipolar coupling effects are anticipated to play a role. In Fig. 5, d=250 nm diameter dots based on 150 nm thick film was arranged in square lattices with various spacing distances, hence a tunable coupling strength. One axis of the square lattice was along the (unpatterned) sample s easy axis of magnetization and the M-H response was measured by the TMOKE technique in the same direction. The interpretation of the observed behavior in Fig. 5 is as follows: Magnetization reversal via multidomain state appear to be observed most commonly. Similar hysteresis shape has been observed in Fe nanomagnets on

4 7530 J. Appl. Phys., Vol. 96, No. 12, 15 December 2004 Zhang et al. FIG. 5. Effects of interaction among the nanomagnets on the easy-axis hysteresis characteristics: the arrays arranged in square lattices were fabricated on as-grown film with film thickness h=150 nm, the spacing distance between the dots was varying from 230 nm down to 30 nm. The SEM pictures of the corresponding array are shown on the left. GaAs (001) (Ref. 14) as well as in patterned permalloy dots. 15 As the edge-to-edge spacing between the dots was reduced from s=230 nm to about s=30 nm, the hysteresis loop experienced dramatic changes due to the dot-to-dot coupling. Both the coercivity and remanent magnetization have increased, while the saturation field was reduced. In the case of smallest spacing array, the strength of the coupling a dot experienced from its surrounding neighbors can be evaluated by the reduction of saturation field from widely separated dot array case with negligible interaction, which was about 355 Oe. We employed high resolution SEM to carefully examine the highest density array s=30 nm to ensure that the fabrication process did not leave any residue material between pairs of the dots which might form a conducting bridge and create an additional means of direct (exchange) ferromagnetic coupling. Hence the argument that the interparticle coupling must be of magnetostatic origin, with nearest neighbor interaction as the dominant term for the square lattice. Magnetostatic dipolar coupling can, of course, be either parallel (i.e., ferromagnetic) or antiparallel (antiferromagnetic), depending on the relative geometry of the separation of the interacting particles to their magnetic moments. When the dot-to-dot separation is parallel/normal to the dot moment, the coupling is parallel/antiparallel. We show examples from the CrO 2 arrays in Fig. 6 acquired through the TMOKE measurements. In Fig. 6(a), d=200 nm size nanomagnet elements (film thickness h =25 nm) were arranged in rectangular, as opposed to earlier square arrays. The coupling between the dots along the long axis (sample hard axis) was designed to be negligible. The short axis (sample easy axis) FIG. 6. Experimental evidence for parallel coupling and antiparallel coupling: easy-axis hysteresis characteristics of interacting nanomagnet array are shown (a) parallel coupling: nanomagnet d=200 nm, h=25 nm arranged in rectangular lattice, the short axis is along the sample easy axis and spacing of the dots are varied from s=120 nm to s=30 nm, the dot spacing along the long axis was large enough so that the interaction in this direction are negligible (b) antiparallel coupling: a rectangular lattice array d =230 nm, h=40 nm with short axis along the easy-axis direction is compared to a square lattice array with identical dot diameters and dot-to-dot spacing s=40 nm. dot spacing was changed to systematically vary the coupling strength. As the edge-to-edge nearest neighbor spacing was reduced from s = 120 nm to s = 30 nm, the coercivity and squareness of the loop increased monotonically. Semiquantitatively at least, this measurement yields a value for the parallel coupling strength of about 40 Oe. The contrasting example of antiparallel coupling is shown in Fig. 6(b). In this case, the individual nanomagnets were 230 nm in diameter and 40 nm tall, with the short axis of the rectangular array was still along the sample easy axis. The easy axis response was compared with the corresponding square lattice array of the same dot-to-dot spacing s=40 nm. We interpret the clear difference between the coercivity of these two arrays as being due to the antiparallel coupling present in the square lattice array, which can thus be estimated to be about 22 Oe. IV. CONCLUSION In summary, submicron nanomagnet arrays based on single crystal epitaxial CrO 2 films were fabricated by electron-beam lithography and reactive ion etching. A wide range of magnetic behavior was seen through magnetooptical measurements, with details depending on the specific array size parameters. In dilute patterned arrays fabricated on as-grown films, multidomain state and magnetization reorientation transition appear to be important and aided by the built-in lattice mismatch strain. On the other hand, manipulation of the strain by etch-thinning the films prior to patterning shows behavior close to single domain with similar film

5 J. Appl. Phys., Vol. 96, No. 12, 15 December 2004 Zhang et al thickness. In high-density arrays, dipolar interactions among the nanomagnets were also observed, including both parallel coupling and antiparallel coupling depending on the lattice geometry. With CrO 2 as the particular test material system studied here, we have shown how dimensions, film thickness, and interaction are all playing an important role to determine the magnetic behavior of the nanomagnet arrays. While phenomenological arguments have been given in the above, it is clear that considerably more detailed and systematic future studies are required, to appreciate the opportunities for these types of nanomagnet arrangements in future applications. ACKNOWLEDGMENTS Research at Brown was supported by grants from the National Science Foundation (Grant No. NSF/DMR , Materials Research Science and Engineering Center at Johns Hopkins University, and Grant No. NSF/DMR ). 1 Y. Ji, G. J. Strijkers, F. Y. Yang, C. L. Chien, J. M. Byers, A. Anguelouch, Gang Xiao, and A. Gupta, Phys. Rev. Lett. 86, 5585 (2001). 2 A. Anguelouch, A. Gupta, G. Xiao, G. X. Miao, D. W. Abraham, S. Ingvarsson, Y. Ji, and C. L. Chien, J. Appl. Phys. 91, 7140 (2002). 3 F. Y. Yang, C. L. Chien, E. F. Ferrari, X. W. Li, G. Xiao, and A. Gupta, Appl. Phys. Lett. 77, 286 (2000). 4 Q. Zhang, A. V. Nurmikko, A. Anguelouch, G. Xiao, and A. Gupta, Phys. Rev. Lett. 89, (2002). 5 R. P. Cowburn, D. K. Koltsov, A. O. Adeyeye, M. E. Welland, and D. M. Tricker, Phys. Rev. Lett. 83, 1042 (1999). 6 D. S. Rodbell, J. Phys. Soc. Jpn. 21, 1224 (1966). 7 J. S. Parker, S. M. Watts, P. G. Ivanov, and P. Xiong, Phys. Rev. Lett. 88, (2002). 8 A. Gupta, X. W. Li, S. Guha, and G. Xiao, Appl. Phys. Lett. 75, 2996 (1999). 9 R. P. Cowburn, D. K. Koltsov, A. O. Adeyeye, and M. E. Welland, Appl. Phys. Lett. 73, 3947 (1998). 10 G. X. Miao, G. Xiao, and A. Gupta (unpublished). 11 K. Y. Guslienko, S. B. Choe, and S. C. Shin, Appl. Phys. Lett. 76, 3609 (2000). 12 W. C. Uhlig and J. Shi, Appl. Phys. Lett. 84, 759 (2004). 13 G. X. Miao (unpublished). 14 R. Pulwey, M. Zölfl, G. Bayreuther, and D. Weiss, J. Appl. Phys. 91, 7995 (2002). 15 M. Schneider and H. Hoffmann, J. Appl. Phys. 86, 4539 (1999).