JOURNAL OF APPLIED PHYSICS VOLUME 84, NUMBER 3 1 AUGUST 1998 Growth temperature dependence of the magnetic and structural properties of epitaxial Fe layers on MgO 001... S. M. Jordan, J. F. Lawler, R. Schad, H. van Kempena) Research Institute for Materials, University of Nijmegen, Toernooiveld 1, NL-6525 ED Nijmegen, The Netherlands. Received 10 November 1997; accepted for publication 29 April 1998 We have studied the growth and magnetic properties of molecular beam epitaxy grown layers of bcc Fe 001 on MgO 001 substrates at a wide range of temperatures. For growth temperatures in the range 80 595 K, the iron forms islands which increase in lateral size with increasing temperature. Completed films in the same temperature range show the magnetic properties expected for a system with biaxial anisotropy, and a coercivity of 10 Oe. The value of the first cubic anisotropy constant divided by the magnetization (K1 /M) remained constant. No evidence for uniaxial magnetic anisotropy in the films was found. Above 595 K, the films' structure and magnetic properties changed dramatically to those characteristic of a particulate system. © 1998 American Institute of Physics. S0021-8979 98 06315-4 I. INTRODUCTION electrical contact between the deposited film and the Mo MgO(001) forms an ideal substrate for the growth of sample holder. The substrates were heated to 1070 K in ul- iron layers for several reasons. Firstly, the lattice mismatch is trahigh vacuum UHV for 1 minute and then analyzed by only 4%, and the substrate is robust, transparent and easily Auger electron spectroscopy AES . A KLL C peak was seen obtained in large pieces. A monolayer of Fe on MgO is pre- corresponding to 6% of 1 monolayer. Heating the MgO to dicted to display a highly enhanced magnetic moment.1 A temperatures as high as 1400 K did not reduce this contami- uniaxial magnetic anisotropy in films evaporated at an ob- nation. No traces of Au were found in the center of the lique angle of incidence has also been seen,2,3 and attributed sample. Atomic force microscopy investigations showed the to in-plane distortion of the Fe. The Fe(001) surface is also substrates to be of exceptional flatness; single atom high ter- expected to display a highly spin-polarized surface state.4 races of width up to 200 nm were seen. Interesting points are that the predicted enhanced magnetic Fe layers were grown using a Knudsen cell at a rate of moment has not been confirmed experimentally;5 neither has 0.13 nm per minute. The iron atoms were incident at angle of the spin-polarization of the surface state. 15° to the sample normal, the flux being directed along the The molecular beam epitaxy MBE growth of one ma- Fe 110 axis. The sample was maintained at the required terial upon another is influenced strongly by surface diffu- temperature by electron heating of the sample holder; liquid sion, which allows transport of material during the ordering nitrogen cooling was also available. process. Above a critical temperature, the Schwoebel barrier6 is overcome, and atoms can diffuse not only in an upward III. LOW-ENERGY ELECTRON DIFFRACTION LEED... but in a downward direction at step edges. Thu¨rmer et al.7 INVESTIGATION found Fe/MgO to form an ideal Schwoebel system. The completed films were examined by LEED confirm- Several growth8­11 and magnetic12,13 studies of thin ep- ing that the Fe 110 axis is parallel to MgO 100 , and the itaxial films have already been reported. This study differs in (001) planes of the two materials are parallel. Only weak that we report direct measurements of the surface morphol- correlation with temperature was found, the diffracted spots ogy using scanning tunneling microscopy STM over a wide becoming less diffuse as the temperature was increased. A deposition temperature range. We were also able to measure sample grown at 80 K was warmed to room temperature the magnetic properties of samples immediately after their in during LEED observations. It was observed that the pattern situ STM investigation. In addition, we have used a novel changed in a nontrivial way, some diffracted orders becom- technique to measure the magnetic anisotropy magneto- ing more diffuse with others becoming sharper. This indi- optically. cates that the film changes in structure as the temperature is increased. The samples grown below room temperature II. SAMPLE PREPARATION should be considered as annealed samples, since the STM Commercial substrates were cleaned by washing first and magneto-optic Kerr effect MOKE studies were carried with hexane, then with acetone and finally rinsed twice with out at room temperature. propan-2-ol. 50 nm thick Au stripes were deposited on two sides close to the edges of the MgO to provide a reliable IV. STM INVESTIGATION STM investigations of the samples were made in situ a Corresponding author, Electronic mail: hvk@sci.kun.nl using a locally developed STM. Representative images of 5 0021-8979/98/84(3)/1499/5/$15.00 1499 © 1998 American Institute of Physics 1500 J. Appl. Phys., Vol. 84, No. 3, 1 August 1998 Jordan et al. FIG. 1. STM images for growth at 80 top and 295 K bottom . Image sizes FIG. 3. 50 50 nm STM image for growth at 595 K. Black-white contrast 50 25 nm. Black-white contrast is 2.1 nm top , 2.5 nm bottom . 0.9 nm. Parameters: Vtip-86 mV, setpoint 90 pA. Parameters: top, Vtip100 mV, setpoint 59 pA; bottom, Vtip 200 mV, setpoint 150 pA. creased step edge diffusion has begun to dominate, causing square islands. nm thick films with scan size 50 nm are shown in Figs. 1­3. Several authors have reported pyramid growth of Fe on It is clear that as the growth temperature is increased, the MgO7 at a temperature of 400­ 450 K and GaAs.14 They islands both increase in size and become squarer. Table I found pyramids with facet angles of 27° and 13° respec- gives the average island diameter and the overall roughness tively, the formation of which was explained in terms of averaged over several 200 nm images. surface diffusion. For a deposition temperature of 395 K we The difference caused by increasing the growth tempera- obtained pyramidal islands with facet angles of 20°; films ture from 80 to 295 K is mainly a small increase in island grown at higher temperature had much reduced facet angles size, together with an increase in roughness Fig. 1 . The with clearer atomic steps. islands also appear to become less rounded, and less uniform At the highest growth temperature at which continuous in size. It should be noted that the lower temperature sample films are produced, 595 K, large ( 15 nm terraces are was annealed to room temperature before measuring, and formed. The step edges are aligned along Fe 100 . Two definite changes in the LEED pattern were observed. A much atomic steps are visible along the white line in Fig. 3, with more dramatic change is observed as the growth temperature heights corresponding to a/2. These large flat terraces may is increased to 395 K Fig. 2 . The islands begin to become form a basis for further study, such as scanning tunneling square, with a decrease in roughness being observed. spectroscopy; we conclude that this temperature is the opti- Between 395 and 495 K a subtle change occurs in the mum to produce flat films, since still higher temperatures island shape. At the lower temperature, round structures produce discontinuous films. within the islands are still visible, but at the higher tempera- ture, the structure changes to square pyramidal islands with V. MAGNETIC BEHAVIOR clearly defined steps on their faces. At this point the in- A. Hysteresis After STM investigation the films were assessed for magnetic properties by in situ MOKE. Hysteresis loops were taken with H at various angles to the substrate lattice direc- tions. The incident light was p polarized, so that the magne- tooptic signal was due solely to magnetization in the longi- tudinal direction,15 which was parallel to H. A typical hysteresis loop with H applied along the mag- netic easy axis is shown in Fig. 4 a . The low coercivity and TABLE I. Average island size and rms roughness summarized. The stan- dard deviation in island sizes was approximately 15% in all cases. Growth temp. Average island size nm rms roughness nm 80 5.0 0.37 295 7.4 0.52 395 9.1 0.42 FIG. 2. STM images for growth at 395 top and 495 bottom . Images 50 495 15.5 0.58 25 nm. Both images have a black-white height of 2.1 nm. Parameters: top, 595 30.7 0.28 Vtip250 mV, setpoint 190 pA; bottom, Vtip140 mV, setpoint 235 pA. J. Appl. Phys., Vol. 84, No. 3, 1 August 1998 Jordan et al. 1501 FIG. 6. Ex situ MOKE loop for a 10 nm Fe film protected by 2 nm of Au grown at 695 K. hard axis, then no secondary steps in the hysteresis loop are seen. We believe that this is due to domain formation, the hysteresis behavior being then due to domain motion, rather than abrupt changes of magnetization. Alternatively, the steps may have moved to a field greater than 600 Oe, the maximum field attainable. If is slightly greater than /4, then M will lie initially along I, since this easy axis is closer to negative H. Easy axis FIG. 4. In situ MOKE hysteresis loops from a 5 nm Fe film grown at 295 K. a is with H along Fe 100 , b with H almost along the magnetic hard axis A will become slightly preferable over B at positive H, since Fe 110 ). M will have to turn through a smaller distance to reach it from the initial state, I. However, M will eventually lie along B since A is further from H. The motion of M from A to B the steepness of the reversals at the coercive points indicate a causes the secondary jumps seen in Fig. 4 b . Gu et al.18 high quality epitaxial film with a low concentration of inclu- have investigated this process by Lorentz microscopy and sions or defects. determined that this jump occurs by domain motion. Mea- Figure 4 b shows a typical loop along the magnetic hard surements of the hysteresis loops support the assertion that axis, which was found to lie along Fe 110 . The secondary M in Fig. 4 b lies along A between 10 and 400 Oe. Experi- jumps seen at 400 Oe in Fig. 4 b are a consequence of the mentally, the position, but not the height of the steps is found biaxial magnetic anisotropy; similar loops have been re- to depend strongly on . ported by Postava et al.16 in Fe/MgO and Daboo et al.17 in A uniaxial anisotropy revealed in the hysteresis loop Fe/GaAs. All films with growth temperatures at or below taken along a particular easy axis has been reported by 595 K showed this behavior. Postava et al.16 and Durand et al.3 The loops displayed steps The processes governing the presence of the jumps men- where the longitudinal magnetization was zero between 0 tioned in the paragraph above are explained in Fig. 5. The and approximately 10 Oe. The loops were perfectly sym- diagram shows the evolution of the magnetization, M, for a metrical. We also saw steps between 7 and 10 Oe in our small positive H, just before the coercive point is reached. If hysteresis loops when H was applied within 20° of either /4, then the two directions I and I will both be equiva- easy axis in our case. The loops were asymmetric, with the lent. The transitions I A and I B are energetically step appearing only on one side of the loop. The longitudinal equivalent, hence at the coercive point (H 0) A and B are magnetization at the steps was also nonzero. We believe that both equally likely as final destinations for M. Experimen- our steps are due to domain motion rather than switching tally, we found that when H is applied exactly along either processes due to an inequivalence of the magnetic easy axes. Figure 6 shows a hysteresis loop from a film grown at 695 K, at which the films become discontinuous. This shape is characteristic of a particulate system. The sample dis- played no magnetic anisotropy. Paramagnetic behavior has also been reported by Park et al.19 at growth temperatures of 700 K. B. Determination of K1 /M A second series of samples was grown for the ex situ determination of the ratio of the first cubic magnetic anisot- FIG. 5. Magnetization processes in a biaxial film. As H increases from negative saturation, M will move from I to A. A further jump can then occur ropy parameter K1 see Eq. 1 to the saturation magnetiza- when M moves from A to B for certain values of . tion, M. The samples were 10 nm thick to provide a larger 1502 J. Appl. Phys., Vol. 84, No. 3, 1 August 1998 Jordan et al. TABLE II. Summary of the values of K1 /M in Oe for 10 nm Fe films The values of K1 /M were found from the versus protected by 2 nm Au at various temperatures. curves by making a discrete Fourier transform of the values of sin as described by Pastor and Torres,23 whose equations Au growth temp. K 80 160 295 595 are equally applicable to biaxial anisotropy if their periodici- Fe growth temp. K ties are doubled. The component with a periodicity of 4 in 80 216 10 data over the range 0 is given by 160 225 6 295 232 8 N 383 232 8 f 4 595 190 6 208 8 190 6 2 Hk /H 12 Hk /H 3 , 2 where N is the number of data points and Hk is K1/2M. Here, Hk /H 12. Analysis of the curves yields the re- magnetooptic signal , protected from oxidation by 2 nm of lation between f 4 and H, which can then be fitted to Eq. 2 Au. The growth temperatures of the layers were varied inde- to give Hk . The quality of fit was used to estimate the error pendently to isolate effects due to increased interdiffusion in Hk . Tests performed on using data from numerical solu- between the Au and Fe; the temperatures and measured val- tions of Eq. 1 demonstrated the correctness of this method. ues of K1 /M are summarized in Table II. The values of K1 /M are all equal within experimental The amount of interdiffusion between the Au and Fe error, indicating that the effect of varying the deposition tem- layers was assessed by AES of the completed sample. Typi- perature on the magnetic anisotropy is too small for this cally, the Fe peak due to the underlayer was 4% of the technique to measure. One sample was grown with the Fe height of the Au overlayer peak. Little correlation of this flux incident along the Fe 100 directions. This gave a Hk ratio with growth temperature was seen; however it increased value of 148 10, scant evidence that the direction of inci- to 30% when the Au layer was grown at 595 K, indicating dent atoms relative to the substrate lattice affects the biaxial significant intermixing. anisotropy. We could not detect any difference in the hyster- The technique used relies on measuring hysteresis loops esis loops from this sample and one grown at the same tem- magnetooptically as the sample is rotated in-plane. Use of perature with the Fe flux along Fe 110 . The values with flux vectorial MOKE15,20 allows the components of M normal along Fe 110 are quite close to the bulk value, which lies but in-plane and parallel to the applied field, H, to be mea- between 260 and 280 Oe.24 sured independently. The angle between M and H can then Vibrating sample magnetometer VSM measurements be found, and plotted against the in-plane angle, see Fig. gave the saturation magnetization of the films to be 1800 5 . This yields similar information to that produced by a 200 emu/cc, consistent with the bulk value. Our technique torque magnetometer,21 and is used to determine K1. Typical was not sensitive enough to make any difference between the versus curves are shown in Fig. 7, together with simu- samples apparent, however. lations found by minimizing the free energy equation,22 Goryunov et al.25 measured K1 /M over a wide range of which consists of an anisotropy and a magnetostatic term, thicknesses at various temperatures. They found that this pa- K rameter varies strongly with film thickness, the difference at E 1 low thicknesses also depending on temperature. The values 8 cos4 MHcos c.g.s. units assumed . 1 at 300 and 475 K were 185 and 236 Oe respectively for a 10 In practice, 8 curves with H ranging between 50 and 3000 nm sputtered film. Kohmoto and Alexander12 reported a Oe were measured. The curves show the expected behavior value of 300 Oe for an epitaxial film. for a biaxial system - no evidence of uniaxial anisotropy was seen at these fields. VI. CONCLUSIONS We have found an interesting relation between island shape and size with growth temperatures between 80 and 595 K; at still higher temperatures the Fe became particulate. This forms a useful system, since rough, smooth or discon- tinuous oriented films can be chosen simply by selection of the deposition temperature. However, the magnetic proper- ties of the Fe appear to be fixed within the epitaxial tempera- ture range. It is possible that a change in magnetic anisotropy with growth temperature occurs, since it has been reported that the lattice parameter of Fe has a complex thickness dependence with the most pronounced changes around 10 monolayers.8 A change in deposition temperature will affect the Fe/MgO interface, and perhaps also the lattice parameter and mag- netic properties of a thin film. However, our films were much FIG. 7. Curves showing angle between H and M) as a function of in-plane angle for several values of H. The biaxial anisotropy is clearly thicker than this in order to provide sufficient magnetooptic seen. The solid lines are simulations, the points experimental values. signal. This perhaps explains our failure to find a correlation J. Appl. Phys., Vol. 84, No. 3, 1 August 1998 Jordan et al. 1503 of the value of K 6 1 / M with deposition temperature. The ef- R. Schwoebel, J. Appl. Phys. 40, 614 1969 . fect of the step edges on K 7 K. Thu¨rmer, R. Koch, M. Weber, and K. Rieder, Phys. Rev. Lett. 75, 1767 1 is likely to be small, since their effect is inversely proportional to film thickness.14 Atomic 1995 . 8 steps on the substrate surface have been shown to affect the B. Lairson, A. Payne, S. Brennan, N. Rensing, B. Daniels, and B. Clem- ens, J. Appl. Phys. 78, 4449 1995 . anisotropy of a film,26 but only at a deposited thickness 9 T. Urano and T. Kanaji, J. Phys. Soc. Jpn. 57, 3403 1988 . equivalent to 2 monolayers. The use of thinner films will 10 J. Lawler, R. Schad, S. Jordan, and H. van Kempen, J. Magn. Magn. increase the perturbing effects of the Au capping layer, ne- Mater. 165, 224 1997 . 11 cessitating the use of in situ MOKE for anisotropy measure- P. Thibado, E. Kneedler, B. Jonker, B. Bennett, B. Schanabrook, and L. Whitman, Phys. Rev. B 53, R10481 1996 . ments. This is made difficult by the high H field required. 12 O. Kohmoto and C. Alexander, Jpn. J. Appl. Phys., Part 1 31, 2101 1992 . ACKNOWLEDGMENTS 13 H. Ohta, S. Imagawa, M. Motokawa, and E. Kita, J. Phys. Soc. Jpn. 62, 4467 1993 . The work here has been financially supported by the 14 M. Gester, C. Daboo, S. Gray, and J. Bland, J. Magn. Magn. Mater. 165, Dutch Foundation for the Fundamental Research of Matter 242 1997 . 15 FOM , which is, in turn, financially supported by the Dutch S. Jordan and J. Whiting, Rev. Sci. Instrum. 67, 4286 1996 . 16 K. Postava, H. Jaffres, F. Nguyan van Dau, M. Goiran, and A. Fert, J. Organization for Scientific Research NWO . One of the au- Magn. Magn. Mater. 172, 199 1997 . thors J.F.L. would like to acknowledge the assistance of the 17 C. Daboo, R. Hicken, D. Eley, M. Gester, S. Gray, A. Ives, and J. Bland, Human Capital and Mobility program of the European Com- J. Appl. Phys. 75, 5586 1994 . 18 munity Contract No. ERBCHBICT 94 1207 . R.S. would E. Gu, J. Bland, C. Daboo, M. Gester, L. Brown, R. Ploessl, and J. Chap- man, Phys. Rev. B 51, 3596 1995 . like to acknowledge the assistance of the Brite-Euram pro- 19 Y. Park, S. Adenwalla, G. Felcher, and S. Bader, Phys. Rev. B 52, 12779 gram of the European Community contract number BRE 1995 . 2-CT93-0569 . The help of A. M. Keen is also acknowl- 20 C. Daboo, J. Bland, R. Hicken, A. Ives, M. Baird, and M. Walker, Phys. edged. Rev. B 47, 11852 1993 . 21 S. Jordan and C. Prados, J. Magn. Magn. Mater. 172, 69 1997 . 22 R. Bozorth, Phys. Rev. 50, 1076 1936 . 1 C. Li and A. Freeman, Phys. Rev. B 43, 780 1991 . 23 G. Pastor and M. Torres, J. Appl. Phys. 58, 920 1985 . 2 Y. Park, E. Fullerton, and S. Bader, Appl. Phys. Lett. 66, 2140 1995 . 24 Magnetic Properties of Metals, d-elements, Alloys and Compounds, Data 3 O. Durand, J. Childress, P. Galtier, R. Bisaro, and A. Schulhl, J. Magn. in Science and Technology, edited by H. Wijn Springer, Berlin, 1991 . Magn. Mater. 145, 111 1995 . 25 Y. Goryunov, N. Garifyanov, G. Khaliunnin, I. Garifullin, L. Tagirov, F. 4 J. Stroscio, D. Pierce, A. Davies, R. Celotta, and M. Weinert, Phys. Rev. Schreiber, T. Muhge, and H. Zabel, Phys. Rev. B 52, 13450 1995 . Lett. 75, 2960 1995 . 26 D. Huang, J. Lee, G. Mulhollan, and J. Erskine, J. Appl. Phys. 73, 6751 5 Y. Huang, C. Liu, and G. Felcher, Phys. Rev. B 47, 183 1992 . 1993 .