Magnetoelastic coupling of compressively stressed FeÕGaAs001 G. Wedler and B. Wassermann Institut fu ¨r Experimentalphysik, Freie Universita ¨t Berlin, Arnimallee 14, D14195 Berlin, Germany R. Koch Paul-Drude-Institut fu ¨r Festko ¨rperelektronik, Hausvogteiplatz 5 7, D10117 Berlin, Germany Received 12 March 2002; published 14 August 2002 The magnetoelastic coupling, a property of major importance in heteroepitaxy, describes the dependence of the free energy of magnetic materials on strain and stress. Using our versatile UHV cantilever beam magne- tometer we have investigated the magnetoelastic coupling constants B 1 and B 2 of Fe/GaAs001, a system yielding compressed Fe001films. Both constants exhibit a strong dependence on the compressive film stress, which so far is not explained by theory: At stress values of about -0.5 GPa the sign of B 1 is positive and thus opposite to the bulk value; B 2 is about 15% smaller than the bulk value. DOI: 10.1103/PhysRevB.66.064415 PACS numbers: 75.70.-i, 75.80.+q, 68.60.Bs I. INTRODUCTION The magnetoelastic MEcoupling interrelates the state of magnetization of ferromagnetic materials with lattice distor- tions. It is responsible for the well-known phenomenon of magnetostriction and can be quantified either via the ME coupling constants B i Ref. 1or the magnetostriction con- stants i Ref. 2. For typical magnetostrictive distortions of 10 -3 –10 -5 the changes in the ME energy are negligibly small compared with the other magnetic contributions to the free energy. In heteroepitaxial thin films, however, intrinsic distortions due to lattice mismatch between film and sub- strate usually amount to a strain of several percent ( 10 -2 ), thus affecting the ME energy substantially. It may even become comparable to the magnetocrystalline energy, which offers the challenging opportunity to stabilize new magnetic anisotropies different from the bulk 3,4,5 in het- eroepitaxial thin films. In the recent years it was convincingly shown by various experimental 6–11 and theoretical 12–17 studies that at strain values of a few percent a linear description of the ME energy ( F ME B ) is no longer sufficient. It is convenient to use a series expansion in strain for the experimentally determined ME coupling constants B i expt : B i expt =B i +D i +••• . 1 Here B i is a ME coupling constant of the unstrained state, whereby good agreement with the corresponding bulk con- stants was found. 7–9 D i is a second-order constant. In the case of Fe the second-order expansion of B 1 is valid for a strain of at least 4% according to theoretical calculations. 15 Experimental studies, however, point to the necessity of third- or higher-order terms already at a tensile strain above +0.5%, 9 whereas films under compression have not been investigated so far. In this study we report on the ME coupling constants B 1 and B 2 of Fe/GaAs001, a technologically important het- eroepitaxial system, where the influence of compressive strain on the ME coupling can be investigated. The misfit is -1.36% ( a Fe =0.2866 nm, a GaAs /2=0.2827 nm); i.e., ideal coherent growth is accompanied by a compressive stress of -2.8 GPa. Being one of the promising candidates for spin injection into semiconductors, 18 growth, epitaxy, and magnetism of Fe/GaAs001have been studied intensively in the past. It is well established that Fe grows epitaxially in its bcc modification with the 001plane parallel to the substrate 19 both at elevated temperatures 420–450 KRefs. 20–22and at room temperature Refs. 23–26. However, due to diffusion of As and Ga into the Fe film, no sharp interface is formed. 22,27,28 Recent stress investigations 29 re- vealed that interdiffusion is reduced in films deposited at 300 K, but definitely not negligible. II. EXPERIMENT The experiments were performed in a multiple chamber UHV system base pressure 1 10 -10 mbar) equipped with a sensitive cantilever beam magnetometer CBMfor in situ stress and magnetic measurements 30 and a four-grid low- energy electron diffraction LEEDoptics for in situ control of the substrate and film quality as well as a homebuilt UHV scanning tunneling microscope STMfor in situ structural investigations. The substrates were cut from 100-m-thick As-capped GaAs001wafers, which were prepared in a separate molecular beam epitaxy MBEchamber by re- moval of the oxide, deposition of an 0.5-m-thick GaAs001buffer layer, and coating with a thin As layer. Prior to the Fe deposition the As capping layer was desorbed at about 650 K until 001LEED spots were obtained; some of the substrates were additionally annealed for 1 h at 810 K, which exhibit also faint (2 4) and c (4 4) superstructure spots. Fe was electron beam evaporated from a Knudsen- type tungsten source at a pressure better than 2 10 -9 mbar; the deposition rate determined by a quartz crystal microbalance was 0.0080.001 nm/s. Immediately after the film preparation the magnetic measurements were performed in situ. For the measurements of B 1 the substrate orientation has to be chosen so that the 100and 010di- rections of the Fe001film are parallel to the length and width of the cantilever beam; the difference in magnetostric- tive stress upon saturation magnetization along 100and PHYSICAL REVIEW B 66, 064415 2002 0163-1829/2002/666/0644155/$20.00 ©2002 The American Physical Society 66 064415-1