- Email: [email protected]
Atomic structure and magnetic properties of Mn on InAs(1 0 0)
Applied Surface Science 212–213 (2003) 17–25
Atomic structure and magnetic properties of Mn on InAs(1 0 0) K. Hricovinia,b,*, P. De Padovac, C. Quaresimac, P. Perfettic, R. Brochiera, C. Richtera, V. Ilakovaca, O. Heckmanna, L. Lechevalliera, P. Bencokd, P. Le Fevreb, C. Teodorescub a
LMPS, Universite´ de Cergy-Pontoise, Neuville/Oise, 95031 Cergy-Pontoise, France b LURE, Bat. 209 d., 91898 Orsay, France c CNR-ISM, Via Fosso del Cavaliere, 00133 Rome, Italy d ESRF, Grenoble, France
Abstract We have studied Mn/InAs(0 0 1)(4 2)c(8 2) interfaces by high-resolution core-level spectroscopy, surface EXAFS and by X-ray magnetic circular dichroism (XMCD). The growth of manganese is found to depend strongly on the temperature of InAs. At low temperatures (room and liquid nitrogen temperature (LNT)), Mn atoms form the compound MnAs in the surface region leaving metallic In islands on the surface. Manganese deposition at 530 K results in strong diffusion of Mn into the substrate followed by formation of the compound In1xMnxAs in which Mn atoms are substituted in indium-sites. We show by XMCD that In1xMnxAs is ferromagnetic with a relatively high value of an average magnetic moment per Mn atom, exceeding 0.7 mB at low temperatures and with an individual magnetic moment of about 2.4 mB. The temperature dependence of the Mn magnetic moment follows the Brillouin function and it drops to zero at temperature of about 150 K for measurements in an external magnetic field of 5 T. # 2003 Published by Elsevier Science B.V. PACS: 61.10.Ht; 68.35.Fx; 75.70.Pa; 75.30.Et Keywords: Indium arsenide; Manganese; Diluted magnetic semiconductor; XMCD
1. Introduction A new category of semiconductors, called diluted magnetic semiconductors (DMS), are alloys between a nonmagnetic semiconductor and a magnetic element (manganese in most cases). An important property of these materials is that the carrier density can be controlled over a wide range between n- and p-types. This opens up the possibility to control magnetic *
Corresponding author. Tel.: þ33-1-3425-7029; fax: þ33-1-3425-7071. E-mail address: [email protected] (K. Hricovini).
properties simply by changing the carrier density. This behaviour is generally called ‘‘carrier-induced ferromagnetism’’ because hole carriers introduced into the system mediate the ferromagnetic coupling between Mn ions, although its microscopic mechanism has been controversial until now. So, in such systems, both, the charge and the spin of the electron can potentially be used to create new types of semiconductor devices. III–V based DMS have attracted considerable interest because of their relatively high Curie temperature (Tc 110 K in the case of GaMnAs [1]). Theoretical predictions indicate [2] that Tc can be raised above
0169-4332/03/$ – see front matter # 2003 Published by Elsevier Science B.V. doi:10.1016/S0169-4332(03)00013-8
18
K. Hricovini et al. / Applied Surface Science 212–213 (2003) 17–25
300 K for the materials containing a high concentration of magnetic ions or consisting of lighter elements. In the case of In1xMnxAs, the Curie temperature reported to date is relatively low Tc < 55 K. The major obstacle in making III–V semiconductors magnetic has been the low solubility of magnetic elements (such as Mn) in the compounds. Introduction of a large controllable amount of manganese atoms into the III–V semiconductor InAs was first accomplished in 1989 by using molecular beam epitaxy (MBE) (e.g. see review [1]). Structural studies of MBE prepared In1xMnxAs [3–5] done by EXAFS and fluorescence have revealed the local order around Mn atoms. It was found that in the high-growth-temperature samples (temperature of the substrate Ts ¼ 550 K) Mn atoms are primarily incorporated in the form of MnAs clusters with NiAs structure. No significant disorder is observed. In the low-growth-temperature samples (Ts ¼ 490 K), the majority of Mn atoms form small, disordered, sixfold-coordinated centers with As. According to studies in [4], only a very small fraction of Mn atoms may substitute for In in the zinc-blend InAs structure. On the contrary to the thin layers, in case of growth of quantum dots [5] the majority of the Mn atoms were substituted in the indium-sites (In-sites) of InAs. It is believed that the growth in a 3D mode relaxed the strain on (In, Mn)As, which may have suppressed the formation of NiAs-type MnAs. The electronic structure of In1xMnxAs is interesting because it has been reported that In1xMnxAs grown on GaSb substrates exhibits field-induced ferromagnetism [6]. Unlike Ga1xMnxAs, the impurity-band-like states near the top of the valence band (VB) have not been observed by angle-resolved photoemission spectroscopy, in spite of the fact that in both systems Mn 3d states are in the Mn2þ configuration. The electronic structure of the diluted magnetic semiconductor (In, Mn)As has recently been calculated by Akai [7] by using the Korringa–Kohn–Rostoker coherent-potential and local density approximation (KKR–CPA–LDA). The calculations show that the ground state of (In0.94Mn00.06)As is ferromagnetically stabilized by double exchange and the local Mn magnetic moment has a value of about 4.2 mB. On the contrary, (In0.90Mn0.06As0.04)As has no magnetic order and the density of states at EF is primarily due to
impurity bands composed of As electrons. An important fact is that the ferromagnetic state is half-metallic, namely, the Fermi surface exists only in the majority spin band. In this paper, we report photoemission core-level spectroscopy, surface EXAFS and X-ray magnetic circular dichroism (XMCD) studies on In1xMnxAs samples prepared by an alternative technique, which consists in manganese deposition on the semiconductor substrate at higher temperature. We show that the diffusion of Mn in the substrate can be controlled and we observe by XMCD ferromagnetic ordering of manganese atoms diluted into InAs(1 0 0). The temperature of the substrate was chosen to be high enough to allow the diffusion of Mn atoms, without being so high that segregation of MnAs clusters could occur [8]. We found that in the literature there are still controversies concerning the magnetism of MnInAs. XMCD is an appropriate technique to investigate individual and macroscopic magnetic moments of Mn atoms: it is element sensitive and it provides quantitatively the local spin and orbital magnetic moments through the application of the sum rules [9]. We found a Mn macroscopic moment of about 0.7 mB, and for the individual Mn moment we found from the L3/L2 intensity ratio [10] a value of about 2.4 mB (theoretical result 4.2 mB [7]).
2. Experimental Photoemission core-level spectroscopy experiments were carried out in ultra high vacuum conditions (base pressure 8 1011 mbar) at the synchrotron radiation beamline VUV of ELETTRA (Trieste, Italy). All spectra reported were collected at normal emission. Undoped InAs(0 0 1) (3.5 O cm) single crystals were used for all the measurements. First, the Interminated InAs(0 0 1) clean surfaces were prepared after several cycles of Ar-ion bombardment at room temperature and subsequent heating to a temperature of 620 K, followed by some additional annealing at temperatures of 700 K. The study of LEED figures showed that we achieve a high-quality In-terminated InAs(0 0 1)c(8 2) reconstructed surface. Mn was deposited from a Knudsen cell evaporator and calibrated with a quartz monitor.
K. Hricovini et al. / Applied Surface Science 212–213 (2003) 17–25
Photoemission spectra were measured after cooling the sample to 150 K using an angle-resolved hemispherical electron energy analyser (acceptance 28), and the total energy resolution was better than 50 meV. The Fermi level was measured on a gold foil in electrical contact with the sample. Surface EXAFS and XMCD experiments were performed at the synchrotron radiation laboratory LURE in Orsay, France. We used a molecular beam epitaxy chamber with standard surface preparation techniques, LEED, RHEED and Auger spectroscopy facilities. For XMCD measurements, the MBE chamber was connected to the experimental chamber equipped with a liquid He cooled sample holder and a super conducting magnet generating magnetic fields up to 5 T. The whole system was installed on the SU 23 beamline of the Super ACO storage ring, which provides photons in the energy range from 80 to 1000 eV with about 60% degree of circular polarisation. The XMCD spectra were performed by recording the total electron yield signal from the sample.
3. Results and discussion 3.1. Core-level spectroscopy By photoemission core-level spectroscopy of In 4d and As 3d, we have first studied the growth mode of Mn on InAs(1 0 0)c(8 2) at three different temperatures: 530 K, room temperature and liquid nitrogen temperature (LNT). We did not find any detectable signal from Mn 3p core-levels for deposition at any temperature. In Fig. 1, we display In 4d core-level spectra for different Mn coverage deposited at RT (Fig. 1a) and at 530 K (Fig. 1b). The growth mode, as seen through core-level photoemission spectra, is very similar at both temperatures, LNT and RT. So, in the following, we will mostly compare Mn growth at RT and at 530 K. The decomposition of the direct photoemission peak In 4d from a clean InAs(1 0 0)c(8 2) surface is now well-established [11]: in addition to the bulk component (the component B with the highest spectral intensity shown in Fig. 1a and b, upper panel), there is a clear evidence of two surface peaks S1 and S2. S1 having lower kinetic energy (0.31 eV) than the bulk peak is attributed to surface In atoms. The peak S2 (þ0.25 eV from the bulk peak) corresponds to In atoms of self
19
organized In chains running in the (1 1 0) direction of the InAs(1 0 0)(4 2)c(8 2) surface [11]. Dramatic changes appear in the core-level spectra after deposition even of the very first thin Mn layers. For both substrate temperatures a new component lying at higher kinetic energy then the bulk one (þ0.75 eV) appears in In 4d spectra (black colour component S3 in the middle and in the lower panel of Fig. 1; in order to follow in a clear way the evolution of the spectra, in the middle and bottom panels we do not show the bulk, S1 and S2 components any more, only S3 is plotted schematically). S3 is attributed to free In atoms (i.e. atoms forming In–In bonds, rather than In–As ones) that are the result of the reaction between Mn and As. This is immediately evident when one compares the enthalpies of compound formation: 8 and 50 kJ/mol in the case of MnIn and MnAs, respectively. This is why, from the energy balance point of view, the MnAs compound formation is much more probable. Now, it is interesting to determine the dynamics of the reaction and of the Mn atom diffusion as a function of substrate temperature. In spite of the fact, that deposited quantities of Mn are not exactly the same for RT and 530 K depositions, from Fig. 1 it is clear that the amount of free In atoms on the surface is much larger in the case of RT depositions. After deposition of 3.5 Ml at RT the In 4d photoemission signal becomes dominant on the surface (as seen in the bottom panel of Fig. 1a) and As 3d photoemission is strongly reduced (not shown). For 7 Ml deposited at liquid nitrogen temperature, no As 3d signal is detectable and the surface is completely covered by In. For high Mn coverages (as shown in Fig. 1a, bottom panel) In 4d exhibits Doniah–Sunjic function-like shape, typical for a metallic surface. The growth of Mn at 530 K shows only very slight increase of In 4d signal due to free In atoms, so it appears that the Mn atoms can penetrate inside the bulk (Fig. 1b). 3.2. Valence band spectroscopy This situation is further confirmed by the valence band (VB) spectra in Fig. 2. The spectra at RT depositions (Fig. 2a) are difficult to analyse, because they are a superimposition of signals from free In atoms and Mn þ InAs compounds. As seen in Fig. 2a, the metallicity of the surface is confirmed by valence band spectra: electron emission from Fermi level is
20
K. Hricovini et al. / Applied Surface Science 212–213 (2003) 17–25
Fig. 1. In 4d core-level photemission spectra of clean InAs(1 0 0)(4 2)c(8 2) and Mn/InAs(1 0 0)(4 2)c(8 2) for different Mn coverages. The spectra were measured at RT (a) and at 530 K (b) with a photon energy of 84 eV. Vertical bars indicate energy position of the component attributed to free indium atoms on the surface. See the text for other details.
observed after 0.5 Ml of Mn at RT deposition. On the contrary, the VB spectra measured after the deposition at 530 K are comparable to spectra measured on MBE grown In1xMnxAs samples [6]. It is interesting to note that the VB spectrum does not exhibit important changes with increasing Mn amount at 530 K (Fig. 2b), bottom panel). For 7 Ml, however, an emission from the Fermi level is observed, which is in agreement with increasing intensity of the S3 component in In 4d spectrum (Fig. 1b, bottom panel). From the photoemission spectra we can draw a preliminary image of Mn growth on InAs(0 0 1)(4 2)c(8 2) at both low (RT and LNT) and high (530 K)
temperatures. In both cases, the growth is governed by strong chemical reaction between Mn and As atoms. Consequently, an indium metallic layer is formed on the surface for low temperature deposition. For deposition at 530 K, however, we find an important difference in growth mode. It appears that at 530 K Mn atoms diffuse rapidly into the substrate leaving the surface much less perturbed than for low temperature depositions. From the electronic structure point of view it seems also, that the compound formed at 530 K deposition is close to MBE grown In1xMnxAs sample, deduced from the valence band spectra. In order to confirm this hypothesis, we have
K. Hricovini et al. / Applied Surface Science 212–213 (2003) 17–25
21
Fig. 2. Valence band photoemission spectra of clean InAs(1 0 0)(4 2)c(8 2) and Mn/InAs(1 0 0) (4 2)c(8 2) for different Mn coverages measured at RT (a) and at 530 K (b) with a photon energy of 84 eV. Vertical bars indicate the position of the Fermi level.
measured EXAFS spectra that give information about local order and about the number of near neighbours. 3.3. Surface EXAFS In the surface EXAFS experiments we have measured 1, 2 and 4 Ml of Mn deposited at 530 K. In order to be sensitive to the environment of the Mn atoms, EXAFS spectra have been recorded at the Mn K-edge at 77 K. The raw spectra for 1 and 4 Ml after background subtraction are shown in Fig. 3. One can immediately notice that the general shape of the EXAFS oscillations is different (in the low k region)
in these two films, consequently these two samples have different structures. The Fourier Transform (FT) of the EXAFS oscillations gives a series of peaks corresponding to the different shells of neighbours of the excited atom [12]. The FT of our spectra after deposition of 1, 2 and 4 Ml of Mn are shown in Fig. 4. The crystallographic parameters R (first nearest neighbour distance), s2 (mean square relative displacement, s2 contains two contributions: the thermal vibration, so called the Debye–Waller factor and the static disorder) and N (apparent number of first nearest neighbours) were obtained by fitting the experimental data with the simple EXAFS formula using the single-scattering formalism.
22
K. Hricovini et al. / Applied Surface Science 212–213 (2003) 17–25
Fig. 3. EXAFS spectra after background subtraction of Mn/InAs for 1 and 4 Ml of manganese recorded at the Mn K-edge. The overall shapes of the EXAFS oscillations are different in these two films.
The most important information for our study is ˚ ) that contained in the main peak (close to R 2:5 A shows the distances to the first nearest neighbour shell. Peaks located at higher atomic distances R (clearly appearing namely for the 1 Ml sample) are due to the
more distant neighbours shells and we will not analyse them. When Mn atoms diffuse into the InAs crystal, they can occupy two possible crystallographic sites and form: (i) In1xMnxAs compounds by substitution of In
Fig. 4. Fourier transforms of the EXAFS spectra of Mn/InAs for three different Mn amounts. The vertical bar indicates the first nearest neighbour positions in the sample with 1 Ml of Mn. This distance is obviously longer than for 4 Ml of Mn.
K. Hricovini et al. / Applied Surface Science 212–213 (2003) 17–25 Table 1 Structural parameters of MnAs and InAs Central atom
First neighbour shell
N
˚) R (A
Mn in MnAs
As Mn
6 2
2.57 2.85
In in In1xMnxAs
As
4
2.61
N is the apparent number of first nearest neighbours and R the first nearest neighbour distance.
atoms (in-site-substitution structure) and maintain in such a way the zinc blend structure of InAs, or (ii) MnAs compounds with a stable hexagonal compact structure. In Table 1, we display the crystallographic parameters of the MnAs structure (NiAs structure, hexagonal compact) and InAs zinc blende structure. We can see from Table 1 that the first nearest neigh˚ ) is longer than in MnAs bour distance in InAs (2.61 A ˚ (2.57 A). The FTs of the EXAFS spectra in Fig. 4 show that the distance of the first nearest neighbour decreases with increasing an amount of Mn atoms. Quantitative analysis of the spectra (see Table 2) indicates that for low Mn deposition (1 Ml) manganese atoms replace In atoms in the InAs crystal, whereas for larger amounts the MnAs compound is preferentially formed: a better fit is obtained supposing the InAs-like structure for samples with 1 and 2 Ml of Mn, whereas the simulation is much better when a MnAs structure is used in the case of 4 Ml. 3.4. X-ray magnetic circular dichroism Now, having determined the structure of the Mn þ InAs system, we can study its magnetic properties by XMCD. As already pointed out in the experimental section, the XMCD is an appropriate technique to investigate the magnetic moment of Mn atoms. Fig. 5 shows two X-ray absorption spectra (XAS) measured Table 2 Structural parameter from least square fits of the first shell of neighbours for different samples of Mn/InAs(0 0 1) Thickness of Mn layer (Ml)
N
1 2 4
4 4 6 2
˚) R (A Mn neighbours Mn neighbours Mn neighbours As neighbours
2.64 2.62 2.55 2.83
23
in applied magnetic fields (H ¼ 5 T) parallel to the helicity of the soft X-rays impinging on the sample at the angle of 458. The high value of applied field ensures that the sample magnetisation is saturated. In an independent experiment [13] (not shown here), we determined that the sample magnetisation remains saturated in magnetic fields down to about 0.5 T. Before studying the magnetism, we consider only the L2,3 absorption of X-rays shown in Fig. 5 (upper panel) to estimate the dilution of Mn in the InAs. The relevant parameter in this case is the ratio between the L3 maximum intensity s(L3) and the background B intensity s(B) determined for the highest photon energies in the spectra: sðL3 Þ=sðBÞ 1. Here sðL3 Þ=sðBÞ 1 ¼ 0:83, whereas for 0.5 Ml Mn/ Cu(1 0 0) sðL3 Þ=sðBÞ 1 ¼ 0:243 [14] and for 1 Ml Mn/Fe(1 0 0) sðL3 Þ=sðBÞ 1 ¼ 0:5 [15], i.e. s0 ðL3 Þ=s0 ðBÞ 1 0:5 for 1 Ml of Mn independent of the substrate used (provided no absorption edge lies just before the L2,3 absorption edges). The probing depth of the total electron yield technique we used in ˚ [16] which is absorption measurements is l 20 A equivalent to about 14 Ml of Mn. The effective Mn concentration in our case is thus estimated as ½sðL3 Þ=sðBÞ 1 =½ðs0 ðL3 Þ=s0 ðBÞ 1Þl 7:2% in a thickness range of l. However, as deduced from the photoemission experiments, Mn can migrate deeper into the substrate, especially for depositions at higher temperatures (530 K). The difference between two XAS spectra (Fig. 5, bottom panel) represents the dichroism signal, which is proportional to the net magnetic moment. Theory predicts that spin and orbital magnetic moments can be derived by using XMCD sum rules [9]. From our dichroic signal, using the sum rules, we find total effective magnetic moment of about 0.7 mB per Mn atom. It is often argued that the XMCD sum rules are not valid in the case of Mn at the L2,3 edges, due to the spin–orbit splitting which is too small to avoid mixing of the j levels [10]. It was suggested, that the apparent spin moment obtained from the integrated intensities by applying the sum rule should be multiplied by 1.5 to give the real value of the spin moment. XMCD measurements will give information also about the orbital magnetic moment. Our results show very low values of the orbital moment (about 0.04 mB), comparable to bulk ferromagnets like Fe, Co and Ni. So, in our sample the total apparent magnetic moment is
24
K. Hricovini et al. / Applied Surface Science 212–213 (2003) 17–25
Fig. 5. Mn L2,3 X-ray absorption and X-ray magnetic circular dichroism recorded at a temperature of 4 K with a magnetic field of H ¼ 5 T. The geometry of the experiment is shown in the insert.
determined mostly by the spin moment and if we apply the correction factor, the total magnetic moment will be about 1 mB. Fig. 6 represents the temperature dependence of the effective magnetic moment of Mn measured in an external magnetic field of 5 T. It follows the Brillouin
function and it drops to zero at a temperature of about 150 K. From the XAS spectrum, one can also estimate the individual Mn moment calculating the L3/L2 intensity ratio [10]. The absorption spectrum measured on our sample with linearly polarised light (not shown) gives a value of the individual Mn moments of about 2.4 mB (note the theoretical result of 4.2 mB for the individual Mn moment [7]). Taking into account the effective magnetic moment found from the sum rules, we can conclude that roughly one-third of Mn spins are ferromagnetically coupled. We suggest that one of the origins of this discrepancy between individual and effective moments is an irregular spatial distribution of Mn atoms due to diffusion processes. Another reason is probably that the concentration of Mn atoms is not at its optimal value.
4. Conclusion
Fig. 6. Temperature dependence of the Mn total magnetic moment measured in a magnetic field of H ¼ 5 T.
Small amounts (typically 1 Ml) of Mn atoms diluted into InAs(1 0 0) by simple deposition at high temperature (T ¼ 530 K) form In1xMnxAs compounds in
K. Hricovini et al. / Applied Surface Science 212–213 (2003) 17–25
which In atoms are substituted by Mn atoms and thus the zinc-blende structure of InAs is conserved. For higher deposition amounts of Mn atoms, MnAs compounds with hexagonal packed structure are formed. In1xMnxAs presents ferromagnetic ordering and the magnetization drops to zero at TC ¼ 150 K when measured in a magnetic field of 5 T. The Mn effective moment determined from the sum rules is found to be about 0.7 mB per atom and the atomic Mn moment is determined to be about 2.4 mB. We suggest that the technique of Mn deposition at high temperature of the InAs substrate is suitable for the preparation of the In1xMnxAs diluted magnetic semiconductor. More work is needed to optimise the Mn concentration and the homogeneity of the spatial distribution of the Mn atoms in order to improve the magnetism of In1xMnxAs. References [1] H. Ohno, Science 281 (1998) 95. [2] T. Dietl, H. Ohno, F. Matsukura, J. Cibert, D. Ferrand, Science 287 (2000) 101.
25
[3] H. Munekata, H. Ohno, S. von Molnar, A. Segmu¨ ller, L.L. Chang, L. Esaki, Phys. Rev. Lett. 63 (1989) 1849. [4] A. Krol, Y.L. Soo, S. Huang, Z.N. ming, Y.A. Kao, H. Munekate, L.L. Chang, Phys. Rev. B 47 (1993) 7187. [5] H. Ofuchi, T. Kubo, M. Tabuchi, Y. Taked, F. Matsukura, S.P. Guo, A. Shen, H. Ohno, J. Appl. Phys. 89 (2001) 66. [6] J. Okabayashi, T. Mizokawa, D.D. Sharma, A. Fujimori, Phys. Rev. B 65 (2002) R161203 (and references therein). [7] H. Akai, Phys. Rev. Lett. 81 (1998) 3002. [8] H. Ofuchi, et al., J. Appl. Phys. 89 (2001) 66. [9] P. Carra, B.T. Thole, M. Altarelli, X. Wang, Phys. Rev. Lett. 70 (1993) 694. [10] H.A. Du¨ rr, G. van der Laan, D. Spanke, F.U. Hillebrecht, N.B. Brookes, Phys. Rev. B 56 (1997) 8156. [11] P. De Padova, C. Quaresima, P. Perfetti, R. Larciprete, R. Brochier, C. Richter, V. Ilakovac, P. Bencok, C. Teodorescu, V.Y. Aristov, R.L. Johnson, K. Hricovini, Surf. Sci. 482–485 (2001) 587. [12] D.E. Sayers, E.A. Stern, F.W. Lythle, Phys. Rev. Lett. 27 (1971) 1204. [13] C.M. Teodorescu, K. Hricovini, F. Chevrier, C. Richter, V. Ilakovac, O. Heckmann, R. Brochier, L. Lechevallier, P. De Padova, Phys. Rev. B., in press. [14] Y. Huttel, C.M. Teodorescu, F. Bertran, G. Krill, Phys. Rev. B 64 (2001) 094405. [15] S. Andrieu, E. Foy, H. Ficher, M. Alnot, F. Chevrier, G. Krill, M. Piecuch, Phys. Rev. B58 (1998) 8210. [16] W.L. O’Brien, B.P. Tonner, Phys. Rev. B50 (1994) 12672.
Atomic structure and magnetic properties of Mn on InAs(1 0 0) K. Hricovinia,b,*, P. De Padovac, C. Quaresimac, P. Perfettic, R. Brochiera, C. Richtera, V. Ilakovaca, O. Heckmanna, L. Lechevalliera, P. Bencokd, P. Le Fevreb, C. Teodorescub a
LMPS, Universite´ de Cergy-Pontoise, Neuville/Oise, 95031 Cergy-Pontoise, France b LURE, Bat. 209 d., 91898 Orsay, France c CNR-ISM, Via Fosso del Cavaliere, 00133 Rome, Italy d ESRF, Grenoble, France
Abstract We have studied Mn/InAs(0 0 1)(4 2)c(8 2) interfaces by high-resolution core-level spectroscopy, surface EXAFS and by X-ray magnetic circular dichroism (XMCD). The growth of manganese is found to depend strongly on the temperature of InAs. At low temperatures (room and liquid nitrogen temperature (LNT)), Mn atoms form the compound MnAs in the surface region leaving metallic In islands on the surface. Manganese deposition at 530 K results in strong diffusion of Mn into the substrate followed by formation of the compound In1xMnxAs in which Mn atoms are substituted in indium-sites. We show by XMCD that In1xMnxAs is ferromagnetic with a relatively high value of an average magnetic moment per Mn atom, exceeding 0.7 mB at low temperatures and with an individual magnetic moment of about 2.4 mB. The temperature dependence of the Mn magnetic moment follows the Brillouin function and it drops to zero at temperature of about 150 K for measurements in an external magnetic field of 5 T. # 2003 Published by Elsevier Science B.V. PACS: 61.10.Ht; 68.35.Fx; 75.70.Pa; 75.30.Et Keywords: Indium arsenide; Manganese; Diluted magnetic semiconductor; XMCD
1. Introduction A new category of semiconductors, called diluted magnetic semiconductors (DMS), are alloys between a nonmagnetic semiconductor and a magnetic element (manganese in most cases). An important property of these materials is that the carrier density can be controlled over a wide range between n- and p-types. This opens up the possibility to control magnetic *
Corresponding author. Tel.: þ33-1-3425-7029; fax: þ33-1-3425-7071. E-mail address: [email protected] (K. Hricovini).
properties simply by changing the carrier density. This behaviour is generally called ‘‘carrier-induced ferromagnetism’’ because hole carriers introduced into the system mediate the ferromagnetic coupling between Mn ions, although its microscopic mechanism has been controversial until now. So, in such systems, both, the charge and the spin of the electron can potentially be used to create new types of semiconductor devices. III–V based DMS have attracted considerable interest because of their relatively high Curie temperature (Tc 110 K in the case of GaMnAs [1]). Theoretical predictions indicate [2] that Tc can be raised above
0169-4332/03/$ – see front matter # 2003 Published by Elsevier Science B.V. doi:10.1016/S0169-4332(03)00013-8
18
K. Hricovini et al. / Applied Surface Science 212–213 (2003) 17–25
300 K for the materials containing a high concentration of magnetic ions or consisting of lighter elements. In the case of In1xMnxAs, the Curie temperature reported to date is relatively low Tc < 55 K. The major obstacle in making III–V semiconductors magnetic has been the low solubility of magnetic elements (such as Mn) in the compounds. Introduction of a large controllable amount of manganese atoms into the III–V semiconductor InAs was first accomplished in 1989 by using molecular beam epitaxy (MBE) (e.g. see review [1]). Structural studies of MBE prepared In1xMnxAs [3–5] done by EXAFS and fluorescence have revealed the local order around Mn atoms. It was found that in the high-growth-temperature samples (temperature of the substrate Ts ¼ 550 K) Mn atoms are primarily incorporated in the form of MnAs clusters with NiAs structure. No significant disorder is observed. In the low-growth-temperature samples (Ts ¼ 490 K), the majority of Mn atoms form small, disordered, sixfold-coordinated centers with As. According to studies in [4], only a very small fraction of Mn atoms may substitute for In in the zinc-blend InAs structure. On the contrary to the thin layers, in case of growth of quantum dots [5] the majority of the Mn atoms were substituted in the indium-sites (In-sites) of InAs. It is believed that the growth in a 3D mode relaxed the strain on (In, Mn)As, which may have suppressed the formation of NiAs-type MnAs. The electronic structure of In1xMnxAs is interesting because it has been reported that In1xMnxAs grown on GaSb substrates exhibits field-induced ferromagnetism [6]. Unlike Ga1xMnxAs, the impurity-band-like states near the top of the valence band (VB) have not been observed by angle-resolved photoemission spectroscopy, in spite of the fact that in both systems Mn 3d states are in the Mn2þ configuration. The electronic structure of the diluted magnetic semiconductor (In, Mn)As has recently been calculated by Akai [7] by using the Korringa–Kohn–Rostoker coherent-potential and local density approximation (KKR–CPA–LDA). The calculations show that the ground state of (In0.94Mn00.06)As is ferromagnetically stabilized by double exchange and the local Mn magnetic moment has a value of about 4.2 mB. On the contrary, (In0.90Mn0.06As0.04)As has no magnetic order and the density of states at EF is primarily due to
impurity bands composed of As electrons. An important fact is that the ferromagnetic state is half-metallic, namely, the Fermi surface exists only in the majority spin band. In this paper, we report photoemission core-level spectroscopy, surface EXAFS and X-ray magnetic circular dichroism (XMCD) studies on In1xMnxAs samples prepared by an alternative technique, which consists in manganese deposition on the semiconductor substrate at higher temperature. We show that the diffusion of Mn in the substrate can be controlled and we observe by XMCD ferromagnetic ordering of manganese atoms diluted into InAs(1 0 0). The temperature of the substrate was chosen to be high enough to allow the diffusion of Mn atoms, without being so high that segregation of MnAs clusters could occur [8]. We found that in the literature there are still controversies concerning the magnetism of MnInAs. XMCD is an appropriate technique to investigate individual and macroscopic magnetic moments of Mn atoms: it is element sensitive and it provides quantitatively the local spin and orbital magnetic moments through the application of the sum rules [9]. We found a Mn macroscopic moment of about 0.7 mB, and for the individual Mn moment we found from the L3/L2 intensity ratio [10] a value of about 2.4 mB (theoretical result 4.2 mB [7]).
2. Experimental Photoemission core-level spectroscopy experiments were carried out in ultra high vacuum conditions (base pressure 8 1011 mbar) at the synchrotron radiation beamline VUV of ELETTRA (Trieste, Italy). All spectra reported were collected at normal emission. Undoped InAs(0 0 1) (3.5 O cm) single crystals were used for all the measurements. First, the Interminated InAs(0 0 1) clean surfaces were prepared after several cycles of Ar-ion bombardment at room temperature and subsequent heating to a temperature of 620 K, followed by some additional annealing at temperatures of 700 K. The study of LEED figures showed that we achieve a high-quality In-terminated InAs(0 0 1)c(8 2) reconstructed surface. Mn was deposited from a Knudsen cell evaporator and calibrated with a quartz monitor.
K. Hricovini et al. / Applied Surface Science 212–213 (2003) 17–25
Photoemission spectra were measured after cooling the sample to 150 K using an angle-resolved hemispherical electron energy analyser (acceptance 28), and the total energy resolution was better than 50 meV. The Fermi level was measured on a gold foil in electrical contact with the sample. Surface EXAFS and XMCD experiments were performed at the synchrotron radiation laboratory LURE in Orsay, France. We used a molecular beam epitaxy chamber with standard surface preparation techniques, LEED, RHEED and Auger spectroscopy facilities. For XMCD measurements, the MBE chamber was connected to the experimental chamber equipped with a liquid He cooled sample holder and a super conducting magnet generating magnetic fields up to 5 T. The whole system was installed on the SU 23 beamline of the Super ACO storage ring, which provides photons in the energy range from 80 to 1000 eV with about 60% degree of circular polarisation. The XMCD spectra were performed by recording the total electron yield signal from the sample.
3. Results and discussion 3.1. Core-level spectroscopy By photoemission core-level spectroscopy of In 4d and As 3d, we have first studied the growth mode of Mn on InAs(1 0 0)c(8 2) at three different temperatures: 530 K, room temperature and liquid nitrogen temperature (LNT). We did not find any detectable signal from Mn 3p core-levels for deposition at any temperature. In Fig. 1, we display In 4d core-level spectra for different Mn coverage deposited at RT (Fig. 1a) and at 530 K (Fig. 1b). The growth mode, as seen through core-level photoemission spectra, is very similar at both temperatures, LNT and RT. So, in the following, we will mostly compare Mn growth at RT and at 530 K. The decomposition of the direct photoemission peak In 4d from a clean InAs(1 0 0)c(8 2) surface is now well-established [11]: in addition to the bulk component (the component B with the highest spectral intensity shown in Fig. 1a and b, upper panel), there is a clear evidence of two surface peaks S1 and S2. S1 having lower kinetic energy (0.31 eV) than the bulk peak is attributed to surface In atoms. The peak S2 (þ0.25 eV from the bulk peak) corresponds to In atoms of self
19
organized In chains running in the (1 1 0) direction of the InAs(1 0 0)(4 2)c(8 2) surface [11]. Dramatic changes appear in the core-level spectra after deposition even of the very first thin Mn layers. For both substrate temperatures a new component lying at higher kinetic energy then the bulk one (þ0.75 eV) appears in In 4d spectra (black colour component S3 in the middle and in the lower panel of Fig. 1; in order to follow in a clear way the evolution of the spectra, in the middle and bottom panels we do not show the bulk, S1 and S2 components any more, only S3 is plotted schematically). S3 is attributed to free In atoms (i.e. atoms forming In–In bonds, rather than In–As ones) that are the result of the reaction between Mn and As. This is immediately evident when one compares the enthalpies of compound formation: 8 and 50 kJ/mol in the case of MnIn and MnAs, respectively. This is why, from the energy balance point of view, the MnAs compound formation is much more probable. Now, it is interesting to determine the dynamics of the reaction and of the Mn atom diffusion as a function of substrate temperature. In spite of the fact, that deposited quantities of Mn are not exactly the same for RT and 530 K depositions, from Fig. 1 it is clear that the amount of free In atoms on the surface is much larger in the case of RT depositions. After deposition of 3.5 Ml at RT the In 4d photoemission signal becomes dominant on the surface (as seen in the bottom panel of Fig. 1a) and As 3d photoemission is strongly reduced (not shown). For 7 Ml deposited at liquid nitrogen temperature, no As 3d signal is detectable and the surface is completely covered by In. For high Mn coverages (as shown in Fig. 1a, bottom panel) In 4d exhibits Doniah–Sunjic function-like shape, typical for a metallic surface. The growth of Mn at 530 K shows only very slight increase of In 4d signal due to free In atoms, so it appears that the Mn atoms can penetrate inside the bulk (Fig. 1b). 3.2. Valence band spectroscopy This situation is further confirmed by the valence band (VB) spectra in Fig. 2. The spectra at RT depositions (Fig. 2a) are difficult to analyse, because they are a superimposition of signals from free In atoms and Mn þ InAs compounds. As seen in Fig. 2a, the metallicity of the surface is confirmed by valence band spectra: electron emission from Fermi level is
20
K. Hricovini et al. / Applied Surface Science 212–213 (2003) 17–25
Fig. 1. In 4d core-level photemission spectra of clean InAs(1 0 0)(4 2)c(8 2) and Mn/InAs(1 0 0)(4 2)c(8 2) for different Mn coverages. The spectra were measured at RT (a) and at 530 K (b) with a photon energy of 84 eV. Vertical bars indicate energy position of the component attributed to free indium atoms on the surface. See the text for other details.
observed after 0.5 Ml of Mn at RT deposition. On the contrary, the VB spectra measured after the deposition at 530 K are comparable to spectra measured on MBE grown In1xMnxAs samples [6]. It is interesting to note that the VB spectrum does not exhibit important changes with increasing Mn amount at 530 K (Fig. 2b), bottom panel). For 7 Ml, however, an emission from the Fermi level is observed, which is in agreement with increasing intensity of the S3 component in In 4d spectrum (Fig. 1b, bottom panel). From the photoemission spectra we can draw a preliminary image of Mn growth on InAs(0 0 1)(4 2)c(8 2) at both low (RT and LNT) and high (530 K)
temperatures. In both cases, the growth is governed by strong chemical reaction between Mn and As atoms. Consequently, an indium metallic layer is formed on the surface for low temperature deposition. For deposition at 530 K, however, we find an important difference in growth mode. It appears that at 530 K Mn atoms diffuse rapidly into the substrate leaving the surface much less perturbed than for low temperature depositions. From the electronic structure point of view it seems also, that the compound formed at 530 K deposition is close to MBE grown In1xMnxAs sample, deduced from the valence band spectra. In order to confirm this hypothesis, we have
K. Hricovini et al. / Applied Surface Science 212–213 (2003) 17–25
21
Fig. 2. Valence band photoemission spectra of clean InAs(1 0 0)(4 2)c(8 2) and Mn/InAs(1 0 0) (4 2)c(8 2) for different Mn coverages measured at RT (a) and at 530 K (b) with a photon energy of 84 eV. Vertical bars indicate the position of the Fermi level.
measured EXAFS spectra that give information about local order and about the number of near neighbours. 3.3. Surface EXAFS In the surface EXAFS experiments we have measured 1, 2 and 4 Ml of Mn deposited at 530 K. In order to be sensitive to the environment of the Mn atoms, EXAFS spectra have been recorded at the Mn K-edge at 77 K. The raw spectra for 1 and 4 Ml after background subtraction are shown in Fig. 3. One can immediately notice that the general shape of the EXAFS oscillations is different (in the low k region)
in these two films, consequently these two samples have different structures. The Fourier Transform (FT) of the EXAFS oscillations gives a series of peaks corresponding to the different shells of neighbours of the excited atom [12]. The FT of our spectra after deposition of 1, 2 and 4 Ml of Mn are shown in Fig. 4. The crystallographic parameters R (first nearest neighbour distance), s2 (mean square relative displacement, s2 contains two contributions: the thermal vibration, so called the Debye–Waller factor and the static disorder) and N (apparent number of first nearest neighbours) were obtained by fitting the experimental data with the simple EXAFS formula using the single-scattering formalism.
22
K. Hricovini et al. / Applied Surface Science 212–213 (2003) 17–25
Fig. 3. EXAFS spectra after background subtraction of Mn/InAs for 1 and 4 Ml of manganese recorded at the Mn K-edge. The overall shapes of the EXAFS oscillations are different in these two films.
The most important information for our study is ˚ ) that contained in the main peak (close to R 2:5 A shows the distances to the first nearest neighbour shell. Peaks located at higher atomic distances R (clearly appearing namely for the 1 Ml sample) are due to the
more distant neighbours shells and we will not analyse them. When Mn atoms diffuse into the InAs crystal, they can occupy two possible crystallographic sites and form: (i) In1xMnxAs compounds by substitution of In
Fig. 4. Fourier transforms of the EXAFS spectra of Mn/InAs for three different Mn amounts. The vertical bar indicates the first nearest neighbour positions in the sample with 1 Ml of Mn. This distance is obviously longer than for 4 Ml of Mn.
K. Hricovini et al. / Applied Surface Science 212–213 (2003) 17–25 Table 1 Structural parameters of MnAs and InAs Central atom
First neighbour shell
N
˚) R (A
Mn in MnAs
As Mn
6 2
2.57 2.85
In in In1xMnxAs
As
4
2.61
N is the apparent number of first nearest neighbours and R the first nearest neighbour distance.
atoms (in-site-substitution structure) and maintain in such a way the zinc blend structure of InAs, or (ii) MnAs compounds with a stable hexagonal compact structure. In Table 1, we display the crystallographic parameters of the MnAs structure (NiAs structure, hexagonal compact) and InAs zinc blende structure. We can see from Table 1 that the first nearest neigh˚ ) is longer than in MnAs bour distance in InAs (2.61 A ˚ (2.57 A). The FTs of the EXAFS spectra in Fig. 4 show that the distance of the first nearest neighbour decreases with increasing an amount of Mn atoms. Quantitative analysis of the spectra (see Table 2) indicates that for low Mn deposition (1 Ml) manganese atoms replace In atoms in the InAs crystal, whereas for larger amounts the MnAs compound is preferentially formed: a better fit is obtained supposing the InAs-like structure for samples with 1 and 2 Ml of Mn, whereas the simulation is much better when a MnAs structure is used in the case of 4 Ml. 3.4. X-ray magnetic circular dichroism Now, having determined the structure of the Mn þ InAs system, we can study its magnetic properties by XMCD. As already pointed out in the experimental section, the XMCD is an appropriate technique to investigate the magnetic moment of Mn atoms. Fig. 5 shows two X-ray absorption spectra (XAS) measured Table 2 Structural parameter from least square fits of the first shell of neighbours for different samples of Mn/InAs(0 0 1) Thickness of Mn layer (Ml)
N
1 2 4
4 4 6 2
˚) R (A Mn neighbours Mn neighbours Mn neighbours As neighbours
2.64 2.62 2.55 2.83
23
in applied magnetic fields (H ¼ 5 T) parallel to the helicity of the soft X-rays impinging on the sample at the angle of 458. The high value of applied field ensures that the sample magnetisation is saturated. In an independent experiment [13] (not shown here), we determined that the sample magnetisation remains saturated in magnetic fields down to about 0.5 T. Before studying the magnetism, we consider only the L2,3 absorption of X-rays shown in Fig. 5 (upper panel) to estimate the dilution of Mn in the InAs. The relevant parameter in this case is the ratio between the L3 maximum intensity s(L3) and the background B intensity s(B) determined for the highest photon energies in the spectra: sðL3 Þ=sðBÞ 1. Here sðL3 Þ=sðBÞ 1 ¼ 0:83, whereas for 0.5 Ml Mn/ Cu(1 0 0) sðL3 Þ=sðBÞ 1 ¼ 0:243 [14] and for 1 Ml Mn/Fe(1 0 0) sðL3 Þ=sðBÞ 1 ¼ 0:5 [15], i.e. s0 ðL3 Þ=s0 ðBÞ 1 0:5 for 1 Ml of Mn independent of the substrate used (provided no absorption edge lies just before the L2,3 absorption edges). The probing depth of the total electron yield technique we used in ˚ [16] which is absorption measurements is l 20 A equivalent to about 14 Ml of Mn. The effective Mn concentration in our case is thus estimated as ½sðL3 Þ=sðBÞ 1 =½ðs0 ðL3 Þ=s0 ðBÞ 1Þl 7:2% in a thickness range of l. However, as deduced from the photoemission experiments, Mn can migrate deeper into the substrate, especially for depositions at higher temperatures (530 K). The difference between two XAS spectra (Fig. 5, bottom panel) represents the dichroism signal, which is proportional to the net magnetic moment. Theory predicts that spin and orbital magnetic moments can be derived by using XMCD sum rules [9]. From our dichroic signal, using the sum rules, we find total effective magnetic moment of about 0.7 mB per Mn atom. It is often argued that the XMCD sum rules are not valid in the case of Mn at the L2,3 edges, due to the spin–orbit splitting which is too small to avoid mixing of the j levels [10]. It was suggested, that the apparent spin moment obtained from the integrated intensities by applying the sum rule should be multiplied by 1.5 to give the real value of the spin moment. XMCD measurements will give information also about the orbital magnetic moment. Our results show very low values of the orbital moment (about 0.04 mB), comparable to bulk ferromagnets like Fe, Co and Ni. So, in our sample the total apparent magnetic moment is
24
K. Hricovini et al. / Applied Surface Science 212–213 (2003) 17–25
Fig. 5. Mn L2,3 X-ray absorption and X-ray magnetic circular dichroism recorded at a temperature of 4 K with a magnetic field of H ¼ 5 T. The geometry of the experiment is shown in the insert.
determined mostly by the spin moment and if we apply the correction factor, the total magnetic moment will be about 1 mB. Fig. 6 represents the temperature dependence of the effective magnetic moment of Mn measured in an external magnetic field of 5 T. It follows the Brillouin
function and it drops to zero at a temperature of about 150 K. From the XAS spectrum, one can also estimate the individual Mn moment calculating the L3/L2 intensity ratio [10]. The absorption spectrum measured on our sample with linearly polarised light (not shown) gives a value of the individual Mn moments of about 2.4 mB (note the theoretical result of 4.2 mB for the individual Mn moment [7]). Taking into account the effective magnetic moment found from the sum rules, we can conclude that roughly one-third of Mn spins are ferromagnetically coupled. We suggest that one of the origins of this discrepancy between individual and effective moments is an irregular spatial distribution of Mn atoms due to diffusion processes. Another reason is probably that the concentration of Mn atoms is not at its optimal value.
4. Conclusion
Fig. 6. Temperature dependence of the Mn total magnetic moment measured in a magnetic field of H ¼ 5 T.
Small amounts (typically 1 Ml) of Mn atoms diluted into InAs(1 0 0) by simple deposition at high temperature (T ¼ 530 K) form In1xMnxAs compounds in
K. Hricovini et al. / Applied Surface Science 212–213 (2003) 17–25
which In atoms are substituted by Mn atoms and thus the zinc-blende structure of InAs is conserved. For higher deposition amounts of Mn atoms, MnAs compounds with hexagonal packed structure are formed. In1xMnxAs presents ferromagnetic ordering and the magnetization drops to zero at TC ¼ 150 K when measured in a magnetic field of 5 T. The Mn effective moment determined from the sum rules is found to be about 0.7 mB per atom and the atomic Mn moment is determined to be about 2.4 mB. We suggest that the technique of Mn deposition at high temperature of the InAs substrate is suitable for the preparation of the In1xMnxAs diluted magnetic semiconductor. More work is needed to optimise the Mn concentration and the homogeneity of the spatial distribution of the Mn atoms in order to improve the magnetism of In1xMnxAs. References [1] H. Ohno, Science 281 (1998) 95. [2] T. Dietl, H. Ohno, F. Matsukura, J. Cibert, D. Ferrand, Science 287 (2000) 101.
25
[3] H. Munekata, H. Ohno, S. von Molnar, A. Segmu¨ ller, L.L. Chang, L. Esaki, Phys. Rev. Lett. 63 (1989) 1849. [4] A. Krol, Y.L. Soo, S. Huang, Z.N. ming, Y.A. Kao, H. Munekate, L.L. Chang, Phys. Rev. B 47 (1993) 7187. [5] H. Ofuchi, T. Kubo, M. Tabuchi, Y. Taked, F. Matsukura, S.P. Guo, A. Shen, H. Ohno, J. Appl. Phys. 89 (2001) 66. [6] J. Okabayashi, T. Mizokawa, D.D. Sharma, A. Fujimori, Phys. Rev. B 65 (2002) R161203 (and references therein). [7] H. Akai, Phys. Rev. Lett. 81 (1998) 3002. [8] H. Ofuchi, et al., J. Appl. Phys. 89 (2001) 66. [9] P. Carra, B.T. Thole, M. Altarelli, X. Wang, Phys. Rev. Lett. 70 (1993) 694. [10] H.A. Du¨ rr, G. van der Laan, D. Spanke, F.U. Hillebrecht, N.B. Brookes, Phys. Rev. B 56 (1997) 8156. [11] P. De Padova, C. Quaresima, P. Perfetti, R. Larciprete, R. Brochier, C. Richter, V. Ilakovac, P. Bencok, C. Teodorescu, V.Y. Aristov, R.L. Johnson, K. Hricovini, Surf. Sci. 482–485 (2001) 587. [12] D.E. Sayers, E.A. Stern, F.W. Lythle, Phys. Rev. Lett. 27 (1971) 1204. [13] C.M. Teodorescu, K. Hricovini, F. Chevrier, C. Richter, V. Ilakovac, O. Heckmann, R. Brochier, L. Lechevallier, P. De Padova, Phys. Rev. B., in press. [14] Y. Huttel, C.M. Teodorescu, F. Bertran, G. Krill, Phys. Rev. B 64 (2001) 094405. [15] S. Andrieu, E. Foy, H. Ficher, M. Alnot, F. Chevrier, G. Krill, M. Piecuch, Phys. Rev. B58 (1998) 8210. [16] W.L. O’Brien, B.P. Tonner, Phys. Rev. B50 (1994) 12672.