Nature of magnetic coupling between Mn ions in as-grown Ga Mn As studied by 1−x x x-ray magnetic circular dichroism Y. Takeda,1,∗ M. Kobayashi,2 T. Okane,1 T. Ohkochi,1 J. Okamoto,1,† Y. Saitoh,1 K. Kobayashi,1 H. Yamagami,1 A. Fujimori,2 A. Tanaka,3 J. Okabayashi,4,‡ M. Oshima,4 S. Ohya,5,6 P. N. Hai,5 and M. Tanaka5 1Japan Atomic Energy Agency, Synchrotron Radiation Research Center SPring-8, Mikazuki, Hyogo 679-5148, Japan 2Department of Physics, The University of Tokyo, ,Hongo, Tokyo 113-0033, Japan 3Graduate School of Advanced Sciences of Matter, Hiroshima University, Higashi-Hiroshima 739-8530, Japan 4Department of Applied Chemistry, The University of Tokyo, Hongo, Tokyo 113-8656, Japan 8 5Department of Electronic Engineering, The University of Tokyo, Hongo, Tokyo 113-8656, Japan 0 6PRESTO JST, Kawaguchi, Saitama 331-0012, Japan 0 2 (Dated: February 2, 2008) n The magnetic properties of as-grown Ga1−xMnxAs have been investigated by the systematic a measurements of temperature and magnetic field dependent soft x-ray magnetic circular dichroism J (XMCD). The intrinsic XMCD intensity at high temperatures obeys the Curie-Weiss law, but 9 residualspinmagneticmomentappearsalreadyaround100K,significantlyaboveCurietemperature (TC), suggesting that short-range ferromagnetic correlations are developed above TC. The present ] results also suggest that antiferromagnetic interaction between the substitutional and interstitial i c Mn (Mnint) ions exists and that theamount of the Mnint affects TC. s - PACSnumbers: 75.50.Pp,78.70.Dm,75.25.+z,79.60.Dp l r t m Ga Mn As is a prototypical and most well- we have extended the approach and performed system- 1−x x . t characterized diluted magnetic semiconductor (DMS) atic temperature (T) and magnetic field (H) dependent a m [1]. Because Ga1−xMnxAs is grown under thermal non- XMCD studies in the Mn L2,3 absorption edge region equilibrium conditions, however, it is difficult to avoid of Ga Mn As. We have found that short-range fer- - 1−x x d the formation of various kinds of defects and/or disor- romagnetic correlations develop significantly above TC n der. Infact,Rutherfordbackscattering(RBS)channeling and that AF interaction between the Mn and Mn sub int o experiments for as-grown Ga Mn As samples has is important to understand the magnetic properties of 0.92 0.08 c shown that as many as ∼ 17 % of the total Mn ions Ga Mn As. [ 1−x x may occupy the interstitial sites [2]. It is therefore sup- We prepared two as-grown samples with different Mn 2 posed that antiferromagnetic (AF) interaction between v concentrations; x = 0.042 and 0.078, whose TC was the substitutional Mn (Mn ) ions and interstitial Mn 5 sub ∼ 60 and 40 K, respectively, as determined by an Ar- 5 (Mnint) ions may suppress the magnetic moment [3, 4]. rott plot of the anomalous Hall effect. To avoid sur- 1 In addition, the random substitution of Mn ions may face oxidation, the sample had been covered immedi- 1 create inhomogeneous Mn density distribution, which ately after the growth of Ga Mn As films by cap lay- . 1−x x 1 may lead to the development of ferromagnetic domains ers without exposure to air [As cap/GaAs cap (1nm)/ 0 above Curie temperature (T ) [5]. The characterization C Ga Mn As(20nm)/GaAs(001)]. TheX-rayabsorption 8 of non-ferromagneticMn ions is therefore a clue to iden- 1−x x spectroscopy(XAS)andXMCDmeasurementswereper- 0 tifyhowtheyarerelatedwiththeferromagneticordering : formed at the helical undulator beam line BL23SU of v and eventually to improve the ferromagnetic properties SPring-8 [9]. The XAS spectra were obtained by the Xi of Ga1−xMnxAs samples. However, it has been difficult total-electron yield mode. The measurements were done to extract the above information through conventional r without a surface treatment and H was applied to the a magnetizationmeasurementduetothelargediamagnetic sample perpendicular to the film surface. responseofthesubstrateandtheunavoidablemixtureof Figures1(a)and1(b)showtheXASspectra(µ+ and magnetic impurities. µ−) in the photon energy region of the Mn L absorp- 3 X-ray magnetic circular dichroism (XMCD), which is tionedge andthe correspondingXMCD spectra,defined an element specific magnetic probe, is a powerful tech- as µ+ −µ−, at T = 20 K and H =0.5 T for x = 0.042 nique to address the above issues. So far, several results and 0.078. Here, µ+ (µ−) refers to the absorption coef- of XMCD measurements on Ga Mn As have been re- ficient for the photon helicity parallel (anti-parallel) to 1−x x ported [6, 7, 8]. From H dependent XMCD studies, the Mn 3d majority spin direction. The XAS spectra the enhancement of XMCD intensity by post-annealing for both Mn concentrations have five structures labeled implies AF interaction between the Mn and Mn as a, b, c, d and e. The average XAS spectra [defined sub int ions [8]. In the present study, in order to character- by (µ+ + µ−)/2] have been normalized to 1 at struc- ize the magnetic behaviors of the Mn and Mn , ture b. The intensity ratio c/b is very different between sub int 2 ntensity (arb. units) --1100000.......5050052 MTTH x((X =nT =a==M C 2 L20) 0C0~003..5.60 D1KK 04TaT K2a) b c0d.0d0su.5bTµtr a+c t-eeeµµ ( d µµ+- ( -+X ( +XAd MSµ)-C)/D2 ) MTHTXx((n T =M==b=C L 20 C0)2 ~030..05D40 .K1 07TK TK8aa) bc 0.d0d0su.5bTtr aµc+teee d-µµ ( µµ+- ( +-X (+AXd SµM)-C)/2D ) Intensity (arb. units)-11000.....50505 TH(X a==A )20S0.5 K T x(T =rEICTHXna 0x t~=Mw=r.t06 ri n2i6Cd40n0s 2aKsT iDcKti)ca TH(X =b=A 20)S0.5 K T x( T rEITHX=naCx tw =M0=rt~ri. n042i6dCns700 asTi8K DctKica) IntXrinMsCicD c Taxxot ===mXH A200 p=0..S(n00 6cK47e 28Tn)t --00110.....05005Intensity (arb. units) I 0.5T 0.5T --00..64 (c) 1246TTTT b c--00..42638 b64c0 1246TTTT642 (d) 1246TTTT b c---000...321638 b64c0 1246TTTT642 635 640 Phot6o4n5 Energy (eV6)40 645 P6h4o0ton6 4E5ner6g5y0 (eV6)55 FIG.2: (Coloronline)DecompositionoftheXASandXMCD 636 638 640 642 644 646 636 638 640 642 644 646 Photon Energy (eV) spectra of Ga1−xMnxAs intothe intrinsic and extrinsic com- ponentsforx=0.042(a)andforx=0.078 (b)intheMnL3 FIG. 1: (Color online) Mn L3-edge XAS(µ+, µ− and (µ+ + edge region. Panel (c) shows comparison of the line shapes µ−)/2) andXMCD(µ+ -µ−)spectraofGa1−xMnxAstaken of the intrinsic XAS and XMCD spectra between x = 0.042 and 0.078, normalized to thepeak heights. at T = 20 K and H = 0.5 T for x = 0.042 (a) and x = 0.078 (b). Panels (c) and (d) show the H dependence of the XMCD spectra for x = 0.042 and x = 0.078, respectively. Inset shows the difference XMCD spectra obtained by sub- nant in the XMCD spectrum at low H. The intrinsic tracting the XMCD spectrum at H = 0.5 T. XMCD spectrum was then obtained as (XMCD at each H)−β×(extrinsicXMCDspectrum),whereβ waschosen sothatstructurecvanished. Figures2(a)and2(b)show x = 0.042 and 0.078, indicating that the spectra consist the results of the decomposition of the XAS and XMCD of two overlapping components. Figures 1 (c) and 1 (d) spectra into the intrinsic and extrinsic components for showthe H dependence ofthe XMCD spectra. As H in- x= 0.042 and 0.078, respectively. While the XMCD in- creases,XMCDstructurescorrespondingtostructuresc, tensity is enhanced as H increases and T decreases, the d and e are enhanced, particularly, strongly for the x = lineshapesoftheintrinsicXMCDspectraareunchanged 0.042 sample. One can see this behavior more clearly withH andT. The line shapes ofthe intrinsic XASand in the difference XMCD spectra obtainedby subtracting XMCD spectra for both Mn concentrations thus agree the XMCD spectrum at 0.5 T from the spectra at H = with each other as shown in Fig. 2 (c), indicating that 1, 2, 4 and 6 T as shown in the inset of Fig.1 (c) and the decomposition procedure was valid. (d). Recent XAS and XMCD studies have revealed that Using the intrinsic XAS and XMCD spectra, we have these structures (c, d, e) are ascribed to contamination applied the XMCD sum rules [12, 13], assuming the Mn of out-diffused Mn ions on the surface [7, 10, 11]. The 3d electron number N =5.1 [8], and estimated the spin d differenceintheXASintensityratioc/bisthereforenat- magnetic moment (M ) at T = 20 K and H = 0.5 T to S urallyascribedtothedifferenceintheamountofMnions be M =2.5±0.2and 1.7±0.2 (µ per Mn) for x= 0.042 S B diffused into the cap layer or the surface region during and0.078,respectively. TheseM valuesaremuchlarger S the growth of GaAs on Ga1−xMnxAs. In the following, than those obtained in the early studies on oxidized sur- therefore, we shall neglect those extrinsic signals and fo- faces[6]andcomparabletotherecentonesonetchedsur- cus only on intrinsic signals, particularly structure b, to faces[8],indicatingthatthecaplayerprotectedtheferro- investigate the intrinsic magnetic behavior. magneticpropertiesofGa Mn As. TheratioM /M 1−x x L S In order to extract the intrinsic XAS spectrum, we is estimated to be 0.07 for both concentrations, where assumed that structure b could be ascribed to the in- ML is the value of the orbital magnetic moment, show- trinsic Mn ions as mentioned above. Therefore, we ingthattheintrinsicMnionhasafinite,althoughsmall, first obtained the extrinsic XAS spectrum as (XAS x = ML,probablybecauseofcertaindeviationfromthepure 0.042)−p×(XASx= 0.078),where p was chosen so that Mn2+ (d5) state. structure b vanished. The intrinsic XAS spectrum was The T dependence of M from the XMCD signal for S then obtained as (raw XAS)−q×(extrinsic XAS), where H = 6 T is plotted in Fig. 3 (a). As T decreases, the q was determined so that the line shape of the intrin- XMCD signal is increased monotonously except for the sic XAS spectrum agreed with that obtained from the discontinuityataroundT (∼60Kforx=0.042,∼40K C fluorescence yield measurements [10, 11]. Next, in or- for x = 0.078). This discontinuity probably reflects the der to extract the intrinsic XMCD spectra, we first ob- ferromagnetic ordering which aligns the magnetization tained the extrinsic XMCD spectrum as (XMCD at 6 parallel to the sample surface, the easy axis of magneti- T)−α×(XMCD at 0.5 T), where α was chosen so that zation in the films [14]. It should be noted that M in- S an XMCD structure corresponding to structure b van- creasesmonotonouslyevenwellbelowT asT decreases, C ished by utilizing the fact that the ferromagnetic sig- indicatingthatfullspinpolarizationisnotachievedeven nals and hence the intrinsic signals should be domi- wellbelowT . Forx=0.078,theT dependenceforH = C 3 M (per Mn)µSB 4321 TC T(7C.8 (%4.)2%) 47477.....28288%%%%% ee661xx TTTtt(.. a66 )TT 184020 5 077..881%%0 016TT150 74..82(%%b )66TT 7654321 1/SM MS ( per Mn)µB 5432 (a) 2711 ee0005xxKK00tt..KK x(21T 05=CK0 K 0~.6004 2K) (b) 235611ee000005xxttKKKK00.. 1x(KK2T5 0=0CKK 0~.4007 8K) TC(7.T8C%(4).2T%C)(4.2%) 00ee00e..xx..x0000ttt..47.47 0028028...000((474dc282)) 0221100.3......5050500Bµ (Bµ ( per Mn)H(cid:129)ÝS ( 0 T)MHM(cid:129)Ý 0 0 ext. 0.078 0.20/T S 0 50 T1e0m0pe1r5a0ture2 0(K0)250 0 50 T1e0m0pe1r5a0tur2e0 (0K)250 1 TC(7.8%) 0.10 per Mn)> 0.5 TH 0 0.00 FIG. 3: (Color online) T dependence of the spin magnetic 0 2 4 6 0 2 4 6 0 50 100 150 moment MS. (a) T dependenceof MS for H = 6 T. For x= Magnetic Field (T) Temperature (K) 0.078, results for H = 1 T are also plotted. Open symbols show that of the extrinsic component at H = 6 T. (b) T FIG.4: (Coloronline)H dependenceofMS forx=0.042(a) and for x = 0.078 (b) at several temperatures. Dashed lines dependence of the inverse of MS. Inset shows comparison showfittedstraightlinesabove0.5T.(c)T dependenceofthe between 1 and 6 T for x= 0.078. residual magnetization MS|H→0T (MS for H → 0 T). Open symbols show that of the extrinsic component. (d) T depen- 1 T shows essentially the same behavior as that for 6 T. dence of the slope of the MS-H curve above 0.5 T, i.e., the high-field magnetic susceptibility (∂MS/∂H|H>0.5T). Open Figure 3 (b) shows the inverse of M plotted in Fig 3 S symbols show that of theextrinsic component. (a). The high-temperature part is well described by the Curie-Weiss (CW) law, independent of H as shown in the inset of Fig. 3 (b). developaboveT whenthere is magneticinhomogeneity Figure 4 shows the H dependence of M at several C S [5]. temperatures for x = 0.042 [panel (a)] and 0.078 [panel (b)]. M of the intrinsic component is increased rapidly The suppression of the CW-like increase of S from H = 0.1 to 0.5 T, due to the re-orientation of ∂M /∂H| below T in both samples indi- S H>0.5T C the ferromagnetic moment from the in-plane to out-of- catesthat AFinteractionbetweenthe ferromagneticMn plane directions [14]. Above 0.5 T, M is increased i.e.,Mn andnon-ferromagnetic(orparamagnetic)Mn S sub almost linearly as a function of H. We have plot- such as Mn . The recent H dependent XMCD study int ted the T dependence of M | obtained from the of Ga Mn As shows that ∂M /∂H| becomes S H→0T 1−x x S H>0.5T linear extrapolation of M at high fields to H = 0 small and M | becomes large after post-annealing, S S H→0T T and ∂M /∂H| (µ /T per Mn) (the suscepti- suggesting that the changes are caused by a reduction S H>0.5T B bility of the paramagnetic component) in Fig. 4 (c) of Mn [8]. In the present study, ∂M /∂H| and int S H>0.5T and (d), respectively. For the extrinsic component, M | are smaller for x = 0.078 than for x = 0.042 S H→0T M | is vanishingly small at all temperatures and [Fig. 4 (c) and (d)], suggesting that AF interaction S H→0T ∂M /∂H| is increased as T decreases following becomes stronger for x = 0.078 than that for x = S H>0.5T the CW law, indicating that the extrinsic component is 0.042. This is reasonable because the number of Mn int paramagneticanddecoupledfromthe ferromagnetismof is expected to be larger for larger Mn concentration. the intrinsic component. As for the ferromagnetic com- Assuming that M per the Mn is 5 (µ per Mn) S sub B ponent, M | is steeply increased below ∼ 100 K. and M of the Mn is antiparalleled to that of Mn , S H→0T S int sub i.e., from somewhat above T . The T dependence of the ratio of Mn atoms in the intrinsic component C int M | [Fig. 4 (c)] is correlated with the deviation (R ) is estimated as 0.26 for x = 0.042 and 0.33 for S H→0T int from the CW law below ∼ 100 K [Fig. 3 (b)]. Well x = 0.078 from M | at 20 K. This is consistent S H→0T below T , M | still continues to increase with de- with the result of the RBS experiment [2], which R C S H→0T int creasing T, indicating the inhomogeneous nature of the is estimated as 0.17 for an as-grown sample with T = C ferromagnetism. As for ∂M /∂H| , unlike the ex- 67 K, indicating that T is strongly correlated with S H>0.5T C trinsic component, it saturates around T and is not in- the amount of Mn . We have fitted the susceptibility C int creased as T further decreases. The appearance and in- ∂M /∂H| (µ /T per Mn) of the intrinsic compo- S H=6T B crease of M | between T and ∼100 K [Fig. 4 (c)] nent above 100 K [Fig. 3 (b)] to the CW law with an S H→0T C strongly suggest that short-range ferromagnetic correla- offset, ∂M /∂H| = N C/(T − Θ) + ∂M /∂H| , S H=6T x S 0 tions start to develop and ferromagnetic domains form where C = (gµ )2S(S +1)/3k is the Curie constant, B B beforethelong-rangeorderisestablishedatmacroscopic Θ is the Weiss temperature, ∂M /∂H| is the constant S 0 T . Each ferromagnetic domain may have different fer- offset, N is the number of magnetic Mn ions in the C x romagneticbehaviorduetothespatialdistributionofT sample with Mn concentration x, and g is the g factor. C in the as-grown samples. Those results may correspond Θ is estimated to be 68±5 K for x= 0.042 and 69±3 K to the theoretical predictionthat ferromagneticdomains for x = 0.078. ∂M /∂H| is estimated to be of order of S 0 4 ∼10−3 for both samples. Assuming g = 2, S = 5/2 and † National Synchrotron Radiation Research Center, 101 Θ= 68 K, one obtains N = 0.97 and N = 0.67. Hsin-Ann Road, Hsinshuu Science Park, Hsinchu 30077, 0.042 0.078 This result strongly suggests that most of the intrinsic Taiwan, R.O. C. ‡ Department of Physics, Tokyo Institute of Technology, Mn ions in the x = 0.042 sample participate in the Ookayama, Meguro-ku, Tokyo152-8551, Japan paramagnetism above ∼100 K and the paramagnetism [1] H. Ohno,Science 281, 951 (1998). in the x = 0.078 sample is suppressed even at high [2] K. M. Yu and W. Walukiewicz, T. Wojtowicz, I. temperatures, again implying that the AF interaction is Kuryliszyn,X. Liu,Y. Sasaki, and J. K. Furdyna,Phys. stronger and more influential in the x=0.078 sample. Rev. B 65 201303(R) 2002. In conclusion, we have investigated the T, H and Mn [3] J.BlinowskiandP.Kacman,Phys.Rev.B67121204(R) concentration dependences of the ferromagnetism in as- (2003). [4] J. Ma˘sek and F. Maca, Phys.Rev.B 69 165212 (2004). grownGa Mn AssamplesbyXMCDmeasurementsto 1−x x [5] M. Mayr, G. Alvarez, and E. Dagotto, Phys. Rev. B 65 extract the intrinsic magnetic component. The XMCD 241202(R) 2002. intensity deviates from the CW law below∼100 K,indi- [6] H.Ohldag,V.Solinus,F.U.Hillebrecht,J.B.Goedkoop, cating that the ferromagnetic moment starts to form at M.Finazzi,F.MatsukuraandH.Ohno,Appl.Phys.Lett. ∼100K and that the short-range ferromagnetic correla- 76, 2928 (2000). tionsdevelopsignificantlyaboveT . Thehigh-fieldmag- [7] K. W. Edmonds, N. R. S. Farley, R. P. Campion, C. T. C neticsusceptibilitybecomesT-independentbelowT ,in- Foxon, B. L. Gallagher, T. K. Johal, G. van der Laan, C M. MacKenzie, J. N. Chapman and E. Arenholz , Appl. dicating that the AF interaction between the Mn and sub Phys. Lett. 84, 4065 (2004). Mn ions, which becomes strong as the Mn concentra- int [8] K.W.Edmonds,N.R.S.Farley,T.K.Johal,G.vander tion x increases, plays an important role to determine Laan, R. P. Campion, B. L. Gallagher and C. T. Foxon, the magnetic behavior ofGa1−xMnxAs. In addition, the Phys. Rev.B 71 064418 (2005). amountoftheMnint ionsshouldbestronglyrelatedwith [9] A. Yokoya, T. Sekiguchi, Y. Saitoh, T. Okane, T. TC. The present experimental findings should give valu- Nakatani, T. Shimada, H. Kobayashi, M. Takao, Y. able insightinto the inhomogeneous magnetic properties Teraoka, Y.Hayashi,S.Sasaki, Y.Miyahara, T.Harami of many DMS’s. In future studies, it is very important and T. A. Sasaki, J. SynchrotronRad. 510 (1998). [10] Y. Ishiwata, T. Takeuchi, R. Eguchi, M. Watanabe, Y. toperformadetailT andH dependentXMCDstudyfor Harada, K. Kanai, A. Chainani, M. Taguchi, S. Shin, a post-annealed samples. M. C. Debnath, I. Souma, Y. Oka, T. Hayashi, Y. This work was supported partly by Grants-in-Aids Hashimoto, S. Katsumoto and Y. Iye, Phys. Rev. B 71, for Scientific Research in Priority Area “Semiconductor 121202(R) (2005). Spintronics” (14076209), “Creation and Control of Spin [11] D. Wu, D. J. Keavney, Ruqian Wu, E. Johnston- Current” (190481012) and by PRESTO/SORST of JST Halperin, D. D. Awschalom and Jing Shi, Phys. Rev. fromthe MinistryofEducation,Culture,Sports,Science B 71 153310 (2005). [12] B. T. Thole, P. Carra, F. Sette and G. van der Laan, and Technology. Phys. Rev.Lett. 68 1943 (1992). [13] P. Carra, B. T. Thole, M. Altarelli and X. Wang, Phys. Rev. Lett.70 694 (1993). [14] H. Ohno, A. Shen, F. Matsukura, A. Oiwa, A. Endo, S. Katsumoto,andY.Iye,Appl.Phys.Lett.69,363(1996). ∗ [email protected]