Spin- and angle-resolved photoemission ... - Science Direct

JOURNAL OF ELECTRON SPECTROSCOPY and Related Phenomena

ELSEVIER

Journal of Electron Spectroscopyand Related Phenomena 88-91 (1998) 207-212

Spin- and angle-resolved photoemission spectroscopy of ferromagnetic MnAs K. Shimada a'*, O. Rader h, A. Fujimori b, A. Kimura c, K. Ono d, N. Kamakura c, A. Kakizaki c, M. Tanaka d, M. S h i r a i e aHiroshima Synchrotron Radiation Center, Hiroshima University, Higashi-Hiroshima 739, Japan bFaculty of Science, University of Tokyo, Bunkyo-ku, Tokyo 113, Japan Clnstitutefor Solid State Physics, University of Tokyo, Minato-ku, Tokyo 106, Japan dFaculty of Engineering, University of Tokyo, Bunkyo-ku, Tokyo 113, Japan eGraduate School of Engineering Science, Osaka University, Toyonaka, Osaka 560, Japan

Abstract Spin-resolved photoemission spectroscopy has been performed on thick magnetized ferromagnetic single crystal layers of MnAs(il01) and MnAs(1100) grown onto GaAs(001). We observed large spin polarization from Mn 3d states in the valence band of these materials and strong intensity enhancement in the majority-spin spectrum at the photon energy slightly above the Mn 3p core excitation threshold. Band positions are determined for MnAs(i 101), which are consistent with the LAPW bandstructure calculation. Exchange splittings at the L-point derived for two different bands from data measured at MnAs(i 101) are 2.1 +_ 0.3 and 2.7 _+ 0.5 eV. They are slightly larger than the calculated values. © 1998 Elsevier Science B.V. Keywords: Itinerant ferromagnetism; MnAs; Spin-resolved photoemission spectroscopy

1. Introduction The manganese pnictides (MnX: X = As, Sb, Bi) have been studied extensively for their various physical properties associated with their ferromagnetism [1-5]. Recently, single crystals of these compounds have been grown on GaAs by molecular beam epitaxy (MBE) [6,7]. A hetero-junction between semiconductor and ferromagnetic metal has attracted much interest as a new type of device utilizing the spins of the charge carriers [6,7]. M n A s ( i l 0 0 ) grows epitaxially on an As-precovered GaAs(001) surface and M n A s ( l l 0 1 ) on Mn-precovered * Corresponding author. Fax: +81-824-24-6294; e-mail: kshimada @hisor.material.sci.hiroshima-u.ac.jp.

GaAs(001) as shown in Fig. l(a) [6]. MnAs has various structural and magnetic phases in the pressure (p) versus temperature (T) plane [3]. At atmospheric pressure, there are two structural transformations at Ttl = 318 K ( = Curie temperature: Tc) and at Tt2 = 390 K [2,3]. For T < Ttl and Tt2 ~ T, MnAs has the crystal structure of the NiAs-type and for Ttl < T < Tt2, the MnP-type [2,3]. The observed ordered magnetic moment is 3.45/xa/formula [2]. According to polarized neutron scattering studies [ 10], a small magnetic moment, - 0.23 -+ 0.05/xB, is induced on the As sites, which is antiparallel to the moment on Mn, indicating the importance of p - d hybridization. In order to explain the unusual magnetic properties, several attempts have been made using a molecular field approximation [4,5] or using a band model [8,9].

0368-2048/98/$19.00 © 1998 Elsevier Science B.V. All rights reserved PII S0368-2048(97)00129- 1

208

K. Shimada et al./Journal of Electron Spectroscopy and Related Phenomena 88-91 (1998) 207-212

(a)

a (i O0)

MnAs(llOl)

~¢ ~,. c=-5.713 a=3.725 . / ~ / ~ (boo) ~ plane ~ " - ~

/ 6.452

[1100] _ ~ ¢ /

1

(ilOl)

plane

[ O 0 0 1 ] ~ [ 11201 © Mn • As 0 Ga

[OOl1 []lO]~lOl

[001l [h 0 1 ~ l lOl GaAs(001 )

GaAs(001 )

(b)

Fig. 1. (a) Structures of MnAs(il01)/GaAs(001) and Mn(] 100)/GaAs(001) [6]; (b) first Brillouin zone for the NiAs-type crystal structure. Katoh and Motizuki have calculated the density of states of MnAs and discussed the physical properties at finite temperatures taking into account spin fluctuation [9]. In the present paper, we have performed spinand angle-resolved photoemission measurements on MnAs films in order to clarify the electronic states. We will discuss the observed spectra in comparison with new band-structure calculations performed using the linearized augmented-plane-wave method (LAPW) (M. Shirai et al., unpublished data).

2. Experiment Single crystals of 2.5/xm thick MnAs(il00)/ GaAs(001) and 0.44/~m thick MnAs(1101)/ GaAs(001) were synthesized by MBE. After growing the crystal, an As cap was deposited for protecting the surface against contamination, while the crystal was

transported through air and mounted in the photoemission apparatus. After reaching ultra-high vacuum conditions, the crystals were decapped from As by heating to about 460°C. The sample surfaces were, however, contaminated by carbon and oxygen which could not be removed by heating only. To obtain a clean surface we have applied repeated Ne-ion bombardment and subsequent annealing at about 460°C. After the cleaning process, Auger electron spectra showed no contaminants and almost one to one atomic ratio of Mn and As [12]. As for MnAs(ll01), we could observe LEED patterns, which are consistent with the expected surface symmetries. As shown below, we could observe a clear and strong spin-polarization for the present samples, but we could not see a LEED pattern for MnAs(i 100), which may indicate disorder in the topmost surface atomic layers. With the (3101) surface, the band dispersions along the I'L-line in the Brillouin zone can be


resolved by photoemission using a normal emission geometry [Fig. 1 (b)]. For MnAs(il01), the photon energies hu = 40 and 74 eV correspond to the the Land F-points, respectively. 1 The spin- and angleresolved photoemission spectroscopy (SARPES) measurements were done at the Revolver undulator beamline BL-19A at the Photon Factory, High Energy Accelerator Research Organization [14]. We used a hemispherical electron energy analyzer with the acceptance angle of _ 2 ° connected to a 100 kV Mott detector [15]. We measured SARPES in the normal emission mode with the energy resolution of 0.3-0.6 eV for hu = 30-74 eV. The polarization direction of light and the c-axis of the sample are on the same plane and the incidence angle of the light was 20 ° measured from the surface normal. The base pressure in the analyzer chamber was smaller than 4 × 10 -l° mbar. We fixed samples within the gap of a/~-metal yoke which was mounted on a cryostat. A heater was attached to the back of the sample for annealing. We applied a magnetic field along the [1120] easy magnetization axis in the film plane. SARPES measurements have been performed with the remanently magnetized samples. The sample magnetization was controlled in situ by the magnetooptical Kerr effect (MOKE). As shown in Fig. 2, the magnetization saturates and the residual magnetization is equal to the saturation magnetization. Since Tc ( = 318 K) of MnAs is close to room temperature (--0.95 Tc), we cooled the samples to 1 3 0 - 1 8 0 K (0.43-0.57 Tc) during the photoemission and MOKE measurements. Energy calibration was done with the Fermi edge of spin-integrated photoemission spectra. In order to eliminate apparatus asymmetries, we have averaged the spin-resolved photoemission spectra over the upward and downward magnetization direction.

'M n A s ' M O/EK (1100) 1~0 K

t/) t-"

---(1101)

'

209

'

138K

vt~ t/J t"

///

//

¢-

I

I

-4

-2

0

I

I

2

4

Current (A) Fig. 2. Hysteresis curves of MnAs(il01) and MnAs(]100) measured using the magneto-optical Kerr effect.

The majority-spin intensity always dominates as clearly shown in the figure. The cross-section ratio As 4p/Mn 3d is about 0.1 at h~ -- 30 eV and about 0.01 at hl, = 55 eV [16]; therefore, the photoemission spectra mostly reflect the Mn 3d states. Taken at h~ -55 eV which is slightly above the on-resonance I

I

I

M n A s ~1100~

• Majority v Minority

135 K C1-1

"~

~=-.

C1"3

a1"1

at,

3. Results and discussion

b$1 Fig. 3 shows the spin-resolved photoemission spectra of MnAs(ll00) taken at h~ = 30 and 55 eV.

hv =

8

b,[.~ )

30 eV

*~=

1

I

I

6

4

2

0

Binding Energy (eV) t To calculate photon energies for the high symmetrical points, we assumed a free-electron final-state model with an inner potential of 10 eV and a work function of 4.5 eV. For the detailed procedures, see for example, Ref. [13].

Fig. 3. Spin-resolved photoemission spectra of MnAs(1100) taken at hp = 30 and 55 eV. Note the strong intensity enhancement in the majority-spin spectrum taken at h~ = 55 eV.

210


photon energy of the Mn 3 p - 3 d excitation [17], the spin polarization at EB = 3.7 eV (c T2) and 1.9 eV (c 13) are significantly enhanced as is for the higher binding energy region (EB > 5 eV). As shown in the figure, the resonant enhancement of the photoemission intensity at EB = 3.7 eV is mainly derived from the majority-spin spectrum. Interference effect between direct photoemission and 3 p - 3 d excitations followed by a super-Coster-Kronig decay should contribute to the strong enhancement of the majority-spin spectrum [18]. Since most of the unoccupied Mn 3d states are the minority-spin band, a down-spin electron in the Mn 3p core level is likely to make a transition to the unoccupied state within the minority-spin band. The strong enhancement in the majority-spin state suggests that the Auger matrix elements are large for the two hole states left behind being coupled to a singlet. Fig. 4 shows the spin-resolved valence-band spectra of MnAs(ll01) taken at hu = 30, 40 and 74 eV. The peak positions were searched by taking the second derivative after smoothing (not shown) the raw data. As shown in the figure, the peaks in the majority-spin spectra and those in the minorityspin spectra have different binding energies reflecting the different band structures. In the h~ = 30 eV spectra, the minority-spin spectrum has a significant peak (f ~5) just at the Fermi level (EF) cut-off, suggesting the existence of a minority-spin band there. It should be noted that the hu = 3 0 e V spectra of MnAs(il00) and MnAs(1101) have indeed different structures. A significant structure at EB--4 eV (b 12) observed in the minority-spin spectrum of MnAs(il00), for example, is absent in the corresponding MnAs(il01) spectrum. The majority-spin spectrum taken at hp = 40 eV, which corresponds to the L-point at EB = 3 eV, has peaks at EB = 6.2 eV (g T 1), 3.8 eV (g r 2), 3.0 eV (g T3), 1.9 eV (g r4) and 0.33 eV (g Ts). On the other hand, the minorityspin spectrum has peaks at EB = 5.8 eV (h I l), 3.4 eV (h 12), 1.7 eV (h 13) and 0.33 eV (h 14). Fig. 5 shows observed peak positions by large symbols as a function of wave number reduced to the first Brillouin zone of MnAs(ll01) and the calculated points by small symbols from the LAPW bandstructure calculation (M. Shirai et al., unpublished data). The band-structure calculation shows a minority-spin band just below EF near the L-point

i

i

i

MnAs (i101) 135 K hv=74 eV (F-point)

j,l,2

. ~ l~v~

il,3

I

J$1

=.

Majority

V Minority

~I~ v~e,

~A~'~" J ~

il, 1

~'"

I •

il,4

el',

4

. I~' i,

or)

t-s

.6 u)

t(!) t-

8

6

4

2

0

Binding Energy (eV) Fig. 4. Spin- and angle-resolved photoemission spectra of MnAs(ll01) taken at hv = 30, 40 and 73 eV. Binding energies from the band-structure calculation (M. Shirai et al., unpublished data) are also shown. and there is no majority-spin band, which is consistent with the observed peak structure in the minority-spin spectrum taken at hu -- 30 and 40 eV as shown in Fig. 4. At the L-point (hg = 40 eV), however, the observed minority-spin peak near EF is not so sharp as that of the hu = 30 eV spectrum. This may be explained if a part of the spectral weight is above EF since the minority-spin band is very close to EF. Table 1 lists calculated binding energies and partial weights of wave functions for the majority-spin and

211


MnAs (1101)

".Maiori-p.i 0 !!;i: . . . . . ~

~AJlJ~

•

"~:, 2 . . . . . : : " ; " i;

>'

.... : ..... ,!!

>

~L

v v v

>

d'

r~gvv

v

'~

,~,v

l

I~ 4

t-

LLJ

E

J~JAAJ

uJ

', ,." A ". il -

6

6#"

-A'~• •

~

F

i

ii

i

i

" v

i r i

(a)

v ,,:/

Yv]

8 7~eV 3'0eV 4Q"e

74 eV 30 eV 40,'eV 8

v v

F

L

(b)

L

Fig. 5. Observed band points of MnAs(1101) (large symbols) and calculated band points (small symbols) (M. Shirai et al., unpublished data). (a) Majority-spin band dispersions; (b) minority-spin band dispersions. The dashed lines show the points on which spin- and angle-resolved photoemission spectra taken at hu = 30, 40 and 73 eV lie.

minority-spin bands at the L point (M. Shirai et al., unpublished data). The bands at the L-point are classified into two irreducible representations denoted here as L1 and L2. Here the Llo, (a = T, I ) denotes the nth band numbered from the lowest band (L1 oi) in the L1 representation. At the bottom of Fig. 4, we have shown calculated binding energies of bands at the L-point. Comparing the spectra measured at

hu = 4 0 e V w i t h t h e c a l c u l a t e d b a n d p o s i t i o n s at t h e L - p o i n t , in t h e m a j o r i t y s p i n s p e c t r u m w e a s s i g n e d s t r u c t u r e s g T 2, g T 3, a n d g r 4 to L1 l 3/L2 t ], L1 l 4 a n d L1 T 5/L2 T 2, r e s p e c t i v e l y . F o r t h e m i n o r i t y - s p i n s p e c t r u m , w e a s s i g n e d h 13 a n d h 14 to L1 13/L2 I l a n d L1 14, r e s p e c t i v e l y . T h e e x p e r i m e n t a l e x c h a n g e s p l i t ting can therefore be estimated from the difference of t h e b i n d i n g e n e r g i e s o f g 12 a n d h 13 to b e 2.1 ± 0.3 e V

Table 1 The calculated binding energies and compositions of wave functions for the minority-spin and majority-spin bands (M. Shirai et al., unpublished data). We have listed bands below the Fermi level. At the L-point, each band has two-fold degeneracy Binding energy (eV)

Weight within the Muffin-tin sphere (%) Mn s

As p

d

s

p

d

Majority spin L1 T~ L1 r 2 L1 13 L1 ]4 L1 15 L2 i e L2 T2

10.75 5.79 3.21 2.38 1.77 3.30 2.01

1.00 10.32 1.49 0.18 3.09 0 0

3.00 4.9 3.59 1.58 0.19 1.97 0.01

4.23 9.32 60.03 80.11 86.4 62.92 94.57

72.05 0.02 0.11 0.36 0.07 0 0

0.04 31.74 17.36 4.98 2.97 20.31 0.06

0.14 0.18 0.80 1.28 0.32 0.33 0.23

Minority spin L1 t ~ L1 I z L1 ~3 L1 14 L2 I ]

10.62 5.33 1.79 0.04 1.89

0.87 9.82 1.64 0.51 0

2.81 4.96 4.93 2.19 4.25

3.08 4.46 32.29 71.4 32.2

73.83 0.06 0.09 0.04 0

0.06 35.2 35.9 3.6 39.78

0.14 0.15 0.85 3.07 0.13

212


and from g T 3 and h I 4 to be 2.7 + 0.5 eV. The calculated exchange splitting between L1 I 3 and L1 3 and between L2 T1 and L2 11 are 1.42 and 1.41 eV, respectively (M. Shirai et al., unpublished data). In these bands, the weight of Mn 3d is comparable or slightly larger than the one of As 4p. The calculated exchange splitting between the L1 T4 and L1 14 bands is 2.34 eV (M. Shirai et al., unpublished data). In these bands the Mn-3d derived states dominate. Larger exchange splitting in Llo4 bands than those in L1 o3 and L2ol bands is consistent with larger magnetic moments on Mn than on As [9,10] (also M. Shirai et al., unpublished data). The observed exchange splitting in Llo4 band is indeed larger than those in LIo3 and L2~1 bands. However, the obtained exchange splittings are slightly larger than the calculated values (M. Shirai et al., unpublished data) suggesting the underestimation of exchange splitting in the bandstructure calculation.

4. Conclusions

exchange splittings at the L point were estimated as 2.1 ___0.3 and 2.7 ___0.5 eV for different bands, which are slightly larger than the calculated values, 1.4 and 2.3 eV, respectively.

Acknowledgements O.R. acknowledges a research fellowship by the Japan Society for the Promotion of Science.

References [1] [2] [3] [4] [5] [6]

[7] [8]

We have examined the electronic states of MnAs(ll01) and MnAs(i 100)single crystals grown onto GaAs(001) using photoemission spectroscopy. At hu = 55 eV, which is slightly above the onresonance photon energy of the Mn 3 p - 3 d excitation, the measured spin integrated intensity and the measured spin polarization both are enhanced for the binding energy of about EB--3.7eV. For MnAs(1101) with the ordered surface, we have determined band positions along the F-L-direction, which generally agree with the results of the LAPW bandstructure calculation (M. Shirai et al., unpublished data). In the minority-spin spectrum, we have observed a sharp peak just below the Fermi level cut-off near the L-point in accordance with the prediction of the band-structure calculation. The

[9]

[10] [11] [12] [13] [14]

[15] [16] [17]

K. Adachi, J. Phys. Soc. Jap. 16 (1961) 2187. H. Ido, J. Appl. Phys. 57 (1985) 3247. N. Menyuk et al., Phys. Rev. 177 (1969) 942. C.P. Bean, D.S. Rodbell, Phys. Rev. 126 (1962) 104. J.B. Goodenough, J.A. Kafalas, Phys. Rev. 157 (1967) 389. M. Tanaka et al., Appl. Phys. Lett. 65 (1994) 1964; M. Tanaka et al., J. Appl. Phys. 76 (1994) 6278; M. Tanaka, Mater. Sci. Eng. B 31 (1995) 117. H. Akinaga et al., J. Cryst. Growth 150 (1995) 1144; H. Akinaga et al., Appl. Phys. Lett. 67 (1995) 1. L.M. Sandratskii, R.F. Egorov, A.A. Berdyshev, Phys. Stat. Sol. (b) 103 (1981) 511. K. Katoh, A. Yanase and K. Motizuki, J. Magn. Magn. Mater. 54-57 (1986) 959; K. Motizuki, J. Magn. Magn. Mater. 70 (1987) 1. Y. Yamaguchi, H. Watanabe, J. Magn. Magn. Mater. 31-34 (1983) 619. L.E. Davis et al., Handbook of Auger Electron Spectroscopy, Physical Electronics Industory, Inc., Minnesota, 1976, p. 1. S. Hiifner, Photoelectron Spectroscopy, Springer, Berlin, 1995. A. Kakizaki et al., Rev. Sci. Instrum. 63 (1992) 367. J. Fujii et al., in: Proceedings of the International Workshop on Polarized Electron Sources and Electron Spin Polarimeters in Nagoya, Universal Academy Press, Tokyo, 1992, p. 885. J.J. Yeh, I. Lindau, Atom. Data Nucl. Data Tables 32 (1985) 1. T. Matsushita et al., Jap. J. Appl. Phys. 31 (1992) L1767; A. Kimura et al., Solid State Commun. 81 (1992) 707. L.C. Davis, L.A. Feldkamp, Phys. Rev. B 23 (1981) 6239.