Page 1

Magnetic excitations of Fe1 + ySexTe1 − x in magnetic and superconductive phases

This article has been downloaded from IOPscience. Please scroll down to see the full text article.

2010 J. Phys.: Condens. Matter 22 142202

(http://iopscience.iop.org/0953-8984/22/14/142202)

Download details:

IP Address: 192.33.126.162

The article was downloaded on 22/03/2010 at 16:01

Please note that terms and conditions apply.

The Table of Contents and more related content is available

HomeSearchCollectionsJournalsAboutContact usMy IOPscience

Page 2

IOP PUBLISHING

JOURNAL OF PHYSICS: CONDENSED MATTER

J. Phys.: Condens. Matter 22 (2010) 142202 (6pp)doi:10.1088/0953-8984/22/14/142202

FAST TRACK COMMUNICATION

Magnetic excitations of Fe1+ySexTe1−xin

magnetic and superconductive phases

P Babkevich1,2, M Bendele3,4, A T Boothroyd1, K Conder5,

S N Gvasaliya2, R Khasanov3, E Pomjakushina5and B Roessli2

1Department of Physics, Clarendon Laboratory, Oxford University, Oxford OX1 3PU, UK

2Laboratory for Neutron Scattering, Paul Scherrer Institut, CH-5232 Villigen PSI, Switzerland

3Laboratory for Muon Spin Spectroscopy, Paul Scherrer Institut, CH-5232 Villigen PSI,

Switzerland

4Physik-Institut der Universit¨ at Z¨ urich, Winterthurerstrasse 190, CH-8057 Z¨ urich,

Switzerland

5Laboratory for Developments and Methods, Paul Scherrer Institut, CH-5232 Villigen PSI,

Switzerland

E-mail: peter.babkevich@physics.ox.ac.uk

Received 15 February 2010, in final form 1 March 2010

Published 19 March 2010

Online at stacks.iop.org/JPhysCM/22/142202

Abstract

We have used inelastic neutron scattering and muon-spin rotation to compare the low energy

magnetic excitations in single crystals of superconducting Fe1.01Se0.50Te0.50and

non-superconducting Fe1.10Se0.25Te0.75. We confirm the existence of a spin resonance in the

superconducting phase of Fe1.01Se0.50Te0.50, at an energy of 7 meV and a wavevector of

(1/2,1/2,0). The non-superconducting sample exhibits two incommensurate magnetic

excitations at (1/2,1/2,0) ± (0.18,−0.18,0) which rise steeply in energy, but no resonance is

observed at low energies. A strongly dispersive low energy magnetic excitation is also observed

in Fe1.10Se0.25Te0.75close to the commensurate antiferromagnetic ordering wavevector

(1/2 − δ,0,1/2), where δ ≈ 0.03. The magnetic correlations in both samples are found to be

quasi-two-dimensional in character and persist well above the magnetic (Fe1.10Se0.25Te0.75) and

superconducting (Fe1.01Se0.50Te0.50) transition temperatures.

(Some figures in this article are in colour only in the electronic version)

1. Introduction

Considerable effort has been devoted over the past two

years to investigating the basic properties of the Fe-based

family of superconductors [1–3].

whether magnetism plays an important role in the formation

of the superconducting state. A useful strategy for tackling

this problem is combining neutron scattering and muon-

spin rotation (μSR) measurements on one and the same

sample. Neutron scattering provides information on magnetic

correlations and on the nature of the magnetic excitations,

while μSR can determine whether static magnetic order and/or

bulk superconductivity exists.

The Fe1+ySexTe1−x system is a convenient one to study

with this methodology as large high quality crystals can be

A central question is

grown [4, 5] and the tetragonal crystallographic structure is

relatively simple to analyse and model. Single crystals are

easier to grow if there is a small excess of Fe (i.e. y > 0),

especially for small x [6–9].

The pure FeSe compound is a superconductor with a

transition temperature Tc∼ 8 K [6]. The Tccan be increased

by partial substitution of Te for Se such that Tc ∼ 14 K for

0.4 ? x ? 0.8 and y ≈ 0 [10, 11]. The application of pressure

has also been found to raise Tc, with values as high as 37 K

observed for FeSe [12–15]. Compounds with x ? 0.4 do not

exhibit bulk superconductivity but order magnetically below

a temperature which has a maximum of 67 K at x = 0 and

which decreases with x and vanishes at x ∼ 0.4. We recently

found evidence for coexistence of incommensurate magnetic

order and partial superconductivity for x ∼ 0.25 [11].

0953-8984/10/142202+06$30.00

© 2010 IOP Publishing LtdPrinted in the UK & the USA

1

Page 3

J. Phys.: Condens. Matter 22 (2010) 142202 Fast Track Communication

Figure 1. (a) Temperature dependence of the zero-field-cooled magnetization of Fe1.01Se0.50Te0.50and Fe1.10Se0.25Te0.75normalized to the

ideal 1/4π value. The onset Tonset

c

of the superconducting transition is determined from the intersection of straight lines fitted to the data

above and below the transition. (b) Representative ZF and TF μSR time spectra (upper plots) and temperature-dependentinitial TF

asymmetry of the slowly relaxing component (ATF

onset (Tonset

N

) and the mid-point (Tmid

and below the transition and as the point where the asymmetry decreases by a factor of two from its maximum value, respectively.

0and ATF

1, lower plot) for single crystals of Fe1.01Se0.50Te0.50and Fe1.10Se0.25Te0.75. The

N) of the magnetic transition are determined from the intersection of straight lines fitted to the data above

In this study, we present neutron scattering and μSR mea-

surements on two single-crystal samples: (i) Fe1.01Se0.50Te0.50,

a bulk superconductor, and (ii) Fe1.10Se0.25Te0.75, a non-

superconducting sample which exhibits magnetic order. For

superconducting Fe1.01Se0.50Te0.50 we observe a resonant

magnetic excitation consistent with that reported previously

for compounds with similar compositions [16–19].

Fe1.10Se0.25Te0.75we observe strongly dispersive, low energy

magnetic excitations associated with the magnetic ordering

wavevector (1/2 − δ,0,1/2), δ ≈ 0.03, and also with the

wavevectors (1/2,1/2,0) ± (0.18,−0.18,0).

evidence for a resonance peak in the excitation spectrum of

Fe1.10Se0.25Te0.75.

For

We find no

2. Experimental details

Single crystals of Fe1+ySexTe1−x were grown by a modified

Bridgman method as reported by [4].

measurements were carried out on the triple-axis spectrometer

TASP [20] at the SINQ [21] spallation source (Paul Scherrer

Institut, Switzerland). Bragg reflections from a pyrolytic

graphite PG(002) monochromator and analyser were used

at a fixed final wavevector of 2.66 ˚ A−1.

placed after the sample to reduce contamination from higher

order harmonics in the beam and the instrument was set up

in the open–open–open–open collimation with the analyser

focusing in the horizontal plane.

Neutron scattering

A PG filter was

The crystals were single

rods with masses of approximately 4 g. The Fe1.10Se0.25Te0.75

sample was orientated in two settings to give access to (h,0,l)

and (h,k,0) planes in reciprocal space. Measurements on

Fe1.01Se0.50Te0.50 were made in the (h,k,0) plane only. In

this report we index the reciprocal lattice with respect to the

primitive tetragonal unit cell described by the P4/nmm space

groupwith unitcell parameters a ≈ 3.8˚ A and c ≈ 6.1˚ A along

lines joining the nearest neighbour Fe atoms.

Zero-field-cooled magnetization measurements were per-

formed on a Quantum Design MPMS magnetometer with a

measuring field μ0H = 0.3 mT using the direct current

method. To reduce the effects of demagnetization, thin plate-

like pieces of Fe1+ySexTe1−x, cleaved from the main single

crystals, were oriented with the flat surface (ab plane) parallel

to the applied field.

Zero-field (ZF) and transverse-field (TF) muon-spin

rotation (μSR) experiments were performed on the πM3 beam

line at SμS (Paul Scherrer Institut, Switzerland).

experimentsa magneticfieldof 11.8mTwasappliedparallelto

the crystallographic ab plane of the crystal and perpendicular

to the muon-spin polarization.

In TF

3. Results

3.1. Magnetization and μSR measurements

Zero-field-cooled magnetization data normalized to the ideal

1/4π value (Mnorm) are shown in figure 1(a). The

2

Page 4

J. Phys.: Condens. Matter 22 (2010) 142202Fast Track Communication

Figure 2. Elastic neutron measurements of Fe1.10Se0.25Te0.75at the magnetic order propagation vector q = (0.47,0,0.5). (a) Map showing the

incommensurate peak at q in the (h,0,l) plane at 2 K. (b) Temperature dependence of the intensity at q. The inset shows scans along

(h,0,0.5) measured at 2, 20, 30 and 50 K. A sloping background function has been subtracted from the data and the dashed lines show a

Gaussian fit through the peaks. For clarity, the scans have been displaced vertically.

Fe1.01Se0.50Te0.50sample is seen to be a bulk superconductor

with the onset of the transition Tonset

−0.8 at T

exhibits superconductivity (Tonset

c

superconducting fraction of order 10% at low temperature.

For the x = 0.5 sample, the ZF time spectra measured

at T = 1.7 and 20 K are almost identical, thus suggesting

that the magnetic state of this sample is the same above

and below the superconducting transition temperature. The

solid lines correspond to a fit with the function AZF(t) =

AZF

0

is the initial asymmetry and ?ZFis

the exponential relaxation rate. Such behaviour is consistent

with dilute Fe moments as observed recently for FeSe1−x[22].

The TF data for x

=

ATF(t) = ATF

135.5 MHz T−1is the muon gyromagnetic ratio, φ is the

initial phase of the muon-spin ensemble, and σ is the Gaussian

relaxation rate.Figure 1(b) shows that the TF asymmetry

ATF

relaxationofthemuon-spinpolarizationat1.7Krelativetothat

at 20 K is due to the formation of the vortex lattice at T < Tc.

Static (within the μSR time window) magnetism develops

in Fe1.10Se0.25Te0.75as signalled by a fast drop of both AZFand

ATFwithin the first 100 ns (see the upper panel of figure 1(b)).

The solid lines correspond to fits with AZF(t) = AZF

AZF

ATF

initial ZF (TF) asymmetry and the exponential depolarization

rate of the slow (fast) relaxing component, respectively. The

temperature evolutionof ATF

below 20 K magnetism occupies more than 95% of the whole

sample volume. The corresponding values of the onset and the

mid-point of the magnetic transitions, determined as shown in

the figure, are Tonset

N

? 33.7 K and Tmid

that althoughthe magnetic order is shown to extend throughout

virtually the entire volume of the sample, μSR cannot be used

to determine whether the magnetic order is long range. The

c

? 14.0 K and Mnorm?

?

2 K. The Fe1.10Se0.25Te0.75 sample also

? 8.6 K) but has a small

0e−?ZFt, where AZF

0.5 fit well to the function

0e−(?TFt+σ2t2)cos(γμBt + φ). Here, γμ/2π =

0is almost temperature independent. The slightly stronger

1e−?ZF

1t+

2e−?ZF

2e−?TF

2tand ATF(t) = e−σ2t2/2[ATF

2tcos(γμB2t +φ)]. Here, AZF(TF)

1e−?TF

1tcos(γμB1t +φ)+

and ?ZF(TF)

1(2)

1(2)

are the

1, showninfigure 1(b), reveals that

N

? 27.6 K. We note

neutron diffraction data presented in the next section show that

the magnetic order is in fact relatively short range.

3.2. Neutron scattering results

Elastic neutron scattering measurements on Fe1.10Se0.25Te0.75

in the (h,0,l) scattering plane at 2 K, as shown in figure 2(a),

reveal a diffuse magnetic peak centred on (1/2 − δ,0,1/2)

with δ ≈ 0.03. The incommensurate peak is much broader

than the resolution of the instrument. From Q scans through

the peak we obtain correlation lengths along the a and

c axes of 11.4(6) ˚ A and 7.5(4) ˚ A respectively at 2 K.

Figure 2(b) shows that the magnetic peak develops below

TN∼ 50 K. The correlation lengths did not change measurably

upon warming through the TN (figure 2(b): inset).

magnetic propagation vector q = (1/2 − δ,0,±1/2) is

similar to that found previously for the similar composition

Fe1.03Se0.25Te0.75 compound.

confirmed that the peak described by q is magnetic in character

using neutron polarization analysis [11].

consistent with measurements on Fe1.07Se0.25Te0.75for which

the incommensurability is found to be δ ≈ 0.04 [23].

The magnetic scattering cross-section is directly pro-

portional to the magnetic response function S(Q, E)—the

Fourier transform of the space- and time-dependent spin–spin

correlation function. According to the fluctuation-dissipation

theorem, S(Q, E) is in turn related to the imaginary part of the

dynamical susceptibility χ??(Q, E) by [24]

The

For the latter compound we

Our results are

S(Q, E) =1

π[n(E,T) + 1]χ??(Q, E).

(1)

TheBose–Einsteinpopulationfactorn(E,T) = [exp(E/kBT)

−1]−1(where kBis the Boltzmannconstant) takes into account

the increase in scattering from bosonic excitations due to the

thermal population at temperatures T > 0. Correction for this

factor allows the temperature dependence of χ??(Q, E) to be

studied.

3

Page 5

J. Phys.: Condens. Matter 22 (2010) 142202Fast Track Communication

(a)(b)

(c)(d)

Figure 3. Inelastic neutron scattering from Fe1.10Se0.25Te0.75in the vicinity of the magnetic ordering wavevector q = (0.47,0,0.5).

(a) Constant energy scans collected at 2, 4 and 6 meV and 2 K along (h,0,0.6). The data have been shifted in χ??(Q, E) by arbitrary amounts

for clarity. (b) Constant energy scans collected at 2 meV and the temperature of 2 K showing χ??(Q, E) along (h,0,0.5), (h,0,0.7) and

(h,0,0.9). The plots have been displaced and the dashed lines show Gaussian peaks through the spectra. (c) Constant energy scans at 2 meV

at temperatures of 2, 40, 150 and 300 K showing χ??(Q, E) along (0.5,0,l). Note that a linear background has been subtracted in all scans.

(d) Diagram of the (h,0,l) plane to show scan directions denoted by roman numerals.

Figure 3(a) shows background corrected scans along

the (h,0,0.6) direction at energy transfers of 2, 4 and

6 meV for the Fe1.10Se0.25Te0.75 crystal.

q is present in each scan, indicating a strongly dispersing

excitation. The broadening of the dispersion in Q may be due

to unresolvable splitting of the mode into two excitations at

higher energies. The measured magnetic response at 2 meV

parallel to (1,0,0) for l = 0.5, 0.7 and 0.9, as shown in

figure 3(b), reveals considerable broadening of χ??(Q, E) in

the out-of-plane direction. Such broadening is characteristic of

a quasi-two-dimensional system with weak interactions along

c. Figure 3(c) shows that spin fluctuations persist up to at least

150 K, well into the paramagnetic state. At 40 K, i.e. close

to the magnetic ordering temperature, χ??(Q, E) is almost the

same as at 2 K.

We now turn to the low energy excitation spectrum in

the vicinity of the wavevector (1/2,1/2,0).

and (b) show maps of χ??(Q, E) measured along (h,1 −

h,0) for Fe1.10Se0.25Te0.75 at 2 and 40 K. The fluctuations

measured at 2 K are consistent with the magnetic excitation

spectrum at higher energiesreported for Fe1.03Se0.27Te0.73[25].

The excitation spectrum at 2 K is characterized by steep

incommensurate branches arising from (1/2 ± ?,1/2 ∓ ?,0)

where ? ≈ 0.18. The incommensurate excitations are still

present at 40 K. The scans shown in figure 4(c) reveal that

at E = 7 meV, the system response is nearly the same at

2 K as at 40 K. The background corrected χ??(Q, E) for

the Fe1.10Se0.25Te0.75 sample does not appear to change for

energies in the 2–7 meV range measured at these temperatures.

This is also the case for measurements along (1/2,0,l) in

A peak at Q =

Figures 4(a)

figure 3(c) that show χ??(Q, E) data at 2 meV to be similar

at 2 and 40 K.

The results obtained for Fe1.01Se0.50Te0.50 are in stark

contrasttothose forthe non-superconductingFe1.10Se0.25Te0.75

sample just described. Figures 4(d) and (e) show maps of

the magnetic spectrum as a function of wavevector along

(h,1 − h,0) for energies between 2 and 7 meV at 2 and

40 K. At 2 K we find a strong signal in χ??(Q, E) centred on

Q = (1/2,1/2,0) and E ∼ 7 meV. This feature corresponds

to the spin resonance reported previously in superconducting

FeSe0.4Te0.6[16], FeSe0.46Te0.54[18] and FeSe0.5Te0.5[19]. At

higher energies, the excitations have been found to disperse

away from (1/2,1/2,0) along (1,−1,0) [18]. However, it is

the low energy response of the system which shows the most

dramatic change on transition into the superconducting state,

as may be seen in figure 4(f). As the sample is cooled from 40

to 2 K, the integrated dynamical susceptibility of the peak at

7 meV increases by more than a factor of two and decreases in

width along (1,−1,0) by ∼30%. Fluctuations continue to be

observed well above Tc.

4. Discussion

In combination with earlier measurements,

presentedhereestablishthatthelowenergymagneticdynamics

of Fe1+ySexTe1−xvary strongly with x. The magnetic spectra

of the magnetically ordered compound (x = 0.25) and the

bulk superconductor (x

= 0.5) both contain low energy

magnetic fluctuations in the vicinity of the antiferromagnetic

wavevector (1/2,1/2,0). However, at x

the results

=

0.25 the

4

Page 6

J. Phys.: Condens. Matter 22 (2010) 142202Fast Track Communication

Figure 4. Variation of χ??(Q, E) in the (h,1 − h,0) direction for energies between 2 and 7 meV at temperatures of 2 and 40 K. Data in

(a)–(c) are from Fe1.10Se0.25Te0.75and data in (d)–(f) are from Fe1.01Se0.50Te0.50. Constant energy cuts at 7 meV along (h,1 − h,0), measured

at 2 and 40 K for Fe1.10Se0.25Te0.75and Fe1.01Se0.50Te0.50are shown in (c) and (f), respectively. A flat background has been subtracted in all

scans and dashed lines through the data are fits with a Gaussian lineshape.

fluctuations are incommensurate with wavevector (1/2 ±

?,1/2 ∓ ?,0), ?

strongest magnetic signal is commensurate.

x = 0.5 the magnetic spectrum has a gap of ∼6 meV and

the size of the signal just above the gap increases strongly

at low temperatures. This behaviour is consistent with the

superconductivity-induced spin resonance reported recently in

bulk superconducting samples of Fe1+ySexTe1−x of similar

composition to ours [16–19], and also in related Fe pnictide

superconductors [26–30].

A further difference is that the x = 0.25 sample exhibits

short-range, static (within the μSR time window) magnetic

order with a characteristic wavevector q = (1/2−δ,0,±1/2),

δ ≈ 0.03, whereas according to our μSR data there is no

staticmagnetic order inthe bulk superconductor. The magnetic

ordering wavevector q found at x = 0.25 is the same as that

in the parent phase Fe1+yTe. The slight incommensurability is

thought to be caused by the small excess of Fe accommodated

ininterstitialsitesinthecrystal structure[8, 31,32], althoughit

is interesting that the incommensurability is the same to within

experimental error at y = 0.10 (the present sample) and at

y = 0.03 (the sample studied by us previously [11]).

Our results suggest that there are two distinct magnetic

ordering tendencies at x = 0.25, one with wavevector (1/2 −

δ,0,±1/2)andtheother withwavevector (1/2±?,1/2∓?,0).

The μSR data indicate thatthe volumefraction of magnetically

ordered phase is close to 100%, but we cannot say whether the

two characteristic magnetic correlations coexist on an atomic

scale or whether the sample is magnetically inhomogeneous.

Finally, we comment on the fact that for the x = 0.25

sample diffuse peaks are observed in the elastic (within energy

resolution) channel below T ≈ 50 K by means of neutron

≈

0.18, whereas at x

=

0.5 the

Moreover, at

scattering but static magnetic order is only detected below

T ≈ 35 K by μSR. These observations can be reconciled

by means of the difference in fluctuation rates observable by

using muons (∼GHz) and neutrons (∼THz) below which spin

freezing is measured. We infer from this that the characteristic

fluctuations of the spin system lie between ∼GHz and ∼THz

for 35 K ? T ? 50 K.Such a gradual slowing down

of the fluctuations could be a consequence of the quasi-two-

dimensional nature of the spin system, which is also indicated

by the persistence of spin correlations to temperatures well

above the ordered phase. It is also interesting that the size

of the magnetically ordered domains does not significantly

increase with decreasing temperature, which suggests that the

short-range order is never truly static but fluctuates down to the

lowest temperature investigated. This picture is consistent with

therecent observationofspin-glassbehaviourinFe1.1SexTe1−x

for 0.05 < x < 0.55 [33].

5. Conclusion

We have observed a resonance-like peak at the antiferromag-

netic wavevector (1/2,1/2,0) in the low energy magnetic

spectrum of Fe1.01Se0.50Te0.50, and shown that this feature is

absent from the magnetic spectrum of Fe1.10Se0.25Te0.75which

instead shows incommensurate peaks flanking (1/2,1/2,0).

Our results reveal a clear distinction between the magnetic

excitation spectra of Fe1+ySexTe1−xsamples which are mag-

netically ordered and those which are bulk superconductors.

We conclude that the existence of a resonance peak at the

commensurate antiferromagnetic wavevector is a characteristic

of bulk superconductivity in Fe1+ySexTe1−x.

5

Page 7

J. Phys.: Condens. Matter 22 (2010) 142202Fast Track Communication

Acknowledgments

Thiswork wasperformed atthe PaulScherrer Institut,Villigen,

Switzerland. PB is grateful for the provision of a studentship

from the UK Engineering and Physical Sciences Research

Council.

References

[1] Ishida K, Nakai Y and Hosono H 2009 J. Phys. Soc. Japan

78 062001

[2] Lynn J W and Dai P 2009 Physica C 469 469–76

[3] Norman M R 2008 Physics 1 21

[4] Sales B C, Sefat A S, McGuire M A, Jin R Y, Mandrus D and

Mozharivskyj Y 2009 Phys. Rev. B 79 094521

[5] Chen G F, Chen Z G, Dong J, Hu W Z, Li G, Zhang X D,

Zheng P, Luo J L and Wang N L 2009 Phys. Rev. B

79 140509

[6] Hsu F-C et al 2008 Proc. Natl Acad. Sci. 105 14262–4

[7] Grønvold F, Haraldsen H and Vihovde J 1954 Acta Chem.

Scand. 8 1927–42

[8] Bao W et al 2009 Phys. Rev. Lett. 102 247001

[9] Yeh K W, Ke C T, Huang T W, Chen T K, Huang Y L, Wu P M

and Wu M K 2009 Cryst. Growth Des. 9 4847–51

[10] Fang M H, Pham H M, Qian B, Liu T J, Vehstedt E K, Liu Y,

Spinu L and Mao Z Q 2008 Phys. Rev. B 78 224503

[11] Khasanov R et al 2009 Phys. Rev. B 80 140511

[12] Margadonna S, Takabayashi Y, Ohishi Y, Mizuguchi Y,

Takano Y, Kagayama T, Nakagawa T, Takata M and

Prassides K 2009 Phys. Rev. B 80 064506

[13] Mizuguchi Y, Tomioka F, Tsuda S, Yamaguchi T and

Takano Y 2008 Appl. Phys. Lett. 93 152505

[14] Medvedev S et al 2009 Nat. Mater. 8 630–3

[15] Garbarino G, Sow A, Lejay P, Sulpice A, Toulemonde P,

Mezouar M and Nunez-Regueiro M 2009 Europhys. Lett.

86 27001

[16] Qiu Y et al 2009 Phys. Rev. Lett. 103 067008

[17] Iikubo S, Fujita M, Niitaka S and Takagi H 2009 J. Phys. Soc.

Japan 78 103704

[18] Argyriou D N et al 2009 arXiv:0911.4713v1

[19] Mook H A et al 2009 arXiv:0911.5463v2

[20] Semadeni F, Roessli B and B¨ oni P 2001 Physica B 297 152–4

[21] Fischer W E 1997 Physica B 234–236 1202–8

[22] Khasanov R et al 2008 Phys. Rev. B 78 220510

[23] Wen J, Xu G, Xu Z, Lin Z W, Li Q, Ratcliff W, Gu G and

Tranquada J M 2009 Phys. Rev. B 80 104506

[24] Shirane G 2002 Neutron Scattering with a Triple-Axis

Spectrometer: Basic Techniques (Cambridge: Cambridge

University Press)

[25] Lumsden M D et al 2010 Nat. Phys. 6 182–6

[26] Christianson A D et al 2008 Nature 456 930–2

[27] Lumsden M D et al 2009 Phys. Rev. Lett. 102 107005

[28] Chi S et al 2009 Phys. Rev. Lett. 102 107006

[29] Li S, Chen Y, Chang S, Lynn J W, Li L, Luo Y, Cao G,

Xu Z and Dai P 2009 Phys. Rev. B 79 174527

[30] Inosov D S et al 2009 Nat. Phys. 6 178–81

[31] Fang C, Bernevig B A and Hu J 2009 Europhys. Lett. 86 67005

[32] Ma F, Ji W, Hu J, Lu Z-Y and Xiang T 2009 Phys. Rev. Lett.

102 177003

[33] Paulose P L, Yadav C S and Subhedar K M 2009

arXiv:0907.3513v1

6