Self Assembled II-VI Magnetic Quantum Dot as a Voltage-Controlled Spin-Filter

0.44 MB·

by  C. Gould

Self Assembled II-VI Magnetic Quantum Dot as a Voltage-Controlled Spin-Filter.
C. Gould1, A. Slobodskyy1, T. Slobodskyy1, P. Grabs1, D. Supp1,
P. Hawrylak2, F. Qu2, G. Schmidt1, and L.W. Molenkamp1
1
Physikalisches Institut (EP3), Universit¨at Wu¨rzburg, Am Hubland, D-97074 Wu¨rzburg, Germany
2Institute for Microstructural Sciences, NRC, Ottawa, K1A OR6, Canada.
5
(Dated: February 2, 2008)
0
0 A key element in the emergence of a full spintronics technology is the development of voltage
2 controlledspinfilterstoselectivelyinjectcarriersofdesiredspinintosemiconductors. Wepreviously
demonstrated a prototype of such a device using a II-VI dilute-magnetic semiconductor quantum
n
wellwhich, however,still requiredan externalmagnetic fieldtogenerate thelevelsplitting. Recent
a
J theorysuggeststhatspinselectionmaybeachievableinII-VIparamagneticsemiconductorswithout
external magnetic field through local carrier mediated ferromagnetic interactions. We present the
5
first experimental observation of such an effect using non-magnetic CdSe self-assembled quantum
2
dotsin a paramagnetic (Zn,Be,Mn)Se barrier.

]
l PACSnumbers: 72.25.Dc,85.75.Mm
l
a
h
- Nanomagnetics has over the past few years produced in Fig. 1. Standard optical lithography techniques were
s
e a series of fascinating and often unanticipated phenom- usedto pattern the structure into 100 µm squarepillars,
m ena. Tonameafew,molecularmagnetsexhibitquantum and contacts were applied to the top and bottom ZnSe
tunnelling of the magnetization[1], magnetic atoms on a layersinordertoperformtransportmeasurementsverti-
.
t
a surface exhibit giant magnetic anisotropies[2], and mag- cally through the layer stack. More details of the fabri-
m neticdomainwallsarebeingharnessedasdatacarriers[3]. cation procedure are given in Ref. [5]. From the size of
Here, we report on another remarkable phenomenon: thepillars,andthetypicaldensityofthedots,onewould
-
d self-assembled quantum dots, fabricated from II-VI di- expectsomemilliondotswithinourdevice. However,de-
n lute magnetic semiconductors (DMS) that macroscopi- spite this number, transportthrough similar III-V SAD-
o
cally exhibit paramagnetism, possess a remanent mag- RTDsisusuallydominatedbyonlyafewdotsthatcome
c
netization at zero external field. This allows us to op- into resonance at lower bias voltages [11, 12, 13]. We
[
erate the dots as voltage controlled spin filters, capable therefore interpret the low bias transport through our

1 of spin-selective carrier injection and detection in semi- sampleascorrespondingtoelectronstunnelling fromthe
v
conductors. Such spin filter devices could provide a key injectorintoasinglequantumdotandoutofthedotinto
7
elementinthe emergenceofafull spintronicstechnology the collector as schematically depicted in Fig. 1. Based
9
[4]. Wepresentthefirstexperimentalobservationofsuch oncalculationsofenergylevelsofstrainedquantumdots,
5
1 adeviceusinganapproachbasedontheincorporationof wefindseveralquantumdotlevelspopulatedbyelectrons
0 non-magneticCdSeselfassembledquantumdots(SADs) at zero bias. Hence the electrons tunnel through excited
5 in paramagnetic (Zn,Be,Mn)Se statesofquantumdotscontainingafinitenumberofelec-
0 Wepreviouslydemonstratedaprototypeofsuchaspin trons.
/
at filter usingaII-VI DMS-basedresonanttunnelling diode The experiments were carried out in a magnetocryo-
m [5]. However,while thatdevicewastuned byabiasvolt- stat and studied at temperatures down to 1.3 K and in
age, the spin filtering mechanism still required an exter- fields from 0 to 6 T. Fig. 2 shows a full current voltage
-
d nal magnetic field. Moreover, ferromagnetic III-V semi- curve up to a bias voltage of 170 mV. A first feature is
n conductors like (Ga,Mn)As are not suitable for resonant observedatabiasof55mV,associatedwiththefirstdot
o tunnelling devices due to the short mean free path of coming into resonance. At bias voltage above 100 mV,
c
holes [6]. Recent theoretical works [7, 8, 9] have sug- several resonances due to the ensemble of dots can also
:
v gested that spin selection may be achievable in II-VI be observed. We will first focus on the low bias feature
Xi DMS without anyexternalmagneticfield by creatinglo- which is shown in the inset to the figure. These more
calizedcarriersthatmightmediate a localferromagnetic detailed curves taken at 0 and 4T clearly show that the
r
a interaction between nearby Mn atoms. feature actually has a complex structure, consisting of
Our sample is an MBE-grown all-II-VI resonant tun- four distinct peaks, which evolve with the magnitude of
nellingdiode(RTD)structureconsistingofasingle9nm the applied magnetic field. We verified that the evolu-
thick semi-magnetic Zn0.64Be0.3Mn0.06Se tunnel barrier, tion of the features does not appreciably depend on the
sandwiched between gradient doped Zn0.97Be0.03Se in- direction of the magnetic field, indicating that the mag-
jector and collector. Embedded within the barrier are netic response of the system cannot be associated with
1.3 monolayers of CdSe. The lattice mismatch between artefacts such as two dimensional states in the injector
the CdSe and the Zn0.64Be0.3Mn0.06Se induces a strain or wetting layer [5, 11, 12, 13], and that it must be a
in the CdSe material, which is relaxed by the formation property of the dot or the barrier. We also verified that
of isolated CdSe dots [10]. The full layer stack is given the sample does not exhibit any magnetic hysteresis.
2
AAll//TTii//AAuu
3300nnmmZZnnSSee(1(1.5.5xe110919))
1155nnmm ZZnn BBee SSee ((11001188))
00..9977 00..0033
1100nnmmZZnnSSee((ii))
55nnmm ZZnn BBee MMnn SSee ((ii))
00..6644 00..33 00..0066
1 .3MLQDCCddSSee((ii))
55nnmm ZZnn BBee MMnn SSee ((ii))
00..6644 00..33 00..0066
1100nnmmZZnnSSee((ii))
1100nnmm ZZnn BBee SSee ((11001188))
00..9977 00..0033
110000nnmmZZnnSSee(1(1.5.5xe110919)) TTii//AAuu
330000nnmm ZZnn BBee SSee((88xe11081)8)
00.9.977 00.0.033
GGaaAAss SSuubbssttrraattee
FIG. 1: Full layer structure of the device, and schematic of
the transport mechanism. As electrons tunnel through the
quantum dot, they mediate a local magnetic interaction be-
tween nearby Mn ions causing them to align ferromagneti-
cally.
3.0
80 0T
2.5
70
A
A2.0 p
n nt,60
e
t,1.5 urr50 4T
n C
e
40
rr1.0
u 45 50 55 60 65
C Voltage, mV T=4K
FIG. 3: a) A surface plot of the current through the device
0.5
B=0T as a function of magnetic field and bias voltage in theregion
of the first resonance feature. b) A colour scale image of
0.0 the resonances as a function of magnetic field and voltage
0 20 40 60 80 100 120 140 160 for higher bias resonances. Since these higher resonances are
Voltage, mV
weakerandonasignificantbackground,thecolourscaleinb)
isproportionaltothevoltagederivativeofthecurrentinorder
to better resolve the position of the resonances. In both a)
FIG. 2: Current-voltage characteristic of the device, with
andb),thedataathighermagneticfieldsclearlyhasBrillouin
a high resolution view of the first resonance feature in the
like behaviour, as evidenced in b), where Brillouin functions
inset, clearly showing a strong magnetic field dependence of
are plotted as lines. However, at fields below 500 mT, the
theresonances.
behaviourdepartsfromaBrillouinfunction,withthesplitting
becomingconstantandremainingfiniteevenatzeromagnetic
field.
A better understanding of the evolution of the fea-
tures with magnetic field can be obtained from Fig. 3.
In Fig. 3(a), we plot the current through the device as a biasresonancespresentedinFig. 3(b). Thefirstofthese
colour-scalesurfacewithrespecttobiasvoltageandmag- effects, that the levels should split following a Brillouin
netic field. This puts into evidence two very important function is not all that surprising. It was previously ob-
features of the data. Firstly, that as the magnetic field servedinIII-Vdevicesthatresonancessplitinamagnetic
is increased, features split apart with a behaviour remi- field following the Land´e g factor of the material in the
niscent of the Brillouin function, and secondly, that the dots [13].
splitting remains finite in zero external magnetic field. The main difference here is that in the present exper-
The same behaviour can be seen for many of the higher iment, it is the barrier, and not the dots that are mag-
3
netic. Since the effect of the giant Zeeman splitting on a Mn ion. Even for such a simplified Hamiltonian the
theheightofthebarriersisnegligible,thepresenceofMn number of configurations is very large in the number of
should have little effect on the barrier properties. How- Mnions. ThephysicsofMn-Mninteractionsmediatedby
ever, given that electrons are not perfectly localized in electron spin can however be understood by examining
the dots, but rather have wave functions which extend anexactlysolvableproblemoftwoanti-ferromagnetically
into the barrier, it is not surprising that the quantum coupled Mn ions. The energy spectrum of the cou-
levels in the dots spin-split following the magnetization pled Mn-spin system is characterized by the total spin
of the Mn in the barriers, yielding results reminiscent of J=M±1/2 where M is the total Mn spin and the ±1/2
those previously observed [5] for the case of tunnelling corresponds to the directions of the electron spin. The
through a dilute magnetic quantum well. evolutionoftheenergyofthe systemasafunctionofthe
The observation that the splitting remains finite at total Mn spin depends on the direction of the electron
B=0 is however more surprising since there is a priori spin in the following way:
nothing ferromagnetic in the sample. This observation
can be understood by considering the effect of interac- JˆC JNN′ 35
E(M,+)=−( )M +( )[M(M +1)− ]
tions between electrons in the dot and the Mn atoms in 2 2 2
the vicinity of the dot. JˆC JNN′ 35
Electrons populate quantum dot levels according E(M,−)=( )(M +1)+( )[M(M +1)− ]
2 2 2
to the Pauli exclusion principle, and Hunds rules
[14] whenever there is orbital degeneracy. For a as shown in Fig. 4(a). In the absence of coupling to
parabolic dot, the total electron spin follows the se- the electronspin(J =0),itis obviousthatthe minimum
c
quence S = {1/2,0,1/2,1,1/2,0,1/2,1,3/2,...} with in- energy state for either electron spin corresponds to the
creasingelectronnumbers. Hence,foralmostallelectron total Mn spin M=0, i.e. an antiferromagnetic arrange-
numbers, the total spin of the dot is finite. The interac- ment. However, as shown in Fig. 4(a), with coupling to
tionofthistotalnetspinwiththespinofMnionsinduces the electron spin, the E(M,+) ground state of the com-
an effective ferromagnetic Mn-Mn interaction. This can binedsystemhasfinitetotalMnspinM∗ =( JˆC −1)/2.
be seenby consideringthe totalHamiltonianof the elec- ∗ JNN′
To estimate the value of M we approximate our quan-
tronic and Mn system[7, 8, 15, 16]:
tum dot by a spherical CdSe dot with radius R= 4 nm
H =He+g∗µBB~ ·XS~i−JCXM~R¯·S~iδ(~ri−R~) aanndd tahebaBrirri-ePrikpuosteHntaimalilotofn1iaenV. Teshteimeaffteecdtivfreomelecsttrroanin-
i R~,i MnexchangeinteractionforMnionsonthesurfaceofthe
+XgMnµBB~ ·M~R~ + 21 X JR~,R~′M~R~·M~R~′ sapthyepreicaisltMhennsgeipvaernatbiyonJˆCin=thJeCb|Φar(rRie)r|2o=f R4.152=µ1e.V2.nFmor,
R~ R~,R~′
we estimate the antiferromagnetic interaction strength
∗
Here M~ is the spin of Mn ions (M=5/2) at position J12= 1µeV. Hence forourmodel systemwe findM = 2
R~ and the coupling to electron spin aligns spins of nearest
R~, S is the spin of the i-th electron (S=1/2). J is the
i c neighbour Mn ions. Independent mean field calculations
sp-dexchangeconstantbetweentheconductionelectrons
involving tens of Mn ions randomly distributed in the
and the d-electrons of the Mn shell and JRR′ is the anti- barrier around a spherical or disk shaped quantum dots
ferromagnetic Mn-Mn interaction. The first term is the
confirm the existence of ferromagnetic ordering of Mn
spin independent Hamiltonian of electrons confined to a
ionsinthe vicinityofquantumdots[9]. InFig. 4(b), we
quantum dot in a magnetic field, and interacting via a
showthe calculatedaveragedMnandelectronspinmag-
pair wise potential. The full interaction between elec-
netization as a function of temperature for Mn ions lo-
tron spins and Mn ions in the barrier is an extremely
calizedinthebarriersurroundingasphericalCdSequan-
complicated problem. We restrict ourselves here to a
tum dot with radius of 4 nm and Mn concentration of
demonstration that the electron spin is capable of com-
4%. We find the existence of the magnetic polaron,with
pensating the anti-ferromagnetic interaction among Mn
the Mn magnetization decaying as one moves away from
ions and lead to their ferromagnetic arrangement. We
the quantum dot. These finding are in agreement with
consideronlyasingleelectroninthe groundstateandin
previous calculations of magnetic polarons [7, 8, 9], and
the absence of external magnetic field, leaving the prob-
for reasonable parameters for our system shows that the
lemofinteractingmany-electrondotsforfutureanalysis.
presence of electrons in the dot will mediate a local fer-
The effective spin Hamiltonian now reads:
romagnetic interactionbetween Mn atoms near this dot.
The interpretationof our experimental observationsis
1 therefore clear. Electrons localized in the dot mediate
H =E0−JCX|Φ(R)|2M~R¯·S~+ 2 X JR~,R~′M~R~·M~R~′ a local ferromagnetic interaction which causes a finite
R~ R~,R~′ spin splitting even in the absence of an external applied
(1) field. Our experiment is therefore tantamount to mea-
where E0 is the electron energy and |Φ(R)|2 is the suringtransportthroughasinglemagneticpolaron. The
probability of finding an electron at the position R~ of localinteractionhasastrengthcorrespondingtoaneffec-
4
randomly oriented. When an external magnetic field is
thenapplied, the ferromagneticorderwillfirstrotate to-
8
wardsthedirectionoftheappliedfield,butthiswillhave
6 J =4.5meV,
C no effectonthe transport,whichexplains why inthe ex-
V) 4 J =1meV perimentaldata,theresonancepositionsareindependent
12
e 2 of magnetic field for fields below ∼500 mT. However, as
µ
( 0 the magnetic field is further increased, it will start to
gy -2 E(M,-) dominate, and the spin splitting will grow following the
er -4 normalparamagneticinteractionofthediluteMnsystem
n [5]. Aquestionremainsastowhythezeromagneticfield
-6
E
-8 E(M,+) M* splitting isobservedherewhile itwasnotseeninthe op-
tical measurements of Ref. [16, 17]. This however can
-10
be understood by the fact that once currentbegins flow-
0 1 2 3 4 5 ingthroughthe dot,afeedbackmechanismsetsinwhere
M
spinpolarizationofthecurrentenhancesthepolarization
1.6 of Mn spins which in turn enhances the polarization of
N=12,J =1.0meV,
1.4 NN' the current. [7, 8] This dynamical effect also explains
x=0.04,R=40Å,L=1Å why spin polarization is observed at much higher tem-
1.2
perature than the predicted temperature dependence of
)
>1.0 magneto-polaronof Fig. 4(b).
S
<0.8 Inconclusion,wehaveshownthatelectronsinaquan-
( tumdotcanmediate a localferromagneticinteractionin
>0.6
M a surrounding dilute Mn system, and that this leads to
<0.4 a finite energy splitting of spin levels in the dot in the
0.2 <M> absence of an external magnetic field. Coupled with the
<S> resonant tunnelling scheme which allows the bias con-
0.0
trolled selection of which dot level is used in tunnelling,
0 10 20 30 40 50 60 70 80
T(mK) ourresultsopenupexcitingnewpossibilitiesofavoltage
controlledspinfilterwhichcanoperateinabsenceofany
external magnetic field, without relying on an inherent
ferromagnetism of the component materials.
FIG.4: a)TheenergylevelsE[+,M],E[-,M]fortwodifferent
electronspinorientationasafunctionoftotalspinoftwoMn
ions M localized on a surface of spherical quantum dot and
b)averagemagnetization asafunctionoftemperatureofMn Acknowledgments
ions randomly distributed on a surface of spherical quantum
dot.
The authorswish to thank V. Hock for help in sample
fabrication, and to acknowledge the financial support of
tivefieldoftheorderofsomehundredsofmT,andcanbe ONR, DARPA, SPINOSA, and the SFB
[1] L. Thomas et al., Nature 383, 145 (1996). [9] J. Fern´andez-Rossier, L. Brey, Phys. Rev. Lett. 93,
[2] P.Gambardella et al., Science 300, 1130 (2003). 117201 (2004).
[3] C. Ru¨ster et al., Phys.Rev.Lett. 91, 216602 (2003). M. [10] SeeP.M¨ocketal,.Journ.Elect.Materials30,748(2001).
Yamanouchi, D. Chiba, F. Matsukura, H. Ohno, Nature [11] R.J.A. Hill et al., Appl. Phys.Lett. 79, 3275 (2001).
428, 539 (2004). [12] E.E. Vdovin et al., Science 6, 122 (2000)
[4] S. Wolf, Science 294, 1488 (2001), D.D. Awschalom, D. [13] A.S.G.Thornton,T.Ihn,P.C.Main,L.Eaves,M.Henini,
Loss & N. Samarth, Ed. Semiconductor Spintronics and Appl. Phys.Lett. 73, 354 (1998).
Quantum Computation, (Springer, Berlin, 2002) pp. 93- [14] P. Hawrylak, M. Korkusinski, in Single quantum dots:
105, H. Ohno, Ed., Special issue on semiconductor spin- Fundamentals,Applications,andNewConcepts,P.Mich-
tronics, Semicond. Sci. Tech. 17 number4, (2002). ler,Ed.(Springer,Berlin,2003),TopicsinAppl.Phys.90,
[5] A.Slobodskyyetal.,Phys.Rev.Lett.90,246601(2003). pp25-92 (2003).
[6] R.Mattana et al., Phys Rev.Lett. 90, 166601 (2003) [15] A.K.Bhattacharjee,C.Benoit laGuillaume,Phys.Rev.
[7] P.Hawrylak,M.Grabowski,J.J.Quinn,Phys.Rev.B44, B55, 10613 (1997).
13082 (1992). [16] D.M. Hoffman, Solid StateComm. 114, 547 (2000).
[8] A.L.Efros, E.I. Rashba, M. Rosen, Phys. Rev. Lett. 87, [17] S. Mackowski et al., Appl.Phys. Lett. 84, 3337 (2004).
206601 (2001).