Comment on "Observation of Spin Injection at a Ferromagnet-Semiconductor Interface, by P.R. Hammar et al

2 Pages·0.05 MB·English

by  B. J. van Wees

Comment on ”Observation of Spin Injection at a Ferromagnet-Semiconductor
Interface”, by P.R. Hammar et al.
B.J. van Wees
Department of Applied Physics and Materials Science Centre, University of Groningen, 9747 AG Groningen, The Netherlands
In a recent Letter Hammar et al. [1] claim the obser- connected by the low resistance ferromagnetic electrode,
vation of injection of a spin-polarized current in a two- and interchanging this pair should make little difference
dimensional electron gas (2DEG). This is an important on the measured resistance. Therefore the observed be-
observation, since, despite considerable effort of several haviour is consistent with the reciprocity theorem, and
groups,allattemptstorealizespin-injectionintoa2DEG thechangeinresistancecanpossiblybecausedbyaHall
0 using purely electrical measurements have failed sofar. effectduetothereversalofthemagneticfield(e.g. result-
0 However,in my opinion the claim made in [1] is not cor- ingfromthefringefieldsattheedgesoftheferromagnetic
0
rect, and the observed behaviour can be explained by a electrode). An estimate of the effect on the resistance is
2
combination of a magneto resistance (Hall) effect with a difficult however, due to the uncertainty where the cur-

n spin-independent rectification effect due to the presence rent from the 2DEG actually enters the ferromagnet.
a
of a metal-semiconductor junction. Thesecondimportantaspectisthattheauthorsstudy
J
The interpretation of the data depends crucially on transport through a semiconductor-metal system, which
3
1 the theoretical description formulated in [2]. A 2DEG is containsabarrier. Judgedfromthemeasuredresistance,
considered, connected to a ferromagnetic electrode. The which is low and comparable to the resistance of the

1 key ingredient is that the electron spin is conserved in 2DEGitself,thebarrierisnotveryeffective. However,on
v the 2DEG, but the spin-orbit interaction of the Rashba thevoltagescaleof0.1V whichtheauthorsuse,somesort
7
type induces an asymmetry betweenelectrons moving in of rectification might occur, leading to V(−I)6=−V(I).
8
1 a particular direction with different spin directions. As This effect will not depend on the direction of the mag-
1 a result rectification is predicted, which depends on the netization M, since the electron spin is not relevant for
0 direction M of the magnetization of the ferromagnetic rectification in metal-semiconductor junctions. In other
0 electrode. words: V(I)M =V(I)−M.
0
However, at low currents linear response theory re- The presented data therefore does not exclude that
/
t quiresthatV(−I)=−V(I). Rectificationcanonlyoccur the observedeffects are due to a combination of a (mag-
a
m for currents I beyond the linear transport regime, under netization independent) rectification, combined with a
the condition that the transport properties of the elec- magneto resistance (Hall) effect. It is however possible
-
d tronsareenergydependent,andthatthecurrentinboth to distinguish between this explanation and that of the
n directions is carried by electrons with different energies. authors. To support their case they should show that
o However, no energy scale which determines the onset of the rectification behaviour observed in the V(I) char-
c
: rectificationisdiscussedin[2],wheretherectificationde- acteristics changes sign, when the magnetization M is
v
pends onthe directionofthe currentonly. Thereforethe reversed. Inotherwords,forafixedconfigurationofcur-
i
X theory of [2] cannot be correct, and it is not possible to rent and voltage leads, they should show explicitly that
r detect spin injection in this way. the relation V(I)M =−V(−I)−M is obeyed, for the full
a Turning attention to the experiment, it should be current range. In particular this implies that when M
noted first that the authors test the prediction of spin- switches direction when the coercive field is exceeded,
dependent current rectification by performing a four- and a change ∆R is observed for positive I, an oppo-
terminal measurement. They reverse the current di- sitechange−∆Rshouldbe observedatnegativeI. This
rection by interchanging one current and one voltage behaviour can then clearly be distinguished from a Hall
lead. This however is not a critical test of the theory, effect, which predicts that the change in resistance ∆R
since[2]predictsachangeinresistancewhenthecurrent will have the same sign for both current directions.
direction is reversed, but the current and voltage leads
themselves are not changed. The reciprocity theorem for
multi-terminal measurements [3] states that (in the lin-
ear transport regime) the resistance should be invariant
under interchange of the current and voltage leads, ac-
companied by a reversal of the magnetic field B. Al-
[1] P.R. Hammar, et al. Phys. Rev.Lett. 83, 203 (1999)
though in the experiment only one current and voltage
[2] M. Johnson, Phys.Rev. B58, 9635 (1998)
probeisinterchanged,the resultshouldbealmostequiv-
[3] M.Buttiker,Phys.Rev.Lett.57,1761(1986) Thispaper
alent to interchanging both pairs, since the other pair is
also discusses the Onsager-Casimir reciprocity relations,
1
which apply to the general case of multi-terminal resis-
tance measurements
2