PHYSICAL REVIEW B VOLUME 47, NUMBER 2 1JANUARY 1993-II
Vortex-lattice
transition
in
superconducting
Nb/NbZr
multilayers
P.
Koorevaar, W. Maj, *P.
H.
Kes, andJ.
AartsKamerlingh Onnes Laooratorium der Rijksuniversiteit Leiden, P.O.Box9506,2300RA Leiden, Thenetherlands (Received 28 July 1992)
Multilayers ofNb/NbZr are known to show two dimensional crossovers [three dimensional (3D) to
two dimensional (2D) to 3D]in the behavior ofthe parallel critical field H,z~~ as afunction of
tempera-ture T. Here we report a systematic study ofthe behavior ofthe critical current
I,
as function ofthe magnetic field H in the different regimes. Varied were layer thicknesses as well as the angle between Hand the sample surface, and the angle between H and the current
I.
Apart from earlier reported non-monotonic behavior ofI,
(H),which constitutes the continuation ofthe 2Dbehavior in the K~I(T) phase diagram, we find strong maxima inI,
at fields H~ far below the 2D phase line. An analysis in terms of Lorentz forces on theflux lines shows that H~ signifies a transition from alow-field region where straight vortices move freely in adirection perpendicular to the layers, to ahigh-field region where the vortices have developed kinks perpendicular to the layers. In this field regimeI,
isdetermined by motion ofthe kinks along the layers, while the portions ofthe vortex parallel to the layers are pinned by an intrinsic mechanism.I.
INTRODUCTIONSince the discovery
of
high-temperature superconduc-tors there has been renewed interest in the magnetic propertiesof
layered superconductors. ' The anisotropy induced by the layering has important consequences for the structure and behaviorof
the vortex lattice, which is reAected in the temperature and angular dependenceof
the upper and lower critical fields
H,
2andH„,
or in the field dependenceof
the critical currentI,
.
Superconduct-ing multilayers are very convenient model systems for in-vestigating such properties since the anisotropy can be changed more or less continuously by changing the layer thicknesses. Also, the whole field-temperature phase dia-gram is accessible, whereas in the high-T, materials,H,
2 is too high except for temperatures close toT,
. Simple examplesof
phenomena due to the layering are the three-dimensional (3D) to two-dimensional (2D) cross-overs in the parallel critical fieldH,
2~~
of
superconductor-normal metal (SN)or superconductor-insulator (SI) multi-layers ' or, more recently, the decoupling line in the
per-pendicular field in the phase diagram
of SI
multilayers. 'In this paper we want to deal with a third class
of
mul-tilayers where both components are superconductors, which in our case will be Nb and NbZr. Even when these components have the sameT,
but different valuesof
their zero-temperature coherence lengthg(0),
theH,
zt~(T) line can exhibit two crossovers in dimensionality. This be-havior was predicted by Takahashi and Tachiki and has since been demonstrated on Nb/NbTi and Nb/NbZr multilayers. ' ' Going down in temperature fromT,
onefinds a crossing from average
3D
(3D)
to
2D and from 2Dto 3D
behavior(3D-2D-3D).
The crossovers are well understood and signify changesof
the nucleation positionof
the order parameter from one layer to another. We note the following points, which are illustrated inFig. 1:
(I)
In the3D
regime nearT,
the order parameter is spread out over several layers, yielding an averagedbe-"c2/!
3D 20 30
H2D
Hp
H)
FIG.
1. A sketch oftheH-T
phase diagram for Nb/NbZrmultilayers for parallel fields. Indicated are the different
re-gimes ofthe H,2~~-Tline and the different phases below the criti-calfield (see text in Sec.
III
Cfor denotation). A schematic rep-resentation ofthe vortex structure in these phases is given aswell.
havior,
of
the constituting layers and linear behaviorof
H,
zl(T).(2) In the so-called 2D regime between
T,
zD andTzD&D,
H,
zi(T)
can be accurately described by assumingthat the Nb layers, having the larger
g(0),
behave like two-dimensional superconducting thin films which are completely decoupled by nonsuperconducting NbZr lay-ers. They have an effective thicknessd,
z slightly larger than the nominal Nb thickness (db), due to the extensionof
superconductivity into the NbZr layers by the proxim-ity effect. The temperature dependenceof H,
2~~ is givenby the well-known thin-film formula
H
2~iT)=H,
D(0)V
1T—
/T„D,
(1)with
H2D(0)=$0&12/[2~d, sg(0)].
Note that theT,
2Dused in Eq. (1) is slightly lower than the
T, of
the multi-layer.(3)
For
temperatures below TQD3D superconductivity nucleates in theNbzr
layers, having the smallerg(0),
which can now individually behave as
3D
bulk NbZr. The temperature TQD30 at which the transition occurs is found by the criterion thatg,
„(T2D&D )=0.
3A, whereg,
„
follows from the perpendicular critical field
H,
2i=
Pol(2vrg,„)
and A=
d&+
d,
is the multilayerperiodicity. ' At
H,
2~~ in this3D
regime itisthought thatthe superconducting NbZr layers are not coupled by the Nb layers, since the fields involved are higher than the critical fields for the thin Nb films. So in both the 2D and the
3D
regime atH,
2~~ the layers carryingsupercon-ductivity are decoupled toa high degree.
Although we will be mainly concerned with parallel fields, it is useful to remark that the perpendicular fields show one crossover at a temperature around T2D3D. Above this temperature the behavior is linear with a 3D-like slope,
"
below the crossover it is also linear, with a NbZr-like slope. The apparent anisotropy as measured by the ratioH,
2~~/H, 2~ is therefore always small,of
theorder
of 1.
5—2.
The behavior
of H,
2~~ in these multilayers is well
under-stood, but the properties
of
the superconducting state belowH,
2are less clear.For
temperatures in the 3D re-gime, measurementsof
the critical currentI,
as a func-tionof
field' showed the existenceof
a 2Dphase line (see Fig. 1) at whichI,
suddenly increases with decreasing field. An exampleof
this will be given subsequently. The 2D phase line is simply the continuationof H,
2~~ from the2D regime according to
Eq. (1). It
was found in Ref. 12 that both above and below the 2D phase line the angle between field and current did not inAuenceI„which
remained unexplained at the time. Still, the results sug-gested that below the 2D phase line in the3D
regime su-perconductivity is mainly confined to the Nb layers,i.
e., the situation is equivalent to that in the 2D regime and the superconducting order parameter is strongly modu-lated.The question now arises whether this state holds down to the Meissner state; a transition at low fields to amore isotropic state would also seem possible, since in zero field both Nb and NbZr are superconducting. In that case the modulation
of
the order parameter would de-crease with decreasing field, which could affect the struc-tureof
the vortices and have consequences for the pin-ningof
the vortices and the critical current. Especially interesting in this respect is the questionof
intrinsic pin-ningof
vortices by the layers. In this paper we will ex-tensively study these possibilities.We have investigated the phase diagram
of
Nb/NbZr multilayers by measurementsof
I,
as a functionof
field for different temperatures and field orientations. BelowH,
2 in the 2D regime or below the aforementioned 2D phase line in the3D
regime, for fields parallel to thelay-ers and current parallel to the layers but perpendicular to the field, we observe strong nonmonotonic behavior
of
I,
with a peak at a temperature-dependent field
H .
The valueof
I,
atH
can be very large, almost two ordersof
magnitude higher than in perpendicular fields.
For
a better understandingof
this effect we then changed a numberof
parameters, as will be detailed in different sec-tions.Following Sec.
II
on the experimental details, inSec.
III
we will study the mechanism governing the nonmono-tonic behaviorof
I,
by investigating the infiuenceof
the magnitude and the directionof
the Lorentz force on the vortices. From this we propose that aboveH
the vortex structure existsof
long vortex cores (strings) parallel to the layers, combined with small vortex discs perpendicu-lar to the layers,i.e.
, a kinked vortex lattice. This statebears an obvious resemblance to the pancake vortices found in highly anisotropic high-T, superconductors, ' and it isquite surprising to find the same picture holding for these layered structures with small anisotropies and no Josephson coupling. In the regime above
H
any motionof
the strings perpendicular to the layers is im-peded by the intrinsic pinning related to the layered structure, although the conceptof
matchingof
the vortex lattice to the layer periodicity''
does not play a role. The disks on the other hand can move along the layers and the pinningof
the disks determines the critical current. BelowH
movementof
the vortices normal to the layers ispossible. There are experimental indications that this is due to a change from a kinked vortex lattice to a straight Abrikosov vortex lattice and not from a sim-ple change in the strength by which the strings are pinned. We henceforth denote this phenomenon as a structural vortex lattice transition (SVLT) and take it to becharacterized by the fieldH
.
In Sec. IV we investigate the inhuence
of
the sample parameters and the orientationof
the sample with respect to the field on thisSVLT.
The thicknessof
the Nb and NbZr layers and the numberof
layers in the sample is varied. We show that the SVLTisonly dependent on the thicknessof
the Nb layers and not on the angle between field and layers. The SVLT therefore does not appear to be alock-in transition as described inRef.
16,although it is due to a competition between the superconducting con-densation energy and the line energyof
the kinked vor-tices.II.
EXPERIMENTALThe Nb/NbZr multilayers are prepared by magnetron sputtering in an Ar pressure
of
5X10
mbar on sap-phire substrates at room temperature in an UHV system with abase pressureof
10 mbar. Sputtering rates were calibrated by Rutherford backscattering (RBS)on single filmsof
Nb and NbZr. The compositionof
the NbZr lay-ers was checked by electron microprobe measurements and found to be 55at.
%
Nb and 45at.
%
Zr, in accor-dance with theRBS
results. The multilayer characterof
936 P.KOOREVAAR, W. MAJ, P.H. KES,AND
J.
AARTS 47samples are made in a symmetrical configuration
of
1Vdouble layers
of
Nb-NbZr and one extra Nb layer, with thicknessesof
db andd,
for Nb and NbZr, respectively.The samples are referred to as db/d,
/X.
To avoid sur-face superconductivity' and to prevent oxidationof
the Nb layers we added4.
5-nm-thick NbZr bottom and top layers. As was shown inRef.
10these thin NbZr layers do not inAuence the measurements in any way. Various sample parameters are given in TableI.
The behavior
of H,
~~~(T)andH,
2i(T)was equivalent tothat
of
similar samples in previous work.'
From this we infer that theGL
coherence lengths atT
=0
for Nb and NbZr areg&(0)=12
nm and g,(0)=5.
5 nm, respec-tively.The samples were wet etched into strips with a width
of
0.
15 mm and mounted on a rotatable sample holder. The angular resolution is better than0.
3' and sample alignment is better than 1.
For
our current-voltage(I
V) experim-ents three diFerent configurations were used, as shown inFig.
2. The current is always applied along the layers, and the sample holder can only be rotat-ed over one angle. In configuration 1(Cl)
the angle be-tween field and sample surface0,
can be changed over 200 while the angle between current and field OIII remains fixed at90'.
This implies that the magnitudeof
the Lorentz force FL on possible vortices remains un-changed, but the direction
of
Fl
does change with chang-ingH„FL
being perpendicular to the layers for0,
=0
and along the layers for9,
=90'.
In configuration 2 (C2) the field is always parallel to the layers(8,
=0)
but Ol~ can be changed over200'.
FI
will always be perpendicular to the layers and will vary assin(9I~).
In configuration 3 (C3) both angles are equal:0,
=Ol~.
For Fz
this means that it is constantly directed along the layers but its mag-nitude changes~
sin(8, ).As the concept
of
kinked vortices consistingof
vortex strings parallel to the layers and vortex disks perpendicu-lar to the layers will play an important role later on, it is useful to remark on the componentsof
FL on strings and disks. Since the current is applied along the layers, the disks experience a Lorentz forcefz
in thex
direction for all configurations. The Lorentz force per unit length on a string,f„points
perpendicular to the layers and is pro-portional to sin(gl„„.
„),
with Ol„„„s
the angle betweencurrent and the string. In
Cl,
therefore,f,
is always atmaximum; in C3 it will be zero, while in C2
f,
varies continuously. Note also that the angle betweenfz
and the directionof
the strings is(90'
—
Hl„„„s),
so thatfz
isalong the string direction in
C1.
A critical current
I,
was defined by an arbitrary 8-pV/cm criterion. The choiceof
this criterion does not affect the qualitative behaviorof
the data, as was con-cluded fromI-V
curves taken in the whole relevant field regime for some samples.I,
was measured by varyingH
at constant
T
for several8,
(inCl
and C3) or several Ol~ (in C2and C3), or sometimes by varying0,
at constantT
and
H.
III.
THEPEAK FIELDAND AN ANALYSIS OF THELORENTZ FORCESA. The peak field Hp
Typical results for
I,
as a functionof
H
in configuration C1 for various temperatures, all in the 2D regime, are shown inFig.
3for sample 24-24-7. All these measurements are taken at0,
=6',
since an alignmentof
the field exactly parallel to the layers yields such high critical currents at lower fields and temperatures not close to
T,
zD that heating through the contacts becomesimportant. As we will show in Sec. IV
of
this paper, the qualitative behaviorof
I,
does not depend on0,
in alarge angular regime, but it strongly decreases with increasingO„which
enables us to follow the behaviorof
I,
in a larger temperature range by choosing0,
=6.
Concen-trating onFig.
3, we observe strong nonmonotonic be-haviorof
I,
with a peak and a dip. The peak fieldH
ap-pears to be strongly temperature dependent. The peak and dip inI,
are observed in both the 2D and the3D
re-gime, as far as measurements in the3D
regime were pos-sible in viewof
the heating problem.For
temperatures in the3D
regime,i.e.
,close toT„a
peak-dip effect was nev-er observed at any0,
.
For
sample 24-42-3 the temperature dependenceof
Hz (measured for4'& 8,
&14')isshown in aH~~-Tphase dia-gram inFig. 4.
The data can be described well byHp( T)
=Hq(0)+1
—
T/T,
2D .This dependence turns out to hold for all samples. We shall come back to this point when we discuss the
depen-TABLE
I.
Various sample parameters. The sample names denote the thickness ofthe Nb and NbZr layers (db andd„respective-ly)and the number ofdouble layers Nb-NbZr. Note that one extra Nb layer was added toobtain a symmetrical configuration, aswell
C1
50
kkkk k k k k k k ~ k k kkk0
~aa k k kk kk kkk0.
00
0.
2
5
0.50
0.
7
5
1.OO[T]
YFIG.
3.I,
(H) for sample 24-24-7 in configuration Cl at 0,=6'
for temperaturesT=9.
6 K(f),
T=9.
3 K (~
), and T=8.
8K(A)
dence
of
H
on the layer thicknesses in the second partof
this paper.
B.
The inAuence ofthe Lorentz forceIH
The first question to be answered with respect to the peak in
I,
isif
I,
is determined by vortex motion. One way to test this is to see how 0IH influencesI,
.
There-fore, we mounted samples in configuration C2, so that0,
=0
and 0IH can be changed.It
should be noted that in our experimental setup the sample holder can only be ro-tated over one angle. This means that0,
may slightly(
=
1')
change during the experiments, having asignificant effect on the magnitude
of
I„as
will be shown in Sec.IV.
Still, much insight can be gained. InFig.
5 we show the resultsof
I,
versusK
for sample 24/42/7 for various0I~
atT
=9.
8K,
which is in the 2D regime for this sample.First we note that for fields above
H
=0.
18T,
I,
is al-most independentof
0IH, showing that the componentof
I"I
perpendicular to the layers is ineffective in this field regime.For
fields belowK
it is clearly seen thatI,
is influenced by 0IH. The peak-dip effect is stronger forV V V V V V V V V V V V V ~2D-5D ~ego
FIG.
2. The three possible measurement configurations with in the upper part the vector diagrams for currentI,
field H,andLorentz force FL. The layers are parallel to the xy plane and the current is always supplied along the y axis. In the lower part are indicated the Lorentz forces on a string
(f,
)and a disk (fd) when the vortex is kinked, and has field components FI, (parallel) and Hd (perpendicular).4
5
6
7
8
9
1031
[K]
FIG.
4. The H~~-Tphase diagram for sample 24-42-3. Opentriangles
(0)
denote H,2~~ and open circles(0)
the 2D phaseline, where an abrupt change of
I,
occurs. The line isafitofthe H,2~~ data from the 2D regime to Eq. (1). fhe crosses(+)
938
P.
KOOREVAAR, W. MAJ, P.H.KES,ANDJ.
AARTS 470.04
0.
03
0.02
0.0
I 0 o V 0 0 V 0 a 0 CI 00 &8 8& e +a'8,
0.
00
0.0
0.
2
0.4
FIG.
5.I,
(H) for sample 24-42-7 and T=
9.8 K inconfiguration C2 at parallel field for 8&H=0'(+),
HE=8'(0),
~iH= l5('7)
»=30'(o),
OiH=60'(6), and OrH=90
(larger O~H and it disappears for
Os=OrH=0
This leadsto the conclusion that for fields below
H,
I,
indicates motionof
vortices perpendicular to the layers. AboveH
such motion does not occur, and the peak therefore ap-pears to signify the presence
of
a strong intrinsic pinning mechanism. AboveH,
then,I,
must be determined by another mechanism. Here we propose that not only strings but also disks exist, andI,
aboveH
is due to the motionof
the disks. In the ideal situation for0,
=0
one would not expect vortices perpendicular to the planes to be present. However, besides the fact that the sample can always have a small misalignment from mounting, more important isthat our multilayers are not grown epitaxial-ly, and that the planeof
the layers therefore can vary, making it impossible to align the sample exactly with the field.Similar measurements
of
I,
as functionof
0&H in the3D
regime at fields well aboveH
around the 2D phase line are shown for sample 24-24-7 in Fig.6.
SinceI,
is the same for 0&~=90'
and Oz~=0
this experimentconfirms the observation in
Ref.
12 that above the 2D phase line the Lorentz force perpendicular tothe layers is ineffective. This indicates that both above and below the5 E LJ LLl
0
7 O eg~«g
3444
L =Y uX ~1 at ge q a+eL ~asia ~ 4 ~ ~ ~ ~ ~e e—
9—
1.
65 I I—
1.
60 log [I/(A)j—
1.
50 E LJ LLI O O—
7
6o g4o 2O V I~ 0 0.0 0.1 0.20.3 0.4~,
H [T] .--t
kk La~ k g ~ ~ ~ ~2D phase line
I,
isdetermined by the Howof
disks along the layers and the phase line then denotes a shiftof
the disks to the other typeof
layer. This will be discussed in the next section.The conclusion that
H
separates a regime where vor-tices move across the layers from a regime where they only move in the layers is an important one. Therefore we made another direct testof
the perpendicular motion by measuring the transverse voltage over the widthof
a sample, in the case that Oz„„„
isnonzero.If
the current with densityj
is along the y axis,j=(0,
j,
O) andB=(B„,
B,
B,
), then the Lorentzforce
FL=jXB
hascomponents
j(B„O,
B,
—
).If
the.vortices would move along FL with a velocity v=
(v„O,
U,)this would yield anelectric field
E=(v,
B», U,—
B„+U„B„v,
B»—
). Results described previously indicate that aboveH
no vortex motion perpendicular to the layers exists, so U,=O
andE„=
0.
BelowH
vortex motion along the z axis is ex-pected, and a transverse voltage should be seen when By is nonzero. ThereforeE„
is expected to rise drastically belowH . To
measure the transverse voltage wepat-5
Ot 2D Q- Gf-l/ c2//—
2.7
—
2.
2
log fl/(A)]—
l.
7
3.
0
2.
0
2.
5
FIG.
6.I,
(H) at fields around the 2D phase line, atT=8.
8K, for sample 24-24-7, for 0,
=0,
and for8»=90(O)
or ~JH=0(+).
FIG.
7. (a)Longitudinal(A)
and transverse (~
)electric fieldE
as function ofcurrentI
with afieldpoII=0.
09T. The inset of(a) shows the measurement configuration with an indication ofthe directions ofE,I,
HI~ and FL. (b) Same as (a), for afieldC3
4
E ~ ~~~&~ ~~g ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~5 ~ ~ ~ ~ sI~a O.OI.
OFIG.
8. A comparison betweenI,
C,'H) curves measured inconfiguration C1 and C3 for sample 24-24-7 at 8,
=14
andT=9.
25K.
terned a sample with db
=d,
=24
nm andX
=7
into theHall pattern shown in the inset
of Fig.
7(a). During the measurements the componentof
H
along the layers al-ways made an angleof
45 with the current direction [see insetof Fig.
7(a)],so thatE
=E
is expected when the sample behaves isotropically and8,
is small. The sample showed usual multilayer behavior with aT,2D=9.
80K
and usual
I,
(H)
behavior. This is shown in the insetof
Fig.
7(b) for a measurement atT=9.
75K
and0,
=10'
where a clear peak and dip are seen.For
fields ranging from above the peak to below the lo-cal minimum inI,
the transverse(E„)
and longitudinal(E~)
electric field were measured simultaneously as a functionof
applied current. Typical results forH
=0.
09T
andH
=0.
22T
are shown in Figs. 7(a) and 7(b), re-spectively.It
isobserved that forH
=0.
22T,
E„
ismuch smaller thanE
in the whole current regime, whereas forH
=0.
09T
the difference has almost disappeared for all but the smallest currents. Interesting in this respect isto compare the ratio R=E,
/E
at the current where the 8-pV/cm criterion forI,
is fulfilled. This ratio increases rapidly from less than7X10
atH
=0.
22T,
via0.
13 forH
=0.
15T
to0.
75 atH
=0.
09T
again illustrating that vortex movement perpendicular to the layers is im-portant belowH
.
The Lorentz force perpendicular tothe layers, working on the strings, can be eliminated by mounting the sample so that the strings and the current are parallel to each other,
i.
e, configuration C3 as described inSec.
II.
The force parallelto
the layers is determined byO„and
is equal to the parallel force in configuration C1 at the same8,
. Figure 8 shows a typical comparison betweenI,
(H)
curves measured in configuration C1and C3, for samples 12-42-7 at
T
=9.
25K
and8,
=
14'.
AboveH
the curves are practically identical, as expected for disks moving along the layers only. BelowH
the results in C1 clearly deviate from those inC3.
This might be expected from the absenceof
string motion inC3,
but the C3curve still shows a clear plateau starting at Hz, indicating that the disk motion undergoes a change. We will discuss the im-portanceof
this below.C. The HII-Tphase diagram
From the results shown previously we have already drawn some conclusions concerning the nature
of
the nonmonotonic behaviorof
I,
as a functionof
field. Here we give the full picture from the measurements presented until now. Our starting point is the Takahashi-Tachiki model, which explains the behaviorof H,
&~i(T) by shiftsof
the maximum in the order parameter from one typeof
layer to the other. In our case this means that at
H,
2II in the 2D regime the maximum lies in the Nb layers but shifts to the NbZr layers in the3D
regime. We assume that the same shift occurs belowH,
2 in the3D
regime at the 2D phase line. Together with the line constituted by the fieldsH
(T)and the Meissner phase we then suggestfive regions in the
H,
2II-Tdiagram where the vortexstruc-ture may differ, as indicated in
Fig. 1.
For
allT
(
T,
a Meissner phase (M) exists at very low fields. In the anisotropic Abrikosov (AA) region just above the Meissner phase for temperatures belowT,
2D,we suppose that the vortices are straight, as in an ordi-nary anisotropic superconductor. The modulation
of
the order parameter ~gI isweak in this region. The3D
cou-pled regime above the Meissner phase for temperatures between
T,
andT,
2D is similar to the AA region, and thedistinction might be only artificial.
Above the AA region a kinked vortex lattice exists for temperatures below
T,
2D in the kinked (K) region (seeFig. 1). The vortices now consist
of
disks with cores per-pendicular to the layers and strings with cores along the layers. The fieldH
(T)
separates the AA andK
region, and aroundH
the SVLT occurs, yielding a peak and dip in theI,
(H)
curves. In theK
region Ip~ is stronglymodulated, being maximum in the Nb layers. We there-fore suggest that the strings are in the NbZr layers to minimize the loss in condensation energy by admitting the
core.
The disks must then be located in the Nb lay-ers, which can also be viewed as aconsequenceof
the fact that the Nb layers behave asif
decoupled: in asingle thin film with d(2g
in a slightly inclined field, currents can only Aow parallel to the surface, resulting in a vortex lat-tice with an area per vortex (disk) proportional to1/
sin(0, ), as was shown by Thompson. 'At temperatures below T2D30 the
K
region is separat-ed from the decoupled NbZr (DNZ) region by the 2D phase line, as shown inFig. 1.
At this line the maximum in If~ shifts from the Nb to the NbZr layers when in-creasing the field. A question is whether in the DNZ re-gion ~f is fully zero in the Nb layers. Around the 2D phase line this may not be the case, but especially at low temperaturesH,
2II is so much higher thanH»
that=0
in the Nb seems probable. Then strings in the Nb canno longer exist, since they would require an inter-layer supercurrent. On the other hand, strings might ex-ist in the NbZr layers, which behave bulklike, but the ex-periments indicate the presenceof
disks. FollowingRef.
an-940 P.KOOREVAAR, W. MAJ, P.H.KES,AND
J.
AARTS nihilationof
such entities, in contrast toparallel motion.With this phase diagram the experiments on
I,
are ex-plained in the following way. The peak and dip are de-scribed to the SVLT setting in at a fieldH
.
AboveH
the strings are pinned by an as yet unspecified intrinsic pinning mechanism related to the layering, and they do not move normal to the layers. The pinning
of
the disks moving along the layers determinesI„which
explains the independenceof
I,
on OIH above Hz. We have alsoruled out the possibility that
I,
is the depairing current forH
)
H,
on which we will further remark in Sec.IV.
Also relevant in this discussion are measurements
of
I,
as a functionof
parallel field which we performed on single layersof
Nb and NbZr with thicknessesof
24 nm (not shown). In these monolayers OIH does not infiuenceI,
ei-ther, which should be the case when in the multilayer the layers are decoupled as argued above.
From the measurements with
(Cl)
and without (C3) perpendicular Lorentz force presented in Sec.III
B,
we conclude that atH
a real structural transition in the vor-tex lattice takes place. The plateau found in configuration C3 shows that the parallel force belowH
acts on entities other than the bare disks.
For
this, a change from kinked vortices to straight vortices is the most natural explanation.IV. THEMECHANISM FORHp
Several important issues are left unexplained by the model sketched previously. Especially, the mechanism which causes the intrinsic pinning above
H
and the valueof
H
have not been addressed, and we have also not specified whether aboveH
the strings and disks can be described as separate entities, orif
they are partof
a rigid vortex. The experiments discussed subsequently, treating the infl.uenceof
the layer thicknesses andof
the angle between field and layers onH,
are meant to clarify these points.A. The inAuence ofthe layer thicknesses
The observed square-root dependence isnot strong, how-ever, and a more unequivocal test lies in verifying the proportionality to A under the assumption that
H
is the simplest matching configuration (n=1,
p=0).
We prepared setsof
samples where db was varied withcon-stant
d„
for different valuesof d,
as given in TableI.
InFig.
9 we compare the values forH
(T),
scaled on(1
—
T/T,
2D)',
for the different sets. The figure shows that H~(T) follows the temperature dependence as de-scribed byEq.
(2), but more striking is that they group according to the thicknessof
the Nb layer only. In TableI
we list the values forH (0)
as well as for the ratioHz(0)/HzD(0),
which is=0.
4 and almost sample in-dependent. In other words, since HzD(0) is proportional tod,s-)
db, so isH
(0).
This is visualized in Fig. 10 where1/H (0)
is plotted against db. This dependenceof
H
on Nb layer thickness again excludes the possibility that the peak inI,
is caused by a matching effect.The question can still be raised whether matching was observed in similar experiments, especially in those on Pb/PbBi sinusoidally modulated multilayers, ' or on Nb/Ta multilayers. ' This appears not to be the case, other claims notwithstanding. In the experiments on Pb/PbBi, strong and temperature-dependent peaks were found in
I,
at fieldsH
which scaled with 1/A, not with 1/A.
Since layer thickness could not be varied separate-ly, this would be indistinguishable from a 1/d(Pb) depen-dence, equivalent to our 1/d& dependence. Peaks were also found in Nb/Ta multilayers and again did not follow matching conditions, but other systematics were not re-ported. Although not matching, the mechanism causing the peak inI,
is clearly avery general one.B.
The dependence of&~ on8,
We stated before that
I,
drastically decreases with in-creasing0,
but thatH
is independentof
0, .
This will be shown here. Typical results forI,
as a functionof H
in configurationCl
at various8,
are shown in Fig. 11(a)for sample 32-32-7 atT
=9.
3E,
which is the 2D regime for this sample. At I9,=
3.
8,
the peak atH
is clearly visible.An often discussed possibility for enhanced pinning in a multilayer isthe concept
of
matching''
the vortex lat-tice to the underlying multilayer periodicity A, which would lead to enhanced pinning by the layered structure at a matching field. In the isotropic case the matchingcondition takes the simple form Q
CU
0 L
H
=(&3/0/2A
)(n+p
+np)
(3)where
H
is the matching field, $0is the fiux quantum, and n and pare integers. In the anisotropic case the equi-lateral triangles are compressed andEq.
(3) takes the form3.
0
0.
5
0.
7
0 + 0 + 0 0 0 IH'
=H
i/m/M
(4)where m and
M
are the effective masses parallel and per-pendicular to the layers. 'Since no temperature-dependent quantities are in-volved in Eqs. (3) and (4), the matching field
H'
is tem-perature independent, which is not what we find forH .
FIG.
9.0~(T)
scaled by (1—
t,2D)' vs reduced temperature t,»
for samples 12-42-7(D),24-42-7 (A),24-24-7 (~
), 42-42-70.
8
04
—
10
0
10
20
30
40
50
d,
tern]
FIG.
10. The dependence of1/H~(0) on the thickness ofthe Nb layers db forall samples consisting ofseven building blocks.The line ismeant toguide the eye.
0.
05
0.00
0.0
0.
2
a)0.6
~,
H[T]
2 ~ ~ ~ ~ CU CU o U 0 ~~ ~ Ogg~ ~~.
y$isg l~~ -~inst ~igS e'ice~~~ski&~ VOa1lzJlo
~gl+~ ~pat~ a~+4g~+s 4y~ Oeg 0.25 0.50 (b) O141 I I
—
30
0
30 60 90
120
150
6
(deg)FIG.
11. (a)I,
(H) at T=9.
3 K for sample 32-32-7 inconfiguration C1and for diFerent values of0,
.
(a)0,=3.
8',(b)0,
=16,
( ) 0,=24',
(d) 0,=32',
(e) 0,=40',
(f) 0,=60',
(g)0,
=90.
(b)I,
atH~=0.
22 T versus 0, for sample 32-32-7atT=9.
3K.
The line shows afit to 1/sin(0, ). The inset shows all data except 0,=90,
scaled onI,
atH~=0.
22T.
Its magnitude decreases strongly with increasing
O„but
the field value does not change. This is shown more clearly in the inset
of
Fig. 11(b), whereI,
curves are scaled onI,
(H~).It
is remarkable that the dip inI,
ex-ists for angles0,
far away from parallel, up to 50' for some samples. The angle0„
for which a dip cannot be observed anymore depends on temperature—
the angle being smaller for higher temperatures.For
all samples at all temperatures investigatedO„was
smaller than60,
i.e.
, a dip for perpendicular fields as observed in the Pb/Ge system is never found. The decreaseof
I,
atH
as a function
of 8,
is shown in Fig. 11(b). The data fitI,
~
1/sin(8,
)very well, as indicated in the figure. Only around0,
=90,
or sometimes a few degrees away from perpendicular, a small deviation with a maximum occurs. This feature isalso observed in the angular dependenceof
H,
2of
most samples and is probably due to columnargrowth
of
our films with a preferential direction away from perpendicular. Not only atH,
but in a rather broad field regime aroundH
the critical currents scale as1/sin(8,
). Furthermore,Fig.
11(a) shows that for fields below the local minimum inI,
the magnitudeof
I,
does not depend on
0,
for4'(0,
(60'.
The strongly enhancedI,
for0,
close to parallel at these small fields may be due to surface pinning effects.From the angular dependence
of
H
andI,
we can again draw some important conclusions on the mecha-nism for the peak inI,
.
As mentioned before in Sec.III,
a possibility for explaining the peaks in
I,
might be that belowH,
I,
isruled by vortex motion, but that aboveH
no vortices exist because the core diameter is
of
the orderof
the layer thickness and the layers are decoupled. In that caseI,
would be the depairing current, which, how-ever, should not have the very strong1/sin(8,
) angulardependence, observed also for fields well above
H .
Moreover, the shape
of
theI-V
curves below and aboveH
is similar, again discouraging an interpretation in termsof
depairing.The final point in this paragraph is whether the main finding, the
1/
sin(8, ) dependence, is consistent with theearlier sketched picture
of
strings and disks. This obvi-ously depends on the pinning envisaged for the disks, since it is their movement which determinesI,
atH
.
For
instance, for a rigid latticeof
vortex disks, where only a small numberof
disks is pinned by planar struc-tures perpendicular to the layers, such as grain boun-daries in a columnar structure, it was shown by Takahashi and Tachiki ' thatI,
~
1/+sin(8,
). Using such amodel for point pins in the planeof
motion yieldsI,
~ 1/sin(8,
), but again onlyif
a small fractionof
the disks is pinned. Since at small angles the numberof
disks becomes small, this does not seem a good assumption. On the other hand,if
the disks are connected to strings in a kinked vortex, the freedomof
the disks for finding a point pin may be severely limited and the condition might actually hold.V. DISCUSSION
942
P.
KOOREVAAR, W. MAJ, P.H.KES,ANDJ.
AARTSexperiments described in Sec. IV favor this interpreta-tion. We will now turn to the question
of
the precise mechanism governing the transition,i.e.
, why above Hz the strings are pinned in the NbZr layers, while belowH
straight vortices exist which can move both normal and along the layers.
Apart from the possibility
of
matching, which was al-ready discarded, another possible explanation which can besimply put aside isthatH
signifies the transition from proximity coupled Nb layers to decoupled Nb layers. This would involve a field dependenceof
the proximity length in NbZr. The possibilityof
such a dependence has been predicted, but to our knowledge never observed. However,H
would then obviously depend ond„which
iscontrary to the experimental results.Pinning
of
the strings can in principle be furnished by the Nb/NbZr interface. In the same way as the interac-tionof
a vortex with its image field at a superconductor-vacuum interface leads to the so-called Bean-Livingston barrier, the interfaceof
two superconductors with different penetration depths k and Ginzburg-Landau pa-rameters ~ can pin a vortex. This was shown by Mkrtchyan etal.
, who calculated the change in energyof
a vortex as afunctionof
its position with respect to an interface between two half infinite layers. In itself, this mechanism isnot enough to explain our data, since also a change from a pinning to a nonpinning interface would have to occur atH .
In the model this is only possible through the field dependenceof
A,, which is different forthe two layers. The penetration depth in the Nb layers,
A.b, is expected
to
have a 2D field dependence,kb(H)=k(0)[1
—
(H/H,
2zb) ]'~,
where the penetra-tion depth in the NbZr,k„has
a3D
field dependence,A,
,
(H)=&(0)(1
—
H/H,
2zbz,) '~ . Here,H,
2zb is theparallel upper critical field for athin Nb film and
H,
2zbz, the critical field for bulk NbZr. Provided thatA,b(0)&A,
,
(0),
the divergenceof
A,„(H)
would reverse thissituation at some field
H*
belowH,
2~b. The numbers found for A,of
our Nb and the NbZr layers stronglydiscourage such an explanation. Using values for the slope
S
= —
d8,
2/dT
atT„and
the residual resistivity atT
=0,
pofor single filmsof
Nb and NbZr, combined with the relation for weak-coupling amorphous superconduc-tors (Ref. 24)v=3.
54X10
[pP']',
we obtain mb=4,Ab
(0) =48
nm for the Nb layers and~,
=21,
A,,
(0)
=
116
nm for the NbZr layers. We see that although indeed A&(0)&A,
,
(0),
both values are larger than the individual layer thicknesses in the multilayer, which will lead to some kindof
averaging. Also the values differ relatively little. The variation over the interface will therefore be small, leading toweak pinning properties. Finally, due to the small differences involved, the crossing fieldH*
lies nearH /H,
2zb=0.
9, faraboveH .
Until now we have been considering mechanisms which lead to the pinning
of
the string portionsof
the vortex without regarding the kinks; in other words, the kinks are not relevant for the pinningof
the strings, but they just happen to be observable after the strings are pinned. Another pointof
view is that the formationof
the kinked vortex structure is itself the pinning
mecha-nism for the strings. The kinked vortex, once formed, cannot move perpendicular to the layers because itis par-ticular to the NbZr/Nb/NbZr sequence. The formation energy
of
the kink would then serve as an effective pin-ning barrier, which disappears below afieldH
where the kinked vortex structure is no longer favorable over straight vortices.It
should be remarked here that this kink formation bears resemblance to the lock-in transi-tion proposed by Feinberg and Villard, ' but actually is not the same; the lock-in transition is driven by the per-pendicular field component and the lock-in field therefore strongly depends on the angle between field and layers. Rather, we believe that in our case the line energyof
the straight vortices should be compared to the elastic energy connected with the kink. This involves averaging over different parts in the multilayer, in which the modulationof
the order parameter plays a role. In such a competi-tion, perpendicular field components are hardly involved, which would explain the observation thatH
is indepen-dentof
0,
.
The full model should also explain thatH
is inversely proportional tothe effective thicknessof
the Nb layers, but this model is still lacking.VI. CONCLUSION
In conclusion, arather surprising picture has emerged. We have presented strong experimental evidence that in Nb/NbZr multilayers, in a field regime above the Meiss-ner phase and for a wide range
of
angles between field and layers, a transition takes place in which straight vor-tices change tokinked vortices consistingof
strings in the NbZr layers and disks in the Nb layers. Below the transi-tion fieldH
the straight vortices can freely move perpen-dicular to the layers, while aboveH
the strings are in-trinsically pinned and the disks can move parallel to the layers. This leads to a sometimes huge increase in the critical currentI,
.
The fieldH
depends on the thicknessof
the Nb layers and on the temperature, but not on the angle between field and layers. The transition therefore appears to be caused by a competition between the line energyof
the straight vortices (favored at low fields when the modulationof
the order parameter is low) and the formation energyof
the disks (involving the thicknessof
the Nb layers). At the parallel critical field for the Nb layers another phase line is encountered, where the disks shift from the Nb layers to the NbZr layers and the strings probably disappear. This isrejected
in nonmono-tonic behaviorof
the critical current.field clearly is a less relevant parameter. An open ques-tion at the moment is whether these phenomena may also be witnessed in SN multilayers such as Nb-Cu. Further experiments and the development
of
a theoretical description, which is still lacking, will establish the pa-rameter ranges where these e8'ects occur.ACKNOWLEDGMENTS
%e
would like to acknowledge useful and stimulating discussions withJ.
Mydosh,S.
Takahashi, andA.
Koshelev. This work is part
of
the research programof
the Dutch Foundation for Fundamental Research on Matter.Permanent address: Institute ofPhysics, Polish Academy of Sciences, A.Lotnikow 32/46, 02-668 Warsaw, Poland. D. S.Fisher, M.P.A.Fisher, and D.A.Huse, Phys. Rev.B43,
130 (1991)~
~I. Banerjee, Q. S.Yang, C. M. Falco, and
I.
K.
Schuller, Phys. Rev. B28,5037(1982).3S.
T.
Ruggiero, T.W. Barbee, and M. R.Beasley, Phys. Rev. Lett. 45,1299(1980).4D. Neerinck,
K.
Temst, M. Baert,E.
Osquiguil, C. Van Haesendonck, Y. Bruynseraede, A. Gilabert, andI.
V. Schuller, Phys. Rev.Lett. 67,2577(1991).5W.
R.
White, A. Kapitulnik, and M.R.
Beasley, Phys. Rev. Lett.66, 2826(1991).S.Takahashi and M. Tachiki, Phys. Rev. B33, 4620 (1986); 34, 3162 (1986).
~M. G.Karkut, V.Matijasevic, L.Antognazza,
J.
M.Triscone, N. Missert, M.R.
Beasley, and Q. Fischer, Phys. Rev. Lett. 60,1751(1988).J.
Aarts,K.
J.
De Korver, and P.H. Kes, Europhys. Lett. 12, 447(1990}.Y.Kuwasawa, U.Hayano,
T.
Tosaka, and S.Matuda, Physica C 165,173(1990).oW. Maj and
J.
Aarts, Phys. Rev. B44, 7745(1991}.J.
Aarts, W.Maj,K.
J.
de Korver, P.Koorevaar, and P. H.Kes,Physica C 185-189,2071(1991).
W.Maj,
K.
J.
De Korver, P.Koorevaar, andJ.
Aarts, Super-cond. Sci.Technol. 5, 483(1992).I
J.
R.
Clem, Phys. Rev.B 43, 7837(1991).' H. Raffy,
J.
C.Renard, andE.
Guyon, Solid State Commun. 11,1679 (1972); 14,427(1974); 14, 431 (1974).S.Ami and
K.
Maki, Prog. Theor. Phys. 53,1(1975).D.Feinberg and C.Villard, Phys. Rev. Lett. 65,919(1990). '7R. S.Thompson, Zh. Eksp. Teor. Fiz. 69, 2249 (1975) [Sov.
Phys. JETP42, 1144 (1975)].
B.
I.
Ivlev and N.B.
Kopnin, Phys. Rev.Lett.64, 1828(1990).~9P.R.Broussard and
T.
H.Geballe, Phys. Rev.B37,68(1988). A. S.Sidorenko, A.E.
Kolin'ko, L.F.
Rybal'chenko, V. G. Cherkasova, and N. Ya Fogel', Fiz.Nizk. Temp. 706,(1980) [Sov.J.
Low Temp. Phys. 6,341(1980)].IM. Tachiki and S.Takahachi, Solid State Commun. 72, 1083 (1989).
G.Deutscher and P.G.de Gennes, in Superconductivity, edit-edby
R.
D.Parks (Marcel Dekker, New York, 1969); Vol.II,
Chap. 17.
G.S.Mkrtchyan,