Charles Reichhardt,1
Alejandro B. Kolton,2 Daniel Dominguez,2
and Niels Gronbech-Jensen3,4
1Applied Physics Division and Center for Nonlinear Studies,
Los Alamos National Laboratory, Los Alamos, New Mexico 87545
2Centro Atomico Bariloche, 8400 S. C. de Bariloche,
Rio Negro, Argentina
3Department of Applied Science, University of California,
Davis, California 95616
4NERSC, Lawrence Berkeley National Laboratory,
Berkeley, California 94720
(Received 23 May 2001; published 5 September 2001)
For a vortex lattice moving in a periodic array we show analytically and numerically that a new type of phase locking occurs in the presence of a longitudinal dc driving force and a transverse ac driving force. This phase locking is distinct from the Shapiro step phase locking found with longitudinal ac drives. We show that an increase in critical current and a fundamental phase-locked step width scale with the square of the driving ac amplitude. Our results should carry over to other systems such as vortex motion in Josephson-junction arrays.
DOI: 10.1103/PhysRevB.64.134508 PACS number(s): 74.60.Ge, 74.60.Jg
An important difference of vortex arrays from charge-density-wave systems and single-degree-of-freedom systems (like a small Josephson junction) is that the displacement field acting on vortices is a two-dimensional (2D) vector. This means that displacements can be induced in two different directions, and therefore a new kind of phase locking is possible when an ac force is applied transverse to the direction of the dc force.
In this work we show analytically and numerically how phase-locking phenomena may occur when vortices are moving in a periodic potential and the ac force is applied transverse to the direction of the longitudinally applied dc force. The type of phase locking observed in this case is qualitatively different from the Shapiro-type phase locking observed when dc and ac forces are in parallel. We first show analytically the possibility of the phase locked states for certain commensurate vortex configurations and predict that the critical current increases quadratically with the ac amplitude and that the width of some of the phase-locked steps scale as the square of the ac amplitude. We also find scaling of the critical current as well as the first phase-locked region as a function of the ac frequency and the pinning geometry. The predictions of the perturbation analysis are confirmed with numerical simulations. We analyze the validity of the perturbation approach and demonstrate how deviations occur. Our results suggest that more pronounced transverse phase locking may be observed at lower commensurate magnetic fields, such as B=1.5 or B=1.25. The general model and perturbation results are easily generalizable to other systems which exhibit Shapiro-step-like phase locking such as Josephson-junction arrays and Frenkel-Kontorova-type models.
Our studies are directly relevant for several contemporary efforts since vortex matter interacting with nanostructured pinning arrays of holes10-12 and dots13,14 has been attracting increasing attention due to the easily tunable pinning properties. Pronounced commensuration effects are observed in these systems when the density of vortices matches to integer or fractional multiples of the density of pinning sites. In addition to square pinning arrays, recent experiments have been conducted on rectangular pinning arrays.14 Simulations have shown several interesting dynamical phases of dc-driven vortices in periodic pinning systems.15,16 Imaging12 and transport experiments4,11 along with simulations15 have found that the vortex motion above the first matching field can occur by the flow of interstitial vortices between the pinning sites. These vortices still experience a periodic potential created by the vortices located at the pinning sites. Recently, phase locking was observed for dc-and ac-driven vortices interacting with periodic pinning at B=2, where is the field at which the number of vortices equals the number of pinning sites,4 where the dc and ac drives were in the same direction. The values of the voltage response in Ref. 4 strongly suggest that is is the interstitial vortices that are mobile while the vortices on pinning sites remain immobile.
We are performing simulations with periodic boundary conditions (periodicity Lx, Ly), and with a logarithmic interaction potential [uvv(rij)=-ln rij] valid when is of the order of the sample size.
Combining the logarithmic interactions with the periodic boundary conditions
results in17
(4)
where (xij,yij) are the Cartesian coordinates,
r2ij=x2ij+y2ij,
and C(.) is a function determined by
the energy normalization of the long-range interactions.
We have previously demonstrated that phase locking of the vortex motion
can exist if a certain periodicity of the global pinning potential is
present. This was shown for driving forces of the form,
fd=(+sint),
where
is a unit vector in the x direction. Analogous to the
well-known Shapiro steps1
in the current-voltage characteristics
of an ac- and dc-driven Josephson junction, it was demonstrated that nonzero
intervals of
will result in phase locking of vortex motion with
vortex velocities resonating with the external ac field. We will in this
paper demonstrate that one can also expect phase locking in the transversely
ac-driven case, i.e., for
(5)
where and
are the magnitudes of dc and ac forces and
is the frequency of the ac drive.
The relevant equations of motion are, therefore,
, (6)
. (7)
A. Critical dc force without ac drive
The first result to evaluate for the purely dc-driven case is the "critical
dc current", ,
below which the interstitial vortex
is trapped by the pinned vortex lattice. As
is increased from
=0,
the interstitial vortex will find an equilibrium point for which the
pinned vortex lattice provides a cancelling force to the dc bias. However,
for a given (critical) value of
, namely,
, the pinned vortex
lattice can no longer provide enough force to resist
the motion of the interstitial,
which therefore begins to propagate in the x
direction. Assuming that the
pinned vortices do not move and that an interstitial vortex only interacts
with pinned vortices, this critical force can be found as the
maximum gradient value of Eq. (4) for
yij=mly.
The corresponding equation of motion is given by
Eq. (6):
(8)
This equation can produce a critical current by requiring
=0 and
optimizing the left hand side through varying the position xi.
In general, this must be done numerically.
The result is shown in Fig. 1, where the critical dc force (thick solid)
is shown together with the optimized value of the equilibrium value of
xi=x*. In the limit of
sech(ly/lx)1,
we can derive the approximate expression
of the critical current by noticing that only two terms (n=-1,0) are
required to describe the critical current, which is then always given
by xi=x*=(3/4)lx:
(9)
(10)
The approximate expression, Eq. (9), is shown in Fig. 1 as a thin solid curve, validating that the interaction potential between interstitial and pinned vortices can be simplified to the terms n=-1,0 for lx1.5ly. One can also see that in the extreme opposite limit lxly, the critical dc force can be approximated by /ly. This is shown as a dashed line in Fig. 1.
It is here important to remember that the analysis, which is based on a magnetic field of "matching field plus one flux quantum," is actually valid for magnetic fields < B 2, for which the vortex configuration results in forces between interstitial vortices cancelling due to symmetry; such magnetic fields can be, e.g., B=2, (3/2), (5/4), ...
B. Transverse ac drive included
Let us assume that
sech(lylx)1,
so that we can write the approximate equations of motion for the
interstitial vortices as
+[1+(1/2)k2yi2] sin(xik)= (11)
(12)
where we have retained terms in yi only to lowest order.
We have required
|yi|lx
in order to obtain
Eqs. (11) and (12).
The solution to the second of those equations is easily found,
yi = sin(t + ), (13)
with
, (14)
. (15)
Inserting Eq. (13) into Eq. (11) yields
+ [ 1 + (1/4)k2] sin k xi-(1/4)k2 cos 2(t + ) sin k xi = . (16)
This equation is the effective equation of motion for the vortex behavior
in the longitudinal (x) direction given an ac force in the
transverse (y)
direction, derived within the approximations listed above. The equation
is equivalent to that of an overdamped pendulum (
k xi being the pendulum phase) with dc torque
() and a pivot
vertically oscillating with frequency 2
and amplitude proportional to
- i.e., an overdamped
equivalent of the classic Kapitza
problem18 for underdamped
and parametrically driven pendulums.
We will now consider a few separate cases of vortex responses to the transverse ac force.
(i) < > 0 xi(t)=x0.
We will omit the last term on the left-hand side of Eq. (16).
The remaining (dc) terms are
sin k x0 = (17)
|| =[1+(1/4) k2]=+
, (18)
, (19)
where is the critical dc force.
It is here worthwhile to notice that the
critical current
increases
(
positive)
quadratically with the transverse ac amplitude
.
This is in direct contrast to the case when the ac drive is longitudinal,
in which case the critical dc force decreases
quadratically.2
We remember that validity of the expressions requires
lx.
(ii) < > > 0 || > .
Without the parametric term in Eq. (16), the solution,
for
,
is given by2
This range in bias current for which the average speed (voltage)
of the interstitial vortex is constant will manifest itself as a step in
the dc current-voltage characteristics of the system. It is important
to emphasize that this step is very different in origin from the Shapiro
steps5
recently demonstrated for longitudinal ac drive. The
essential difference lies in the fact that the Shapiro steps arise from
a "direct" driving term in an effective pendulum equation, whereas the
present phase locking for the transverse ac drive arises from an effective
parametric ac-driving term in the longitudinal equation of motion.
If one chooses a better (and more complicated) ansatz for the vortex
motion [e.g., Eq. (21) instead of Eq. (22)], it
becomes evident that an
phase-locked step also exists
for kv0=
and that higher-order phase-locked steps may exist at
any subharmonic and superharmonic of the driving frequency.
However, we will not
go into detail with other phase-locked modes in this presentation.
In order to validate the analysis of Sec. III predicting some basic effects of
a transverse ac force on vortices, we have conducted numerical simulations
of driven interstitial vortices in a rectangular lattice of pinned vortices.
Figure 3 shows a series of simulated (normalized) current (dc force) voltage
(average vortex velocity v0) for different values of the transverse
ac drive. The system is a square array of 4 x 4 pinned vortices with
periodicity (lx,ly)=(2,2)
in a computational simulation box of size
(Lx,Ly)=4lx,4ly). The simulated
magnetic field corresponds to B/=1.5
such that each pinning center is
occupied by a vortex and every other plaquette has an interstitial vortex. The
applied normalized frequency is =4.
The four curves (shifted vertically
for clarity) in Fig. 3 are for
(from top) =0,1,2,3. According to the
analysis above, the critical dc force
resulting in the onset of voltage
(vortex transport) should increase quadratically
for increasing .
This is clearly visible from Fig. 3.
A phase-locked step in dc bias current should develop for
increasing around
k v0=2. This is also clearly
visible for
2.5.
As the ac amplitude is increased we observe
more steps in the IV characteristics. However, while the simplest of these,
k v0=,
can be analyzed in some detail, we will be
concerned with only the steps at kv0=0
and kv0=2 here.
Two characteristically different vortex evolutions are shown as insets
to Fig. 3. At bottom right we show a snapshot
() of a vortex
configuration within the parameter range of the phase-locked step for
=2
together with a long time trace (thin line) of the vortex
trajectories. The vortex paths are obviously periodic and the interstitial
vortices seem to move in a geometrically simple configuration where all
interactions between the interstitials cancel due to symmetry.
The top left inset shows the traces outside of the phase-locked region.
Here we observe the pinned vortices as static dots, while the interstitials
move in a chaotic or quasiperiodic band between the pinning sites.
A detailed investigation of the validity of the above perturbation results
was performed for several different sets of system parameters. For a system of
a single interstitial vortex, we show in Fig. 4 the detailed comparisons
between the analytical predictions (lines), Eq. (19),
and numerical simulations (markers) of the
increase
in critical dc force as a function of ac amplitude.
Figure 4(a) shows data for square pinning arrays of different plaquette areas
(lxly=4,16) and with different driving
frequencies (=4,8),
while Fig. 4(b) shows comparisons for rectangular systems
(lx/ly=1/2,2) of one plaquette area
(lxly=4) for
different frequencies (=4,8).
The details of the parameter sets are
given in the caption. The agreement between simulation and analysis is
obviously good overall for these different parameters. Certainly the
quadratic relationship between the increase of the critical current and
the ac amplitude is confirmed. The quantitative agreement is also rather
good. The simulation results seem underestimated by the predicted values.
However, this is to be expected since the analysis only includes the transverse
vortex motion as moving in a harmonic potential. The vortex model will
provide a stronger nonlinearity for larger ac amplitudes, resulting in
a larger than predicted increase of the critical current.
For the same parameter sets as shown in Fig. 4, we have performed
simulations of the phase-locked step at
kv0=2 and compared the
results to the predicted range
given in Eq. (25).
The results, shown in Fig. 5, are also here in good agreement with the
predictions. Thus, we are confident that the above analysis of the single
interstitial behavior can be a useful tool for understanding phase locking
in this system.
Figure 6 shows, again for the same parameter sets as in Fig. 4, the
comparison between the magnitude of the phase-locked step,
,
and the predicted value, but in this case for
B=1.5, i.e., for
one interstitial vortex per two plaquettes of pinned vortices. Once again
we find rather good agreement between simulations and the simple prediction.
However, we notice that rectangular systems with
lx=2ly [top curves
in Fig. 6(b)] cannot produce a measurable phase-locked step for very small
ac amplitudes.
If one simulates the system for the same parameter values, but for
B=2
(not shown), the agreement will get rather poor for all rectangular cases
where lx=2ly
and even for some of the square array cases as well. The
reason for these discrepancies is of course found in the assumptions behind
the perturbation analysis. Particularly, the validity of the
assumption of noninteracting interstitial vortices will break down for
densely populated
(B=2) systems since the vortices are mutually
repulsive. The result is sketched in Fig. 7, where the vortex repulsion
is deforming the geometrically simple lattice into a configuration where
the effective interstitial vortex-vortex interaction is nonzero. The lattice
deformation is characterized by the two parameters,
and .
Figure 8 shows simulation results for
Lx=4lx=Ly=4ly=8,
=4, and
B=2.
The circular markers ()
show the simulated range
in phase locking and the solid straight line represents the prediction
of Eq. (25). It is apparent that the agreement between simulations
and prediction is only fair
()
for relatively large values of
. Measuring the average (in
time and space) "phase difference"
< || > and
< || >, we can correlate the magnitude of
lattice deformation with the comparison between simulation and analysis
of phase locking. The measures of lattice deformation are defined as
Finally, these averages are averaged over the phase-locked step to provide
a single measure for a given
( < || >
and
are typically smallest near the center of the
locking range). Thus,
0 < || > lx,y/2
is a measure of the average spatial lattice deformation and
is a measure of the average temporal fluctuations in the lattice deformation,
||.
The minimum averaged
lattice deformation parameters for a given phase-locked step are shown
in Fig. 8 as solid markers
(,
< || >;
,
< || >). We observe large lattice
deformations at small ac amplitudes, where the agreement between the
observed range of phase locking is poor, while the deformations "collapse"
for larger values of ,
thereby giving rise to better agreement
between simulations and analysis. We have also shown the average of
the temporal fluctuations
and in order to
illustrate that the lattice deformations depend on time when
< || > and
< || >
are nonzero. Thus, it indicates that internal modes of the lattice
deformation are excited and may contribute significantly to the
phase-locking range in the low ac amplitude cases. For large ac amplitudes
we observe a saturation of the range in phase locking accompanied by
the reappearance of the deformation lattice. This is consistent with chaotic
behavior which are expected for strong ac-driven nonlinear systems.
We have developed a simple analysis of the behavior of interstitial
vortices in systems with periodic pinning and transverse ac drive. The
result indicates that the dc-driven longitudinal motion can phase lock
to the transverse ac signal and that the range of phase locking
(in dc current) is
quadratic in the ac amplitude and we have developed a quantitative
expression providing detailed dependencies of also other relevant system
parameters, such as pinning geometry and driving frequency.
We have further demonstrated that the
critical current increases quadratically with the ac amplitude. The
perturbation results have been validated by numerical simulations which
show good agreement with the analytical predictions in the expected
range of system parameters. The mechanism of phase locking discussed in
this paper is distinct from phase locking to a longitudinally applied
ac signal. The latter case was studied in
Ref. 5 and exhibits
qualitatively different responses to ac perturbations, such as decreasing
critical dc current with increasing ac amplitude and phase-locked steps
that grow linearly with ac amplitude. The results shown in this paper
are for commensurate fields. Incommensurate fields, where no simple
geometrical relationship can exist between the pinned and interstitial
lattices, are characterized by noncancelling interactions between
interstitial vortices. Thus, phase-locking ranges for the noncommensurate
fields usually have magnitudes less than predicted by the single
interstitial analysis.
Our results and analyses indicate that the most important phase locking
appears at relatively small magnetic fields (with respect to the first
matching field) where the inter-vortex repulsion
is not deforming the interstitial vortex lattice.
The predictions in this paper should be directly applicable for experimental
verification in superconductors (or Josephson-junction arrays) where
dc and ac fields are orthogonal.
Back to Home CR
(20)
(21)
where
is the dc force when the parametric term in Eq. (16)
is neglected. Let us assume the following simple ansatz for the vortex motion:
xi(t) = v0 t + x0, (22)
where x0 is a constant.
Inserting this ansatz into the effective equation of
motion, Eq. (16), while only maintaining dc terms,
yields
v0-
With the ansatz of Eq. (22) we can therefore expect the term
<.> to contribute to
if
kv0=2 (see Fig. 2),
and the resulting relationship between the internal phase
k x0 - 2
and the dc current
+
is
2/k- (1/2) sin(k x0-2) = +. (24)
As the phase k x0-2
can be adjusted to balance this equation
for different choices of
,
we can argue for a nonzero range
in dc force for which the average speed of the interstitial vortices is
unchanged. This phase-locking range has the magnitude
= . (25)
Thus, we can predict phase locking in the transversely ac-driven vortex
system and the predicted total range in phase locking is equal to the increase
in the critical dc force due to the ac drive. This prediction is correct
up to and including terms
.
IV. NUMERICAL SIMULATIONS
FIG. 3.
Simulated IV characteristics for different values of transverse
ac amplitude,
=0,1,2,3 (top down).
IV curves are vertically
offset for clarity. System parameters are
Lx=Ly=4lx=4ly=8,
=4,
and B=1.5. Lower inset shows vortex trajectories
in a phase-locked state while upper inset shows trajectories outside
a phase locked regime.
FIG. 4.
Increase of critical dc current,
, as a function
of ac amplitude .
Markers represent numerical simulations
of Eqs. (6) and (7). Solid lines are the
corresponding predictions from Eq. (19). Simulations are
conducted for a single interstitial vortex.
(a)
, lx=ly=2
and =4;
, lx=ly=2
and =8;
, lx=ly=4
and =4;
, lx=ly=4
and =8.
(b)
, lx=2ly=4
and =4;
, lx=2ly=4
and =8;
, 2lx=ly=4
and =4$;
, 2lx=ly=4
and =8.
FIG. 5.
Magnitude
of the phase-locked step at
kv0=2
as a function
of ac amplitude .
Markers represent numerical simulations
of Eqs. (6) and (7). Solid lines are the
corresponding predictions from Eq. (25). Simulations are
conducted for a single interstitial vortex.
(a)
, lx=ly=2
and =4;
, lx=ly=2
and =8;
, lx=ly=4
and =4;
, lx=ly=4
and =8.
(b)
, lx=2ly=4
and =4;
, lx=2ly=4
and =8;
, 2lx=ly=4
and =4;
, 2lx=ly=4
and =8.
FIG. 6.
Magnitude, ,
of the phase-locked step at
kv0=2
as a function
of ac amplitude .
Markers represent numerical simulations
of Eqs. (6) and (7). Solid lines are the
corresponding predictions from Eq. (25). Simulations are
conducted for B=1.5,
Lx=4lx, and Ly=4ly.
(a)
, lx=ly=2
and =4;
, lx=ly=2
and =8;
, lx=ly=4
and =4;
, lx=ly=4
and =8.
(b)
, lx=2ly=4
and =4;
, lx=2ly=4
and =8;
, 2lx=ly=4
and =4;
, 2lx=ly=4
and =8.
FIG. 7.
Sketch of interstitial vortex lattice deformed due to the
repulsive vortex-vortex interaction at B=2.
The parameters
and are measures of the magnitude of
deformation from a rectangular
lattice.
FIG. 8.
Magnitude
of the phase-locked step at
kv0=2
as a function
of ac amplitude
for Lx=4lx=Ly=4ly=8,
=4,
and b=2 ().
Markers represent numerical simulations
of Eqs. (6) and (7). Solid line represents the
corresponding prediction from Eq. (25). Also shown are the
lattice deformation measures
< || > and
< || >, as well as their standard deviations
and .
= (xi mod lx), (26)
= (yi mod ly), (27)
, (28)
, (29)
< || > = < || >t,i,j (30)
< || > = < || >t,i,j, (31)
, (32)
. (33)
V. CONCLUSION
ACKNOWLEDGMENTS
Parts of this work were supported by
ANPCYT (Proy. 03-00000-01034), Fundacion Autochas (Proy. A-13532/1-96),
Conicet, CNEA and FOMEC (Argentina),
and by the Director, Office of Advanced Scientific Computing Research,
Division of Mathematics, Information and Computing Sciences, U.S. Department
of Energy Contract No. DE-AC03-76SF00098.
Phys. Rev. B 64, 134508 (2001).
Back to Home CJOR
Last Modified: 7/5/02