Periodic_Forcing.pdf

PHYSICAL REVIE%' E VOLUME 51, NUMBER 6 JUNE 1995
Periodic forcing of a Brownian particle
Luc P. Faucheux, Gustavo Stolovitzky, and Albert Libchaber
Center for Studies in Physics and Biology, Rockefeller Uniuersity, 1230 York Auenue, New Fork, New Fork 10021
(Received 28 October 1994;revised manuscript received 17 February 1995)
We study the effect on a Brownian particle (2 pm diameter polystyrene sphere in water) of an infrared
optical tweezer moving in a circle. For a given potential depth of the optical trap, three different re-
gimes for the particle motion are observed as a function of the trap velocity. For small velocity of the
tweezer (typically (100pm/s), the particle is trapped and moves with the beam. For intermediate ve-
locities (between 100 pm/s and 3 mm/s), the particle escapes but is caught by the returning trap: its
mean angular velocity scales asymptotically as the inverse of the trap rotation frequency. For 1arge
tweezer velocities ( & 3 mm/s), the particle diffuses along the circle but is confined in the radial direction.
We describe these observations by a simple deterministic model. We justify the use of this model solving
the corresponding Fokker-Planck equation.
PACS number(s): 05.40.+j
I. INTRODUCTION
Focusing a laser beam creates an inhomogeneous inten-
sity profile. This attracts a small dielectric particle near
the beam focal point (region of highest intensity), as
shown in Fig. 1. The trapping force for a 2 pm diameter
polystyrene sphere in water is of the order of 1 pN I1,2].
As one moves this optical trap, the viscous drag (Stokes
force), proportional to the particle velocity, will eventual-
ly cause the particle to escape from the trap. This hap-
pens in our experiment for a critical trap velocity of the
order of 100pm/s. The question we address in this paper
is the following: What happens to a Brownian particle
subjected to an optical trap moving faster than the criti-
cal velocity? In particular, does one still induce particle
motion, or is the particle free to diffuse'? To answer these
questions we chose to study the particle response when
moving the trap along a 12.4 pm diameter circle. In this
geometry, the trap always returns before the particle
diffuses away. It confines the particle motion to the cir-
cle. The particle angular displacernent over a long time
period is easily recorded, and is the main observation of
our paper.
Three regimes are observed as a function of the trap
velocity. For small velocities (below 100 pm/s), the
viscous drag is smaller than the maximum trapping force:
the particle is trapped and follows the beam. For veloci-
ties between 100 pm/s and 3 mm/s, the trapping force
does not overcome anymore the viscous drag and the par-
ticle escapes. However, it does not have enough time to
diffuse away before the trap returns and kicks it for a
brief instant. The mean particle angular velocity scales
asymptotically as the inverse of the trap rotation frequen-
cy. Finally for velocities above 3 mm/s, the effect of the
'Also at NEC Research Institute, 4 Independence Way,
Princeton, NJ 08544.
trap is smaller than the thermal noise, and no net drift of
the particle is observed on time scales of the order of 2
min. The particle diffuses freely, still conPned in one di
mension along the circle. To summarize, the system
evolves from a synchronous rotor to an asynchronous
one, and finally to a free diffusive motion, always local-
ized on the circle.
We present a theoretical treatment of the problem. As-
suming the particle motion to be one dimensional, we
study the response of a Brownian particle to a periodic
forcing by a potential of arbitrary shape. Solving the
equations of motion in the deterministic limit, where the
stochastic term is assumed to be negligible, we recover
the observed behavior for the mean particle angular fre-
quency and for the mean residence time inside the trap.
This is a surprising result considering the highly erratic
FICx. 1. A 2 pm diameter polystyrene particle trapped near
the focal point of a Gaussian laser beam propagating down the
vertica1 axis. The gradient forces are shown on each axis as the
thin solid arrows. The radiation pressure is indicated by the
thick solid arrows.
1063-651X/95/51(6)/5239(12)/$06. 00 51 5239 1995 The American Physical Society

FAUCHEUX, STOLOVITZKY, AND LIBCHABER
appearance of the actual particle trajectories. Solving the
Fokker-Planck equation, we study the effect of tempera-
ture and potential depth on the particle's motion. We
justify the use of a zero temperature (deterministic) ap-
proach.
The paper is organized as follows. In Sec. II, we de-
scribe the optical tweezing technique and measure the
profile of the trap. The trap is kept fixed in space. In
Sec. III, we move the beam along a circle and record the
angular displacement of the particle. We discuss the
three regimes previously described. In Sec. IV, we
present the deterministic and Fokker-Planck equations
for this problem. We compare their predictions with the
experimental results.
II. FIXED OPTICAL TRAP
We first present the sample preparation and the optical
setup. We then describe the measurements on a fixed op-
tical trap. Moving the microscope stage, we measure the
maximum trapping force. Watching a particle fall into
the trap, we extract the potential profile.
A. Experimental setup
Commercial suspensions of 2 pm diameter polystyrene
spheres [3] are diluted in pure water [4]. The final
volume fraction (10 ) is such that hydrodynamic in-
teractions are negligible. Typically only a few beads are
seen in the microscope field of view (50X50 pm ). Mylar
sheets, 50 pm thickness, are cut and used as spacers be-
tween a microscope slide [5] and a coverslip [6].
G1assware are cleaned and dried using a nitrogen ionizing
gun [7]. The cells are then filled with a suspension of
spheres and sealed with fast epoxy [8] to avoid any con-
vective Qow. The sample, placed on the translation stage
of an upright microscope [9], is observed under bright-
field illumination. The particles tend to stay confined
close to the bottom glass plate because of the gravitation-
al field g. An estimate of this sedimentation effect is the
Boltzmann length scale: L~=k~T/hmg, where hm is
the relative mass of the particle. Lz is the average verti-
ca1 excursion of the particle caused by thermal Quctua-
tions. It is of the order of 2 pm. The vertical density of
particles within the sample is then exponentially peaked
within the first 2 pm close to the bottom plate. The im-
age is recorded by a charge coupled device (CCD) camera
[10]. The particle appears dark on a clear background.
The video signal is processed in real time via a video
board [11]. The pixel intensity is thresholded around the
m.ean image value. From computing the center of gravity
of the intensity distribution, the X and Y coordinates of
the center of mass of the particle are deduced. The sam-
pling rate is the video acquisition rate 30 Hz. The spatial
resolution is 0.1 pm. The Z coordinate is not recorded:
the particle never gets out of focus when running the ex-
periment.
B. C)ptical trap
500
400-
300
ZOO
100
0
0 0.5
output power (W)
'I.5
1064 nm) is inserted into the microscope's optical path
via the beam splitter B shown in Fig. 5. An 100Xoi1 im-
mersed objective (QBJ) [13]focuses the beam to a sharp
focal point inside the sample. The refractive index of the
polystyrene sphere is larger than the water index; the par-
ticle is thus attracted to high electric field regions. The
transverse Gaussian profile of the beam pulls the particle
towards the beam axis. The converging beam develops a
field gradient along its axis which opposes the radiation
pressure. The forces acting on the sphere are shown in
Fig. 1. The gradient forces localize the particle to the fo-
cal point, and the radiation pressure pushes the particle
along the beam axis. When the gradient forces overcome
the radiation pressure, the beam focal point becomes a
trap, an optical tweezer, for the polystyrene sphere. This
requires a microscope objective with a high enough nu-
rnerical aperture, 1.3 in our case. The strength of the
trap depends linearly on the output laser power. Because
of optical losses due to rejections from the lenses and
beam splitters along the optical path, the laser power at
the level of the sample is 15 times smaller than the laser
output.
We measure the maximum trapping force by moving
the sample in a direction transverse to the optical axis.
The particle escapes when the sample velocity is such
that the Stokes force exceeds the trapping force. The
critical velocity Vz as a function of the output laser
power is shown in Fig. 2. The maximum trapping force
is 6~gaVC, where g is the room temperature water
viscosity (10 kg/ms) and a the particle radius. The
value of the force is of the order of 1 pN for an output
laser power of 150 mW. Taking 1 pm as a typical length,
the trapping potential is then of the order of 300k~T.
The maximum trapping force depends critically on the
position of the beam focal point with respect to the bot-
tom glass plate. This is due to the laser reAection from
the water-glass interface, which reduces the overall radia-
tion pressure on the particle. Shifting upward the verti-
cal position of the beam focal point by less than 5 pm de-
creases the rnaxirnurn trapping force by a factor of 5.
%'hen running the experiment, we constantly check the
distance between the beam focal point and the glass plate
by imaging the laser reAection from the glass plate, and
The TEMOO mode of a neodymium-doped yttrium
aluminum garnet (Nd:YACx) laser [12] (wavelength
FIG. 2. The critical velocity Vc as a function of the output
laser power.

PERIODIC FORCING OF A BRO%'NIAN PARTICLE
adjust it to within 0.1 pm.
Another measure of the potential is obtained by ob-
serving the fall of a particle into the trap [14]for an out-
put laser power of 150 mW (laser power at the level of the
sample of the order of 10 mW). Figure 3(a) shows the
time series of the particle position, when initially close to
the trap. The trap is fixed and defines the axis origin.
When random diffusion brings the particle to within a
few micrometers of the trap, the particle is attracted and
falls into it. The equation of motion for the particle is
d-=
m Vp= —
6~gaVp VU+—
f„„h,
dt
where Vz is the particle velocity, U the trapping poten-
tial, and f„„h
the stochastic thermal forcing. The parti-
cle velocity when falling into the trap is of the order of 6
pm/s. The viscous drag force is then of the order of
10 ' N. The particle acceleration is of the order of 10
pms . The inertial term is then of the order of 10
N. At any instant, the velocity of the particle is such that
the inertial term (left hand side of the equation) can be
neglected: the particle is in the overdamped regime (low
Reynolds number). Also the stochastic forcing f„„h
is
assumed to be much smaller than the trapping force. As
a result, at each point of the trajectory shown in Fig. 3(a),
the Stokes force equals the trapping force. A spatial in-
tegration along the trajectory recovers the trapping po-
tential:
U(p) = 6n—
ala I Vp(p')dp' .
It is plotted in Fig. 3(b) as a function of the position of
the center of mass of the particle relative to the center of
the trap. The symmetric of this curve with respect to the
origin is also plotted for visual convenience. The depth
of the potential is 250k&T for an output laser power of
150 mW, consistent with our previous estimate. This
value justifies a posteriori neglecting the stochastic forc-
ing. Measuring the potential profile for higher output
laser power is not possible because of the finite video sam-
pling rate, 30 Hz: a stronger trap pulls the particle fas-
ter, and decreases the experimental resolution on the par-
ticle trajectory. %'e thus assume that the potential depth
scales linearly with the output laser power. This is
confirmed by the linear dependence of the critical veloci-
ty Vc on the output power (Fig. 2).
Heating effects are discussed in Appendix A and are
shown to be negligible.
III. OPTICAL TRAP MOVING ON A CIRCULAR PATH
We now move the beam along a circle of diameter 12.4
pm, as shown in Fig. 4. %'e first describe the experimen-
tal setup and the three observed regimes for the particle
motion.
~ N ~
0
4
0
II
2 3 4
time (s)
A. Experimental setup
Two mirrors M1 and M2 are mounted on galvanome-
ters [15] oscillating around two perpendicular axes (Fig.
5). This moves the trap in the XYplane transverse to the
optical axis. Two telescopes T1 and T2 allow the beam
to pivot about the back iris diaphragm of the microscope
objective as the mirrors oscillate (thus preserving the
Gaussian beam profile) [16]. The mirrors are synchro-
nously driven by a sine wave signal from a function gen-
-200—
O -3OO 1
p (p,m)
FIG. 3. (a) Time series of the horizontal displacement of a
particle falling into the optical trap. The trap is Axed at the ori-
gin. Output laser power 150 mW. (b) Trapping potential in
units of k~T for a 2 pm diameter polystyrene sphere. The X
axis represents the relative horizontal distance between the
particle's center of mass and the beam focal point. Output laser
power 150 mW. Laser power at the entrance of the cell 10 mW.
FIG. 4. Rotation of the optical trap along a circle of diame-
ter 12.4 pm.

5242 FAUCHEUX, STOLOVITZKY, AND LIBCHABER 51
T2
14—
12—
z/2
Phase
shift
I
I
I
I
I
I
I
I
I
OBJ
I
~I
10—
LASER
Ml
FIG. 5. The beam splitter (8) introduces the infrared laser
beam into the microscope's optical path. The microscope objec-
tive {OBJ)focuses the beam to a focal point, creating an optical
trap. Two mirrors (M1,M2) oscillate around two perpendicu-
lar axes. A m/2 phase shift is applied between the two oscilla-
tors. The beam focal point moves along a circle. Two tele-
scopes {T1,T2) pivot the beam about the back iris diaphragm
of the objective (OBJ). The beam is circularly polarized, using a
quarter wave plate (A./4).
0—
I
0
I I
6 8
x (ttrn)
I
10
I
12
I
14
I
16
FIG. 6. Trajectory of the particle's center of mass. Record-
ing time 60 s. Trap rotation frequency 14 Hz. Output laser
power 700 mW.
erator [17]. A m /2 phase shift between the mirror oscil-
lations results in a circular motion for the beam in the XY
plane. The rotation frequency is the control parameter,
which can reach up to 1 kHz. In order to keep the beam
intensity constant along the trajectory, circular polariza-
tion is used. This is achieved using a quarter wave plate
(A, /4). The potential afFecting the particle in the plane
transverse to the beam axis is then a symmetric, well
shaped potential, rotating around a circle of diameter
12 4 pm (Fig. 4).
10—
0 I
0
I
20
I
40 60
time (s)
I
80
34
50
I
100
B. Particle dynamics
Let us denote the trap rotation frequency by vz. The
particle trajectory is recorded using the image analysis
described in Sec. II A. A typical trajectory is shown in
Fig. 6. The particle radial excursions are small compared
to the circle diameter ( & 3%%uo) and to the particle diame-
ter ( &20%). We thus approximate the particle motion
to be one dimensional and confined to the circle. From
the X-Y coordinates of the particle, we extract the parti-
cle angular displacement as a function of time. The time
series for difFerent trap frequencies are shown in Fig. 7(a).
The mean angular displacement of the particle is shown
in Fig. 7(b). The average is done on periods of the order
of 400 s. Fitting the curves by straight lines going
through the origin, a well defined mean angular frequen-
cy vz is measured. It is plotted in Fig. 8 on a logarithmic
scale as a function of vT for di6'erent output laser powers.
Each curve presents three distinct regimes: a phase-
locked regime where the particle rotates synchronously
20—
15—
a
A 10—
bQ
V
50
60
p
I
20
I I
40 60
time (s)
I
80
FIG. 7. (a) Time series of the particle's angular displacement
for trap rotation frequencies 5.5, 8, 10, 14, 20, 34, 50, and 60 Hz.
Output laser power 700 mW. (b) Mean angular displacement of
the particle as a function of time. Average over periods of
400 s.

PERIODIC FORCING OF A BROWNIAN PARTICLE 5243
101
I I III
I I I I I III(
D. Regime II: Phase-slip regime
10
1 0
10
s i ical
10
s c sisal I I I II ~ l I I
10 10
v (Hz)
FIG. 8. The particle's mean angular frequency vp as a func-
tion of the trap angular frequency v&. The output laser power is
700 mW (circles), 300 mW (triangles), and 1SO mW (crosses).
The three regimes are indicated above the graph for an output
laser power of 7M m%' (circles).
with the trap; a phase-slip regime starting first with a
very sharp decrease of vp followed by the power law
vz ~ vz,' and finally a third regime where the particle's
motion along the circle is diffusive, the mean angular fre-
quency v~ is zero and is not plotted in Fig. 8. These
three regimes are delimited by straight vertical lines in
Fig. 8 for an output power of 700 mW (circles). In Sec.
IV, by rescaling the two frequency axes, we show that all
the curves fall onto a single universal function.
Let us now describe in more detail the curve corre-
sponding to a potential depth of 1250k~ T (output laser
power 700 mW, circles in Fig. 8). For vr smaller than 5
Hz (regime I), the particle is trapped and follows the
tweezer. For vz. between 5 and 70 Hz (regime II) the par-
ticle escapes but still rotates around the circle; a mean
drift is measured. For vz above 70 Hz (regime III), no
net angular motion is observed and the particle diffuses
freely along the circle: the mean angular frequency is
zero. We describe these regimes by comparing the three
forces present in this problem. The trapping force has a
maximum value of a few piconewtons. The Stokes force
is proportional to the particle velocity Vz and opposes
any particle motion. The stochastic force is random in
time and averages out to zero on a Smoluchowski time
scale of the order of 10 s. A more meaningful measure
of it is the average thermal energy k~ T.
In this regime the trap is not strong enough to hold the
particle, but kicks it regularly at each revolution. Be-
tween kicks the particle diffuses, but does not have
enough time to difFuse away from the circle before the re-
turn of the trap. The particle is swept along the circle
with a well defined mean angular velocity. Evidently the
angular motion is no longer smooth: the fluctuations in
the particle angular position increase with the trap rota-
tion frequency [Fig. 7(a)]. The particle mean angular fre-
quency decreases with the trap frequency: the kicking
rotor becomes less and less effective.
One notes just above the critical frequency vz a sharp
decrease of vz. In this region, vz is small enough that
one can observe individual trapping and escape events of
the particle from the moving trap. We now show that
these events are characteristic of a stochastic process.
Figure 9 shows the time series of the particle angular dis-
placement, for vz- equals 5.5 Hz. When the particle is
trapped and entrained by the trap, the trajectory exhibits
almost vertical jumps in Fig. 9. The size of the trapped
periods is unevenly distributed, and can be as large as a
complete rotation of the particle around the circle (2~
rad on the ordinate axis). From direct observation on the
video screen of the particle and trap motions, a histogram
of the residence times ~ in the trap is computed and
shown in Fig. 10. The error bars represent the statistical
uncertainty. The total number of recorded events is 203.
The straight line is a fit of an exponential curve to the ex-
perim. ental points, which gives a mean residence time ( r )
of 50 ms (1.83 video frames). This exponential distribu-
tion of residence times is a signature of a stochastic es-
20—
10—
C. Regime I: Phase-locked regime
At low velocity (vr ~ 5 Hz), the Stokes force is smaller
than the maximum trapping force: the particle is
trapped, its angular frequency is vz-. This is indicated by
the power law of exponent 1 in Fig. 8. This phase-locked
regime disappears when reaching a trap rotation frequen-
cy higher than the critical frequency vc, 5 Hz. The criti-
cal particle velocity is then V&=2+Rv&=190 pms '. It
agrees with the value obtained in Sec. II B by translating
the stage, keeping the trap Axed in space.
0— l
2
time (s)
FIG. 9. Time series of the particle angular displacement, trap
frequency 5.5 Hz, output laser power 700 mW.

5244 FAUCHEUX, STOLOVITZKY, AND LIBCHABER
I
2
10 +
10
10
0.2 0 4
(s)
0.6 0.8
0.01 +
FICx. 10. Histogram of the residence times ~ inside the trap,
frequency 5.5 Hz, output laser power 700 mW. The vertical er-
ror bars represent the statistical uncertainty. The straight line
is an exponential fitted to the experimental points. Inset: The
particle's mean residence time (r) in the trap as a function of
(v&/vT). Output laser power 700 mW. Trap rotation frequency
vT ranging from 4.8 to 5.5 Hz.
0.1
time {s)
FICz. 11. The root mean square value of the particle angular
displacement as a function of time, trap rotation frequency 100
Hz, output laser power 700 mW. The straight line indicates a
power law with exponent —,
'.
cape [14,18]. The mean residence time is a function of
the trap frequency. Because of the scatter in the experi-
mental points (see Fig. 8), it is more convenient to look at
the dependence of (r) on the ratio (v„/vT). This is
shown in the inset of Fig. 10. The mean residence time
diverges when (v~/vT ) goes to 1 (the particle is locked to
the trap rotation and never escapes), and goes to zero
when (vt /vT ) goes to zero (the kicking rotor goes too
fast to induce a net angular motion of the particle).
K. Regime III: Diffusive regime
Since the kick amplitude decreases with vT, it will
eventually become smaller than k~T and thermal effects
will dominate the particle's motion, at least on small time
scales. This happens for frequencies larger than 70 Hz,
for an output laser power of 700 mW. The particle is still
conPned in the radial direction because it always experi-
ences the same side of the optical trap when displaced
away from the circle, whereas in the azimuthal direction
it experiences both sides of the trapping potential. The
particle is essentially free to diffuse along the circle. The
particle's mean angular velocity is zero. Figure 11 shows
the root mean square value of the angular displacement
as a function of time for vT =100 Hz. The straight line is
a power law of exponent —
„
indicating a diffusive motion.
Fitting a square root to the experimental points, the an-
gular diffusion constant is D 6)
=0.008 rad s '. The
theoretical particle diffusion constant is [19]
D =(k~T)/(6vrga ) =0.2 JLtm s '. The trap circular tra-
jectory has a radius R of 6.2 pm, thus one estimates
DO=D/R =0.006 rad s ', consistent with the above
measurement.
Figure 11 shows that the particle's motion is diffusive
over 30 s only. Long time motions are still sensitive to
drifts, however small. We used this one dimensional
diffusion to build an optical thermal ratchet [20], i.e., a
system where the time modulation of a periodic asym-
metric potential induces a directed motion of Brownian
particles. This configuration could also be used to study
the statistics of a one dimensional chain of Brownian par-
ticles.
IV. THEORY
Xb
VT
U
0 J( XQ = X
Xo
Jl
F
X0=x
FIG. 12. (a) Particle subjected to the one dimensional motion
of an attractive potential, moving at velocity VT. (b) Triangular
potential and corresponding force profile.
We now turn to the response of a Brownian particle to
an arbitrarily shaped potential, moving at a velocity VT
[Fig. 12(a)]. The potential and particle motion are one di-
mensional. The potential has a finite support [X„X&]
and is attractive ( U & 0), such that U(X, )=U(Xb ).
other words, the spatial average of the force is zero. The
equation of motion is
mx +m yx =F(x,t )+f„„h,

where y=6mga/m is the damping rate and F(x,t) the
force associated with the trapping potential:
F(x,t) =( a/—
ax) U(x, t).
The physical parameters are such that the system is in
the dissipative limit, i.e., the inertial terms are neglected.
We first study the case where the stochastic forcing f„„h
is neglected. We then turn to a complete Fokker-Planck
treatment of this equation. The opposite limit (conserva-
tive case, y =0) is treated in Appendix B.
A. Deterministic ease
The equation of motion is m yx =F(x,t ) =E(x—
Vz t ).
The potential, moving along the x axis with a velocity
V&, induces a displacement hx of the particle during a
time ht. Before and after, the particle is at rest. In the
referential frame where the potential is fixed, the new po-
sition variable is y=x —
V&t. To the potential support
[X„X&
] corresponds [Y„
I'& ]. The equation of motion is
dy =dt [ [F(y)/m y ]
—
Vz ] and the particle total displace-
ment
or
Y
b.x=hy+Vrb, t= f 1+ dy,
F y /my —
Vz
b F(y)
Ax=
my Vz. F(y)—
Note that the integration limits in the first integral go
from Fb to Y„
the order in which the particle experi-
ences the trap.
The period over which the potential produces a motion
(2)
1S
Yb F y
Y my
dy . (3)
The particle displacement diverges (the particle is
trapped) if there is a y such that F(y) =my Vz. Defining
E as the maximum value of F(y), the ratio F/my is the
critical trap velocity for escape (maximum velocity for
trapping). We now study the large velocity case
(m y Vz. & F), where the particle always escapes.
I.et us first show that Ax is positive. In other words,
the particle is always displaced in the direction of the
moving potential, regardless of the potential shape. Note
that we obtain the opposite result in the conservative case
(Appendix B). One can check that
b F(y)
Ax= dy &
m y Vr F(y)—
XQ
nR
(vr/vc )
(vz /vc) —
1
(Xo/~R )
1+
(vr/vc )
—
1
(v~/vc) is a universal function of the rescaled trap rota-
tion frequency (vz. /vc), as shown in Fig. 13. The solid
line is computed from Eq. (4) for X0=0.6 pm. The com-
plete agreement shows that thermal noise can be neglect-
ed, at least for the mean particle angular frequency. It is
surprising given the stochastic character of the actual
trajectories (see Fig. 9). It is not surprising since the po-
10—
0
v~ ~vz-', in agreement with the asymptotic power law
shown in Fig. 8.
We now turn to a specific potential shape. We approx-
imate the bell-shaped potential in Fig. 3(b) by the triangu-
lar symmetric potential shown in Fig. 12(b). It is defined
by its width 2XO and its slope F. We associate with F a
critical velocity for escape Vc =FIm y and a critical fre-
quency vc=(VC/2mR) on a circle of radius R. vc is
directly proportional to the output laser power (Fig. 2).
It is a known parameter. On the other hand, the poten-
tial half width Xo is a free parameter, such that the tri-
angular potential shown in Fig. 12(b) is equivalent to the
optical trap. The two potentials are said to be equivalent
when having identical values of fi' [F(y)/m y ] dy.
They then induce the identical particle s rotation fre-
quency vz at high enough trap rotation frequency vz
(Appendix C).
From the potential profile in Fig. 3(b) we integrate nu-
merically the square of the force to find
Yb
fz'[F(y)/my] dy=1060 pm s . The maximum value
a
of the force being 6m.gaV&=0. 54 pN, the value of Xo
such that a triangle potential is equivalent to the experi-
mental one is of the order of 0.6 pm.
We now estimate the particle angular frequency for
this triangular potential. From Eqs. (2) and (3),
Ax =2XO Vc~/( Vr —
Vc ) and At =2XO Vz /( Vz. —
Vc ).
From Appendix C, the particle angular frequency is
Yb
whatever the sign of F(y). As fr'E(y)dy =0, the parti-
cle displacement is always positive.
The asymptotic behavior is obtained by expanding Eqs.
(2) and (3) as a function of the small parameter
F(y)/my Vz. This yields
10
0 x s ~ ~ l
I
1
~ ~ 1 I I I
10
bx =(1/Vz. ) fi,'[F(y)my] dy
and b,t =(1/Vz-) fz dy. We derive in Appendix C the re-
Yb
a
lation between Ax, At, and the particle angular frequency
v& along the circle. In this limit, Ax ~ Vz- and thus
FIG. 13. The nondimensional ratio (vz/v~) as a function of
(v&/vz). The output laser power is 700 mW (circles), 300 mW
(triangles) and 150 mW (crosses). The solid line is computed
from Eq. (4), Xo =0.6 pm.

5246 FAUCHEUX, STOI OVITZKY, AND LIBCHABER
tential depth is larger than 250k~T for an output laser
power of 150 mW. Under these conditions, the first mo-
ment of a trajectory, even stochastic, is still deterministic.
In this deterministic model, the residence time for the
particle in the trap is fixed at a constant At. In the exper-
iment, the residence time is distributed on an exponential,
characteristic of a stochastic process (Figs. 9 and 10).
The first moment of the distribution, the mean value ( r),
is plotted in the inset of Fig. 10. We compute in our
deterministic model b, t as a function of (v~lvz. ) from
Eqs. (3) and (4). The result is shown in the inset of Fig.
16 for Xo =0.6 pm, and agrees with ( ~). Again, noise is
negligible.
As a final test of the deterministic model, we have stud-
ied the e8'ect on v~ of the potential depth (output power
of the laser) for a given trap rotation frequency. Results
are shown in Fig. 14(a) for vz of 2, 5, and 10 Hz. To an
output power we associate a critical frequency v& from
Fig. 2. In Fig. 14(b) we show that by rescaling by vr, the
results fall onto a universal curve (solid line, X0=0.6
pm), which can be derived from Eq. (4).
B. I okker-Planck approach
J(y, t)= '
[F(y)-—
myVz ]
— W(y, t) . (6)
W(y, t) ka&
mg mf By
As we are looking for stationary solutions for W(y, t ) we
set to zero the left hand side of Eq. (5); the current is then
a constant [J(y, r ) =J]. The steady state solution is then
also a general solution of Eq. (6):
W(y) = C — I dy'exp
k~T 0
U(y')+my Vz-y'
k~T
We now study the e6'ect of temperature. In the re-
ferential frame where the potential is fixed, the Fokker-
Planck equation [21,22] corresponding to the Langevin
equation of Sec. IV A is written
B B
W(y, t ) = — J(y, t ),
Bt
'
By
where W(y, t ) is the probability density of the particle
and the probability current J(y, r ) is
Xexp
U(y ) +m y Vry
k~T
&10
-2
10
V V T V
g % % X
0
I e
Jmy
W(y) = 1 —
exp
k~T
my V~I.
k~T
Because the trap motion occurs on a circle of radius R,
we set the stationary probability density W(y) to be
periodic of period I =2mB. This implies a relation be-
tween the particle current J and the undefined constant
C. Inserting this relation back into the expression of
W(y) leads to
10
0.5
output power (W)
L
X dy'exp
0
U(y+y') —
U(y )+m y Vry'
k~T
10
10
10 '-
A A ~ A
The particle current J is now obtained by demanding
that the particle probability density W(y) be normalized
to 1. This involves a second integral of W(y) over the
circle perimeter L, and yields
k~T my VII.
1 —
exp
Jmy k~ T
U(y+y') —
U(y)+my Vzy'
dy exp
0 0 k~T
10
0.1
FIG. 14. (a) The ratio (v&/v~) as a function of the output
laser power. The trap rotation frequency is 2 Hz (filled trian-
gles), 5 Hz (open triangles), and 10 Hz (circles). (b) The ratio
(vp/vz. ) as a function of (vc/vz). Trap rotation frequency 2
Hz (6lled triangles), 5 Hz (open triangles), and 10 Hz (circles),
output laser power ranging from 100 mW to 1.3 W. The solid
line is computed from Eq. (4), Xo =0.6 pm.
The above equation relates the particle current J to the
physical parameters of our system: the potential U(y),
the trap velocity V~, and the temperature T. The mean
value of the particle velocity in the referential frame
where the potential is fixed is simply Vz =JI-. In the ab-
sence of potential, this mean velocity would be
V~ = —
Vz-. The presence of a force, however, delays the
motion of the particle, and induces a positive drift
hV= Vz+ Vz-, whatever the referential frame. Since the
motion takes place on a circle of radius R, the mean an-

gular frequency of the particle is given by vz =b, V/2mR.
We show in Appendix D that Eq. (7) implies that vz
scales asymptotically as vT '. The resulting expression for
v~ is independent of the temperature, and is identical to
the one derived in Appendix C within the deterministic
model of Sec. IV A.
%'e now approximate the trapping potential to be tri-
angular as shown in Fig. 12(b). For such a potential, one
can solve analytically for the probability current J and
the particle rotation frequency vz. The complete expres-
sions are given in Appendix E.
The value of (v~/vc), as computed in Appendix E, is
shown in Fig. 15 as a function of (vT/vc) for difFerent
temperatures. The deterministic curve (T=O K, solid
line) is computed from Eq. (4). The room temperature
(T=300 K) curve is almost indistinguishable from the
deterministic one: we show in the inset of Fig. 15 the de-
viation of the (T=300 K) solution from the (T=O K)
solution, close to the transition from a phase-locked re-
gime to a phase-slip regime. It is close to this transition
that the effect of noise is the strongest: the small thermal
fluctuation can make the particle escape from the trap
where it was barely held. The almost complete agree-
ment between the two curves explains why the deter-
ministic model is so powerful in describing the experi-
mental behaviors: the potential depth is so large com-
pared to the noise (250k+ Tfor the minimum output laser
power of 150 mW), that the temperature can be set to
zero when computing the particle's rotation frequency.
As shown in Appendix E, the end result for vz depends
= f w(y)dy,
&~&+
T
which yields
f w'(y )dy
I.—
2XO
&) ()
1 —
f W(y)dy
This expression is computed using Eq. (El). It is plotted
in Fig. 16 as a function of (vT/vc ) for Xo =0.6 pm,
only on myVr/k&T and myVC/ksT. Thus, reducing
the potential depth (and therefore Vc) is equivalent to in-
creasing the temperature. Curves at higher temperatures
(T=3000, 10000, and 30000 K) are also plotted in Fig.
15 to illustrate the effect of thermal noise on vz. v~ al-
ways decreases when T increases: the kicking rotor be-
comes less efficient when more noisy. Also, as vT goes to
infinity, all curves collapse on the asymptotic scaling dis-
cussed in Appendixes C and D. Finally, the trap rotation
frequency at which the particle escapes from the trap also
decreases when T increases: thermal fluctuation favors
the escape of a Brownian particle from a potential trap,
moving or not.
We now estimate the mean residence time & r) of the
particle inside the moving trap. The probability of
finding the particle inside the trap is equal to the integral
of 8'(y) between F, and Yb f.r'W(y)dy. This is also
equal to the fraction of time the particle is trapped. Over
one cycle, the particle is trapped during a time &~) and
the trap comes back under it in a time (I.—
2Xo)/Vz. .
The fraction of time the particle is trapped is thus
I I I I I I I
10
I I I I I I II
102
10 10
10
—
2
100
10
10 10 10 10
10 I I I I I I I I I I I I
FICr. 15. The ratio (v&/v&) as a function of (vz/vc), for
di6'erent temperatures. The solid line (T=O K) is from Eq. (4).
The dashed lines are computed from Appendix E, temperatures
T=300, 3000, 10000, and 30000 K. The higher T, the farther
the dashed curves from the solid one. The parameters are
X0=0.6 pm, Vc =100pms ', R =6.2 pm. The inset is a blow-
up of the ( T=0 K) and ( T=300 K) curves close to the transi-
tion from the phase-locked to the phase-slip regime
(vT/v~ =1).
i'c
FICy. 16. The mean residence time &r) of the particle in the
moving trap as a function of (vT/vz). The solid line ( T=O K)
is computed from Eq. (3), the dashed lines are computed from
Eq. (8) for temperatures T=300, 3000, 10000, and 30000 K.
The parameters are X0=0.6 pm, V&=100 pms ', R =6.2 pm.
The inset reproduces the inset from Fig. 10, with the (T=O K,
solid line) and (T=300 K, dashed line) theoretical values for
(~).

5248 FAUCHEUX, STOLOVITZKY, AND LIBCHABER
Vz =100 pm s ', R =6.2 pm, and for temperatures
T=300, 3000, 10000, and 30000 K. Again the (T=300
K) curve is almost indistinguishable from the determinis-
tic (T=O K) curve from Eq. (4). Note that (r) diverges
as (vz /vc ) approaches 1 in the deterministic model. For
any nonzero temperature, ( ~) is always finite and smaller
than the Kramers time [23] (mean residence time of the
particle in a fixed trap). As the temperature or the trap
rotation frequency increases, (r) decreases. It reaches
for large vr the deterministic limit: (r ) =b,t
=(1/vz )(XQ/mR), as illustrated by the collapse of all
the curves for large vz.
Theoretical considerations similar to the ones present-
ed in this section have been studied in the context of re-
versible computation [24].
U. CONCLUSION
For a given output power (given trapping force), we ob-
serve three regimes for the response of a Brownian parti-
cle (2 pm diameter polystyrene sphere in water) to an op-
tical trap rotating in a circle. These three regimes are
determined by the trap rotation frequency vz-.
When the drag force is smaller than the maximum
trapping force, the particle follows the trap up to a criti-
cal frequency vc (vc =5 Hz for an output power of 700
mW). For larger drag forces the particle escapes from
the trap. Nevertheless a net mean angular velocity vz is
still measured when the kick amplitudes are larger than
k&T. It scales asymptotically as the inverse of the trap
frequency. In the thermal noise dominated regime
(vz )70 Hz for an output power of 700 mW), we observe
the one dimensional diffusion of the particle along the cir-
cle. These three regimes are characteristic of a noisy
asynchronous rotor: to a phase-locked regime succeeds a
phase-slip regime, and then a diffusive regime.
We describe the experimental observations within a
simple deterministic (T=O K) model. We show that the
dependence of (v~/vc) on (vr/vc) for diff'erent output
powers falls onto a universal curve. Solving the Fokker-
Planck equation (which takes into account the effect of a
nonzero temperature), we justify the (T=O K) model:
the potential depth is always much larger than the
thermal noise kz T. The first moments are deterministic,
even though the trajectories themselves are noisy.
ACKNOWLEDGMENTS
It is a pleasure to thank Laurent Bourdieu, Peter
Kaplan, and Albrecht Ott for their help in designing and
building the experimental setup. We thank Marcelo O.
Magnasco and Armand Ajdari for enlightening discus-
sions. We thank R. Landauer for bringing Ref. [24] to
our attention.
APPENDIX A: HEATING EFFECTS
We now discuss thermal gradients inside the cell. PQ is
the laser power in the cell (P0=50 mW at most, see Sec.
II B) and y0 the thermal conductivity of water (g0=0.6
W m ' K '). Polystyrene and water have approximately
the same value for the absorption length l0 in the infrared
(l0 =10 cm). The power absorbed by the sphere of radius
a is P,&, =PD(a/10), of the order of 10 W. In
the steady state, this power is dissipated in the surround-
ings by the heat current: J& =—
gee T, where T is the
temperature field in water. Because of spherical symme-
try, T is a function only of the radial coordinate r, and so
the heat current is radial and equal in magnitude to
y0[dT(r)/dr]. The heat fiux across a spherical shell of
radius r enclosing the polystyrene sphere gives the ab-
sorbed power:
dT(r)
P,q,
= Jg.n =4~r y0 dr
From spatially integrating this equation and fixing the
temperature far away from the sphere to be equal to the
room temperature, the temperature rise at the surface of
the polystyrene sphere is ET=P,&, /4vray0=10 ' K.
The relative change in temperature (b,T/T) is thus of
order 0.03%%uo. Taking the thickness of the cell to be 50
pm, one can scale the above analysis to check that
the total temperature rise is no more than 5 K
( 5T/T = 1.5%%uo ).
APPENDIX B: CONSERVATIVE CASE, y =0
In the conservative case, the equation of motion reads
mx=F(x, t). The total energy of the system is con-
served. In the referential frame where the potential is
fixed, this conservation law reads —,
'my + U(y) =—,
'm Vz,
or
dt=- dy
Q Vz. —
[2U(y) /m ]
The particle displacement is
Ax =Ay+ V~ht
Y
dg 1
Q Vz
—
[2U(y)/m ]
= J,'ay
'
Ql —
[2U(y)/mV ]
It is worth noting that since U(y) & 0 for all y, so is b,x,
whatever the potential shape. This somehow counterin-
tuitive result means that the particle displacement is in
the direction opposite to the trap motion. Also, for
large enough trap velocities, ~2U(y)/m Vz ~
&& 1 and
the above equation can be Taylor expanded as
bx =[Jr'U(y)dy](1/mVr).
Here again the particle displacement scales asymptoti-
cally as the inverse square of the trap velocity. If the
motion occurs in a circle, the particle rotation frequency
will scale as the inverse of the trap rotation frequency.
The overall rotation of the particle will, however, be in
the direction opposite to the trap rotation.
APPENDIX C: MOTION IN A CIRCLE
On a line, the particle displacement bx [Eq. (2)] in-
duced by the moving potential occurs during a time At
[Eq. (3)]. On a circle of radius R (6.2 pm in our experi-
ment), the particle angular displacement is b,8=(hx /R ).
Once the particle escapes from the trap, it is caught again

after the trap comes back: this takes an extra time
2mR —
2XO 1 X()
1—
Vz- vz- m R
1 1 rb F(y)
vr (2mR)' r. my
(C2)
Ax Xo
Vp- At+ 1—
v& mR
(Cl)
In the large trap velocity regime, Eq. (3) yields
b, t =(1/Vz ) fi'dy =(1/vz. )(XO/mR ). Equation (C 1)
reduces then to
v~=(b,x/2mR )vr .
From Eq. (2), the particle's rotation frequency is then
where 2Xo is the total width of the potential
(2XO= I'& —
Y', ). The particle angular frequency along
the circle is then
The particle's rotation frequency scales as the inverse
of the trap rotation frequency. This is the asymptotic
power law shown in Fig. 8 for each curve, and in Fig. 13
when rescaling the frequencies by the critical frequency
&c
APPENDIX D: ASYMPTOTIC SCALING
OF THE FOKKER-PLANCK EQUATION
We show here that the asymptotic scaling of v~ for
large vz- predicted by the Fokker-Planck approach is
identical to the deterministic one [Eq. (C2)]. We first as-
sume that my Vz )&F(y) for all y in the finite support of
the potential, and rewrite Eq. (7) as
k~T my VzL
1 —
exp
Jmy k~ T
m y Vzy' U(y+y') —
U(y)
dy dy'exp
k.T
'+ -y V,y
In this expression, [ U(y+y') —
U(y)]/my Vzy' is a small
parameter that we approximate by F(y)/my —
Vz. Ex-
panding the exponential as a function of this smaH pa-
rameter and integrating over y', we obtain
1 i. F(y) F(y)
dy 1+
V,
' '+-yv. -yv.
Since fOdy F(y) =0, the particle current is then given by
J~ Vz. r, F(y)
dy
L 0 myVz-
L
The particle's mean rotation frequency is
v&=XV/2mR =(JL+Vz )/2mR. Within the above ap-
proximation, it reduces to
1 1 rb F(y)
vr (2mR)' i' ~y
This expression is identical to Eq. (C2): in the asymp-
totic limit of infinite trap velocity the Fokker-Planck
equation and the deterministic Langevin equation give
the same result for the mean particle's rotation frequency
vz. It is independent of temperature and scales as vz '.
APPENDIX E: FOKKER-PLANCK RESULTS
FOR A TRIANGULAR POTENTIAL
The potential shape shown in Fig. 12(b) is assumed to
be centered at L/2, where L is the circle perimeter
(L =2rrR ). We define the following constants:
0.
'=my V~/k&T is a measure of the trap velocity and
s =my VC/k~T a measure of the trapping force. The in-
tegral I=f exOp[U(y)+ y]ady is given explicitly as
a Ya
1
(a —
s){g/2) +sYa aYa
I= +
e ' — e e —
e
a A' S
The four terms in the denominator are
1 —
e
+
—
aYa
I
X (s —
a)XO
0 e
+
5 —
aY 1
X —
+e ' I„——
a(s —
a) CL
(s+a)XO
P3= +
u+s X+A
5 aY 1
+e In
a(s+a) " a
aY
e
+ I
a CK
P& is written formally as
+
b b (a+s )(L /2) aL b
+ e —
e
0,'+S a
We also define I„=I[e /(1 —
e )]. With these no-
tations, the particle's rotation frequency is given by
v~ =( JL + Vz ) /2m. R, where the probability current Jis
—
k~T I
my [P, +P2+P3+P4]

5250 FAUCHEUX, STOLOVITZKY, AND LIBCHABER 51
U S+X' —
U X -ay
dp exp e
aL, '
p p
k~T
f w(y)dy .
—
Jmy p
The terms P2, P3, and P~ are obtained in a similar way by integrating, respectively, from F, to L /2, from L/2 to Yb,
and from Y& to I.. A useful relation for the mean residence time computation is
Y~
f W(y)dy =(P2+P3)
a
(El)
[1]A. Ashkin, Phys. Rev. Lett. 24, 156 (1970).
[2] A. Ashkin, J. M. Dziedic, J. E. Bjorkholm, and S. Chu,
Opt. Lett. 11,288 (1986).
[3] Bangs Laboratories, Inc., Carmel, IN.
[4] The water was purified by a Millipore ultrapure water sys-
tem (Millo-RO 10 plus and Milli-Quf plus) from Millipore
Corp. , Bedford, MA.
[5] 75X50X 1 mm' precleaned microscope slides from Corn-
ing Glass Works, Inc., Corning, NY.
[6] 22 X 22 X0.05 mme coverslips number 0 from Fisher
Scientific, Inc., Pittsburgh, PA.
[7] Simco Topgun from Simco, Inc., Hatfield, PA.
[8] Five minute fast epoxy from Devcon Corp. , Danvers, MA.
[9]Upright microscope model Axioskop from Carl Zeiss,
Inc., Thornwood, NY.
[10]TI-24 CCD camera from NEC Electronics, Inc., Moun-
tainview, CA.
[11]Perceptics PixelBuffer model PTB 425 from Perceptics
Corp. , Knoxville, TN, used with a MacIIFx computer
from Apple Computers, Inc., Cupertino, CA.
[12]Model C-105 from CVI Laser Corp. , Albuquerque, NM.
[13]Zeiss Plan-Neotluar 100.
[14]A. J. Simon, Ph.D. thesis, University of Chicago, 1992
(unpublished).
[15]Mirror positioning system model 650X from Cambridge
Technology, Inc., Watertown, MA.
[16]K. Svoboda and S. Block, Annu. Rev. Biophys. Biomol.
Struct. 23, 247 (1994).
[17]Synthesizer model 3325B from Hewlett Packard Corp. ,
Sunnyvale, CA.
[18]A. Simon and A. Libchaber, Phys. Rev. Lett. 68, 3375
(1992).
[19]A. Einstein, Ann. Phys. (N.Y.) 17, 549 (1905).
[20] L. P. Faucheux, L. S. Bourdieu, P. D. Kaplan, and A. J.
Libchaber, Phys. Rev. Lett. 74, 1504 (1995).
[21]H. Risken, The Fokker Planck Equa-tion (Springer-Verlag,
New York, 1989).
[22] N. G. Van Kampen, Stochastic Processes in Physics and
Chemistry, North-Holland Personal Library (Elsevier Sci-
ence, Amsterdam, 1981).
[23] H. A. Kramers, Physica (Utrecht) 7, 284 (1940).
[24] R. Landauer and M. Biittiker, Phys. Scr. T9, 155 (1985).

Periodic_Forcing.pdf

More Related Content

Similar to Periodic_Forcing.pdf (20)

More from luc faucheux (20)

Recently uploaded (20)

Periodic_Forcing.pdf