Вы находитесь на странице: 1из 16

Pure and Applied Optics: Journal of the European Optical Society Part A

Related content
- Trapping and levitation of a dielectric
Trapping and levitation of a dielectric sphere with sphere with off-centred Gaussian beams.
I. Experimental
off-centred Gaussian beams. II. GLMT analysis M I Angelova and B Pouligny

- A concise presentation of the generalized


Lorenz-Mie theory for arbitrary location of
To cite this article: G Martinot-Lagarde et al 1995 Pure Appl. Opt. 4 571 the scatterer in an arbitrary incident profile
B Maheu, G Gouesbet and G Grehan

- Expressions to compute the coefficients


gmn in the generalized Lorenz-Mie theory
using finite series
View the article online for updates and enhancements. G Gouesbet, G Grehan and B Maheu

Recent citations
- Generalized Lorenz-Mie theories and
mechanical effects of laser light, on the
occasion of Arthur Ashkin’s receipt of the
2018 Nobel prize in physics for his
pioneering work in optical levitation and
manipulation: A review
Gérard Gouesbet

- Optical tweezers: theory and modelling


Timo A. Nieminen et al

- A scientific story of generalized Lorenz-


Mie theories with epistemological remarks
G. Gouesbet

This content was downloaded from IP address 14.139.38.11 on 15/04/2019 at 13:00


Pure Appl. Opt. 4 (1995) 571-585. Printed in the UK

Trapping and levitation of a dielectric sphere with


off-centred Gaussian beams: 11. GLMT analysis
G Martinot-Lagardet, B Poulignyt, M I Angelovats, G Grkhant and
G GouesbetJ
t Cenue de Recherche Paul-Pascal. CNRS, Avenue A Schweitzer, 33600 Pessac, France
t Labomtoire d'bergktique des systk"e et proc6d6s. CNRS. INSA de Rouen. BP 08,
76130 Mont-Saint-Aignan. F m c e

Received 20 December 1994. in final form IO April 1995

Abstract. We study the forces exerted by a Gaussian laser beam on a small dielectric sphere in
water as a function of beam-off centring, i.e. the distmce between the beam axis and the sphere
wve. Our experimental data are f"the first part work of Angelova and Pouligny who studied
the levitation of polystyrene latex spheres, a few ndcrometres in radius, by a couple of vefiical
laser b w n s in water. We also repon on newly acquired results, particularly the observation of
sustained oscillations of the levitated sphere in water.
These data are analysed from the viewpoint of radiation pressure forces, which we calculate
using the genenlired Larem-Mie theory (GLMT). We find that calculated equilibria fit those
experimentally recorded with beams of unequal intensities. Experiments with equally intense
beams lend to instabilities in the levitated paiicle position. This is also well explained by CLMT,
but the instability threshold is found to be very sensitive to uncontrolled water convection around
the particle.

1. Introduction

This paper addresses the basic problem of measuring the forces exerted by a Gaussian laser
beam illuminating a micrometre-sized transparent dielectric sphere. This is the second part
of the recent work by Angelova and Pouligny [l], which was aimed at directly measuring
such forces as a function of the distance of the beam axis to the particle centre. In [I],
the authors used a four-beam levitation trap and investigated the example of polystyrene
latex spheres in water. As they explained, knowing quantitatively the forces due to the laser
beams is essential for future applications using such levitated spheres as captors in optical
dynamometry experiments.
The main goal of this paper is to compare calculated radiation pressure (RP) forces to
those found experimentally. Such comparisons have been made before in a few situations
[2-51. However, these were restricted to particles located on beam axis or to particles
large enough for the ray optics approximation to hold [?-7]. Here we calculate RP forces
in the most general way, using the so-called 'generalized Lorenz-Mie theory (GLMT)' [SI.
This theory starts from the Maxwell equations and directly gives the electromagnetic field
scattered by a homogeneous sphere illuminated by an arbitrary field distribution. The
solution is set out in the form of an infinite series, just as in the basic Lorenz-Mie theory
5 Permanent address: Institute of Biophysics. Bulgarian Academy of Sciences. I 1 13 Sofia, Bulgaria.

0963-9659/95x)50571+I5$l9,500 1995 1OP Publishing Ltd 57 1


572 G Martinot-Lagarde et a1
[9],but with coefficients (grim) that depend on beam shape and location relatively to the
particle centre.
The paper is organized in the following way. In section 2, we briefly review the
principles of the experiments canied out in [I], i.e. levitation of a sphere by two parallel
vertical Gaussian beams. In particular, we recall the important observation that the lateral
position of the particle (s) is discontinuous compared to beam separation ( x ) when the two
beams have equal powers (p = 1). As we will see, bare application of GLMT reproduces
the experimental equilibria observed with beams of very different powers (p <( I), within
a small but systematic difference. GLMT also predicts the particle position instability when
p = I, but the value of the calculated beam separation threshold is found to be systematically
larger than the measured one, beyond experimental error.
In section 3, we use GLMT to study the equilibrium of a sphere held by two parallel
vertical beams as a function of beam separation ( x ) . particle altitude ( 2 ) and beam intensities
ratio (p). Our results are set out in the form of a three-dimensional stability diagram. Among
other things, this analysis will show that the transition (instability) observed in [l] is not
second order, as previously believed, but first order. However, the calculated stability limit
does not coincide with that found experimentally.
Section 4 is aimed at reconciling the GLMT stability diagram with the experimental
observations. We investigate the importance of previously ignored mechanisms such as
Brownian motion of the particle and parasitic convection flow in water. Indeed we show
that the latter mechanism is strong enough to explain the above mentioned deviations.
Experiments reported in [I] were made at constant altitude ( z = 0) and variable beam
separation ( x ) . In section 5, we investigate the conjugate (constant x, variable z ) situation.
The first-order character of the transition studied in section 3 leads to hysteresis and, in
special situations, to particle sustained oscillations. We report on a few novel experiments,
where t h e e oscillations are observed for the first time.
Our main conclusions are given in the final section 6, together with remarks about
related but still unsolved problems.
In this paper, GLMT calculations are carried out on the basis of the theoretical framework
set out in [8]. RP forces are calculated using the following equation [8]:

Here n is the refractive index of the continuous medium around the sphere, c is the velocity
of light in vacuum. Ei, Hi is the source field, i.e. the field in the absence of the particle.
E , If is the total field (incident + scattered) when the particle is present. S is a spherical
surface which is concentric of the particle and whose radius is large enough for the field to
be in rhe far-field limit. 0 is the unit vector, normal to the local surface element dS and
oriented as dS.
In [IO], ,.g coefficients were calculated using the localized approximation (LA) to assure
fast and accurate prediction of beam shape coefficients,.g [8]. The program we used here
is a slightly modified version of the [IO] version. The only modification was to replace LA
by a 'modified localized approximation' (MLA), as explained in [I I]. However, in the case
of moderately focused beams of interest in this paper, the difference between LA and MLA
is insignificant.
Panicle levitation by off-centred laser beams 573

2. Experimental data: direct comparison with GLMT

The problem that was addressed in [I] is defined in figure l(a). A homogeneous transparent
sphere is levitated in water by a Gaussian beam. z is the altitude of the sphere centre, with
the origin taken at the beam-waist. e is the distance from sphere centre to beam axis. The
problem was to measure the vertical (4)and horizontal ( F l ) forces acting on the sphere
as a function of e, at constant i .

Top view

.Front view

(3 (b) (4

Figure 1. (a) The basic problem (see text); @)calculated G L radiation


~ pressure forces with the
following parameters: laser wavelength A = 514.5 nm in air, beam-waist og = 3.7 fim; sphere
radius n = 8.26 fim; sphere refractive index n, = 1.59: water refractive index n, = 1.34; (e)
basic scheme of he experiments carried out in part I (see text and [I]).

If the sphere refractive index n, is larger than that of water (nw = 1.34). RP forces
always pull the sphere Centre towards the beam axis. Experiments carried out [l] used
polystyrene ( P S ) latex spheres (n, N 1.59) with diameters in the 10 + 20 pm range. For
illustration, we show an example of theoretical (GLMT)Fl(e) and F l ( e ) profiles. For the
purpose of the forthcoming discussion, we denote a m the value of e at which F l ( e ) is
maximum.
The principle of the experimental procedure used in [l] is sketched in figure l(c). The
sphere is levitated by two vertical beams, a fixed one (F)and a mobile one (M),which are
made up from the same argon ion laser, operated on the 514.5 nm line. The two beams
are identical in shape, with a beam-waist 00 = 3.7 pm. (In [l], this value was obtained
from the beams’ far-field patterns; recently we obtained an identical value from the cross
correlation of the two beams in the beam-waist plane (z = O).) The beams are linearly
polarized at 45” from the figure plane and perpendicularly to each other.
We define the following quantities: 1, and 1, are the powers of the fixed and mobile
beams, respectively. We denote by XF, XM the distances of the sphere centre to the fixed
+
and mobile beam axes, respectively. x = x p xM is the separation between the two axes.
w = (4/3)na36pg is the sphere weight in water. Here a is the sphere radius, g is the
acceleration of gravity, and Sp = ps - pw is the difference between the sphere density and
574 G Marfinot-lagarde et a1
that of water.
Several experiments were carried out with different values of x and of the ratio
p = ZM ~ Z Fto produce a variety of (XF, XM) configurations. In practice, the experiment
+
amounts to tuning the total power IF ZM so as to keep the sphere always at the same
altitude. Experiments were restricted to the beam-waist plane, i.e. z = 0. Figure 2 shows
+
recent results obtained with p = 0.5, in the form IF ZM against x and X F against x .
The sphere equilibrium is described by the following simple equations:
- IMF.L(xM)
IFFL(XF) =0 (24
+
Z M ~ I ( X M )= W .
ZFF~~(XF) (W
In equation (2) we implicitly assume that each beam produces a force proportional to
its own power and that the two beams act incoherently. The first point was checked for
using a procedure which is described in detail in [l]. Although the two beams are built with
crossed polarizations, it is not obvious that the scattered F and M fields do not interfere
and then that there is no cross (F,M) term in equation (2). We investigated this question
empirically by comparing two situations. The first situation is the basic one, where the two
beams are coherent and constant in time;in the second situation the two beams were quickly
alternated using a specially designed chopper. The switching frequency (about 1 kHz) was
high enough for the sphere to feel only averaged powers. We found that both situations
+
lead to the same results in terms of IF ZM and XF against x . within experimental scatter.
Therefore we may safely ignore (F,M) interference in equation (2).
The full curves in figure 2 are the result of simulation, i.e. were found by injecting
GLMT forces into equation (2). Equation (24 was first solved for ( X F , XM) and equation (26)
+
was consecutively solved for ZF I M . In the fitting procedure values of x were bare
experimental ones. Particle radii and density were allowed to vary within experimental
error, & 0.1 p m and f 0.003, respectively. Values of ps were measured from individual
particle sedimentation velocities, and found to be in the range 1.041 f 1.049. These values
are in the lower half of the range indicated by POLYSCI" (who provided the latex beads)
for PS density, namely 1.04+ 1.065, We also allowed for a systematic shift (*0.2 pm) in
XF values.
It is evident from figure 2 that the behaviours calculated from equation (2) closely
parallel those experimentally recorded. However, there is a slight systematic mismatch,
which is due to the fact that experimental data are not exactly even in x , contrary to what
might be expected from symmetry (changing x in -x amounts to exchanging the two beams,
which are identical). Our interpretation is that this non-evenness is caused by a parasitic
horizontal force which we attribute to large-scale water convection around the levitated
particle. This point will be discussed in section 4. In figure 2 we anticipate the results of
section 4 and show convection corrected responses, I ( x ) and x&) (dotted curves). The
correction very well compensates for the above mentioned deviation from evenness.
We found that the best GLMT fit to experimental data was obtained with n, = 1.64,
a value slightly larger than that commonly accepted for massive PS, i.e. 1.59 [12]. In
fact, anomalously high values of n6 for polystyrene microspheres were mentioned earlier
by^ Ashkin er al [13] and by Baker-Schut et a1 [4]. The data shown in figure 2 correspond
to spheres which were about 1 year old (starting from the batch delivery date). We found
that the data reported in [l] (see figure 5(c) and 5(d) in [l]) were best fitted with ns = 1.59.
As these data were obtained with a fresh batch (less than a month old), we believe that PS
aging is the reason for the evolution in ns.
At this stage, we thus conclude that GLMT quantitatively fits our experimental data;
which in practice means that we are able to quantitatively predict the radiation pressure
Particle levitation by off-cenrred lnser beams 575

-E
Y

I X
P

+
Figure 2. Total levitation intensity (IF IM) and pmicle displacement off lhe mm1 intense
beam axis (XF or X M ) against beam separation ( x ) . Circles are experimental data obtained with
3 different sphere radii, which are indicated in the figw. Full curves and dotted CUNS are
o w best fits lo experimental points using equation (2) and equation (7), mspectively. C is
the convection panmeter, as defined in equation (7). p stands for the ratio IM/ I F . In principle
, = 0.5 and 8 = 2 are equivalent hom symmetry. The most inlense beam is the fixed one in
9
(a) to (d) and the mobile one in (e). (0.

forces acting on the PS spheres in our conditions.


In [I], the authors proposed a procedure to deduce the Fll(e) and FL(e) profiles directly
from experimental data. The procedure is based on levitation with two beams of equal
516 G Martinot-Lagardeet al
powers (p = 1). In this case and when x is small, the sphere is equidistant from both
beams, i.e. x p = XM = x/Z. This simple behaviour allows for a direct determination of
FII(x/2) from equation (26). When x increases the picture becomes more complex because
the sphere suddenly jumps out of the symmetrical position and locks near one of the two
beams. This instability happens for a critical beam separation which was denoted X T R (see
figure 6(c) and (d) in 111). From a qualitative analysis of the instability, Angelova and
Pouligny argued that x m / 2 is equal to a m (see figure I(b)) and took this as a constraint
for the elaboration of F l ( e ) and fil(e) profiles from experimental data For a sphere
which was 16.5 pm in diameter, they found x m / 2 = 5.1 pm. In this case, CLMT gives
6.1 <aTR < 6.5 pm, for ns ranging from 1.59 to 1.64 and for 00 ranging from 3.6 to
3.8 prn. The difference between x T ~ / and 2 aTR is definitely beyond experimental error
(& 0.14 pm). We observed the instability with several other spheres and systematically
found the same type of mismatch. A few examples are given in table 1.

Table 1. Expenmcnul value? of h e km sepmuon threshold (xm)lor larenl inst~biliry


k mismch klseen
mwwed wuh spheres of differen1 ndii ( a ) The 1x1line m the t3blc IS l
rhe cdculated (%)and muswcd thresholds All quvluucs am in " m e t r e 9
a 6.25 7.9 8.25 9.4 9.8 16.6
-12 3.5 5.2 5.1 6.3 6.4 12.2
2nm - r m 2.0 1.3 2.4 1.9 2.2 3.0

As we will see in the following, the observed instability is understandable on the basis
of equation (2). We will proceed in two steps. In the next section, we will first confirm
the above mentioned point that the instability occurs when x = ZQTR, in the conditions
of equation (2), i.e. when parasitic horizontal forces are neglected. In section 4, we will
discuss the influence of convection on the instability and tentatively exptain the differences
between the calculated and measured threshold values.

3. Levitation by two vertical beams: lateral stability diagram

This section is a general study of the solutions of equation (ZQ) and of their stability as a
function of p, x and z . As in [I]. we define s as the distance between the sphere centre
and the vertical plane which is equidistant from the two beam-axes. We thus have
s - X F = XM - -21x '
'x
2 (3)
In equation (2) we fix the values of B. x and z . Equation (2) can then be cast in a form
where s and IF are the unknowns. In other words, we impose the separation between the
two beams and the power ratio 6. Then we look for the lateral position of the sphere and
for the levitation power necessary to hold it at altitude z. Equation (2a) now reads:
Fl(~x-s)-pFL(;x+s)=o (4)
which we write in the following potential form:

..
Here Vl is a potential energy per unit power of the fixed beam (IF). Using GLMT,
we numerically calculated Vl(p,x , z , s) for a few particle sizes. An example is shown in
figure 3 corresponding to p = 1, a = 8.26 pm, n,/n, = 1.19, 00 = 3.7 pm. The column
on the left-hand side of the figure shows the sequence of potentials in the beam-waist
Panicle levitation by off-centred h e r beams 577

Figure 3. General analysis of the lateral positions of a sphere levitated by WOvertical Gaussian
beams. Forces are calculated by o ~ m with 0 = 8.26 flm nJn, = 1.19, A = 348 nm (in
water), og = 3.1 wm. (a) shows the sequence of lateral potentials felt by the sphere in the beam
waist plane (I = 0) for +4 = I and increasing values of beam separation. "he potential energy
is defined as l o V i @ . x, L. s) where Io is a beam power. There is some arbiuariness io the way
of defining IU in the different figures of the sequence. For simplicity, we took 10 = 2.9 mW,
which is half the toWl power necessary to hold the sphere at z = 0 with two identical coaxial
beams (x = 0); (b) is the diagram of stable lateral positions at +4 = I, against x and L (see text);
(c) is the most general diagram, against x, E and +4. Stable and metastable equilibria caexist
..
in the domain between the x' line (- - -) and the x** line (. , , .). The full curve is the me
transition line, in the thermodynamic sense.

plane ( z = 0) for increasing values of x . This sequence corresponds to the conditions of


figure 6 in [l]. When x is small, we find equilibria of the type denoted I in figure 3(a) (the
example shown corresponds to x = 11.6 pm): the levitated particle is in the symmetrical
position (s = 0). Two lateral wells appear when x reaches a threshold value which we
denote x* (= 11.7 fim here). The corresponding positions are metastable up to x = x,
(= 12.1 pm here). We denote by U the type of equilibria found between x" and x,. Above
xc, the symmetric position becomes metastable and survives up to x = x** (= 12.7 pm
here). The region corresponding to x, x < <
x** is denoted 111. In the region x > x**
(region JY)the only equilibrium positions are non-symmetric ones. In the language of
phase transitions [14], x, is equivalent to a transition temperature, with s playing the role
of the order parameter. The transition is first order because s varies discontinuously at x,:
s, = 5.9 pm in this example. x * and x** are the equivalents of an overcooling temperature
578 G Martinot-Lagarde et a1
and of an overheating temperature, respectively. Following this analysis, we predict that,
starting from x = 0, the levitated sphere stays in the symmetric position up to x = x**,
where it abruptly jumps to a lateral position su (= 6.3 pm in this example). A direct
comparison with the GLMT F l ( e ) profile shows that x** = ~ Z Z T Rwith
, our definition of am
(see section 2, figure l(b)).
Qualitatively, this sequence very well reproduces the experimental observation reported
in [l] (see figure 6(c) in [l]): there we observed that the particle made a discontinuous jump
to a non-symmetrical position at x = XTR = 10.2 pm. The discontinuous character of the
transition was not interpreted in [l]. However, the agreement is only qualitative because
the stability limit ( x " ) and the measured one (XTR) do not coincide within experimental
error.
Figure 3(b) contains the results of a similar analysis at different altitudes, i.e. above and
< <
below the beam-waist plane. We find that the transition is first order for Zb z z,,. with
z. N -Zb (= 120 p m here). Outside these limits the transition is continuous (second-order).
The two points where the x*(z), x,(z) and x**(z) lines merge are equivalent to tricritical
points (TCP) [14]. The reason why the transition is first or second order is easily found by
looking at the shape of F l ( e ) (see figure l(b)) and by comparing the slope in the raising
part of the curve, say at e = 0, to that in the decreasing part, say at the inflexion point. The
transition is discontinuous when the former is smaller than the latter (in absolute value).
This criterion is fulfilled only when the beam diameter is small enough compared to that of
the sphere. For given beam characteristics, this may happen only near the beam waist, if
the sphere is large enough. Because diffraction increases the beam diameter, the transition
will necessarily turn second-order far above or below the beam-waist plane. This is the
origin of the TCPS.
We find that, at large 121, the only stable sphere position is the symmetric one (s = 0).
This also can be intuited because diffraction ultimately mixes the two beams into apparently
a single one.
A more subtle feature in the stability diagram is the existence of a minimum in x,
along both branches of the second-order line, at zdn N 148 pm (in absolute value). If, for
example, we tune the beam separation at x N 11.8 pm, we predict that the levitated sphere,
starting from a low altitude, say z = -200 pm, will be at s = 0 up to z = -160 pm,
then in the non-symmetrical position up to z = -140 pm, and then back to s = 0. This
is another consequence of beam diameter evolution due to diffraction. In fact, the second
order line is just the graph of aTR versus z . The value of aTR depends essentially on the
beam size w ( z ) , and weakly on the wavefront curvature [15]. From GLMT one finds that aTR
is minimum for w = 6.5 pm. This is the beam size at z = rt148 pm, which corresponds
to the above given value of zmin.
Another and more obvious influence of wavefront curvature is the fact that the stability
diagram is not reflection-symmetric through z = 0. Indeed the two TCPS correspond to
different values of the beam separation. The same remark holds for the two second-order-
line minima. In practice, this means that the sequences of lateral positions of the levitated
sphere along z depend on the direction in which light propagates.
Figure 3(c) shows the evolution of the stability diagram when the two beams have
unequal intensities, for different values of ,3. Figure 3(b) corresponds to 3, = 1. When
,3 < 1, the 'coexistence zone', i.e. the region between x' and x** is reduced in size and
completely shrinks out at ,3 = Pc E 0.65. When 3, c pc, there is no lateral position
instability. The case ,3 = 0.5, which was experimentally studied in section 2, corresponds
to this situation. Also notice that the second-order line exists only in the limit 3, = 1. To
make an analogy with a magnetic system [14]. the second-order line is the locus of points
Particle levitation b y 08-centred laser b e a m 579
where the order parameter susceptibility x if infinite. Here

y, is singular (infinite) when x = 2aTR. When @ < 1, the singularity in the order parameter
(s) disappears, just as in the magnetic system when a finite external field is applied. Here,
(@- 1 ) F ~ ( x / 2 )is the quantity that plays the role of the external magnetic field.
To conclude this section, we wish to comment briefly on the influence of possible
electromagnetic resonances (also called ‘morphology dependent resonances’, MDR)on the
stability diagram. MDRs happen for special values of the size parameter, 27ralh. and do
affect the RP forces [IO]. For example, in our situation (ns/n, = l.I9), a , = 8.6 pm and
a2 = 8.5 p n are resonant and non-resonant sphere radii, respectively. When comparing
these two examples, we found a significant difference between the Fll(e) profiles and a
negligibly small difference between the F L ( e ) profiles. Particularly the difference between
the corresponding x** values (x;’ - x;* Y 0.15 pm) is much inferior to the mismatch
between theory and experiment, i.e. X I * - XTR Y 2.4 pm in this case. Consequently, we
believe that our conclusion that the CLMT value for XTR is larger than the experimental one
is valid whatever the sphere size is resonant or not.

4. Influenceof convection

RP forces and gravity are not the only forces acting on a levitated sphere in water. For
instance thermal agitation generates a rapidly fluctuating force which causes the Brownian
motion of the particle. Including this stochastic force in equation (2) makes the problem
a statistical mechanical one. In this respect, what we dealt with in sections 2 and 3 is the
zero-temperature limit of this general problem. The correctness of this approximation can be
roughly estimated by comparing the thermal energy kBT (kB is the Boltzmann constant and
T is the absolute temperature) to typical energies found in the zero-temperature l i t . In the
problem of the sphere lateral transition, the relevant energy is the transverse potential barrier
(U) at x = xc (see figure 3(a)). For a sphere 8.26 p m in radius, we find U rr 125 kBT, which
means that Brownian excursions can be safely neglected. Notice that this would not be true
for much smaller particles, because U decreases steeply with a: for instance U rr 24 kBT
for a = I um.
Since the ‘large’ (a = 8.26 pm) sphere is the focus of our discussion, we now ignore the
Brownian motion. Another and, in practice, much more important complication comes from
large-scale convection flow in the volume of the sample cell. This flow can be visualized,
using incoherent illumination, from the motion of tiny dust particles in the water. In our
conditions, (see [I] for a description of the cell) we recorded horizontal velocities of the
order of a few micrometres per second. Over a time period of several minutes the local flow
velocity fluctuates both in amplitude and direction around an average value, which is small
but non-zero (about a micrometre per second is typical). This parasitic flow exists whether
the beams are present or not and consequently it is not attributable to laser induced heating.
What causes the convection is unclear to us. It may be due to residual temperature gradients
around the cell or to a slight evaporation of the water, although the cell is supposed to be
hermetically closed. In practice, we could never suppress convection.
Convection creates a fluctuating force acting on the levitated particle, which
intermittently can be as large as the particle’s apparent weight. To describe the influence
of this parasitic force, we modify equation (2) to
IFFL(XF) - IMFJ.(xM) = c w (74
580 G Martinot-Lagarde et ai

IFFII(~F)+ I M ~ I (=~WM . ) (76)


Here we have just added a Stokes force to the right-hand side of equation (2a). The
amplitude of this force is defined relative to the particle weight ( w ) by the factor C. We
did not do this in equation (26) since this would just amount to rescaling w . Notice that
a' = tan-' C is the angle between the total light-induced force acting on the levitated particle
and vertical.
The dotted lines in figure 2 are the result of simulation with equation (7). Best fits to
experimental responses are obtained with rather small values of C (< 0.2). which means that,
in this example, parasitic Stokes forces are small compared to particle weight and RP forces.
Here we deal with experiments involving beams of very different intensities (p = 0.5), as
we explained in section 2. In this case, we believe that the parasitic force to be considered
in equation ( 7 4 is essentially the average Stokes force, which is small, as we noticed.
There are two reasons for this: (i) when p = 0.5, the particle lateral position (s) is a regular
function of x and C;and (ii) each levitation experiment lasted several hours, and data were
collected by successively increasing and decreasing the beam separation. Consequently,
each experimental point is the result of a statistical average over the fluctuations of the
parasitic force. To a first approximation, we may write =s(x, c), with the overbar
denoting time average values. In other words, fluctuations are averaged out and Z is found
from equation (7). with C corresponding to the average Stokes force.
The existence of a lateral instability when p = I is preserved even in the presence of
a large convection (C N I), as we will now see. Setting IF = IM = I in equation (7) and
eliminating I we obtain
+
FL(xF) - FI(xM) - c[FO(xF) qI(xM)l =o. (8)
We studied the solutions of equation (8) and their stability in the same way as we did for

-
equation (2a) in section 3, i.e. we set the equation in a potential form. We find a first-order
transition as in the limit a' + 0, even for very large values (fgol 1) of the Stokes force.
This is illustrated in figure 4(a) in one example (a' = 29") for a = 8.26 pm. In this case the
particle lateral jump occurs for XTR N 10.2 pm, a value definitely smaller than that found
in the absence of convection (x** = 12.7 pm) and which fits the experimental one (see
table 1 and [I]). The displacement of the transition by convection is illustrated in a more
general way in figure 4(b) in the form XTR against fg(ol).
In this analysis, the value of C that fits the experimental position of the instability
(C = 0.55) is fairly large compared to those found in the p = 0.5 case. This is not
accidental, and not contradictory either. In fact, the particle lateral jump is a single event,
which is taken as the onset of the instability. This single event occurs as soon as the lateral
force is large enough to produce it and is then mostly sensitive to the peak value of the
fluctuating Stokes force. In fact C = 0.55 corresponds to U N 4 pm s-I for the flow
velocity, which is well of the order of what we have observed as instantaneous velocities.
These rough arguments may explain the trends observed in comparing CLMT equilibria to
experimental ones. In practice, our most important prediction in relation to the experiments
carried out in [ I ] is that convection moves the experimental instability threshold (XTR) to
a value lower than the theoretical ( x * * ) one. The difference between these two values is
found to be very sensitive to the fluctuating convection flow around the particle.

5. Particle sustained oscillations

In section 3, we described situations where the levitated particle lateral position goes through
a first-order transition when the beam separation is increased. In principle, hysteresis should
Particle levitation by off-centred larer beams 581

0.0 0.. 0.1


c = t*(u)

Figure 4. Effect of horizontal convection on particle lateral stability. (a) Potentials Felt by the
same particle (a = 8.26 p n ) with C = 0.55 (see equation (7)). forx = IO pm i-),10.2 pm
..
(- - -) and 10.4 pm (. . , .). The Vansition occurs at r?r( E 10.2 pm. (b) Evolution of the
sepadon threshold (.qdagainst horizontal convection.

be observed between x* and x**. Indeed we found that bringing the sphere back to the
symmetrical position requires decreasing x definitely beIow XTR (in the exampIe discussed
hereafter the difference was 0.8 pm).
Up to now we have restricted the discussion to static equilibria. In this section, we want
to describe a remarkable cyclic movement of a particle levitated by two vertical beams in
a viscous medium. We could observe this phenomenon with a few particles, whose radii
were in the 6.2 f 7 p m range (approximately), at the onset of lateral instability ( x I IXTR).
Such a movement is sketched in figure 5. Experimental conditions were: sphere radius
= 7.0 pm; x = 9.5 fim. The laser power was tuned about 2.5% higher than that necessary
to hold the sphere at z = 0, s = 0, near the lateral stability limit.
The initial sphere position was s = 0, z = 0. The laser power excess made the particle
move upwards, up to about f10 pm. There, it shifted to the unsymmetrical position
(s II4.5 pm) and started losing altitude, down to about -30 pm. There it went back
to the symmetrical position and rose again. The movement was cyclic, i.e. repeated itself
indefinitely. The period was about 28 s.
These sustained oscillations can be interpreted qualitatively from the variations of the
FA and 41forces against both s and z. In principle, the cycle takes place at constant x and
variable z. Figure 6(a) is the lateral stability diagram of a latex particle 7 p m in radius.
In this diagram, we consider the path numbered 1 to 7. The evolution of the lateral forces
along this path is sketched in figure 6(b) in the form of a sequence of V i @ )potentials. The
corresponding evolution of the levitation force is shown in figure 6(c).
The movement starts at z = 0, i.e. at step 4 in the sequence. Here the sphere is in the
582 G Martinot-Lugardeel a1
20

v
i i

N
-20

-4 0
0

ili
Figure 5. Sketch of the cyclic trajectory followed by a panicle of radius 7.02 fim between
the two vertical laser beams, near the onset of lateral instability. The panicle movement was
observed fmm above and video recorded. with the video camera (see [I] for details) fwused
at a few diffwent altitudes around the beam-waist plane. This allowed us to find the laleral
positions of the particle as a function of z and then to roughly reconstruct the trajectory in the
vertical plane.

Figure 6. Qualitative explanation of the panicle sustained oscillations. (a) lateral stability
diagram of the sphere (0 = 7.02 pm). @) Sketch of the hysteresis cycle of Ihe particle lateral
position along the 1 to 7 path. (c) Evolution of the toW levitation force against particle lateral
position. Here lo is t&en equal to 1.4 lp., where 1, is the pawer per bevn necessary to hold
the sphere in the symmeuical position (s = 0) at the onset of the instability (xm = 10.18 pm
io this case).

symmetrical position and experiences a levitation force larger than its weight. The particle
Panicle levitation by off-centred loser beams 583

moves up, reaches the xxx line and shifts to the unsymmetrical position (step 6). There the
levitation force takes on a value lower than w (figure 6(c)). The sphere then falls down and
remains in the unsymmetrical position until it reaches the X * line (step 3). Transition hack
to s = 0 then happens, the particle levitates and lifts up through step 4 again.

(a) (b)

Figure 7. This figure is made with p = 0.998. (a) Calculated positions (0)of the particle along
the hysteresis cycle sketched in figure 6@). The broken curve is the trajectory followed by the
paRicle centre, as found 6” integration of the kinetic equations of the movement. @) Here
one beam has been tilted by O S x IO‘’ rad from venical. The trajectory (full curve) is now
cyclic and nearly follows the hysteresis cycle (A).

The cycle sketched in figure 6@) was numerically calculated from the GLMT forces and
is displayed in figure 7(a). Strictly speaking, the existence of this hysteresis cycle is not
a proof for the existence of a periodic solution to the dynamical equations of the system.
These can be set out in the form:
FL(Z.S) = z‘s (9a)
F l I ( 2 . S ) = cz. (9b)
Here 3; and FIIare the total horizontal and vertical RP forces, respectively. 5 = 6irqa is
the sphere friction coefficient in the fluid of viscosity q. Equation (9) is the Stokes limit
of the movement, i.e. inertial terms are neglected. We integrated equation (9) numerically
with the GLMT versions of Fl and Fll by means of a classical RungeKutta routine 1161
and looked for a periodic solution (limit cycle). In fact no such solution was found in the
conditions of figure 6(a). In this case, the particle escapes the hysteresis cycle after step 3
and falls down in the non-symmetrical position (see the broken curve in figure 7(a)). This
behaviour can be understood as the result of a competition between two kinetic processes:
the lateral transition of the particle back to s = 0 at step 3 and the movement of the
particle in the non-symmetrical position from step 3 down to step 2 (see the broken curve
in figure 6(b)). In this case the latter is faster than the former and then prevents the sphere
from coming back into the hysteresis cycle. The particle ultimately moves back to s = 0
but an attitude where the Levitation force is smaller than w . Incidently, we added inertial
terms in equation (9) and found no modification of the result.
Of course the adequacy of equation (9) to represent the sphere movement can be
questioned for exacily the same reasons as those discussed in section 4. A general analysis
should include parasitic convection flows, thermal noise (in figure 6(b) the energy harrier
584 G Martinot-Lugarde et ai
at step 5 in only 2.5 ksT) and imperfections in the set-up configuration. In fact any
small horizontal perturbation can considerably modify the picture. Interestingly, a slight
imperfection in the parallelism between the two beams produces a true cycle, as illustrated
in figure 7(b). In this example, we tilted one beam by 0.5 x rad from vertical. Such a
defect is quite realistic when compared to the experimental error on beam parallelism inside
the optical trap, which is about 0.01 rad. This result proves the existence of limit cycles
in the model in the absence of any fluctuating force. However, the vertical amplitude of
the calculated cycle (about 160 pm) is much larger than that found experimentally (about
40 pm), a mismatch that makes our result only a qualitative one. As we noticed, an
exhaustive analysis should include convection and particle Brownian motion. Though we
are unable to handle this terribly complex problem, it seems obvious that such mechanisms
will help the particle to move over energy barriers (see figure 6(b), steps 4 and 5) and then
will reduce the amplitude of the cycle.

6. Conclusion

We have analysed in detail the equilibrium of a dielectric sphere held by two vertical laser
beams in water, Basically we wanted to compare theoretical off-axis radiation pressure
forces to those found from experiments carried out with polystyrene latex spheres, a few
micrometres in size, and moderately focused visible Gaussian beams. The forces were
calculated in the most general way, using generalized Lorenz-Mie theory [8,10] (the particle
size domain made simple approximations, such as Rayleigb, Rayleigh-Gans or ray optics
[9] inapplicable).
We observed that the presence of the fluid around the levitated particles makes the
comparison delicate, because of Brownian motion and residual large-scale convection.
Fortunately, Brownian excursions are negligible for ‘large’ particles (a z 8 pm). In this
case, and in experiments involving beams of very different intensities (p << l), we also find
that parasitic forces due to convection do not complicate the picture too much, as they just
cause a slight lateral shift in the I ( x ) and x&) responses. In such conditions, theoretical
(GLMT) responses quantitatively fit to experimental data, from which we conclude that GLMT
gives the correct values of the radiation pressure forces in our experimental conditions.
When the sphere is levitated by two beams whose intensities are equal or nearly equal,
the lateral position of the particle is discontinuous compared to the beam separation. This
feature is well explained from the shape of the GLMT forces, but the position of the instability
( X T R ) is greatly affected by convection. Without an independent knowledge of the parasitic
flow velocity around the particle, the value of xm/2 should be taken just as a lower boundary
of the off-centring parameter aTR (see figure I(b)). In [I], XTR/2 and aTR were supposed to
be equal, which led to a systematic error in the ‘experimental’ 4 ( e ) and F s ( e ) profiles.
We observed that the above mentioned lateral instability leads to hysteresis in the particle
position and also to sustained oscillations, which we could explain from the GLMT model.
Interestingly, this situation is an example of a laser permanently dissipating energy inside a
finite system, in the absence of any absorption (all refractive indices are real). Here, energy
is taken from the light source not by suppressing photons but by changing their frequency
through Doppler effect.
To conclude, it is useful to comment on the characteristic sizes involved throughout
this work. At the experiments, we used spheres which were large enough to be optically
resolved. In the same time, these were small enough to be easily levitated and to limit
GLMT series to an acceptable number of terms (Zzn,a/n,h .e 200).
We also used ‘moderately focused’ lasers (00 = 3.7 pm), which made the description
Particle levitation by off-centred laser beams 585

of the beams easy, with a simple first-order Davis representation [17].


We are currently extending GLMT to much smaller beam-waists and particle sizes.

-
Because the beam-wavelength is not (or nearly not) down-scalable, a first difficulty is the
proper representation of tightly focused beams (2wo A) [18]. A second problem is the
particle Brownian motion which, apart from exmemely high power densities, may play a
major role in the particle average equilibrium. Particularly, we expect the 'thermodynamic'
radiation pressure forces not to be simply proportional to the beam power.

Acknowledgments

This work is supported by ULTIMATECH program (CNRS) and is a part of a programme


internationul de coopirafion scientifique between CNRS and the Bulgarian Academy of
Sciences. We thank F Nallet and R Elezgaray (CRF'P) for illuminating discussions. One of
us (MIA) is gratefully indebted to CNRS and to the Commission of European Community
for supporting her stay in France.

References
[I] Angelova M I and Pouligny B 1993 Pure AppL Opt. 2 261
[2] Ashkin A 1970 Phys. Rev. Leu. 24 156
[3] Roosen G and lmbut C 1976 Phys. Ldt. 59A 6
Roosen G 1977 Opt. Commun 21 189
[4] B&er Schut T C, Hesselink G,de Groat B G and Greve J 1991 Cylomefty 12 479
[ 5 ] Walz J Y and Prieve D C 1992 Langmuir 8 3073
[6] Gussgard R, Lindmo T and Brevik I 1992 3, Opt. Soc. Am B 9
[7] Asbkin A 1992 Biophys. J. 61 569
[SI Gouesbet G. Maheu B and Gr€han G 1988 J. Opf. Soc. Am A 5 1427
Gouesbet G. Orkhan G and Maheu B 1990 J. Opt Soc. Am A 7 998
[9] Van de Hulst 1964 Light Scatrering by S m N Particles (New York Wiley)
[IO] Ren K F.Gr6han G and Gouesbet G 1994 Opt, Commun. 108 434
[ I l l Lo& J A and Gouesbet G 1994 J. Opt. Soc. Am A 11 2503
Gouesbet G and Lock I A 1994 J. Opt. Soc. Am A 11 2516
Gouesbet G, Lock JA and Gehan G 1994 Pmc. 7th Workshop on Two-Phme Flows ( E r h g e n , Germny.
Jl-14April) to appear
[I21 Bandrup J and lmmergut E H (ed) Polymer Handbook (New York Wiley)
[I31 Ashkin A. Dziedzic J M,Bjorkholm J E and Chu S 1986 Opt Lett. 11 288
1141 Domb C and Green M S (ed) 1976 Phose Trmitions and Critrcol Phenomena vol 6 (London: Academic)
11-51 Kogelnik H and Li T 1966 Appl. Opt. 5 1550
[I61 Press W H,Teukolsky S A, Vetterling W T and Flannery B P (ed) 1992 Numerical Recipes in Fortran
(Cambridge: Cambridge University Press)
[Ill Davis L W 1979 Phys. Rev. A 19 1177
[IS] Gouesbet G, Lock J A and Gehan G 1995 Appl. Opt. to nppear

Вам также может понравиться