Академический Документы
Профессиональный Документы
Культура Документы
net/publication/260069190
Article in Proceedings of The Royal Society A Mathematical Physical and Engineering Sciences · December 2003
DOI: 10.1098/rspa.2003.1164
CITATIONS READS
116 678
3 authors, including:
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Herch Nussenzveig on 02 April 2015.
Email alerting service Receive free email alerts when new articles cite this article - sign up in the box at the top right-hand
corner of the article or click here
10.1098/rspa.2003.1164
Starting from a Debye-type integral representation valid for a laser beam focused
through a high numerical aperture objective, we derive an explicit partial-wave (Mie)
representation for the force exerted on a dielectric sphere of arbitrary radius, position
and refractive index. In the semi-classical limit, the ray-optics result is shown to
follow from the Mie expansion, holding in the sense of a size average. The equilibrium
position and trap stiffness oscillate as functions of the circumference-to-wavelength
ratio, a signature of interference, not predicted by previous theories. We also present
comparisons with experimental results.
Keywords: optical traps; semiclassical limit; Mie scattering
1. Introduction
Single-beam optical traps, commonly known as optical tweezers, have become a pow-
erful tool, with many applications in physics and biology (Ashkin 1997). They have
been used to measure forces in the piconewton range, opening the way for quan-
titative investigations in some fundamental fields of cell biology, such as molecular
motors (Wang et al . 1998; Mehta et al . 1999). The trapped particles are usually
dielectric microspheres, employed as handles in most quantitative applications (Svo-
boda & Block 1994).
In order to achieve trapping, a microscope objective is used to bring the laser
beam to a diffraction-limited focal spot. Typical values of numerical aperture (NA)
are greater than 1, corresponding to beam-opening angles θ0 > 60◦ . Gaussian beam
models are reasonable approximations for near-paraxial conditions (sufficiently small
opening angles; beam waist much larger than the wavelength). However, diffraction-
limited beams cannot be represented by such models: theoretical treatments based
on near-paraxial approximations (Barton et al . 1989; Gussgard et al . 1992; Ren et
al . 1996) are not valid descriptions of optical tweezers (Svoboda & Block 1994).
Alternative approaches, based on geometrical-optics (GO) approximations (Ashkin
1992; Gu et al . 1997), should not be applicable to the microspheres usually employed
as handles, since their sizes are not much larger than the laser wavelength.
Most measurements employ indirect force calibrations against the Stokes law under
complicated boundary conditions for the flow of surrounding fluid past the micro-
sphere, leading to discrepancies among results by factors of two or more. A reliable
derivation of the trapping force is sorely needed, so as to disentangle the effects of
where k = |k(θ, φ)| = n1 ω/c, ˆ(θ, φ) = x̂ + iŷ and the unit vectors x̂ and ŷ
are obtained from x̂ and ŷ, respectively, by rotation with Euler angles (φ, θ, −φ).
According to (2.1), the incident electric field is a superposition of circularly polar-
ized plane waves. The focus is at r = −R, the point where all plane waves are in
phase, yielding constructive interference and maximum intensity. As the dephasing
among the plane waves increases, the intensity at the focal plane decreases, becom-
ing localized to a transverse dimension of the order of the wavelength. The factor
exp(−γ 2 sin2 θ) results from the Gaussian intensity profile of the laser beam at the
entrance pupil of the objective, I = Im exp(−2ρ2 /w2 ) (where w is the waist, Im is
the intensity at the beam axis, and ρ is the distance to the beam axis), and from the
Abbe sine condition ρ = f sin θ, where f is the focal length, so that γ = f /w. Note
that we take a Gaussian beam before the passage through the objective. The actual
trapping beam, given by (2.1), is not Gaussian, and arbitrary values of θ0 may be
considered, in a consistent way.
The scattered electric and magnetic fields Es and Hs are computed in appendix A.
We denote by n2 the absolute refractive index of the sphere. We neglect absorption,
assuming that both n1 and n2 are real. We employ a spherical coordinate system,
with origin at the centre of the sphere. For each plane-wave component of the input
field (2.1), the scattered field is a corresponding Mie partial-wave series, and the
resulting total scattered field is a linear superposition of Mie components. They are
given in terms of the Mie coefficients aj and bj , which correspond (when multiplied
by i/k) to the scattering amplitudes for electric and magnetic multipole waves. The
Mie coefficients are functions of the size parameter β, and of the relative refractive
index n = n2 /n1 (Bohren & Huffman 1983).
The optical force is computed through the Maxwell stress tensor. Because of
momentum conservation, we can take a spherical surface at infinity:
F = lim − 12 r dΩ r(εE 2 + µ0 H 2 ) . (2.2)
r→∞ S(r)
Following Ashkin (1992), we define the dimensionless force through a vector efficiency
factor,
F
Q= , (2.3)
n1 P/c
where P is the laser power at the sample and c is the speed of light. Q quantifies
how efficiently the available field momentum is transferred to the sphere. Except for
possible nonlinear optical effects (not considered in this paper), Q does not depend
on the laser power P . For normal incidence of a plane wave on a totally reflecting
mirror, Q = 2. In optical tweezers, the maximum value of Q is usually smaller by at
least one order of magnitude.
We assume that the transmission loss through the objective is uniformly dis-
tributed within the focused beam, so that the intensity distribution is not affected
by this process. We may then ignore losses when considering the ratio in (2.3), and
replace P by the power filling the objective aperture,
P = 12 πw2 AIm , (2.4)
where A = 1 − exp(−2γ 2 sin2 θ0 ) is the fraction of available beam power that fills the
objective aperture.
It is convenient to express the force in terms of cylindrical coordinates, with the
origin at the focus and the centre of the sphere at the point R = (ρR , ϕR , zR ). Since
we take circular polarization, the cylindrical components do not depend on ϕR (by
axial symmetry). The final expressions are sums over the variables j, for the total
angular momentum J 2 (eigenvalues j(j + 1)) and m for its z component Jz , of the
form
∞ j
= .
j,m j=1 m=−j
where Gj,m and Gj,m are multipole coefficients of the focused incident beam:
θ0 √
dθ sin θ cos θ e−γ sin θ djm,1 (θ)Jm−1 (kρR sin θ)eikzR cos θ , (2.9)
2 2
Gj,m (ρR , zR ) =
0
∂Gjm
Gj,m (ρR , zR ) = −i , (2.10)
∂(kzR )
where djm,m (θ) are the matrix elements of finite rotations (Edmonds 1957) and
Jm are Bessel functions (Abramowitz & Stegun 1972). When the sphere is aligned
along the optical axis (ρR = 0), we may use the result Jm−1 (0) = δm,1 to sim-
plify (2.7)–(2.10). The resulting expressions, as expected, agree with those obtained
by Maia Neto & Nussenzveig (2000).
For the transverse components, we find
γ2 j(j + 2)(j + m + 1)(j + m + 2)
Qsρ = 4 Im
A j,m
j+1
γ2
Qeϕ = −2 Re (2j + 1)[(aj + bj )Gj,m (G− + ∗
j,m+1 + Gj,m−1 ) ], (2.14)
A j,m
where
θ0 √
G± dθ sin2 θ cos θ e−γ sin θ djm±1,1 (θ)Jm−1 (kρR sin θ)eikzR cos θ .
2 2
jm (ρR , zR ) =
0
(2.15)
The transverse force contains products of multipole coefficients for consecutive values
of m.† Hence, as was remarked following (2.10), it vanishes when ρR = 0, as expected.
We have also considered polarization σ−. The force in this case may also be com-
puted from (2.7)–(2.15), provided that we change the sign of the azimuthal compo-
nent, Qϕ [σ−] = −Qϕ [σ+].
An important check of (2.7)–(2.15), performed next, is to consider their semi-
classical limits, which must be related to the GO approximation.
3. GO limit
In this section, we obtain a closed analytical representation for the force in the GO
approximation, starting from the exact results (2.7)–(2.15), in the form of an integral
over θ and φ of elementary functions. Previous approaches relied on time-consuming
numerical vector sums of the forces exerted by each ray (Ashkin 1992; Gu et al .
1997). Moreover, our approach takes into account the correct pre-factors describing
the intensity distribution in the focused beam, verifying the Abbe sine condition.
The derivation of the GO approximation from the exact partial-wave series is
a textbook example of the subtle nature of semi-classical approximations (Berry
& Mount 1972). It involves the following stages (Nussenzveig 1992; Nussenzveig &
Wiscombe 1980).
† Equations (2.7), (2.11) and (2.12), giving the scattering component Qs , agree with the results
of Barton et al. (1989) (see also Farsund & Felderhof 1996) if we replace Gjm by paraxial multipole
coefficients. In the same sense, the extinction component Qe as given by (2.8), (2.13) and (2.14) is also
formally equivalent to the result of Barton et al. To make the connection for the axial component Qez ,
we write cos θdjm,1 (θ) as a linear combination of dj+1 j j−1
m,1 (θ), dm,1 (θ) and dm,1 (θ), allowing us to express
Gjm in terms of Gj+1,m , Gjm and Gj−1,m . Likewise, for the transverse components Qeρ and Qeϕ , we
expand G± jm as a linear combination of Gj m with different j .
The force turns out to be the sum of two terms, Q = Q + Q⊥ , each one given by
an integral over the rays forming the incident focused beam. The axial components
are given by (Θ is the Heaviside step function)
2π
2γ 2 θ0 −2γ 2 sin2 θ
Qz = 2
dθ sin θ cos θe dφ F (R⊥ )Θ(a − R⊥ ), (3.3)
πA 0 0
2γ 2 θ0
Q⊥ dθ sin θ cos θe−2γ sin θ
2 2
z =
πA 0
2π
F ⊥ (R⊥ )
× dφ [zR sin2 θ − 12 ρR sin(2θ) sin φ]Θ(a − R⊥ ), (3.4)
0 R⊥
,⊥ ,⊥
where F ,⊥ = 12 (FTE + FTM ), and
r cos(2θ1 ) + cos[2(θ1 − θ2 )]
F = 1 + r cos(2θ1 ) − (1 − r )2 , (3.5)
1 + r2 + 2r cos(2θ2 )
r sin(2θ1 ) + sin[2(θ1 − θ2 )]
F⊥ = r sin(2θ1 ) − (1 − r )2 . (3.6)
1 + r2 + 2r cos(2θ2 )
The r ( = TE, TM) are the Fresnel reflection coefficients for a plane interface, and
θ1 represents the angle of incidence of a given ray at the surface of the sphere (angle
between the ray direction and the normal at the point of incidence). It is given by
R⊥
sin θ1 = . (3.7)
a
θ2 is the angle of refraction inside the sphere,
sin θ1
sin θ2 = . (3.8)
n
Proc. R. Soc. Lond. A (2003)
Downloaded from rspa.royalsocietypublishing.org on May 24, 2011
In the axial case, ρR = 0 and equations (3.2) and (3.7) yield R⊥ = |zR | sin θ and
sin θ1 = |zR | sin θ/a. Hence, in this particular case, the results given by (3.3) and (3.4)
become equivalent to those of Maia Neto & Nussenzveig (2000), whereas the radial
component vanishes.
The factors F and F⊥ in the above expressions can be identified with the longi-
tudinal and transverse components of the force exerted by a given ray with respect
to its direction of propagation r̂(θ, φ). Thus we may also derive (3.3)–(3.10) in the
framework of ray-optics theory. Note that the set of rays that hit the sphere and
thereby contribute to the force is defined by the inequality
R⊥ a, (3.11)
which explains the step function in (3.3), (3.4) and (3.9), (3.10). The remaining
pre-factors, not accounted for previously,† describe the intensity distribution among
different rays in the focused beam, as implied by the Abbe sine condition and the
intensity profile at the entrance aperture of the objective. Indeed, the power dP
contained in a solid angle dΩ = sin θ dθ dφ is given by
−2ρ2
dP = Im exp ρ dρ dφ, (3.12)
w2
The force exerted by a ray of given power dP and direction r̂(θ, φ) was computed
by Roosen (1979) (note that R⊥ determines the direction perpendicular to the ray
in the plane of incidence, which contains the centre of the sphere):
⊥ R⊥ dP
dQ = F r̂ + F . (3.14)
R⊥ P
† Q⊥z , which provides the restoring force along the axial direction, is overestimated by Ashkin (1992),
as a consequence of neglecting the sine condition (cf. Gu et al. 1997) and the corresponding factor cos θ
in (3.4). This diminishes the contribution of rays at large angles.
1.5
1.0
zeq / a
0.5
0
0 5 10 15 20 25
ka
Figure 1. Equilibrium position as a function of β = ka, in units of the sphere radius.
The dashed horizontal line represents the GO value zeq /a = 0.286.
Decomposing the unit vectors in the right-hand side of (3.14) into cylindrical com-
ponents, using (3.13) and integrating over dΩ, we recover, now in the context of
ray-optics theory, the results (3.3), (3.4) and (3.9), (3.10).
The integrals yielding the optical force in the GO approximation can be easily
performed by standard numerical techniques. As discussed in the next section, they
provide a reliable check of the partial-wave numerical results.
4. Numerical results
A remarkable qualitative feature of the exact results, not present in the GO theory,
is the oscillatory behaviour of the force as a function of the size parameter β = ka.
When β 1 and |n − 1| 1, the force, with the sphere centre at the focus, is a
nearly sinusoidal function of the phase ∆ = 4n2 ωa/c associated with radial round-
trip propagation inside the sphere (Maia Neto & Nussenzveig 2000),
Q(R = 0) ≈ 8rcos θ sin2 ( 12 ∆),
where · denotes an average over the intensity distribution of the focused beam
and r = 12 (rTE + rT M ) 1. In this approximation, the force is proportional to the
reflectivity of a parallel plate interferometer of optical length 2n2 a. In particular, it
vanishes when the fields back-reflected at the two interfaces interfere destructively.
In this case, the equilibrium position zeq is very close to the focus. More generally,
zeq as a function of β shows oscillatory behaviour, which results from the sinusoidal
variation of the force at R = 0. In figure 1, we plot zeq as a function of β, for a
fixed value of the wavenumber k. The parameters correspond to the experiment of
Ghislain et al . (1994): λ0 = 1.064 µm, A = 0.91, θ0 = 70◦ (angle in water), n1 = 1.32
and n2 = 1.57.†
Very small spheres correspond to the Rayleigh limit, β 1. The force is then
2
proportional to the gradient of Ein , with the dominant contribution coming from the
electric dipole term in (2.8) and (2.13) (Qs is negligible). Hence zeq /a → 0 in this
limit. We see that zeq /a has a peak near β = 1; in this intermediate range, the typical
† We have employed Cauchy’s dispersion formula (Bateman et al. 1959) for computing the refractive
indices of water and polystyrene.
0.5
0.4
0.3
z−/ a
0.2
0.1
0
0 0.2 0.4 0.6
ρR / a
Figure 2. Equilibrium position with a lateral external force. The plots represent the root z̄ of
Qz (ρR , z̄(ρR )) = 0 as a function of ρR . All lengths are expressed in units of the sphere radius.
Solid line, a = 3.05 µm (β = 23.8); dashed line, a = 3.09 µm (β = 24.1); dotted line, a = 3.15 µm
(β = 24.2); dot-dashed line, GO.
−0.1
−0.035
Qρ
Qρ
−0.2 −0.040
−0.045
0.13 0.14 0.15 0.16
−0.3 ρR / a
Figure 3. Transverse force Qρ , calculated at the point (ρR , z̄(ρR )) defined in (4.1), as a function
of ρR /a. Qρ is negative, corresponding to a force pointing towards the optical axis. Solid line,
a = 3.05 µm; dashed line, a = 3.09 µm; dotted line, a = 3.15 µm; dot-dashed line, GO.
length-scale of the optical potential well is still partly governed by the wavelength
λ = λ0 /n1 . For larger values of β, we find the oscillatory behaviour arising from
interference. We also show in figure 1 the value computed in the GO approximation
(horizontal dashed line).
Usually, in optical tweezers experiments, a lateral viscous drag force is applied
to the sphere. The sphere is then displaced from the position (ρR = 0, zR = zeq )
and a new equilibrium position is reached at a given distance ρR from the axis. If
the external force has no axial component, the new equilibrium position (ρR , z̄(ρR ))
satisfies
Qz (ρR , z̄(ρR )) = 0. (4.1)
0.2
Qmax
0.1
0
0 5 10 15 20 25 30
ka
Figure 4. Maximum transverse force Qmax as a function of β = ka. Theoretical results: par-
tial-wave (‘guide-the-eye’ solid line) and GO value 0.275 (horizontal dashed line). Experimental
results of Ghislain et al . (1994) with reported values of radius (squares), and with rescaled radius
(circles).
250 60
50
κ (pN µm−1)
200 40
30
κ (pN µm−1)
150 20
10
100 0 2 4 6 8 10 12 14
k (µm−1)
50
0 5 10 15 20 25 30
ka
Figure 5. Transverse trap stiffness κ as a function of β = ka, for a fixed wavenumber
k = 7.8 µm−1 and power P = 60 mW. Theoretical results: partial-wave (solid line) and GO
result κ = 541/β pN µm−1 (dashed line). Experimental results of Ghislain et al . (1994) with
reported values of radius (squares), and with rescaled radius (circles). Insert: transverse trap
stiffness as a function of k, for a fixed radius a = 2.0 µm. Solid line, partial-wave results; dashed
horizontal line, GO result κ = 34.8 pN µm−1 .
There are no roots of (4.1) beyond a specific value of ρR , close to 0.7a in the numer-
ical example considered in figure 2. We define ρmax as the maximum value of ρR .
It corresponds to the maximum external transverse force that can be equilibrated
by the optical force. In figure 4, we plot Qmax ≡ −Qρ (ρmax , z̄(ρmax )) as a function
of the radius,† as well as the GO value Qmax = 0.275. No fitting is employed when
computing the theoretical values, but we apply a global rescaling of the parameter
β to the experimental values reported by Ghislain et al . (1994). Good agreement
is obtained for both Qmax and the transverse trap stiffness (see figure 5), provided
that we divide by a factor of 2 the values of β that result from the data of Ghislain
et al . (1994). We plot the experimental points with (circles) and without (squares)
rescaling in figures 4 and 5. A possible source of this discrepancy (see § 5 for further
discussion) may be the effect of spherical aberration (Ghislain et al . 1994), which
was theoretically analysed in the Rayleigh limit by Yao et al . (2001).
The transverse trap stiffness is given by
n1 P ∂Qρ
κ=− , ρ = 0, z = zeq . (4.2)
c ∂ρ
To calculate κ, we first derive from (2.11) and (2.13) the partial-wave expansion for
∂ρ Q, and then replace it into (4.2). In the GO approximation, Qρ depends on ρ only
through ρ/a. Hence the derivative in (4.2) is proportional to 1/a,
n1 P ∂Qρ (ρ/a = 0) k
κ=− . (4.3)
c ∂(ρ/a) β
In figure 5, we plot the results, for the same parameters employed above, against
the size parameter β, for a fixed wavelength. As before, we take the parameters of
Ghislain et al . (1994), with a power P = 60 mW. There is good agreement with the
GO result κ = (541/β) pN µm−1 for large β, except for a small-amplitude oscilla-
tion,‡ which is related to the interference effect discussed in connection with figure 1.
For very small spheres, on the other hand, the optical force is proportional to the
static polarizability of the dielectric sphere (Rayleigh regime), and hence κ vanishes
as β 3 . The crossover between these two regimes is associated with a peak near β = 3.
The experimental points of Ghislain et al . (1994) also feature a peak, which agrees
with the theoretical prediction, provided that we take, as in figure 4, one half of the
diameter values reported by Ghislain et al . (1994).
It is probably easier to verify experimentally the oscillations displayed in figures 1
and 5, which result from interference, by scanning the laser frequency. In this case,
the expected variation of the trap stiffness is shown in the insert of figure 5, for a
microsphere with a diameter of 2 µm, as a function of wavenumber k. The oscillations
around the GO value have the expected period δk = π/(2na) = 0.66 µm−1 corre-
sponding to a 2π variation of the round-trip phase ∆. In this example, five periods of
oscillation would be scanned by tuning the vacuum wavelength from 970 to 700 nm,
which is feasible with a Ti:Sapphire laser (trapping with a tunable CW Ti:sapphire
laser has been reported by Neuman et al . (1999)).
5. Conclusion
Partial (qualitative) agreement between wave and GO is found for size parameters
β > 5. However, GO clearly overestimates the maximum transverse force in the
range shown in figure 4, whereas, for the equilibrium position (figure 1) and the trap
stiffness (figure 5), the GO values tend to approach more closely the average of the
partial-wave results over a period of oscillation.
We have not taken into account the spherical aberration of the incident focused
beam, produced by the objective itself (which is corrected for the visible and not
for the infrared wavelength employed by Ghislain et al . (1994)) and by the glass–
water planar interface. Rays at the edge of the focused beam intersect the axis in
the interval between the paraxial (Gaussian) focus and the interface.
Because of spherical aberration, the laser spot size increases with increasing depth
into the specimen chamber, so that both axial and transverse intensity gradients
decrease. Ghislain et al . (1994) reported a decrease of axial stiffness with depth and
Felgner et al . (1995) found a reduction of both axial and transverse maximal trapping
forces. This is in line with figure 4, since the experimental values of the maximum
transverse force lie below the theoretical curve.
The stretching of the focal region by spherical aberration may help us to under-
stand why agreement between theory and experiment was improved by rescaling the
size parameter in figures 4 and 5 (no additional fitting was employed).
Inclusion of spherical aberration in the theoretical treatment thus appears to be
required. We plan to address this question in the context of a comprehensive com-
parison with experimental results based on a new technique for measuring the trap
stiffness (Viana et al . 2002a) and the power available at the sample (Viana et al .
2002b).
We thank Oscar Mesquita, Nathan Viana, Carlos Cesar and Adriana Fontes for valuable dis-
cussions, and Warren Wiscombe for suggestions on numerical integration and Mie scattering
calculations. This work was completed while H.M.N. held a National Research Council Research
Associateship Award at the NASA Goddard Space Flight Center. P.A.M.N. acknowledges sup-
port by CNPq, PRONEX, FAPERJ and the Millenium Institute of Quantum Information. We
thank the José Bonifácio Foundation for its support of the COPEA Optical Tweezers Labora-
tory.
θ0 ∞
E0 4π(2J + 1)
cos θk e−γ
2 2
E sin θk
Πin (r, θ, φ) = dθk sin θk iJ−1 jJ (kr)
k 0 J(J + 1)
J=1
J 2π
× dJM,1 (θk )YJM (θ, φ) dφk eik·R e−i(M −1)φk .
M =−J 0
(A 3)
The integral over φk above is calculated in terms of the cylindrical Bessel function
JM −1 :
2π
dφk eik·R e−i(M −1)φk = 2π(−i)M −1 eikzR cos θk JM −1 (kρR sin θk )e−i(M −1)ϕR .
0
(A 4)
Once the Debye potentials for the incident field are known, it is a simple matter
to obtain the scattered field by considering the boundary conditions at the surface of
the sphere. ΠsE , representing the electric multipole potential for the scattered field,
is written in terms of the Mie coefficients aJ :
E0 J−M
ΠsE (r, θ, φ) = −2π i GJM (ρR , zR )e−i(M −1)ϕR
k
J,M
4π(2J + 1) (1)
× aJ hJ (kr)YJM (θ, φ), (A 5)
J(J + 1)
(1)
where hJ is the spherical Hankel function and GJM is defined in (2.9). The result
for the magnetic Debye potential is computed in a similar way, in terms of the Mie
coefficients bJ .
From the multipole expansions of the scattered field, we calculate the optical force
using (2.2) (a detailed derivation of the optical force from the Debye potentials was
presented by Farsund & Felderhof (1996)).
(B 1)
where
m n
z = cos θ, µ= , ν= , (z − µν)2 < (1 − µ2 )(1 − ν 2 ) (B 2)
j j
and the inequality defines the classically allowed (oscillatory) region. The lower limit
of integration was not determined by Brussaard & Tolhoek (1957), except for n = 0,
when the rotation matrix reduces to an associated Legendre function.
Employing this reduction, as well as results for the asymptotic expansion of Legen-
dre functions of large degree and order (Thorne 1957), we find that (B 1), for n = 1,
becomes
1/2
2
j
dm,1 (θ) ≈ (λ2 sin2 θ − m2 )−1/4 cos[fλ,m (θ) + ξλ,m ], (B 3)
π
where
λ≡j+ 1
2 (B 4)
and
λ cos θ m cot θ
fλ,m (θ) ≡ λ arccos √ − m arccos √ . (B 5)
λ 2 − m2 λ 2 − m2
The domain of validity in (B 2) becomes
m+1
|sin( 12 θ)| < . (B 6)
2j
The phase constant ξλ,m in (B 3) results from the unknown lower limit of integra-
tion in (B 1). In order to determine it, we apply the relationship (Gel’fand & Shapiro
1956)
j(j + 1) sin θ[djm,1 (θ) + djm,−1 (θ)] = −2mdjm,0 (θ), (B 7)
together with the symmetry relation (Brink & Satchler 1968)
djm,n (θ) = (−1)j−m djm,−n (π − θ). (B 8)
Combining these relations for θ = 12 π, we get
m
ξλ,m = (m + 34 )π + arccos . (B 10)
λ
A similar procedure can be applied to dj−m,1 (θ). The final WKB results are
1/2
j m+1 2 2 −1/4 m
dm,1 (θ) ≈ (−1) (λ sin θ − m )
2 2
cos fλ,m (θ) + arccos − 4π ,
1
π λ
(B 11)
1/2
2 m
dj−m,1 (θ) ≈ (λ2 sin2 θ − m2 )−1/4 cos fλ,m (θ) − arccos − 14 π . (B 12)
π λ
These results appear to be new. They have been numerically tested in several
sample cases, showing very good agreement and consistency with expected WKB
typical features.
Gj,m (ρR , zR )
2π θ0
−3/2
= (2π) dφ dθ sin θ(λ2 sin2 θ − m2 )−1/4 H(θ)
0 0
× {exp[ikψλ,m
+
(ρR , zR ; θ, φ)]
−
+ exp[ikψλ,m (ρR , zR ; θ, φ)]}, (C 3)
where
±
ψλ,m (ρR , zR ; θ, φ) ≡ zR cos θ + ρR sin θ sin φ − ηφ ± [fλ,m (θ) + ξλ,m ]/k, (C 4)
with (m 1)
m−1 m
η≡ ≈ . (C 5)
k k
The asymptotic behaviour of (C 3) in the GO limit, k → ∞, is obtained by apply-
ing the method of stationary phase for double integrals (Born & Wolf 1980). The
The relationship between a given ray direction and θ1 , as given by (3.7), follows
from (C 8) and (D 1).
The WKB approximation to the Mie coefficients is obtained by substituting all
cylindrical functions by their Debye asymptotic expansions. Introducing the notation
χ ≡ β[cos θ1 − ( 12 π − θ1 ) sin θ1 ] − 14 π,
(D 3)
ζ ≡ nβ[cos θ2 − ( 21 π − θ2 ) sin θ2 ] − 14 π,
R22 = R = −R11 , T21 = 1+R , T12 = 1−R , T12 T12 = 1−R2 = 1−r ,
(D 5)
where r is the reflectivity.
We also need γλ+1/2 . In order to relate it with γλ−1/2 , we employ recursion relations
among cylindrical functions, with the following result, to dominant order:
1 −2i(χ+θ1 ) (1 − r )e2i(θ2 −θ1 ) e2i(ζ−χ)
γλ+1/2 = 1 + R e − . (D 6)
2 1 − R e2i(ζ+θ2 )
While, according to (D 1), θ1 and θ2 are slowly varying functions of β, χ and ζ
are rapidly varying, by (D 3). Accordingly, by defining · to be an average over one
period 2π in ζ (corresponding to a small range in β), we get, to leading order,
γλ−1/2 = 12 . (D 7)
On the other hand, equations (D 4) and (D 6) yield
∗
γλ−1/2 γλ+1/2
1 1
= 1 − r e2iθ1 + (1 − r )2 e2i(θ1 −θ2 ) .
4 1 − r e−2iθ2 + R e2iζ − R e−2iθ2 e−2iζ
(D 8)
The evaluation of all such averages follows from the formula
α + β cos(2ζ) + γ sin(2ζ) βb + γc βb + γc 1
= 2 + α− a √ , (D 9)
a + b cos(2ζ) + c sin(2ζ) b + c2 b2 + c2 a − b2 − c2
2
where the average is over a period 2π in the variable 2ζ. The result is
∗ 1 (1 − r )2 e2i(θ1 −θ2 )
γλ−1/2 γλ+1/2 = 1 − r e2iθ1 + . (D 10)
4 1 + r e2iθ2
In terms of the definitions (3.5) and (3.6), it follows from (D 10) that
aj a∗j+1 + bj b∗j+1 = 1 − 12 (F + iF ⊥ ). (D 11)
The ‘cross-polarization’ average a∗λ−1/2 bλ−1/2 will not be needed, because the cor-
responding term in the partial-wave series in (2.7), (2.11) and (2.12) is negligible in
the GO limit.
sin θ|H(θ)|2 ∂
≈ exp −ik ψλ,m
2πkR⊥ ∂λ
sin θ|H(θ)| (λ cos θ − i λ2 sin2 θ − m2 )
2
= √ (E 8)
2πkR⊥ λ 2 − m2
and
j(j + 2)(j − m + 1)(j + m + 1) 2
≈ λ − m2 . (E 9)
j+1
The last, cross-polarization term in (2.7) is of order λ−1 times smaller and can be
neglected.
Combining these results with (D 11), we derive, from (2.7),
θ0
4γ 2
Qsz = − dθ cos θ sin θ|H(θ)|2
A
0
2π
dφ λ2 sin2 θ − m2 ⊥
× Θ(a − R⊥ ) 2 − F + F . (E 10)
0 2π kR⊥ cos θ
Adding up (E 7) and (E 10), and taking (E 4) into account, we obtain the final GO
result for Qz as given by (3.3) and (3.4).
By (2.15), G+j,m−1 differs from Gj,m by having an extra sin θ under the integral
sign, as well as by the substitution m → m − 1 in the order of the Bessel function,
which leads to an extra factor e−iφ when replaced in (2.13) and (2.14) (cf. (C 1)),
and similarly for G−j,m+1 . This leads to
sin2 θ
|J|[Gj,m (G− ∗ ∗
j,m+1 ) ∓ Gj,m (Gj,m−1 ) ] =
+
|H(θ)|2 (eiφ ∓ e−iφ ). (E 11)
2πkR⊥
Since 2π
dφ cos φg(sin φ) = 0 (E 12)
0
for any function g, we find (by (3.2), R⊥ depends on φ only through sin φ)
Qeϕ = 0. (E 13)
We also derive from (E 11)
2π
8γ 2 θ0 dφ
Qeρ = dθ sin2 θ|H(θ)|2 Θ(a − R⊥ ) sin φ. (E 14)
A 0 0 2π
On the other hand, similarly to (E 8), we find
∗ sin θ|H(θ)|2 ∂ ∂
|J|Gj,m Gj+1,m+1 = exp −ik + ψλ·m + iφ , (E 15)
2πkR⊥ ∂λ ∂m
and, taking (B 12) into account,
Gj,−m (Gj+1,−m−1 )∗ = Gj,m (Gj+1,m+1 )∗ . (E 16)
Thus, neglecting the cross-polarization terms in (2.12), we find
Qsϕ = 0. (E 17)
With
j(j + 2)(j + m + 1)(j + m + 2)
≈ λ + m, (E 18)
j+1
References
Abramowitz, M. & Stegun, I. 1972 Handbook of mathematical functions. New York: Dover.
Ashkin, A. 1992 Forces of a single beam gradient laser trap on a dielectric sphere in the ray
optics approximation. Biophys. J. 61, 569–582.
Ashkin, A. 1997 Optical trapping and manipulation of neutral particles using lasers. Proc. Natl
Acad. Sci. USA 94, 4853–4860.
Barton, J. P., Alexander, D. R. & Schaub, S. A. 1989 Theoretical determination of the net
radiation force and torque for a spherical particle illuminated by a focused laser beam. J.
Appl. Phys. 66, 4594–4602.
Bateman, J. B., Weneck, E. J. & Eshler, D. C. 1959 Determination of particle size and concen-
tration from spectrophotometric transmission. J. Colloid Sci. 14, 308–329.
Berry, M. V. & Mount, K. E. 1972 Semiclassical approximations in wave mechanics. Rep. Prog.
Phys. 35, 315–397.
Bohren, C. F. & Huffman, D. R. 1983 Absorption and scattering of light by small particles.
Wiley.
Born, M. & Wolf, E. 1980 Principles of optics, 6th edn. Oxford: Pergamon.
Brink, D. M. & Satchler, G. R. 1968 Angular momentum, 2nd edn, p. 147. Oxford: Clarendon.
Brussaard, P. J. & Tolhoek, H. A. 1957 Classical limits of Clebsch–Gordan coefficients, Racah
l
coefficients and Dmn (φ, θ, ψ) functions. Physica 23, 955–971.
Edmonds, A. R. 1957 Angular momentum in quantum mechanics. Princeton University Press.
Farsund, Ö. & Felderhof, B. U. 1996 Force, torque and absorbed energy for a body of arbitrary
shape and constitution in an electromagnetic radiation field. Physica A 227, 108–130.
Felgner, H., Müller, O. & Schliwa, M. 1995 Calibration of light forces in optical tweezers. Appl.
Optics 34, 977–982.
Gel’fand, I. M. & Shapiro, Z. Ya. 1956 Representations of the group of rotations of 3-dimensional
space and their applications. Am. Math. Soc. Transl. 2 2, 207–316.
Ghislain, L. P., Switz, N. A. & Webb, W. W. 1994 Measurement of small forces using an optical
trap. Rev. Scient. Instrum. 65, 2762–2768.
Gu, M., Ke, P. C. & Gan, X. S. 1997 Trapping force by a high numerical-aperture microscope
objective obeying the sine condition. Rev. Scient. Instrum. 68, 3666–3668.
Gussgard, R., Lindmo, T. & Brevik, I. 1992 Calculation of the trapping force in a strongly
focused laser beam. J. Opt. Soc. Am. B 9, 1922–1930.
Maia Neto, P. A. & Nussenzveig, H. M. 2000 Theory of optical tweezers. Europhys. Lett. 50,
702–708.
Mehta, A. D., Rief, M., Spudich, J. A., Smith, D. A. & Simmons, R. M. 1999 Single-molecule
biomechanics with optical methods. Science 283, 1689–1695.
Neuman, K. C., Chadd, E. H., Liou, G. F., Bergman, K. & Block, S. M. 1999 Characterization
of photodamage to Escherichia coli in optical traps. Biophys. J. 77, 2856–2863.
Nussenzveig, H. M. 1992 Diffraction effects in semiclassical scattering. Cambridge University
Press.
Nussenzveig, H. M. & Wiscombe, W. J. 1980 Efficiency factors in Mie scattering. Phys. Rev.
Lett. 45, 1490–1494.
Ren, K. F., Grehan, G. & Gouesbet, G. 1996 Prediction of reverse radiation pressure by a
generalized Lorenz–Mie theory. Appl. Optics 35, 2702–2710.
Richards, B. & Wolf, E. 1959 Electromagnetic diffraction in optical systems. II. Structure of the
image field in an aplanatic system. Proc. R. Soc. Lond. A 253, 358–379.
Roosen, G. 1979 La lévitation optique de sphères. Can. J. Phys. 57, 1260–1279.
Svoboda, K. & Block, S. M. 1994 Biological applications of optical forces. A. Rev. Biophys.
Biomol. Struct. 23, 247–285.
Thorne, R. C. 1957 The asymptotic expansion of Legendre functions of large degree and order.
Phil. Trans. R. Soc. Lond. 249, 597–620.
van de Hulst, H. C. 1957 Light scattering by small particles. Wiley.
Viana, N. B., Freire, R. T. S. & Mesquita, O. N. 2002a Dynamic light scattering from an optically
trapped microsphere. Phys. Rev. E 65, 1–11.
Viana, N. B., Mesquita, O. N. & Mazolli, A. 2002b In situ measurement of laser power at the
focus of a high numerical aperture objective using a microbolometer. Appl. Phys. Lett. 81,
1765–1767.
Wang, M. D., Schnitzer, M. J., Yin, H., Landick, R., Gelles, J. & Block, S. M. 1998 Force and
velocity measured for single molecules of RNA polymerase. Science 282, 902–907.
Yao, X.-C., Li, Z.-L., Guo, H.-L., Cheng, B.-Y. & Zhang, D.-Z. 2001 Effects of spherical aber-
ration on optical trapping forces for Rayleigh particles. Chin. Phys. Lett. 18, 432–434.