Optics Communications 89 ( 1992 ) 363-369 OPTICS

North-Holland COMMUNICATIONS

Theory of Mie resonances and ripple fluctuations

L G . Guimar~es
Centro Brasileiro de Pesquisas Fisicas, Rio de Janeiro, R J 22290, Brazil

and

H.M. Nussenzveig
Departamento de Fisica, PUC, Rio de Janeiro, RJ 22452, Brazil

Received 12 December 1991

CAM (complex angular momentum) results for Mie resonances and for the associated ripple fluctuations in Mie scattering are
reported. Resonance positions and widths are accurately determined. They are described by Regge trajectories, that are found
also to produce strong antiresonant contributions. The background includes a new forward glory term. CAM theory provides a
physical picture of all contributions and the first accurate non-partial-wave fit of the total cross section including both background
and ripple. The backscattering cross section is also reproduced: both background (dominated by tunneling) and resonance con-
tributions must be taken into account.

1. Introduction fluctuations, have been lacking hitherto.

Mie scattering gives rise to extremely sharp reso-

nances, that appear in the form of a rapidly-varying, 2. Physical interpretation
complicated pattern of quasiperiodic fluctuations,
known as the "ripple", in all types of Mie cross sec- For a sphere of radius a with complex refractive
tions [ 1 ]. This pattern depends sensitively on all pa- index N, resonances arise for ReN> 1. Their origin
rameters, showing some quasichaotic features. and physical interpretation become very conspicu-
Observed manifestations of Mie resonances in- ous in CAM theory [ 5 ] with the help of two basic
clude resonance fluorescence, lasing and nonlinear concepts: the effective potential and the localization
optical effects in microspheres [ 2 ]. They have been principle.
applied to optical particle sizing and to the spec- The effective potential appears in the radial equa-
troscopy of levitated particles [ 2,3 ]. Possible appli- tion for the Debye potentials [ 8 ], which, according
cations to cavity QED in the visible [4,5 ] and to the to the well-known analogy between optics and me-
localization of light [6] have been suggested. Re- chanics, has the form ofa Schr6dinger equation with
cently, cavity-QED-enhanced stimulated emission in this potential (in units such that h = 2 m = 1 ). We take
dye-doped ethanol droplets has been detected [ 7 ]. N to be real, for definiteness. For the/th partial wave
It has long been known that the ripple is associated and a wave number k, the effective potential Ux(r)
with resonances in individual partial waves [ 1 ] . is the sum of an attractive square well of depth
However, a physical explanation and a detailed fit (N 2- 1 )k 2 with the centrifugal potential 22/r 2, where
(except by numerical partial-wave summation) of 2 = / + ½represents the angular momentum. As shown
all features of the cross sections, including both the in fig. la, U~takes the form of a potential pocket sur-
smoothly-varying background and the sharp tipple rounded by a centrifugal barrier, giving rise to sharp

0030-4018/92/$05.00 © 1992 Elsevier Science Publishers B.V. All fights reserved. 363
Volume 89, number 5,6 OPTICS COMMUNICATIONS 15 May 1992

U~(r)

_ _ _

k2| ~ ~ n=0

0 ba b r
N
(a) (b)

Fig. 1. (a) Effectivepotential Ua(r) for a transparent sphere with N> I. Resonant wave functions with n=0, l are sketched (the black
circle indicates a node). (b) Correspondingpath for a ray with impact parameter b, that penetrates inside by tunneling and is multiply
reflected between the surface and the caustic r= bin (inner turning point).

resonances between the top T and the bottom B of leads to very large ratios of internal to external am-
the pocket. plitudes (fig. 1a). The resonances are "quasibound"
The localization principle [ 9,10 ] associates a con- states of light, that would become bound in the limit
tribution from angular momentum ;t (extended from of zero leakage. For a given 2, there may exist several
discrete to continuous values) with an incident ray resonances, characterized by a "family number" n
having an impact parameter ( =0, l, 2, ...), the number of nodes of the radial wave
function inside the well in this limit.
b(2) = 2 / k . ( 1)
The lower the value of n, the deeper the quasi-
According to fig. I a, the resonances occur in the "en- bound state lies within the well and the smaller the
ergy" range barrier transmissivity. This may lead to very long
lifetimes for the low-n resonant states, correspond-
(2/Na)2 < k 2 < (2/a)2 (2)
ing to many internal reflections. In a ray picture, this
so that they correspond to impact parameters b such is analogous to orbiting. Because of the low trans-
that a < b < Na, i.e., to incident rays passing outside missivity, the internal field amplitude within the al-
of the sphere. In this range, there are three radial lowed region near the surface can build up to very
turning points (fig. la): r = b / N , r = a and r=b. high values during the large resonance lifetime, lead-
To get across the forbidden region a < r < b re- ing to nonlinear optical effects.
quires tunneling through the centrifugal barrier. In- This process can also be described as the excita-
side the sphere, there are multiple reflections be- tion of natural modes of oscillation, defined as con-
tween r = b / N and r = a (fig. lb), corresponding to taining only outgoing radiation [ 11 ]. The corre-
internal incidence beyond the critical angle, accom- sponding complex eigenfrequencies are poles of the
panied by nearly total reflection: the small leakage to S-function, with real parts associated with resonance
the outside region at each internal reflection arises positions and imaginary parts associated with reso-
from reverse tunneling. nance widths or lifetimes [ 12 ].
Near "resonance energies", the (polarization-de-
pendent) matching between oscillatory solutions i n
the allowed regions and exponential-type ones in for-
bidden regions, with regular behavior at the origin,

364
Volume 89, number 5,6 OPTICS COMMUNICATIONS 15 May 1992

3. Regge trajectories where ~vrepresents a radial phase integral across the

forbidden region.
The resonances are determined as complex poles Higher-order WKB approximations to the roots of
of the S-function, verifying the transcendental equa- eq. (2) are obtained [ 14-16 ] by employing the De-
tion [ 13 ] bye asymptotic expansions of the cylindrical func-
tions, including a subdominant contribution to the
ejaln'Ja(a)-flln'H)l)(fl)=½(1-ej) , (2) Hankel function in order to obtain the imaginary
where J and H are cylindrical functions, In' denotes part. For fl~ 102, a range suitable for applications to
the logarithmic derivative, fl-=ka is the size param- microspheres, they lead to many useful estimates
eter, a - N i l , el = 1 [3= 1, perpendicular (M) polar- about resonance locations a4td spacings [ 15,16 ], but
ization] and e2=N_2 [./'=2, parallel (E) polariza- their accuracy does not suffice to locate very narrow
tion ]. resonances within less than their linewidth. This can
For physical l (l= 1, 2, ...), the real and imaginary be accomplished very efficiently, however, by em-
parts of the roots ofeq. (2) in the fl-plane determine ploying uniform asymptotic expansions of the cylin-
the position and width of the resonances in t h e / t h drical functions [17]. Details are given elsewhere
partial wave, respectively. For a given pair (l, j), the [16].
resonances are labelled by the integer n. According The lowest-n families of Regge trajectories for both
to eq. (1) and fig. la, Reflo ~ lies between 2 / N and polarizations, with N = 1.33 and # in the range from
2, and both Refl0~ and Imfl0~ increase monotonically 50 to 60 are shown in fig. 2, where ReZj, and
with n. logmlm2j, are plotted against p. Resonances occur
For physical fl, the roots of eq. (2) in the complex where Re2j, intersects a half-integral (physical)
2-plane are Regge poles [ 12,14 ]. The integer n is the value; two resonance pairs in the interval 58.2~ p
"family number" [ 5 ] of the Regge trajectories 2j~ (fl) ~<59 are marked by open circles. The linear behavior
that are described by the roots as fl varies. Each Regge of the trajectories in this logarithmic plot shows that
trajectory gives a unified description of all reso- the p-plane trajectories approach the real axis ex-
nances with the same family number [ 12 ]. ponentially fast as fl increases, leading to very large
In lowest-order WKB approximation, for real N, Q factors when losses can be neglected [ 18 ].
the real part of the poles is determined by The well-known physical interpretation of the res-
onance widths in potential scattering [ 19 ] also ap-
plies here: their inverse, proportional to the reso-
2 i keff(r) d r = ( 2 n + l ) T t , ( n = 0 , 1 , 2 .... ) ,
2/Nk
(3) N =.1.33
-1.5 . . . . ~. . . . . . . . . . . . . . . . . . . . . . . LE
:. ................................ :Z.............
where the integral extends over the allowed region
and Re
-2.5 68.5
keff(r) = ~/N 2k 2 --~2/r2 (4) E
........... Im
.=
is the effective radial wave number within the well. 2~ - 3 . 5 64.5 ~.,

This is analogous to a Bohr-Sommerfeld quantiza-

tion condition for "bound" states of light. -4.5 ... E -~ 60.5
The imaginary part of the poles is proportional to i
..... "~......ZZ~?'n-O
the penetration factor of the centrifugal barrier,
-5.5 - - ~ ,~.s
exp[ - 2 7'(or, 2 / k ) ] 50 52 54
p
56 58 60

A/k
Fig. 2. Reggetrajectoryfamilies n=0, 1 for N= 1.33 and 50<fl
< 60, for both polarizations(E and M ). Re 2: fight scale;logmIm
a 2: left scale. The open circles mark the resonances for
58.2~<#~<59.0, at l=65, 71.

365
Volume 89, number 5,6 OPTICS COMMUNICATIONS 15 May 1992

nance lifetime, is of the order of the period of internal remainder term), a sum of residues at Regge poles.
motion within the well multiplied by the average The background amplitude, which varies smoothly
number of internal reflections before transmission, with t, has already been evaluated by CAM theory
which is inversely proportional to the barrier pene- for all ranges of the scattering angle 0 (see ref. [ 10]
tration factor (5). Similarly, Irn2~. can be inter- and references therein). It contains geometrical-op-
preted in terms of a resonance "lifeangle" [ 12 ]. tic (WKB) contributions as well as a variety of dif-
The exponentially fast decrease of In~jn with ]/re- fraction effects (Fock-type diffraction in penumbra
flects the exponential behavior of the barrier trans- regions, rainbows, glories, ...). The ripple fluctua-
missivity. The sensitivity of the resonances to barrier tions, contained in the resonance amplitude, are rep-
deformations explains ~ e sensitive dependence of resented in CAM theory by the contributions from
the cross sections on the parameters and the quasi- Regge trajectories.
chaotic appearance of the ripple fluctuations [20 ].
The radial phase integral 7' across the barrier in eq.
(5) plays the role of a Lyapunov exponent. Least 5. Total cross section
sensitive to changes, and consequently most relevant
for the usual applications, are the broader reso- Application of the optical theorem to eq. (7) for
nances near the barrier top, associated with inci- the forward scattering amplitude leads to a corre-
dence near the edge of the sphere. sponding decomposition of the extinction efficiency
Qex,-a,/na 2 , (8)
4. Background and resonance contributions where at is the total cross section. We find for the
contribution from the Regge pole 2in(t) to Qext,R
According to CAM theory [ 10,13 ], the Mie scat- (there are misprints in the corresponding result in
tering amplitudes can be written in the form ref. [5])
P 2n. [. exp(in2j,)]
Sj(]/, 0 ) = ~ Sip(t, O)+ASje(]/, 0 ) , (6) Qexta.- - f12 l m lL
z j . r j . -cos(n,~j.)
- q (9)
p=o d,
where the first sum corresponds to the Debye mul- where rj. is the residue of the S-function at 2j.(fl).
tiple-internal-reflection expansion and the last term The resonant behavior originates from the cosine
is the remainder of this expansion after P + 1 terms. in the denominator, which becomes small at a nar-
In eq. (6), Sjo represents the effect of direct reflec- row resonance (when 2j. is close to a half-integer).
tion from the surface of the sphere and Sip (p >_ 1 ) In the immediate neighborhood of the peak, eq. (9)
corresponds to transmission following ( p - 1 ) inter- yields a Breit-Wigner (lorentzian) resonance shape
nal reflections. [121.
The Regge pole contributions are entirely con- The peak contribution to Q~t is given by 2 (2l+ 1 ) /
tained in the remainder term. Indeed, Regge poles //2, in agreement with the peak (resonant) contri-
are singularities of the Debye infinite series expan- bution from a given partial wave and polarization
sion, located at the boundaries of its domain of [9]. Since l=O(fl), we see that this contribution is
convergence. O (fl- ~), so that it becomes smaller and smaller, rel-
For large enough P, the only significant contri- ative to the asymptotic limit 2 of Qcxt as fl increases.
butions to ASje arise from the residues at the Regge The background contribution to Q~xt, in a Fock-
poles that are associated with resonances, so that we type approximation, was previously given [5 ] as a
rewrite eq. (6) as sum of terms arising from geometrical direct reflec-
tion, classical diffraction, axial-ray contributions and
Sj=Sj.a + Sj, R , (7) edge diffraction (tunneling [ 17] ). A fit of Q¢xt with
where Sj,B denotes the background amplitude, a sum this background leads to residual errors having a
of contributions from the Debye expansion, and Sj.R dominant sinusoidal behavior reminiscent of for-
denotes the resonance amplitude (arising from the ward glory oscillations [ 21 ]. Indeed, one finds that

366
Volume 89, number 5,6 OPTICS COMMUNICATIONS 15 May 1992

the main contribution to this discrepancy is due to ground represents a size average of Q, xt, with ap-
a previously neglected forward-glory contribution to proximately equal areas above and below it. This
the background, associated with the p = 4 term of the agrees with the results found in extensive numerical
Debye expansion. tests [22 ] that the average of Q~xt over a siie param-
The p = 4 forward glory (for N = 1.33) contains eter interval Aft~ 7r is very closely approximated by
both real glory-ray contributions, corresponding to the background, so that the ripple fluctuations have
paths of the type shown in fig. 3a, with an angle of approximately zero size average.
incidence 01 ~ 49.5 °, and complex-ray contributions The explanation is provided by a remarkable fea-
due to tunneling, of the type shown in fig. 3b, where ture of the Regge trajectory contributions, appar-
( ~ 30 ° is the angular gap bridged by surface waves, ently unnoticed so far. In contrast with ordinary
the sum of arcs that may be described at each of the Breit-Wigner resonances in individual partial waves,
vertices [ 14 ]. This surface-wave term is the forward which are always positive (and which they approx-
counterpart of that conjectured by van de Hulst to imate near the resonance peaks), Regge contribu-
explain the (backward) meteorological glory [ 9 ]. tions can turn negative as a trajectory is passing be-
The resulting CAM fit to Qext for N = 1.33 and tween consecutive physical points, leading to large
58.2 ~<fl~<59.0 (an interval of about one ripple quasi- "antiresonant" drops.
period) is shown in fig. 4. There are three families This feature is illustrated in fig. 4 for the trajectory
of Regge trajectories for each polarization that con- j = 1, n = 2 , that gives rise to the broad resonance
tribute within this interval: n = 0 , l, and 2. The cor- M21: its contribution (offset by 2 and labelled as
responding resonance locations are indicated in fig. 2+M21 ) is separately plotted, and it is seen to con-
4. For an accurate evaluation of 2j, and rj, in eq. (2), tain a broad and substantial negative portion. Can-
it is essential to employ the uniform asymptotic ap- cellation effects in the size average among positive
proximations to the cylindrical functions. and negative Regge contributions arise from the
In fig. 4, we have plotted the background both phase factor exp(ilrReAj,) in eq. (9), explaining the
omitting and including the p = 4 forward glory con- near-coincidence between the size-averaged cross
tribution. The oscillatory character of this contri- section and the background [ 22 ]. Note also that the
bution, computed here by summation of the asso- contributions from the broadest resonances, M21 and
ciated Debye partial-wave series, is clearly seen. Real E21, which are mainly responsible for this effect here,
forward glory rays (that first become possible for are merged together.
p = 4 ) and surface waves give comparable contri- We see that CAM theory leads to a very good fit
butions in this case. of the total cross section, including the ripple. The
It is immediately apparent in the figure that Qext typical accuracy found is of the order of 0.05% to
oscillates around the background, peaking above it 0.075%, becoming worse (by a factor of order 2) near
but also dropping well below it, so that the back- the broad n = 2 resonances. Without the p = 4 term,

(a) (b)

Fig. 3. Forwardglorypaths for the p--4 Debyeterm from (a) real rays; (b) complexrays (surfacewaves). For N- 1.33,one has 01~ 49.5°
and the total surface-waveangle (sum of those describedat each vertex) is ~---30°.

367
Volume 89, number 5,6 OPTICS COMMUNICATIONS 15 May 1992

2.15 i J i i i i ,_ i related to the theory of the meteorological glory [23],

-- Q ,t(Mie) - - - Backgd.- (p=4) is far more difficult to evaluate, because o f the large
M16s ........ Q,xt(CAM) 2 + M2e~ )rl number o f Debye terms that give appreciable
t . . . . Backgd. contributions.
2.1
Here we have retained only the leading Debye

0°2.05 i terms within this range: p = 0 (direct reflection ), p = 2

(van de Hulst surface-wave contribution and axial-
ray one) and p = 11 (10th-order rainbow shadow
contribution). These contributions were evaluated
by numerical summation o f the corresponding par-
tial-wave series.
The resulting CAM approximation, within the
same range as before, is compared to the exact Mie
1.95 result in fig. 5. Although the ripple fluctuations are
58.2 58.4 58.6 58.8 59
p much larger than those found for Qoxt, with the gain
varying by over two orders of magnitude across this
Fig. 4. Comparison between the exact Mie extinction efficiency
range, the quality o f the fit, while not comparable
and the CAM approximation, for N= 1.33, 58.2 ~<fl~<59.0. Res-
onance positions are indicated. The background contribution with that o f fig. 4, remains fairly good, with typical
( both excludingand including the p = 4 term) is shown. The con- errors o f about 20%. The relative error increases to
tribution from the (M, n = 2) Reggetrajectory, offset by 2, is also about twice as much in ranges where G is very small.
plotted (curve labelled 2 +M~ ): portions below ordinate 2 rep- It should be noted that the backward gain arises
resent negative contributions.
almost entirely from complex paths, i.e., from tun-
neling (the geometrical-optic contribution is negli-
typical errors would have been several times larger. gible). Since several significant contributions to the
The accuracy o f the fit is consistent with that o f the background [23 ] have not been included in the pres-
asymptotic approximations (to order fl- 2 ) employed.
For the background, the results can be improved
12~j . . . . . . . . . . ~ ........... 1.4
by resorting to the uniform CAM approximation , N= 1.3
[ 17 ]. For the broader resonances, which seem to be G(~.=) (Mie) / ." : -'-" I%,,~ef EO71 1.2
responsible for most o f the error, corrections of or- 10 ........ G(~.n)(CAM)f "
der (Im2j,) 2, not taken into account in the evalua- I
tion of eq. (9), would have to be included. 8 Miss}
: ISB/Sal
0.8 _
t (1)
(/)

6. Backscattering cross section :;t O,6 ~--

:it

~
0.4
While ripple fluctuations are o f relative order fl- E, !,. ~', ....>~< ......
in Qext, they become o f order unity and play a very
4201L__~'/1[ 0.2

large role in backscattering. The backscattering gain

factor .... --j,, ;, ~; :- ~ 0
58.2 58.4 58.6 58.8 59
p
G(fl, n) - Ig(/~,n)12
is defined [ 9 ] as the ratio o f the backscattering cross Fig. 5. Comparison between the exact Mie backward gain and the
section to its limiting value a2/4 for an ideal iso- CAM approximation, N= 1.33, 58.2 ~<fl~<59.0; the correspond-
tropic scatterer. ing results for the background term alone and for the Reggeterm
alone are also plotted (left scale). The ratio of background to
The Regge backscattering amplitude Sj, R(fl, ~t) is Reggebackward scattering amplitudes is also shown (fight scale).
given [ 16 ] by a sum of contributions similar to eq. The background is approximated by the sum of the Debye terms
(9). The background amplitude Sj.s (fl, n), which is p=0,2 and 11.

368
Volume 89, number 5,6 OPTICS COMMUNICATIONS 15 May 1992

ent approximation, as well as corrections to the b r o a d References

resonances, this result is very reasonable. N o other
non-partial-wave fit o f the backscattering cross sec- [ 1] M. Kerker, The scattering of light and other electromagnetic
tion with c o m p a r a b l e accuracy seems to be available. radiation (Academic, New York, 1969).
Resonances m a y a p p e a r to be the only significant [2] P.W. Barber and R.K. Chang, eds., Optical effects associated
contributions to backscattering [ 15 ]. However, fig. with small particles (World Scientific, Singapore, 1988 ).
5 shows that this is not so: a plot o f the resonant [3] A. Ashkin, Science 210 (1980) 1081.
(Regge) contribution alone gives a much worse fit, [4] S.C. Ching, H.M. Lai and K. Young, J. Opt. Soc. Am. B 4
even at resonance peaks, with errors o f several (1987) 1995, 2004.
[ 5] H.M. Nussenzveig, in: Coherence and quantum optics VI,
h u n d r e d percent; the same applies to a plot o f the
eds. J.H. Eberly, L. Mandel and E. Wolf (Plenum, New York,
b a c k g r o u n d by itself. The net gain G = IgB + gR 12 1990) p. 821; Comments At. Mol. Phys. 23 (1989) 175.
contains strong interference effects, b o t h destructive [ 6 ] S. John, Physics Today 44 ( 1991 ) 32.
a n d constructive, between b a c k g r o u n d (gB) a n d [ 7 ] A.J. Campillo, J.D. Eversole and H.-B. Lin, Phys. Rev. Lett.
Regge (gR) terms; the ratio ISB/SRI, also plotted in 67 (1991) 437.
fig. 5, is ~>0.5 over most o f the interval. [8] M. Born and E. Wolf, Principles of optics (Pergamon,
The C A M t r e a t m e n t applies just as well in the London, 1959).
presence o f material losses ( I m N > 0). It also allows [ 9 ] H.C. van de Hulst, Light scattering by small particles (Wiley,
New York, 1957).
us to obtain accurate a p p r o x i m a t i o n s to the struc-
[10]H.M. Nussenzveig, Diffraction effects in semiclassical
ture o f the resonant field within the sphere [ 16 ].
scattering (Cambridge University Press, Cambridge, 1992).
[ 11 ] P.J. Debye, Ann. Phys. (Lpz.) 30 ( 1909 ) 57.
[12]H.M. Nussenzveig, Causality and dispersion relations
7. Conclusion (Academic, New York, 1972).
[ 13] N. Fiedler-Ferrari,H.M. Nussenzveig and W. J. Wiseombe,
We conclude that C A M theory provides, for the Phys. Rev. A 43 ( 1991 ) 1005.
first time, a physical e x p l a n a t i o n as well as an ac- [ 14] H.M. Nussenzveig, J. Math. Phys. 10 (1969) 82.
curate ( n o n - p a r t i a l - w a v e ) fit o f the full Mie cross [ 15] J.R. Probert-Jones, J. Opt. Soc. Am. A 1 (1984) 822.
sections, including the ripple fluctuations. [ 16] L.G. Guimar~es, Ph.D. thesis (Brazilian Center for Physics
Research, 1991 );
L.G. Guimar~es and H.M. Nussenzveig, to be published.
[ 17] H.M. Nussenzveig, J. Phys. A 21 (1988) 81.
Acknowledgments [18] S.C. Hill and R.E. Benner, J. Opt. Soc. Am. B 3 (1986)
1509.
This work was s u p p o r t e d by the Brazilian agencies [ 19 ] J.M. Blatt and V.F. Weisskopf, Theoretical nuclear physics
CAPES, C N P q a n d FINEP, a n d by the French agency (Wiley, New York, 1952).
C N R S during the stay o f one o f us ( H M N ) at the [ 20 ] S.T. Shipley and J.A. Weinman, J. Opt. Soc. Am. 68 (1978)
Ecole N o r m a l e S u p r r i e u r e in Paris; the hospitality o f 130.
C. C o h e n - T a n n o u d j i a n d S. H a r o c h e is gratefully ac- [21]H.M. Nussenzveig and W.J. Wiscombe, Optics Left. 5
(1980) 455.
knowledged. We t h a n k W.J. W i s c o m b e for the use o f
[22 ] H.M. Nussenzveig and W.J. Wiscombe, Phys. Rev. Lett. 45
several o f his programs. (1980) 1490.
[23] V. Khare and H.M. Nussenzveig, Phys. Rev. Lett. 38 (1977)
1279.

369