Exact Vectorial Model For Nonparaxial Focusing by Arbitrary Axisymmetric Surfaces

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Research Article Vol. 33, No.

5 / May 2016 / Journal of the Optical Society of America A 801

Exact vectorial model for nonparaxial focusing


by arbitrary axisymmetric surfaces
DENIS PANNETON,* GUILLAUME ST-ONGE, MICHEL PICHÉ, AND SIMON THIBAULT
Centre d’optique, photonique et laser, Université Laval, Québec City, Québec G1V 0A6, Canada
*Corresponding author: denis.panneton.1@ulaval.ca

Received 23 December 2015; revised 7 March 2016; accepted 7 March 2016; posted 9 March 2016 (Doc. ID 256227); published 5 April 2016

We present a new approach, based on Richards–Wolf formalism, to rigorously model nonparaxial focusing of ra-
dially and azimuthally polarized electromagnetic beams by axisymmetric systems without a single-point focus. Our
approach is based on a combined method that uses ray tracing and diffraction integrals. Our method is validated by
comparing known results obtained with a parabolic mirror. Our integral representation of the focused beams,
compliant with diffraction theory, is thoroughly discussed and solved for various conics that, so far, have not been
treated analytically. The extension of the method to other polarization states is straightforward. © 2016 Optical
Society of America

OCIS codes: (140.3295) Laser beam characterization; (140.3300) Laser beam shaping; (050.1960) Diffraction theory.

http://dx.doi.org/10.1364/JOSAA.33.000801

1. INTRODUCTION The goal of this paper is to adapt RWT to distributed-focus


Rigorous modeling of nonparaxial beams is required in many systems and to provide researchers with a more flexible tool for
topical areas, such as superresolution microscopy [1,2], optical highly nonparaxial beam manipulation. The method is first va-
trapping [3], or electron acceleration [4]. In general, papers lidated with a known element, the parabolic mirror. We will
have addressed specific beam shapes with prescribed polariza- then consider the spherical, ellipsoid, and hyperboloid mirrors.
tion states that are critical for given applications [5–8].
Focusing systems that are considered as paraxial can usually
2. OVERVIEW OF THE RICHARDS–WOLF
be described by a scalar approach. However, as the numerical
FORMALISM
aperture of a focusing system increases, so does the necessity of
applying a vectorial treatment to the electromagnetic fields in The core of this paper is to provide a general and insightful
the focal volume. A well-known approach to this problem is the approach to focusing by complex systems. The present section
Richards–Wolf theory (RWT), which is described in the first will help the reader understand the basis of the method and
part of this paper. In its current form, the RWT considers how the new treatment is introduced within the well-known
single-point focus systems upon which a collimated beam (flat Richards–Wolf theory.
phase) of arbitrary polarization and field profile is incident. To
A. Richards–Wolf Integrals
the best of our knowledge, there is, to this date, no general
treatment of the focusing of vectorial optical fields under non- Basically, the Richards–Wolf integral representation of focused
paraxial conditions by systems that do not respect the condition beams considers a collimated beam incident on a perfectly
of single-point focus. The usual approach to complex systems focusing system [12]. According to these assumptions, one
involves the introduction of phase aberrations [9]. While this can represent the focusing system as parallel rays incident on
approach can generally be legitimate and sufficient to obtain a reference sphere, which are then deflected toward the focal
physical insight of the so-produced beams, a semi-analytic un- point (see Fig. 1). The rays are incident at a height h, and de-
approximated method would be, if developed with rigor and flected at angles α and β (respectively, the polar and azimuthal
generality, a useful tool for high-precision calculations (as angles in a spherical coordinate system).
needed in superresolution microscopy or particle acceleration, In this representation, a plane wave is converted into a per-
for instance). Recent works also used the combined method fect spherical wave by the focusing system. When considering
based on ray tracing computability and diffraction integrals the vectorial and oscillatory nature of light, the exact shape of
for wave propagation [10]. The first example of the general the electromagnetic field near the focus can be calculated in a
approach we are developing was recently published [11] as a manner self-consistent with Maxwell’s equations. Near the
specific solution to spherical mirrors. focus of this system, the electric field can be expressed as

1084-7529/16/050801-10 Journal © 2016 Optical Society of America


802 Vol. 33, No. 5 / May 2016 / Journal of the Optical Society of America A Research Article

Fig. 2. Phase-term dependency on z for each plane wave constitut-


Fig. 1. Geometry of the classical Richards–Wolf formalism. A col- ing the wavefront. The phase ϕ is associated with the physical
limated ray is incident at a height h and focused with angles α and β distance δs.
with respect to a spherical coordinate system centered at the focal
point. The electromagnetic field can then be represented with cylin-
drical coordinates r, z, ϕ.
the relative phase of a given plane wave at a distance z from the
focus (as seen in Fig. 2).
ZZ As previously mentioned, the other orthogonal axisymmet-
ric polarization state is the azimuthally polarized incident beam.
Ẽ⃗r   âα; βAα; β expi k⃗ · r⃗ dΩ; (1)
Ω By using the duality of electromagnetic fields, one can switch
the definitions in Eqs. (2) and (3) for such a beam by substi-
where d Ω is an element of the solid angle upon which the in- tuting E⃗ → η0 H⃗ and H⃗ → E∕η ⃗ 0 . Any other axisymmetric
tegration is performed. This integral can be interpreted as the polarization state can be obtained by a linear combination of
local combination (at a position r⃗ ) of various plane waves of those two basic states; thus, a linear combination of their field
amplitude Aα; β, polarized along directions α̂ and β̂ with a distribution solution. It should be noted that, for any other
⃗ Equation (1) fully respects Maxwell’s equations,
wave vector k. non-axisymmetric polarization state (e.g., linear, circular, or
according to the superposition theorem, as long as the incident elliptical), the integrals cannot be reduced to Eqs. (2) and (3),
plane wave spectrum considered is also physically valid. and a Cartesian decomposition needs to be done [15].

B. Axisymmetric Systems C. Limitations of the Richards–Wolf Theory


Because we consider axisymmetric systems in this paper, some It can be seen from Fig. 1 that Richards–Wolf integrals actually
simplifications and assumptions can be used in order to obtain assume a common focal point for all incident rays. This “ideal”
better insight into the physical solutions of the RWT. representation can be seen as a perfect spherical wavefront con-
First, when considering an axisymmetric system, the integral verging toward the focal point. If the system introduces a
over β can be automatically computed. Two states of polariza- deformation to the reflected/refracted wavefront in such a
tion form a base for the axisymmetric condition: radial and way that the system has no more than a single-point focus,
azimuthal. For the following section, we will consider an in- the Richards–Wolf integrals can no longer be used as such.
coming beam âα; β with radial polarization. Under RWT, A common way to handle this phenomenon is to add aberra-
this situation leads to a well-known form of the integrals in tion in the phase term. This approach considers a correction in
Eq. (1) [13,14]: the form of Seidel coefficients [16–18]; it still considers a per-
  Z   fectly focusing system with a single focal point but with phase
Ez αmax sin αJ 0 kr sin α modifications at local angles of the focused wavefront.
∝ qαl 0 α sin α
Er αmin i cos αJ 1 kr sin α Physically, one can consider that the introduction of an
aberration is actually a scalar correction within a vectorial
× expiϕα; zdα; (2)
theory [10,16,17]. Thus, this approach works for highly non-
and, through Maxwell’s equations, paraxial systems (because the core of the theory is vectorial), but
Z α only with small deviation from the ideal spherical wavefront.
max Figure 3 illustrates why the aberration treatment ignores any
Hϕ ∝ qαl 0 αi sin αJ 1 kr sin α × expiϕα; zdα
αmin vectorial effect.
(3) Furthermore, the aberration approach does not consider the
influence of the surface on the apodization factor and requires a
with series of aberration terms to adequately describe any given sur-
face. Computations can be made with Eq. (1) by applying a
ϕα; z ≡ −kz cos α: (4) correction to the phase such as
l 0 α is the amplitude profile of the incident beam, qα is the ϕα; z → ϕα; z − kpα; β; (5)
apodization factor, and ϕα; z is the relative phase of each
plane wave in the angular spectrum of the focused beam. where phase corrections for known aberrations are listed in
The phase term −ikz cos α can be geometrically interpreted as Table 1.
Research Article Vol. 33, No. 5 / May 2016 / Journal of the Optical Society of America A 803

A B

Fig. 3. Fundamental difference between aberration and exact


vectorial treatments. A) Exact treatment (ERWT), as considered
and implemented into RWT in this paper. B) Aberration treatment.
One can see that the aberration treatment considers a perfectly focus-
ing system with a correction in the optical path length (OPL), thus
Fig. 4. Geometry used within the formalism for any given focusing
adjusting the relative phase between rays. In a scalar approximation,
surface. Considering the symmetry, only a 2D slice in the fx; zg plane
both methods are alike, but, in a nonparaxial, vectorial approach, the
is shown without loss of generality, where x and z are the coordinates
aberration does not consider any effect on the polarization of individ-
of said surface.
ual waves at the recombination. Mathematically, the phase term in
Eq. (1) is not adequately computed to consider the relative angle
between focused waves if approach B) is used.
To totally describe the surface within our formalism, one
needs to translate the surface coordinates into angular coordi-
nates. We define a function Rθ, the distance between the sur-
Table 1. Aberration Factorsa
face and a point C, for each angle θ. This function greatly
Primary Aberrations Phase Factor pα;β depends on the chosen point C, making it mandatory to wisely
Distortion C 1;1 sin α cos β choose C according to the global system geometry. With this
Field curvature C 2;0 sin2 α definition, one can obtain Rθ by the following substitutions:
Coma C 3;1 sin3 α cos β
z → d − Rθ cos θ (6)
Astigmatism C 2;2 sin2 α cos2β
Spherical aberration C 4;0 sin4 α and
a
Aberration factors can be used in non-axisymmetric geometries. x → Rθ sin θ: (7)
To each angle, θ can be associated with a local focal point
3. EXTENDED RICHARDS–WOLF THEORY f loc and a declination angle α, which corresponds to the reflec-
tion of a ray of light initially parallel to the optical axis. We
This paper hereby proposes an approach to the computation of specify the condition that there is only one height h of the in-
fields produced by axisymmetric focusing systems without the cident ray associated with a given value of α and vice versa. This
need to define a single-point focus. The idea behind our gen- can be rigorously described by ensuring that d 2 x∕d z 2 does not
eralization is that a full vectorial treatment can be implemented change sign over the considered surface. If this condition were
within the RWT for any focusing system without approxima- not satisfied, there would be a degeneracy of plane waves in the
tion because each infinitesimal portion of a complex focusing focusing spectrum. The integral in Eq. (2) should then be sep-
surface can be described as an equivalent surface with a single arately computed for each part of the system where the afore-
local focus. By considering the vectorial influence of the dis- mentioned condition is satisfied.
placement of the local focus and with an appropriate geomet- We will normalize the function Rθ by introducing the
rical formalism, we introduce a novel formalism, which we call function ρθ  Rθ∕d so that ρ0  1.
the extended Richards–Wolf theory (ERWT), self-compliant Geometrically, one can find the declination angle of every
with Maxwell’s equations for any given axisymmetric surface, plane wave, according to the surface definition:
with no requirement of a single-point focus. The ERWT is  −1 
an extension, not a correction, of the RWT. It gives the oppor- dx
α  2 arctan : (8)
tunity to treat an extended range of focusing systems whose dz
solutions were unavailable with the classical RWT formalism. With the definition of ρθ, one can show that
Both formalisms lead to identical results for single-focus
systems. dx ρ 0 sin θ  ρ cos θ
 0 ; (9)
The following sections consider a reflective surface, making d z −ρ cos θ  ρ sin θ
it possible to create extremely nonparaxial beams (α > π∕2). where ρ 0  d ρ∕d θ. We also introduce a dimensionless param-
The principles are still valid for diopters, for instance, and gen- eter ζ (normalized with d ) describing the position of any f loc
eralized formulas for refraction-based focusing systems are with respect to C:
under investigation.
ζ  −ρcos θ − sin θ cot α: (10)
A. Definition of the Reflective Surface
Let us consider a general axisymmetric surface and an incident B. Generalization of the Phase Term
beam (see Fig. 4). To describe the surface, Cartesian coordinates We then describe the extended phase term Φe α that will be
are used, with the origin at the apex of the reflecting surface. used within the integral representation. If we place the origin of
804 Vol. 33, No. 5 / May 2016 / Journal of the Optical Society of America A Research Article

the cylindrical coordinate system at C, any distance to the new where d Ω is an element of solid angle, P being the plane wave
origin will be given by z 0 ≡ z − d. One can write density. To obtain the expression in Eq. (15), we assumed perfect
Φe α  −kz 0 cos α  pe α: (11) axisymmetry and, thus, removed the azimuthal dimension from
the expression of the solid angle. If we now consider the energy
The factor pe α is associated with the path of a ray between within an angular portion of the incident beam, we can write
the phase-zero plane (collimated beam plane that passes on C)
until its conjugated plane wave attains the origin (see Fig. 5). ϵα  2πjE⃗ 1 j2 hd h; (16)
One can find that, geometrically, the following analytical phase where E 1 defines the incident field and d h represents an element
function is valid for all eligible cases: of height h, an infinitesimal annular portion of the incident

sin θ beam (see Fig. 4).
Φe  −k z 0 cos α  R cos θ  R This expression only stands for collimated incident beams.
sin α
 We used hα as the height, with respect to the optical axis, of
 Rcos θ − sin θ cot α cos α : (12) any given ray with postfocusing declination angle α. For every
plane wave within an angular element, we can consider that the
Introducing dimensionless parameters ζ, ρ, γ ≡ kd , and individual element of electromagnetic intensity I pw is actually
Z ≡ z 0 ∕d , the phase term finally can be expressed as ϵα
  I pw  (17)
sin θ Pαd α
Φe  −γ Z cos α  ρ cos θ  ρ − ζ cos α : (13)
sin α
It is important to see that the phase is a function of θ only, jE⃗ 1 j2 hd h∕d α
because α is also a function of θ, thus making it possible to  ; (18)
κ sin α
integrate over α or θ, depending on which one is more suitable
where we introduced a normalizing constant κ with surface
for the geometry under consideration.
units. The intensity of a given element of plane waves after
C. Generalization of the Apodization Factor focusing is related to the incident intensity producing these
plane waves. The corresponding factor between those two
The generalization of RWT, as given by Eq. (1), requires a
quantities is, by definition, the apodization factor. One can
proper formulation of the apodization factor, which balances
then write that
the density of plane waves between the collimated incident
beam and the beam reflected by the axisymmetric system. jE⃗ 2 j2  jE⃗ 1 j2 qα2 ; (19)
This factor is classically defined as a projection of the plane
where E 2 defines the focused field.
over a reference sphere. For a generalized mirror, the variable
This leads to the general expression
focal spot makes this definition inappropriate. The approach sffiffiffiffiffiffiffiffiffiffiffiffiffiffi
hereby proposed is based upon plane wave density. The wave- h dd αh
front produced after focusing can be, as previously mentioned, qα  : (20)
κ sin α
decomposed into plane waves. From this approach, one can
compute the plane wave density comprised within an annular When the system is known, qα can be determined from
portion of the aberrated wavefront. Because one can say that, the geometrical relation between the angle α and the incident
without consideration of the incident field, the number of plane height h. For any typical surface with a flat apex (plane surface
waves within an angular portion is proportional to the solid angle locally perpendicular to the optical axis at the center of the ax-
exactly representing this angular portion, it is possible to write isymmetric system), the normalization is ensured by κ, which
dΩ can be determined by
Pα ∝ (14) lim qα2  1;
dα α→0
(21)

knowing that for a retro-reflected beam α  0, the incident


∝ sin α; (15) and reflected fields must be the same.
To express the apodization factor within the new formalism,
one only needs to use the relation h  Rθ sin θ and substi-
tute it into Eq. (20). This yields
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
R sin θR 0 sin θ  R cos θ dd αθ
qe  (22)
κ sin α

vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi


u
uρ sin θρ 0 sin θ  ρ cos θ d α −1
t dθ
 : (23)
⃗ The κ 0 sin α
Fig. 5. Phase correction for the plane wave of wave vector k.
total length of the solid arrow, for each plane wave, needs to be calcu- κ 0 includes all normalization factors and can still be determined
lated to correctly see the relative path between plane waves (pe α). by imposing
Research Article Vol. 33, No. 5 / May 2016 / Journal of the Optical Society of America A 805

lim q  1: (24) dα 2ϵ cos θ  2


α;θ→0  : (33)
d θ ϵ2  2ϵ cos θ  1
D. Extended Richards–Wolf Integrals
Equation (32) can be inverted in order to have an implicit
For axisymmetric systems and radially polarized incident expression of all quantities involved in the integration over α:
beams, the field is computed by Eq. (2) with the following
modifications: 0
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi


1
cot 2 − tan 2 1 − ϵ   1 cot α2
α 2 α 2
ϕα; z → Φe α; z (25) B C
θα  2 × arctan@ A:
and ϵ−1
qα → q e α: (26)
(34)
After an adequate definition of all parameters according to
the geometry under study, one can solve the integral equation All these expressions are consistent with the known results in
numerically with an adapted quadrature for fast-oscillating in- the limits ϵ  0 (sphere [11]) and ϵ  1 (parabola [13]).
tegrands (e.g., Levin [19]). Gauss–Kronrod quadratures also To test the ERWT, we produced three incident profiles with
proved to have an adequate accuracy for those calculations [20]. radial polarization. Such beams can only be physically valid if
For a better understanding of the numerical results, one can they have zero intensity on the optical axis. The profiles used
also look for an asymptotic analytical solution to the integral. were of the form prescribed in Table 2.
For that approach to be valid, the incident beam must not have The three profiles used are shown in Fig. 6. They were nor-
strong field amplitude near the extrema of the integral (i.e., near malized to unity because the computation is used to validate the
α  0 or α  π). The asymptotic solution could be based profile of the solution field, which will also be normalized later.
upon the stationary-phase method, as described in [11]. It is always possible to normalize the profiles according to their
integral over α  f0; πg. These normalizations have no effect
4. FOCUSING WITH CONIC SECTIONS on the normalized profiles of this section but need to be kept in
mind while doing the computation for real cases.
We have applied the ERWT formalism to nonparaxial focusing
For each case, we computed the intensity profile I in the
by axisymmetric conic sections.
focal region using
Conic sections can usually be described by
p
r ; (27)
1 − ϵ cos θ Table 2. Different Incident Profiles Used for the
where r is the distance from the geometrical focus to the surface Computations Using the ERWTa
and p is the semilatus rectum, defined as the radius of the
Profile ID Profile l 0 α Parameters [rad]
surface in the geometrical focal plane.
When considering the geometry used within this paper, we A α exp−α2 ∕Δα2  Δα  0.1
choose the point C as the geometrical focal point so the expres- B α exp−α2 ∕Δα2  Δα  0.5
C exp−α − α0 2 ∕Δα2  Δα  0.1α0  fπ∕4; π∕2g
sion in Eq. (27) can be rearranged as
a
For Profile C, α0  π∕2 for all cases computed in Section 4, except for the
1ϵ
ρθ  (28) hyperbola. More on this can be found in Section 4.D.
1  ϵ cos θ
and
dρ ϵ sin θ
 ρ2 ; (29)
dθ 1ϵ
where ϵ is the eccentricity. With Eqs. (8) and (9), we find the
relation between α and θ:
 0 
−ρ cos θ  ρ sin θ
α  2 arctan (30)
ρ 0 sin θ  ρ cos θ

 
−ϵ sin θ cos θ  sin θ1  ϵ cos θ
 2 arctan : (31)
ϵ sin2 θ  cos θ1  ϵ cos θ
After simplifications, the angle of every plane wave α for any
given observation angle θ is given by
 
sin θ
αθ  2 arctan ; (32) Fig. 6. Incident profiles used for the computations. The formulas
ϵ  cos θ
and parameters used can be retrieved in Table 2. For profile C, the
and its derivative, which is needed to compute the apodization solid line represents α0  π∕2, while the dotted one represents the
factor, is given by case where α0  π∕4.
806 Vol. 33, No. 5 / May 2016 / Journal of the Optical Society of America A Research Article

I  jE z j2  jE r j2 : (35) This expression is what can be retrieved in the literature


Some linear cuts will also be plotted to see the distribution [13,21] from the standard RW theory. Illustrations of focusing
of intensity along certain major axes (i.e., along the optical axis with a parabola are found in Fig. 7, for the incident profiles
r  0 and in the paraxial focal plane z  0). Finally, it has to previously prescribed.
be noted that the scaling factor will be γ  105 for every case The integrals of RWT are robust and fit for paraxial (inci-
and that the calculations are made in a grid centered on dent profile A), intermediate (incident profile B), and highly
z  ζjαopt , where αopt is the angle α at which the incident nonparaxial beams (incident profile C). ERWT predicts the
same exact behavior for the extreme case of perfect focusing
distribution is maximum. This means that the calculation is
described by Grosjean and Courjon [14].
centered on the expected “main” focus.
We can see the narrowing of the profiles with more non-
A. Parabola paraxial beams. The incident profile C (with α0  π∕2) leads
It is interesting when ensuring that the generalized model to the formation of a needle of light with the narrowest profile
coincides with the classical RW model for the limit case of a accessible, as described in [13,14].
perfect parabola. B. Sphere
For a parabola, ϵ  1. This yields the interesting result
The simplest spread-focus surface treatable is the spherical
α  θ. This actually implies that our system can be centered
mirror.
on the optical focus C (as prescribed by the previously used
This surface can be described as a conic section of eccentric-
conic section formula). According to the Richards–Wolf theory,
ity ϵ  0. The generalized radius is ρθ  1. From this, we
the phase term should be exactly −kz cos α. Equation (13) ac-
obtain α  2θ. We find the generalized phase term:
tually becomes 
  sinα∕2
sin α Φ  −γ Z cos α  cosα∕2 
Φe  −γ Z cos α  ρ cos α  ρ − χ cos α (36) sin α
sin α   
cos α
 cosα∕2 − sinα∕2 cos α : (43)
 −γZ cos α  ρcos α  1 (37) sin α
After some algebraic manipulations, one obtains
Φ  −γZ cos α  2 cosα∕2: (44)
 −γZ cos α  2: (38)
0
As for the apodization factor, with ρ  0, we obtain
Because the phase offset can be chosen arbitrarily, this phase rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
term is exactly the one expected with a parabolic mirror (up to a sinα∕2 cosα∕2
qα  (45)
constant phase offset 2γ). This supplemental phase is due to the 2κ 0 sin α
new origin from which the phase is calculated. In RWT, one
considers that all plane waves are in phase at the focus and the
relative phase between angular contributions is calculated from
this “zero-phase” plane. In ERWT, the “origin” from which the
phase is calculated is located in a plane prior to reflection to
adequately compare the phase difference in the plane wave
spectrum. For a parabola, the accumulation of phase from that
plane to the focus is the same for all plane waves, thus produc-
ing a simple offset.
For the apodization factor, because dd αθ  1, we can find qα
from the generalized definition:
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ρρ 0 sin α  ρ cos α
qe α  (39)
κ0
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2 2 ffi
0−1∕2 ρ sin α
κ ρ  ρ cos α (40)
2

2
 κ 0−1∕2 : (41)
1  cos α
As previously mentioned, we use the limit where q0  1
to fix the value of κ 0 . This yields
 
α Fig. 7. Calculated fields and profiles for different beams (A, B, and
qe α  sec2 : (42)
2 C with α0  π∕2) incident on a parabola.
Research Article Vol. 33, No. 5 / May 2016 / Journal of the Optical Society of America A 807

rffiffiffiffiffiffiffi
1
 (46)
4κ 0

 1: (47)
Those results are consistent with a recent paper [11] in
which a spherical mirror was used to produce an extended nee-
dle of light, from a radially polarized annulus of light. ERWT,
thus, gives us a tool for further analysis of focusing systems.
If we compare the phase term in Eq. (44) with the aberration
factor for spherical aberration in Table 1, we can see a major
difference between phase relations (the normalized phase rela-
tions are plotted in Fig. 8). This difference does influence the
solution at the focus; thus, when using the correct apodization
factor, an asymmetry develops that is not considered with the
aberration approach. The treatment also becomes independent
of the actual width of the incident beam. Aberration treatment
includes an “aberration factor” that is dependent of the effective
aperture of the system. The aberration treatment can be con-
sidered as effective only in a span of ≈0.4 rad near the opti-
cal axis.
The unitary apodization factor is usually associated with the
Herschel condition [21]. This condition describes a system in
which the object and image planes are related by a perfectly Fig. 9. Calculated fields and profiles for different incident beams
stigmatic relation because all incident energy is distributed with (A, B, and C with α0  π∕2) on a sphere.
the exact same configuration postfocusing. While this interest-
ing condition is not achievable with aplanetic elements usually
considered in nonparaxial focusing systems, the spherical mir-
ror naturally respects it. The spherical mirror is commonly used profile with the increase of Δα, as expected by the geometric ray
in the paraxial domain but rarely under nonparaxial conditions. tracing upon a sphere.
We computed the intensity profile generated for the sphere For the annular profile C, we can retrieve the result found in
by the incident angular fields given earlier. The resulting inten- [11] where a needle of light is produced. It is also interesting to
sities in the focal area are shown in Fig. 9. see the similarity between the intensity profile from the inci-
The calculation of these intensity distributions are not pos- dent profile A upon a parabola (preceding section) and the
sible with the standard RWT. With ERWT, they can be readily sphere, with a small asymmetry. These results show the simi-
computed, and it becomes easy to see the stretching of the axial larity between those two surfaces at small angles but still pin-
point the importance of ERWT for high-precision calculations.
The needle of light could not have been correctly computed
with the aberration treatment. The aberration factor is also
physically invalid and needs to be adapted to coincide better
with the real solution. This is a result of the extremely high
numerical aperture (NA) and inappropriate apodization factor.
C. Ellipse
ERWT was also tested for other values of eccentricity. An el-
lipses, whose eccentricity resides between 0 and 1, should pro-
duce fields near the focus somewhere between those of the
parabola and the sphere. For general cases, the relation between
α and θ cannot be simplified, and the integrals need to be com-
puted with the explicit forms of Eqs. (28)–(32). Results of such
calculations for an ellipse of eccentricity ϵ  0.5 are given for
the three different incident fields used in the previous cases.
Fig. 8. Normalized aberration factor (black line) versus normalized Computed intensity distributions are shown in Fig. 10.
ERWT phase factor (gray line). The zero-phase was set to be at the
same place because the definition of the two factors have a noncon-
The computed fields, again, could not have been totally de-
tributing constant offset. It can be seen that the third-order aberration scribed with the aberration approach. The resulting intensity
factor, in its classical definition, is valid only in the paraxial limit. To profiles produced with the ellipse are actually somewhere be-
make this comparison correct, a negative spherical aberration needs to tween equivalent results obtained with a parabola and a sphere.
be used, which is the case for a concave spherical mirror with the The apparent stretching of the needle (incident profile C) is
geometry considered. larger for the ellipse than the sphere. This could be explained
808 Vol. 33, No. 5 / May 2016 / Journal of the Optical Society of America A Research Article

Fig. 11. Apodization factor for different values of eccentricities. For


a hyperbola, there is an asymptotic limit on α for which the apodiza-
tion factor diverges. This limit is shown as the dashed vertical line for
the case at ϵ  1.25.

Fig. 10. Calculated fields and profiles for different incident beams Near the asymptotic limit, the integral becomes ill defined
(A, B, and C with α0  π∕2) on an ellipse. and needs to be reinterpreted. Extremely nonparaxial beams
produced by a hyperbola are described by a function αθ that
tends toward a constant value as θ grows, thus making the in-
tegration over α impractical. This situation can be interpreted
as the results with an axiconic mirror (for instance, [24]). There
by the angular definition of the incident beam because similar is actually no ray over α ≈ 1.85 rad for the hyperbola with
angular profiles are not actually associated with similar spatial ϵ  1.25 (see Fig. 11).
profiles of incident fields.
Interesting calculations have been published for elliptic mir-
rors, considering an illumination focused on one of the ellipse’s
two focal spots [22,23]. Ellipses’ properties make it possible, in
this particular case, to use classical RWT because the focusing
from the mirror converts a perfect spherical wavefront into an-
other spherical wavefront.
D. Hyperbola
Another conic surface is the hyperbola, which is defined by an
eccentricity of ϵ > 1. The computations in this paper are based
on the case where ϵ  1.25. Hyperbolas, though, have an ab-
solute limit for the angle α that is physically valid. The asymp-
totic behavior of the surface makes it so that, for a given ϵ, any
ray coming from an observation angle θ will eventually tend
toward a fixed value of α. This can be seen by tracing the apod-
ization factor for different values of ϵ (see Fig. 11). This graph
shows a value for which qe α diverges, meaning that there is a
reflection angle to which an infinite number of rays can be as-
sociated (because every ray coming from a height sufficiently
large will be reflected at that same angle α).
For this reason, we used an alternative definition of the
incident profile C (with α0  π∕4) to provide a physically
valid situation. The results for the hyperbola are shown
in Fig. 12.
As expected, we can see a similar pattern for the incident
profile B as the one obtained with the ellipse, but with a focus Fig. 12. Calculated fields and profiles for different incident beams
stretching on the other side of the paraxial focus. (A, B and C with α0  π∕4) on a hyperbola.
Research Article Vol. 33, No. 5 / May 2016 / Journal of the Optical Society of America A 809

Our choice of α0  π∕4 does not completely convert the 8. Compute the integrals given by Eq. (25) to obtain the
initial radial polarization into an axial polarization. The dashed electromagnetic solution field for the system under study.
profile in Fig. 6(C) thus produces a profile combining Bessel Further generalizations (e.g., pulsed regimes) and different
beams of 0th and 1 st orders. Profiles A and B clearly give sim- formalisms are still under study to extend the possibilities and
ilar results to those of the ellipse (Fig. 10) but reverted along the the robustness of such tools for computing solutions on non-
z axis. paraxial complex problems. The ERWT hereby proposed could
Another interesting phenomenon is the spreading of the also be further developed by removing the inherent necessity of
focal spot with imperfect systems (sphere, ellipse, and hyper- incident collimation.
bola) with the widening of the incident beam while the oppo-
site is observed with a perfect system (parabola). This can be Funding. Fonds de Recherche du Québec-Nature et
used to engineer the illumination in the focal region for various Technologies (FRQNT) (165091); Natural Sciences and
cases of focusing [11,13]. Engineering Research Council of Canada (NSERC)
E. Alternate Formulation (1173114).
In cases where an alternative form of the integrals (over θ)
would be preferable (e.g., highly nonparaxial hyperbola), the REFERENCES
integrals can be written as
  Z 1. G. Thériault, Y. De Koninck, and N. McCarthy, “Extended depth of field
Ez θmax microscopy for rapid volumetric two-photon imaging,” Opt. Express
∝ q e θl 0 θ sin αθ 21, 10095–10104 (2013).
Er θmin
2. F. O. Fahrbach, V. Gurchenkov, K. Alessandri, P. Nassoy, and A.
 
sin αθJ 0 kr sin αθ Rohrbach, “Light-sheet microscopy in thick media using scanned
× expiΦe θ; zdθ: Bessel beams and two-photon fluorescence excitation,” Opt.
i cos αθJ 1 kr sin αθ Express 21, 13824–13839 (2013).
3. Q. Zhan, “Trapping metallic Rayleigh particles with radial polarization,”
(48) Opt. Express 12, 3377–3382 (2004).
It also has to be noted that the angular distribution for 4. C. Varin, M. Piché, and M. A. Porras, “Acceleration of electrons from
rest to GeV energies by ultrashort transverse magnetic laser pulses in
the incident field (i.e., l 0 α or l 0 θ) is actually dependent
free space,” Phys. Rev. E 71, 1–10 (2005).
on the geometry of the system. For better comparison between 5. R. Boivin and A. Boivin, “Optimized amplitude filtering for superreso-
two focusing surfaces, it would be beneficial to define the lution over a restricted field. I: Achievement of maximum central irra-
incident field with Cartesian, prefocusing parameters diance under an energy constraint,” Opt. Acta Int. J. Opt. 27, 587–610
(e.g., l 0 x). (2010).
6. K. Jahn and N. Bokor, “Solving the inverse problem of high numerical
aperture focusing using vector Slepian harmonics and vector Slepian
5. CONCLUSION multipole fields,” Opt. Commun. 288, 13–16 (2013).
7. T. Liu, J. Tan, J. Liu, and H. Wang, “Modulation of a super-Gaussian
We presented an extension of the Richards–Wolf formalism, optical needle with high-NA Fresnel zone plate,” Opt. Lett. 38, 2742–
which is suitable for focusing surfaces without the restriction 2745 (2013).
of single-point focus. We were able to develop a methodology 8. R. Kant, “An analytical solution of vector diffraction for focusing optical
to adequately describe the electromagnetic field near the focus of systems,” J. Mod. Opt. 40, 337–347 (2007).
9. A. April, P. Bilodeau, and M. Piché, “Focusing a TM(01) beam with
any given axisymmetric focusing system, based upon a rigorous a slightly tilted parabolic mirror,” Opt. Express 19, 9201–9212
vectorial treatment. Three distributed-focus systems have been (2011).
studied for the first time, to the best of our knowledge, within an 10. B. Andreas, G. Mana, and C. Palmisano, “Vectorial ray-based diffrac-
analytical and exact formalism: the sphere, ellipse, and hyper- tion integral,” J. Opt. Soc. Am. A 32, 1403–1424 (2015).
11. D. Panneton, M. Piché, S. Thibault, and G. St-Onge, “Needles of
bola. The extension of the theory has proven to be consistent
light produced with a spherical mirror,” Opt. Lett. 40, 419–422
with the basic theory, as the parabolic system coincides in both (2015).
cases. 12. B. Richards and E. Wolf, “Electromagnetic diffraction in optical sys-
For the computation of the focal field, one needs to proceed tems. II: Structure of the image field in an aplanatic system,” Proc.
as follows: R. Soc. Lond. Ser. A 253, 358–379 (1959).
13. H. Dehez, A. April, and M. Piché, “Needles of longitudinally polarized
1. Define the focusing surface with a function Rθ light: guidelines for minimum spot size and tunable axial extent,” Opt.
(according to the geometry of Fig. 4). Express 20, 14891 (2012).
14. T. Grosjean and D. Courjon, “Smallest focal spots,” Opt. Commun.
2. Determine the functions ρθ and ρ 0 θ.
272, 314–319 (2007).
3. With those functions, find the relation between α and θ 15. L. Novotny and B. Hecht, Principles of Nano-Optics (Cambridge
with the Eqs. (8) and (9). University, 2006).
4. Determine the phase-term function with Eq. (13) and 16. R. Kant, “Vector diffraction in paraboloidal mirrors with Seidel aberra-
the relation θα found previously. tions. I: Spherical aberration, curvature of field aberration and distor-
5. Determine the extended apodization factor with tion,” Opt. Commun. 128, 292–306 (1996).
Eq. (23). 17. R. Kant, “An analytical method of vector diffraction for focusing optical
systems with Seidel aberrations II: astigmatism and coma,” J. Mod.
6. Determine the incident angular definition l 0 α. Opt. 42, 299–320 (1995).
7. Determine the portion of z over which the computation 18. C. Sheppard, “Vector diffraction in paraboloidal mirrors with Seidel
needs to be done with Eq. (10), evaluated at the values of α aberrations: effects of small object displacements,” Opt. Commun.
included in the distribution l 0 α. 138, 262–264 (1997).
810 Vol. 33, No. 5 / May 2016 / Journal of the Optical Society of America A Research Article

19. D. Levin, “Procedures for computing one-and two-dimensional inte- 22. J. Liu, M. Ai, J. Tan, R. Wang, and X. Tan, “Focusing of cylindrical-
grals of functions with rapid irregular oscillations,” Math. Comput. vector beams in elliptical mirror based system with high numerical
38, 531–538 (1982). aperture,” Opt. Commun. 305, 71–75 (2013).
20. S. Ehrich, “Stieltjes polynomials and the error of {G}auss-{K}ronrod 23. J. Liu, M. Ai, H. Zhang, C. Wang, and J. Tan, “Focusing of an
quadrature formulas,” Appl. Comput. Orthogonal Polynomials 131, elliptical mirror based system with aberrations,” J. Opt. 15, 1–7
57–77 (1999). (2013).
21. C. Sheppard and K. Larkin, “Optimal concentration of electromagnetic 24. M. Zhu, Q. Cao, and H. Gao, “Creation of a 50,000 λ long needle-like
radiation,” J. Mod. Opt. 41, 1495–1505 (1994). field with 0.36 λ width,” J. Opt. Soc. Am. A 31, 500–504 (2014).

You might also like