1488 J. Opt. Soc. Am. B / Vol. 25, No. 9 / September 2008 Yan et al.

Decoupling and asymmetric coupling in triple-

core photonic crystal fibers

Yan Yan,1,* Jean Toulouse,1 Iavor Velchev,2 and Slava V. Rotkin1

1
Department of Physics, Lehigh University, 16 Memorial Drive E., Bethlehem, Pennsylvania 18015, USA
2
Laser Center, Fox Chase Cancer Center, 432 Rhawn Street, Philadelphia, Pennsylvania 19111, USA
*Corresponding author: [email protected]

Received May 20, 2008; revised July 8, 2008; accepted July 9, 2008;
posted July 21, 2008 (Doc. ID 96420); published August 19, 2008
We systematically investigate the intercore coupling in triple-core photonic crystal fibers (PCFs). Coupled-
mode equations are developed and solved analytically in the linear approximation. We derive the eigenmodes
in triple-core PCFs, in particular we model the decoupling mode and test its stability against perturbations.
The coupling coefficients and the coupling length are determined experimentally and found to agree extremely
well with the calculated results. We describe the propagation of light in triple-core PCFs for different launch-
ing conditions. © 2008 Optical Society of America
OCIS codes: 060.0060, 060.5295.

1. INTRODUCTION sented. Second, we establish the coupled-mode equations

In recent years, photonic crystal fibers (PCFs) have at- that describe the propagation of light in triple-core PCFs.
tracted substantial interest [1–4]. The PCFs have been We then study propagation of light for different launching
shown to support endlessly single-mode guidance [5], tun- conditions, in particular launching into the central core.
able chromatic dispersion properties [6,7], strong confine- From our simulations and calculations, we also determine
ment [8], and high nonlinearities [9,10]. Multicore PCFs, the coupling length of the triple-core PCF under test. Ex-
because of their unique properties, offer additional advan- perimental results are then reported and shown to be in
tages and provide new opportunities for original applica- excellent agreement with theoretical predictions.
tions [11–14] such as in-fiber frequency shifters and direc-
tional couplers. Multicore fiber lasers [15] or multicore
fiber amplifiers [16,17] that use evanescent field-coupled 2. PHOTONIC CRYSTAL FIBER
arrays are also potential applications. So far, most of the The PCF used in our theoretical analysis and experi-
research on multicore PCFs has focused on theory and ments consisted of three solid cores with an index of re-
computer simulations [18–20]. The few experimental re- fraction of n = 1.46, embedded in a triangular lattice of air-
sults that have been reported are mainly on dual-core holes with a diameter of d = 1.08 ␮m, as sketched in Fig.
PCFs. Blanchard et al. proposed a two-dimensional bend 1(a). The three central missing airholes provided the PCF
sensor with a single microstructured fiber [21]. Fogli et al. with the necessary defect to guide light. The pitch of the
analyzed dual-core PCF couplers with full vectorial beam periodic structure (the hole spacing) was ⌳ = 2.52 ␮m. The
propagation methods [22]. The properties of dual-core distance between the centers of the cores was 3⌳. A mi-
PCF were experimentally characterized by Mangan et al. croscope image of the actual cross section of the fiber used
[23]. Directional coupling in a dual-core PCF was realized in the experiments is shown in Fig. 1(b), which shows
by Kakarantzas et al. using heat treatment [24]. Yet we that the fiber was well fabricated and cleaved. The central
anticipate that triple-core PCFs, because of their unique part, including the core area and the periodic microstruc-
structure, possess distinct characteristics and open new ture zone, did not reveal any obvious sign of deformation.
possibilities beyond those of dual-core fibers. In triple- Nevertheless, it is important to note that even small
core PCFs more eigenmodes are possible and therefore variations in the shape of microcores can affect the over-
more complex light patterns in propagation, which offers lap between optical modes that propagate in neighboring
new opportunities for versatile manipulation and control cores and therefore the coupling between these cores,
of guided light, e.g., switching on the basis of polarization. even though the coupling constant is barely affected by
Nonlinear switching with a pump beam also combines in- such imperfections [25].
tercore coupling with nonlinear aspects in PCFs. In sum-
mary, the triple-core fibers represent the simplest ex-
ample of multicore fibers, allowing us to demonstrate 3. THEORY
specific coupling effects present only in fibers with more In multicore fibers, the neighboring cores must be suffi-
than two cores. ciently close for the optical modes that propagate in the
Here we systematically investigate the coupling behav- adjacent core to overlap the separating region. Such eva-
ior in triple-core PCFs by means of both theory and ex- nescent wave coupling between optical modes associated
periments. First, the particular PCF of interest is pre- with an adjacent core can lead to the transfer of power

0740-3224/08/091488-8/$15.00 © 2008 Optical Society of America

Yan et al. Vol. 25, No. 9 / September 2008 / J. Opt. Soc. Am. B 1489

= ␲ / 共2冑␬12 + ␬22兲, where we have assumed for generality

that coupling coefficients ␬1 and ␬2 between neighboring
cores might not be equal. Starting from z = 0, all the power
is transferred from the central core to the two side cores
over a distance Lc or vice versa.
Since we are considering a low-power cw beam
launched into the triple-core fiber, the time-dependent
Fig. 1. (Color online) (a) Fiber geometry; cores are marked with terms can be set to zero. Therefore, Eq. (1) can be simpli-
large circles for illustration purpose only. (b) A microscope image fied as
of the cross section of the fiber used in the experiments. (c) The
cross section profile. ␬1 and ␬2 are the coupling coefficients be- dA1 dA2 dA3
tween neighboring cores. = i ␬ 1A 2, = i ␬ 1A 1 + i ␬ 2A 3, = i␬2A2 . 共2兲
dz dz dz
from one core to the another under suitable conditions.
Starting with trial solutions
Since the optical modes tend to be localized within each
Ai共z兲 = Ci1ei冑␬1+␬2·z + Ci2e−i冑␬1+␬2·z + Ci3 ,
core, only nearest neighbors communicate with each other 2 2 2 2

[26]. Neglecting direct coupling between cores 1 and 3, the

coupled-mode equations in the linear approximation for where i = 1 , 2 , 3, substituting them into Eqs. (2) and rear-
triple-core PCFs can therefore be written as [27] ranging the terms, we obtain
⳵A1 1 ⳵A1 i ␤ 2 ⳵ 2A 1
A1共z兲 = ␬1C+ei冑␬1+␬2·z + ␬1C−e−i冑␬1+␬2·z + ␬2C0 ,
2 2 2 2
+ + = i ␬ 1A 2 ,
⳵z vg ⳵t 2 ⳵t 2

A2共z兲 = 冑␬12 + ␬22C+ei冑␬1+␬2·z − 冑␬12 + ␬22C−e−i冑␬1+␬2·z ,

2 2 2 2
⳵A2 1 ⳵A2 i ␤ 2 ⳵ 2A 2
+ + = i ␬ 1A 1 + i ␬ 2A 3 ,
⳵z vg ⳵t 2 ⳵t2
A3共z兲 = ␬2C+ei冑␬1+␬2·z + ␬2C−e−i冑␬1+␬2·z − ␬1C0 .
2 2 2 2
共3兲
⳵A3 1 ⳵A3 i ␤ 2 ⳵ 2A 3
+ + = i ␬ 2A 2 , 共1兲 By putting Ai共z = 0兲 = Ai0, we determine the three indepen-
⳵z vg ⳵t 2 ⳵t2 dent constants of the problem C0, C+, and C− in terms of
A10, A20, A30, the initial amplitudes in cores 1, 2, and 3,
where A1, A2, and A3 denote the amplitudes in cores 1, 2,
respectively,
and 3, respectively, at position z along the fiber; vg is the
group velocity; ␤2 is the group velocity dispersion (GVD); ␬2 ␬1
␬1 and ␬2 are the coupling coefficients between cores 1 C0 = A10 − A30 ,
and 2 and cores 2 and 3, respectively, as illustrated in Fig. ␬12 + ␬22 ␬12 + ␬22
1(c). For a perfect fiber, ␬1 and ␬2 should be equal; how-
ever, we found that was not the case in our fibers. 1 A20
The coupling length, which is the main characteristic C+ = 共␬1A10 + ␬2A30兲 + ,
parameter of the coupling system, is defined as the short-
2共␬12 + ␬22兲 2冑␬12 + ␬22
est distance over which the signal fully transfers from one
waveguide to an adjacent one. For triple-core fibers, if the 1 A20
C− = 共␬1A10 + ␬2A30兲 − .
2冑␬12 + ␬22
beam is initially launched into only the central core, it
2共␬12 + ␬22兲
will fully transfer to both side cores as it propagates over
a coupling length along the fiber. For a pair of coupled 共4兲
waveguides, the coupling length is simply given by L
= ␲ / 2␬, where ␬ designates the coupling coefficient be- To determine the eigenmodes for the triple-core fiber, we
tween the two waveguides. For the triple-core PCF used rewrite Eqs. (3) in matrix form T ជ . The matrix T
J • Aជ = A J is
0
in our experiments, the coupling length is found to be Lc thus found to be

冤 冥
␬12 cos共冑␬12 + ␬22 · z兲 + ␬22 i␬1 ␬ 1␬ 2
sin共冑␬12 + ␬22 · z兲 关cos共冑␬12 + ␬22 · z兲 − 1兴
␬12 + ␬22 冑 ␬12 + ␬22 ␬12 + ␬22
i␬1 i␬2
J=
T sin共冑␬12 + ␬22 · z兲 cos共冑␬12 + ␬22 · z兲 sin共冑␬12 + ␬22 · z兲 . 共5兲
冑 ␬12 + ␬22 冑␬12 + ␬22
␬ 1␬ 2 i␬2 ␬22 cos共冑␬12 + ␬22 · z兲 + ␬12
关cos共冑␬12 + ␬22 · z兲 − 1兴 · sin共冑␬12 + ␬22 · z兲
␬12 + ␬22 冑␬12 + ␬22 ␬12 + ␬22
1490 J. Opt. Soc. Am. B / Vol. 25, No. 9 / September 2008 Yan et al.

Fig. 2. Initial amplitudes in three cores of the eigenmodes. ␬1 and ␬2 are equal.

It is worth noting here that the factor i in the off-diagonal to 关1 / 2共Aជ +A ជ 兲+A ជ 兴 and launching into core 3 corre-
s a d
terms of the matrix indicates that the fiber introduces a ជ
sponds to 关1 / 2共As + A ជ 兲−A ជ 兴.
a d
relative phase shift between the coherent beams that
The amplitude distributions that correspond to these
propagate in the three cores, which can be applied to the
three eigenmodes at z = 0 are shown in Fig. 2, where ␬1
design of fiber interferometers. Solving the determinant
and ␬2 are set equal for illustration purpose. By defini-
of this 3 ⫻ 3 matrix, we obtain three eigenvalues of the
tion, light launched according to one of these eigenmodes
冑 2 2 冑 2 2
system: ␭ = 1, ei ␬1+␬2z, e−i ␬1+␬2z. The corresponding eigen- will propagate undisturbed with a constant ratio of the in-
vectors are then easily found to have the form tensities in the three cores, i.e., 兩A ជ 共z兲兩2 = const. We note
i
ជ in Eq. (6), neither the

冢 冣冢 冣冢 冣
␬2 ␬1 ␬1 that, for the decoupling mode A d

0 , 冑
␬12 + ␬22 , − 冑␬12 + ␬22 .
amplitudes nor the phases of the beams in the three cores
change with distance. Decoupling here means that no en-
− ␬1 ␬2 ␬2 ergy transfer occurs between cores. When no beam is
launched into the central core, and the ratio of the ampli-
After normalization, we obtain tudes in the two side cores is equal to that of the coupling

冢 冣
␬2 coefficients but out of phase, these amplitudes will remain
A0 constant as they propagate along the fiber, with no light
ជ ⬅
A 0 , 共6兲
d
冑␬12 + ␬22 − ␬1
being transferred to the central core, as shown in the first
graph in Fig. 2. Figure 3(a) shows the amplitudes in the
three cores along the fiber when launching A ជ by itself.
d

冢 冣
␬1
A0 ei冑␬1+␬2z
2 2 ␬1 / ␬2 is set to 3 / 4 for illustration purpose.
ជ ⬅
A · 冑 ␬12 + ␬22 , 共7兲 In practice, there can be slight variations along the fi-
s
冑2 冑␬12 + ␬22 ber due to imperfections. Hence we have studied the sta-
␬2 bility of the decoupling mode against perturbations. Sup-
posing that the ratio of the amplitudes in the two side

冢 冣
␬1
A0 e−i冑␬1+␬2z
2 2 cores suffers a minor change due, for example, to unequal
ជ ⬅
A · − 冑␬12 + ␬22 , 共8兲 fiber loss, we can write the launching condition as A1
a
冑2 冑␬12 + ␬22 = ␬2A0 − ⌬A, A2 = 0, A3 = −␬1A0, which is equivalent to
␬2 launching into the decoupling mode except for a small ad-
where 兩A0兩2 = P0 is the total initial power launched into the ditional amplitude ⌬A in core 1. Using Eqs. (3), we then
ជ ,Aជ ,A ជ are the eigenmodes, or so-called su- calculate the amplitudes in the three cores after a propa-
fiber and A d s a gation length z:
permodes, of the propagation. They are labeled as decou-
pling, symmetric, and antisymmetric, respectively. An ar- ␬12⌬A ␬22⌬A
bitrary optical amplitude distribution between the three A1共z兲 = ␬2A0 − cos共冑␬12 + ␬22 · z兲 − ,
␬12 + ␬22 ␬12 + ␬22
cores can then be written as a superposition of these three
eigenmodes. For example, launching light into the central
i␬1⌬A
sin共冑␬12 + ␬22 · z兲,
core would correspond only to a superposition of two
A2共z兲 = −
eigenmodes: 共A ជ −A
s
ជ 兲, launching into core 1 corresponds
a 冑␬12 + ␬22

Fig. 3. (Color online) (a) Amplitudes in the three cores for the decoupling mode. (b) Amplitudes in the three cores for the decoupling
mode with small perturbations.
Yan et al. Vol. 25, No. 9 / September 2008 / J. Opt. Soc. Am. B 1491

4. SIMULATION
To study the guided propagation of light with coupling in
multicore PCFs, we developed an efficient and versatile
numerical program [28]. Although many methods have
been employed, such as the modal field expansion method
[29,30], multipole method [31], and beam propagation
method [32,33], our model utilizes a fully three-
dimensional (3-D) numerical method to solve Maxwell’s
equations in vector form. The code is based on applying
the dispersion and propagation operators separately for
each single propagation step in a finite-difference scheme.
We use the fast Fourier transform (FFT) technique for the
calculation of diffraction effects in kជ space. The calcula-
tions are run on a rectangular mesh with a square cross
section. A super-Gaussian instead of a Gaussian model is
used to eliminate the light reflected back from the bound-
aries of the fiber structure. This 3-D code allows for the
determination of coupling coefficients for various fiber ge-
ometries. The high precision of our method is confirmed
by the consistency of our results with values reported in
Fig. 4. (Color online) Decoupling versus coupling. the literature [22].

␬1␬2⌬A ␬1␬2⌬A
A3共z兲 = − ␬1A0 − cos共冑␬12 + ␬22 · z兲 + .
␬12 + ␬22 ␬12 + ␬22
A. Decoupling Mode
共9兲 We first tested the theoretical results on the decoupling
mode by modeling two antiphase beams of equal intensity,
The propagated amplitudes are plotted in Fig. 3(b). This
simultaneously incident into the two side cores, as in de-
result shows that the decoupling mode is stable against
coupling mode A ជ . A symmetric fiber, in which the two
small perturbations. We have also checked the perturba- d
tion that is due to small changes in the coupling coeffi- coupling coefficients were equal, was used for this simu-
cients and have found that they affect the decoupling lation. The results are shown in Fig. 4. After 10 cm, the
mode in a similar way. It is useful to note that, for the intensity distribution indeed remains unchanged. We
other two eigenmodes, symmetric and antisymmetric at compared this with the result obtained for two in-phase
any propagation length z, we always have beams in the same launching conditions (not an eigen-
mode). After 10 cm we see observed that 59% of the en-
兩A1共z兲兩
兩A3共z兲兩
= 冑 P1
P3
=
␬1
␬2
, 共10兲
ergy that propagates initially in the side cores was trans-
ferred to the central core.

where P1 and P3 are the intensities in core 1 and core 3,

respectively. Equation (10) shows that the ratio of the two
coupling coefficients is equal to the square root of the ra- B. Propagation in Guided Cores
tio of the output powers in the two side cores. Note here In the experiments, to measure the coupling lengths and
that both the symmetric and the antisymmetric modes describe the propagation, we found it simple and suffi-
maintain constant intensities in all three cores, whereas cient to launch light into the central core, i.e., core 2. In
the phases change as they propagate along the fiber. this case, the normalized initial amplitudes are

Fig. 5. (Color online) Amplitudes in the three cores versus the propagation length, with the initial light in core 2. (a) ␬1 / ␬2 = 1 (b) and
␬1 / ␬2 = 3 / 4
1492 J. Opt. Soc. Am. B / Vol. 25, No. 9 / September 2008 Yan et al.

tudes. In the simulation, we modeled the propagation of

light launched into the central core. The first complete en-
ergy transfer from the central core to the two side cores is
found at z = 672.54 nm, which is therefore the coupling
length. The result is shown in Fig. 6. Similarly, we ob-
tained the propagation equations for launching with one
side core, core 1 or core 3. The amplitudes in the three
cores along the propagation length are plotted in Fig. 7:
␬1 / ␬2 is set to 3 / 4 for illustration purpose.

C. Offset Sensitivity
Fig. 6. (Color online) At z = 672.54 mm, the light in the central To test the sensitivity of the latter experiment to the
core is entirely transferred to the two side cores. launching conditions, we performed a simulation in which
the light was launched again into the central core, but

冢 冣 冢冣
A10 0 with a 1 ␮m offset from the center of the core. As expected
from theory, this resulted in higher optical modes being
A20 = 1 , excited, although these were found to die out within 5 cm
A30 0 as the beam propagated in the fiber. The results are
shown in Fig. 8.
and the coefficients are
A20
C0 = 0, C+ = − C− = . 5. EXPERIMENTS
2冑␬12 + ␬22
We now present the experimental technique followed by
Equation (3) there becomes the experimental results from which we determined the
coupling coefficients and the coupling length. The cou-
i␬1A20
A1共z兲 = sin共冑␬12 + ␬22 · z兲, pling length is estimated using an empirical formula as
冑␬12 + ␬22 well.

A2共z兲 = A20 cos共冑␬12 + ␬22 · z兲,

A. Experimental Setup
To probe the linear coupling characteristics of the fiber,
we launched light from a helium–neon laser into a single
i␬2A20
sin共冑␬12 + ␬22 · z兲.
core. To do this, the beam was first expanded through a
A3共z兲 = 共11兲
冑␬12 + ␬22 Galilean expansion system, and the divergence and spot
size of the beam matched to the diameter and numerical
As can be seen from Eqs. (11), the amplitudes in the two aperture (NA) of the microcore, using an apochromatic
side cores vary sinusoidally while the amplitude in the aberration and flat field optical correction microscope ob-
central core varies like a cosine wave, as shown in Fig. 5, jective. The cross section of the fiber was illuminated
where we plotted the amplitudes in the three cores along through a glass plate and was projected onto a television
the propagation length. For illustration purpose, ␬1 / ␬2 is screen along with the focal spot of the launched beam. In
set to 1 in Fig. 5(a) and to 3 / 4 in Fig. 5(b). It is worth not- this way proper launching conditions were accurately
ing that the amplitudes in the two side cores reach their monitored. At the opposite end of the fiber, another micro-
maxima or minima at the same points along the fiber. The scope objective was placed at its focal distance from the
difference between the two coupling coefficients ␬1 and ␬2, fiber tip. A magnified image of the near field of the output
if there is any, simply results in different maximum am- was produced through a CCD camera. In this way we
plitudes in the two side cores rather than different cou- managed to record the output and measured the respec-
pling lengths. In another words, ␬1 and ␬2 do not result in tive intensities in each of the three cores. Figure 9 is a
different interaction lengths but only different ampli- schematic of the experimental setup.

Fig. 7. (Color online) Amplitudes in the three cores versus the propagation length with the initial light in one side core. Light launched
into (a) core 1 and (b) core 3.
Yan et al. Vol. 25, No. 9 / September 2008 / J. Opt. Soc. Am. B 1493

Fig. 8. (Color online) Offset dies out at 5 cm through a simula-

tion program.

For a final test that the intensity output from the three
cores would not be affected by small offsets of the
launched beam from the center of a particular core, we Fig. 9. (Color online) A schematic of the experimental setup.
launched light into the central core. The launching spot
was slightly displaced along the x and the y axis [parallel other. This asymmetric coupling phenomenon was ob-
with and perpendicular to the orientation of the three served for all sections of the triple-core fiber and for dif-
cores, respectively, as shown in Fig. 1(c)] so as to test the ferent launching conditions, indicating that the coupling
result of offsets in launching. The normalized transmitted asymmetry was intrinsic to the fiber, i.e., the coupling co-
power through each core was found to be insensitive to efficients ␬1 and ␬2 were not equal.
x-axis offsets to within 8 ␮m and y-axis offsets to within The coupling coefficients and the coupling length were
7 ␮m. This provides convincing background for later ex- determined in the following experiments. As mentioned
periments. The results are shown in Fig. 10. earlier, a simple and efficient method for obtaining the
coupling length is to launch light into the central core and
B. Experimental Results measure the relative output intensities from the three
Using the setup described above, we conducted experi- cores. Substituting the output intensities on the left-hand
ments on the light propagation in a triple-core three-pitch side of Eqs. (11) and the input intensities on the right-
PCF to study the intercore linear coupling. The lengths of hand side, we were able to calculate the coupling coeffi-
the tested fibers ranged from 7 to 65 cm. For longer fi- cients and the corresponding coupling length: ␬1
bers, although nominally isotropic, small twists, bends, = 1.8833 m−1, ␬1 = 2.1778 m−2,
and other stresses can induce unknown and uncontrolled Lc = ␲ / 共2冑␬1 + ␬2兲 = 54.56 cm.
2 2

birefringence on the fiber so that the output becomes un-

predictable. Shorter fibers gave stable results. No visible C. Estimate of the Coupling Length with an Empirical
beam coupling was observed in fibers shorter than 9 cm, Formula
which indicated longer coupling lengths. The results re- For a symmetric coupler, the following empirical formula
ported below were obtained on a 25.2 cm fiber. The output can also be used to estimate the coupling coefficient [34]:
intensity distributions for light launched successively into
each of the three cores are shown in Fig. 11. Comparison ␲V
of these intensity distributions reveals some asymmetry ␬= exp关− 共c0 + c1d̄ + c2d̄2兲兴, 共12兲
2k0n0a2
in the intercore coupling, most clearly seen in Figs. 11(a)
and 11(c): very little power was found in core 2 in Fig. where c0, c1, and c2 are constants: c0 = 5.2789− 3.663V
11(c); some light was transferred into core 2 in Fig. 11(a). + 0.3841V2, c1 = −0.7769+ 1.2252V − 0.0152V2, c2 = −0.0175
When light was launched into one of the side cores, that − 0.0064V − 0.0009V2. V is the fiber parameter, so-called
is, core 1 or core 3, the intensity distributions at the out- normalized frequency V = k0a共n12 − n22兲1/2 = 共2␲ / ␭0兲aNA in
put were not symmetric. This asymmetry persisted when
which a is the core radius and d̄ is the normalized center-
light was launched into the central core; the intensities
transferred to the two side cores also being unequal. After to-center spacing between neighboring cores 共d̄ ⬎ 2兲.
a certain propagation distance, the power in the central Equation (12) has been shown to be accurate to within 1%
core was coupled more into one side core than into the for values of V and d̄ in the range of 1.5艋 V 艋 2.5 and

Fig. 10. (Color online) Normalized transmitted power through cores versus offsets along the x- and y-axes.
1494 J. Opt. Soc. Am. B / Vol. 25, No. 9 / September 2008 Yan et al.

Fig. 11. (Color online) Basic result of the asymmetric coupling between cores. The fiber was 25.2 cm long.

2.0艋 d̄ 艋 4.5. The NA of the fiber was calculated to be under an optical microscope. Future research will focus
NA = sin ␣ ⬃ 0.20 by illuminating the entire input end of on the polarization dependence of the intercore coupling
the fiber and measuring the diverging angle ␣ of the out- and the effect of fiber nonlinearities for higher launched
coming beam. Therefore, for the triple-core three-pitch powers.
PCF under test, we find the coupling coefficients and the
coupling length, as shown in Table 1, where the values ob-
tained from the simulations and the experiments are
ACKNOWLEDGMENTS
listed as well. Considering the discretization errors in the Special thanks to Ivan Biaggio for his useful suggestions
simulations and the approximation in the empirical for- with regard to the experimental setup. We acknowledge a
mula, they are in good agreement. grant from the Optoelectronics Industry Development As-
Further confirmation of the agreement between model- sociation (OIDA) that enabled us to acquire the photonic
ing and experiments was obtained by launching light into crystal fibers used in this study. The research was made
one side core. For example, with all the other input con- possible by a grant from the Electrical Engineering Divi-
ditions fixed but light launched into core 3 and using the sion of the National Science Foundation,
propagation in Eq. (3), our simulation predicted the nor- NSFគECSគ0401269. Additional support was also provided
malized output intensity in core 3 to be 73.22% of the to- by the Center for Optical Technologies at Lehigh Univer-
tal while the experiment yielded 72.49%. sity funded by a grant from the State of Pennsylvania De-
Experiments were also conducted with triple-core four- partment of Community and Economic Development.
pitch PCFs (hole diameter of d = 1.10 ␮m, hole spacing ⌳
= 2.52 ␮m, core spacing of 4⌳). Similar asymmetric cou-
pling and even longer coupling lengths were found in REFERENCES
these fibers. 1. P. St. J. Russell, “Photonic-crystal fibers,” J. Lumin. 24,
4729–4749 (2006).
2. J. A. West, N. Venkataramam, C. M. Smith, and M. T.
Gallagher, “Photonic crystal fibers,” in 27th European
6. SUMMARY Conference on Optical Comunication (IEEE, 2001), paper
Th. A. 2.2.
We have analyzed the mechanism of linear coupling in
3. Y. Jiang, Z. Shi, B. Howley, and R. T. Chen, “Highly
triple-core PCFs and identified and calculated the eigen- dispersive photonic crystal fibers for true-time-delay
values and eigenmodes or supermodes. The amplitudes in modules of an x-band phased array antenna,” Proc. SPIE
the three cores were found to change in a periodic fashion. 5360, 253–258 (2004).
In particular, the decoupling mode was simulated and 4. J. G. Rarity, J. Fulconis, J. Duligall, W. J. Wadsworth and
P. St. J. Russell, “Photonic crystal fiber source of correlated
measured in great detail. The coupling parameters of the photon pairs,” Opt. Express 13, 534–544 (2005).
triple-core PCF were measured and found to be in good 5. J. C. Knight, T. A. Birks, P. St. J. Russell, and D. M. Atkin,
agreement with those from theory. The PCFs studied “All-silica single-mode optical fiber with photonic crystal
were all found to exhibit asymmetric coupling, possibly cladding,” Opt. Lett. 21, 1547–1549 (1996).
6. L. P. Shen, W.-P. Huang, G. X. Chen, and S. S. Jian,
due to slight variations in fiber parameters or fiber geom-
“Design and optimization of photonic crystal fibers for
etry along the length, even though no obvious structural broad-band dispersion compensation,” IEEE Photon.
variations or deformations were visible by observation Technol. Lett. 15, 540–542 (2003).
7. G. K. L. Wong, A. Y. H. Chen, S. W. Ha, R. J. Kruhlak, S. G.
Murdoch, R. Leonhardt, J. D. Harvey, and N. Joly,
Table 1. Coupling Length and the Coupling Coeffi-
“Characterization of chromatic dispersion in photonic
cients Obtained from the Estimation, the Simula- crystal fibers using scalar modulation instability,” Opt.
tion, and the Experiments Express 13, 8662–8670 (2005).
8. J. C. Knight, T. A. Birks, P. St. J. Russell, and J. P. de
Coupling Length Coupling Coefficient Sandro, “Properties of photonic crystal fiber and the
(mm) 共mm−1兲 effective index model,” J. Opt. Soc. Am. A 15, 748–752
(1998).
NA= 0.20 384.29 0.0029 9. A. I. Siahlo, L. K. Oxenlowe, K. S. Berg, A. T. Clausen, P. A.
Estimation Andersen, C. Peucheret, A. Tersigni, P. Jeppesen, K. P.
NA= 0.21 723.69 0.0015
Hansen, and J. R. Folkenberg, “A high-speed demultiplexer
Simulation 672.54 0.0017
based on a nonlinear optical loop mirror with a photonic
Experiments 545.57 0.0019, 0.0022 crystal fiber,” IEEE Photon. Technol. Lett. 15, 1147–1149
(2003).
Yan et al. Vol. 25, No. 9 / September 2008 / J. Opt. Soc. Am. B 1495

10. R. Jiang, R. Saperstein, N. Alic, M. Nezhad, C. McKinstrie, crystal fibers and couplers,” Opt. Express 10, 54–59 (2002).
J. Ford, Y. Fainman, and S. Radic, “375 THz parametric 23. B. J. Mangan, J. C. Knight, T. A. Birks, P. St. J. Russell,
translation of modulated signal from 1550 nm to visible and A. H. Greenaway, “Experimental study of dual-core
band,” in Optical Fiber Communication (Optical Society of photonic crystal fibre,” Electron. Lett. 36, 1358–1359
America, 2006), postdeadline paper 16. (2000).
11. K. Saitoh, N. J. Florous, M. Koshiba, and M. Skorobogatiy, 24. G. Kakarantzas, B. J. Mangan, T. A. Birks, J. C. Knight,
“Design of narrow band-pass filters based on the resonant- and P. St. J. Russell, “Directional coupling in a twin core
tunneling phenomenon in multicore photonic crystal photonic crystal fiber using heat treatment,” in Conference
fibers,” Opt. Express 13, 10327–10335 (2005). on Lasers and Electro-Optics (Optical Society of America,
12. W. N. MacPherson, M. J. Gander, R. McBride, J. D. C. 2001), paper JTuD2.
Jones, P. M. Blanchard, J. G. Burnett, A. H. Greenaway, B. 25. M. Wrage, P. Glas, M. Leitner, T. Sandrock, N. N. Elkin, A.
Mangan, T. A. Birks, J. C. Knight, and P. St. J. Russell, P. Napartovich, and A. G. Sukharev, “Experimental and
“Remotely addressed optical fibre curvature sensor using numerical determination of coupling constant in a
multicore photonic crystal fibre,” Opt. Commun. 193, multicore fiber,” Opt. Commun. 175, 97–102 (2000).
97–104 (2001). 26. R. Beach and L. Zapata, “Tactical laser weapons for defense
13. J. M. Dudley, G. Genty, and S. Coen, “Supercontinuum SI,” Laboratory Directed Research and Development final
generation in photonic crystal fiber,” Rev. Mod. Phys. 78, report (U.S. Department of Energy, 2002).
1135–1184 (2006). 27. G. P. Agrawal, Applications of Nonlinear Fiber Optics
14. D. M. Taylor, C. R. Bennett, T. J. Shepherd, L. F. Michaille, (Academic, 2001).
M. D. Nielsen, and H. R. Simonsen, “Demonstration of 28. I. Velchev and J. Toulouse, “Directional coupling and
multi-core photonic crystal fibre in an optical interconnect,” switching in multi-core microstructure fibers,” in
Electron. Lett. 42, 331–332 (2006). Conference on Lasers and Electro-Optics (Optical Society of
15. W. J. Wadsworth, R. M. Percival, G. Bouwmans, J. C. America, 2004), paper CTuV1.
Knight, and P. St. J. Russell, “High power air-clad photonic 29. T. M. Monro, D. J. Richardson, N. G. R. Broderick, and P. J.
crystal fibre laser,” Opt. Express 11, 48–53 (2003). Bennett, “Modeling large air fraction holey optical fibers,”
16. Y. Huo, P. K. Cheo, and G. G. King, “Fundamental mode J. Lightwave Technol. 18, 50–56 (2000).
operation of a 19-core phase-locked Yb-doped fiber 30. A. Ferrando, E. Silvestre, J. J. Miret, P. Andrés, and M. V.
amplifier,” Opt. Express 12, 6230–6239 (2004). Andrés, “Full-vector analysis of a realistic photonic crystal
17. L. Michaille, C. R. Bennett, D. M. Taylor, and T. J. fiber,” Opt. Lett. 24, 276–278 (1999).
Shepherd, “Multi-core photonic crystal fibers for high 31. B. T. Kuhlmey, T. P. White, G. Renversez, D. Maystre, L. C.
power lasers and amplifiers,” Proc. SPIE 6102, 61020W Botten, C. Martijn de Sterke, and R. C. McPhedran,
(2006). “Multipole method for microstructured optical fibers.
18. K. Saitoh, Y. Sato, and M. Koshiba, “Coupling Implementation and results,” J. Opt. Soc. Am. B 19,
characteristics of dual-core photonic crystal fiber couplers,” 2331–2340 (2002).
Opt. Express 11, 3188–3195 (2003). 32. M. A. van Eijkelenborg, A. Argyros, G. Barton, I. M.
19. J. Lægsgaard, “Directional coupling in twin-core photonic
Bassett, M. Fellew, G. Henry, N. A. Issa, M. C. J. Large, S.
bandgap fibers,” Opt. Lett. 30, 3281–3283 (2005).
Manos, W. Padden, L. Poladian, and J. Zagari, “Recent
20. S. K. Varshney, N. J. Florous, K. Saitoh, and M. Koshiba,
progress in microstructured polymer optical fibre
“The impact of elliptical deformations for optimizing the
fabrication and characterization,” Opt. Fiber Technol. 9,
performance of dual-core fluorine-doped photonic crystal
199–209 (2003).
fiber couplers,” Opt. Express 14, 1982–1995 (2006).
33. B. J. Eggleton, P. S. Westbrook, C. A. White, C. Kerbage, R.
21. P. M. Blanchard, J. G. Burnett, G. R. G. Erry, A. H.
S. Windeler, and G. L. Burdge, “Cladding-mode-resonances
Greenaway, P. Harrison, B. Mangan, J. C. Knight, P. St. J.
Russell, M. J. Gander, R. McBride, and J. D. C. Jones, in air-silica microstructure optical fibers,” J. Lightwave
“Two-dimensional bend sensing with a single multicore Technol. 18, 1084–1100 (2000).
optical fibre,” Smart Mater. Struct. 9, 132–140 (2000). 34. R. Tewari and K. Thyagarajan, “Analysis of tunable single-
22. F. Fogli, L. Saccomandi, P. Bassi, G. Bellanca, and S. Trillo, mode fiber directional couplers using simple and accurate
“Full vectorial BPM modeling of index-guiding photonic relations,” J. Lightwave Technol. 4, 386–390 (1986).