This document summarizes the optical properties of three types of fiber polarizers based on panda-type W lightguides with different lengths: 200 m, 1 m, and 50 mm. For the 200 m and 1 m lightguides, dichroism (polarization dependent loss) is higher than 30 dB due to a difference in cutoff thresholds between the fundamental x and y modes. For the 50 mm lightguide, an additional scattering layer is included near the cladding boundary to achieve dichroism higher than 15 dB. The physical mechanisms behind the dichroism in each case are also described, including the effects of microbending and macrobending losses.

1. ISSN 1063 7850, Technical Physics Letters, 2011, Vol. 37, No. 7, pp. 627–630. © Pleiades Publishing, Ltd., 2011.
Original Russian Text © A.M. Kurbatov, R.A. Kurbatov, 2011, published in Pis’ma v Zhurnal Tekhnicheskoі Fiziki, 2011, Vol. 37, No. 13, pp. 70–77.
627
Optical components transfer to the fiber base
begun many years ago [1]. Here we’ll consider fiber
polarizers on the base of lightguides Panda [1] with
refractive index (RI) W profile [2].
Earlier [3] we reported about 500 m W lightguide
Panda with dichroism. IN the present work it is
reported about dichroism in developed by us light
guides with lengths 200 m, 1 m and 50 mm, and most
probable dichroism physical mechanisms of this
dichroism are also described with brief description of
its calculation ways for the first two cases. Lightguides
were manufactured from 2003 to 2007 according to
technical requirement and technology developed by
authors of the present work in different Russia organi
zations having optical fiber manufacturing base.
On Fig. 1a W lightguide Panda cross section is
shown with germanosilicate core 1 having RI = n1, flu
orine reflective cladding 2 having RI = n2, quartz clad
ding 3 with RI = n3, and circle stress applying rods 4,
which are boron doped and induce the linear bire
fringence. Additional layer 5 takes place only in
third (50 mm) lightguide. In Table parameters of all
three lightguides are listed.
On Fig. 2a x and y modes spectral losses in 200 m
lightguide are shown. On Fig. 2b, related to 1 m light
guide, curves 1 and 2 are x and y modes losses in
straight lightguide and curves 3 and 4 are those in
coiled one with 60 mm diameter (3 turns). In each
case light with õ and ó polarization is by turn was
launched by white light source. Unfortunately ous
measurements were limited by wavelength ~1.7 μm, so
we don’t see the whole dichroism windows. In 200 m
lightguide this window is right hand to operation
region (1.55 μm), but at lower winding radius it will go
leftward at the spectrum.
Consider 200 m lightguide in more details. Having
in [3] excellent agreement with experiment our bend
ing losses models here gave us too slow growth of x
and y modes spectral losses comparing with Fig. 2a.
So we considered one more loss mechanism:
microbends. We generalized microbend loss model in
convenient straight lightguides [4] to the case of bent
lightguide with any RI profile and PML layer [5].
Bent lightguide RI profile is related to straight one
profile n0(x, y) as n2(x, y) = (x, y)(1 + x/R) [5] (R is
bending radius). Light in bent lightguide could be
described in the form of supermodes [6] having fields
ψj. One of them, ψj0 (detailed), has a form similar to
fundamental mode of straight convenient two layer
lightguide and it’s microbending loss coefficient has
the form:
(1)
n0
2
2γ k
2
Cj
2
Φ Δβj( )/ Reβj0Reβj( ).
j
∑=
Fiber polarizer based on W lightguide Panda1¶
A. M. Kurbatov and R. A. Kurbatov
Department of Center for terrestrial space infrastructure objects exploiting,
Kuznetsov Research Institute for Applied Mechanics, Moscow, 111123 Russia
e mail: akurbatov54@mail.ru
Received January 11, 2011
Abstract—Three kinds of fiber W polarizer Panda are described: with lengths 200 m, 1 m and 50 mm. In the
first two cases dichroism is higher than 30 dB, in the third case it is higher than 15 dB. The feature of 50 mm
polarizer is the scattering layer in quartz cladding. In each case a most probable physical dichroism mecha
nisms are described.
DOI: 10.1134/S106378501107011X
Fig. 1. W lightguide Panda cross section. 1 is the core, 2 is
fluorine cladding, 3 is quartz cladding, 4 is stress rods, 5 is
additional scattering layer (absent in 200 m and 1 m light
guides).1The article was translated by the authors.

2. 628
TECHNICAL PHYSICS LETTERS Vol. 37 No. 7 2011
A.M. KURBATOV, R.A. KURBATOV
Here k is vacuum wavenumber, Cj is coupling coeffi
cient of detailed and number j supermodes Cj =
〈ψj|xψj0〉/(〈ψj0|ψj0〉〈ψj|ψj〉)1/2
, 〈A|B〉 = dyA*(x,
y)B(x, y) (* is complex conjugate), Δβj = Re(βj0 – βj)
is propagation constants difference (synchronism) of
detailed and number j supermodes. In Gaussian
microbends model with correlation length we have [4]
(2)
where σ is root mean square of inverse microbends
radius. We applied (1) and (2) to detailed supermodes
with õ and ó polarization (further x and y modes) of
200 m lightguide. Calculations with Lc ~1.5 mm gave
us good resemblance with Fig. 2a graphs.
Physically bend and microbends co operated work
could look like the following. As the wavelength grows
x and y modes synchronism with the rest supermodes
gets better (Δβj decreases). It (and also Lñ) sets the
growth abruptness of microbending spectral loss curve
(2), i.e. dichroism window width. As for its position,
due to x and y modes fields bending distortion [7]
coefficients Cj of their coupling with the rest super
modes are increase and loss curves get leftward at the
spectrum. Thus the bend here may regulate dichroism
window position and microbends may regulate its
width.
Let’s turn now to 1 m lightguide. In [8] W light
guide is described with dichroism window in visible
spectrum range (with relative width 5%) which is
determined by cutoff thresholds of fundamental x
and y modes. Due to bend this window goes leftward
at the spectrum and gets narrower. In [9] a W light
guide with 13% dichroism window is described in the
region 0.85 μm. Here the desired x mode cutoff
threshold is infinite so the dichroism window should
be limited only from below. However as the wavelength
grows õ mode penetrates into quartz cladding touch
ing the coating, its spectral losses grow and the dichro
ism window is also limited from above.
So, we’ve got W lightguide Panda which on the
length ~1 m may give dichroism ~30 dB and more due
to x and y modes cutoff thresholds difference [8]. For
fundamental mode cutoff normalized frequency Vcut
dx
∫
Φ Δβj( ) 2π
1/2
σ
2
Lc ΔβjLc/2( )
2
–[ ],exp=
when fluorine cladding is not very thin (our case)
we’ve got an approximation Vcut ≈ 0.333 + 1.859Λ1/2 +
0.078Λ – 5.035 × 10–4
Λ2
, where Λ ≡ ( – )/( –
). Hereof, assuming that birefringence takes place
only in the core and fluorine cladding we’ve got
dichroism window position in straight lightguide.
Bending gets it narrower basically due to x mode
losses (Fig. 2b) which are well described by our bend
ing losses models [3]. However, birefringence in this
lightguide is not large enough for such kind of applica
tions (see Table), and as it is objectively enough to
increase it up to ~0.001 then one may get substantial
lightguide characteristics improvement with the same
RI profile.
Unfortunately our bending loss models are ade
quate only if these losses are large before the cutoff
threshold, which is probably due to applied PML
layer model imperfections. On Fig. 2b y mode bend
ing losses are not large even after cutoff threshold.
However one may calculate y mode losses in straight
lightguide using other methods [6, 10] and together
with x mode bending losses accept it as worst variant
of dichroism window.
Thus, the bend is almost not replaces dichroism
window of our lightguide (contrary to [8]). To our
point of view for not thin enough fluorine claddings it
could be explained assuming that fundamental mode
bending losses are due to its bending coupling to radi
ation modes [11]. If fo fundamental mode we roughly
have Vcut > 2.4–2.6 then it is packed tightly enough in
the core and has a weak coupling to radiation modes,
i.e. low bending losses even in cutoff regime. Other
wise situation is reverse. In our lightguide one may
assume that Vcut ≈ 2.8 for y mode is large enough and
Vcut ≈ 2.2 for x mode is small enough.
The imperfection of obtained lightguide is the
necessity to coil it without axial twist. To our point of
view it is due to the following. When modeling the
bending losses we saw that they essentially depend on
stress rods orientation at the bending plane because
they have reduced RI. When bending with twist this
orientation angle changes continuously so the turns
have long enough sections with the worst orientation.
In favor of this explanation speaks the fact that bend
ing losses could be substantially reduced coiling this
lightguide without twist. One way to overcome this
problem is again birefringence increasing.
Bending dichroism window reducing leads to idea
of short polarizers (~50 mm). However the experi
ments with such W lightguides Panda sections, where
at the length 1 m dichroism was ~30 dB, gave us
dichroism ~1–3 dB. So for y mode suppression we
applied the scattering layer near the cladding boundary
with the air [12, 13] (Fig. 1). Figure 3 shows the spec
tral losses graphs in such W lightguide having length
50 mm (range 1.15 μm) both ends of which are spliced
with single mode PM lightguides Panda. It is seen
n3
2
n2
2
n1
2
n3
2
Polarizing W lightguides Panda parameters
Parameter
200 m
lightguide
1 m
lightguide
50 mm
lightguide
Core RI 1.465 1.4626 1.462
Fluorine cladding RI 1.451 1.4567 1.4566
Linear birefringence 4.7 × 10–4
7.5 × 10–4
7 × 10–4
Core diameter, µm 9.0 9.5 8
Fluorine cladding
diameter, µm
27 22.8 19.2
Fiber diameter, µm 95 125 125

3. TECHNICAL PHYSICS LETTERS Vol. 37 No. 7 2011
FIBER POLARIZER BASED ON W LIGHTGUIDE PANDA 629
−59.9
−71.9
−77.9
−83.9
−89.9
−65.9
1450 1550 20.00 1650nm/D
nm
2
1
3
(a)
3.0 2.000 500dB/D RES: AVG:
SENS: SMPL:NORMAL 501 (AUTO)
dBm
REF
−100.8
−90.8
−80.8
−70.8
−60.8
−50.8
5.0 dB/D RES:
SENS:MID
2.000 10AVG:
SMPL: 1001 (AUTO)
15001300 40.00 nm/D 1700
nm
2−
4−
1:
2:
3:
4:
1:
3:
1550.0000 nm
1300.0000 nm
−66.92 dBm
−71.80 dBm
−250.000 nm
−4.88 dBm
REF
2
4
2
3
3
1
(b)
dBm
Fig. 2. (a) Fundamental x and y modes spectral losses in lightguide having length 200 m (curves 1 and 2), and light source spec
trum (3). Vertical axis scale factor (power level) is 3 dB, horizontal axis scale factor (wavelength) is 20 nm. (b) Fundamental x
and y modes spectral losses in straight lightguide with the length 1 m (curves 1 and 2) and in the coiled one with diameter 60 mm
(curves 3 and 4). Vertical axis scale factor (power level) is 5 dB, horizontal axis scale factor (wavelength) is 40 nm.
−60.0
−85.0
−110.0
930 30.0 1080 1230in Vacnm/div
nm
5.0 dB/div
dBm
REF
TMkr (Peak)
1120.2 nm
−69.83 dBm
Normal (A & B)
: A : B
2
1
Fig. 3. Fundamental x and y modes spectral losses in lightguide having length 50 mm (curves 1 and 2). Vertical axis scale factor
(power level) is 5 dB, horizontal axis scale factor (wavelength) is 30 nm.

4. 630
TECHNICAL PHYSICS LETTERS Vol. 37 No. 7 2011
A.M. KURBATOV, R.A. KURBATOV
that this time dichroism is not lower than 15 dB in the
range ~90 nm. Here y mode is scattered by additional
layer into other modes decaying in the next ÐÌ light
guide coating. This layer was manufactured by intro
ducing in it of additions (basically of ytterbium). The
latter have their own narrow absorption bands which
are probably can’t be responsible for y mode decaying
within the whole dichroism window which is to our
point of view is determined by the scattering which is
not so sensitive to wavelength.
So, for broad band PZ fibers generally it is neces
sary: 1) to avoid too wide fluorine claddings which are
prevent to suppress the undesirable polarization
y mode, choosing its minimal width when x mode
bending losses are acceptable; 2) to grow the birefrin
gence up to ~0.001. The rest RI profile parameters in
the case of lightguide with the length ≤1 m should be
chosen from fundamental x and y mode cutoff
thresholds calculation (setting the y mode cutoff at
~1.4–1.5 μm). In the case of long lightguides the rest
RI profile parameters should be chosen by fundamen
tal x and y modes bending losses modeling.
REFERENCES
1. J. Noda et al., J. Lightwave Technol. 4, 1071 (1986).
2. S. Kawakami and S. Nishida, IEEE J. Quant. Electron.
QE 10, 12 (1974).
3. A. M. Kurbatov and R. A. Kurbatov, Pis’ma Zh. Tekh.
Fiz. 36 (17), 23 (2010) [Tech. Phys. Lett 36, 789
(2010)].
4. A. Bjarklev, J. Lightwave Technol. 4, 341 (1986).
5. Y. Tsuchida et al., Opt. Express 13, 4770 (2005).
6. P. L. Francois and C. Vassallo, App. Opt. 22, 3109
(1983).
7. S. J. Garth, J. Lightwave Technol. 7, 1889 (1989).
8. J. R. Simpson et al., J. Lightwave Technol. 1, 370
(1983).
9. M. Messerly et al., J. Lightwave Technol. 9, 817 (1991).
10. H. Renner, IEEE Photon. Technol. Lett. 3, 31 (1991).
11. W. A. Gambling et al., Opt. Quant. Electron. 11, 43
(1979).
12. A. M. Kurbatov and R. A. Kurbatov, RF Patent
No. 2250481 (Registered 20.04.2005).
13. A. M. Kurbatov and R. A. Kurbatov, RF Patent
No. 2269147 (Registered 27.01.2006).