Multimode fiber devices with single-mode performance.

Abstract

A taper transition can couple light between a multimode fiber and several single-mode fibers. If the number of single-mode fibers matches the number of spatial modes in the multimode fiber, the transition can have low loss in both directions. This enables the high performance of single-mode fiber devices to be attained in multimode fibers. We report an experimental proof of concept by using photonic crystal fiber techniques to make the transitions, demonstrating a multimode fiber filter with the transmission spectrum of a single-mode fiber grating.

Full text

Page 1

Multimode fiber devices with single-mode
performance
S. G. Leon-Saval and T. A. Birks
Department of Physics, University of Bath, Claverton Down, Bath BA2 7AY, UK
J. Bland-Hawthorn
Anglo-Australian Observatory, P.O. Box 296, Epping, NSW 2121, Australia
M. Englund
Redfern Optical Components, Australian Technology Park, Eveleigh, NSW 1430, Australia
Received May 6, 2005; accepted June 2, 2005
A taper transition can couple light between a multimode fiber and several single-mode fibers. If the number
of single-mode fibers matches the number of spatial modes in the multimode fiber, the transition can have
low loss in both directions. This enables the high performance of single-mode fiber devices to be attained in
multimode fibers. We report an experimental proof of concept by using photonic crystal fiber techniques to
make the transitions, demonstrating a multimode fiber filter with the transmission spectrum of a single-
mode fiber grating. © 2005 Optical Society of America
OCIS codes: 060.2340, 060.2430.
Many optical fiber devices exploit physical interac-
tions that are mode dependent. Although such de-
vices can have high performance in single-mode fiber
(SMF), the corresponding multimode fiber (MMF) de-
vices generally perform poorly. For example, simple
SMF Bragg gratings are highly reflective over a nar-
row band of wavelengths, but a grating written di-
rectly in MMF reflects each mode at a different Bragg
wavelength, giving a response that is spread over a
range of wavelengths and depends on the mode spec-
trum excited in the MMF.1(Sun et al.2demonstrated
narrowed MMF gratings, but only for a particular
subset of modes in a special fiber incompatible with
ordinary MMFs.) This has limited the range of high-
performance devices implemented in MMF.
Here we describe how to attain SMF performance
in MMF devices by using tapered transitions be-
tween an MMF and several SMFs. By way of ex-
ample, we made an MMF Bragg grating device with
the narrow spectral width of an SMF grating. Our
work was motivated by applications in astronomy.
When a ground-based telescope observes the night
sky in the wavelength range 600–1800 nm, the sky
background is overwhelmed by bright narrow spec-
tral lines due to atmospheric OH emission. Filtering
out these spectral features would greatly enhance
contrast in the infrared. Complex Bragg gratings can
achieve the required filter response in SMF.3How-
ever, each pixel of the image is incoherent and highly
multimode and can only be efficiently coupled to
MMF. The ability to implement the SMF response in
MMFs could revolutionize ground-based observations
at infrared wavelengths.
The second law of thermodynamics (brightness
theorem) prohibits the lossless coupling of light from
an arbitrarily excited MMF into one SMF (despite re-
ports to the contrary4), but low-loss coupling to an-
other multimode system with at least as many de-
grees of freedom is possible. If the second multimode
system is degenerate, its modes have the same propa-
gation constant and so share the same Bragg wave-
length for a given grating. Such a grating therefore
behaves like an SMF grating. Reflected light couples
back into the original MMF, and a second, reversed,
coupling arrangement can similarly couple transmit-
ted light into another MMF. The entire experiment
has MMF input and output ports but behaves like an
SMF grating.
We implemented this idea using an array of iso-
lated identical SMF cores as the degenerate multi-
mode system. Its spatial modes are the supermodes
of the array. Their number equals the number of
cores, and their propagation constants equal that of a
core on its own.5Light can be coupled between the ar-
ray and an MMF via a gradual taper transition, Fig.
1(a),conceptuallylike
coupler.5–7If the transition is adiabatic, then modes
of the MMF core evolve into supermodes of the SMF
array, and vice versa. If the number of MMF spatial
modes matches the number of SMFs, the transition is
a reversible low-loss splitter–combiner between the
MMF and the SMFs. Otherwise there are MMF
modes that do not evolve into SMF supermodes, or
SMF supermodes that do not evolve into MMF
modes, causing loss in the forward or reverse direc-
tion, respectively, for arbitrary excitation. This is im-
portant because inevitable unequal path lengths
along the SMFs effectively scramble the supermodes.
The SMF ports of two similar transitions are con-
nected. Identical SMF components in each path per-
form the same function on all the light. Such a device
with Bragg gratings, Fig. 1(b), reflects the same nar-
row band of wavelengths as each grating and trans-
mits the rest, but other functions could be imple-
mented. Indeed, some functions do not need any SMF
devices. For example, connecting the SMF ports of a
single transition yields a mode-scrambling mirror.
Multicore SMFs could be used as transmission fibers
halfa multiportfused
October 1, 2005 / Vol. 30, No. 19 / OPTICS LETTERS
2545
0146-9592/05/192545-3/$15.00 © 2005 Optical Society of America

Page 2

instead of ordinary MMFs, carrying as many modes
while avoiding focal ratio degradation caused by
mode-dependent loss.
The required transitions superficially resemble
integrated-optic mode splitters,8,9where each mode
of a multimode waveguide is channeled into a differ-
ent single-mode waveguide. However, for our pur-
poses it is enough that each MMF mode evolves adia-
batically into a supermode distributed across all the
SMFs, rather than into just one SMF. The SMFs can
then be identical, simplifying fabrication and avoid-
ing a limitation to ?5 SMFs.8
We made 1?19 transitions by using a technique10
for interfacing SMFs to photonic crystal fibers
(PCFs), Fig. 2(a). A silica cane, or ferrule, with an ar-
ray of holes was made, Fig. 2(b). Conventional SMFs
(diameter 125 ?m, cutoff wavelength ?1250 nm)
were inserted into 19 of the holes. Each SMF was
coated for ?2 cm inside the ferrule, to prevent acci-
dental cleaving, but the rest of the fiber in the ferrule
was uncoated. The filled ferrule was treated as a pre-
form and drawn to a length of multimode PCF, Fig.
2(c). The MMF core incorporated material from the
19 SMFs and was supported in an air cladding by a
network of silica webs.11(The irregular shape that
was due to distortion of the holes could be avoided by
using a ferrule with more holes.) By preserving the
neck-down region in the furnace when drawing
stopped, the MMF remained connected via a continu-
ous transition to the SMF pigtails protruding from
the ferrule, Fig. 2(d).
Nineteen nominally identical SMF Bragg gratings
were fusion spliced between two such transitions as
depicted in Fig. 1(b). Light from an erbium-doped fi-
ber amplifier continuum source was launched via a
conventional MMF into one MMF port. The output
from the other MMF port was fed via another conven-
tional MMF into an optical spectrum analyzer. The
measured spectrum is plotted in Fig. 3(a), along with
the average of the 19 SMF gratings. There were no
other features outside the 2 nm range shown. The
?0.07 nm ripples are weak Fabry–Perot fringes from
the coupling optics and not a feature of the device.
The zero on the vertical axis was adjusted for com-
parison with the SMF gratings and does not repre-
sent zero loss.
The shape of the spectrum closely matched that of
the SMF gratings, with a similar notch width and
depth. In contrast, the response of a grating in an
MMF with a numerical aperture (NA) of 0.2 would be
spread over ?15 nm. Heating 9 of the 19 gratings by
60°C shifted the Bragg wavelength of those gratings
and gave a spectrum with two 3 dB (50%) notches
separated by 0.4 nm, Fig. 3(b). Their similar depths
indicate the expected distribution of light between
the two roughly equal sets of SMFs.
The measured transmission away from the grating
notch was only 3.4% (−14.7 dB), but this high loss
can be accounted for by the mode-number mismatch
between the MMF and the SMF array. We made no
attempt to match mode numbers in this proof-of-
principle experiment. The number of MMF modes
was estimated from SEM images like Fig. 2(c). The
core is approximately a disc of diameter 34.5 ?m, and
the supporting webs were 0.5 ?m thick. Hence the ef-
fective NA was120.75, and the fiber supported ?710
spatial modes.13If these modes were equally excited
at the input, the transmission of an otherwise ideal
device would be 19/710, or 2.7% (−15.7 dB).
There are large uncertainties in comparing these
two transmission values: the measured value in-
Fig. 2.
holey ferrule filled with SMFs. (b) Optical micrograph of
the ferrule. The solid outer jacket was ?260 ?m thick. (c)
SEM image of the multimode PCF drawn from the ferrule
after the central 19 holes were each filled with an SMF. (d)
Photo of a complete transition.
(a) An MMF–SMF transition made by drawing a
Fig. 1. (a) A taper transition between an MMF and several
SMFs. Each MMF mode evolves adiabatically into a super-
mode of weakly coupled SMF cores and distributed be-
tween separate SMFs at the output. The process is revers-
ible (SMFs to MMF). (b) An MMF grating device made by
inserting SMF gratings between the SMF ports of two tran-
sitions. (c) A more manufacturable form of the device, made
by tapering a multicore fiber in two places with a low-index
jacket to give MMF ports, and writing gratings in the cores
in between.
2546
OPTICS LETTERS / Vol. 30, No. 19 / October 1, 2005

Page 3

cluded unknown coupling losses to the conventional
MMFs at input and output; the grating and device
SMFs were not identical, giving splice losses of
?0.8 dB; and the MMF modes would not have been
equally excited. Nevertheless, the similarity between
the measured transmission and that estimated by
mode counting suggests that the device would be low
loss if the mode numbers were matched. Unoptimized
ferrule transitions between PCFs and SMFs exhibit
losses10of ?0.6 dB, and analogous 1?19 fused cou-
plers with losses of 0.3 dB have been reported.7Mode
numbers can be matched by including enough SMFs
to match a given number of MMF modes or by adjust-
ing the MMF core diameter or NA to match a given
number of SMFs: 50 SMFs would have matched our
MMF if its NA were 0.2.
In our experiment the SMF cores were separate
SMFs and the MMF was a multimode air-clad PCF.
The device was therefore cumbersome to make, but
more manufacturable implementations can be con-
ceived. An example is shown in Fig. 1(c): the SMF
cores are contained in a multicore fiber, the gratings
are written in one shot across all the cores (research-
ers studying multicore gratings believe this to be
possible14), and the fiber is tapered down either side
(perhaps on a tapering rig) and cleaved to yield mul-
timode ports. Jacketing with low-index cladding
glass makes them conventional MMFs. Construction
would involve far fewer process steps (e.g., no splic-
ing) and be more scalable to higher mode counts,
making it applicable to cost-sensitive applications
such as MMF communication and sensing.
We have shown how transitions between one MMF
and several SMFs can give an MMF device with the
same characteristics as the SMF components within
it. We illustrated this with transitions made by using
PCF techniques, and we implemented an MMF filter
with the response of an SMF grating. The high loss
can be explained by the mismatch between the num-
ber of MMF modes and the number of SMFs, suggest-
ing that a mode-number matched device should be
low loss, and we have described how such devices
could be more readily manufactured.
J. Bland-Hawthorn dedicates this paper to his
brother Simon, who died in December 2004.
References
1. T. Mizunami, T. V. Djambova, T. Niiho, and S. Gupta, J.
Lightwave Technol. 18, 230 (2000).
2. Y. Sun, T. Szkopek and P. W. E. Smith, Opt. Commun.
223, 91 (2003).
3. J. Bland-Hawthorn, M. Englund, and G. Edvell, Opt.
Express 12, 5902 (2004).
4. Y. Yang, J. Lee, K. Reichard, P. Ruffin, F. Liang, D.
Ditto, and S. Yin, Opt. Commun. 249, 129 (2005).
5. F. Ladouceur and J. D. Love, Opt. Quantum Electron.
22, 453 (1990).
6. D. B. Mortimore and J. W. Arkwright, Appl. Opt. 30,
650 (1991).
7. J. W. Arkwright, D. B. Mortimore, and R. M. Adnams,
Electron. Lett. 27, 737 (1991).
8. R. N. Thurston, E. Kapon and Y. Silberberg, IEEE J.
Quantum Electron. 23, 1245 (1987).
9. B.-T. Lee and S.-Y. Shin, Opt. Lett. 28, 1660 (2003).
10. S. G. Leon-Saval, T. A. Birks, N. Y. Joly, A. K. George,
W. J. Wadsworth, G. Kakarantzas, and P. St. J. Russell,
Opt. Lett. 30, 1629 (2005).
11. R. P. Espindola, R. S. Windeler, A. A. Abramov, B. J.
Eggleton, T. A. Strasser, and D. J. DiGiovanni,
Electron. Lett. 35, 32 (1999).
12. W. J. Wadsworth, R. M. Percival, G. Bouwmans, J. C.
Knight, T. A. Birks, T. D. Hedley, and P. St. J. Russell,
IEEE Photonics Technol. Lett. 16, 843 (2004).
13. W. Snyder and J. D. Love, Optical Waveguide Theory
(Chapman & Hall, 1983).
14. I.Bennion (AstonUniversity,
February 2005), G. A. Miller (SFA, Inc., Largo, Md.,
February 2005), separate personal communications.
Birmingham, UK,
Fig. 3.
vice measured when all 19 gratings were at room tempera-
ture (solid curve), together with the average (dashed curve)
of the 19 SMF gratings as calculated from their data
sheets. (b) Same as (a) but measured when 9 of the gratings
were heated by 60°C (solid curve), together with the calcu-
lated average with 9 of the gratings shifted in wavelength
by the measured amount (dashed curve).
(a) Transmission spectrum of the MMF grating de-
October 1, 2005 / Vol. 30, No. 19 / OPTICS LETTERS
2547