eanH.

ressonr de operil prrato

1. Introduction

s) wir manyeratiodeviceequiredispered to m

ber, parametric gain can be achieved [2]. At the same time, a con-verted idler (xi) will be generated at the frequency xi = 2xp xs.In order to obtain efcient and broadband FWM, it is necessary tophase-match the waves involved in the process along the fulllength of the ber. Therefore, bers that combine a high effectivenonlinear coefcient with low dispersion, low dispersion slope

many parametric-based telecommunications devices one wouldideally like to design the waveguide dispersion in such a way thatit exactly compensates the material dispersion over a broad spectralrange, to generate a perfect dispersion-less ber, as schematicallyshown in Fig. 1b. Although a wideband dispersion-less behavior isextremely challenging to achieve in practice, broadband, near-zerodispersion proles have been obtained by exploiting the novelholey or microstructured ber designs discussed here.

In this work, we review our progress towards the design andfabrication of a highly nonlinear dispersion-attened optical berbased on high-index lead silicate glass and designed for operation

Corresponding author.

Optical Fiber Technology 16 (2010) 378391

Contents lists availab

Optical Fiber

. coE-mail address: xif@orc.soton.ac.uk (X. Feng).source. This allows the relative strength of nonlinear effects to bemaximized and can give rise to extreme spectral broadening evenin a short ber length [1]. Optical signal processing applications onthe other hand, such as wavelength conversion which is a crucialoperation in high speed wavelength division multiplexed (WDM)optical networks, often require bers with even tighter specica-tions. Four-wave mixing (FWM) in bers in particular, is regardedas one of the most promising wavelength conversion mechanisms,due to its transparency in terms of both modulation format and bitrate. By combining a strong pump wave at an angular frequency(xp) with a signal at another frequency (xs) in a highly nonlinear

and n2 is the nonlinear refractive index of the glass, respectively.According to Eq. (1), a high nonlinearity c can be achieved by choos-ing a high n2 glass as the host material and/or by targeting thesmallest possible effective mode area Aeff (which is typically ob-tained in a wavelength-sized ber core with sufciently large indexcontrast with the cladding). The second key optical property of a -ber is its dispersion prole, which characterizes the wavelengthdependence of the group velocity of the guided mode. The total dis-persion (D) of a ber can be expressed, to very good approximation,as the sum of the material dispersion (Dm) (see Fig. 1a and b) andthe waveguide dispersion (Dw) (see Fig. 1b), i.e., D = Dm + Dw. ForHighly Nonlinear Fibers (HNLFdispersion are of extreme interest foranging from supercontinuum genmanipulation and optical parametrictinuum generation applications rdispersion prole, so that the zero-nonlinear ber is accurately position1068-5200/$ - see front matter 2010 Elsevier Inc. Adoi:10.1016/j.yofte.2010.09.014th tailored chromaticphotonic applications,

n to all-optical signals. Generally, supercon-a shifted chromatic

sion wavelength of theatch that of the pump

and a short propagation length are typically required for suchapplications.

The effective nonlinear coefcient c of a ber is the primaryparameter used in order to gauge its performance for nonlinear de-vice applications, and can be expressed as:

c 2pn2=kAeff; 1

where k is the wavelength of light, Aeff is the effective mode area,Dispersion controlled highly nonlinear bat telecoms wavelengths

Xian Feng , Francesco Poletti, Angela Camerlingo, FrGiorgio M. Ponzo, Marco Petrovich, Jindan Shi, WeiOptoelectronics Research Centre, University of Southampton, Southampton SO17 1BJ, UK

a r t i c l e i n f o

Keywords:Non-silica glass berMicrostructured optical berHighly nonlinear berChromatic dispersion controlAll-optical processing

a b s t r a c t

We review our recent progtailored near-zero dispersistructured and W-type beare described in detail. Thand through accurate numbers for all-optical signakey experimental demonst

www.elsevierll rights reserved.rs for all-optical processing

cesca Parmigiani, Periklis Petropoulos, Peter Horak,Loh, David J. Richardson

in the development of lead silicate glass bers with high nonlinearity andat telecommunication wavelengths, encompassing holey, all-solid micro-esigns. The fabrication techniques and relative merits of each ber designtical properties of the fabricated bers are assessed both experimentallycal simulations. The signicant potential of lead silicate highly nonlinearocessing at telecommunication wavelengths is shown via a number ofrs.

2010 Elsevier Inc. All rights reserved.

le at ScienceDirect

Technology

m/locate /yof te

cladding.

velele o

X. Feng et al. / Optical Fiber Technology 16 (2010) 378391 379Fig. 2 shows the schematics of four types of typical index-guided nonlinear optical bers: (a) HF with an air-suspended core(ASC), (b) HF with a complex two dimensional (2D) microstruc-tured cladding, and (c) MOF with an all-solid one-dimensional(1D) microstructured cladding. The fourth design (Fig. 2d) is nottypically classied as a MOF and comprises a W-type step-indexprole with high index-contrast; this design is quite common inaround 1.55 lm. Material aspects are discussed as well as the rel-ative merits of various optimized bers. The rst promising resultson the use of our bers in short-length ber based nonlinear de-vices for all-optical processing will also be discussed. In addition,the initial results on splicing the fabricated bers with the com-mercial silica bers have been introduced.

2. Structure and material aspects for dispersion-tailored highlynonlinear bers

2.1. Fiber structures

Over the past decade and a half [3,4], single-material holey bers(HFs), often also referred to as photonic crystal bers (PCFs) ormicrostructured optical bers (MOFs), have attained considerabletechnological maturity. These bers open up the possibility formuch higher nonlinearity and better exibility for dispersion con-trol than is possible using conventional ber technology, particu-larly for realizing at, near-zero dispersion proles over a broadwavelength range. HFs exploit the large index contrast betweenair and glass, while taking full advantage of the design exibility ofwavelength-scale features in their holey or microstructured

Fig. 1. (a) Schematic example of the material dispersion of an optical glass; the wawaveguide dispersion (Dw = Dm) for achieving a at and near-zero dispersion prothe development of germanium-doped silica highly nonlinear -bers. In the following sections the ber design, fabrication tech-niques and ber characterization will be discussed in detail foreach ber type. Some key applications achieved with these berswill also be discussed.

Fig. 2. (a) ASC HF; (b) HF with complex two dimensional (2D) microstructured claddincontrast.2.2. Fiber materials

Eq. (1) indicates that using a glass with high n2 [5,6] will resultin a higher ber effective nonlinearity, c. According to materialchemistry, the linear refractive index, n, and the nonlinear refrac-tive index, n2, of a dielectric material are both attributed to thepolarizability and the hyperpolarizability of the constituent chem-ical ions [79]. Thus glasses consisting of ions with heavy atomicweight and/or large ionic radii will generally exhibit high valuesof n and n2. Fig. 3 illustrates the relation between n and n2 of var-ious optical glasses, including uoride glasses, silica, lead silicateglasses, tellurite glasses, other heavy metal oxide (HMO) glasses,and chalcogenide glasses. It can be seen that the nonlinear refrac-tive index n2 increases with the linear refractive index n, in agree-ment with the empirical prediction of Millers rule: the nonlinearoptical response of a material is related to its linear response. Inparticular, high-index HMO glasses and chalcogenide glasses(based on S, Se, and Te) possess a nonlinear refractive index, n2,which is 13 orders of magnitude higher than silica glass(2.5 1020 m2/W). This indicates that by choosing a high-indexglass as the host material for a nonlinear ber, the ber nonlinear-ity c can be enhanced by 13 orders of magnitude. In addition,glasses of higher refractive indices have higher glass/air index con-trast, thus allowing for better mode connement in the core andsmaller effective areas to be achieved. Consequently, the maximumnonlinearity c of a holey ber based on a high-index glass can be ashigh as 103105 times higher than conventional silica bers. Theuse of such bers opens up the possibility for extremely efcientand compact nonlinear ber devices with meter and sub-meterlengths.

ngth scale shown would normally span over several 100s of nanometers. (b) Idealver a broad wavelength range.2.3. Understanding refractive index and material dispersion of opticalglass

When an electromagnetic wave at optical frequencies interactswith the bound electrons of a dielectric medium, such as a glass,

g; (c) all-solid 1D MOF; and (d) W-type step-index prole ber with high index-

Techthe linear response of the medium in general depends on the opti-cal frequency, x. This effect, often referred to as chromatic disper-sion of the dielectric material, is characterized by a frequency- (orwavelength-) dependent refractive index, n(x) (or n(k)). On a fun-damental level, the origin of chromatic dispersion is related to thecharacteristic resonance frequencies at which the medium absorbsthe electromagnetic radiation through oscillations of bound elec-trons. At wavelengths far from these resonances, the refractive in-dex can be approximately expressed by the well-known Sellmeierequation [10]:

n2 x 1Xm

i1

Bi x2ix2i -2

; 2

wherexi is the resonance frequency and Bi is the strength of ith res-onance. In practice, it is very difcult to precisely determine all theresonant frequencies, and an approximate model is often used,which is based on an average electronic absorption bandgap andan average lattice absorption bandgap. This leads to the simpliedtwo-pole Sellmeier equation [10]:

n2 x 1 B1 x21

x21 x2 B2 x

22

x22 x2; 3

where x1 is the resonance absorption frequency of the electronictransition, x2 is the resonance absorption frequency of the vibra-tion transition, and x1 and x2 typically lie in the ultraviolet (UV)

Fig. 3. Relation between the linear (n) and nonlinear refractive index (n2) of variousoptical glasses.

380 X. Feng et al. / Optical Fiberand in the infrared (IR) region, respectively. B1 and B2 are the corre-sponding resonance strengths. Eq. (3) can be formulated as a func-tion of the wavelength k:

n2k A Bk2

k2 k21 Ck

2

k2 k22; 4

where k1 and k2 are the absorption wavelengths in the UV and IR re-gions, respectively; A, B, C are the constants concerned with theabsorption strength, and k1,2 = 2pc/x1,2 [11].

Fig. 4 shows a schematic of the refractive index n(k) and of thetransmission T(k) of a dielectric material at optical frequencies. Thematerial dispersion is then dened as:

Dm kc d2nkdk2

: 5

According to Eq. (5), the wavelength of zero material dispersion,k0, falls at the inection point of n(k), indicated by the circle inFig. 4. From the viewpoint of material science, the position of k0is determined by the combined effect of both the average elec-tronic absorption bandgap and the average lattice absorptionbandgap, as modeled by the two-pole Sellmeier equation, Eq. (4).Table 1 shows a summary of k0, k1 and k2 for various opticalglasses. It is clear that the position of k0 depends on the positionsof the resonance peaks in the UV and the IR, k1 and k2, as well ason the refractive index. For most high-index, high-nonlinearityglasses such as lead/bismuth silicate glasses (HMO, SiO2 based)and pure HMO glasses (non-SiO2 based), including tellurite (TeO2based), germanate (GeO2 based) and gallate (Ga2O3 based), thezero-dispersion wavelength, k0, typically falls between 2 and3 lm. The material dispersion of such glasses at 1550 nm is there-fore always large and normal. An optical ber with a at and near-zero dispersion prole in the extended telecommunicationwindow (1.31.7 lm) requires a signicant amount of waveguidedispersion to compensate for the material dispersion, which canonly be achieved with a high index contrast between the ber coreand the cladding.

2.4. Properties of selected commercial high-index lead silicate glasses

In this work, we focus on lead silicate glasses, which representan attractive family of glasses due to their superior thermal andcrystallization stability and less steep viscositytemperature char-acteristic curves as compared to most compound glass alternatives[17]. In particular, we employ three commercially available leadsilicate glasses, Schott SF57, SF6, and LLF1, to fabricate variousHNLFs with tailored dispersion at telecommunication wave-lengths. Table 2 shows the basic thermal and optical propertiesof these three glasses. In particular, it can be seen that SF57 andSF6 glasses have high nonlinear indices n2 of 41 1020 and22 1020 m2/W, respectively.

The glass viscosity is paramount for every thermal engineeringstep involved in the ber fabrication, e.g. preform fabrication and/or ber drawing. Fig. 5 shows the viscosity curves of Schott SF57,LLF1, and SF6 glasses measured using the parallel plate method[6]. Note that the temperature corresponding to the viscosity of107.6 poise, T7:610 (shown in Table 2), is the glass softening tempera-ture, around which glass extrusion can be performed. As will bedemonstrated in the following, despite the large difference in vis-cosity between these glasses at any given temperature, we have

Fig. 4. Schematic dispersion curve of refractive index n(k) and transmission curveT(k) of a dielectric material in the optical wavelength range.

nology 16 (2010) 378391consistently been able to draw different combinations of them intobers.

3. Fiber design, fabrication, and characterization

In this section we review our recent progress in developinghighly nonlinear lead silicate glass bers, including holey bers,all-solid 1D MOFs, and W-type index prole bers, with attenedand near-zero dispersion proles around 1550 nm.

3.1. Air-suspended core holey ber

A holey ber is a single-material optical ber where light guid-ance is obtained through an array of longitudinal air holes sur-rounding a solid glass core. Holey ber technology can achieve

Table 1Summary of k0 of various types of optical glass.

Glass k1 (lm) k0 (lm) k2 (lm) References

Low refractive index (n < 1.7) Silica

1550 nm. As will be shown in the following, however, althoughfabrication of ASC HF with submicron core dimensions is possible,the precise position of the ZDW is extremely sensitive to the corediameter, thus posing very stringent constraints in terms of consis-tency of the core diameter along the ber length, which are verychallenging to meet with the current fabrication techniques.

The following procedure is generally employed in order to fab-ricate a non-silica ASC HF with a submicron core diameter: (1) astructured preform is extruded from a glass billet; (2) the struc-tured perform is elongated or drawn into a small cane with theouter diameter of 12 mm; and (3) the cane is inserted into a jac-keting tube and this preform assembly is then drawn into a berwith the targeted core dimension through a method very similarto the rod-in-tube technique [28]. A schematic of the fabricationtechnique for the extrusion of structured non-silica glass preformsis illustrated in Fig. 7a. First the glass billet is heated to above theglass softening temperature (519 C for SF57). Then pressure is ap-plied through the ram onto the glass, thereby forcing a viscousglass ow through the structured die. The structured preform isthen cooled down to room temperature.

Fig. 7b shows an optical photograph of the cross-section of the

tain a ZDW within the C-band, the core diameter needs to be con-trolled with an accuracy better than 10 nm.

It is worth noting that a signicant difference is observed inFig. 9 between the dispersion proles calculated for an idealizedASC HF structure (see Fig. 6) and the actual fabricated structurewith the same core diameter dcore of 0.74 lm (Fig. 8). This discrep-ancy is due to the fact that the two structures have slightly differ-ent core geometries outside the enclosed circle (Fig. 6a) whicharise as a consequence of the surface tension on the submicroncore during the ber drawing. For ASC HFs with a submicron-diam-eter core, the guided mode will inevitably expand beyond the glasscore in the air holes and in the transition area with the glassspokes. Thus the dispersion of the real ber is not only sensitiveto the core diameter but also to the geometry of the transition areabetween the core and the supporting spokes. This makes it verydifcult to precisely control the dispersion prole, e.g., the locationof the ZDW, of the submicron ASC HF. This difculty, in spite of itsextremely high nonlinearity, has to date limited the practical use ofthis ber type.

3.2. Holey ber with complex microstructured cladding

A key requirement for several applications of HNLFs based onfour-wave-mixing (FWM) is that in order to satisfy exactly orapproximately the phase matching condition over a broad wave-length range the bers should have both zero dispersion and zerodispersion slope at the required wavelength of operation, typically1.55 lm. [30]. In the case of applications requiring short (i.e., meter

382 X. Feng et al. / Optical Fiber Techextruded 10 mm outer diameter (OD) SF57 preform. A ScanningElectron Microscopy (SEM) image, Fig. 7b and c, shows that thestructured preform has a glass core with a diameter of 130 lm,and three high aspect ratio supporting spokes, about 1.45 mm inlength and only 7 lm thick.

The preform was then elongated into a small cane and insertedinto a jacketing tube with suitable diameter. The ASC HF with asubmicron-diameter core was then drawn. Negative pressure wasused during the ber drawing in order to eliminate the gap be-tween the structured cane and the jacket tube and avoid the pres-ence of any interstitial holes. Fig. 8 shows SEM images of the HFwith a core diameter dcore = 0.74 lm. The average thickness tspokeand length lspoke of the three spokes are 0.12 lm and 4.0 lm,respectively.

Based on the ber geometry obtained via high resolution SEMimages, the wavelength dependent dispersion curve was calcu-lated using full vector Finite Element Method (FEM) simulations[29]; the results are shown in Fig. 9. The ASC HF has a ZDW at1.59 lm and a dispersion slope of 0.7 ps/nm2/km at 1550 nm.The nonlinearity c of the ber is calculated to be 1850W1km1

at 1550 nm, close to the maximum theoretical value for SF57 HFsFig. 7. (a) Schematic of extruding a soft glass structured preform. (b) Opticalphotograph and (c) SEM photograph of an extruded SF57 preform with 130 lm corediameter and 10 mm OD.[23]. However, Fig. 9 also conrms that the dispersion of this berdesign is extremely sensitive to the core diameter dcore, which isthe primary parameter used to control the dispersion prole inan ASC HF. By reducing dcore by approximately 5%, from 0.74 lmdown to 0.70 lm, the ZDW of the HF shifts by over 100 nm from1.55 lm to 1.45 lm. This observation suggests that, in order to ob-

Fig. 8. SEM photographs of SF57 ASC HF with dcore of 0.74 lm.

Fig. 9. Calculated dispersion curves of the fabricated SF57 glass ASC HF and thecalculated dispersion curves of idealized bers with dcore of 0.70, 0.74 and 0.80 lm.

nology 16 (2010) 378391or sub-meter) device lengths, these conditions can be somewhatrelaxed and near-zero dispersion and dispersion slope at1550 nm are acceptable.

The air-suspended structure described in the previous sectiondoes not allow easy control of both dispersion and dispersion slopeindependently, thus a different and more complex HF structure isrequired [31,32] where additional degrees of freedom are availablein engineering these properties. We have therefore considered anomitted centre-hole structure based on a triangular lattice.Fig. 10 shows a schematic of the ber structure and an opticalproperty map displaying how the chromatic dispersion and thedispersion slope depend on the structural parameters in the caseof a ber made of SF57 glass at 1550 nm. The condition for zerodispersion and zero dispersion slope is highlighted in the graphand corresponds to a hole-to-hole spacing (K of 1.36 lm and a rel-ative hole size (d/K) of 0.454. The ber nonlinearity c of this par-ticular ber is predicted to be 470W1km1 at 1550 nm. Bycomparison, a silica highly nonlinear HF with attened and near-zero dispersion at 1550 nm only has c of 1020W1km1[32,33].

The structure shown in Fig. 10 has ve rings of cladding holes,

X. Feng et al. / Optical Fiber Techand this is required in order to achieve a low connement loss ofthe fundamental mode LP01. Even though reports show that pre-form extrusion should still be possible for the fabrication of sucha structure [34], extruding preforms for structures with a largenumber of moderately sized holes (d/K 0.45) is considered tobe quite challenging, especially when the precise size and overallshape of the structure is of primary importance. However, thenumerical simulation indicates that, (1) the dispersion of the nalholey ber is mainly determined by the hole dimensions of the rstand second ring of holes around the solid core, consequently (2) aHF with slightly graded hole sizes will still provide similar atten-ing and a near-zero dispersion prole as a HF with all identicalholes would. Therefore, we have investigated an alternative meth-od, based on the Structured Element Stacking Technique (SEST)[31,35]. First, a preform with three rings of holes was obtainedby extrusion (note that the complexity in extruding this structureis signicantly lower than the corresponding of the 5-ring struc-ture of Fig. 10). An optical microscope image of its cross sectionis shown in Fig. 11a; the corner-corner length of the preform is16.2 mm. It is important to notice that, in order to pre-compensatefor the non-uniform expansion of the holes that occurs during thesubsequent ber drawing step, the hole-to-spacing ratio d/K of thedifferent rings of holes needs to be graded.

The extruded stacked preform was then elongated to a cane andstacked at the centre of six other thin SF57 capillaries inside anSF57 jacket tube. This assembled preform was then drawn into aber. Fig. 11b shows SEM images of the fabricated HF, where the

Fig. 10. (a) Schematic of a HF based on a triangular lattice of holes and an omittedcentre-hole structure and (b) optical property map (at 1550 nm) for a SF57 HF. K isthe hole-to-hole spacing and d is the hole diameter, respectively. Dispersion (D)

contours (in ps/nm/km) are shown as solid curves, while dispersion slope (DS)contours (in ps/nm2/km) are shown in dashed curves. The circle represents thecross point of the zero dispersion and dispersion slope.two different holey claddings can clearly be identied. The rstcladding, immediately surrounding the core, has an average holespacing (K1 of 1.60 lm; the hole diameter d1 of the rst ring ofholes is 0.56 lm, and d/K varies from 0.35 to 0.5 in the three ringscomposing this inner cladding. The outer cladding, consisting of sixlarge holes, exhibits a large hole-to-spacing ratio d2/K2 = 0.85. Thedispersion of this HF is primarily determined by the structuralparameters of the inner rst cladding, while the connement lossof the LP01 mode is mainly determined by the structure of the outercladding. Although numerical simulations indicate that this HFtheoretically supports a few modes, effective single-mode guid-ance was experimentally observed at 1550 nm as conrmed byanalysis of near eld modal proles. We speculate that this isdue to a combination of a good launch into the fundamental modeand higher differential loss of the higher-order modes. The c valueof this HF was measured as 270W1km1 at 1550 nm, using theBoskovic method [36], whereas the propagation loss was measuredto be 3.0 0.1 dB/m at 1550 nm.

We obtained an accurate estimate of the ber dispersionthrough numerical calculations based on the actual ber structureobtained by high resolution SEM images; the result is shown inFig. 12. Measurements at 1550 nm wavelength have conrmed adispersion value of 17 ps/nm/km (indicated by a circle inFig. 12) and a dispersion slope of 0.10 ps/nm2/km. Fig. 12 alsoshows an optimized dispersion prole which exhibits minimal dis-persion across the wavelength range 15001600 nm. A ber withthese characteristics would be ideal for applications targetingparametric processes. In order to obtain this prole, it would benecessary to (1) reduce the hole spacing in the rst cladding, K1,from 1.60 lm to 1.36 lm and (2) increase the hole-to-spacing ratioin the rst cladding, d1/K1, from 0.35 to 0.45. The second point isparticularly challenging to achieve in practice. First, one has tocarefully choose different values of d/K for each ring of the ex-truded preform, in order to compensate for the distortion of theholey microstructure during ber drawing, where the differentrings of holes generally experience different expansion. Moreover,the actual amount of hole expansion is strongly dependent on thedetailed ber drawing conditions, such as length/shape of the hot-zone of the furnace, diameter of the preform and preform feed rate,making it very hard to obtain a consistent structure. We thereforeconcluded that, for this HF structure, it is currently very difcult toobtain the target parameters required in order to achieve near-zerochromatic dispersion and dispersion slope at 1550 nm. It is how-ever possible that continued improvements of the extrusion and -ber drawing processes may lead to more consistent and usablebers.

3.3. All-solid 1D microstructured ber

We have seen in the previous sections that, when targeting aattened and near-zero chromatic dispersion at the ps/nm/km le-vel at 1550 nm in holey ber structures, it is necessary to achievecontrol of hole size and spacing to the submicron level. This leadsto two fabrication-related problems: (1) submicron control of holesize is very difcult to reliably achieve in practice because of thecomplex interdependence between temperature, surface tensionand internal pressure in the holes. This will almost inevitably causeconsiderable deviations between the initial structure, e.g. in an ex-truded preform or cane, and the nal ber, which will in turn causea signicant deviation from the targeted dispersion prole and (2)for hole dimensions below the wavelength scale, the surfaceroughness of the glass at the holes, in conjunction with the high in-dex contrast between air and glass, will generally result in a high

nology 16 (2010) 378391 383scattering loss at the air/glass interface [4,24].An elegant solution to overcome these drawbacks is to use a

high index-contrast all-solid microstructured optical ber (MOF)

f ho

Techrather than a holey ber: in these microstructured bers holes are

Fig. 11. (a) Optical microscope image of an extruded SF57 preform with three rings ostructure is surrounded by a further ring of expanded air holes.

Fig. 12. Calculated dispersion curve of the optimum SF57 HF design (d/K = 0.454,K = 1.36 lm, solid line), calculated dispersion curve of the fabricated SF57 HF withcomplex holey cladding (dashed line) and measured dispersion at 1550 nm (circle).

384 X. Feng et al. / Optical Fiberreplaced with solid regions of a second glass, which signicantlyreduces the occurrence of structural deformations induced duringthe ber drawing and helps maintain the original relative scale fac-tor of high index/low index features when going from preform tober [3739]. Furthermore, any losses due to interface scatteringcan be substantially reduced in this case by accurately polishingthe surfaces of the glass elements used to fabricate the preform.Following this idea, we recently fabricated an all-solid one-dimen-sional (1D) MOF with low loss, high nonlinearity and low disper-sion at 1550 nm [39]. Two chemically compatible commercialoptical glasses, SF6 and LLF1 (respectively high-index and low-in-dex, see Table 2), were used for this MOF. As shown in Fig. 5, thereis a substantial thermal mismatch in the viscosities of the twoglasses. Although such mismatch affects in principle both preformextrusion (carried out at viscosity ranges of 109107 poise) and -ber drawing (106.5104 poise) processes, we found that the twoglasses are in fact thermally compatible and that it was possibleto produce high optical quality preforms and bers. We pursueda ber design comprising a number of alternating high- and low-index coaxial rings with the layer thickness ultimately determiningthe optical properties of the ber. The structured preform was fab-ricated by co-extruding alternately stacked high- and low-indexglass discs through a circular aperture (see Fig. 13). Note that allthe glass discs used within our extrusion process were of highoptical quality with accurately and nely polished surfaces. Theco-extrusion method employed to form the multiple coaxial-ringstructure in the extruded preform is explained in more detail inRefs. [38,39].

A structured preform with an outer diameter (OD) of 10 mmwas elongated to a cane of 1.09 mm diameter and then insertedinto an extruded SF6 glass jacket tube with an OD of 15.80 mmand an inner diameter (ID) of 1.10 mm. From a single ber draw,bers with core diameters of 5.0, 4.1, 3.7, and 3.3 lm have beenfabricated with excellent yield of over 100 m in length. Fig. 14ales surrounding a solid core. (b) SEM images of the nal double clad HF; the central

nology 16 (2010) 378391shows SEM images of the obtained MOF, where the regions of dif-ferent glass are highlighted via a z-contrast technique: a high-index glass has a higher brightness than a low-index glass, as ex-plained elsewhere [37]. The ber shown has a 150 lm OD and ahigh-index SF6 core of 3.7 lm in diameter. The core is surroundedby alternating LLF1 and SF6 glass rings. The three low-index ringshave thicknesses of 1.1, 0.6, and 0.6 lm, respectively. The twohigh-index rings have thicknesses of 0.4 and 0.3 lm, respectively.Fig. 14b illustrates the prole of the refractive index of the 1DMOF, according to the structural parameters obtained from highresolution SEM images.

Fig. 15 compares the ber structure to that of the 1.09 mm ODcane from which it was fabricated. As a visual aid, four rectangularframes between the cane and the ber are also superimposed onthe images to mark the various boundaries of the two glasses. Notethat on the cane a third high-index ring is present, which becomesmerged with the outer jacketing glass tube during the ber draw. Itis very obvious that the shape and relative size of the high/low in-dex features in the cane and in the nal ber are completely iden-tical, even though the features in the nal ber are micron-scale.This clearly demonstrates the advantages of using all-solid MOFsover air-lled HFs. The former can give far more predictable andreliable control of the microstructure during the various fabrica-tion stages and right up to the nal ber. This is critically impor-tant when targeting MOFs with strongly structure-dependentber properties.

We calculated the optical properties of the solid MOF assumingan idealized structure with parameters matching those measuredfrom high resolution SEM images. It was found that the MOFshown in Figs. 14 and 15 (3.7 lm core diameter) supports a few

Fig. 13. Schematic of the fabrication of a preform with coaxial-ring structures byco-extrusion of alternately stacked high- and low-index glass discs.

speed to minimize the formation of any bulk and interface defectsin the structured preform. By contrast, conventional soft-glass HFfabrication procedures involving extrusion [18] or drilling [4,34],tend to produce signicant surface roughness and/or surface de-fects in the preforms, which inevitably impact the quality of theair/glass interfaces in the nal bers and can give rise to relativelyhigh attenuation due to light scattering.

The effective nonlinearity c of the 1D MOF was again measuredusing theBoskovicmethod [36],whichgavea valueof 120 W1km1

at 1550 nm. Based on the numerical modeling, the effective modearea Aeff was calculated to be 6.7 lm2 at 1550 nm, which corre-sponds (using the nonlinear refractive indices n2 of the two glasses)to a nonlinear coefcient c of 130 W1km1, in good agreementwiththe measured value.

The dispersion of the 1D MOF was measured to be +12.5ps/nm/km at 1550 nm using the FWM method [40] this valueis shown by a cross in Fig. 17. The chromatic dispersion was again

X. Feng et al. / Optical Fiber Technology 16 (2010) 378391 385higher-order modes (HOMs) at 1550 nm. The connement loss ofall HOMs is however of several tens of dB/m and effectively sin-gle-mode guidance was observed from this 1D MOF over meter-

Fig. 14. (a) SEM images and (b) index prole of the solid MOF with 3.7 lm corediameter.scale lengths (see near-eld mode prole in Fig. 16a). Fig. 16a alsoshows the intensity prole of the simulated fundamental LP01mode. Using the cutback method, the propagation loss of the fun-damental mode was measured as 0.8 0.2 dB/m at 1550 nm; forincreased accuracy, the loss value was determined with multiplecutbacks (the total cutback length of 2.4 m) (see Fig. 16b). Thismeasured loss gure is very close to the bulk attenuation deducedfrom the published data of commercial Schott SF6 glass [14]. To thebest of our knowledge, this loss gure is one of the lowest ever re-ported in a non-silica glass MOF. We attribute such a very low lossto the following reasons: (1) the starting glass discs used for pre-form extrusion were nely polished to ensure a high optical qualityand (2) the extrusion was optimized in terms of temperature and

Fig. 15. Comparison between the structural parameters of the cane (left) anddetermined numerically using the structural parameters of the realber as input; we employed a full vector nite-element (FEM)method and independently conrmed the results via a semi-analytic transfer matrix approach. The dispersion curve shown inFig. 17 indicates that this 1DMOF is in practice a dispersion-shiftedber with a ZDW at about 1475 nm and a dispersion slope of 0.16ps/nm2/km at 1550 nm. The calculated dispersion at 1550 nm is ingood agreement with the measured value. For reference, the dis-persion curve of commercial silica SMF28 ber is also illustratedin Fig. 17. At 1550 nm the dispersion value of the MOF is close tothat of SMF28, even though the effective nonlinearity of the MOFis 120 times higher than the SMF28. A few meters of this 1DMOF can therefore exhibit a total nonlinearity equal to that ofseveral hundreds of meters of SMF28 whilst exhibiting much lessnet dispersion an important property for many nonlinear appli-cations such as FWM. Fig. 17 also illustrates the material disper-sion of bulk SF6 glass and the calculated dispersion curves of the1D MOFs with core diameters of 5.0 lm, 4.1 lm, and 3.3 lm fabri-cated from the same cane. It can be seen that reducing the size ofthe wavelength-scale features in the microstructured claddinggenerates strong and positive waveguide dispersion, i.e. oppositein sign to the material dispersion of the core glass, thus shiftingthe ZDW of the ber to shorter wavelengths. For instance, theZDW shifts from 1.57 lm to 1.44 lm when the core diameter isdecreased from 5.0 lm to 3.3 lm.

The characterization results suggest that the 1D MOF is suitablefor applications relying on FWM in the 1.55 lm band. To investi-gate this further we set-up a FWM-based wavelength conversionscheme, in which a cw wave and a pulsed data signal (which actedthe nal ber (right) for the all-solid MOF with coaxial-ring structure.

cations involving parametric effects. In order to further atten thedispersion of these bers, a more accurate control of the ratio be-tween the core diameter and the thickness of the innermost ringswould be required. Although the all solid nature of these bers al-lows to achieve a remarkably low transmission loss (for a soft glassMOF) and excellent structural consistency when drawing from pre-form to cane and nally into ber, it is still very challenging to con-trol the thickness of many rings in the preform independentlyusing the co-extrusion method described above. Therefore we con-sidered a further and simpler fabrication approach, as describedbelow.

3.4. W-type step-index proled ber with high index contrast

We looked for a ber design and a fabrication approach whichwould enable us to maintain the desirable deformation-free prop-erties of all-solid bers, while allowing a better control of the over-all structural features and dispersion properties. In an attempt tosimplify the design, we went back to the most basic optical berdesign possible, i.e. the step-index design. Waveguide theory shows

386 X. Feng et al. / Optical Fiber Technology 16 (2010) 378391Fig. 16. (a) Simulated (left) and observed (right) LP01 mode of 1D MOF; (b) poweras the pump) were launched into a 1.5-m long piece of the 1DMOF.Broadband wavelength conversion with a -3 dB bandwidth of17 nm was observed, covering a substantial part of the C-band[39]. Note, however, that a short length of ber was used in orderto achieve a broad conversion bandwidth. This in practice limitedthe conversion gain achieved and did not fully utilize the benetsof the exceptionally low loss exhibited in this ber. Further optimi-zation of the dispersion properties is required in order to improvethe performance of the ber when considered for broadband appli-

transmission versus ber length obtained from a cutback measurement, and lineart.

Fig. 17. Calculated dispersion curve and measured dispersion (marked by the cross at 1fabricated 1D MOFs with dcore = 5.0, 4.1 and 3.3 lm, as well as dispersion of silica SMF28inset frame in the right gure.that, for any given core material, it is always possible, in principle,to engineer a step-index ber (with diameter d and step index Dn)such that, at a given wavelength, the waveguide dispersion (Dw)exactly compensates for the material dispersion (Dm) in absolutevalue and slope, thus creating an overall dispersion attened prole[41]. Modifying Dn can be regarded as a material-based route tocontrol the ber dispersion, as opposed to the structure based ap-proach followed in HFs where the hole sizes are modied.Although the structure-based route has the advantage of requir-ing only one glass, as already discussed in the previous sections,there are inherent difculties in engineering the desired air holedimensions with the required precision.

In this work, we chose SF57 glass as the core material and sys-tematically searched for a suitable, compatible cladding glassgenerating a near-zero dispersion prole around 1550 nm in astep-index ber structure. Combining practical material con-straints with the results from numerical simulations, we found thatSchott LLF1 glass is an excellent option as a cladding medium[42]. When the core diameter is reduced to wavelength-scaledimensions, the SF57-LLF1 combination provides a high enoughindex-contrast and hence a strong enough waveguide dispersionto compensate for the core material dispersion. As shown in550 nm) of the 1D MOF with core diameter dcore = 3.7 lm, dispersion curves of theber and of bulk SF6. Note that the left gure is corresponding to the rectangular

conditions, bers with good optical and mechanical propertiescould nonetheless be fabricated.

The W-type ber was fabricated using the rod-in-tube method.The inner region was made from a SF57 rod with OD 2.9 mm in-serted into an LLF1 tube with an ID of 3 mm and an OD of13 mm. The rod surface and both inner and outer surfaces of thetube were polished to optical quality. A SF57-LLF1 cane with0.8 mm OD was then drawn from this assembly and inserted intoa SF6 jacketing tube, from which the nal ber was drawn.Fig. 20 shows SEM images of the cross-section of the fabricated -ber. The ber had a core diameter of 1.63 lm and a ratio of core

X. Feng et al. / Optical Fiber Technology 16 (2010) 378391 387Fig. 18a, such a step-index ber with a core diameter d0 of 1.641.68 lm has a low dispersion D(|D| < 5 ps/nm/km) in the wave-length range between 1.45 and 1.65 lm. According to our calcula-tions, however, this particular structure is predicted to support twomodes, as indicated from Fig. 18b.

In order to obtain single-mode operation in such a step-indexber, in analogy to the concept used in conventional W-type silicabers with low index-contrast [43], an outer cladding with a high-er index than the effective index of the LP11 mode is introduced inthe ber structure. By choosing an appropriate diameter ratio forthe rst and second cladding, it is possible to impose a very highconnement loss on the LP11 mode, thus enabling effectively sin-gle-mode operation for a broad wavelength range. For the secondcladding we chose Schott SF6 glass (n = 1.76 at 1550 nm), andfound that the resulting SF57-LLF1-SF6 W-type ber withd0 = 1.68 lm and d1 = 7.4 lm (see Fig. 19), not only is single-modedbut also has a dispersion prole very similar to the SF57-LLF1 step-index ber with the same core diameter (shown in Fig. 18a).

Fig. 18. Numerical simulation of (a) dispersion prole and (b) effective index neff ofthe LP01 and LP11 modes of a SF57-LLF1 step-index ber.In order to realize such a ber design in practice, the crucial is-sue is whether the three proposed glasses exhibit sufcient ther-mal and chemical compatibility to allow for the ber fabricationprocess. As already discussed previously and shown in Fig. 5, thethree glasses (Schott SF57, LLF1 and SF6) have a relatively large vis-cosity mismatch in the temperature range for ber drawing (vis-cosities of 106.5104 poise), indicating a challenge for thefabrication. Based on previous experience and on preliminary tests,we proved however that with careful denition of the drawing

Fig. 19. Refractive index prole of the SF57-LLF1-SF6 W-type ber.diameter to cladding diameter of 1:4.4, the same as the value ofthe initial preform and matching exactly the original target. Thisstraightforward achievement of structural targets clearly presentsan advantage of our all-solid ber technology over the holey bercounterpart.

The propagation loss of the fabricated ber was measured to be2.1 0.2 dB/m at 1550 nm using the cutback method. We expectthat this loss value can be reduced by further optimization of thefabrication process to a value similar to that of the 1D MOF de-scribed above. The predicted effective single-mode guidance wasconrmed by analyzing the near-eld mode proles of the berat 1550 nm.

Despite the inherent difculties in measuring the bers disper-sion due to its low absolute value and short effective length, weobtained a very accurate and broadband measurement using alow-coherence MachZehnder interferometric set-up and asupercontinuum source [44]. High quality interferograms weremeasured for the two fundamental polarization states, one ofwhich is shown in Fig. 21a, from 1.30 to 1.68 lm. The resultingdispersion curve, shown in Fig. 21b, is in excellent agreement withthe numerical predictions, and shows a attened prole at telecomwavelengths. In particular, at1.52 lm the dispersion slope is zeroand D = 2.6 ps/nm/km. Fig. 21b also shows that by drawing a 2%larger ber, a region of small and at anomalous dispersion a fewhundreds of nm wide can be achieved.

The ber effective nonlinearity c at 1.550 lm was measured tobe 820W1km1, in good agreement with a simulated value of854W1km1 (effective area Aeff 2 lm2). Note that this value is1.6 times higher than the predicted nonlinearity of the SF57dispersion attened HF (zero dispersion and zero dispersion slopeat 1550 nm) discussed in Section 3.2, which shows a second advan-tage of the present design as compared to its air/glass counterpart.

We performed further FWM-based wavelength conversionexperiments using this latest ber, achieving markedly improvedresults in comparison to those obtained with the 1D MOFdescribed in the previous Section. In these experiments a quasi-CW pump was placed in the middle of the C-band, (kpump =1550 nm) while the CW signal was tuned across the entire rangeallowed by the sources we were using (15301570 nm). Fig. 22shows the results obtained for a peak pump power of 35dBm anda ber length of 2.2 m. A nearly at conversion prole wasFig. 20. SEM images of the W-ber with d0 of 1.63 lm.

together. In comparison with other alternatives like connectorsand freespace coupling, splicing (mechanical splicing and fusionsplicing) is a key technology to achieve compact integrated berbased devices.

In a mechanical splice, two cleaved ber tips are mechanicallyaligned to each other and then index matching gel is positioned be-tween the ber tips to maximize coupling and minimize reec-tance. Because mechanical splicing is proceeded at the roomtemperature and consequently it is ideal to join silica ber andnon-silica glass ber, which possess very different thermal proper-ties. A low splicing loss of 0.3 dB per point was reported by NTT

Technology 16 (2010) 378391Fig. 21. (upper) observed interferometric pattern of single polarization broadbandlight propagating through the W-ber; (lower) measured and calculated dispersionof W-bers with core diameters d0 of 1.63 and 1.67 lm.

388 X. Feng et al. / Optical Fiberobtained across the full 40 nm span that was available from oursources, and a conversion efciency (dened as the ratio betweenthe idler and signal power) of 0 dB was measured. Fig. 22 alsoshows how the experimental results we obtained match the simu-lated gain prole, which suggests an overall 3 dB conversion band-width in excess of 60 nm can be achieved in this length of ber.

Using the same sample of the fabricated W-type ber we havealso experimentally demonstrated the simultaneous wavelengthconversion of three 40 Gbit/s Differential Phase Shift Keyed (DPSK)signals [45]. The unique ber properties allowed for a uniform con-version efciency between the three wavelength converted signalsof 12 dB and an excellent performance in terms of their noiseproperties as conrmed through eye diagram and bit error ratiomeasurements.

Finally, using a 3-m long sample of the same ber we have dem-onstrated the generation of high quality, high repetition rate pulses(>160 GHz) based on parametric mixing of two phase-locked nar-row-linewidth tones and numerically investigated the potentialof using such a ber for the generation of even higher repetitionrates (>1 THz) [46].

3.5. Asymmetric conguration for fusion splicing silica ber and softglass HNLFs

In the reality, it is necessary to integrate the above non-silicaglass HNLFs into the existing silica ber based telecommunicationsystem by joining these two different types of optical bers

Fig. 22. Measured (scatter symbols) and simulated (solid line) curves for the FWMconversion efciency in the W-type ber.(Japan) between a high-NA small-core tellurite glass ber and ahigh-NA silica ber, using the mechanical splicing method [47].However, the refractive index of most index matching compoundsunfortunately varies with temperature so that the optical perfor-mance of a mechanical splice can be sensitive to the ambienttemperature and the laser power launched into the ber [48]. Inparticular, in the case of FWM-based nonlinear applications, therequired average power level of the pump laser (CW or pulsed)launched into the small-core nonlinear ber is typically between1 and 10 W. Therefore, fusion splicing is obviously the best optionto obtain long-termly reliable splices between silica bers andhighly nonlinear non-silica glass bers.

Compared with other fusion splicing techniques like lamentfusion splicing, laser fusion splicing and gas fusion splicing, electricarc fusion splicing is the most commonly used technique, due tothe compact and economic hardwares, the quick and efcient oper-ation procedures, and the well accepted reliability to achieve lowsplicing loss and robust mechanical strength on the spliced berjoints. In this work, a fusion splicer (Model: Ericsson FSU 975)was employed to splice the commercial silica bers with our softglass HNLFs. In the normal procedure to splice two silica bers to-gether, the heating element, i.e., the arc electrodes here, is placedsymmetrically in the gap between the two bers. In this work, inorder to overcome the difculty arising from the low softeningtemperatures of the lead silicate glasses (1600 C), an asymmetricconguration [49], i.e., moving the heating element along the silicaber from the gap (see Fig. 23), was adopted to splice the non-silicaglass HNLF and the commercial silica ber. This asymmetric con-guration heats but does not soften the silica ber and heats thesoft glass ber to above its softening temperature directly via theheating elements and indirectly via the silica ber. Thus, the tem-perature at the end of the silica ber is greater than the tempera-ture at the end of the multi-component ber. This temperaturegradient serves to improve thermal diffusion between the two -bers when brought into contact thereby strengthening the fusionsplice.

Table 3 summarizes our initial results on fusion splicing thecommercial high-NA silica ber (UHNA3, Nufern, mode eld diam-eter (MFD): 3.3 lm) with the lead silicate glass HNLFs. It is seenthat for splicing the 1D MOF with the UHNA3, the initial splicingFig. 23. Photograph of the spliced high-NA silica ber (Nufern UHNA3) (left) and a1D MOF (right) after the fusion splicing with asymmetric conguration.

ghly

rame

: 3.7lm: 1lmlm

Techloss is 2.2 dB per point at 1550 nm. The origin of this loss is mainlydue the imperfections formed by the arc fusion, while both bershave the similar MFD. For splicing the W-type HNLF with theUHNA3, the initial splicing loss is 9 dB per point at 1550 nm, whichshould arise from both MFD mismatch and imperfections formedby the arc fusion. In addition, the splicing loss of the UHNA3 berand the standard SMF28 silica ber is currently reduced down to0.2 dB per point level in our experiment.

A recorded best splicing loss of 1.76 dB was achieved when fu-sion splicing a high bismuth oxide containing borosilicate glassHNLF (MFD: 2.1 lm) with the high-NA silica ber (UHNA4, Nufern,MFD: 3.3 lm) [50]. Therefore, we expect that the splicing loss levelof 12 dB per point, can be achieved for splicing the W-type HNLFwith the high-NA silica ber. However, it is known that fusionsplicing these two types of dissimilar glass bers together withlow splicing loss and reasonably good mechanical strength is chal-lenging and also a laborious and time-consuming task. We are cur-rently exploring the splicing condition and congurations in orderto obtain the total splicing loss less than 2 dB level from a standardSMF28 ber to the W-type lead silicate glass HNLF.

4. Conclusion

We have reviewed our recent progress in the development ofhighly nonlinear lead silicate glass bers with a tailored near-zerodispersion prole at telecommunication wavelengths. Investigatedber types include: (i) air-suspended core holey bers, (ii) holey -bers with complex holey cladding, (iii) all-solid 1Dmicrostructuredbers, and (iv) W-type step-index bers with high index-contrast.Various aspects relevant to the choice of ber structure and of theglass materials were discussed in detail, with particular emphasison the implications for the ber fabrication. The conclusions aresummarized in the following.

(i) In air-suspended core HFs, the maximum theoretical nonlin-earity c for a given glass can be achieved, as well as a ZDW at1.550 lm, although in general with the penalty of a veryhigh dispersion slope. The position of the ZDW is stronglydependent on the core diameter and also on the ne detailsof the core shape, indicating that, though not impossible, it ispractically very challenging to engineer a precise dispersionprole around 1550 nm with such a ber design.

(ii) In small core HFs based on a triangular lattice of holes, a at-

Table 3Summary of initial results on fusion splicing commercial silica ber with soft glass hi

A Fiber parameters B Fiber pa

Silica ber (UHNA3Nufern)

MFD: 3.3 lm at 1.31 lm; OD:125 lm

1D MOF SF6/LLF1 Core diaOD: 150

Silica ber (UHNA3Nufern)

MFD: 3.3 lm at 1.31 lm; OD:125 lm

W-type bre SF57/LLF1/SF6

Core diaAeff = 2OD: 115

X. Feng et al. / Optical Fibertened and near-zero dispersion prole can in principle beachieved in the 1.55 lm region. However, several rings ofholes, or complex double clad structures are required inorder to obtain low connement loss. Again this require-ment makes it very challenging to precisely control the holespacing K and the hole-to-spacing ratio d/K to the degree ofaccuracy required in order to achieve the target dispersionproles. We believe that this approach is more promisingthan the ASC HF, and is especially interesting for wavelengthranges where suitable pairs of compatible glasses cannot befound. Further improvements of the fabrication process arerequired in order to achieve improved control in bers witha complex structure.(iii) All-solid 1D MOFs with a coaxial-ring structure can achievesub-dB/m attenuation levels at 1550 nm, among the lowestever reported for lead silicate glasses bers, by utilizingnely polished glass discs and optimized fabrication condi-tions. In contrast to the small core HF, it is easy to accuratelypreserve the relative size of the high/low index features inthe all-solid MOFs during the various stages of fabrication,i.e. going from preform, to cane, and nally to ber. A dif-culty is represented by the fact that the initial aspect ratioof the glass discs is not faithfully reproduced in the extrudedpreform, and suitable correction factors are required.

(iv) High index-contrast, W-type bers provide the best compro-mise between ber nonlinearity and accurate control of berstructure amongst all the designs described here. We havedeveloped a fabrication technique that allows for an extre-mely precise (submicron) control of the ber structure.Consequently, control of chromatic dispersion to the ps/nm/km level can be achieved in this ber, and attenedand near-zero dispersion prole around 1550 nm have beendemonstrated. Reasonably low loss of 2.1 dB/m, high nonlin-earity of 820W1km1, and single-mode operation havebeen achieved. Thus, these bers offer an overall better per-formance as compared to all the other ber types, includingthe coaxial all-solid 1D MOFs. Numerical simulations predictthat bers with a few percent larger core diameter than theber reported here would achieve a attened and near-zerodispersion prole with around 200 nm bandwidth of anom-alous dispersion in the telecom window, opening up thepossibility for a host of applications. The fabricated 1DMOF and the W-type ber were used for FWM-basedfrequency conversion experiments at 1550 nm, indicatingthe great potential of non-silica highly nonlinear bers forapplications in all-optical processing at telecommunicationwavelengths.

(v) We have experimentally demonstrated that it is feasible tosplicing the 1D MOF or the W-type ber with commercialsilica ber, by the commonly used fusion splicing method.In our initial experiments, the splicing loss of 2.2 and 9 dBper point have been achieved, when respectively splicingthe 1D MOF and the W-type ber with the commercialhigh-NA silica ber. It is expected that the splicing loss canbe lower than 2 dB per point in the coming future.

nonlinear bers.

ters Splicing loss per point(A? B) at 1550 nm

Causes of splicing loss

lm, Aeff = 6.7 lm2; 2.2 dB Mainly due to imperfections atsplicing point formed by arc fusion

.7 lm,2;

9 dB Due to both MFD mismatch andimperfections at splicing pointformed by arc fusion

nology 16 (2010) 378391 389In comparison with the silica counterpart, i.e., the silica glasshighly nonlinear dispersion attening ber (HNL-DFF), a gure ofmerit (FOM), cLeff, which is the product of the ber nonlinearityc and the effective length Leff (Leff = [1exp(aL)]/a, where L isthe ber length, a is the ber attenuation respectively) is com-monly used to evaluate the performance of the nonlinear ber. ThisFOM depends on both the nonlinearity and the propagation loss ofthe used ber. A larger cLeff product means that in order to inducethe same nonlinear phase shift, a lower pump or shorter length ofthe ber is required. Thus it is deduced that our W-type ber with2 dB/m loss has the comparable nonlinear optical performance to200-m long commercial silica-based dispersion attening nonlin-ear bers [51]. Moreover, it is noted that for the FWM-based

Technonlinear applications, the stimulated Brillouin scattering (SBS)threshold should be a secondary concern for evaluating the nonlin-ear optical ber with attening and near-zero dispersion prole[32]. Note that even with very low attenuation of 1 dB/km [51],the usable length of a silica HN-DFF for the FWM-based applica-tions is limited by the deteriorative nonlinear effects like the SBSunder a certain pump power level. In the case of the W-type berwith the SF57 core, the Kerr effect induced optical nonlinearity c is70 times higher than that of a commercial silica HNL-DFF [51],but the threshold of the acousto-optic effect induced SBS is only3 times higher than that of the silica HNL-DFF [32]. This is oneof the key advantages of using short-length non-silica glassHNL-DFFs over its silica counterpart. Thus, in the long term, furtherlowering the propagation loss and the splicing loss of the non-silicaglass HNLFs and exploring different combinations of glasses offer-ing even higher nonlinearity will be our main targets in order to letthe non-silica glass HNLFs compete with the well developed, highperformance silica counterpart.

Acknowledgments

The research leading to these results has received funding inpart from the Engineering and Physics Science Research Council(UK) and the European Communities Seventh Framework FP/2007-2013 under Grant Agreement 224547. Dr. Francesco Polettigratefully acknowledges support from the Royal Society througha University Research Fellowship.

References

[1] J.M. Dudley, J.R. Taylor (Eds.), Supercontinuum Generation in Optical Fibers,Cambridge University Press, 2010.

[2] J. Hansryd, P.A. Andrekson, M. Westlund, J. Li, P.-O. Hedekvist, Fiber-basedoptical parametric ampliers and their applications, IEEE J. Select. Topics QE 8(2002) 506520.

[3] J.C. Knight, T.A. Birks, P.St.J. Russell, D.M. Atkin, All-silica single-mode brewith photonic crystal cladding, Opt. Lett. 21 (1996) 15471549.

[4] T.M. Monro, D.J. Richardson, N.G.R. Broderick, P.J. Bennett, Holey optical bres:an efcient modal model, J. Lightw. Technol. 17 (6) (1999) 10931102.

[5] T.M. Monro, Y.D. West, D.W. Hewak, N.G.R. Broderick, D.J.R. Richardson,Chalcogenide holey bres, Electron. Lett. 36 (2000) 19982000.

[6] X. Feng, A.K. Mairaj, D. Hewak, T.M. Monro, Nonsilica glasses for holey bers, J.Lightw. Technol. 23 (6) (2005) 20462054.

[7] H. Nasu, O. Sugimoto, J. Matsuoka, K. Kamiya, Non-resonant-type third-orderoptical nonlinearity of alkali silicate and alkali aluminosilicate glasses-contribution of individual chemical species in the glasses to v(3), J. Non-Cryst. Solids 182 (1995) 321327.

[8] I. Thomazeau, J. Etchepare, G. Grillon, A. Migus, Electronic nonlinear opticalsusceptibilities of silicate glasses, Opt. Lett. 10 (1985) 223225.

[9] R. Adair, L.L. Chase, R.A. Payne, Nonlinear refractive index of optical crystals,Phys. Rev. B 39 (1989) 33373350.

[10] D. Marcuse, Light Transmission Optics, Van Nostrand Reinhold, New York,1982.

[11] D.W. Pohl, P.F. Meier, Multiphonon absorption in NaF, Phys. Rev. Lett. 32(1974) 5861.

[12] I.H. Malitson, Interspecimen comparison of the refractive index of fused silica,J. Opt. Soc. Am. 55 (1965) 12051209.

[13] T. Miyashita, T. Manabe, Infrared optical bers, IEEE Trans. Microw. TheoryTech. MTT-30 (10) (1982) 14201438.

[14] Schott Optical Glass Catalogue Version 1.8 e, 2009, .

[15] G. Ghosh, Sellmeier coefcients and chromatic dispersions for some telluriteglasses, J. Am. Ceram. Soc. 78 (10) (1995) 28282830.

[16] W. Dumbaugh, J. Lapp, Heavy-metal oxide glasses, J. Am. Ceram. Soc. 75 (9)(1992) 23152326.

[17] S. Fujino, H. Ijiri, F. Shimizu, K. Morinaga, Measurement of viscosity of multi-component glasses in the wide range for bre drawing, J. Jpn. Instrum.Met. 62(1998) 106110.

[18] S.R. Friberg, P.W. Smith, Nonlinear optical glasses for ultrafast optical switches,IEEE J. Quant. Electron. QE-23 (1987) 20892234.

[19] D.N. Nikogosyan, Properties of Optical and Laser-related Materials: AHandbook, John Wiley & Sons, 1997.

[20] T.M. Monro, K.M. Kiang, J.H. Lee, K. Frampton, Z. Yusoff, R. Moore, J. Tucknott,

390 X. Feng et al. / Optical FiberD.W. Hewak, H.N. Rutt, D.J. Richardson, High nonlinearity extruded single-mode holey optical bers, in: OFC2002, OSA, Washington, DC, PostdeadlineFA1, 2002, pp. 13.[21] V.V.R.K. Kumar, A.K. George, W.H. Reeves, J.C. Knight, P.St.J. Russell, F.G.Omenetto, A.J. Taylor, Extruded soft glass photonic crystal ber for ultrabroadsupercontinuum generation, Opt. Express 10 (2002) 15201525.

[22] P. Petropoulos, H. Ebendorff-Heidepriem, V. Finazzi, R.C. Moore, K. Frampton,D.J. Richardson, T.M. Monro, Highly nonlinear and anomalously dispersive leadsilicate glass holey bers, Opt. Express 11 (2003) 35683573.

[23] H. Ebendorff-Heidepriem, P. Petropoulos, S. Asimakis, V. Finazzi, R.C. Moore, K.Frampton, F. Koizumi, D.J. Richardson, T.M. Monro, Bismuth glass holey berswith high nonlinearity, Opt. Express 12 (2004) 50825087.

[24] S. Afshar V, W.Q. Zhang, H. Ebendorff-Heidepriem, T.M. Monro, Small coreoptical waveguides are more nonlinear than expected: experimentalconrmation, Opt. Lett. 34 (2009) 35773579.

[25] J.Y.Y. Leong, P. Petropoulos, J.H.V. Price, H. Ebendorff-Heidepriem, S. Asimakis,R.C. Moore, K.E. Frampton, V. Finazzi, X. Feng, T.M. Monro, D.J. Richardson,High-nonlinearity dispersion shifted lead-silicate holey bers for efcient 1-lm pumped supercontinuum generation, J. Lightw. Technol. 24 (2006) 183190.

[26] H. Ebendorff-Heidepriem, S.C. Warren-Smith, T.M. Monro, Suspendednanowires: fabrication, design and characterization of bers with nanoscalecores, Opt. Express 17 (2009) 26462657.

[27] L. Dong, B.K. Thomas, L. Fu, Highly nonlinear silica suspended core bers, Opt.Express 16 (2008) 1642316430.

[28] B. Elliott, M. Gilmore, Fiber Optic Cabling, second ed., Newnes, 2002.[29] S. Asimakis, P. Petropoulos, F. Poletti, J.Y.Y. Leong, R.C. Moore, K.E. Frampton, X.

Feng, W.H. Loh, D.J. Richardson, Towards efcient and broadband four-wave-mixing using short-length dispersion tailored lead silicate holey bers, Opt.Express 15 (2007) 596601.

[30] K. Inoue, Experimental study on channel crosstalk due to ber four-wavemixing around the zero-dispersion wavelength, J. Lightw. Technol. 12 (6)(1994) 10231028.

[31] J.Y.Y. Leong, S. Asimakis, F. Poletti, P. Petropoulos, X. Feng, R.C. Moore, K.E.Frampton, T.M. Monro, H. Ebendorff-Heidepriem, W.H. Loh, D.J. Richardson,Towards zero dispersion highly nonlinear lead silicate glass holey bres at1550 nm by structured-element-stacking, in: ECOC 2005 Glasgow, 2529September, 2005, Th4.4.5 (Postdeadline).

[32] F. Poletti, K. Furusawa, Z. Yusoff, N.G.R. Broderick, D.J. Richardson, Nonlineartapered holey bers with high stimulated Brillouin scattering threshold andcontrolled dispersion, J. Opt. Soc. Am. B 24 (9) (2007) 21852194.

[33] F. Poletti, V. Finazzi, T.M. Monro, N.G.R. Broderick, V. Tse, D.J. Richardson,Inverse design and fabrication tolerances of ultra-attened dispersion holeybers, Opt. Express 13 (10) (2005) 37283736.

[34] H. Ebendorff-Heidepriem, T.M. Monro, Extrusion of complex preformsfor microstructured optical bers, Opt. Express 15 (23) (2007) 1508615092.

[35] X. Feng, A. Camerlingo, F. Poletti, P. Petropoulos, K.E. Frampton, N.M. White,C.J. Oton, W.H. Loh, D.J. Richardson, Highly nonlinear non-silica glassmicrostructured optical bers with near-zero dispersion slope for1.55 micron applications, in: CLEO/Europe-EQEC 2009, Munich, 1419 June,2009, CE3.4.

[36] A. Boskovic, S.V. Chernikov, J.R. Taylor, L. Gruner-Nielsen, O.A. Levring, Directcontinuouswave measurement of n2 in various types of telecommunicationber at 1.55 lm, Opt. Lett. 21 (1996) 19661968.

[37] X. Feng, T.M. Monro, P. Petropoulos, V. Finazzi, D. Hewak, Solidmicrostructured optical ber, Opt. Express 11 (18) (2003) 22252230.

[38] X. Feng, T.M. Monro, P. Petropoulos, V. Finazzi, D.J. Richardson, Extrudedsingle-mode high-index-core one-dimensional microstructured optical berwith high index-contrast for highly nonlinear optical devices, Appl. Phys. Lett.87 (2005) 81110.

[39] X. Feng, F. Poletti, A. Camerlingo, F. Parmigiani, P. Horak, P. Petropoulos, W.H.Loh, D.J. Richardson, Dispersion-shifted all-solid high index-contrastmicrostructured optical ber for nonlinear applications at 1.55 lm, Opt.Express 17 (22) (2009) 2024920255.

[40] T. Hasegawa, T. Nagashima, N. Sugimoto, Determination of nonlinearcoefcient and group-velocity dispersion of bismuth-based high nonlinearoptical ber by four-wave-mixing, Opt. Commun. 281 (2008) 782787.

[41] A.W. Snyder, J.D. Love, Optical Waveguide Theory, Chapman and Hall, London New York, 1983.

[42] X. Feng, G.M. Ponzo, F. Poletti, A. Camerlingo, F. Parmigiani, M.N. Petrovich, P.Petropoulos, N.M. White, W.H. Loh, D.J. Richardson, A single-mode, high index-contrast, lead silicate glass bre with high nonlinearity, broadband near-zerodispersion at telecommunication wavelengths, in: ECOC 2010, Torino, Italy,September 1923, 2010, Tu.4.D.2.

[43] T. Miya, K. Okamoto, Y. Ohmori, Y. Sasaki, Fabrication of low dispersion single-mode bers over a wide spectral range, IEEE J. Quant. Electron. QE-17 (6)(1981) 858861.

[44] H. Shang, Chromatic dispersion measurement by white-light interferometryon metre-length single-mode optical bres, Electron. Lett. 17 (17) (1981) 603605.

[45] A. Camerlingo, F. Parmigiani, X. Feng, F. Poletti, P. Horak, W.H. Loh, P.Petropoulos, D.J. Richardson, Multichannel wavelength conversion of 40 Gbit/sNRZ DPSK signals in a highly nonlinear dispersion attened lead silicate bre,in: Nonlinear Photonics 2010, Karlsruhe, Germany, 2124 June, 2010, NThA3.

[46] A. Camerlingo, F. Parmigiani, R. Slavik, X. Feng, F. Poletti, P. Horak, W.H. Loh, P.

nology 16 (2010) 378391Petropoulos, D.J. Richardson, Generation of ultra-high repetition rate pulses ina highly nonlinear dispersion-tailored compound glass bre, in: IEEE SummerTopical 2010, Cancun, Mexico, 1921 July, 2010, WC4.

Giorgio M. Ponzo was born in Tourin, Italy. He received a B.Sc. degree (2006) and aM.Sc. degree (2008) with honors in electronic engineering from the University ofPalermo, Italy. He worked on the development of a DC/DC converter for VRMapplications at ST Microelectronics (STM), Catania, Italy, for 4 months during his

X. Feng et al. / Optical Fiber Technology 16 (2010) 378391 391[47] A. Mori, H. Masuda, K. Shikano, M. Shimizu, Ultra-wide-band tellurite-basedber Raman amplier, J. Lightw. Technol. 21 (5) (2003) 13001306.

[48] A.D. Yablon, Optical Fiber Fusion Splicing, rst ed., Springer, 2005.[49] S. Jiang, J. Wang, Method of fusion splicing silica ber with low-temperature

multi-component glass ber, US Patent 6705771.[50] N. Sugimoto, Erbium doped ber and highly non-linear ber based on bismuth

oxide glasses, J. Non-Cryst. Solids 354 (2008) 12051210.[51] D.J. Richardson, X. Feng, F. Poletti, S. Dasgupta, A. Camerlingo, F. Parmigiani, P.

Petropoulos, W.H. Loh, S. Herstrom, L. Gruner-Nielsen, Recent advances inhighly nonlinear bres, in: ECOC 2010, Turin, 1923 September, 2010, Tu.4.D.1(invited).

Xian Feng was born in Hangzhou, China. He received his Doctor of Engineering,majoring in Material Science, from Shanghai Institute of Optics & Fine Mechanics,Chinese Academy of Sciences in 1999. He joined the soft-glass group of Optoelec-tronics Research Centre, University of Southampton in 2001, after the experience asthe post-doctoral researcher in Kyoto University, Japan and Rutgers, The StateUniversity of New Jersey. His current research interests include the fabrication andthe applications of microstructured specialty optical bers based on non-silicaoptical glasses.

Francesco Poletti graduated with honours in Electronics Engineering at the Uni-versity of Parma (Italy) in 2000. From 2000 to 2003 he worked on the design ofoptical telecoms networks at Marconi Communications (Italy and UK), and in 2007he obtained a PhD in optoelectronics from the Optoelectronics Research Centre,University of Southampton (UK) for research on the direct and inverse design ofmicrostructured optical bers. His research, supported by a Royal Society UniversityResearch Fellowship, currently focuses on the study of micro and nano-structuredoptical waveguides, on the modeling of photonic bandgap structures, nonlinearoptical processes in gas, liquid and semiconductor lled holey bers, and on inverseelectromagnetic problems.

Angela Camerlingo was born in Naples, Italy. She graduated with honours inElectronics Engineering at the Second University of Naples in 2005. From January2006 to July 2006 she worked at the Department of Electrical Engineering andInformation Technology, Second University of Naples. Her research included thedevelopment of optical-ber based pressure sensors. In October 2006 she joined theOptoelectronics Research Centre, University of Southampton, (UK), where she iscurrently working towards her Ph.D. degree in optical communication systems. Hercurrent research interests include the development of optical ber technologies andtheir applications for highly nonlinear systems.

Francesca Parmigiani was born in Milan, Italy. She graduated with honours inElectronic Engineering at Politecnico di Milano, Milano, Italy, in 2002, and receivedthe Ph.D. degree in optical communication systems at the Optoelectronics ResearchCentre (ORC), University of Southampton, Southampton, UK in 2006. She is cur-rently a Research Fellow at the ORC. In April 2010, she was awarded a prestigiousPostdoctoral Research Fellowships from the Royal Academy of Engineering, insupport of her research on the combination of all-optical signal processing andadvanced modulation formats. Her research interests include ultra-fast all-opticalsampling techniques, pulse shaping using specialized ber Bragg gratings, all-optical nonlinear processing and switches mainly in optical bers, as well asadvanced modulation formats. Dr. Parmigiani is a member of the Optical Society ofAmerica (OSA).

Periklis Petropoulos was born in Patras, Greece. He graduated from the Depart-ment of Electrical Engineering and Information Technology, University of Patras, in1995, received the M.Sc. degree in communications engineering from the Universityof Manchester Institute of Science and Technology, Manchester, UK, in 1996, andthe Ph.D. degree in optical telecommunications from the Optoelectronics ResearchCentre (ORC), University of Southampton, Southampton, UK in 2000. His researchinterests include all-optical processing and switching in optical bers; pulsemanipulation for optical communications using ber Bragg gratings, includingapplications in optical correlation systems for the implementation of optical codedivision multiple access and optical packet switched systems; silica and compoundglass holey bers and their nonlinear applications; and ber lasers. His research hasproduced more than 240 papers in journals and conferences in the eld of opticalphysics and optical communications. He is currently a Reader at the ORC.

Peter Horak obtained a M.Sc. degree in theoretical physics in 1993 and a Ph.D.degree in theoretical quantum optics in 1997 from the University of Innsbruck,Austria. He held research positions at the University of Innsbruck, at the EcoleNormale Superieure in Paris, France, and at the University of Strathclyde in Glas-gow, UK, before joining the ORC in 2001 where he is now a Senior Research Fellow.He heads the Computational Nonlinear Optics group at the ORC, focusing on theoryand modeling of nonlinear and quantum optical systems. He currently holds aResearch Council UK Academic Fellowship.B.Sc. degree thesis, and on the development of an innovative DC/DC converter fortelecom applications, at Texas Instrument (TI), Freising, Germany, for 6 monthsduring his M.Sc. degree thesis. In 2008, he held a research position at University ofPalermo working on high efcient DC/DC conversion. From the end of 2009 he is aPh.D. student at the Optoelectronics Research Centre (ORC), Southampton, UK. Hisresearch currently focuses on the development and characterization of micro-structured optical bers for several applications.

Marco Petrovich received a Laurea degree in experimental Solid State Physicsfrom the University of Padua, Italy in 1998 and a Ph.D. in Optoelectronics from theUniversity of Southampton in 2003, for work on the fabrication of chalcogenidebers based on gallium lanthanum sulphide glass. In 1999, he was a researchassistant at the INFM MDM Labs working on characterisation thin insulator andconductive layers for MOS technology. He currently holds a position as SeniorResearch Fellow at the ORC, which he joined as a research staff in 2003. His researchinterests encompass ber fabrication technologies, the development of novel silica,non silica and composite microstructured bers, and the application of such bersfor nonlinear optics, novel laser sources, gas sensing and spectroscopy. He hasauthored or co-authored over 80 journal and conference papers. Dr. Petrovich is amember of the Optical Society of America.

Jindan Shi was born in Lanxi, Zhejiang Province, China. She received a B.Sc. degreewith honors in telecommunication engineering from Beijing University of Post andTelecommunications (BUPT), China, in 2006. From October 2008, she is a Ph.D.student at the Optoelectronics Research Centre (ORC), Southampton, UK. Herresearch currently focuses on periodic ber devices for advanced applications in alloptical systems.

Wei H. Loh obtained his PhD in Electrical Engineering from Cornell University, NY.From 1991 to 1997, he was with the Optoelectronics Research Centre at Universityof Southampton. As a Senior Research Fellow working on application-specic ber-optic devices, he made pioneering contributions to the design and fabrication ofber Bragg gratings, dispersion compensation, semiconductor saturable absorbersfor short pulse ber lasers and soliton transmission, rare earth-doped ber laserdynamics, distributed feedback ber lasers and ber laser phase noise. From 1997to 2003, he worked in industry with various companies. From 1997 to 1998, he waswith QPS Technology/Bragg Photonics (Montreal), where he oversaw the R&Dprogram for the commercialisation of ber Bragg gratings and grating array sensorsfor both the telecommunications and oil sensing industries. At the end of 1998, hejoined E-Tek Dynamics (San Jose), which then merged with JDS Uniphase to becomethe largest optical component vendor in the global marketplace. As Director forNew Product Development at JDSU, he oversaw the new product developmentprogramme at its Silicon Valley site in San Jose, including the formulation oftechnology roadmaps and the controlled introduction of new products to themarket. The range of products developed included passive optical components (freespace, ber as well as planar waveguides), switches/attenuators, network monitors,and ampliers. He has been a Principal Research Fellow with the ORC since 2004.He has over 100 refereed journal and conference publications, and 8 patents.

David J. Richardson was born in Southampton, UK, in 1964. He received the B.Sc.degree and the Ph.D. degree in fundamental physics from Sussex University, UK, in1985 and 1989, respectively. In May 1989, he was a Research Fellow at the thenrecently formed Optoelectronics Research Centre (ORC), Southampton University,UK. He is currently a Deputy Director at the ORC, responsible for much of the ORCsber-related activities. He is one of the founders of Southampton Photonics Incor-porated, a university spin-off venture that has successfully commercialized ele-ments of high-power laser technology developed within the ORC. His currentresearch interests include amongst others: microstructured bers, high-power berlasers, short pulse lasers, optical ber communications, and nonlinear ber optics.Prof Richardson has published more than 500 conference and journal papers in histime at the ORC, and produced over 20 patents. He is a frequent invited speaker atthe leading international optics conferences in the optical communications, laserand nonlinear optics elds and is an active member of both the national andinternational optics communities. He was awarded a Royal Society University Fel-lowship in 1991 in recognition of his pioneering work on short-pulsed ber lasers,and was made a Fellow of the Optical Society of America in 2005. He was elected asa Fellow of the Royal Academy of Engineering in recognition for his contributions toengineering in 2009.

Dispersion controlled highly nonlinear fibers for all-optical processing at telecoms wavelengthsIntroductionStructure and material aspects for dispersion-tailored highly nonlinear fibersFiber structuresFiber materialsUnderstanding refractive index and material dispersion of optical glassProperties of selected commercial high-index lead silicate glasses

Fiber design, fabrication, and characterizationAir-suspended core holey fiberHoley fiber with complex microstructured claddingAll-solid 1D microstructured fiberW-type step-index profiled fiber with high index contrastAsymmetric configuration for fusion splicing silica fiber and soft glass HNLFs

ConclusionAcknowledgmentsReferences