Introduction

Methods have been devised to modify the optical and/or acoustic properties of an optical fiber at points along its length either permanently (e.g., by modifying fiber diameter during the draw [1] or by tapering) or transiently (e.g., using temperature [2] or strain [3]). Such perturbations influence the modal or propagation characteristics of the optical (or acoustic) field in the fiber and, consequently, have been used to control nonlinearities [4], including chromatic dispersion and soliton propagation [1], and suppress stimulated Brillouin scattering (SBS [5]) and four-wave mixing (FWM [6]), to name just a few applications. For completeness, it is worth noting that the benefits to suppressing Brillouin oscillations in fibers using compositional gradients had been conjectured over 30 years ago [7].

Fibers possessing longitudinal changes in composition previously have been realized [8,9] using vapor-axial deposition (VAD). However, since the composition was changed along the length of the entire preform, the resultant longitudinal compositional gradient in the drawn fiber was very low, just 0.4% over 28 km of length. While such a length is acceptable for long-haul telecommunication applications, many other applications exist where shorter lengths of fibers possessing higher longitudinal gradients are of interest.

In this work, a simple method is employed in order to controllably shape the compositional profile of the core along a short length of optical fiber; presently about 10 – 20 m, but conceivably longer or shorter depending on the gradient and preform design. By controlling the properties of the fiber along its length through the core glass composition, rather than dimension, strain, or temperature, a new family of property-enhanced optical fibers is realized.

Experimental

A depiction of the general process employed here is shown in Fig. 1 and follows a scheme described by Rice, et al. [10]. Conceptually, a radial index profile is generated in a silica preform using conventional chemical vapor deposition methods. This radial profile is transformed into a longitudinal profile by core-drilling a rod through the side of the preform. This rod, which contains a portion of the core as well as two end sections of the cladding glass (see Fig. 1(b)), is then sleeved inside a lower index tube which now acts as the preform for the subsequent drawing of the longitudinally-graded optical fiber.

 

Fig. 1 Idealized representation of the process employed. (a) Conventional GeO₂-doped SiO₂ preform fabricated with specific radial refractive index profile using a chemical vapor deposition process; (b) a rod is core-drilled out through the side of the preform such that radial gradient of the preform becomes a longitudinal gradient in the rod; (c) rod from (b) is sleeved into a lower refractive index inner cladding tube such that an index-guiding core/clad geometry is achieved; (d) preform from (c) is drawn into fiber such that longitudinal refractive index profile is now present in the optical fiber. Also shown in (d) are the idealized longitudinal refractive index and compositional profiles of the fiber; which are correlated and are defined by the initial radial profile of the preform in (a). The vertical green dotted lines in (d) are guides to the eye.

Download Full Size | PDF

More specifically, a doped SiO₂ preform was fabricated using an SG Controls modified chemical vapor deposition (MCVD) lathe at Clemson University. The preform was fabricated from a pure silica tube that was about 450 mm in length and had an inner and outer diameter of 17 and 21 mm, respectively. The core region consisted of four deposition layers: the first three were doped with germanium from a GeCl₄ vapor source. The GeCl₄ flow rate was increased by 2 standard cubic centimeters per minute (sccm) per layer from about 30 sccm for the first layer to about 34 sccm for the third layer. A constant flow of SiCl₄ at about 60 sccm was maintained throughout the first three layers. The fourth layer was co-doped with GeCl₄ at about 100 sccm and SiF₄ at about 70 sccm. The use of fluorine-doping in this last core layer was to investigate whether an additional dopant could measurably influence either the longitudinal optical or acoustic properties of the resultant fiber. During the preform collapse stage, a low flow of SiF₄ was maintained in order to lessen fluorine losses. For completeness, it is noted that the fluorine is too light of an element to be measured using the energy dispersive x-ray analytical methods described below. Accordingly, the fluorine is not discussed with respect to the chemistry of the core but could influence both the optical and acoustic behavior. However, since the fluorine level is fairly low, based on the SiF₄ flow rates, its impact on the performance of these proof-of-concept fibers is not expected to be large compared to that of the GeO₂ doping. The radial index profile of the resultant preform was analyzed using a Photon Kinetics PK2600.

Rods comprising the 1.7 mm diameter core then were transversely-drilled through the full diameter of the preform (Ceramare, Piscataway, NJ) at the same locations as index profiled on the PK2600. This side-core-drilled slug was then sleeved into a fluorinated silica cladding tube. This fluorinated silica case was then, in turn, sleeved inside a 10mm inner diameter by 26 mm outer diameter F300 pure silica tube (Heraeus, Buford, GA). The fluorinated silica inner cladding tube provides a lower refractive index relative to the longitudinally-varying core slug to enable light-guiding in the resultant fiber. The outer pure silica cladding tube, though not necessary in theory, was employed since fluorinated silica tubes with the desired core/clad ratio were not available. Pure silica rods, whose outer diameter matched well the inner diameter of the fluorinated silica inner cladding tube, were placed above and below the doped-SiO₂ core slugs to prevent the regions of interest from being lost during the initial stages of drawing. The entire billet was consolidated at 2300C on the lathe under a vacuum of 1.5 torr in order to limit motion of the individual pieces of glass comprising the preform during fiber fabrication.

The side-core-drilled slug was sleeved inside a F320 cladding tube (Heraeus, Buford, GA), which had a refractive index of 0.001 index units below that of silica. The fully-consolidated preform was drawn at 1950C on a Heathway optical fiber draw tower (Clemson University) into a 1700 m length of 125 μm diameter fiber. The fibers were coated with a standard single coating (Desolite 3471-3-14, DSM Desotech Inc., Elgin, IL) to a final outer diameter of about 235μm.

Electron microscopy of selected preform and fiber samples was performed using a Hitachi S-3400 scanning electron microscope (SEM) operating at 20kV under variable pressure. Elemental analysis was conducted using a Hitachi TM-3000 tabletop SEM operating at 15kV. Spectral attenuation measurements were conducted using a cut-back method on a Photon Kinetics PK2500 (Beaverton, OR). The refractive index profile at arbitrary position along the length of the fiber was measured at 980 nm by Interfiber Analysis Inc. using a spatially resolved Fourier transform technique [11]. The experimental apparatus used to measure the Brillouin gain spectrum (BGS) was a heterodyne system [12] operating at 1534 nm similar to that described in Ref [13], with a standard single mode fiber (SMF-28^TM, Corning Incorporated) used as a control standard. A characteristic signature from the circulator fiber is typically observed in the measured spectra.

Results and discussion

The correlation between the radial refractive index profile from the original preform and the average germania concentration along the length of the longitudinally-varying optical fiber is shown in Fig. 2 . There are several different longitudinal gradients in this particular fiber as a function of length. Emphasized in Fig. 2 are two such gradients where the measured GeO₂ content changes by 5.46% or 2.45% over a 10 meter length yielding gradients of 0.546%/m and 0.245%/m, respectively. These values are higher by a factor of about 1900 times the gradients achieved in Ref. 8 by virtue of the fact that the compositional gradient is built into a smaller region of the preform and not the entire rod.

 

Fig. 2 Refractive index profile of the as-made MCVD preform at the position where the core slug was side-drilled out (dashed red line) and the average germania [GeO₂] concentration in the core measured at a variety of positions along the length of the as-drawn fiber (solid blue line; specific data points shown as diamonds). Note that the germanium content along the length of fiber follows the refractive index of the as-made preform. The circles and arrows denote the corresponding ordinate and abscissa for each curve. Also provided, in the shaded areas, are examples of length-wise GeO₂ gradients in the as-drawn fibers of about 0.55 and 0.25 weight % GeO₂/meter.

Download Full Size | PDF

The refractive index profile of the initial as-made preform (prior to side-core drilling) and also the germania (GeO₂) concentration as a function of position along the fiber also is shown in Fig. 2. The central dip in the index profile of the original preform results from burn-out of the germania and fluorine. The use of fluorine during the collapse of the original preform leads to the negative index values, relative to pure silica, that are observed. Additionally, while the GeO₂ level in this fiber is higher than that in SMF-28^TM (~8.4 versus 6.7 weight percent), the fluorine co-doping yields an index difference that is essentially equivalent ( + 0.005 relative to pure silica).

Figure 3 provides the refractive index profiles, relative to silica and measured at 970 nm, at two different locations over an arbitrarily chosen 20 meter length of the LGF. As can be seen, the index difference changes by about 0.001, representing a numerical aperture change of 0.013 (~13%), over a distance of only 16.6 m verifying that the fiber indeed possesses a gradient in refractive index along its length. The numerical aperture, NA, governs the modal properties of the fiber. For the fibers developed in this work, minimum NA is defined by the fluorinated silica inner cladding and the pure silica ends of the side-core-drilled slug. Accordingly, the NA is 0.054, for these pure silica core lengths of fibers drawn using the F320 inner cladding tube. The maximum NA then would be defined by the fluorinated silica inner cladding relative to the refractive index associated with the maximum germanium content portion of the longitudinally-varying optical fiber. Accordingly, for this specific fiber, the maximal change in numerical aperture is 0.067 (0.120 maximum versus 0.054 minimum).

 

Fig. 3 Refractive index profiles at 0.3 and 16.9 m positions along a 20 m length of the longitudinally-graded optical fibers. Profiles were taken at a wavelength of 970 nm.

Download Full Size | PDF

The lowest loss of the LGF was 82 dB/km at a wavelength of 1550 nm (Fig. 4 ). Thus there would be less than 1 dB of loss over a 10 m segment where the germania gradient is greatest. A very large hydroxyl ion absorption peak was observed at 1380 nm indicating water contamination, likely due to the wet-cutting and grinding of the core-drilled slugs that then were re-sleeved into the fluorinated cladding tube. Regardless of source at this proof-of-concept level, the measured losses are dominated by extrinsic factors since the spectral attenuation of the as-prepared (prior-to-side-core drilling) preform had a minimum value of about 23 dB/km at a wavelength of 1550 nm. The origin of the fairly high loss in the as-prepared preform is unknown as (unpublished) next generation analogs exhibit losses below 5 dB/km away from OH and active dopant absorption peaks.

 

Fig. 4 Spectral attenuation of the longitudinally-graded optical fiber and the as-made original MCVD preform. The minimum loss of the longitudinally-graded fiber was about 82 dB/km where as for the original preform, the minimum loss was about 23 dB/km.

Download Full Size | PDF

Since material composition influences both the optical and acoustic properties of the fiber, Brillouin gain spectra (BGS) were measured from both ends of the same length of the LGF (ends arbitrarily designated as ‘A’ and ‘B’) and compared to that from a conventional (Corning SMF-28^TM) optical fiber. The results are shown in Fig. 5 . The LGF spectra are normalized roughly to the highest-frequency peak, with reasoning to be discussed later. The multiple peaks observed in the LGF are believed to be modes of the acoustic waveguide and the peak near 11 GHz is a contribution by the measurement apparatus.

 

Fig. 5 Stimulated Brillouin scattering spectrum of the longitudinally-graded fiber, interrogated from both ends (arbitrarily A and B), and a conventional single mode fiber (Corning SMF-28^TM) measured at a wavelength of 1534nm.

Download Full Size | PDF

Due to the background loss at the test wavelength (~80 dB/km), there is slightly less pump power to drive Brillouin scattering from the fiber end opposite (far-end) the optical launch end (near-end). The back-scattered Stokes’ signal from the far-end also experiences more total loss than the signal scattered from the near-end. Thus, the Stokes’ signal from the far-end contributes less to the measured spectrum than that from the near-end, which can be used to explain the observed differences in the Side A and Side B measurements. Looking just at the L₀₁ mode (~10.71 GHz), since there is relatively more signal at lower frequencies for Side A than Side B, it can be concluded that Side A has more GeO₂ than Side B. The structure in the spectrum probably results from a non-linear length-wise change in germania content in the test segment.

The measured spectral width (FWHM) of the SMF-28^TM fiber is about 29 MHz. In contrast, the spectral width (FWHM) of the LGF is approximately 80 MHz, representing a broadening by about 50 MHz, or increase of about 4.4 dB. In order to compare the Brillouin gain coefficients, the acousto-optic overlap integral (i.e., the relative strength of higher-order acoustic modes, or HOAMs) must be taken into account. There is only one weak HOAM (L₀₂) observed for the SMF-28^TM, whereas there is a significant presence of HOAMs in the LGF. Coarsely integrating the spectra such that the total integrated Brillouin gain is conserved [14], not including the contribution by the apparatus, the LGF has a Brillouin gain coefficient about 6.7 dB below that of the standard commercial single-mode longitudinally-invariant optical fiber. Modeling results indicate that >10 dB suppression of SBS is viable with a compositional gradient corresponding to [GeO₂, F] = [6.33, 0] and [10.0, 0.61] weight percent at two ends of a linearly-graded optical fiber. The fabrication and characterization of this next generation LGF is underway and will be reported separately.

The calculated acoustic mode frequencies, 10.72 GHz, 10.92 GHz, and 11.20 GHz, are in reasonable, though not excellent, agreement with measured data. The presence of these modes is most likely due to perturbations from core circularity arising from the core-drilling process, and not from an intrinsic refractive-index dip usually associated with Ge-doped fibers. The measured mode residing near 11.3 GHz is assigned to the cladding interaction with the tails of the optical mode, and is well-predicted by the theory. The LGF spectra have therefore been normalized to this peak for comparison since its strength is invariant of the launch end. The reasonable agreement provides confidence in the aforementioned assignment of the measured peaks to be the L_0m guided acoustic modes of the waveguide. The HOAMs are found to have a greater frequency separation as the width of the waveguide is reduced.

As noted above the results on these initial fibers that should be considered, at best, proof of concept. As with any nascent effort much work remains towards achieving higher performance and complexity. The purpose of this section is to offer recommendations as to next steps as well as potential applications that could gain benefit from the findings of this work.

Even though a 20 meter length of the LGF would only impose an added loss of < 2 dB, further reduction in attenuation is warranted. The measured losses in the LGF (Fig. 4), with respect to the as-fabricated initial preform, likely result principally from the core-drilling and rod/tube stacking process. That said, as an expedient, the core-drilled slugs were not subsequently mechanically- or flame-polished which would improve surface quality and likely reduce loss. It is worth noting that a wide variety of specialty optical fibers used today employ core-drilling and stacking processes (e.g., boron stress rods in polarization-maintaining fibers, rod/tube stacking in photonic crystal and microstructured optical fiber) and so these general processes are amenable to higher quality, lower loss fiber.

Additionally, it should be possible to both increase the gradient and change its longitudinal shape, both by enhanced doping and preform design. The present fiber exhibits a maximum germania gradient of 0.546%/m. Given the generalized process, higher doping levels within smaller core size initial preforms, sleeved inside narrower wall-thickness cladding tubes, would be effective in enhancing the gradient. Very high germania-content fibers have been fabricated using similar MCVD processes [15] and so significant opportunities exist for greater doping levels; hence greater longitudinal gradients (i.e., shorter LGF lengths). Additionally, multiple side-core-drilled slugs can be sequentially stacked atop one another in the secondary preform to permit one draw to yield multiple LGF sections. Further, if scaled to preform dimensions typical of commercial preforms (> 300 mm [16]), then the same process would yield nearly 1 km of LGF. Even at the 10 meter lengths where the greatest gradients are realized in these initial fibers, a variety of novel applications abound for the LGFs and are discussed here forth.

The longitudinal refractive index of the LGF could significantly impact the efficiency of processes such as four wave mixing (FWM). In high power fiber amplifier arrays where multiple signal frequencies are simultaneously operating, FWM has been observed to yield a parasitic degradation of mutual amplifier coherence. This has been a particularly significant issue for passively phase-locked fiber amplifier arrays [17], which has been partially mitigated by a counter-pumping strategy. The strong longitudinal gradients in refractive index, and expected changes in modal propagation constants, demonstrated in this work, can greatly reduce the coherence length for processes like FWM that are strongly impacted by phase-matching of the interacting waves. While a complete analysis of this relatively complex phenomenon has not yet been conducted, large longitudinal phase-mismatch should be possible for narrowly-spaced optical frequencies in high power amplifiers employing LGFs. The extensive treatments of FWM suppression in telecom fibers [6] can be adapted for modeling high power fiber amplifiers. Such high power amplifiers, which are typically about 10 meters long, are well within the lengths of the initial LGFs reported here.

Optical fibers whose dimensions have been modified by means of a tapering process are useful for modifying or otherwise controlling numerical aperture and dispersion and have been used for supercontinuum generation among other all-fiber devices [¹8-20]. In the LGFs, the numerical aperture can be tailored as a function of length along the fiber (e.g., Fig. 3) and, as such, the modal characteristics of the fiber can be influenced without changing the core size. Interestingly, changes to dispersion are analogous to gain and loss to solitons [21] and so these LGFs might be useful for pulse shaping in the nonlinear regime, which require lengths of tens of meters.

Fiber Bragg gratings (FBGs) are useful in-fiber devices [22] that reflect wavelengths, λ, that meet the Bragg condition of λ = 2 ∙ n∙ Λ, where Λ is the grating period and n is the effective index of the optical mode. Chirped fiber Bragg gratings (CFBGs), which have been employed for dispersion compensation [23], exhibit grating periods that change along the fiber length, z; i.e., Λ = Λ(z). The LGFs treated here, where n = n(z), adds an extra degree-of-freedom in designing chirped fiber Bragg gratings. For example, a CFBG can be produced using a LGF where Λ is constant, which generally is a simpler proposition from a manufacturing perspective.

In another example, typical CFBGs are a few tens-of-centimeters long which sets delays to be between 10 – 100 ps. With an longitudinal refractive index gradient and constant Λ, manufacturing CFBGs on the order of meters, lengths achieved in this work, becomes feasible. This longer fiber length increases the amount of dispersion or delay that can be imparted on a signal to about a nanosecond.

Conclusions

A new simple and versatile method for fabricating optical fibers with a longitudinal composition gradient was developed. MCVD-derived germanosilicate fibers were fabricated with a gradient of up to about 0.55 weight % GeO₂ per meter and refractive index difference of about 0.001 over lengths of less than 20 m. The spectral attenuation was 82 dB/km at a wavelength of 1550nm, which resulted from extrinsic factors and should be diminished with continued optimization. The MCVD-derived germanosilicate LGFs exhibited a Brillouin spectral width broadened by about 3 MHz/m relative to industry standard fibers though broadening as large as 52 MHz/m is possible with further reduction in fiber attenuation. These gradients enable the possibility of large-scale SBS suppression relative to conventional fibers in fiber lengths < 10 m, suitable for fiber laser applications. The measured BGS exhibited a 4.4 dB broadening, relative to a standard single-mode fiber, over a 17-meter length of fiber. Simulated BGS on an idealized, yet manufacturably-reasonable design, disclose that greater than 10dB reduction in SBS gain is possible over short (< 20 m) lengths, and forms the basis for the next generation of fibers based on this novel manufacturing technique. More generally, these novel LGFs show significant promise for SBS suppression in high energy laser systems as well as a range of other novel applications including constant diameter “tapers” and constant grating period chirped Bragg gratings.