1. Introduction

Microstructured optical fibers (MOFs) have found extensive use over the last decade in a range of sensing applications [1, 2], in part at least due to the increased evanescent overlap available with MOFs [3] compared to alternative sensing geometries such as D-fibers [4]. The large parameter space over which the fiber geometry can be tailored to allow for the fibers to be optimized for sensing has allowed for a variety of designs to find successful use [5, 6]. MOFs can be employed as sensors using a variety of techniques, included fluorescence [7, 8], absorption measurements [9], gratings [10, 11] or whispering gallery modes [12].

Initial work on evanescent sensing using MOFs was limited to filling the transverse holes along the fiber [13, 14], however the potential interaction length using this technique is limited by the sizes of the holes, and the corresponding time to fill these holes via capillary action or diffusion [15]. More recently exposed core fibers (ECFs) have been demonstrated as a potential sensing platform that shows considerable advantages over standard MOFs. In ECFs one of the holes is opened to the surrounding environment, such that the fiber core interacts directly with its environment, rather than relying on liquids or gases that have been passed through the fiber from the end face.

The first demonstration of an ECF fabricated during the preform stage was made from polymer by Cox et al. [16], and later from lead silicate glass by Warren-smith et al. [15]. The high fabricated loss of these fibers (2.2 dB/m at 900 nm in F2 [15]) somewhat restricts their practical use for distributed measurements, as the potential interaction length is limited by the high attenuation, and their guidance at shorter wavelengths is relatively poor.

Further work in this field lead to demonstrations of ECFs from silica glass [10, 17], with losses ranging from 1.1 dB/m at 900 nm for a 10 µm core [17] to 0.18 dB/m for a 7.5 µm core [18], opening up possibilities for using these fibers for long-distance distributed measurements. However, while the loss of fibers fabricated from silica was low compared to the initial demonstrations in polymer and lead silicate glass, the core size of these fibers is significantly larger than what has been demonstrated with suspended core fibers (SCFs) [19].

For evanescent sensing applications the overlap of the guided mode with the holes is an important parameter to consider [3]. For a silica:air interface at 532 nm the fraction of power within the holes of the fiber is 2.2 × 10⁻² for a 1.5 µm core fiber, with this value dropping significantly to 1.1 × 10⁻⁴ for a 10 µm core fiber. Core sizes as small as 1.7 µm have been demonstrated previously in silica ECFs [20], however the loss of these fibers is prohibitively high, with applications to date utilizing very short lengths of fiber [21].

In this work we show the first demonstration of low-loss, small-core ECFs for potential sensing applications. Through a combination of modified fabrication techniques, and improvements to the fiber geometry we are able to demonstrate that these fibers can be successfully drawn with cores as small as 1.6 µm, with a loss comparable to that demonstrated previously in significantly larger core fibers.

2. Experimental method

2.1 Fiber fabrication

Fabrication of the small core ECFs was based on existing techniques described in detail in previous work [17, 19]. 120 mm lengths of F300 fused silica (Heraeus Quartzglass) rods were drilled using an ultrasonic mill to create the three holes (Fig. 1(a)). A slot was cut along the length of the preform to expose one side of the core of the fiber (Fig. 1(b)). The preform was cleaned in nitric acid overnight to remove drilling contaminants, washed with acetone and dried using nitrogen. This initial preform was then caned using a 6 m drawing tower to form the inner structure of the final preform (Fig. 1(c) and 1(d)) before being pulled to fibre (Fig. 1(e)).

 

Fig. 1 Fabrication method for small core ECF a) The preform is drilled on an ultrasonic mill (b) A slot is cut along the length of the preform (c,d) The preform is caned, and inserted into a jacket tube (e) Jacket/cane is drawn to fiber.

Download Full Size | PDF

The initially fabricated preform had an outer diameter of 12 mm, with a drilled hole diameter of 2.8 mm, and a spacing between the holes of 0.4 mm, similar to that used in [17], which was then caned to a diameter of 1-2 mm. A second rod was drilled with an inner hole diameter closely matching that of the cane, and a slot cut along its length using an identical method to that used on the initial preform to form the jacket.

The cane was then inserted into the jacket and drawn to fiber, with a scanning electron microscope (SEM) image of the resulting fiber (Fiber #1) shown in Fig. 2(a). Active pressurization was used during the draw on the central cane piece.

 

Fig. 2 Small core ECFs, showing (a) Fiber #1, with thin struts and larger holes (b) Fiber #2, with thicker struts and thus more robust geometry. Main scale bars – a) 100 µm b) 50 µm and 5 µm for insets.

Download Full Size | PDF

Fibers #2 and #3 were fabricated using a similar technique to that used for Fiber #1, with the spacing between the holes increased from 0.4 mm to 0.7 mm, and for Fiber 3 an alignment tool was utilized to ensure correct orientation of the cane within the outer jacket (Fig. 1(d)).

2.2 Characterization

Broadband characterization measurements were performed using a Koheras SuperK Extreme supercontinuum, coupled into the fiber using a 60x objective. An Ando AQ6315E optical spectrum analyzer was used to examine the output light, with loss measurements obtained using a standard cutback technique at a series of locations along the fibre length.

3. Results

3.1 Experimental

The initially fabricated fiber (Fiber #1) had a core size of 2.6 µm, and an outer diameter of 222 µm. Preliminary testing showed that the fiber guided light, however when coupling visible light into the core a large number of scattering sites were observed along the length of the fiber. Measurements across a range of fibers showed that in some cases the scattering was sufficient that no light propagated to the output end, even when using short (<50 cm) lengths.

Two possible explanations for this high observed loss are proposed. The first possibility is contamination, most likely of the cane. The exposed core fiber is more prone to contamination, as the entire core structure is open to the external environment. The standard preform cleaning process after mechanical milling involves cleaning in 70% Nitric acid overnight, and rinsing with HPLC grade acetone. This process is also applied here, which removes contaminants resulting from the initial preform milling phase. However, after the preform is caned cleaning becomes challenging. Previous experience demonstrated that cleaning the cane at this stage introduces additional complications, as it is difficult to dry a one-meter length of cane, leaving the risk of liquid or residue from cleaning within the cane structure. Following initial results with Fiber #1 showing a high loss, care was taken with subsequent draws to minimize exposure of the cane to dust and other contaminants by enclosing them within a protective sheath at all times in preference to physical cleaning of the cane. Fiber drums were also rigorously cleaned, and immediately covered in plastic after the fiber had been spooled.

Note that the preform and cane were not etched, as literature shows that this process can introduce additional surface roughness that could significantly degrade the loss [22].

The second potential source of the observed high loss was the fragility of the struts, with SEMs showing that the strut thickness of the exposed struts was only 100 nm. The narrowness of these struts, combined with their long length increases the risk of mechanical damage, while the large exposed slot allows for direct interaction of these fibers with potential contaminants. While previous SCFs have been successfully used with struts thinner than these [19], the struts in these fibers are considerably more protected from physical damage or contamination than they are in the ECF.

During draws and subsequent characterization several mechanisms for physical damage were noted. Regions of intense scattering were marked, and the corresponding region collected for later SEM analysis. Some regions showed evidence of direct physical damage or contamination, such as that shown in Fig. 3(a). This damage is likely due to dust or salt coming in contact with the fiber core/strut region, potentially involving physical contact with this area resulting in damage to the fiber structure and loss of the guided light.

 

Fig. 3 Physical damage to Fiber #1, showing damage to thin struts and core structures. Scale bars show 40 µm.

Download Full Size | PDF

A second damage mechanism was also observed, where contamination or defects on the fiber surface resulted in a reduction in transmitted power, and damage to the fiber core/struts following illumination with the supercontinuum source (Fig. 3(b)). This could be observed during experiments when using an un-attenuated supercontinuum source, where the transmitted power would decrease rapidly after coupling was optimized. Attenuating the source by 10 × , or using a 20 mW helium neon laser resulted in no measurable power drop over several hours.

As a potential solution to the issue of fragility limiting the practicality of the ECF, a new preform was fabricated, with larger gaps between the holes (0.7 mm compared to 0.4 mm, as described earlier in methods. The first trials of this fiber (Fiber #2) showed that although the core geometry was good, and loss measurements demonstrated a more uniform attenuation along the length, the inner structure was rotated in the cladding jacket. A SEM image of the final fiber structure can be seen in Fig. 2(b).

For some applications, this may be desirable (surface functionalization, evanescent field interaction) as the core is still exposed along the length of the fiber, with the rotation of the core structure and the ‘ridge’ providing an effective buffer against physical damage.

While the rotation of the core presents some potential advantages, it also prevents the potential application techniques that require direct access to the core, such as femtosecond laser writing to create gratings along the length of the fiber. For the next fiber draw the cane structure was aligned in the jacket with the open hole at the center of the slot. With the fiber at the correct alignment the lower end of the cane was fused into the preform using a hydrogen torch, with the resultant fibre (Fiber #3) sho.wn in Fig. 4.

 

Fig. 4 Fiber 3, showing (a) 2.35 µm and (b) 1.65 µm core diameters, and (inset) magnified view of the core structure. The scale bars show 50 µm and 3 µm on the main and insets respectively.

Download Full Size | PDF

By varying the outer diameter of Fiber #3 a range of core sizes from 1.65 µm (Fig. 4(b)) to 2.35 µm (Fig. 4(a)) were fabricated. Loss measurements were performed on each band, with larger (2.08 µm) and smaller (1.65 µm) bands shown in Fig. 5(a) and 5(b), respectively.

 

Fig. 5 Loss measurements of Fiber 3, for (a) 2.08 µm and (b) 1.65 µm core fibers. Red lines show loss results, blue lines show the error in the fit for each wavelength point.

Download Full Size | PDF

From these results we can see that there does not appear to be a strong core-size dependence of the loss in the visible/near infra-red, as the spectra obtained are of comparable shape and magnitude from 400 to 1300 nm. Beyond this point the loss of the 1.65 µm core fiber increases compared to the 2.08 µm fiber, however even with a 5 dB/m loss at 1550 nm this could still find use as a sensor with lengths up to one meter.

3.2 Theory

The experimental results from Section 3.1 show that fibers can be fabricated with both low-loss, and good mechanical strength. To examine the optical properties the fiber was modeled in Comsol Multiphysics with a constant core size (1.65 µm) at 1550 nm, showing an approximately 100 × increase in the overlap between the guided mode and the air holes when comparing these 1.65 µm fibre (5%) to previous 10 µm core fibre (0.05%). Confinement loss was also modeled, with the strut thickness and length varied using an idealized model of the fiber geometry. The results of this are shown in Fig. 6. Note than the hole diameter here is a function of the strut length, as can be seen in the diagram below.

 

Fig. 6 – (left) Theoretical confinement loss for varied strut thicknesses, core diameter 1.65 µm at 1550 nm, and (right) modeled fiber geometry. Grey line shows 0.1 dB/m, as below this value confinement loss will not have a large contribution to sensor performance.

Download Full Size | PDF

From these results, we can see that as expected the confinement loss increases with increasing strut thickness and decreasing hole size. Measurements on the fabricated fiber band with the smallest outer diameter show a minimum strut thickness of 300 nm for the two exposed struts, and an approximate hole diameter of 9 µm. For example, if a similar hole diameter was desired, the strut thickness could be increased by more than 50% to 430 nm while still maintaining a 0.1 dB/m CL.

4. Conclusions

We have successfully demonstrated low-loss (<1 dB/m) ECFs, with core sizes as small as 1.65 µm. These fibers would be suitable for distributed measurements over lengths of several meters before loss in the fiber significantly degrades the signal. By adjusting the preform geometry, the robustness of the fibers has been improved over initial trials such that these fibers can now find practical use in sensing applications. The increased evanescent field overlap with the holes arising from the smaller core size will allow for increased sensitivity for fluorescence applications, or for gratings to be written with significantly shorter lengths for refractive index or temperature measurements. The loss of these fibers has been improved significantly over previously published results, such that much longer lengths fibers can be used in sensing applications for distributed applications, or in applications in which longer fibers will allow for integration of fluorescence signals.

Theoretical studies of the CL have shown that the strut thickness could potentially be increased compared to the fibers shown here, with potential corresponding improvements in the durability of the final fibers. This is especially true for applications at visible wavelengths, where confinement loss will be considerably lower than that seen at telecoms wavelengths.

Future work on improving the practical strength of these fibers could examine the fabrication of fibers with thicker or shorter struts to further enhance the durability of these fibers. Additional improvements to the strength of the fiber could also result from fabricating a fiber with a rotated core structure, such that the core itself is protected from direct physical damage and yet still open to interaction with gases or liquids along the length.