Hermetic Optical Fibers: Carbon-Coated Fibers

This is a continuation from the previous tutorial - X-ray optics.

 

1. Introduction

Hermetically coated optical fibers have a thin layer of an impervious material applied over the surface of the glass fiber. Hermetic coatings are used to improve the reliability of fibers by preventing strength degradation caused by moisture attack on the fiber surface, and by preventing the diffusion of hydrogen into the core of the fiber.

The typical polymer coatings that are used on optical fibers are able to prevent liquid water from contacting the glass surface but are not able to stop the diffusion of H\(_2\)O molecules or even-smaller H\(_2\) molecules.

Water molecules accelerate crack growth on the glass–fiber surface in a strength degradation process known as fatigue. Hydrogen molecules can quickly diffuse into the core of a nonhermetic optical fiber where they can cause significant increases in optical loss.

Although the effects of water and hydrogen permeation are minor for typical telecommunication environments, they can cause significant problems when optical fibers are used in environments in which the mechanical stresses and/or hydrogen levels are higher than normal.

For instance, fibers used in oil-well data logging are exposed to hydrogen and high stress levels at high temperatures, conditions that can quickly cause failure of standard fibers.

For fibers exposed to demanding conditions, it is often necessary to ‘‘seal’’ the fiber with a hermetic coating. Although many materials have been considered for use as hermetic coatings on optical fibers, carbon coatings have been the most successful solution and are the focus of this tutorial.

Figure 14.1 shows a schematic representation of a single-mode optical fiber with a hermetic carbon coating and a single-layer polymer coating. When the hermetic layer is a nonductile material such as carbon, a polymer coating, similar or identical to that used on nonhermetic fibers, is used to protect the thin hermetic layer from scratches and other mechanical damage.

 

 

Many materials have been evaluated for use as hermetic coatings on optical fibers. Because an effective hermetic coating has to block the diffusion of small molecules—H\(_2\)O and H\(_2\)—the material used for the coating should have a close-packed structure.

Early efforts to develop hermetic coatings for glass fibers focused on metallic coatings. Although fibers drawn through a molten pool of metal have been shown to be resistant to static fatigue and hydrogen ingression, it can be difficult to achieve long lengths of low-loss fiber with thin pinhole-free metal coatings.

Thicker metal layers can be used to avoid pinholes, but the increased thickness usually causes microbending losses to increase to unacceptable levels. Fibers with short metallized sections, on the other hand, are successfully used for sealing fibers into hermetic packages because the loss increases are acceptably low when the metallized fiber sections are short.

Carbon-coated optical fibers are made by depositing thin carbon layers onto the surface of the silica during the fiber-draw process. Layer thicknesses of only 20–50 nm of carbon have been shown to be sufficient to make the fiber hermetic against both water and hydrogen diffusion, without causing microbend losses.

Because of these advantages, carbon-coated optical fibers provide a good engineering solution for specialty fibers used in demanding applications. Carbon-coated optical fibers have been used to solve hydrogen loss problems in underwater cables, to protect sensitive erbium-doped amplifier fibers, to prevent fracture and darkening in fibers used in sensor systems for oil drilling, to protect fibers in automotive applications, and to prevent fatigue in tightly routed fibers in avionics.

While optically transparent diamond-like carbon coatings have been used to provide hard abrasion-resistant coatings for fiber devices, this type of carbon coating is not able to block water or hydrogen diffusion and, thus, is not useful for providing hermetic protection in most applications.

The target audience for this tutorial is the end-user (i.e., the engineer who needs to specify a hermetic fiber to solve a reliability problem). The different materials that have been evaluated as candidate hermetic coatings are described to give historical perspective. However, the primary emphasis is on carbon-coated hermetic fibers, because this was the primary type of hermetic fiber as this tutorial was being written.

The tutorial describes the primary reliability risks—fatigue and hydrogen losses—and how these problems are solved using hermetic fibers. The processes used to deposit carbon coatings onto fibers are discussed, as are some of the material studies done to characterize the nature of the carbon structure in carbon-coated fibers.

Measurement techniques used to evaluate mechanical and optical properties of carbon hermetic fibers are described because these methods are used to evaluate the specifications for a hermetic fiber, and to determine its suitability for use. Hydrogen testing is discussed in detail because this places the greatest demands on a carbon coating.

Techniques for handling, cleaving, and splicing hermetic fibers are also described. Finally, we discuss some of the applications of hermetic fibers to solve critical reliability problems, as well as the methods that can be used to specify the properties of hermetic fibers.

 

2. History

Hermetic coatings are used to protect fibers against fatigue and hydrogen-induced loss increases (i.e., against degradation by water and hydrogen, respectively). The development history of hermetic coatings, therefore, has been tied to the need to address these reliability issues.

Hermetic coatings for modern optical fibers have been investigated since the 1970s. It was recognized early on that the mechanical reliability of fibers was controlled by moisture-assisted bond breaking at the tips of microscopic cracks on the glass surface.

Since it was recognized that this was an inherent reliability issue for silica fibers, a variety of hermetic coatings were considered potential ways of reducing or eliminating the water-related fatigue of fibers.

Several early patents described, in general terms, the advantages of protecting the glass surface using metals, ceramics, or carbon. A paper gives a good historical and scientific overview of strength and fatigue issues relating to hermetic carbon-coated fibers.

Metal coatings were one of the first approaches to be considered. The general approach was to apply a metal coating during fiber drawing. While metal-deposition methods such as vacuum evaporation, sputtering, and deposition from glow discharges were mentioned in early work, most of the practical methods were based on drawing fibers through molten metals.

Pinnow et al. reported 15- to 20-μm aluminum coatings that were applied during fiber draw. Although the average fiber strength was low (~3.4 GPa [500 kpsi]), the fracture stress was found to be independent of the strain rate, indicating that the metal coating was effectively blocking water from reaching the fiber surface.

Some individual samples had strengths as high as 6.3 GPa (900 kpsi). Some success was achieved by others in stabilizing the coating deposition, thus eliminating coating discontinuities (‘‘pinholes’’) and improving the mechanical properties.

One important problem with these metal-coated fibers was increased microbending loss. Problems associated with microbending losses were widespread in metal-coated fibers for several fundamental reasons: typical metal coatings were relatively thick.

Therefore, bends imparted to the fiber during normal handling tended to cause semipermanent deformations of the metal coatings and of the fiber, resulting in localized microbends and associated losses.

In addition, the thermal contraction of a metal coating as it cooled from the melting point resulted in compressive forces on the fiber, which in turn caused small lateral displacements leading to microbend losses.

Winding and coiling the metal-coated fibers also tended to change the fiber loss. These losses were particularly problematic for long-length telecommunication applications with tight loss budgets.

An example of work on metal-coated fibers is that on indium-coated fibers. Indium coatings, with their lower melting points (156\(^\circ\)C), were investigated as a way to lower the microbend losses. They showed fatigue parameters of \(n=32\), significantly higher than the \(n=20\) value usually seen for nonhermetic fibers. The excess optical losses were 0.1–0.2 dB/km, tolerable for some short-distance applications. The fibers showed as-drawn fiber strengths of about 3.5 GPa (500 kpsi).

Although metal coatings were successfully made and did provide a way to prevent water diffusion, the optical losses were too high for many applications. Nonmetallic coatings based on materials such as Si\(_3\)N\(_4\), SiC, SiON, TiC, and carbon were all investigated as potential low-loss hermetic coatings.

These high-temperature coatings were typically applied by using different types of vapor-phase deposition during fiber drawing. The development of these and other hermetic coatings was accelerated in the 1980s by the realization that hydrogen-induced loss increases could occur in deployed optical fibers.

Because H\(_2\) molecules are smaller than H\(_2\)O molecules, the requirements on coatings became more stringent. Metals such as aluminum could stop H\(_2\) diffusion, but they suffered from the previously mentioned problem of microbending losses. Appropriately deposited films of inorganic coatings such as SiC, Si\(_3\)N\(_4\), SiO\(_x\)N\(_y\), and several forms of carbon, were shown to be capable of blocking H\(_2\) (and H\(_2\)O) without causing high losses.

Work at Hewlett Packard Laboratories used mixtures of various gases to deposit coatings of the general formula Si\(_x\)C\(_y\)N\(_z\)O\(_w\). One of the goals was the protection of fibers from fatigue in oil-well data-logging applications, where fibers can see 2% strains and temperatures in excess of 200\(^\circ\)C in a high-humidity corrosive environments.

Mixtures of SiH\(_4\), CO\(_2\), NH\(_3\), hydrocarbon gases, N\(_2\), and He were used to deposit thin films of various compositions. Although not all of the coating compositions were successful, several compositions showed good fatigue resistance with little or no adverse effect on fiber loss.

The patent discussed the importance of the Si-C bond in achieving fatigue resistant hermetic fibers. The inclusion of O and N in the coating composition was thought to help match the physical properties of the film to the underlying silica fiber, but the coatings were basically SiC according to the claims in their patent.

The reported values for the fatigue parameter, \(n\), were between 8 and 256, with film thicknesses ranging from about 10 to 67 nm. As was seen in much of the later work on ceramic and carbon coatings, there was a noticeable reduction in the as-drawn strength of some of the fibers.

Results in one paper described 3.2 GPa (464 kpsi) fiber strengths for hermetic fibers [19]. Although this was acknowledged to be lower than the typical 5 GPa (725 kpsi) as-drawn strength for nonhermetic fibers, it was pointed out that the hermetic fiber strength was maintained at this 3.2 GPa level and that of the nonhermetic fiber steadily degraded due to moisture attack.

Analysis showed that after about 2.5 hours, the strength of the hermetic fiber would exceed that of the nonhermetic fiber. The coatings were also shown to slow the diffusion of hydrogen into the fiber core. No hydrogen was detected after 1000 hours at room temperature, but some H\(_2\) diffusion was noted at 75\(^\circ\)C, albeit at a much lower level than in nonhermetic fibers.

British Telecomm Labs developed practical silicon oxynitride hermetic coatings in the early 1980s. These coatings were tested at both low (0.74 atm) and high (65 atm) H\(_2\) partial pressures. In neither case was it possible to detect H\(_2\) absorption peaks in the fibers. This allowed the maximum room temperature diffusion coefficient for H\(_2\) to be estimated as being about 5 x 10\(^{-19}\) cm\(^2\)/sec.

This H\(_2\) diffusivity is more than seven orders of magnitude lower than that for SiO\(_2\). Because the experiments were done only at 21\(^\circ\)C, it was not possible to predict the H\(_2\) diffusivity at higher temperatures. It is likely that the same coatings would have been effective barriers to H\(_2\)O molecules and, thus, would have improved fatigue and prevent H\(_2\) losses.

A study in 1986 described fatigue and strength results for SiON, C, TiC, SiC, and for C-SiON and C-TiC composite coatings. The SiON fibers had low strengths of 1.4–1.7 GPa (200–250 kpsi) but had high \(n\) values of about 90.

The carbon-coated fibers in that study had strengths of 4.1–4.5 GPa (600–650 kpsi), but \(n\) values in the range of 23–25 indicated that these carbon coatings were only partially effective in blocking H\(_2\)O diffusion. The results of that study were typical in demonstrating the need to tradeoff fracture strength against hermetic properties.

Starting in about 1988, there was significant progress in carbon-coated fibers. The carbon-coated fibers developed by a number of companies showed good engineering properties with acceptable fracture strengths and good resistance to fatigue and hydrogen loss.

Corning reported results for thin (<50-nm) amorphous carbon coatings, showing good fatigue and hydrogen properties and usable fiber strengths. Dynamic fatigue values of \(n\) ~ 110 were achieved with fiber strength of about 3.5 GPa (500 kpsi).

Analysis showed that the improved fatigue characteristics of hermetic fibers allowed the hermetic fibers to be used at 80% of their proof-test level, versus only about 30% for a nonhermetic fiber.

Hydrogen diffusion in the same fibers resulted in loss changes of only about 0.25 dB/km at 1240 nm after 200 days at 21\(^\circ\)C, indicating that the coatings were effective in retarding H\(_2\) diffusion at room temperature. As expected, H\(_2\) diffusion was faster at higher temperatures. Carbon coatings are still used on Corning erbium-doped fibers, both to permit tight coiling of the Er fibers in compact amplifier modules and to stabilize the fibers against long-term loss changes.

In 1985 AT&T reported results for hermetic fibers with approximately 50- to 100-nm carbon coatings having a pyrolytic graphite structure. The coatings were deposited by thermally decomposing carbon-containing gases onto the fiber surface during fiber draw.

These fibers had relatively high as-drawn strengths of 4.1 GPa (600 kpsi). Fiber fatigue was minimal as evidenced by dynamic fatigue values from 350 to 500 and static fatigue values of about 200.

Hydrogen permeation through the coatings was not measurable at room temperature but could be detected by exposing fibers to high pressure H\(_2\) at 100–150\(^\circ\)C.

By characterizing the diffusion constant as a function of temperature, it was possible to quantitatively predict fiber loss increases in H\(_2\) atmospheres as a function of time and temperature. Similar to the work on SiON, the room temperature diffusion coefficient for H\(_2\) in the film was extremely low, about 1 x 10\(^{-20}\) cm\(^2\)/sec at 21\(^\circ\)C for the carbon films.

These fibers incorporated reactive gettering sites in the optically inactive silica cladding layers. These sites aided in protection against hydrogen loss by reacting with and immobilizing the trace levels of H\(_2\) that penetrated the carbon coating at high temperatures.

Several Japanese groups investigated the relationship between the properties of carbon-coated fibers and the carbon morphology. Workers at Sumitomo predicted long-term loss increases using accelerated H\(_2\) testing and correlated the predicted behavior with the surface morphology of the carbon. They concluded that a smooth surface structure leads to improved performance of the fiber in hydrogen.

Work at NTT Laboratories came to a similar conclusion regarding fiber strength. Fiber strengths of more than 4.8 GPa (700 kpsi) were obtained, with dynamic fatigue \(n\) values as large as 670.

These high values were attributed to an ultrasmooth carbon surface (as measured by a Scanning Tunneling Microscope). Hydrogen measurements at 75\(^\circ\)C showed that the coatings did have the ability to retard H\(_2\) diffusion, although long-term predictions were not made.

Field-test experiments evaluating cabled carbon-coated fibers were conducted at Furukawa and showed that no hydrogen losses occurred over a 1.5-year period even for fibers in a water-filled cable.

Various mechanical tests on the cabled fibers show that loss increases did not occur even under simulation of harsh handling conditions. A similar study by NTT concluded that the performance of carbon-coated fibers was excellent in field tests, and that the fibers offered improved reliability against fatigue.

A study by BTRL pointed out, in a 1991 paper, the differences between carbon-coated fiber obtained from different manufacturers, and discussed the materials properties of the carbon films as well as the testing techniques used to characterize strength, hydrogen permeation, and electrical resistivity of the coatings.

This paper discussed the use of electrical resistivity measurements to characterize carbon coating properties, and the correlations of resistivity with coating properties affecting H\(_2\)O and H\(_2\) diffusion.

The authors discussed the sensitivity of the coating properties to the details of the carbon bonds— diamond-like versus graphitic—and the methods used to deposit the coating. They concluded that of the fibers studied, only a subset would be suitable for use in environments where H\(_2\) losses might be a problem.

They did find a good correlation between electrical resistivity and hydrogen performance for some of the fibers in the study. They also concluded that thinner coatings tended to be associated with higher fiber strengths.

Overall, the paper highlighted the importance of having clearly defined test procedures to quantitatively determine the suitability for use of a carbon-coated ‘‘hermetic’’ fiber.

By about 1990, the engineering of carbon-coated fibers had resulted in fibers that met stringent optical performance requirements and had significantly improved reliability.

There were, however, several engineering and perception tradeoffs associated with these carbon-coated fibers. The black color of the carbon coatings altered the visual appearance of colored inks applied over the polymer coating for fiber identification purposes. Fusion splicers required frequent cleaning when used to splice hermetic fibers.

Finally, there was a perception that all carbon-coated fibers had inadequate fracture strengths, probably stemming from the fact that early hermetic fibers had been much weaker than both nonhermetic fibers and the improved carbon-coated fibers that came later.

Since 2005, hermetic fibers have been used predominantly in specialty applications. One reason that the coatings are not used more widely is that problems of fatigue and hydrogen aging have been solved for many applications without the use of hermetic fiber coatings.

Design rules limit the stresses seen by fibers, greatly decreasing fatigue-induced weakening of nonhermetic fibers. Similarly, optical fiber cable designs and materials have been improved so that H\(_2\) evolution in a cable is a relatively uncommon event, and the extra protection offered by a carbon coating is not usually required.

However, there are a growing number of specialty applications that do require that the fiber be protected from H\(_2\) aging and/or fatigue. Fiber sensors used in oil-well data logging need the protection offered by carbon coatings, to protect the fibers from both fatigue and hydrogen aging.

Tightly routed fibers, for instance, in airframes or ultracompact fiber modules, can avoid fatigue and maintain their reliability even at very small bend radii when they have hermetic carbon coatings.

Finally, in some applications (e.g., space or undersea), the stringent system reliability requirements are more readily met when the fiber reliability is enhanced with a protective carbon coating.

 

3. Deposition of Carbon Coatings on Fibers

The deposition of the carbon layer on an optical fiber occurs during the fiber-draw process. Immediately after the glass fiber reaches its final size (e.g., 125 μm), it is introduced into a hydrocarbon gas such as dilute acetylene, where the retained heat of the fiber causes the hydrocarbon to ‘‘crack’’ (thermally decompose) in a pyrolytic reaction.

This reaction on the surface of the glass leads to the chemical vapor deposition of carbon onto the fiber. Although the deposition process is relatively straightforward, it has many possible permutations based on the deposition reactor design, precursor hydrocarbon gas, and fiber-draw speed.

All of these factors affect the pyrolysis of the reactant gas, the deposition rate, and the hermeticity of the resulting carbon layer.

Because the carbon layer is a thin brittle coating, it does not protect the fiber from mechanical damage such as scratches. For this reason, carbon-coated fibers are always protected by polymer coatings, which are applied using standard coating applicators located below the hermetic coating reactor on the draw tower.

Standard dual-layer acrylate coatings can be used, as can specialty polymer coatings such as polyimide coatings for high-temperature applications. The draw speed needs to be compatible with both the carbon and the polymer coating process.

For instance, a carbon coating process that requires a high draw speed will not be compatible with a polymer coating process that needs a low draw speed.

A carbon deposition reactor is designed to strip the boundary layer of air from around the glass fiber and to deliver a hydrocarbon gas to the glass surface at a reaction temperature around 700–900\(^\circ\)C.

For fibers drawn at high speeds (>5 m/sec), the fiber exits the draw furnace at a high temperature and the reaction zone is very long (tens of centimeters). Carbon reactors used at this speed are typically elongated chambers located below the draw furnace, where an organic gas is introduced at one end and unreacted gases are exhausted from the other.

As the draw speed decreases, the reaction zone shrinks and the carbon reactor may need to be attached directly to the draw furnace to ensure the fiber has enough retained heat for carbon deposition.

Reactant gases include a variety of hydrocarbons, and indeed, many different carbon-containing gases have been investigated. Reactants have included methane, acetylene, ethylene, propane, butadiene, trichloroethylene, and benzene.

In addition, some researchers have reported that the addition of chlorine gas to the reactor improves the carbon fiber’s strength and hermetic characteristics, possibly by acting as a hydrogen scavenger during the pyrolytic reaction.

Hydrogen, which is a byproduct of the pyrolysis reaction, can, in some cases, become trapped under the carbon coating, resulting in increased fiber loss. The choice of precursor gas depends on the reactor design and the temperature of the fiber in the reaction zone, which in turn is determined by the fiber draw speed.

The carbon layer applied to the fiber is black and is usually shiny in appearance. It is electrically conductive, which allows measurement of the coating’s electrical resistance and inference of the coating thickness.

The carbon layer thickness can be estimated from the electrical resistance with the following equation:

\[\tag{14.1}R=\frac{\rho}{2\pi{r}\delta}\]

where \(R\) is the linear electrical resistivity (\(\Omega\)/cm), \(\rho\) is the resistivity of the applied carbon layer (\(\Omega\)-cm), \(r\) is the fiber radius, and \(\delta\) is the carbon layer thickness.

By correlating dynamic fatigue measurements with measurement of electrical resistance, it is sometimes possible to use the linear resistivity as a measure of the coating’s ability to block water diffusion.

For instance, one manufacturer has shown that a linear resistivity of R < 25 k\(\Omega\)/cm is sufficient to guarantee that the carbon layer will be hermetic from the standpoint of fatigue. The measured resistance in this case corresponded to a thickness of about 20 nm.

The correlation between thickness and resistance depends on the resistivity of the applied carbon, which in turn is a function of graphitic crystallite size, orientation, and perfection.

Because the material properties are sensitive to processing conditions, the resistance values for a given coating thickness may be different for fibers made by different manufacturers.

Coating resistivity can be measured directly using an ohmmeter and liquid metal contacts. The polymer coating needs to be stripped from the fiber for these off-line measurements.

Online measurements, conducted during the drawing of a hermetic fiber, are advantageous because they give immediate information about the coating properties and do not require destructive testing (i.e., polymer coating removal) after the fiber is drawn.

Noncontact measurements of electrical resistance and laser scattering techniques have been used to measure resistivity and coating thickness, respectively.

One analysis of the applied carbon material indicated that the layer consists of disordered graphitic platelets that are randomly oriented and bonded on the surface of the glass fiber in a continuous structure.

Thus, to close down pathways between the graphite platelets for water or hydrogen ingression, the carbon layer needs to have a thickness more than some minimum value.

One study concluded that the carbon thickness should be more than 20 nm to prevent fatigue and 25 nm to delay hydrogen ingression. It is important to realize that that thickness per se is not always a useful parameter for coating characterization, especially when comparing fibers made by different processes.

The direct measurement of coating thickness requires careful sample preparation and examination of the fiber in an electron microscope.

In addition, the material properties of a carbon coating depend on the processing conditions. For instance, a thick diamond-like carbon coating will offer much less protection from water and H\(_2\) than a thinner coating with a graphitic structure.

 

4. Fatigue Properties of Carbon-Coated Fibers

By excluding water from the surface of the glass, the carbon layer prevents the onset of moisture-assisted crack growth, known as fatigue.

Resistance to fatigue is measured by the stress corrosion resistance parameter \(n\). A high \(n\) value indicates a greater resistance to fatigue and, thus, a greater capability for a fiber to maintain its original strength.

The static fatigue parameter, \(n_s\), can be determined by measuring time to failure for fibers subjected to different static loads. Alternatively the dynamic fatigue parameter, \(n_d\), can be determined by measuring failure stress as a function of strain rate.

The values for \(n_s\) and \(n_d\) typically have similar values for a given fiber. Because dynamic fatigue tests are quicker to perform, they are more often used in characterizing fibers.

Conventional fibers tend to have \(n\) values around 20, whereas hermetic fibers demonstrate \(n\) values of 100 or more. The area of fiber strength and fatigue in nonhermetic optical fibers has been studied extensively.

The strength retention of carbon-coated fibers makes them ideal for use in adverse environments, in which water or other corrosive chemicals would normally lead to premature failure in unprotected fibers.

Carbon-coated fibers immersed in water have demonstrated no loss in strength, even after many months at elevated temperatures. In laboratory experiments, carbon coatings have been shown to be impervious to concentrated hydrofluoric acid and to hot sodium hydroxide solutions.

In addition to providing a long lifetime at standard application stresses, hermetic fiber will permit operation at elevated mechanical stresses. A rule of thumb for nonhermetic fibers is that the application stress should not exceed about 20–30% of the proof-test level.

However, because the carbon layer prevents the crack growth that leads to failure, hermetic fibers can operate at up to 80% of the proof-test level. This is a very attractive feature for fibers used in high stress applications such as compact fiber-based components that require tight fiber bend radii.

Fatigue is defined as crack growth on a stressed fiber in the presence of water. The stress applied to a fiber during use may be tensile stress, such as in aerial cables, or bending stress, such as those found in tight enclosures.

On a molecular level, the mechanics of fatigue are straightforward: an H\(_2\)O molecule at a crack tip in the SiO\(_2\) glass matrix will rupture the silicon–oxygen bond, leading to crack growth.

Fatigue occurs preferentially at the crack tip because this is the point of maximum stress concentration (i.e., where the silicon–oxygen bonds are most strained). Thus, to mitigate the effect of fatigue, it is best to minimize

1. The size of cracks on the surface of the glass fiber
2. The stress applied to the fiber
3. The presence of water

The role of the hermetic coating is to affect factor (3) by preventing water molecules from reaching the fiber surface and the crack tip.

The fatigue resistance of hermetic fibers can be assessed using dynamic tensile testing. In this procedure, fibers are strained to failure at different strain rates. Nonhermetic fibers show decreased strengths at lower strain rates because the lower strain rates give more time for water reaction at the crack tip.

Figure 14.2 shows data for hermetic carbon-coated fibers and a typical nonhermetic fiber. The fracture stresses were determined for both fiber types using a range of strain rates: 25%/min, 2.5%/min, 0.25%/min, and 0.025%/min. The fracture stresses were plotted at each strain rate, as shown in Fig. 14.2.

 

 

The slope (\(m\)) of each line is calculated by linear regression and the dynamic fatigue resistance factor \(n_d\) for each fiber was derived from Eq. (14.2):

\[\tag{14.2}n_d=\frac{1}{m}-1\]

where \(n_d\) is the dynamic fatigue factor and \(m\) is the slope of stress–strain curve.

A flatter slope (i.e., a slope approaching zero) will produce a higher dynamic fatigue value, \(n_d\). The graph demonstrates the advantages of a high \(n\) value.

For the nonhermetic fiber, the breaking strength declines as the strain rate is decreased and the fiber spends more time under strain. The longer time under strain increases the time for water molecules to react at the crack tip.

In actual use, a nonhermetic fiber that is held under stress will gradually weaken. For hermetic fibers, the absence of water at the crack tip prevents fatigue, so there is no appreciable decrease in fracture strength at slower strain rates.

Similarly, the hermetic fiber will maintain its strength over long times even when exposed to water or humidity in a stressed state.

Although there are several models to estimate a fiber’s lifetime under stress, they all generally derive from the Weiderhorn model

\[\tag{14.3}t_f=BS_\text{int}^{n-2}\sigma^{-n}\]

where \(t_f\) is the time to failure, \(B\) is a crack growth parameter, \(S_\text{int}\) is the intrinsic strength of the fiber, and \(\sigma\) is the applied stress.

The difficulty in forming a standard equation for fiber lifetime lies in the fact that, for any given fiber, \(B\) and \(S_\text{int}\) are unknown. However, if a fiber is proof-tested, then the term \(S_\text{int}\) may be reasonably replaced in Eq. (14.3) by \(\sigma_p\), the proof-test stress. Long-term stress levels for nonhermetic fibers, with \(n\) values of about 20, have to be maintained at much lower levels than those for high \(n\)-value hermetic fibers.

The ability of hermetic fibers to retain their strength over long periods is demonstrated in Figs. 14.3 and 14.4.

In this study, standard nonhermetic acrylate-coated fibers and carbon-coated fibers (also with acrylate coatings) were exposed to various aqueous environments over a 9-month period.

The nonhermetic fibers exhibited strength degradation that was accelerated with water temperature. For fibers hermetically sealed with a thin layer of carbon, no change from initial strength was observed for any of the aged samples (Fig. 14.4).

Similar results were observed for fibers soaked in conventional chemicals, indicating that carbon coatings can protect fibers from harsh environments and elevated temperatures more effectively than standard polymer coatings.

 

 

 

5. Hydrogen Losses in Optical Fibers

5.1. Hydrogen-Induced Losses in Nonhermetic Fibers

In the early 1980s, it was realized that hydrogen-induced losses could impair optical fiber systems. Fibers exposed to hydrogen, even at low levels, showed increased losses that, in some cases, caused fiber losses to exceed the system-aging margin, resulting in system failure.

Hydrogen was found to originate from certain silicone-based polymers and from the generation of H\(_2\) by corrosion of metal parts inside the cables. H\(_2\) molecules can diffuse through polymer materials and through a fiber’s silica cladding in a matter of days. At room temperature, the H\(_2\) is detectable at the core in about a day and losses reach their equilibrium levels in less than 2 weeks.

As hydrogen gas diffuses into the light-carrying portion of the optical fiber, it can lead to both reversible and permanent optical loss increases. Figure 14.5 shows hydrogen-induced loss changes in an accelerated experiment with a nonhermetic fiber exposed to pure H\(_2\) for 3 days at 150\(^\circ\)C.

The reversible loss increases are due to unreacted H\(_2\) molecules dissolved in the silica matrix. H\(_2\) molecules are not infrared active in the gas phase but become absorbing when dissolved in SiO\(_2\). There are several localized loss peaks in the 1000- to 1700-nm spectral region, including a prominent first overtone absorption at about 1240 nm.

 

 

Figure 14.6 shows the molecular H\(_2\) spectrum for a fiber that was equilibrated in a high-pressure H\(_2\) gas environment. There is a loss edge that increases at wavelengths longer than about 1500 nm, affecting losses in the 1550-nm transmission window.

The absorbing strength of H\(_2\) in the 1550-nm region is about 0.6 dB/km/atm at room temperature. The effects in the 1310-nm window are about three times lower than at 1550 nm.

The molecular H\(_2\) losses are proportional to the local H\(_2\) partial pressure. The losses are reversible in that the H\(_2\) losses will recover if the hydrogen is removed from the fiber environment. The time scale for loss recovery is the same as for growth; 95% of the H\(_2\) will leave the fiber core in 2 weeks at room temperature once the fiber is removed from the H\(_2\) atmosphere.

Hydrogen diffusion rates increase with temperature, but the strength of the absorption actually falls off with increases in temperature because the solubility of H\(_2\) decreases as the temperature is raised.

Molecular hydrogen losses are approximately the same for all silica-based fibers. These losses are avoided only by carefully controlling hydrogen in the fiber environment (e.g., the cable) or by using a hermetically coated fiber that is designed to block H\(_2\) diffusion.

 

 

At elevated temperatures, H\(_2\) can react with point defects in the fiber core, giving rise to OH absorption peaks at 1390 nm, 1240 nm, and other wavelengths, and to a short wavelength edge (SWE) that causes spectrally broad loss increases in Ge-doped fibers (Fig. 14.5).

Unlike molecular H\(_2\) losses, the reaction rates for hydrogen in different fiber types can differ greatly and tend to increase rapidly as the temperature is increased. These reactions are usually not reversible.

The absorbing strength of some of the lossy species can be quite high. For instance, 1 ppm\(^2\) of OH will cause about 15 dB/km of loss increase at 1390 nm. By lowering the core dopant levels (e.g., Ge and P), it is sometimes possible to lower the fiber’s reactivity, but such dopant changes are not always practical.

For instance, fibers with undoped silica cores are less prone to hydrogen reactions, but the variety of fiber designs that can be achieved without Ge and/or P doping is limited. As with molecular H\(_2\) losses, it is possible to eliminate or at least decrease the losses due to hydrogen reactions by using a hydrogen-blocking carbon coating.

 

5.2. Hydrogen Losses in Carbon-Coated Hermetic Fibers

At room temperature, most carbon coatings can block H\(_2\) to such an extent that no loss changes will be detectable over the time scale of days. However, at elevated temperatures (e.g., 100\(^\circ\)C), some H\(_2\) diffusion may be detectable in accelerated aging experiments or under aggressive use conditions.

Whether a carbon-coated fiber is suitable for use in a hydrogen environment depends on the tolerable loss changes, the operating temperature, the concentration of hydrogen around the fiber, and the material properties of both the fiber and the carbon coating.

Determining whether a coating is adequately hermetic requires a quantitative analysis. No carbon fiber coating made to date has been shown to be perfectly hermetic. Increases in ambient hydrogen levels, pressure, temperature, and fiber length all lead to increased demands on the fiber coating.

Similarly, if the fiber’s core glass material is highly reactive with hydrogen (e.g., some erbium-doped fibers and Ge-P co-doped multimode fibers), the demands on the hermetic coating are increased.

 

5.3. Testing of Hermetic Fibers in Hydrogen

To characterize a hermetic coating’s ability to retard hydrogen diffusion, it is necessary to use accelerated aging experiments.

In these experiments, a length of hermetic fiber is exposed to hydrogen, usually at an elevated temperature and sometimes using high-pressure hydrogen. Loss changes are detected either by measuring the changes in fiber loss in real time (in situ measurements) or by measuring the fiber loss before and after the exposure of the fiber to hydrogen.

Measurement of loss changes can be done via OTDR loss measurements (1310, 1550, and/or 1625 nm), by using a single-wavelength laser source and an optical detector, or by using a broadband light source and an optical spectrum analyzer (OSA) to obtain a full loss spectrum.

Diffusion of hydrogen through carbon depends strongly on the temperature. Therefore, unless the temperature dependence is already known, experiments need to be done at two or more temperatures to characterize the temperature dependence of H\(_2\) diffusion through the coating.

 

 

Figure 14.7 shows an example of an experimental setup that can be used for in situ characterization of a hermetic fiber. In this case, the measurement equipment allows a choice between OTDR and optical loss measurements, although typically only one method would be used.

Loss-change data are obtained periodically throughout the accelerated aging experiment, generally by computer-controlled test equipment. Before testing, one should wind the fiber onto a spool that will not be damaged at the test temperature and that will not impose stresses on the fiber due to thermal expansion of the spool material.

Because of the flammable and explosive nature of hydrogen gas, it is essential to recognize that hydrogen gas must be handled with appropriate safety precautions, especially if it is used at high pressures.

Because hydrogen can cause brittleness in some types of steel, it is important to verify that the materials and fittings used to handle the hydrogen are compatible with the temperatures and pressures used in the experiments.

A fiber with a ‘‘good’’ hermetic coating may exhibit only small loss changes when tested in hydrogen, especially if the test temperature is under about 100\(^\circ\)C. Improved measurement sensitivity can be obtained by increasing the fiber length and/or by measuring the loss change at a wavelength where the hydrogen absorption is strong.

Fiber lengths of about 1 km are usually sufficient for experiments run in the 100–200\(^\circ\)C range. The 1240-nm H\(_2\) peak is distinct and readily measured. It is generally the easiest feature to monitor if spectral measurements are being made. (There is small contribution at the same wavelength from an OH absorption band. The OH contribution at 1240 is about 1/20 that of the OH overtone at 1390 nm.)

When an OTDR is used to measure loss changes, it is best to use 1550- or 1625-nm wavelengths because the H\(_2\) absorption is significantly stronger at these wavelengths than at 1310 nm.

For weakly guiding fibers, microbend loss increases can sometimes complicate the data analysis because these losses may also contribute to 1550- and 1625-nm losses. For this reason, it is usually preferable to characterize fibers with spectral loss measurements since this allows the tracking of hydrogen-specific features such as the 1240-nm H\(_2\) overtone.

 

 

Figure 14.8 shows before-and-after loss measurements for a carbon-coated fiber tested at 152\(^\circ\)C for 15 hours in 144 atm of H\(_2\). The increase in the 1240 H\(_2\) peak and the rising loss beyond 1600 nm are clear evidence of H\(_2\) in the fiber core.

The overall upward shift in loss values is due to spectrally broad SWE loss increases, associated with hydrogen reaction at Ge sites.

Figure 14.9 shows the growth of the 1240-nm H\(_2\) peak in three sections of fiber tested at different temperatures, with 140 atm of H\(_2\) pressure. As is typical for carbon-coated fibers, there is an initial ‘‘lag’’ period followed by a period where the losses increase linearly with time. The explanation for this behavior is discussed in the following section.

 

 

5.4. Diffusion of Hydrogen in Hermetic Fibers

The time and temperature dependencies for hydrogen diffusion through carbon coatings can be accurately predicted using classic diffusion equations once the coating properties have been determined.

When a hermetic fiber is exposed to a hydrogen-containing atmosphere, the hydrogen will initially diffuse into the carbon and later into the silica part of the fiber.

Characteristic time constants, \(\tau_i\) and \(\tau_f\), can be used to describe the durations of the initial and final stages of diffusion. Loss increases occur only when the H\(_2\) reaches the silica core of the fiber.

Figure 14.10 shows hydrogen concentration profiles at different times for a hermetic fiber exposed to hydrogen.

 

 

In Fig. 14.10a, the fiber is first exposed to H\(_2\) and no diffusion has yet occurred. In Fig. 14.10b, the hydrogen has partially diffused through the carbon but has not reached the silica part of the fiber.

In Fig. 14.10c, the H\(_2\) concentration has started to increase in the silica, and H\(_2\) losses will start to become measurable in the fiber.

In Fig. 14.10d–f, the concentration of H\(_2\) in the fiber gradually reaches equilibrium (Fig. 14.10f). Because of the slow diffusion of H\(_2\) in carbon coatings near room temperature, the values of \(\tau_i\) and \(\tau_f\) can be quite large, from months to years for \(\tau_i\) and more than 10\(^4\) years for \(\tau_f\).

The H\(_2\) concentration gradients are flat in the gas phase and in the silica because H\(_2\) diffusion in these materials is much faster than in the carbon layer. Because of differing solubilities of H\(_2\) in SiO\(_2\) and carbon, there are offsets in the concentrations at the interfaces, as seen, for instance, in Fig. 14.10d–f.

The rate of hydrogen diffusion into the fiber can be mathematically formulated in a manner similar to that described by Crank. For the early stages (Fig. 14.10a–c), the loss changes associated with molecular H\(_2\) losses are proportional to the concentration of H\(_2\) in the fiber core and are given by

\[\tag{14.4}\Delta\alpha_{H_2}(t)=L_{H_2}P_{H_2}K_s\left[\frac{t}{\tau_f}-\frac{\tau_i}{\tau_f}-\frac{12\tau_i}{\pi^2\tau_f}\sum_{n=1}^{\infty}\frac{(-1)^n}{n^2}\exp\left(\frac{-n^2\pi^2t}{6\tau_i}\right)\right]\]

where \(L_{H_2}\) is the optical loss due to a given concentration of H\(_2\), P\(_{H_2}\) is the hydrogen partial pressure, and \(K_s\) is the solubility of H\(_2\) in SiO\(_2\) per unit partial pressure.

The elapsed time is given by \(t\), and the two characteristic time constants, \(\tau_i\) and \(\tau_f\), depend on material properties and the carbon thickness as follows:

\[\tag{14.5}\tau_i=\frac{\delta^2}{6D_c}\]

\[\tag{14.6}\tau_f=\frac{r\delta{K_s}}{2D_cK_c}\]

where \(\delta\) is the coating thickness, \(D_c\) is the diffusivity of H\(_2\) in the carbon-coating material, \(r\) the fiber radius, and \(K_c\) is the solubility of H\(_2\) in carbon per unit partial pressure.

The value of \(\tau_i\) indicates how long it takes H\(_2\) to penetrate the coating, whereas \(\tau_f\) is a measure of how fast the hydrogen losses will increase once the H\(_2\) is present in the fiber.

The molecular H\(_2\) losses follow a simple exponential time dependence after an initial diffusion lag time, \(\tau_i\):

\[\tag{14.7}\frac{\Delta\alpha_{H_2}(t)}{\Delta\alpha_{H_2}(\infty)}=1-\exp\left[\frac{-(t-\tau_i)}{\tau_f}\right]\]

where \(\Delta\alpha_{H_2}(\infty)\) is the equilibrium hydrogen loss and is equal to L\(_{H_2}\)P\(_{H_2}\)K\(_s\).

To predict the hydrogen loss for a carbon-coated fiber, the values for \(\tau_i\) and \(\tau_f\) need to be known. Because the coating properties (\(\delta\), \(D_c\), and \(K_c\)) that determine \(\tau_i\) and \(\tau_f\) depend on manufacturing methods and will typically not be known, it is necessary to have a practical method for experimentally determining \(\tau_i\) and \(\tau_f\).

By doing in situ loss measurements on a fiber exposed to hydrogen, one can determine the values for \(\tau_i\) and \(\tau_f\) based using short-duration accelerated experiments.

A loss that is associated with molecular H\(_2\), typically the 1240-nm peak, is measured at different times. The results will show an initial lag period followed by a period where the H\(_2\) losses increase linearly with time.

A schematic representation is shown in Fig. 14.11. Extrapolation of the straight portion of the curve back to the time axis gives \(\tau_i\). The slope of the curve is inversely proportional to \(\tau_f\) and directly proportional to the value of the hydrogen loss, \(\Delta\alpha_{H_2}(\infty)\), which would be seen in a nonhermetic fiber (or for a hermetic fiber at infinite time). The value for \(\tau_f\) is therefore \(\tau_f=\Delta\alpha_{H_2}(\infty)/m\), where \(m\) is the linear slope of the loss change versus time curve.

 

 

The value for \(\Delta\alpha_{H_2}(\infty)\) depends strongly on the wavelength, as shown in Fig. 14.6. It decreases weakly with temperature and is proportional to the H\(_2\) partial pressure used in the experiment.

Molecular H\(_2\) losses are similar for different fiber types and values can generally be obtained from published literature. A useful expression for estimating equilibrium hydrogen losses is

\[\tag{14.8}\Delta\alpha_{H_2}(\infty)=P_{H_2}A_{H_2}=P_{H_2}\left[\alpha_{H_2}\exp\left(\frac{8.67kJ/mole}{RT}\right)\right]\]

where \(\alpha_{H_2}\) is a wavelength-dependent hydrogen absorption with units of dB/(km-atm). Values for \(\alpha_{H_2}\) at 1240 nm (the first H\(_2\) overtone peak) and at 1550 nm are 0.355 and 0.017 dB/(km-atm), respectively.

Exact loss values depend on the resolution bandwidth setting of the optical spectrum analyzer. Values at other wavelengths can be obtained by scaling the \(\alpha_{H_2}\) value using the spectral shape of the H\(_2\) spectrum (e.g., Fig. 14.6).

Typical results from in situ testing of a carbon-coated fiber are shown in Fig. 14.12.

 

 

Sections of a typical carbon-coated fiber were tested at high hydrogen pressures (137–144 atm) at different temperatures. Results like those in Fig. 14.9 were obtained by monitoring the growth of the 1240-nm peak at temperatures from 50 to 150\(^\circ\)C and analyzing the data to determine \(\tau_i\) and \(\tau_f\).

As shown in Fig. 14.12, the temperature dependencies for \(\tau_i\) and \(\tau_f\) were consistent with Arrhenius relations and had similar activation energies. The values for \(\tau_i\) and \(\tau_f\) as functions of temperature were

\[\tag{14.9}\tau_i=6.91\times10^{-7}\exp\left[\frac{82.07\text{ kJ/mole}}{RT}\right]\text{ sec}\]

\[\tag{14.10}\tau_f=1.7\times10^{-5}\exp\left[\frac{99.21\text{ kJ/mole}}{RT}\right]\text{ sec}\]

where \(R\) is the gas constant (8.314 J/mol-K).

By using Eq. (14.7) with the experimental values for ti and tf , the effective hydrogen level in the fiber core can be calculated as a fraction of the external hydrogen pressure.

Figure 14.13 shows predictions based on the data in Fig. 14.12, compared to a nonhermetic fiber. The H\(_2\) loss increase is the ratio from the vertical axis in Fig. 14.13 multiplied by the outside hydrogen pressure and the hydrogen absorption (\(A_{H_2}\)) at the wavelength and temperature of interest (Eq. [14.8]).

 

 

The hydrogen-blocking properties of a specific carbon coating depend on how the coating is deposited. The reactants and the processes used are specific to individual fiber manufacturers.

The range of \(\tau_i\) and \(\tau_f\) values measured for different carbon coatings can be quite large. A survey was made of some commercially available carbon-coated fibers and of fibers made by varying the processing conditions on a single draw tower.

The results showed that most of the fibers had very good hydrogen-blocking properties, with \(\tau_fs\) (at 150\(^\circ\)C) ranging from 100 days to more than 10,000 days.

The range of activation energies within this grouping of fibers was fairly tight: 84–105 kJ/mol (20–25 kcal/mol). Values of \(\tau_f\) at room temperature for this group of fibers were from 8 x 10\(^6\) to 2 x 10\(^9\) days.

Some of the experimental fibers, though having black carbon coatings of normal appearance with high \(n\) values, had significantly lower \(\tau_f\) values. The \(\tau_f\) values for these fibers ranged from 1 to 8 days at 150\(^\circ\)C and 2000 to 30,000 days at 21\(^\circ\)C.

The activation energies for these low \(\tau_f\) fibers were 63–67 kJ/mol (15–16 kcal/mol). The lower \(\tau_f\) and activation energy values suggested that the carbon layers for these fibers were different in structure or below a critical thickness. The initial lag times, \(\tau_i\), increased as \(\tau_f\) increased. This was the expected behavior because both \(\tau_i\) and \(\tau_f\) increase as the coating thickness (\(\delta\)) increases and as the diffusivity of H\(_2\) in carbon coating (\(D_c\)) decreases.

 

5.5. Effects of Glass Composition on Hermetic Fiber Behavior

The diffusion analysis in the previous section assumes that H\(_2\) molecules diffuse inertly through the fiber’s silica cladding. It also assumes that there are no sources of hydrogen in the fiber’s cladding material.

Most of the glass in a single-mode fiber is made up of the undoped SiO\(_2\) that is outside the fiber’s core. The properties of this optically inactive glass can vary and depend on the materials and the manufacturing processes used in making the fiber.

In some cases, silica glasses derived from the fusion of natural quartz crystals can contain metastable OH impurities, which can later convert into mobile forms of hydrogen that can diffuse into the core and cause loss.

When fibers made using such glasses are hermetically coated, the hydrogen impurities are trapped under the hermetic coating. At high temperatures, the hydrogen can be liberated from the cladding glass. It can then diffuse into the fiber core and cause hydrogen-related loss increases such as OH growth or SWE increases.

The magnitude of the trapped-hydrogen losses will depend on time, temperature, wavelength, and the composition of both the core and the cladding glasses.

Although the loss increases due to trapped hydrogen may be acceptably small for some applications, it is best to avoid fibers made with these natural fused-quartz glasses when a hermetic coating will be applied to the fiber.

Other types of silica have the opposite effect and can scavenge trace levels of hydrogen. When the cladding of a hermetic fiber is made of silica that contains reactive ‘‘gettering’’ sites, the fibers can show improved resistance to hydrogen losses.

The small amounts of H\(_2\) that diffuse through the coating quickly react with the gettering sites and are immobilized in the cladding. Drawing-induced silica defects are known to exist at concentrations of about 100 ppb in some types of undoped silica. These defects react quickly with H\(_2\), even at room temperature, and thus provide a practical secondary hydrogen barrier to the hermetic coating.

The duration of this gettering-induced lag time in a hermetic fiber, \(t_{gh}\) is

\[\tag{14.11}t_{gh}=\frac{\tau_fC_gf_g}{P_{H_2}(ext)K_s}\]

where \(C_g\) is the concentration of gettering sites and \(f_g\) is the fraction of the silica material that contains the gettering sites.

When coupled with a good hermetic coating, the reactive sites can greatly increase the duration of the lag period (i.e., the period where loss increases are zero). For instance, assuming coating properties like those in Fig. 14.12, 100 ppb of reactive sites will result in a lag time of about 7.5 years at 100\(^\circ\)C and \(P_{H_2}=0.01\) atm. The normal lag time without reactive gettering sites, \(\tau_i\), would be about 2 days at this temperature.

The protective effect provided by the reactive silica material will remain until the reactive sites are depleted. The duration of this extended lag period, therefore, depends on the concentration of reactive sites in the glass, the coating properties (i.e., the value of \(\tau_f\)), and the external hydrogen pressure.

For a hermetic fiber that does not contain reactive sites, the length of the lag period will be \(\tau_i\), which does not depend on hydrogen pressure. For a hermetic fiber that does have reactive sites in the cladding, the length of the lag period will be inversely proportional to the hydrogen pressure used in the accelerated test experiment (or to the ambient hydrogen pressure seen by the hermetic fiber in actual use).

Whether a hermetic fiber contains a useful concentration of reactive sites is most easily determined by monitoring the in situ loss changes at two different hydrogen pressures. For example, if a fiber is tested under 1.0 atm and 0.1 atm H\(_2\) pressure, the lag period should be approximately 10 times longer for the fiber tested at 0.1 atm. Similar lag periods at both pressures indicate that reactive sites are either absent or low in concentration.

The figure of merit for hermetic fibers that contain reactive sites is the product \(\tau_fC_g\), where \(C_g\) is the concentration of reactive sites in the cladding. The length of the reactive lag time is proportional to the quantity \(\tau_fC_g/P_{H_2}\).

Once the lag period, \(t_{gh,exp\;t}\), is measured in an accelerated experiment, the results can be used to estimate the duration of the lag period at the anticipated use condition, \(t_{gh,use}\):

\[\tag{14.12}t_{gh,\;use}=t_{gh,exp\;t}\frac{P_{exp\;t}}{P_{use}}\exp\left[\frac{(E_f-E_s)}{R}\left(\frac{1}{T_{use}}-\frac{1}{T_{exp,\;t}}\right)\right]\]

where \(P_{exp\;t}\) and \(P_{use}\) are the hydrogen pressures for the experiment and for the use condition, \(E_f\) is the activation energy in the exponential term that characterizes \(\tau_f\) (e.g., 99.21 kJ/mol for the fiber of Fig. 14.12), and \(E_s\) ( 8.67 kJ/mol) is the term accounting for the decreasing H\(_2\) solubility with temperature.

For example, if an experimental lag time of 21 hours was observed at test conditions of 150\(^\circ\)C and 1.0 atm of H\(_2\), the expected reactive lag period would be 6 years for assumed use conditions of 75\(^\circ\)C and a hydrogen pressure of 0.1 atm.

While there can be significant differences in \(E_f\) values for different carbon-coating types, the value is often in the range of 84–105 kJ/mol for high-quality carbon coatings. Ideally the value for \(E_f\) should be determined experimentally for a new coatings or for coatings with unknown properties.

 

6. Use and Handling of Carbon-Coated Hermetic Fibers

Carbon-coated fibers can be worked with in much the same way as standard nonhermetic fibers. However, there are a few differences, as discussed in the following section.

 

6.1. Fiber Strength

As mentioned earlier, the fracture strengths of carbon-coated fibers are lower than those of standard nonhermetic fibers. The root cause of the strength reduction is at least partially understood, and it is reasonably clear that some tradeoff in fiber strength is necessary to achieve improved hermetic properties, especially with respect to hydrogen diffusion.

Nonetheless, fibers with high strength and good hydrogen properties can be made. For instance, the carbon-coated fibers whose hydrogen data are shown in Fig. 14.12, had reasonably high strengths of about 4.1 GPa (600 ksi) and were capable of blocking hydrogen diffusion over a wide range of conditions.

Higher strength carboncoated fibers can be made, but care should be taken to verify that their hermetic properties are adequate for the application.

 

6.2. Fiber Handling

Bend radii guidelines for nonhermetic fibers are determined by fiber fatigue considerations. Typical fiber handling practice for nonhermetic fibers requires that the long-term stresses imposed on a fiber be limited about 20% of the prooftest level.

Because a carbon-coated fiber is virtually immune to fatigue, it is permissible to use tighter bend radii when storing fibers on spools or in splice trays. One application note on hermetic fibers allows the fiber to be used at 80% of its proof-test value.

In this case, the specification of a value less than 100% was done to allow for less-than-perfect control of the stress values seen by a deployed fiber.

For a carbon-coated fiber proof-tested at 0.7 GPa (100 kpsi), the allowable stress would, therefore, be 0.55 GPa (80 kpsi), corresponding to a minimum bend radius of about 8 mm. The corresponding value for a nonhermetic fiber would be about 33 mm.

Different manufacturers may provide different handling guidelines for their hermetic fibers.

 

6.3. Fiber Stripping, Cleaving, and Connectorization

Polymer coatings can be stripped using normal fiber stripping tools. The carbon coating itself will usually adhere well to the glass fiber and, therefore, will not be removed by the stripping tool.

Once the polymer coating is removed, the fiber can be cleaved using a conventional fiber cleaver.

Because the carbon coating is very thin, the overall fiber diameter of a stripped fiber will be very close to the standard 125-μm size, allowing the fiber end to be inserted and bonded into standard fiber ferrules.

In general, the only reason to remove the carbon coating is for fusion splicing, described in the next section.

 

6.4. Fusion Splicing

Carbon-coated fibers can be fusion spliced using the same commercial splicers that are used for nonhermetic fibers. Both the polymer coating and the black carbon coating should be removed from each fiber end in preparation for splicing.

The thin carbon layer is best removed using a pre-fusion arc. This ‘‘burn-off’’ of the carbon coating allows the optics in an automatic fusion splicer to ‘‘see’’ the fiber core and align the two sides.

Once the carbon is removed from each end, standard fusion splicer programs can be used to splice the fiber. A proof-test should be done on the splice region, identical to the way that a nonhermetic fusion splice would be tested for strength.

In principle, it is desirable to have some way to reapply the carbon coating in the splice region. Although some work was done to develop methods for local deposition of carbon onto bare fibers, the equipment for carbon recoating is not readily available.

Further, it is rarely necessary to reapply the carbon on the short, approximately 10-mm, section of spliced fiber. Hydrogen losses on such a short section of fiber are negligible. H\(_2\) molecules that diffuse into this section of fiber will be confined to the splice region because lengthwise diffusion away from the unprotected splice region will be slow.

Mechanical fracture of the unprotected splice region can be avoided by use of a standard splice protector or by simply avoiding the application of high stresses to this region.

It is important to note that the minimum bend radius for the spliced section of hermetic fiber will be the same as for a nonhermetic fiber unless the carbon coating is reapplied.

Fusion splicers that are used to splice carbon-coated fibers may require more frequent cleaning, particularly of the electrodes. Some users have found it advantageous to have a separate piece of equipment that is used to remove the carbon from the fiber ends before fusion splicing, helping to reduce extra maintenance on the fusion splicer itself.

An example of such equipment is a small high-temperature oven that heats the fiber end to a temperature at which the carbon reacts with oxygen from the air, vaporizing the carbon as CO\(_2\).

 

6.5. Fiber Color

Semi-opaque inks are commonly used to color-code and identify optical fibers. These fiber colors are applied as thin coatings (several micrometers) on top of the acrylate polymer coatings.

The black color of a carbon coating can affect the appearance of a color-coded fiber, causing the fiber to appear to be darker. This may require that special color charts be used for hermetic fibers or can require that certain colors be avoided to prevent misidentification.

 

7. Specifying Carbon-Coated Fibers

When specifying the properties of a carbon-coated fiber, one must remember that different carbon-coated fibers have different properties and that no coating will be 100% impervious to all species, especially small molecules like H\(_2\)O or H\(_2\).

For instance, it is possible to have a fiber with a layer of carbon of normal appearance that is nonetheless as susceptible to hydrogen loss as a nonhermetic fiber. Therefore, some level of testing needs to be conducted to show that a fiber is adequately hermetic.

The properties of carbon coatings are known to be sensitive to the processing conditions used in drawing and coating the fiber. This sensitivity may require that individual lots of fibers be tested to verify that they meet minimum requirements for resistance to fatigue and/or hydrogen sensitivity. The extent of the testing depends on how sensitive the coating properties are to normal variations in manufacturing parameters.

A fiber manufacturer will typically have qualified a hermetic fiber during its development. Qualification experiments will have been used to show that the fiber and its coating are capable of limiting fatigue and/or hydrogen-induced loss increases.

It is desirable that the qualification experiments be done under accelerated test conditions. For characterizing a fiber’s resistance to fatigue, accelerated test conditions will be elevated temperatures and humidity levels. For characterizing a fiber’s resistance to hydrogen aging, the accelerating factors will be elevated temperature and hydrogen partial pressure.

Qualification tests tend to be time consuming, often requiring weeks to months to complete, making them expensive to conduct and not repeated for each lot of manufactured fiber. However, because the properties of hermetic coatings depend on the details of the processing steps (e.g., draw conditions), it is common to conduct certification tests as part of the manufacturing process.

In a certification test, a subset of tests is conducted to verify that the fiber meets some minimum standards. Examples of certification tests for carbon-coated hermetic fibers are room-temperature measurements of the dynamic fatigue parameter \(n_d\) and measurement of fiber loss before and after exposure to hydrogen at an elevated temperature for a designated time.

A primary question is whether the fiber needs to be protected from fatigue or from hydrogen-induced optical loss. In some cases, both types of protection are needed. A coating that is capable of blocking H\(_2\) diffusion will generally be impervious to H\(_2\)O diffusion and fatigue.

The converse, however, is not true. There are carbon coatings that can effectively eliminate fatigue but that are not cable of blocking H\(_2\) diffusion. Improving a coating’s H\(_2\) blocking ability may have an adverse effect on the fiber’s fracture strength, so overspecifying H\(_2\) resistance may not be advisable.

The ability of a carbon coating to prevent fatigue is usually ensured by showing that the fiber has a high \(n\) value. A high \(n\) value means that the fiber’s fracture stress is insensitive of strain rate or the time that the fiber is under load. Typical \(n\) values for nonhermetic fibers are on the order of 18–22. The value of \(n\) for a hermetic fiber will typically be determined using dynamic fatigue measurements. Good hermetic carbon coatings usually have \(n\) values that are 80 or higher.

Values much in excess of 100 become sensitive to measurement error and can be misleading. For instance, a carbon-coated fiber with a nominal \(n\) value of 250 should not automatically be assumed to be better than one with an \(n\) value of 150 unless significant care has been taken in the data acquisition and analysis.

Dynamic fatigue measurements should be obtained using at least three strain rates. In most cases, dynamic measurements are done at room temperature. Obtaining fatigue data at elevated temperature and humidity is most readily done via static fatigue measurements. However, because static fatigue measurements are time consuming, they are not commonly used as certification tests.

A primary engineering tradeoff that needs to be made in improving the hermetic properties of a carbon-coated fiber is that of initial fiber strength. The general trend is that a carbon-coated fiber’s strength will decrease as the hermetic properties improve.

It is reasonable to expect fiber strengths of about 4.1 GPa (600 kpsi) for carbon-coated fibers that have excellent fatigue resistance (\(n_d\) > 100) and very low hydrogen permeability (\(\tau_f\) ~ 10\(^6\) days at 50\(^\circ\)C).

Significant improvements in hydrogen blocking ability can be achieved if fiber strengths of 2.8–3.4 GPa (400–500 kpsi) are acceptable. In almost all cases, a reasonable compromise between fiber strength and hermetic properties can be reached for a given application.

Showing that a hermetic coating can limit hydrogen losses requires that the fiber be tested in a controlled hydrogen environment and that loss changes be below some critical value.

While a full characterization of the coating (i.e., \(\tau_i\) and \(\tau_f\) as functions of temperature) is advantageous, it is not always feasible for a fiber manufacturer to provide such detailed data.

A more common procedure is to test a hermetic fiber in hydrogen for a given time and temperature and show that loss increases are below some specified value at a designated wavelength. Ideally, the basis of the test (i.e., the rationale for the pass–fail criterion) will be provided by the fiber’s manufacturer.

Alternatively, the passing criterion for such a test can be set using the approximation shown in Eq. (14.13). Equation (14.13) is based on Eq. (14.7) and is derived using the approximation that \(\exp[-(t-\tau_i)/\tau_f]\sim{t}/\tau_f\) for small values of \(t\), and assuming that \(\tau_i\) is short in comparison to \(t\).

This expression is appropriate when losses due to molecular H\(_2\) are expected to be the dominant loss change. The predicted change in loss (dB/km) at system conditions, \(\Delta\alpha_{sys}\), is calculated based on measured results in an accelerated experiment.

\[\tag{14.13}\Delta\alpha_{sys}=\Delta\alpha_{exp\;t}\frac{t_{sys}}{t_{exp\;t}}\frac{\alpha_{sys}}{\alpha_{exp\;t}}\frac{P_{sys}}{P_{exp\;t}}\exp\left[\frac{(E_f-E_s)}{R}\left(\frac{1}{T_{exp\;t}}-\frac{1}{T_{sys}}\right)\right]\]

The \(sys\) and \(exp\;t\) subscripts refer to the deployed system and to the accelerated experiment, respectively. \(\Delta\alpha_{exp\;t}\) is the loss change (dB/km) measured in the experiment. The parameters \(a_{sys}\) and \(a_{exp\;t\) are the pre-exponential terms from Eq. (14.8) and express the H\(_2\) absorption strength at the wavelengths of interest. \(P_{sys}\) and \(P_{exp\;t}\) are the hydrogen partial pressures for the system conditions and for the experiment.

In most cases, the molecular H\(_2\) peak will be monitored during the experiment and \(a_{exp\;t}\) will be the value for 1240-nm peak (0.355 dB/ km-atm). Similarly, a typical value for \(a_{sys}\) is 0.017 dB/km-atm, corresponding to a system wavelength of 1550 nm. The exponential term accounts for the temperature dependence of \(\tau_f\) and of H\(_2\)’s solubility in silica.

The acceptable loss change in an experiment, \(\Delta\alpha_{exp\;t}\), is determined by calculating the predicted system loss change, \(\Delta\alpha_{sys}\) and verifying that it is less the value allocated for system aging. If it is not, the acceptable value for \(\Delta\alpha_{exp\;t}\) needs be lowered, which in turn will usually require an improved fiber coating.

If the pass–fail criterion is ‘‘no’’ loss change, then \(\Delta\alpha_{exp\;t}\) should be the minimum detectable loss change, which will depend on the sensitivity and stability of the loss test set and on the length of the fiber under test.

The only term in Eq. (14.13) that might not be known is \(E_f\), the activation energy characterizing \(\tau_f\)’s dependence on temperature. This value needs to be established during the qualification process of a new coating because it has a major influence on the predictions.

For the carbon coatings described earlier, \(E_f\) ranged from 84 to 105 kJ/mol. Note that Eq. (14.13) is not appropriate for predicting loss increases if the system losses are dominated by hydrogen reactions, for instance, by OH formation or reaction of H\(_2\) at Ge sites in the core. These reactions become increasingly important factors at high temperatures.

When a fiber is used at an elevated temperature, or when it has a core composition that is known to be reactive with hydrogen, there are several approaches that can be used to qualify the hermetic fiber.

The most thorough approach is to fully model the hydrogen diffusion through the coating and its reaction in the fiber’s core. The loss change is integrated over time, accounting for the gradually increasing level of hydrogen inside the fiber.

This requires detailed knowledge of the reaction kinetics for the particular fiber core composition, values for \(\tau_i\) and \(\tau_f\), and information about the concentration of gettering sites in the cladding.

When done correctly, this approach results in an accurate prediction of loss change as a function of time, temperature, \(P_{H_2}\) and wavelength. However, the amount of work entailed can be significant, and for this reason such studies are carried out only when necessary.

A more practical approach is to separately determine a tolerable partial pressure of H\(_2\), \(P_{tol}\), such that loss changes in a nonhermetic version of the fiber would be suitably small at the operating conditions.

The hermetic coating then needs to ensure that the internal hydrogen partial pressure, \(P_{int}\), stays below this critical \(P_{H_2}\) value. A modification of Eq. (14.13) permits the internal H\(_2\) pressure to be predicted on the basis of an experiment that monitors changes, \(\Delta\alpha_{exp\;t}\), in an H\(_2\)-related loss feature (e.g., the 1240-nm peak):

\[\tag{14.14}P_{int}=P_{sys}\frac{\Delta\alpha_{sys}(t)}{\Delta\alpha_{sys}(\infty)}=\frac{\Delta\alpha_{exp\;t}}{\alpha_{exp\;t}}\frac{P_{sys}}{P_{exp\;t}}\exp\left[\frac{(E_f-E_s)}{RT_{exp\;t}}-\frac{E_f}{RT_{sys}}\right]\]

The value for \(P_{int}\) is calculated for \(t_{sys}\) equal to the system lifetime and using the observed loss change, \(\Delta\alpha_{exp\;t}\), from an accelerated test. As long as the value calculated for \(P_{int}\) is less than the tolerable \(P_{H_2}\) value, \(P_{tol}\), the carbon coating will provide a useful barrier.

Yet another alternative is to show that the net lag time, due to reactive sites in the cladding, (\(t_{gh}\)) and due to the diffusion lag (\(\tau_i\)), is greater than the system lifetime. For instance, Eq. (14.12) can be used to calculate the expected lag time, \(t_{gh}\), caused by reactive sites in the cladding.

If this lag time is longer than the system lifetime, the combination of the coating and the reactive sites is sufficient to protect even a highly reactive core composition. For the example given immediately after Eq. (14.11), the lag time attributable to reactive cladding sites was predicted to be 7.5 years at 100\(^\circ\)C and \(P_{H_2}=0.01\) atm, sufficiently long for many applications.

If the reliability of a carbon-coated fiber depends on the presence of reactive gettering sites in the fiber, periodic hydrogen tests should be conducted by the fiber manufacturer to verify the \(\tau_fC_g\) figure of merit for the carbon-coated fibers. The concentration of the reactive sites in the glass is not readily measured by other means.

As previously discussed, the electrical resistance of a carbon coating can be used to measure carbon-coating properties and as an indirect measure of coating thickness and quality.

If qualification experiments show a good correlation between a fiber’s electrical resistance and its fatigue and/or hydrogen properties, then the measurement of electrical resistance can be used as a valid quality-control parameter. There is limited value in comparing electrical resistance measurements when comparing fibers made by different methods or by different manufacturers.

Some types of carbon can have low electrical resistance (suggesting a thick carbon layer) but can still have poor hermetic properties. The material properties of carbon layers deposited on fibers depend on the details of the gas reactants used and the temperature of deposition.

Because these factors will not be the same for different manufacturers but will influence the coating’s electrical resistance, the comparison of resistance values across fiber types is of limited value.

On the other hand, used as a quality-control metric for a stable coating deposition process, the electrical resistance can be a useful parameter.

 

8. Applications for Carbon-Coated Hermetic Fibers

8.1. Fibers in Underwater Cables

Fibers in underwater cables have sometimes exhibited optical loss increases due to the evolution of hydrogen gas. Failures have been due to H\(_2\) gas evolved from the galvanic corrosion of metallic components in the submerged cables or from outgassing of H\(_2\) from certain types of silicone-based polymers.

In most cases, these problems were solved by redesigning the cables to avoid dissimilar metals and by avoiding the use of hydrogen-generating polymers. However, hermetic carbon coatings were successfully used in some cables and were shown to be a viable solution to the hydrogen-aging problem.

One advantage of using carbon-coated fibers to protect against hydrogen is that it avoids the expense and delay associated with the redesign and qualification of a cable. The low fatigue properties of carbon-coated fibers are an additional factor influencing the decision to use carbon-coated fibers in underwater cables.

Because the typical temperature of an underwater cable is low (3–20\(^\circ\)C), hydrogen reactions in the fiber core will be negligible for most types of single-mode fibers. The only losses that are likely to affect the fibers are those due to molecular H\(_2\).

The low temperatures also result in slow hydrogen diffusion through the carbon coatings. For instance, the results for the carbon-coated fiber shown in Fig. 14.13 at 10\(^\circ\)C are \(\tau_i\) = 30 years and \(\tau_f\) = 8 x 10\(^5\) years. This fiber would be immune to H\(_2\) losses over times much longer than a typical 25-year lifetime.

Equation (14.13) can be used to estimate long-term loss changes in underwater cables. The exact hydrogen level in the cable will usually not be precisely known.

Nonetheless, because the H\(_2\) permeation through most carbon coatings is very slow, it is usually possible to assume highly pessimistic values for the hydrogen pressure inside the cable and still justify very low hydrogen aging losses over the system lifetime.

Likewise, even though carbon coatings with low \(E_f\) values may not be adequate for high-temperature applications, they may be perfectly adequate for the low temperatures encountered in underwater cables.

 

8.2. Amplifier Fibers

Carbon coatings have been used to protect erbium-doped amplifier fibers. These fibers play a critical role in commercial telecommunications and in specialty applications.

In their nonhermetic version, these fibers have the same bending limitations as other fibers. This limits the physical size of amplifier modules because the bend radii used in winding and routing the fiber cannot be less than a critical value, typically about 33 mm. By using hermetic carbon coating, the problem of fatigue is eliminated, allowing tighter fiber coils and more compact amplifier modules.

The carbon coating also protects the Er-doped fiber from hydrogen aging. Many Er-doped fibers use core compositions that are highly reactive with hydrogen. A hermetic carbon coating is one way to protect the fiber from loss increases that could otherwise occur due to unexpectedly high H\(_2\) levels.

 

8.3 Avionics

The enhanced mechanical reliability of carbon-coated fibers also makes them attractive for avionics applications, where restricted spaces can require that fibers be routed with tight bend radii.

In addition, the fibers see significant environmental stresses due to cycling between temperature and humidity extremes. Although these harsh conditions can cause fatigue-induced failures in nonhermetic fibers, carbon-coated fibers will retain their strength even in the presence of high humidity and stress levels.

Carbon-coated fibers are preferable to metal-coated fibers in these applications both because of their low optical loss and because they are significantly lighter than metal-coated fibers.

 

8.4 Geophysical Sensors

Arguably the most challenging environment for specialty optical fibers is the environment encountered in oil-well down-hole data logging. Fibers used in sensor systems in oil wells can be exposed to temperatures up to 300\(^\circ\)C, high pressures, water, corrosive chemicals such as H\(_2\)S, and high levels of H\(_2\) gas.

The sensor fibers are often installed into stainless-steel conduits using high-pressure water or other liquids. For the purpose of strength retention and chemical resistance, the carbon coating plays a vital role.

When H\(_2\) gas is present at the high temperatures encountered in the down-hole environment, the optical aging of the fibers can be rapid. The temperatures encountered in the application of these fibers can be comparable to the highest temperatures used in accelerated aging experiments.

At these temperatures, H\(_2\) diffusion becomes rapid and the dopants in the fiber core become increasing reactive with hydrogen. The losses due to reacted hydrogen can greatly exceed those due to unreacted molecular H\(_2\).

The extent to which H\(_2\) molecules react in the fiber core to form lossy OH and Ge defects is highly dependent on the core glass composition. The use of a carbon coating with low H\(_2\) permeability can greatly extend the useable lifetime of fibers used for data logging.

Expressions such as Eq. (14.13) are generally not useful for predicting the loss changes in a fiber used in a high-temperature environment because the only losses that are accounted for are those due to unreacted molecular H\(_2\).

An alternative approach is to identify a carbon-coated fiber and show that its properties are sufficient to give a reliability advantage in the actual field environment. Once the desirable fiber parameters are identified, they need to be tightly controlled in the manufacturing process.

By establishing an appropriate certification test, where the loss changes are shown to be below a critical limit, it is possible to detect changes in the manufacturing process that could adversely affect the fiber reliability.

As discussed earlier, it is possible to make carbon-coated fibers that have relatively slow H\(_2\) permeation even at high temperatures. When a fiber must survive H\(_2\) exposure in a high-temperature down-hole environment at 150\(^\circ\)C, a carbon coating that has a \(\tau_f\) of 10,000 days will be a better choice than a coating with \(\tau_f\) in the 1- to 100-day range.

The increased hydrogen protection might require a tradeoff in fiber fracture strength of up to 0.7 GPa (100 kpsi), but this is likely to be acceptable in view of the improved optical lifetime.

Using low-dopant low-reactivity fiber core compositions and increasing the concentration of reactive sites in the cladding can also improve the reliability of a fiber exposed to hydrogen at high temperatures.

 

9. Conclusion

Carbon-coated optical fibers play an important role among specialty optical fibers. The thin carbon layer provides a hermetic barrier to water and hydrogen without inducing microbending losses.

A well-designed carbon coating stops water from diffusing to the silica surface of the fiber and, thus, prevents latent fiber fractures associated with fatigue. Because of this, carbon-coated fibers can be configured with tight bend radii, without the risk of fatigue-induced fracture.

Carbon-coated fibers can also be designed to prevent hydrogen-induced loss increases, an effect that can be problematic for fibers used in aggressive applications such as sensor systems in oil-well data logging.

The properties of carbon-coated fibers depend on the properties of both the carbon and the underlying glass, which in turn depend on the processes used in making the hermetic fiber.

By establishing suitable qualification and certification tests, we can ensure that carbon-coated fibers will meet a user’s requirements for providing enhanced long-term fiber reliability. 

 

 

The next tutorial discusses laser oscillation dynamics and oscillation threshold.