Open Access
Add to BibSonomyAbstract
A process leading to a stunningly beautiful and distinctive propagating plasma emission in optical fibers was discovered by the author 25 years ago. The genie that escaped its glass bottle leaves a trail of destruction. This paper traces the history and impact of the effect, which can threaten the security of all modern communication systems.
©2013 Optical Society of America
1. Introduction
Optical fibers form the backbone of all communication systems. Their ubiquitous use was clearly described recently in a presentation by Bergano [1], in which the author described how heavily all undersea cable systems and long haul optical links form the framework on which modern communications depend: satellite, wireless and optical. Our world could not function properly without the use of optical fibers even for minutes, and yet their security is fragile, compromised by the material nature of the glass used to fabricate them. It would be unimaginable if suddenly parts of the network were accidentally disabled, and required replacing on a timescale of days - which is what could happen without adequately protecting the most vulnerable element in communications – optical fiber. Twenty-five years ago, the author discovered a phenomenon [2], which revealed how light used for communications is a double-edged sword: it’s enormous benefits, is only available as a package deal: an implicit pact, much like Goethe’s Faust [3].
This paper revisits a discovery of low power optical damage in waveguides. The threat of impending disaster is constantly there when using high power optical transmission systems, especially since the advent of amplified systems and wavelength division multiplexing (WDM) has pushed the power carrying capacity of optical fibers upwards and well into the hazardous and unsafe operation regimes. The question is: which threat is worse- a severe cyber-attack or a seriously disabling event based on this damage mechanism, which may take days to recover? The observation, our current understanding and measures possible for protection are thus covered in this article. Curiously, most researchers trying to propose a theory for this effect continue to ignore the fact that oxygen is released and remains under high pressure in cavities, as pointed out by the author some 25 years ago.
2. Discovery of the low power optical damage
The Fiber Fuse, as it has come to be known is a damage mechanism observed in waveguides, especially optical fibers, which almost always causes irreversible destruction of the core region, and sometimes also to the cladding. Indeed, transmitting high power in optical fibers is fraught with difficulties, as it is well known that bending the fiber, which causes leakage loss can cause the polymer coating to catch fire, as a result of absorption. Thus, it is important to handle optical power carefully, since the intensity of light can be very high in the core. Low power optical damage [2] was discovered at ~2am on Tuesday the 14th of July 1987 almost by chance when, during the course of experiments on second harmonic generation (SHG) in optical fibers at BT Research Laboratories. High voltage was applied to a side polished fiber in the form of a “half-coupler block” (HCB), via a spatially periodic metal electrode. An average power of ~1 W, Q-switched (QS) from an Nd:YAG laser was used to observe electric field induced SHG. Unfortunately, a blue-white plasma like emission was initiated at the HCB, which terminated the experiment [4]. The plasma emission travelled the meter of fiber in less than a second, and close examination of the fiber began the discovery of this eerily silent, visually stunning, yet devastating phenomenon. This was unprecedented, as it was assumed till then that the fiber was immune from interaction with electromagnetic radiation.
3. Observations and developments
The surface of the embedded fiber in a half-coupler-block (HCB) is shown in Fig. 1 . The metal from the electrodes had come in contact with the evanescent field outside the core heated it and caused the damage initiation. The trail of destruction, manifest as bullet-shaped features, indicated that some kind of self-focusing damage had occurred leading to the coining of the process as “self-propagating self-focusing damage” (SPSF) by the author. Figure 2 shows the tracks of self-focusing plasma generation as observed by Maker et al. [5] in air, and which has also been investigated by others [6]. The filament first moves forward before retracing its trajectory, leaving damage voids behind. The principle differences are that the self-focusing filament has a velocity of ~105 m/s with an optical intensity for initiation of 100GW/cm2 whereas my observations showed a velocity of a ~0.5-1m/s and with an intensity of only 5MW/cm2! Interestingly, the ratios of intensities and velocities for the two effects are within an order of magnitude of each other.
Fig. 1 A magnified photograph of the HCB used in the first observation of propagating damage in optical fibers. The dimensions of the bullet marks: ~3 microns in the vertical direction by ~10 microns long (Note: Vertical scale is different to horizontal scale).
Fig. 2 Self-focusing induced plasma in air. The velocity of propagation is 105 m/s and the optical intensity is ~1011 W-cm−2. Adapted from Ref [7].
The difference in our case was that the damage travelled backwards. The original movie of the optical damage is shown in Fig. 3(a) . The heating was so severe, that on reaching the input end, the propagating plasma blew out the glass in the core with tremendous force, causing a hole to remain and the ejected glass to form a long frozen whisker attached to the end-face (Fig. 3(b))! This dramatic event could have been easily missed since the plasma was short-lived, traversing the 0.5m of optical fiber in around half-a-second, as the fiber was not being observed directly! Thus, began a set of investigations into this peculiar effect.
Fig. 3 (a) Original movie of optical damage from 1987 (Media 1). (b) 125 micron diameter fiber, with a hole in the core after the damage terminated at the input end - adapted from Ref [8].
In 1987, it was immediately noted that this interesting process was heat-driven and could be initiated by heating the fiber [8]. Several hundred meters of BT optical fiber were enjoyably destroyed in replicating the effect and in detailed investigations to understand the mechanisms driving the process [9].
Several different techniques were tried to initiate the process, including one of the most successful – contact with cream-yellow enamel paint on a standard BT issue trolley, paper, metal surfaces, plastics, and photographic paper. In the following months, I was contacted by a researcher in Israel, who claimed that this process was started when his finger accidentally came in contact with the end of fiber carrying power from an Argon-ion laser! This is certainly not recommended as a scheme to initiate the damage, as it could lead to severe burns. Typically, the power required to start the damage was over 1W optical, although higher powers are desirable for reliable initiation. The fiber fusion jointing machine created the right environment to start the plasma, when struck while the fiber was carrying over a watt of power. The output end after such an event is shown in Fig. 4 .
Fig. 4 The output end of the optical fiber at the point where the fusion arc from a splicer was struck.
The large bubble formed close to the hottest part of the arc indicated sudden expansion of a gas. The size is mitigated by the closeness to the external source of heat – the fusion splicing arc. The subsequent cavities stabilize into a smaller more regular size (almost identical) form, since the external heat source is too far away to alter the temperature.
The propagation velocity of the plasma was measured using the simple scheme shown in Fig. 5 . Indeed, measurements of the velocity of propagation soon revealed that the damage velocity was dependent not on the absolute power in the core, but its density, see Fig. 6 , indicating that to melt a given volume of glass requires a minimum amount of energy, as will be discussed later.
Fig. 6 Velocity of plasma propagation vs power density. (From Ref. 9). Also shown are the three measured threshold values (a, b, c) below which the damage does not propagate. Fibers A, B & C are with 1064nm radiation, Fiber B: with 514nm radiation.
The curves of velocity of propagation for three fibers made at two different wavelengths (1064nm and 514nm) are shown in Fig. 6. The graphs indicate near identical slopes for all three fibers for the two very different wavelengths but with higher velocity for the shorter wavelength. The graphs also indicate a minimum power density, below which the plasma will not propagate. As we will see later, at some power density, the heat generated by the source is equal to the heat loss, and hence the plasma is unable to propagate, however it can remain stationary. The question raised was: What was the reason for this thermal runaway? There was no apparent damage to the outer cladding. However, the cladding region close to the core was certainly affected, as may be seen in Fig. 1, where it appears that the core region has extended into the cladding.
The clue lay in the observation, that industrial processing of glass uses its higher conductivity at elevated temperatures to initiate resistive heating in order to melt it without the supply of an additional external heat source. Since purified glass is an extremely good insulator, it offers ultra-low loss transmission as an optical fiber. This assumption, while true at room temperatures, is no longer valid at higher temperatures, and dielectric loss is dominated by the imaginary part of the dielectric constant, leading to increased absorption of light in the glass. Figure 7 shows the absorption loss of an optical fiber when heated to above 1000 °C. There is a dramatic increase in the optical absorption above 1050 °C. These measurements were made with low optical power in a short length of optical fiber inside a furnace.
Fig. 7 Transmission loss vs. temperature of various fibers: (1) Chalcogenide, (2) Fluoride, (3) Silica/Germania fiber [7]. Adapted from Ref [10].
In the comparison in Fig. 7 curve (3), Dianov et al. [10], shows my original data [9] for silica/germania optical fiber. The other two curves (1-2) in Fig. 7 will be discussed later in the Section 5. The absorption increases exponentially, and therefore any optical power propagating in the fiber is absorbed in a very short distance, increasing the temperature further. This is the thermal runaway process, which makes the fiber so vulnerable to electro-magnetic radiation, and no longer safe. The external heat may be removed once the damage starts, since the heat source is replaced by the propagating optical power in the core, which is now absorbed, causing heating. As the light is confined to the core, the heat continues to spread outwards from the core in all directions from the point of initiation. As the temperature rises and spreads closer to the source, the absorption point moves in that direction. It should be noted that the rise in temperature is extremely fast in this thermal runaway process, with little heat dissipation through conduction into the bulk of the fiber. Hence the temperatures rise beyond all expectations to thousands of °C, giving rise to the plasma like emission seen during this process.
One of the first experiments attempted immediately after the discovery, was to capture the damage formation using high speed photography (30k fps) in November 1987.
In order to photograph the morphing of the core in real time during plasma formation, microscope images were filmed as the plasma traversed the field of view. Unfortunately, the cost of hiring the camera at the time was several thousand British pounds for half-a-day, and it made it impossible to try this more than a few times. The only image that was captured by the camera at the microscope was a bright plasma emission, by triggering the camera using a photodiode a few cm away! The intense plasma overexposed the camera film, and attempts had to be abandoned before appropriate neutral density filters could be found. However, Todoroki [11], in a set of beautiful experiments years later, captured the image in real time. A frame of the video of the formation of the enigmatic damage [11], is shown in Fig. 8 .
Fig. 8 A frame of the video taken by Todoroki [11], showing how the long void self assembles into periodic voids. The void just being formed detaches itself from the long section, and freezes as the heat zone moves on. Note that the laser light is propagating from the left.
Since it was serendipity that discovered the effect, it could have been missed easily, had it not been for the IR radiation used for the experiments. With visible radiation, the plasma emission is masked due to scatter of visible light at the point of damage.
Immediately after its discovery, a video of the moving plasma was shown at the Eighth Quantum Electronics Conference at St Andrews in 1987; it turned out to be a huge attraction and became a highlight [2], much to the chagrin of the other presenters! Detailed papers were also published [9, 12], in which the cavities were shown to contain molecular oxygen, a remnant of the damage mechanism, see Fig. 9 . The release of oxygen indicates a reduction and not an oxidation process, implied by the “fiber fuse” coined at Southampton University by Philip Russell in 1988 [13], which indicates a burning process. The name was chosen for its striking resemblance, but is of course not an accurate description of the reduction process.
Fig. 9 (a) A micro-Raman measurement of the damage cavity, showing the presence of molecular oxygen’s vibrational mode [9] estimated to be under 4 atmospheres pressure. The release of oxygen is from a breakdown of the silica and Germania into sub-oxides, SiOx and GeOx. (b) Formation of a capillary after heat is applied to the cavities in the fiber.
The effect of the high pressure of the oxygen in the cavities can be seen in Fig. 9(b) when a region with cavities is heated with a fusion splicer. The cavities expand and coalesce to form a capillary. As expected, there is indeed a change in the refractive index of the core measured after the damage as shown by Dianov et al. [14], consistent with the much higher refractive index of sub-oxides of germanium as suspected earlier [8]. More recent results, kindly pointed out by one of the Reviewer, showed a 10 × increase in the refractive index from 0.002 to 0.02 [15], as seen in Fig. 10(a) . The authors report that approximately half the increase in refractive index is caused by compression and densification of the fiber material after the propagation of the plasma. Also reported was the formation of the bubbles in regions of high intensity of mode beat between the LP01 and LP02 modes propagating simultaneously in the core.
Fig. 10 Refractive index profile of the fiber before and after damage. The increase in refractive index is clearly differentiated and seen to be 10x in region 1 in (b), decreasing towards the laser in region 2 and 2 in (b). Adapted from Ref [15].
The resulting damage is in the form of a region of periodic bubbles followed by an expanded core region shown in Fig. 10(b). This periodic trace continues over a distance determined by the coherence length of the modal interference, a function of linewidth of the laser. When the fiber is subsequently heated, the refractive index change reduces dramatically, indicating a relief of compressive stress.
The next item for investigation was the temporal dependence for the cavity formation. Indeed, since the effect was concluded to be thermally driven, it was suspected that there would be a dependence of the damage on the nature of the input light – CW or pulses (mode-locked). The effect on the formation of the cavities on the pulse width is shown in Fig. 11 . In each case the input power was kept at 2 W average. When subjected to CW light the cavities appeared to be oblong and cylindrically symmetric, indicating some kind of thermal equilibrium. For the mode locked pulses (100ps and 190ps) there seems to be a dependenceon the pulse width; the shorter pulses forming shorter asymmetric cavities, while the longer mode-locked (ML) pulses form longer bullet shaped voids. The time dependence of the nonlinearity in the formation of the cavities is evident from their shape. For 1kHz 150ns Q-switched pulses, it was also noted at the time that the damage could be sustained over distances of ~100mm with as little as 200mW average [12]. It is rather surprising that with a pulse mark-space ratio of ~104, and 1ms between pulses, there is sufficient heat remaining to sustain the effect.
It has been difficult if not impossible to predict the formation of the cavities as a result of the damage. In the first instance, experiments to capture the periodic formation of the cavities was attempted by using the scheme shown in Fig. 12(a) with a transient digitizer and a photomultiplier tube (PMT). The fiber was fed through a capillary into an enclosure housing the PMT. A photodiode also placed close to the entrance of the fiber allowed a transient digitizer to be triggered as soon as the plasma entered the housing. This allowed a large number of signals to be recorded as the plasma plume traversed the 50mm diameter of the PMT. With suitable neutral density filters, signals were recorder and one such is shown in Fig. 12(b) [9]. The burst of signals as the plasma passed in front of the PMT shows a correlation between the periodic emission and the spacing of the cavities. While it was stated that this might be due to an artifact [16], the fact that the spacing correlates so well with the observed temporal signal is curious. This shall be discussed further later.
Fig. 12 The scheme used for the measurement of periodic emission from the plasma. b). The periodic emission correlated with the spatial distance between cavities in the fiber [9].
Hand and Russell suggested in their paper [13] that there was indeed the possibility of self-focusing due to the positive refractive index change with temperature in silica. Their simulations suggested a collapse in the mode diameter to almost half, thereby increasing the generation of heat and therefore temperature. The generation of the cavities was not discussed, however they did suggest the possibility of a temperature instability with large fluctuations in temperature. So far, the only proposed reason is the chemical decomposition (reduction) of the glass at elevated temperatures suggested by the author, based on the evidence of the release of oxygen shown in Fig. 9(a), through enhanced self-focusing. Indeed, Driscoll et al. [17], rejected this idea and alluded to the strange phenomenon of spontaneous bubble formation in optical fibers at relatively low temperatures (900-1000C).
These bubbles form in the presence of a temperature gradient, for example at the entrance to a furnace. They suggest an exothermic reaction at lower temperatures which causes the generation of heat ignoring the fact that very high temperatures achieved in the fiber, and it is difficult to imagine the exact chemical nature of glass and its constituent elements. Further, the simulations made so far do not take into account any exothermic reactions in the glass at high temperatures, which is the suggested cause for the effect [17]. When one combines this with the video pictures of the formation of the bubbles, one notices that the extended cavity (in Fig. 8) appears to start at the hottest region in the front of the plasma (See Fig. 14), and then coalesces into smaller ones at the tail end. If Driscoll et al.’s argument was correct then the bubbles would also form during the fiber drawing process, which is clearly not the case.
Just after the author’s paper reporting this discovery, Hand and Russell reported the phenomenon with their finding [13] along with a model of the temperature profile of the moving plasma, which they termed as a solitary shock wave. The results of their simulation based on heat input from the source and dissipation by conduction (ignoring radiation loss) is shown in Fig. 13 . They note that the temperature in the core could rise to several thousand degrees, and the thermal runaway led to a moving constant temperature profile whose details change with input energy, balancing heat input with heat loss.
Fig. 13 The temperature profile of the fiber in the inset, for different input power densities. T is the core and θ, is the cladding temperature. From Ref [13].
Our own simulations, which included radiative, convective and conductive heat loss and a fixed absorption coefficient, resulting in similar conclusions, was prevented from being published by management at BT in 1988 due to the potential damage (sic!) it could cause to the fledgling optical fiber industry. Our predictions showed a peak temperature of 10,000K in the core, based on the absorption data in Fig. 7. The simulations were fitted to the measured curves in Fig. 6. Additionally, our predictions also indicated a non-propagating damage threshold of 125mW for standard telecom fibers, easily confirmed by the ease with which optical fiber connectors are routinely damaged. At this power, the heat input is equal to heat loss, resulting in a non-propagating plasma. However, the formation of the plasma modifies the core as well as damages the fiber facet. Our paper was eventually published in 1996 [18]! A significant result was the temperature profile, made in 1988, now shown in Figs. 14(a) -14(b).
Fig. 14 (a) The temperature contour of a plasma in a fiber generated with a power density of 10MW-cm−2. with a lateral dimension of 125 microns. The total horizontal length scale is approximately100 microns. (b) Solitonic temperature profile of the plasma. Adapted from Ref [18].
The calculation for Figs. 14(a) and 14(b) was based on a conduction, both radially and longitudinally along the fiber, convective heat loss from the surface of the cladding, and radiation from the core-cladding boundary. The assumptions for the model include a lumped absorption factor (of between 0.040 – 0.094 mm−1 depending on the fiber) to average the temperature dependence, for all temperatures above 1100C. In addition, other room-temperature data used were: 1.4Wm-K−1, a thermal capacity of silica of 788 J-kg−1 K−1 and a density of 2660 kg m−3. Heat loss at the surface of a 120 micron diameter fiber was estimated to be 46 W m−2 K−1.
The energy emitted from the core of the fiber is restricted to wavelengths between 0.35 and 1.7 μm, which is the transparency window of silica, and at temperatures above 1000 C the fiber is assumed to be a Blackbody radiator. Calculations showed that radiation from the core is not significant, and the main mechanism for the heat flow is conductive heat spreading, even at high temperatures [18]. Most of the heating is adiabatic at high power densities; at low power densities, conduction plays a significant role, limiting the rise in temperature and thus limiting damage propagation. The modification of the fiber due to the damage resulted in a granted patent [19].
Recently Facão et al. [20] and Rocha et al. [21] presented a 2D heat flow model with similar results. Interestingly, the first very simple relationship proposed [7] gives an excellent idea of the temperature rise, if the velocity of propagation is known. The relationship, slightly modified assuming the affected mass is the entire length within the region of the mode-field diameter, then,
Where, Tr is the temperature rise, r is the mode-field radius, ρ is the density, cp is the specific heat capacity, and v is the observed velocity of propagation. k(r, λ) is a constant dependent on the mode-field radius and the wavelength, and W is the input power. For each fiber, the constant of proportionality will change depending on the wavelength and fiber parameters. The slopes vary with input power for the different fibers and wavelengths as seen in Fig. 15 .Fig. 15 Comparison between plasma velocity made by difference researchers. Adapted from Davies et al. [17], with the author’s annotations in color.
Davies et al. [22] had gathered data on the various measurements on this phenomenon, and their compilation for the velocity of damage propagation are shown in Fig. 15, in which they make a comparison between their own, the data from Hand & Russell, and the authors’.
Significantly, the very different slope of velocity propagation for Hand and Russell seems startling. Although they made no comment on this discrepancy, it is clear that the smaller mode volume for their fiber means lower heat dissipation due to smaller surface area, and hence a faster plasma velocity. There are small differences in slopes for the other fibers, due to the larger mode volumes in the respective fibers. This conclusion is supported by the observation made by Bufetov [7], which will be discussed later.
Davies et al. [17] also compare the different models, and conclude that Hand and Russell’s seems most appropriate and state that our original model based on self-focusing was not plausible. In order to set the record straight, the author wishes to spell out the original model, based on absorption of energy from the source due to the increased temperature in the core. It is reasonable to assume the core temperature follows the mode intensity profile as confirmed by the plasma intensity measured by Dianov et al. [23, 24]. Indeed it is clear that there is an increase in the refractive index of the core with temperature. It was also suggested at the time that the generation of the plasma may lead to an increase in χ(3). However, when the glass melts above around 1773 C, the temperature coefficient of the refractive index becomes negative, and this leads to de-focusing of the guided mode. Figure 16 shows the computed change in the refractive index of silica using published data [9, 12, 25(Chapter 10), 26]. What is clear is that the mode v-value more than doubles by the time the temperatures reaches 1773C, increasing the power density between two and four-fold.
Fig. 16 Idealized refractive index of the core (Solid red line) and the v-value (Long dashed black line) of a standard optical fiber as a function of temperature, assuming a constant positive refractive index coefficient below and a negative refractive index coefficient above 1773°C. The cladding refractive index (short dashed blue line) is assumed to remain constant.
Thus, a greater power density increases heat generation on a short timescale, before the temperature exceeds the melting point, where upon the refractive index undergoes a change in slope to negative. Since heat generation causes the temperature to continue to rise, the diameter of the core mode continues to increase until the mode is cutoff at around 3850C. However this exceeds the chemical breakdown temperature of glass releasing oxygen to form bubbles. These are thought to form into a shape determined by the history of heat generation (CW light, ML, QS etc.), due to an interplay with surface tension and heat loss. The observations are being re-visited and shall be reported elsewhere in the future. The departure from linear dependence on the power density only occurs close to threshold, where heat loss becomes a predominant factor in countering temperature rise. On the other end of the temperature spectrum, careful measurements have shown the Blackbody (BB) nature of the plasma emission. Dianov et al. [23, 24] measured the spectral emission of the plasma at various power densities, and showed remarkably good agreement with BB radiation.
Figure 17(a) shows the fit of the emitted spectra to the BB radiation also demonstrating the temperature at which this agreement occurs. Clearly, temperatures in excess of 10,000K are achieved albeit at much higher powers than predicted by the simple model [Eq. (1)]. Theplasma velocity was also found to be linearly related to temperature even up to the maximum temperature recorded, as shown in Fig. 17(b). According to Eq. (1), using the velocity and temperature data, one can calculate an effective mode radius for Fiber 1 to be 7.79 μm, instead of the actual mode radius of 2 microns, ~4 × larger.
Fig. 17 (a) Blackbody radiation and measured spectra of the plasma emission at various powers for fibers with mode diameters of 4 (Fiber 1) and 5.75 microns (Fiber 2). (b) Plasma propagation velocity and its corresponding plasma temperature. Adapted from Ref [23].
In a subsequent paper, Dianov et al. [27], used 250ns QS pulses to investigate the velocity of propagation of the plasma at very high powers (up to 3kW peak). Surprisingly, they found the velocity of propagation of the plasma at 3kW QS to be ~100 × faster than expected from previous observation. This indicates a different regime, one that may be described as a detonation like shock-wave, which travels at around half the speed of sound. This unexpected result requires further investigation as the physics of the phenomenon is not understood. However, one can speculate that the shock-wave generated at the high optical intensity, causes further heating through adiabatic compression and plastic deformation [28]. The considerable deformation reported by Dianov et al. [27] is consistent with the explanation I have suggested. Under these circumstances, one expects that as the optical intensity increases, the velocity of propagation will asymptotically approach the sound velocity. A complete model should also take into account the heat of plastic rupture of the glass.
Askins et al. [16] proposed a explanation for the formation of the cavities, based on Rayleigh instabilities in fluids, similar to the breakup of water into droplets when dripping from a tap. An interplay between the cohesive forces of water and the pull of gravity, and in the case of a fiber, it is the change in the viscosity close to the glass transition temperature, when the glass simply separates into small cavities. An example of this is shown in Fig. 18 [29, 30]. However, the reality is different, since the cavities are not voids but filled with gas – they contain gas under some considerable pressure (~4 atmosphere [9]), although it would be interesting to see what the pressure in the cavity is with higher optical powers. Therefore, their model remains phenomenological, and does not fully explain the observation.
The issue related to an increase in the absorption is clearly seen in Figs. 19(a) and 19(b), which is a calculation based on the generation of electrons as a result of chemical breakdown of the silica molecule into Si and O ions [31]. This absorption at high temperatures is more than sufficient to provide the heat necessary for optical damage propagation from a laser source. The model estimates the number of free electrons released as being >1022 cm−3 above 4500K.
Fig. 19 Absorption increase due to electron (plasma) generation in silica (a). (b) shows the contributions to absorption due to different mechanisms as a function of temperature. From Ref [31].
Three sources for absorption are also suggested: point defects, through the formation of SiO, and generation of the plasma. Sufficient absorption is necessary for the initiation of the fiber fuse, and once started, it is sustained by the heat generated through increased absorption. From Fig. 19(b), the formation of SiO at temperatures above ~1500C is seen.
Shuto et al. [31] also states that the heat generation goes through an instability, leading to a periodic fluctuation in temperature. This conclusion certainly supports the earlier observation made by the author on the periodic light emission from the propagating damage (See Fig. 12(b). It would be interesting to confirm Shuto’s calculations through independent experimental means.
Further, Abedin et al. [29], observed that the moving damage and the subsequent formation of bubbles produces a periodic reflection, which generates a Doppler shift in reflected light [30], shown in Fig. 20(b) , based on the experimental arrangement in Fig. 20(a).
Fig. 20 The experimental arrangement to observe backscattered light from the moving plasma (a). Periodic moving reflection from the cavities produces a spectra of Doppler shifted frequencies (b). From Abedin et al. [30].
This arrangement may also be used to detect the onset of damage and used to arrest itby turning off the laser source [29, 30]. Indeed it is clear from the data summarized by Dianov et al. [9],that the velocity of propagation of the damage is an inverse function of mode-field diameter, and shown in Fig. 21 .
As the mode diameter gets larger, the velocity decreases inversely at a given input power. This observation was clearly concluded early after the discovery of the phenomenon [9], and therefore led to the technique to limit the damage from the fuse by incorporating a small section of large mode area fiber or a tapered fiber section in the transmission link [32].
4. Models
Golyatina et al. [33] calculated the velocity and threshold values of the damage based on a 2-D heat conduction model, which gave a square-root dependence on the intensity. Their results concur with the model of Hand and Russell [13] close to threshold, assuming a damage threshold at a velocity of 0.1 m/s. The real threshold, a non-propagating plasma with vp = 0 m/s, in which the heat generation equals to conductive heat loss was accurately predicted by the author in a model based a 3D simulation of heat loss by all three mechanisms – conductive, convection and radiation, with conduction playing a principle role, as already discussed earlier [18]. Akhmediev et al. [34] also arrived at qualitative agreement with the Hand and Russell model. However, all the measured data presented so far indicates a linear dependence on intensity, as shown in Figs. 3 and 6. The author’s simulations show a linear dependence [18]. This may be possibly because the other models do not include all the heat loss mechanisms in operation, or because of the value of the diffusion coefficient assumed in the models.
5. Different fibers
Damage in fluoride and chalcogenide fiber was first reported by Dianov et al. [35]. The author noted in 1988 that damage in fluoride fiber does not behave the same way as silica fiber. This is due to the very low melting temperature in fluoride glass; the glass melts and loses guidance and then decomposes. In chalcogenide fiber, the effect leads to a thermal decomposition of the fiber, and the damage point moving towards the laser.
Dianov et al. [36] also reported the first observation of catastrophic damage in microstructured fiber at 9W. They observed a higher threshold for damage in these fibers, however, the catastrophic nature of damage is evident Fig. 22 . This also leads to a more recent and interesting observation of damage in in hole assisted optical fiber (HAF) by Hanzawa et al. [37]. They found that damage in hole assisted fiber could obliterate the holes, albeit periodically, as shown in Fig. 23(b) and 23(c). Cutting the fiber at different positions after damage shows how the azimuthal distribution of holes slowly disappear, periodically moving to the core region [37]. They found a regime in which the period of the bubbles became long when the diameter of the holes was increased. They reported that above a certain power (in this case above 3.93W), they could not observe any damage propagation, and that the fiber remained intact. It is unclear to which power the authors tested their fiber as this was not reported. The authors reported that work was continuing on the subject to elucidate the observations. It is hard to imagine that damage can be avoided in any optical fiber design, since basically heat will always cause damage.
Fig. 22 The end of a fluoride fiber after damage at a power of 0.5 W at 1064 nm wavelength (a). Adapted from Dianov et. al. [35]. SEM of microstructured fiber before damage (b). Damage sustained by 9W of 1064nm radiation (c). Adapted from Dianov et. al. [36].
Fig. 23 (a) shows the pristine HAF before damage. (b) shows the collapse of the holes into one central hole at a certain position in the fiber after damage. (c) shows an intermediate region. Adapted from Ref [37].
Recently, Jeong et al. [38] investigated optical damage propagation in a piece of hollow-core optical fiber, concluding that a short piece of it between two sections of solid core fiber may not prevent the damage from jumping across.
6. Notes on self-focusing
The effects of self-focusing in optical fiber were demonstrated by Alfano and Shapiro [39]. By launching high peak power light into a multimode optical fiber, they were able to observe the beam diameter collapse into a small spot, as the power was increased. The refractive index change in their case was in the order of only ~10−6 across the beam, shown in Fig. 24 in a sequence of increasing power. The refractive index changes shown in Fig. 16 can easily result in strong self-focusing if transverse heat dissipation is slow, which is indeed the case in the fiber fuse. Thus, we can conclude that self-focusing does play a role in the fiber fuse, given that the process is so fast as to be adiabatic.
Fig. 24 Collapse of a laser beam in a multimode optical fiber as a function of peak energy launched. The change in refractive index in the core is only ~10−6. Adapted from Ref [39].
Recently, a most interesting observation of self-focusing and optical damage was reported by Kanehira et al. [40]. By launching an fs laser pulse in to a microscope slide, the authors were able to self-guide light, forming a channel in the slide. The self-guided light travels ~1mm until it reaches the edge of the microscope slide. At this point dielectric breakdown occurs in air. The high temperatures generated at this point causes the fuse to propagate in the glass, forming bubbles as it retreats towards the laser source, becoming larger close to the focal point, as the beam diameter spreads. This technique could lead to the fabrication of interesting devices in the future. Figure 25 shows the results of the process. Indeed, one needs to draw connections between fs laser pulse induced refractive index changes in glass and optical damage as different faces of the same coin [25 (Chapter 11)].
Fig. 25 Damage propagation initiated by self-guidance in a glass slide. A fs pulse self-writes a waveguide which allows the pulse to propagate to the end of the slide, at which point dielectric-breakdown in air generates heat, causing the fiber-fuse to be initiated. Adapted from Ref [40].
The original observation of pulsed high power damage in optical fiber reported by Archambault et al. [41] shows that certain spontaneous damage thresholds exist. In their experiment, they imprinted a periodic structure in the core by physically damaging the fiber with high optical power. This points to an upper limit before optical damage sets-in and is therefore unavoidable. The clear threshold they noted at ~30mJ in a 20ns pulse indicates that no optical fiber is safe from optical damage. Figure 26 shows the result of the diffractive element formed asymmetrically in the core, resulting in large short wavelength loss.
Fig. 26 Damage diffractive optical element fabricated in optical fiber above a critical optical threshold. The graph shows the “effective” induced refractive index change as a function of pulse energy, indicating a distinct threshold above which damage ensues. Adapted from Ref [41].
Finally, in order to understand how much optical power optical fibers are capable of handling, a graph outlining the various limits from Raman threshold to thermal lensing as a function of input power, core radius and propagation distance, is shown in Fig. 27 [42].
The ultimate limit in silica fiber over long distance seems to be ~36kW CW with the appropriate core-radius for the power.
7. Protection
So how does one save the day when fiber is being operated above a dangerous level [43]? Perhaps one can use a real optical fuse as shown in Fig. 28 ! Todoroki’s optical fuse blows spontaneously when power exceeds a certain value, and thus acts as a traditional fuse protecting sensitive components down the line [44]. This may not be a fail-safe way of protecting an optical fiber system, but collectively with active detection [45, 46], automatic termination with a large mode-field [32] width component, dissipation of the plasma at a surface [47] or the real “fuse” [44] may help reduce extensive damage to telecommunications and in the high power fiber amplifier and laser industry.
Fig. 28 A real fiber fuse, which damages above a certain optical power and effectively breaks the optical circuit. From Ref [44].
8. Epilogue
Many years after the first report of propagating damage in optical fibers in the West [2], the cold war ended and revealed publications from the old Soviet Union. In particular, Bunkin et al. [48] reported a laser spark regime in 1969, which appears to be the first reference to the “slow burning” in air, however, not with guided waves. Raizer [49] discusses this and other works including some aspects of the propagation of a “conflagration wave” in glass. One of the Reviewers kindly pointed out a very early reference to similar plasma propagation in RF waveguides at a wavelength of ~100mm [50]. Interestingly, the authors, Beust and Ford reported the generation of a plasma in what appears to be ~50mm × 50mm cross-section metal-clad, air waveguide at power levels of 2kW. I estimate the power density they used in their experiments to be ~1MW-cm−2, somewhat lower than that required for damage in optical fibers (~5MW-cm−2). The plasma generated by the introduction of foreign objects into the waveguide, was confined in air inside the metal waveguide, and which propagated towards the source at a velocity of ~3m-s−1, damaging its ceramic window. Although they demonstrated the near linear power dependence on power, they did not propose a model for the phenomenon. However, they did propose a scheme to protect the source similar to one now shown recently for the fiber fuse [46]. Despite these early reports, none revealed the impending disaster awaiting optical fibers. The tell-tale periodic signature of the fiber-fuse had to wait nearly another decade [2] to be observed!
Finally, optical waveguides are also not immune to plasma damage as has already been observed [51], and likely to remain an issue with high-Q nanophotonic resonators [52] in which the low injected power is magnified by the very high-Q of the cavity. The optical pump powers used for Raman amplification in present-day long-haul optical fiber systems is in the watts regime, and the output power from Raman amplifiers for communications is close to 1 W [53], and even higher for submarine system amplifiers [54]. These powers are more than sufficient to cause damage to connectors connected or disconnected live. Indeed, all pump lasers operate at power levels at which accidental damage can easily occur - one of the very early observations in semiconductor lasers was reported by Henry et al. [55], in which the damage began at the facet and propagated backwards, in a similar fashion to what later became known as the fiber-fuse. The fiber-fuse has also been triggered by the mere action of fiber cleavage [56], although the reasons for this are not clear. Further, recently IPG Photonics reported singlemode 7kW CW power from a fiber laser [57]! While this fiber laser has a large mode area, it nonetheless has a power density well above the safe-limit and has to be handled with extreme care, respecting minimum bend radii and avoiding spurious reflections. It should be noted that increased demand for optical bandwidth in communication systems is being addressed using multi-cored fiber [58]. As one reaches the capacity limit per fiber-core from power considerations, it will put at risk all the fiber cores should accidental damage occur in one. As different operating WDM windows are deployed (S-C-L etc.) in communication systems, it is the average power that will determine the probability of failure through accidental damage. Of course, dark or live fiber may also be damaged willfully through sabotage.
Owing to limitations in space, the author would like to direct the reader to Todoroki’s review article [59], which provides many references and other details on the “fiber fuse”.
Unfortunately, despite all the research to date, the most important observation of the damage trail made by the author in 1987 - the cavities containing molecular oxygen under high pressure, continues to be ignored. Thus, some more work needs to be done to include the breakdown products of silica glass in the model for the propagating damage, to fully elucidate the effect.
9. Conclusions
This paper revisits the area of optical fiber damage at low powers, often referred to as the “fiber fuse”. The review covers influencing observations and progress in the field in the 25th year of the discovery. It is hoped that parallels drawn in different areas will allow the reader to appreciate the issues at hand and allow more secure systems to be designed in the future, as the power demands in optical fiber increase, bringing them ever closer to spontaneously induced optical damage.
Acknowledgment
The author gratefully acknowledges support from the Canada Research Chairs Program of the Govt. of Canada.
References and links
1. N. S. Bergano, “Undersea cables make the web world wide,” Plenary, OECC2012, Busan, Korea (2–6 July, 2012).
2. R. Kashyap and K. J. Blow, “Spectacular demonstration of catastrophic failure in long lengths of optical fiber via self-propelled self-focusing,” Post deadline paper PD7, 8th National Quantum Electronics Conf., QE8, St. Andrews, Scotland, UK, (21–25 September 1987).
3. J. W. Goethe, Faust: Texte und Kommentare: 2 Bände, with commentary by Albrecht Schöne (Insel Verlag, 2003).
4. R. Kashyap, “Phase-matched periodic-electric-field-induced second-harmonic generation in optical fibers,” J. Opt. Soc. Am. B 6(3), 313–328 (1989). [CrossRef]
5. P. D. Maker, R. W. Terhune, and C. M. Savage, Quantum Electronics, v.III., P. Grivet and N. Blombergen (Eds.)., (Columbia Univ. Press, 1964).
6. R. R. Alfano and S. L. Shapiro, “Observation of Self-Phase Modulation and Small-Scale Filaments in Crystals and Glasses,” Phys. Rev. Lett. 24(11), 592–594 (1970). [CrossRef]
7. I. A. Bufetov, POWAG '2002, Lecture 16, St. Petersburg, Russia, (June 21–24, 2002).
8. R. Kashyap and K. J. Blow, “Observation of catastrophic self-propelled self-focusing in optical fibers,” Electron. Lett. 24(1), 47–49 (1988). [CrossRef]
9. R. Kashyap, “Self-propelled self-focusing damage in optical fibers,” in Proceedings of The Xth International Conference on Lasers, 859–866, Stateline, Lake Tahoe, Nevada, USA, Ed. F. J. Duarte, (STS Press, 1987)
10. E. M. Dianov, I. A. Bufetov, A. A. Frolov, V. M. Mashinsky, V. G. Plotnichenko, M. F. Churbanov, and G. E. Snopatin, “Catastrophic destruction of fluoride and chalcogenide optical fibers,” Electron. Lett. 38(15), 783–784 (2009).
E. M. Dianov, I. A. Bufetov, A. A. Frolov, V. G. Plotnichenko, V. M. Mashinsky, M. F. Churbanov, and G. E. Snopatin, “Catastrophic destruction of optical fibers of various composition under the laser radiation,” Quantum Electron. 32(6), 476–478 (2002).
11. S. Todoroki, “Origin of periodic void formation during fiber fuse,” Opt. Express 13(17), 6381–6389 (2005). [CrossRef] [PubMed]
12. R. Kashyap, “High average power effects in optical fibers and devices,” SPIE 4940, Reliability of Optical Fiber Components, Devices, Systems, and Networks, ed. H. G. Limberger, M. J. Matthewson, (SPIE, Bellingham, WA, 2003), 108–117 (2003).
13. D. P. Hand and P. St. J. Russell, “Solitary thermal shock waves and optical damage in optical fibers: the fiber fuse,” Opt. Lett. 13(9), 767–769 (1988). [CrossRef] [PubMed]
14. E. M. Dianov, V. M. Mashinskii, V. A. Myzina, Y. S. Sidorin, A. M. Streltsov, and A. V. Chickolini, “Change of refractive index profile in the process of laser-induced fiber damage,” Sov. Lightwave Commun. 2, 293–299 (1992).
15. I. A. Bufetov, A. A. Frolov, A. V. Shubin, M. E. Likhachev, S. V. Lavrishchev, and E. M. Dianov, “Propagation of an optical discharge through optical fibers upon interference of modes,” Quantum Electron. 38(5), 441–444 (2008). [CrossRef]
16. R. M. Atkins, P. G. Simpkins, and A. D. Yablon, “Track of a fiber fuse: a Rayleigh instability in optical waveguides,” Opt. Lett. 28(12), 974–976 (2003). [CrossRef] [PubMed]
17. T. J. Driscoll, J. M. Calo, and N. M. Lawandy, “Explaining the optical fuse,” Opt. Lett. 16(13), 1046–1048 (1991). [CrossRef] [PubMed]
18. R. Kashyap, A. Sayles, and G. F. Cornwell, “Heatflow modeling and visualisation of catastrophic self-propelled damage in single mode optical fibers,” Optical Fibers Measurement Symposium, Boulder, SPIE 2966, 586–591 (1996).
19. R. Kashyap, US patent Number: 5,022,734 “Method of modifying an optical waveguide and waveguide so modified,” (filed 21 September 1988.)
20. M. Facão, A. M. Rocha, and P. S. Andre, “Traveling solutions of the fuse effect in optical fibers,” J. Lightwave Technol. 29(1), 109–114 (2011). [CrossRef]
21. A. M. Rocha, M. Facão, A. Martins, and P. S. André, “Simulation of fiber fuse effect propagation,” International Conf. on Transparent Networks – Mediterranean Winter, Angers, France, paper FrP.12 (2009). [CrossRef]
22. D. D. Davis Jr, S. C. Mettler, and D. J. DiGiovanni, “A comparative evaluation of fiber fuse models,” Proc. SPIE 2966, 592–606 (1997). [CrossRef]
23. E. M. Dianov, V. E. Fortov, I. A. Bufetov, V. P. Efremov, A. E. Rakitin, M. A. Melkumov, M. I. Kulish, and A. A. Frolov, “Temperature of optical discharge under action of laser radiation in silica-based fibers,” Paper We3.4.4, Proceedings of ECOC 2005, 3, 469–470, Glasgow, U.K., (Sep. 25–29, 2005). [CrossRef]
24. E. M. Dianov, V. E. Fortov, I. A. Bufetov, V. P. Efremov, A. E. Rakitin, M. A. Melkumov, M. I. Kulish, and A. A. Frolov, “High-speed photography, spectra, and temperature of optical discharge in silica-based fibers,” IEEE Photon. Technol. Lett. 18(6), 752–754 (2006). [CrossRef]
25. R. Kashyap, Fiber Bragg Gratings, Second Edition (Optics and Photonics Series), (Academic Press, 2009).
26. I. H. Malitson, “Interspecimen comparison of the refractive index of fused silica,” J. Opt. Soc. Am. 55(10), 1205 (1965). [CrossRef]
27. E. M. Dianov, V. E. Fortov, I. A. Bufetov, V. P. Efremov, A. A. Frolov, M. Ya. Schelev, and V. I. Lozovoĭ, “Detonation-like mode of the destruction of optical fibers under intense laser radiation,” JETP Lett. 83(2), 75–78 (2006). [CrossRef]
28. E. L. Ruden and G. F. Kiuttu, “Adiabatic, shock, and plastic work heating of solids and exploding metal cylinders,” IEEE Trans. Plasma Sci. 30(5), 1692–1699 (2002). [CrossRef]
29. K. S. Abedin, M. Nakazawa, and T. Miyazaki, “Backreflected radiation due to a propagating fiber fuse,” Opt. Express 17(8), 6525–6531 (2009). [CrossRef] [PubMed]
30. K. S. Abedin, T. Miyazaki, and M. Nakazawa, “Measurements of spectral broadening and Doppler shift of backreflections from a fiber fuse using heterodyne detection,” Opt. Lett. 34(20), 3157–3159 (2009). [CrossRef] [PubMed]
31. Y. Shuto, S. Yanagi, S. Asakawa, M. Kobayashi, and R. Nagase, “Evaluation of High-Temperature Absorption Coefficients of Optical Fibers,” IEEE Photon. Technol. Lett. 16(4), 1008–1010 (2004). [CrossRef]
32. R. Wyatt, R. M. Percival, and R. Kashyap, “Optical communication system and method of protecting an optical route,” US Patent 7,162,161 B2 (2007).
33. R. I. Golyatina, A. N. Tkachev, and S. I. Yakovlenko, “Calculation of velocity and threshold for a thermal wave of laser radiation absorption in a fiber optic waveguide based on the two-dimensional nonstationary heat conduction equation,” Laser Phys. 14(11), 1429–1433 (2004).
34. N. Akhmediev, P. St. J. Russell, M. Taki, and J. M. Soto-Crespo, “Heat dissipative solitons in optical fibers,” Phys. Lett. A 372(9), 1531–1534 (2008). [CrossRef]
35. E. M. Dianov, I. A. Bufetov, A. A. Frolov, V. M. Mashinsky, V. G. Plotnichenko, M. F. Churbanov, and G. E. Snopatin, “Catastrophic destruction of fluoride and chalcogenide optical fibers,” Electron. Lett. 38(15), 783–784 (2002). [CrossRef]
36. E. M. Dianov, I. A. Bufetov, A. A. Frolov, Y. K. Chamorovsky, G. A. Ivanov, and I. L. Vorobjev, “Fiber fuse effect in microstructured fibers,” IEEE Photon. Technol. Lett. 16(1), 180–181 (2004). [CrossRef]
37. N. Hanzawa, K. Kurokawa, K. Tsujikawa, K. Takenaga, S. Tanigawa, S. Matsuo, and S. Tomita, “Observation of a propagation mode of a fiber fuse with a long-period damage track in hole-assisted fiber,” Opt. Lett. 35(12), 2004–2006 (2010). [CrossRef] [PubMed]
38. W. Ha, Y. Jeong, and K. Oh, “Fiber fuse effect in hollow optical fibers,” Opt. Lett. 36(9), 1536–1538 (2011). [CrossRef] [PubMed]
39. R. R. Alfano and S. L. Shapiro, “Observation of self-phase modulation and small-scale filaments in crystals and glasses,” Phys. Rev. Lett. 24(11), 592–594 (1970). [CrossRef]
40. S. Kanehira, J. Si, J. Qiu, K. Fujita, and K. Hirao, “Periodic nano void structures via femtosecond laser irradiation,” Nano Lett. 5(8), 1591–1595 (2005). [CrossRef] [PubMed]
41. J.-L. Archambault, L. Reekie, and P. St. J. Russell, “100% reflectivity Bragg reflectors produced in optical fibers by a single excimer laser pulse,” Electron. Lett. 29(5), 453–455 (1993). [CrossRef]
42. J. W. Dawson, M. J. Messerly, R. J. Beach, M. Y. Shverdin, A. K. Sridharan, P. H. Pax, J. E. Heebner, C. W. Siders, and C. P. J. Barty, “Ultimate power limits of optical fibers,” Paper OMO6, NFOEC/OFC (2010).
43. K. Seo, N. Nishimura, M. Shiino, R. Yuguchi, and H. Sasaki, “Evaluation of high-power endurance in optical fiber links,” Furukawa Review 24 (2003).
44. S. Todoroki and S. Inoue, “Optical fuse made of silica glass optical fibers spliced through low-melting glass with carbon-coating,” Proceedings of XX Int. Cong. on Glass, paper O-14–010, Kyoto, Japan (2004).
45. A. M. Rocha, P. F. Da Costa Antunes, M. D. F. F. Domingues, M. Facão, and P. S. De Brito André, “Detection of fiber fuse effect using FBG sensors,” IEEE Sens. J. B11(6), 1390–1394 (2011), doi:. [CrossRef]
46. K. S. Abedin and M. Nakazawa, “Real time monitoring of a fiber fuse using an optical time-domain reflectometer,” Opt. Express 18(20), 21315–21321 (2010). [CrossRef] [PubMed]
47. E. M. Dianov, I. A. Bufetov, and A. A. Frolov, “Destruction of silica fiber cladding by the fuse effect,” Opt. Lett. 29(16), 1852–1854 (2004). [CrossRef] [PubMed]
48. F. V. Bunkin, V. I. Konov, A. M. Prohorov, and V. B. Fedorov, “Laser spark in the regime of slow burning,” JETF Letts. 9, 609–612 (1969).
49. Y. P. Raizer, Laser-Induced Discharge Phenomena (Plenum Publishing, 1977).
50. W. Beust and W. L. Ford, “Arcing in CW transmitters,” Microwave J., MTT. 10, 91 (1961).
51. R. Kashyap, B. J. Ainslie, and G. D. Maxwell, “Second-harmonic generation in a GeO2 ridge waveguide,” Electron. Lett. 25(3), 206–208 (1989). [CrossRef]
52. M. Ferrera, L. Razzari, D. Duchesne, R. Morandotti, Z. Yang, M. Liscidini, J. E. Sipe, S. Chu, B. E. Little, and D. J. Moss, “Low-power continuous-wave nonlinear optics in doped silica glass integrated waveguide structures,” Nat. Photonics 2(12), 737–740 (2008). [CrossRef]
53. http://www.finisar.com/technical-information, “Operational issues in the deployment of Raman amplifiers (Updated November 2012),” Technical Papers: Optical Amplifiers.
54. http://www.redcinc.com/?CategoryID=267&ArticleID=301
55. C. H. Henry, P. M. Petroff, R. A. Logan, and F. R. Merritt, “Catastrophic damage of AlxGa1-xAs double-heterostructure laser material,” J. Appl. Phys. 50(5), 3721–3732 (1979). [CrossRef]
56. P. St. J. Russell, Private Communication (2012).
57. V. P. Gapontsev, “Advances in power scaling of fiber lasers,” Paper 8601–13, Photonics West (2013).
58. R.-J. Essiambre and R. W. Tkach, “Capacity trends and limits of optical communication networks,” P.I.E.E.E 100(5), 1035–1055 (2012).
59. S. Todoroki, “Fiber fuse propagation behavior,” Optical Fibers/Book 4, ed. Y. Moh, (InTech, 2011) [Online] http://www.intechweb.org/books/).
