Progress in low-loss and high-bandwidth plastic optical fibers


  • Yasuhiro Koike,

    Corresponding author
    1. Keio Photonics Research Institute, Keio University, 3-14-1 Hiyoshi, Kouhoku-ku, Yokohama 223-0061, Japan
    2. ERATO-SORST Koike Photonics Polymer Project, Japan Science and Technology Agency, 7-1 Shinkawasaki, Saiwai-ku, Kawasaki 212-0032, Japan
    • Keio Photonics Research Institute, Keio University, 3-14-1 Hiyoshi, Kouhoku-ku, Yokohama 223-0061, Japan
    Search for more papers by this author
  • Kotaro Koike

    1. Keio Photonics Research Institute, Keio University, 3-14-1 Hiyoshi, Kouhoku-ku, Yokohama 223-0061, Japan
    2. Polymer Research Institute, Polytechnic Institute of New York University, 6 Metrotech Center, Brooklyn, New York 11201
    Search for more papers by this author


Plastic optical fibers (POFs) are highly promising transmission media for future home networking.In comparison to glass optical fibers (GOFs), which are commonly used in core and metropolitan networks, POFs offer many advantages such as great flexibility and easy handling. This review begins with the basic concepts of optical fibers and moves on to the early history of loss reduction in POFs. What drastically changed the status of POFs in the communications field was a graded-index technology that improved the bandwidth to over 1 gigabits per second. However, even after the loss and bandwidth were enhanced to their limits, the performances of POFs were insufficient for market demand when using conventional optical polymer materials such as poly(methyl methacrylate). Recently, this problem has been solved by several lines of material research using fluorinated polymers. As a result, high-speed optical home networking by POFs has become more realistic. © 2010 Wiley Periodicals, Inc. J Polym Sci Part B: Polym Phys, 2011


One of the Nobel Prizes in Physics in 2009 was awarded to Charles Kao for “groundbreaking achievements concerning light transmission in fibers for optical communication.” In the 1960s, glass optical fibers (GOFs) that could be used for transmitting light had already been developed.1, 2 However, these fibers exhibited significant propagation losses above 1000 dB/km; the transmission distance was severely limited. Kao made a discovery that led to a breakthrough in fiber optics. After closely examining the possibilities for reducing propagation losses, he realized that the attenuation of existing fibers were orders of magnitude above the fundamental limit.3 He theorized that by using a fiber of the purest glass, it would be possible to transmit light signals over 100 km, in contrast to fibers available in the 1960s that could only realize a distance of 20 m. His prediction of the future potential of low-loss optical fibers inspired other researchers to make great efforts toward its realization. Just 4 years later, in 1970, the first ultrapure fiber with an attenuation of 20 dB/km was successfully fabricated.4 Others soon followed, and losses were pushed down to the theoretical limit (0.2 dB/km).5–7 A first-generation fiber-optic communication system was successfully deployed in 1976, and since then, GOFs have steadily connected the world. Today, over a billion kilometers of GOFs, sufficient to encircle the globe more than 25,000 times, have been placed underground and in oceans, enabling global broadband communication media such as the Internet.

We believe that the next big step in optical fiber technology will be plastic optical fibers (POFs). In advanced countries, fiber-to-the-home (FTTH) services that interconnect homes with a GOF backbone are well established. However, intrabuilding networks such as home networks have yet to be adequately developed. With the increase in the number and use of home PCs, high-definition (3D) TVs, blue-ray devices, digital cameras, storage capabilities, and intelligent home appliances, the demand has increased for optical data-link connections not only up to the end-users' buildings but also within them.8 If we compare communication networks to human blood vessels, the GOF backbone functions like the arteries and veins, whereas the home network is similar to the capillaries. Surprisingly, the entire length of all terminal networks accounts for 95% of that of all optical networks. In such short-distance applications, the extremely low attenuation and enormous capacity of a single-mode GOF is unnecessary. Instead, simpler and less expensive components, greater flexibility, and higher reliability against bending, shocks, and vibrations are considerably more valuable properties. Naturally, polymers are overwhelmingly superior to glasses in all of these requirements.

However, for POFs, which have been overshadowed by the success of GOFs in previous decades, the road has never been smooth. In particular, many studies have sought to reduce fiber attenuation as had been done for GOFs. To be used as the core base material, polymers must ultimately be transparent. Given its amorphous structure and easy processing via a free-radical bulk polymerization that requires no metal catalysts and avoids contaminants, poly(methyl methacrylate) (PMMA) has mainly been utilized for POFs.9 Coupled with low material costs, its adequate thermal stability and excellent corrosion resistance have also made PMMA the representative material for POF. However, even with highly purified PMMA-based POFs, the intrinsic problem of attenuation caused by C[BOND]H molecular vibrations remained unresolved. Furthermore, POFs with a large core and modal dispersion had severe limitations in bandwidth. In this review, we begin with a brief explanation of optical fibers, and proceed to describe technical developments that have solved the aforementioned problems and propelled POFs into the main stream of modern home networking applications.


The principle of light propagation through POFs is the same as that for GOFs. POFs are basically composed of two coaxial layers: the core and the cladding. The core is the inner part of the fiber that guides light, whereas the cladding completely surrounds the core. The refractive index of the core is slightly higher than that of the cladding. Hence, when the incident angle of the light input to the core is greater than the critical angle, the input is confined to the core region because of total internal reflection at the interface between the different dielectric materials comprising each layer. Although this simple concept is a useful approximation of light guidance in many kinds of fibers, it does not provide a full explanation. Light is really an electromagnetic wave with a frequency in the optical range. An optical fiber guides the waves in distinct patterns called modes, which describe the distribution of light energy across the waveguide. Commonly used optical fibers can be separated into two classes based on their modal properties: single-mode and multimode fibers. Single-mode fibers are step-index (SI) fibers, whereas multimode fibers can be divided into SI and graded-index (GI) fibers. SI and GI refer to patterns of variation in the refractive index with the radial distance from the fiber axis. Figure 1 schematically shows these three types of fibers: (a) the single-mode fiber, (b) the SI multimode fiber, and (c) the GI multimode fiber.

Figure 1.

Ray trajectories through basic types of optical fibers: (a) Single-mode fiber, (b) Step-index multimode fiber, and (c) Graded-index multimode fiber.

The information-carrying capacity of an optical fiber is determined by its impulse response. The impulse response and hence bandwidth are largely determined by the modal properties of the fiber. Given that single-mode transmission avoids modal dispersion and other effects that occur with multimode transmission, single-mode fibers with core diameters of 5–10 μm can carry signals at considerably higher speeds than multimode fibers.10, 11 Modal dispersion can be understood by referring to the SI multimode fiber [Fig. 1(b)], where different rays are shown to travel along paths with different lengths. Even if they are coincident at the input end and travel at the same speed within the fiber, these rays disperse at the output end because of their different path lengths. As a result, the impulse signal broadens. This becomes a serious restriction on transmission speed because pulses that overlap can interfere with each other, making it impossible to receive the signal. On the other hand, GI guides have less modal dispersion and greater transmission capacity than SI guides.12–15 The refractive index of the core in GI multimode fibers is not constant but decreases gradually from its maximum at the core center to its minimum at the core/cladding boundary. From Figure 1(c), it is easy to understand qualitatively why the modal dispersion decreases for GI fibers. As in the case of SI fibers, the path is longer for more oblique rays (higher-order mode). However, the ray velocity changes along the path because of variations in the refractive index. The speed of light in a material, v, is equal to the velocity of light in vacuum, c, divided by the refractive index, that is, v = c/n. More specifically, ray propagation along the fiber axis takes the shortest path but has the slowest speed since the index is largest along this path. Oblique rays have a large part of their path in a lower refractive index medium in which they travel faster. The difference in the refractive index is small but sufficient to compensate for the time delay. By carefully controlling the refractive index profile in the core region, the modal dispersion can be drastically reduced.

Currently, most GOFs are fabricated with single-mode structures and are commonly used in long-distance applications such as core and metropolitan networks. With respect to coupling loss, the small core of single-mode fibers is a serious disadvantage; the smaller the core diameter, the harder it is to couple light into the fiber. Hence, GI multimode-type fibers were also extensively studied12–15 and deployed in some telecommunication applications up until the mid-1980s. However, as single-mode GOFs were far superior in both attenuation and bandwidth, they gradually shifted to short-length applications such as storage area networks. Coupling light into a single-mode fiber inevitably requires considerably tighter tolerances than doing so into the larger cores of a multimode fiber. However, such tighter tolerances were achieved; nowadays, the single-mode GOF has become the standard choice for virtually all kinds of telecommunications that involve high bitrates or span distances longer than a couple of kilometers.

On the other hand, POFs have attracted attention as optimal candidates for short-distance networks such as intrabuilding networks.9 Optical fibers must be able to connect various devices in the individual rooms of a building; hence, they must be flexible, easy to bend, and connected at several points. Predictably, the single-mode GOF is unsuited for this application given its brittleness and small core. In contrast, POFs can be enlarged to ∼1 mm in diameter without losing flexibility or ease of fiber alignment (Fig. 2). Therefore, apart from a few exceptions,16–20 a great deal of effort has been focused on developing multimode POFs with large diameters.

Figure 2.

Cross-sectional views of representative optical fibers and a photo of a POF with a 3-mm knot.


The first POF, Crofon™, which was invented in the mid-1960s by Du Pont, was a multimode fiber with an SI profile in the core region.21, 22 Compared to GOFs, the SI POF was also advantageous in terms of mass production; it was not only inexpensive to fabricate, but also easy to mold and manufacture. The first commercialized SI POF was Eska™, which was introduced by Mitsubishi Rayon in 1975;23, 24 subsequently, Asahi Chemical and Toray also entered the market. However, the first fibers were insufficiently transparent to be used as an intrabuilding communication medium, and their application was severely limited to extremely short-range areas such as light guides, illuminations, audio data-links, and sensors.

Fiber attenuation limits how far a signal can propagate in the fiber before the optical power becomes too weak to be detected. The oldest and most common approach for determining fiber attenuation is measuring the optical power transmitted through long and short lengths of the same fiber using identical input couplings,25 a method known as the cutback technique. First, the optical power at the output (or far end) of the fiber is measured. Subsequently, the fiber is cut off a few meters from the light source, and the output power at this near end is measured without disturbing the input. When Pout and Pin represent the output powers of the far and near ends of the fiber, respectively, the attenuation α, which is normally expressed in decibels, is defined as follows:

equation image(1)

where L (in kilometers) is the distance between the two measurement points. To estimate the acceptable transmission losses of the fiber, the total optical link from the light source to the receiver must be considered. As a typical example for home networking, the minimum output power of the light source is −6 dBm, and the lowest optical sensitivity of the receiver is −18 dBm. Considering other factors such as coupling loss (1.5 dB), bending loss (0.5 dB), power penalty (1.5 dB), and margin (2.5 dB), the acceptable transmission loss for the fiber is 6 dB. Accordingly, in the case of deploying POFs in a home, in which the maximum laying distance is ∼30 m, the fiber attenuation must be less than 200 dB/km. In comparison, the attenuation of the first prototype was over 1000 dB/km. Faced with this challenge, many researchers have attempted to reduce the attenuation. Although the various mechanisms contributing to losses in POFs are basically similar to those in GOFs, their relative magnitudes differ. Figure 3 shows the loss factors for POFs, which are divided into intrinsic and extrinsic factors. These are further classified into absorption and scattering losses.

Figure 3.

Classification of intrinsic and extrinsic factors affecting POF attenuation.

Through analyses of each factor of fiber attenuation described below, the limitations of POFs based on several materials and their theoretical groundings have been steadily clarified. Meanwhile, it has been shown that the major factors causing high attenuation were not intrinsic but rather extrinsic—those such as contaminants becoming mixed into the polymer during the fiber fabrication process. By preparing the fiber in an all-closed system from monomer distillation to fiber drawing, low-loss POFs with losses near theoretical limits were fabricated. In 1981, Kaino reported a poly(styrene) (PS)-based SI POF with an attenuation of 114 dB/km at 670 nm,26 and in the following year, his group also succeeded in obtaining a PMMA-based SI POF with an attenuation of 55 dB/km at 568 nm.27 Table 1 is a brief early chronology of SI POFs with reduced attenuation.

Table 1. Early History of Loss Reduction for SI POFs
inline image

Absorption Loss

Every material absorbs light energy, whose amount depends on the wavelength and material. Intrinsic absorption loss in POFs is caused by electronic transitions and molecular vibrations. Electronic transition absorption results from transitions between electronic energy levels of bonds within the materials; the absorption of photons causes an upward transition, which leads to excitation of the electronic state. Typically, electronic transition peaks appear in the ultraviolet wavelength region, and their absorption tails influence the transmission losses in POFs. For example, when azo compounds are used as the initiator, a PMMA core POF, one of the commercially available POFs, has the n-π* transition of the ester groups in MMA molecules, the n-σ* transition of the S[BOND]H bonds in chain transfer agents, and the π-π* transition of the azo groups. The most significant absorption is the transition of the n-π* orbital of the double bond within the ester group. The relationship between the electronic transition loss αe (dB/km) and the wavelength of incident light λ (nm) can be expressed by Urbach's rule:28

equation image(2)

Here A and B are substance-specific constants. In the case of PMMA, A and B have been clarified to be 1.58 × 10–12 and 1.15 × 104, respectively.29 Hence, the αe value of PMMA is less than 1 dB/km at 500 nm. On the other hand, PS and poly(carbonate) (PC), which are also universal polymers for POFs, exhibit considerably larger absorption losses (Fig. 4).29, 30 This is because of the π-π* transition of the phenyl groups in PS and PC. The energy band gap between the π and π* levels is sufficiently small to be excited by visible light. The tails are drastically shifted to longer wavelengths as the conjugation length increases.

Figure 4.

Electronic transition losses for PMMA, PS, and PC.

The effect of molecular vibrational absorptions becomes strong at wavelengths in the visible–to-infrared region. In the case of PMMA, which shows negligibly small electronic transition absorption, this molecular vibration is the predominant factor contributing to fiber attenuation. The energy level of the absorption wavelength, which reflects the light absorption at each wavelength, is expressed by31, 32

equation image(3)

Here, υ is the quantum number (= 0, 1, 2, 3…), χ is the anharmonic constant, and v0 is the fundamental frequency. The first and second terms on the right side are the harmonic and anharmonic vibrations, respectively. The overtone frequency is expressed as follows:

equation image(4)

Setting the quantum number as 1, υ0 is transformed as follows:

equation image(5)

Equation 5 can then be substituted into eq 4 to yield

equation image(6)

The overtone absorption frequencies can be calculated using eq 6 and χ, which can be obtained from measurements of v1 and the first overtone v2. Groh estimated the attenuation by the overtone vibrational absorptions of the C[BOND]H and C[BOND]X bonds as follows:32

equation image(7)

Here αv (dB/km) is the attenuation; ρ (g/cm3), the polymer density; M (g/mol), the molecular weight of the monomer unit; NCX, the number of C[BOND]X bonds per monomer; and Emath image/Emath image, the vibration energy ratio of each bond to the fundamental frequency of the C[BOND]H bond. Figure 5 shows the spectral overtone positions and normalized integral band strengths for the C[BOND]H, C[BOND]D, and C[BOND]F vibrations. If we set ρ = 1.19 g/cm3, M = 100 g/mol, and NCH = 8 as the PMMA values, it follows from eq 7 that Emath image/Emath image = 3.3 × 10–8, which corresponds to an attenuation of 1 dB/km. In the visible to near infrared region, the overtones for C[BOND]D and C[BOND]F are several orders of magnitude lower than the overtone for C[BOND]H. This implies that fiber attenuation can be reduced drastically by substituting the hydrogen atoms for heavier atoms such as deuterium and fluorine with lower energy absorption bands.

Figure 5.

Calculated spectral overtone positions and normalized integral band strengths for different C[BOND]X vibrations.

Scattering Loss

Scattering losses in polymers arise from microscopic variations in material density. When natural light of intensity I0 passes through a distance y, and its intensity is reduced to I by scattering loss, the turbidity τ is defined by

equation image(8)

Since τ corresponds to the summation of all light scattered in all directions, it is given as

equation image(9)

Here, V and H denote vertical and horizontal polarizations, respectively. The symbol A and the subscript B in the expression for a scattering component, AB, represent the directions of the polarizing phase of a scattered light and an incident light, respectively. θ is the scattering angle in relation to the direction of the incident ray. In structureless liquids or randomly oriented bulk polymers, these intensities are given by the following equations:

equation image(10)
equation image(11)

Here the isotropic part Vmath image of Vv is given as follows:33

equation image(12)

By substituting eqs 1012 into eq 9, τ can be rewritten as follows:

equation image(13)

Furthermore, the intensity of the isotropic light scattering, Vmath image, and the anisotropic light scattering, HV, can be expressed by eqs 1434 and15, 33 respectively

equation image(14)
equation image(15)

Here λ0 is the wavelength of light in vacuum; n, the refractive index; k, the Boltzmann constant; T, the absolute temperature; β, the isothermal compressibility; N, the number of scattering units per unit volume; and 〈δ2〉, the mean square of the anisotropic parameter of polarizability per the scattering unit. Finally, from the definition of the turbidity τ in eq 8, the light scattering loss αs (dB/km) is related to the turbidity τ (cm−1) by

equation image(16)

Through analyses of light scattering in several optical polymers such as PMMA,35, 36 PS,37 and PC,38, 39 these equations have proven to be suitable for amorphous polymers with heterogeneous structures small in size relative to the wavelength of the incident light. In particular, PMMA, which has the lowest scattering loss among such polymers, has received considerable attention. Using published data of β = 3.55 × 10–11 cm2/dyn at around Tg for bulk PMMA40 and assuming freezing conditions, the value of Vmath image at room temperature and a wavelength of 633 nm is 2.61 × 10–6 cm–1. As a result, the theoretical light scattering loss can be estimated from the Vmath image value as 9.5 dB/km by using eqs 13 and 16; this is almost identical to the experimental value of 9.7 dB/km. It should be noted here that even in highly purified PMMA, if it is polymerized below the glass transition temperature (Tg), the scattering loss inevitably increases to several hundred dB/km.35 This is because of large-sized heterogeneities with dimensions of ∼1000 Å formed by volume shrinkage during polymerization.


After the studies on transparency, a complete high-speed optical network based on POFs became realistic. However, despite these advances, the SI POF had a limitation in bandwidth, requiring the excitation of several tens of thousands of modes for transmission. The large core increases modal dispersion and drastically degrades the bandwidth to approximately several hundred megahertz over 100 m. The concept of fiber bandwidth originates from the general theory of time-invariant linear systems.41 If the optical fiber can be treated as a linear system, its input and output powers in the time domain are described simply as follows:

equation image(17)

In other words, the output pulse response pout(t) of the fiber can be calculated through the convolution of the input pulse pin(t) and the impulse response function h(t) of the fiber. The period T between input pulses should be wider than the expected time spread of the output pulses. In the frequency domain, eq 17 can be expressed as follows:

equation image(18)

Here H(f), the power transfer function of the fiber at the baseband frequency f, is the Fourier transform of h(t)

equation image(19)

and Pout(f) and Pin(f) are the Fourier transforms of the output and input pulse responses pout(t) and pin(t), respectively, that is,

equation image(20)

The optical bandwidth of the fiber is defined by the Fourier-transformed H(f). This is normally done in terms of the −3 dB bandwidth, which is the modulation frequency at which the optical power of H(f) falls to one-half the value of the zero frequency modulation. The larger the −3 dB bandwidth f−3dB, the narrower the output pulse and the higher the possible transmission capacity.

Typically, modal dispersion is the dominant factor degrading the bandwidth for multimode optical fibers such as SI POFs. However, as mentioned earlier, it can be minimized by forming a quadratic refractive index profile in the core region and by controlling the propagation speed of each mode. This method was first developed by Nishizawa et al. for glass fibers,12, 15 and subsequently, it was adapted to POFs. The first GI POF was reported in 1982.42 Initially, the GI profile was formed by copolymerizing two or three kinds of monomers with different refractive indices and monomer reactivity ratios.43, 44 In this system, several properties such as attenuation, bandwidth, refractive index profile, and numerical aperture were limited by the differences in the refractive indices and reactivity ratios between the monomers. To solve the problem, a simpler method of forming the GI profile—the low-molecular doping method—was developed in 1994.45 By adding a dopant with a higher refractive index than the base polymer and forming a concentration distribution in the radial direction, it became easy to control the refractive index profile. Furthermore, this groundbreaking method enabled the reliable fabrication of low-loss GI POFs. Since the first GI POF was reported, various methods for forming a GI profile have been proposed.46–51 Among these, here we discuss two major techniques using the low-molecular doping method.

Interfacial-Gel Polymerization Technique

In this method, a rod with a GI profile called a “preform” is first prepared, after which the preform is heat-drawn to the GI POF. The interfacial-gel polymerization technique52 refers to the method used to form a parabolic refractive index profile in the core region. In this review, we use a typical system as an example. The base polymer of the core and cladding layers is PMMA and the low-molecular-weight dopant is diphenyl sulfide (DPS). First, a glass tube charged with MMA monomer mixtures including benzoyl peroxide (BPO) and n-butyl mercaptan (n-BM) as the initiator and chain transfer agent, respectively, is rotated on its axis at 3000 rpm in an oven at 70 °C for 3–6 h. The MMA monomer comes to coat the inner wall of the glass tube because of centrifugal force, and is gradually polymerized [Fig. 6(a)]. After heat treatment at 90 °C for 24 h, the polymer tube based on PMMA is obtained as the cladding layer of the GI preform. The tube is then filled with a core solution containing a mixture of MMA monomers as well as DPS, di-tert-butyl peroxide (DBPO), and n-lauryl mercaptan (n-LM) as the dopant, initiator, and chain transfer agent, respectively. The tube filled with the mixture is heated in an oil bath at 120 °C for 48 h under a nitrogen pressure of 0.6 MPa [Fig. 6(b)]. Previous studies have clarified that an amorphous polymer glass such as PMMA polymerized at a temperature lower than the Tg exhibits an extremely high scattering loss because of density fluctuations.35–37 Hence the core solution has to be polymerized at a temperature higher than the Tg, which is 105–115 °C. However, such a high temperature leads to numerous bubbles during the polymerization reaction because it is above the boiling point of MMA (100 °C). Therefore, polymerization must be performed under appropriate pressure. Finally, the preform is heat-drawn to the fiber at 220–250 °C [Fig. 6(c)].

Figure 6.

Fabrication process of GI POFs by the preform method: (a) preparation of polymer tube, (b) core polymerization, and (c) heat-drawing.

The mechanism for forming a GI profile is described in Figure 7. At the beginning of core polymerization, the inner wall of the PMMA tube becomes slightly swollen by the monomer-dopant mixture to form a polymer gel phase. The reaction rate of polymerization is generally faster in the gel phase than in the monomer mixture because of the gel effect,53 and the polymer phase grows from the inner wall of the tube to the core center. During the growth process, the MMA monomer can diffuse into the gel phase more easily than the dopant molecules because the molecular volume of the dopant, which contains benzene rings, is larger than that of the monomer. Hence, the dopant is concentrated in the middle region to form a quadratic refractive index profile. The shape of the profile can be adjusted by various methods, and precise control is possible.54 Above all, the amount of initiator, polymerization temperature, and ratio of the diameter of the hole to that of the tube strongly affects the profile.

Figure 7.

Formation process of GI distribution by the interfacial-gel polymerization technique.

The interfacial-gel polymerization technique is particularly common in acrylic GI POF studies and enables the precise control of the refractive index profile, leading to a maximally high bandwidth. However, this batch process requires many complicated procedures. Furthermore, the fiber length obtained at any one time is completely dependent on the preform size. This is a serious limitation in terms of fabrication costs.

Coextrusion Process

To realize mass production, a new process for GI POF fabrication called coextrusion was investigated.55–57 In this process, polymers doped with low-molecular-weight dopants and homogeneous polymers are first prepared as the base materials of the core and cladding layers, respectively. In PMMA-DPS systems, polymerizations are performed at 90 °C for 24 h. The bulk polymers are then further heated at 110 °C for 48 h. The two materials are melted in their respective extrusion sections at 210–250 °C and compounded in a die to fabricate a POF with a concentric circular core/cladding structure. At this point, the fiber does not have a GI distribution. However, by heating the fiber in the diffusion section, a radial concentration profile of low-molecular-weight dopant is formed as a result of molecular diffusion. Finally, the GI POF is obtained by winding it onto a take-up reel. The operational procedure of coextrusion is illustrated in Figure 8. This process is very simple, and hence, GI POFs are expected to be manufactured continuously at a relatively low cost.

Figure 8.

Schematic diagram of GI POF fabrication by the coextrusion procedure.

In previous studies, it was presumed that a GI POF fabricated by this process would have a refractive index distribution with a tail at the core/cladding boundary, not as steep as that prepared by the conventional batch process.49 Therefore, the bandwidth of the GI POF obtained by dopant diffusion co-extrusion would be inferior to that obtained by the batch process. However, contrary to this expectation, a GI POF without any tail prepared by the process was reported in 2007.56 If low-molecular-weight molecules diffuse in a radial direction, their diffusion can be expressed by Fick's diffusion equation:

equation image(21)

Here C is the concentration of low-weight molecules; t, the diffusion time; r, the distance from the fiber center; and D, the mutual diffusion coefficient of low-weight molecules. The concentration profile calculated using eq 21 has a gradual shape around the diffusion front because D is almost constant in general systems. However, in the polymer-dopant system, the plasticization that occurs as a result of adding the dopant into the polymer matrix must be considered. As the dopant concentration increases, the Tg, which determines the mobility and diffusivity of the dopant, decreases linearly. In other words, diffusion within the diffusion section of the coextrusion equipment obeys Fick's law with a diffusion coefficient that varies as a function of dopant concentration. Figure 9 shows the comparison of the refractive index profiles of experimental and simulated data. The simulation was performed with a diffusion coefficient calculated from the results of a one-dimensional diffusion experiment. The polymer matrix and dopant are PMMA and DPS, respectively. The experimental and simulated data show good agreement, indicating a high possibility that a high-bandwidth GI POF can be prepared by the dopant diffusion coextrusion process. Currently, this continuous fabrication method is well established and is employed for some commercially available GI POFs.

Figure 9.

Refractive index profiles. Broken line is the refractive index profile calculated from eq 21 using the measured dependence of the mutual diffusion coefficient for a PMMA-DPS system. Solid line is the measured refractive index profile for a PMMA-DPS-based GI POF fabricated by the coextrusion method. From Asai et al., J Lightwave Technol 2007, 25, 3062–3067, © IEEE, reproduced by permission.


With various improvements in the fabrication process enabling the avoidance of contaminants and the formation of GI profiles, low-loss and high-bandwidth POFs have become leading candidates for short-distance networks (e.g., home networks). For practical applications, however, the intrinsic problem of attenuation still remained.

Since the first SI POF was commercialized, most POFs have been manufactured using PMMA, a mass-produced commercially available polymer that demonstrates high light transmittance and provides excellent corrosion resistance to both chemicals and weather. These properties, coupled with low manufacturing costs and easy processing, have made PMMA a valuable substitute for glass in optical fibers. PMMA-based SI POFs have been used extensively in short-distance datacom applications such as digital-audio interfaces. They have also been utilized for data transmission equipment, control signal transmissions for numerically controlled machine tools, railway rolling stocks, and optical data buses in automobiles. In particular, the increasing complexity of in-vehicle electronic systems has led to POFs becoming indispensable to the automobile industry.58 Today, it is not uncommon to find 10 or 20 consumer electronic devices, such as main units, DVD (blue-ray) players, navigation systems, telephones, Bluetooth interfaces, voice recognition systems, high-end amplifiers, and TV tuners, all connected inside a car. To meet all the necessary requirements for data transfer between such devices, POFs have provided a great solution.

On the other hand, when high-speed data transmission of more than 1 Gbps is required, PMMA-based POFs cannot be used. In the short-distance datacom or telecom applications above, systems using SI POFs as the transmission medium and red light-emitting diodes (LEDs) at 650 nm as the light source have been employed. In contrast, to realize gigabit in-home communications, GI POFs and vertical-cavity surface-emitting lasers (VCSELs),59 which offer fast modulation, is the most reasonable combination.60 VCSELs constitute a relatively new class of semiconductor lasers, which are monolithically fabricated. They are now considered to be key devices for gigabit Ethernet, high-speed LANs, computer links, and optical interconnects. As mentioned before, an acceptable level of fiber attenuation in home networks is ∼200 dB/km. Furthermore, since the emission wavelengths of long-life and inexpensive VCSELs are 670–680 nm, the fiber must satisfy the limitations at this wavelength region. However, for PMMA-based POFs, the wavelengths satisfying the requirement are limited to 570 and 650 nm because of absorption losses from C[BOND]H stretching vibrations.37

Another widely used base polymer for POF is PS. With respect to mechanical properties and chemical resistance, PS is slightly inferior to PMMA. However, PS has some features appropriate for the core material of POF as well as an attractive price. Among them, the biggest advantage of using PS is the low attenuation at wavelengths of 670–680 nm. While PMMA-based POFs have a high attenuation of over 200 dB/km in this region, the attenuation of PS-based POFs is as low as 114 dB/km.26 Although the visible spectrum of a PS-based POF is dominated by high harmonic C–H absorptions as in the case of a PMMA-based POF, the shape of the spectrum is completely different. The attenuation spectrum of a PS-based POF is shown in Figure 10. In the styrene unit, there are three aliphatic and five aromatic C[BOND]H bonds. The absorption wavelengths of the aliphatic C[BOND]H bonds corresponding to the 5th, 6th, and 7th overtones are 758, 646, and 562 nm, respectively, whereas the same overtones of aromatic C[BOND]H bonds appear at 714, 608, and 532 nm, respectively.26, 31 It is noteworthy that the positions of the aliphatic C[BOND]H overtones are slightly shifted to longer wavelengths relative to those for PMMA. As a result, the emission wavelength of a VCSEL is centered between the 5th aromatic and 6th aliphatic C[BOND]H peaks and is not significantly affected by C[BOND]H absorption losses. Most studies on PS-based POFs have considered SI-type POFs.29, 31 If a high-speed GI POF based on PS can be fabricated with low attenuation, the fiber could be a highly promising candidate for a home network medium. However, this is not easy. The difficulty arises from the high refractive index. The refractive index of PS is around 1.59 and is higher than that of PMMA by 0.1. The importance of this difference can be observed in Figure 9. When a GI profile with a sufficient difference in refractive indices between the core and cladding is formed by doping DPS, the Tg of the core center decreases to below 60 °C because of the plasticization effect. Clearly, the thermal stability is insufficient for practical use.

Figure 10.

Attenuation spectrum of PS-based SI POF. ν0 and νmath image correspond to absorption wavelengths of aliphatic and aromatic C[BOND]H bonds, respectively. Shoulders in the region of lower wavelength of each peak are assigned to the combination bands nν0 + δ and nνmath image + δ, where +δ is the fundamental aliphatic C[BOND]H bending vibration. From Kaino et al., J Appl Phys 1981, 52, 7061–7063, © AIP, reproduced by permission.

Partially Fluorinated Polymers

The high attenuation of conventional POFs is dominated by C[BOND]H overtone stretches and combinations of stretch and deformations. Hence, the most effective method to obtain a low-loss POF is substituting hydrogen with heavier atoms such as fluorine, as was shown in Figure 5. However, if vinyl monomers such as MMA are perfluorinated, the polymerization rate drastically decreases. Boutevin et al. studied various partially fluorinated polymers which are easily prepared by free-radical polymerization and calculated the influence of the molar number of C[BOND]H bonds per unit volume.61 The contribution of the fluorine substituent is considerably larger than expected from the chemical structure. For instance, poly(2,2,2-trifluoroethyl methacrylate) (poly(TFEMA))-based GI POF provides a surprisingly low attenuation—127–152 dB/km at 670–680 nm62—relative to that for PMMA-based GI POF. Although the number of C[BOND]H bonds per monomer unit between TFEMA and MMA differ by only one, the trifluoroethyl group of TFEMA possesses a large volume, and the number of C[BOND]H bonds per unit volume in poly(TFEMA) is only 64% of that in PMMA. However, the long side chain not only reduces fiber attenuation but also lowers Tg.

Recently, a novel GI POF based on a partially fluorinated polymer with a lower attenuation and higher Tg than those of PMMA-based GI POF was reported.63 In this study, a copolymer of MMA and pentafluorophenyl methacrylate (PFPMA) was employed as the core polymer. In general, copolymers tend to have extremely high scattering losses because of their large heterogeneous structure and corresponding heterogeneity of the refractive index; consequently, they have not been leading candidates for POF base materials. In this copolymeric system, however, the increase in scattering loss is negligibly small since the refractive indices of both homopolymers are almost identical (PMMA: nD = 1.4914, poly(PFPMA): nD = 1.4873). Moreover, the number of C–H bonds per unit volume in poly(PFPMA) compared with that in PMMA is only 34%; hence, smaller absorption losses and lower attenuation than those for PMMA can be obtained by copolymerization. When the core composition is 65/35 = MMA/PFPMA (mol %), the attenuations are 172–185 dB/km at 670–680 nm; this satisfies the required attenuation for optical home network systems.

Regarding thermal stability, the copolymer exhibits a higher Tg than PMMA when the PFPMA content in the monomer feed is 0–50 mol%. The relationship between the Tg of bulk MMA-co-PFPMA and the PFPMA content in the monomer feed is shown in Figure 11(a). The Tg for copolymers can generally be described by the Gordon-Taylor equation and changes linearly with the weight fractions of the monomer content.64 Interestingly, in the case of this copolymeric system, it is observed that the Tg plot exhibits a positive deviation. Although the abscissa is expressed as the molar fraction of PFPMA in the monomer feed, the result is similar even if it is converted to weight fractions. This result can be explained by examining Figure 11(b), in which the Tg of precipitated copolymers and the amount of residual monomers in the bulk copolymer are shown. As Boutevin et al. reported, poly(PFPMA) has an intrinsically high Tg of 130 °C, whereas the bulk shows a low Tg of 91 °C. The reactivity of the carbon–carbon double bond of phenyl methacrylate is known to be significantly reduced by substituting the hydrogen of the benzene ring with larger molecules.65, 66 In this case, the pentafluorophenyl group disturbs the propagation reaction with steric hindrance, thereby leading to low polymerization conversion. As a result, numerous unreacted monomers remain in the bulk, and the Tg is degraded by the plasticization effect. That the Tg of the copolymer increases as the PFPMA content decreases is related to changes in the amount of PFPMA residues.

Figure 11.

(a) Tg plots of bulk MMA-co-PFPMA polymerized at 110 °C for 48 h (○). (b) Tg plots of the purified MMA-co-PFPMA (•), and the amount of residual MMA (⋄) and PFPMA (♦) in bulk copolymers against the PFPMA content in the monomer feed. From Koike, K.; Koike, Y., Polymer 2010, 51, 1377–1385, © Elsevier, reproduced by permission.

The amount of MMA remaining in all bulks except poly (PFPMA) is ∼1 wt % and nearly independent of the monomer feed composition. In contrast, the residual PFPMA drastically decreases after copolymerization with MMA. This derives from the monomer reactivity ratios. Figure 12 shows the changes in the copolymer composition with the polymerization conversion calculated by the Mayo-Lewis equation. Here, r12 and r21 (M1: MMA, M2: PFPMA) are 0.56 and 1.30, respectively,67 and the initial monomer content in the feed is 50/50 mol %. As shown in the figure, PFPMA polymerizes preferentially given ideal polymerization. As the polymer initially consists of 60 mol% PFPMA units, PFPMA conversion can be improved in the presence of MMA. On the other hand, the Tg of the purified copolymer with no residual monomers linearly increases with the PFPMA content in the monomer feed because of the high Tg of poly(PFPMA). Therefore, the Tg of the bulk copolymer increases with increasing PFPMA content when the bulk has a small amount of residual monomers, whereas it decreases when the bulk has numerous residual monomers because of the plasticization effect. This is why the Tg plots for the bulk copolymers exhibit a positive deviation. The most important aspect of these experiments is that the Tg of the bulk copolymers is sufficiently high, and that the bulks—across a wide range of copolymer compositions—can be used as the base material for GI POFs as a transmission medium in home networks. In the copolymer study, both the core and cladding layers were designed to have good stability against humidity, and the minimum Tg of a GI POF was controlled to be over 90 °C, ensuring long-term usage in home environments.

Figure 12.

Change in accumulated copolymer composition with polymerization conversion when the initial MMA and PFPMA contents in the monomer feed are 50 mol%. From Koike, K.; Koike, Y., Polymer 2010, 51, 1377–1385, © Elsevier, reproduced by permission.

Perfluorinated Polymer

For further reduction of fiber attenuation, perfluorinated polymers have also been intensively investigated. Compared to a large number of radically polymerizable hydrocarbon monomers, only a few classes of perfluoromonomers can homopolymerize under normal conditions via the free radical mechanism. The most typical example of a perfluoromonomer that can be prepared via free radical polymerization is tetrafluoroethylene (TFE), developed by DuPont in 1938. As is well known, poly(TFE) (Teflon™) is opaque despite a lack of C[BOND]H bonds. In general, perfluorinated resins are rigid and easily form partially crystalline structures.68 Hence, light is scattered at the boundary between the amorphous and crystalline phases, causing haziness. To avoid the crystalline form, an effective method is to introduce aliphatic rings into the main chain, which becomes twisted and incapable of forming a crystalline structure. The most famous examples of amorphous perfluorinated polymers are Teflon™ AF and CYTOP™,69 which were developed by DuPont and AGC, respectively. Their chemical structures are shown in Figure 13. Both have excellent clarity, solubility in fluorinated solvents, thermal and chemical durability, high electrical isolation, low water absorption, and low dielectric properties. In particular, their high transparencies arise from cyclic structures existing in the polymer main chains. Teflon AF is a copolymer of perfluoro-2,2-dimethyl-1,3-dioxole (PDD), which posses a cyclic structure in its monomer unit, and TFE. On the other hand, CYTOP is a homopolymer of perfluoro(4-vinyloxyl-1-butene) (BVE), and cyclopolymerization yields the cyclic structures (penta- and hexa-cyclic) on the polymer chain.70

Figure 13.

Chemical structures of Teflon AF and CYTOP.

Although both have sufficient properties as the core material, Teflon™ AF has mainly been utilized as the cladding layer of optical fibers or waveguides because of its extremely low refractive index (nD = 1.29). Given that this value is lower than even that of water, liquid-core optical fibers (LCOFs) have been studied for other purposes such as Raman spectroscopy analysis.71, 72 In 2000, AGC commercialized the first CYTOP-based GI POF (Lucina™), which has been adopted by condominiums, hospitals, data centers, and other facilities in Japan.73 Figure 14 is a comparison of the attenuation spectra for PMMA- and CYTOP-based GI POFs. CYTOP molecules consist solely of C[BOND]C, C[BOND]F, and C[BOND]O bonds. The wavelengths of the fundamental stretching vibrations of these atomic bonds are relatively long; therefore, vibrational absorption loss of CYTOP at the region of the light source wavelength is negligibly small. In addition, it has a fairly low light scattering property because of the low refractive index (nD = 1.34) (see eq 14). The attenuation of CYTOP-based GI POF is ∼10 dB/km at 1.0 μm. Given that the theoretical limit of attenuation is 0.7 dB/km at this wavelength,74 it is expected that the attenuation can be lowered further by preventing contamination during the fabrication process. The reason why the attenuation at 670–680 nm is slightly higher than that at around 1.0 μm is that the intrinsic scattering loss expressed by eq 14 is inversely proportional to λ4. In this case, the effect of the anisotropic scattering expressed by eq 15 need not be considered because the anisotropy of the polarizability in CYTOP is negligibly small.

Figure 14.

Comparison of attenuation spectra between PMMA- and CYTOP-based GI POFs. From Koike, Y.; Ishigure, T. J Lightwave Technol 2006, 24, 4541–4553, © IEEE, reproduced by permission.

In fact, the excellent low-loss characteristic of CYTOP-based GI POF (e.g., 10 dB/km) is far beyond the requirement of home networking. However, its real uniqueness is the low material dispersion derived from the low refractive index. In the past, there have been several reports on perdeuterated PMMA (PMMA-d8) as a novel base material for low-loss POFs.75–77 The replacement of deuterium for hydrogen in PMMA also results in considerable reduction in the C[BOND]H vibrational absorptions in the infrared region and in its overtones in the visible-to-near infrared region. As a result, loss reductions of 20 and 63 dB/km at 670–680 nm for SI POF76 and GI POF,77 respectively, have been successfully achieved, but the substitution of hydrogen with deuterium does not lower the refractive index and material dispersion. The bandwidth of optical fibers that excite many modes (e.g., POFs) is predominantly influenced by modal dispersion. However, once the modal dispersion is reduced by forming a GI profile, the influence of material dispersion on bandwidth can no longer be ignored. Material dispersion is induced by the wavelength dependence of the refractive index of the fiber material and the finite spectral width of the light source.

The refractive index profiles of GI POFs can be approximated by the following power law:78

equation image(22)

Here n(r) is the refractive index n at radius r of the fiber; n1, the refractive index at the core center; n2, the refractive index of the cladding; R, the core diameter; g, the refractive index profile coefficient; and Δ, the relative index difference defined as

equation image(23)

The crucial factor defining the refractive index profile in GI POFs is the coefficient g, and the optimum value for maximizing the bandwidth can be determined from the modal and material dispersions.79–81 From analyses using the Wentzel-Kramers-Brillouin (WKB) method, the modal dispersion σinter, material dispersion σintra, and total dispersion σtotal can be expressed as follows:

equation image(24)
equation image(25)
equation image(26)
equation image(27)

Here, the spectral width σs is the wavelength dispersion of the input pulse; C, the velocity of light; λ, the wavelength of light; and L, the transmission distance. Before turning to the material dispersion of CYTOP, we first consider the relationships among the refractive index coefficient g, modal dispersion, material dispersion, and −3 dB bandwidth by using a typical example. Figure 15 shows the relationship between the theoretical bandwidth and the index exponent g for GI POFs based on PMMA at an operating wavelength of 650 nm. The theoretical bandwidth was calculated using the wavelength dispersion of the light source at σs = 1.0 and 3.0 nm with fiber length = 100 m. Figure 15 indicates that the theoretical limit of the bandwidth is ∼4.0 GHz over 100 m when the spectral width of the light source is 1.0 nm. When the index exponent g differs from the optimum value, the spectral width dependence of the bandwidth is negligibly small since σinter is considerably larger than σintra. However, when g ranges between 2 and 3, there is a significant difference between these two curves; this is the effect of the material dispersion of PMMA and the dopant included in the core. The experimentally measured bandwidth shown by the closed circles is well predicted by taking into account the material dispersion.

Figure 15.

Relationship between index profile coefficient g and −3 dB bandwidth over 100 m of PMMA-DPS-based GI POF at a wavelength of 650 nm. Solid lines are the calculated results. Closed circles are the measured bandwidths of GI POFs prepared by the interfacial-gel polymerization technique (spectral width is 3.0 nm). From Koike, Y.; Ishigure, T., J Lightwave Technol 2006, 24, 4541–4553, © IEEE, reproduced by permission.

Figure 16 shows a comparison of material dispersions among PMMA, CYTOP, and silica glass. The material dispersion of CYTOP is even lower than that of silica glass. The theoretical dependence of the bandwidth of the CYTOP-based GI POF and a multimode GOF on wavelength is shown in Figure 17. CYTOP-based GI POFs are predicted to have a higher bandwidth than multimode GOFs.54 Indeed, in 1999, AGC and Bell Laboratories reported successful experimental transmission at 11 Gbps over 100 m using CYTOP-based GI POFs.82 Furthermore, CYTOP-based GI POFs capable of 40 Gbps transmission over 100 m were subsequently reported in 200783 and 2008.84, 85 Figure 18 is the eye-pattern of 40 Gbps data transmission through a CYTOP-based GI POF. The eye-pattern method is described in greater detail elsewhere. Suffice it to say that the good eye opening clearly shows that this fiber sufficiently ensures 40 Gbps transmission. These are highly significant results that demonstrate that POFs can achieve a higher bandwidth than multimode GOFs.

Figure 16.

Comparison of the material dispersion of fiber materials for pure CYTOP, PMMA, and silica. From Koike, Y.; Ishigure, T., J Lightwave Technol 2006, 24, 4541–4553, © IEEE, reproduced by permission.

Figure 17.

Dependence of the theoretical −3 dB bandwidth on wavelength in the CYTOP-based GI POF link compared to that in a multimode GOF (σs = 1.0 nm). Bandwidth is calculated on the assumption that each fiber has an optimal refractive index profile at 850 nm. From Koike, Y.; Ishigure, T., J Lightwave Technol 2006, 24, 4541–4553, © IEEE, reproduced by permission.

Figure 18.

Eye diagram of 40 Gbps data transmission at 1550 nm propagated through the 100-m GI POF based on CYTOP.

Recently, demands for the replacement of electrical with optical signals are rapidly increasing not only for in-home networks but also for those of shorter reach, such as in-device and on-board networks. In these areas, the required bit rate is forecasted to be tens of Gbps in the next decade.86, 87 Although multimode GI GOFs are currently considered for the transmission medium, the above data show that POFs can offer better performance even in such extremely high-speed connections. Moreover, as we have mentioned several times, POFs can also provide ease of handling and flexibility in wiring design with its features of plastic-specific robustness. In 2010, AGC released another CYTOP-based GI POF called Fontex™.88 By employing a double cladding structure—a thin layer with considerably lower refractive index is placed around the first cladding—the bending loss was further reduced while high-speed capacity was maintained,89 thereby enabling various wiring designs to become possible (Fig. 19). In addition, the continuous fabrication of GI POF has also been established by the coextrusion process.

Figure 19.

Photographs of the double-cladding CYTOP-based GI POF (FONTEX) provided by AGC.


This article reviewed the status of POF developments in the last half century that have focused on loss reduction and bandwidth enhancement. Today, optical fibers are ubiquitous and can be encountered anywhere—in networks such as submarine links, long-distance terrestrial networks, metropolitan and access loops, and FTTH drop architectures. However, they are still not used in home networks. Currently, several technologies are available for broadband home networking. In particular, coaxial cables, twisted-pair cables, and wireless local area network links are being intensively investigated. However, from the standpoints of transmission speed, reliability, ease of handling, and safety, GI POFs appear to be the best solution. Furthermore, POFs are now ready not only for use in home networking but also for device and storage interconnections where data transmissions over 40 Gbps will be required in the near future. POFs are no longer just alternatives to GOFs; their advantages for short-range networks are obvious.

In the future beyond FTTH services, there awaits a world in which all information will be directly transferred by light signals, bringing us back to real-time face-to-face communications with large-screens and clear motion pictures. We believe that POF technologies will contribute in no small way to accelerate this paradigm shift.


This work was supported by the Japan Science and Technology Agency through a grant from the ERATO-SORST Koike Photonics Polymer Project.

Biographical Information

original image

Yasuhiro Koike was born in Tokyo, Japan, in 1954. He received his MS and PhD degrees in applied chemistry from the Graduate School of Keio University in 1979 and 1982, respectively. He was a visiting researcher in AT&T Bell Labs from 1989 to 1990. Since 1997, he has held the post of professor at Keio University. He received an Honorary Doctorate at Eindhoven University of Technology in 2007, and he has been an affiliate professor of the University of Washington since 2009. He has also been a general chair of the POF Consortium since 1994. He led the “High-Speed POF Project” of the Kanagawa Academy of Science and Technology (1995–1998), and was the project leader of the “POF Project” of the Telecommunications Advancement Organization of Japan (1998–2001). He conducted the GigaHouse Town™ Project of Keio Engineering Foundation (2002–2007). Since 2002, he has been the research director of the Koike Photonics Polymer Project of ERATO, Japan Science and Technology Agency (JST). He is a Member of the Science Council of Japan since 2006. His project “Face-to-Face Communication Business by Ultra High-Speed Plastic Optical Fiber and High-Definition Photonics Polymer” funded by the Cabinet of Japan began in April 2010.

Biographical Information

original image

Kotaro Koike was born in Toyama, Japan, in 1982. He received his BS degree in applied physics and physico-informatics from Keio University in 2006. He received his MS and PhD degrees in integrated design engineering from the Graduate School of Keio University, in 2007 and 2010, respectively. During his doctoral studies, he worked as a research assistant for the Exploratory Research for Advanced Technology and Solution-Oriented Research for Science and Technology (ERATO-SORST) project of the Japan Science and Technology Agency (JST). He is presently a research associate at the Keio Photonics Research Institute, Keio University and a visiting research scholar at the Polytechnic Institute, New York University. His current research interests are in developing novel photonics polymers with excellent transparency and achieving thermal stability for high-speed POFs, which will be key technologies for automobile and aircraft networking in the future.