Properties and Radiation Response of Optical Fibers: Role of Dopants

Transcription

1 UNIVERSITÉ JEAN MONNET OF SAINT-ETIENNE (FRANCE) and UNIVERSITÁ DEGLI STUDI OF PALERMO (ITALY) Cotutelle Ph.D. Thesis Giusy Origlio Properties and Radiation Response of Optical Fibers: Role of Dopants TUTORS: Prof. Youcef Ouerdane Prof. Marco Cannas Ph.D. RAPPORTEURS: Prof. Roberto Boscaino Prof. Linard Skuja

2

3 To my new and to my old family

4

5 Contents Contents 1 Introduction 1 I State of the art 3 1 The silica optical fibers General structure Light propagation in step-index optical fibers Single-Mode and multi-mode optical fibers Dispersion and losses in fibers Fiber Fabrication Dopants in optical fibers Germanium doped optical fibers Fluorine doped optical fibers Phosphorus doped optical fibers The optical fibers under irradiation exposure Point defects in optical fibers Intrinsic point-defects Oxygen Deficient Centers

6 3.1.2 Oxygen associated hole centers Extrinsic point-defects Ge-related defects P-related defects II Materials and methods 37 4 The canonical samples Tested optical preforms and fibers Experimental set-ups Irradiations UV laser irradiations γ-ray and X-10 kev irradiations Absorption Photoluminescence and Raman spectroscopy Photoluminescence Stationary and time resolved luminescence setup Photoluminescence under synchrotron radiation excitation Raman measurements Confocal Micro-spectroscopy setup Electron Paramagnetic Resonance measurements III Ge-doped fibers and preforms 59 6 Measurements on non-irradiated samples Discussion: the drawing effect Effects of the UV and X-ray irradiation 69

7 7.1 EPR results Optical absorption Discussion: radiation effects Localization of defect species Generation processes of GECs defects The drawing effect IV Influence of further dopants: fluorine and phosphorus 83 8 F-doped fibers and preforms Raman results EPR measurements on irradiated samples Results Discussion: generation of E centers P-doped fibers and preforms Optical activity of P-related point defects Absorption and photoluminescence analysis Discussion: luminescent P-defects structure Conclusions Conclusions 107 List of related papers 109 List of communications to congresses 111 Bibliography 113

8

9 Introduction Today the circulatory system that sustains our communication society is made up by optical fibers. These low-loss glass fibers facilitate worldwide broadband communication such as the Internet. Light travels in thin guides of glass, and it carries almost all of the telephony and data traffic in every direction. Text, music, images and video can be transferred around the world in a fraction of second. If unraveled, all of the glass fibers that wind around the globe would turn into a single thread over one billion kilometers long-which is sufficient to encircle the globe more than times-and which is still increasing by thousands of kilometers every hour. Global communication, and in particular internet and long-distance telephony, is now based mainly on optical fiber technology. The main benefit resulting from the use of optical waves with respect to radio waves is the high frequencies that allow high data transmission rate. Today, it is possible to transmit several terabits per second in a single fiber and that represents an improvement by a factor of one million to what could be obtained fifty years ago with radio signal transmission. The number of optical fiber cables being installed all over the world is increasing rapidly. Fiber optics is also important for a huge number of other applications in medicine, laser technology and sensors. In order to be developed and manufactured, the optical fiber needed modern glass technology. Furthermore, a reliable light source was also needed and this was provided by semiconductor technology. Finally, a clever network needed to be assembled and extended, consisting of transistors, amplifiers, switches, transmitters and receivers, as well as other units, all working together. This telecommunications revolution was made possible by the work of thousands of scientists and inventors from all around the world. Even if the optical fibers have been so intensively investigated over the years, the interest of the scientific community is still alive: in fact the Nobel Price in Physics 2009 was awarded to C. K. Kao, whose discoveries have paved the way for optical fiber modern technology. In 1966, Kao understood that it was not imperfections in the fiber thread that was the main responsible for losses, instead it was the glass that had to be purified, because of the presence of defects. He admitted that this would be feasible but very difficult. The goal was to manufacture glass of a transparency that had

10 2 Introduction never been attained before. Even in the glass fiber of the highest purity, the signal, however slightly, is reduced along the way and needs reinforcement when it is transmitted over longer distances. This task previously required electronics, while it is nowadays performed by optical amplifiers. This has allowed to overcome the unnecessary losses that occur in the transformation of light to and from electronic signals. Furthermore, choosing which fiber to use is subject to so many different technical considerations, communication needs and costs, that it is not possible to speak of only one single kind of fiber. The fibers are based on a complex interplay between size, material properties, and wavelengths of light. Following the interest in the field, this Thesis deals with the experimental study of the spectroscopic properties of three types of prototype preforms and associated fibers. The samples have been designed and fabricated to investigate the role of germanium (Ge), fluorine (F) and phosphorus (P) doping elements on the fiber attenuation and eventually on the radiation sensitivity of silica-based glasses. We characterized the behaviors of these canonical samples before, during and after irradiation through several spectroscopic techniques, to obtain global information (electron paramagnetic resonance) or spatially-resolved information (confocal microscopy, absorption and luminescence on preform). The Thesis is organized in four parts. Part I, comprising Chapters from 1 to 3, deals with the general system of optical fiber communication providing an extensive overview of the history, construction, operation, and benefits of optical fiber, with particular emphasis on the importance of the doping procedure to enhance the fiber characteristics. An overview of the main intrinsic and extrinsic defects in silica is also presented. Part II includes Chapters 4 and 5 and it is devoted to the description of the prototype samples and of the adopted experimental techniques. Part III, including Chapters 6 and 7, reports on the experiments on Ge-doped samples and their main results. The results concerning F-doped and P-doped samples are reported and discussed in Part IV, comprising Chapters 8 and 9. Finally, the most relevant "conclusions" are summarized. A "list of the scientific papers" comprising the results presented in this Thesis and a few others on closely related topics are reported at the and of the manuscript, together with a list of communications to congresses.

11 Part I State of the art

12

13 Chapter 1 The silica optical fibers Optical fibers lie at the very heart of modern society, providing the information superhighways required within our global communication systems. Fiber-optic communication is based on the principle that light in a glass medium can carry more information over longer distances than electrical signals can do in a copper or coaxial medium or radio frequencies through a wireless medium. The purity of today s glass fiber, combined with improved system electronics, enables fiber to transmit digitized light signals hundreds of kilometers without amplification. With few transmission losses, low interference, and high bandwidth potential, optical fiber is an almost ideal transmission medium. Thanks to high transmission speed, low attenuation and interference and the large bandwidth, optical fibers represent at now the major progress in data transfer. The purpose of this chapter is to introduce the main basic fiber features starting with a description of some general properties of silica optical fibers and showing the reasons of the massive scientific investment in optical fiber telecommunications technology nowadays. 1.1 General structure An optical fiber is essentially a dielectric waveguide used for information transfer. The fiber communication is based on the principle that the light in a glassy medium can carry more information and at longer wavelengths in comparison to the electrical signal transferred by classical cables. The usual telecommunication optical fibers are made of two cylindrical parts: the interior part is called core, the exterior one is the cladding (Figure 1.1). Core and cladding have different refractive index: the exterior part has a smaller refractive index than the inner one, allowing light reflection according to the classical geometrical optics laws. The two fiber portions are generally made of the same glassy material where the

14 6 1. The silica optical fibers jacket cladding n 2 buffer core n 1 Figure 1.1: Structure of a classical optical fiber for telecommunications. refractive indexes are varied and accurately controlled during the fiber fabrication through dopant incorporation in the silica-based matrix. A classical fiber for telecommunication has an exterior diameter of about 125 µm, while the core diameter varies from few µm to 60 µm, depending on the network requirements a. Such a fiber could prove to be mechanically fragile, so it is necessary to cover it with two protective coatings: a first vitreous buffer and a polymeric exterior jacket Light propagation in step-index optical fibers The most basic function of a fiber is to guide light, i. e., to keep light concentrated over longer propagation distances despite the natural tendency of light beams to diverge, and possibly even under conditions of strong bending. A very important concept in fiber optics is that of waveguide modes. These are field configurations which maintain their intensity profile during propagation, apart from possible power losses. Of highest interest are usually the guided modes, i. e. those modes which have significant intensity only in or near the core. Depending on the fiber design and the optical wavelength, some number of guided modes may exist, or only a single one, or even no guided mode at all. A fiber with only one guided mode is called a single-mode fiber, and multi-mode fibers support several guided modes (section 1.1.2). a We can also find multimode fibers with cores of 100, 200 µm, or more. Such fibers are used for peculiar applications (like sensors) and they are not routinely used for telecommunication networks.

15 1.1. General structure 7 The propagation mechanism inside an optical fiber can be approximatively described by geometrical optics principia. The used ray picture cannot be applied to fibers with a small core or a small refractive index contrast between core and cladding: the approximation scale improves with reduction in λ/r ratio, where r is the optical fiber core and λ is the light wavelength propagating inside. The reason is that wave effects occur: a real beam has some finite width, and the incident and reflected wave interfere with each other. Furthermore, the optical field somewhat extends beyond the core/cladding interface. Therefore, the ray picture is only a rough approximation for strongly guiding large core fibers, while a wave analysis through Maxwell equations is required for the more general case. The research activity described in this PhD Thesis is related to multi-mode optical fibers, for which the ray optics approximation is effective: therefore only a description of light propagation on fiber by means a geometrical approach is here supplied. The loss of generality that such choice implies is partially balanced by the immediate physical interpretation of the results and the simple visualization of the propagation processes. It is common to explain the guiding effect as a result of total internal reflection. A light beam S approaching the separation interface between two transparent and homogenous media, with refractive index n 1 and n 2 respectively, is partially reflected and partially refracted. If θ 1 is the incident angle against the normal direction, the refractive beam will propagate in n 2 medium in accordance with the Snell law: n 1 sin(θ 1 ) = n 2 sin(θ 2 ) (1.1) The same mechanism is involved in the optical fiber: the core refractive index n 1 is greater than the cladding one n 2 and the refracted angle is greater than the incident one (θ 1 <θ 2 ). For incidence angle values greater than θ 1 =θ c =arcsin(n 2 /n 1 ), called critical angle, there is no refracted angle. This is the total internal reflection phenomenon which is at the basis of the optical fibers working. All the rays propagating inside the core with an angle θ > θ c, will be totally reflected and trapped inside the fiber. Typical values for the optical fibers are n 2 = 1.475, n 1 = 1.5; θ c = It is possible to define an acceptance cone (Figure 1.2) containing all the rays propagating inside the core through total internal reflection. The cone vertex is the center of the entry face of the fiber and the vertex angle is called acceptance angle θ A. It is also possible to get a measure of the coupling efficiency between the source and the fiber defining the so called numerical aperture (N.A.) defined by b : N.A. = sin(θ A ) = n 2 1 n 2 2 (1.2) b If the core radius a is much larger than the wavelength λ, a geometrical-optics description for the propagation of light is valid. However, when a is in the order of λ, a wave-propagation theory is needed.

16 8 1. The silica optical fibers Ray outside the acceptance cone θ A Ray lost into the cladding Acceptance cone Figure 1.2: Acceptance cone in an classical optical fiber Single-Mode and multi-mode optical fibers As anticipated in section 1.1.1, based on the number of modes propagating through the fiber, there are multi-mode and single mode fibers [1]. Figure 1.3: Optical fiber sizes Multi-mode fibers Multi-mode fiber was the first type of fiber to be commercialized. It has a much larger core than single-mode fiber, allowing hundreds of modes of light to propagate through the fiber simultaneously. Multi-mode fibers routinely used for telecommunication networks have core sizes of 50 to 62.5 µm in diameter, while the overall diameter is about 125 to 200 µm

17 1.1. General structure 9 (Figure 1.3). Based on the refractive index profile we have two types of fibers: (a) Step index fiber (b) Graded index fiber. (a) Step index fiber: in the step index fiber, the refractive index of the core is uniform throughout and undergoes an abrupt or step change at the core cladding boundary. The light rays propagating through the fiber are in the form of meridional rays which will cross the fiber axis during every reflection at the core cladding boundary and are propagating in a zig-zag manner as shown in Figure 1.4a. When light is launched into a multi-mode fiber, multiple guided modes can be excited, and at the fiber exit, there is an intensity profile which arises from the interference of light in all these modes. In this kind of fiber there is a considerable modal dispersion (section 1.1.3): even rays with the same wavelength but emitted at a different incident angles (lower than the acceptance angle) propagate with the same speed into the fiber but across different length paths. They will arrive at the fiber end at distinct times, producing a temporal broadening of the transmitted pulse. (b) Graded index fiber: in the graded index fiber, the core refractive index is made to vary in a parabolic manner so that the maximum value of refractive index is at the center of the core. The light rays propagating through it are in the form of skew rays or helical rays which will not cross the fiber axis at any time and are propagating around the fiber axis in a helical or spiral way as shown in Figure 1.4b. In the case of multi-mode graded index fiber, signal distortion is very low because of self-focusing effects. Here the light rays travel at different speeds in different paths of the fiber because of the parabolic variation of refractive index of the core. As a result, light rays near the outer edge travel faster than the light rays near the center. In fact, the rays are continuously refocused as they travel down the fiber and almost all of them reach the exit end of the fiber at the same time due to the helical path of the light propagation. Multi-mode fibers are strongly required when light from a source with poor spatial coherence has to be transported. As an example, the output of a high power diode bar contains thousands of modes, and requires a correspondingly large number of fiber modes. Additionally, the larger core diameter facilitates the use of lower-cost optical transmitters and connectors. In most applications, the standard multi-mode graded index optical fibers have significant performance advantages over conventional copper-based systems: they are very useful when multimodal transmission is needed for relative long distances. However, performance requirements and cost restraints may prohibit the use of these fibers in certain applications. First of all it is very complex and expensive to realize a graded-index fiber in which the refractive index varies continuously during all the fabrication process. Sometimes fiber manufacturers modify standard material composition and structural design to meet these additional requirements. The intent of each change is to increase performance and reduce cost. For instance it

18 10 1. The silica optical fibers Figure 1.4: Different propagation modes in (a): multi-mode step index, (b): multi-mode graded-index and (c): single-mode step-index fibers. is possible to obtain the optimal characteristics of a graded-index fiber in the so called multistep index fiber. As its name indicates, the structure, showed in Figure 1.5, uses multiple step indexes which approximate the parabolic curve of the refractive index profile. Although the Figure 1.5: Refractive index profile in multi-step index fiber. basic principle is the same as that of step index fiber because the index of refraction changes in multiple steps, the locus of the light is shifted toward the center at the same time. In any case, with enough number of steps, differences from a graded-index become small and they could be neglected. So it is possible to reconcile the advantage of a little modal dispersion with a more easy fiber production at reasonable prices. Single-mode Fibers

19 1.1. General structure 11 In a single mode fiber, only one mode can propagate through its core (Figure 1.4c). The single mode fiber has a smaller core diameter (10 µm, Figure 1.3) and the difference between the refractive indices of the core and the cladding is very small. Its fabrication procedure could be very difficult and the launching of light into single mode fibers is also hard. The advantages of single mode optical fibers lie in the very low transmission loss and dispersion or degradation, thus resulting very useful in long distance communication Dispersion and losses in fibers Dispersion in the fiber means the broadening of the signal pulse width due to dependence of the refractive index of the material of the fiber on the wavelength of the carrier. If we send digitized signal pulses in the form of square pulses, they are converted into broadened gaussian pulses due to dispersion. The dispersion leads to the distortion or degradation of the signal quality at the output end due to overlapping of the pulses. There are two kinds of dispersion mechanisms in the fiber: intramodal dispersion and intermodal dispersion. The first one arises due to the dispersive properties of the optical fiber material (material dispersion) and the guidance effects of the optical fiber (waveguide dispersion). Further it increases with the increase in spectral width of the optical source. Intermodal dispersion or multi-mode dispersion arises due to the variation of group velocity for each mode at a single frequency. Different modes arrive at the exit end of the fiber at different times. So there is multi-mode dispersion and hence there is broadening of the signal pulses. The multi-mode step index fibers exhibit a large value of dispersion due to the enormous amount of multi-mode dispersion which gives the greatest pulse broadening. At the same time the multi-mode graded index fiber exhibits an overall dispersion which is 100 times lesser than the multi-mode step index fiber s dispersion. This is due to the shaping of the refractive index profile in a parabolic manner. In the case of single mode step index fibers, they have only intramodal dispersion. Attenuation is the reduction of signal strength or light power over the length of the lightcarrying medium. Fiber attenuation is measured in decibels per kilometer (db/km) and it is a function of wavelength as shown in Figure 1.6. Attenuation is caused by several different factors, but primarily diffusion (Raileigh scattering) and absorption. It can be classified into two types: intrinsic and extrinsic losses generated by several mechanisms: Tail of infrared (IR) absorption by Si-O coupling that it is present at higher wavelengths around 1.4 µm to 1.6 µm. Tail of ultraviolet (UV) absorption due to electron transitions and present at lower wavelengths near 0.8 µm. This produces a loss of 0.3 db/km.

20 12 1. The silica optical fibers Figure 1.6: Spectral attenuation of a silica optical fiber. Rayleigh scattering (Figure 1.7) originates from microscopic irregularities in the glass Figure 1.7: Illustration of Rayleigh scattering effect. structure; it is inversely proportional to λ 4 and in many cases it can be expressed as c : ( ) α R [db/km] = 1.7 (1.3) λ[µm] It produces high losses mainly in the ultraviolet region. In the wavelength region around 0.8 µm to 1 µm, it gives a loss of 0.6 db/km. Absorption: conversion process of electromagnetic wave energy into other forms (i. e. lattice vibration). Intrinsic silica glass absorption occurs in both ultraviolet and infrared c Equation 1.3 is sample-dependent: actually Rayleigh losses depend on the core composition. The formula predicts 0.15 db/km at 1.57 µm, while lower Rayleigh losses of 0.12 db/km have been reported by Nagayama et al. [2].

21 1.1. General structure 13 bands, in particular infrared absorption tail causes attenuation for the wavelengths longer than 1.6 µm. Further attenuation is caused by light absorbed by residual species, such as metals or OH ions, within the fiber core and inner cladding. In particular OH causes the water peak region on the attenuation curve, typically around 1.4 µm. The removal of OH ions is of primary interest to fiber manufacturers as this water peak has a broadening effect and contributes to attenuation loss for nearby wavelengths. Figure 1.8 shows the spectral attenuation of different material fibers. Figure 1.8: Spectral attenuation of different material fibers. For silica fiber, the lowest losses of about 0.12 db/km can be obtained in the region around 1.55 µm [2]: at longer wavelengths, the attenuation increases. An optical signal transmitted through fiber, could travel more than 100 km without regeneration or amplification. Other attenuation mechanisms are due to macroscopic bends, occurring when installing fibers, microscopic bends, due to local distortions of the fiber geometry, and nonlinear scattering. Optical power propagating in a fiber decreases exponentially with distance: P (z) = P 0 exp( α z) (1.4) where P is the optical signal power and α is the attenuation coefficient [1/km]. Using a logarithmic scale we obtain: log P (z) = αz/10db + log P 0 (1.5) where α is the logarithmic attenuation coefficient measured in [db/km]. Overall optical fibers offer superior performances over other transmission media because they combine high bandwidth with low attenuation. These properties allow the transmission of signals over longer distances while using fewer regenerators or amplifiers, thus reducing cost and improving signal reliability.

22 14 1. The silica optical fibers Fiber Fabrication The manufacture of an optical fiber takes place into two steps: the preform fabrication and the drawing process. Preform is a cylinder of silica composition from 10 mm to some cm in diameter and from 60 to 120 cm length d. It consists of a core surrounded by a cladding with a desired refractive-index profile; in other words, this is a desired optical fiber, but on a much larger scale. The main reason a preform is prepared is to have a drawable material that is clean, low in OH concentration, low in metallic-ion contaminants, and inexpensive. Many techniques have been developed to prepare these preforms. Some common commercially used methods are Outside Vapor-Deposition (OVD), Modified Chemical Vapor Deposition (MCVD), Vapor Phase Axial Deposition (AVD), and Plasma Chemical Vapor Deposition (PCVD) and Plasma Modified Chemical Vapor Deposition (PMCVD) [3]. All these methods are based on thermal chemical vapor reaction in which two gases, SiCl 4 and O 2, are mixed at a high temperature (>800 C) to produce silicon dioxide (SiO 2 ): SiCl 4 + O 2 SiO 2 + 2Cl 2 (1.6) Silicon dioxide, or pure silica, is usually obtained in the form of small particles (about 0.1 µm) called soot. This soot is deposited on the target rod or tube layer upon layer and it forms a homogeneous transparent cladding material. To change the value of the cladding s refractive index, some dopants are used. For example, fluorine (F) is used to decrease the cladding s refractive index in a depressed-cladding configuration. The soot for the core material is made by mixing several gases which results in a mixture of SiO 2 and of the core dopant. The degree of doping is controlled by changing the amount of dopant gas added to the mixture. Since deposition is made by the application of silica layers, the manufacturer can control the exact amount of dopant added to each layer, thus controlling the refractive-index profile. The different preparation methods differ mainly by the way the soot is deposited. The preforms studied in this PhD thesis were all made by MCVD process which provided a simple and straightforward means of manufacturing high-quality optical fibers. This method was developed by Bell Laboratories [4]. The soot is deposited on internal wall of the tube (Figure 1.9) and then vitrified by the traversing burner to provide a thin glass layer. The procedure is repeated many times as the cladding and core layers are formed. When the deposition is finished, the temperature of the burner is increased ( 1700 C) to collapse the tube into a solid preform [5]. The entire process is highly automated and all process parameters are precisely controlled. Optical fibers are obtained by drawing from the preform at high temperature ( 2000 C). The drawing process must be integrated with the coating process to avoid contamination of fiber surface. These processes are shown schematically in Figure The tip of the preform is d It remains difficult to have an idea of the maximum preform diameter as this is confidential for the fiber manufacturers

23 1.1. General structure 15 Figure 1.9: Deposition by modified chemical vapor deposition (MCVD) process. Figure 1.10: Optical Fiber Drawing Process heated in a furnace to a molten state. Formed molten gob falls down under the force of gravity while shrinking in diameter into a proper diameter strand. It is controlled continuously during the drawing process. Diameter drift cannot exceed 0.1%. The strand is threaded through a series coating applicators immediately after drawing. Liquid prepolymer coatings are cured by thermal or ultraviolet apparatus. Dual coating, soft inner and hard outer, is needed to protect against impact and crushing forces in both manufacturing process and installation.

24 16 1. The silica optical fibers The fiber with coatings is pulled down and wound on a winding drum. The drawing process must take place in controlled atmosphere, because air pollution influences fiber attenuation. Both stages of fiber manufacturing are fully automated and are performed in a clean, climatecontrolled room. Obviously, the manufacturers use high-precision measuring equipment to automatically control each step of the fabrication process. For example, preform analyzers measure the critical characteristics of the optical-fiber preform. Also, specific measurement systems control fiber geometry, the refractive-index profile, and the coating geometry.

25 Chapter 2 Dopants in optical fibers As reported in section 1.1, fundamental condition to having light propagation in optical fibers is the different refractive index between core and cladding. To realize an index variation in a SiO 2, dopants are usually added in the glass matrix. Depending on the use and characteristics of the optical fibers, several elements can be added to modify the fiber characteristics. GeO 2 and P 2 O 5 are dopants commonly used for doping the core region, raising the refractive index. On the other hand B 2 O 3 or F are dopants chosen for the cladding region that in turn lower the refractive index (see Figure 2.1). Several Figure 2.1: Refractive index as a function of dopant materials and their concentration (from ref. [6]). dopants can be added in more special fibers to functionalize the glass, as rare-earth ions (erbium, ytterbium [7]) for fiber-based amplifiers, or fluorine to improving the fiber radiation hardness (see section 2.4). The nature of the elements (impurities or dopants) contained in fibers deeply influences

26 18 2. Dopants in optical fibers the optical properties of the fibers themselves: dopants can modify the fiber hardness under radiation exposure or simply influence the drawing process. The dopants need to have the following characteristics [8]: It is of high purity and easily available. It is easy to liquify. It differs from the transition metal in vapor pressure. It is easy to vitrify with silica and gives a proper refractive index. After being vitrified, its coefficient of thermal expansion is nearly equal to that of SiO 2. When vitrified, it has stable properties. The following sections are devoted to a review of the influence of three particular dopants often used in optical fiber technology: germanium, fluorine and phosphorus. 2.1 Germanium doped optical fibers The addition of germanium in the silica matrix, disguised as GeO 2, allows to increase the refractive index of the glass. This property is often used for the elaboration of the optical fiber core and Ge has been the first traditional dopant used in fiber. Ge-presence does not affect the fiber losses in the telecommunication windows ( nm) (save for the increase of Rayleigh scattering due to density fluctuations), but it can produce the apparition of new energy levels within the silica band gap, thus leading to detrimental losses of part of the transmitted signals into the fibers (see section for details). The scientific interest for germanosilicate glass increased even more after the experimental discovery of the property of photosensitivity of this material. Photosensitivity of a medium is defined as its capacity to have its refractive index permanently changed by a modification of its physical or chemical properties through UV light exposure. Photosensitivity is a complex phenomenon and it is not well understood yet because of the influence of many parameters: fiber composition, fabrication process, operation wavelength and even light sources. Photosensitivity was first observed in 1978 by Hill et al. [9] at the communication Research Centre in Canada. The experiment consisted of injecting light from a single frequency Argon laser (514 nm) into the core of a Ge-doped silica fiber. Hill observed that a fraction of the input power was reflected by the fiber itself and this phenomenon was attributed to the formation of a permanent index grating. Progress in optical fiber photosensitivity research developed rapidly after the discovery of the possibility to write Fiber Bragg Gratings (FBG) into the

27 2.1. Germanium doped optical fibers 19 fiber illuminating the core from the fiber s side with the interference pattern of two beams of coherent UV radiation [10], as shown in Figure 2.2. Spectroscopic studies of Ge-doped fibers Figure 2.2: Inscription of a Fiber Bragg Grating on the core of an optical fiber before and after intense UV exposure have been interpreted as pointing to a color center model for photosensitivity, in which a Ge-related defect optical activity (see section 3.2.1) at the exposure wavelength (242 nm) is bleached [10, 11]. Nevertheless many studies provided some additional clues about the microscopic mechanisms of photosensitivity, such as laser-induced densification [12, 13, 14, 15]. Apart from the photoinduced change of the isotopic refractive index, it was discovered in 1985 by Parent et al. that photoinduced birefringence could also be written into fibers by polarized radiation [16]. Additionally, in 1986 Osterberg et al. discovered that a prolonged exposure of an optical fiber to 1064 nm light from a Nd:YAG laser results in the generation of second harmonic light at 532 nm [17,18]: optical nonlinearity was discovered in germanosilicate glasses. These features have a deep impact in several applications and therefore they stimulate the strong interest in the study of Ge-doped amorphous silica (a SiO 2 ), to understand the microscopic mechanisms at the basis of the properties of the material with the aim to control and enhance them. Usually the largest part of scientific investigation on Ge-related glasses for optical fibers consists in the direct study of bulk samples and the subsequent transfer of information to the fibers [19, 20, 21]. However, this approach cannot take into account the peculiarities implied in the fiber preparation procedure, such as the drawing process, which can generate precursors and influence the defect generation [22] and the necessity of direct studies on fibers samples strongly emerges.

28 20 2. Dopants in optical fibers 2.2 Fluorine doped optical fibers Fluorine is an important dopant in optical fiber technology. In contrast to most of other dopants, F decreases the refractive index of silica glass [23] (Figure 2.1). This property is of great practical importance for designing optical fibers with an undoped high-purity silica core. Such fibers exhibit the best performance in the UV and IR spectral regions and have a better durability in environments with an increased level of ionizing radiation. So fluorine-doped silica is a promising key material in optical fiber technology directed to applications requiring high and stable transmission over a broad spectral range from infrared to ultraviolet. Indeed, recent studies have shown that radiation toughness of silica samples is achieved by incorporating Si F groups (Figure 2.3) whose positive effect is assumed to be the reduction of defect precursors [24, 25], such as strained bonds ( Si O) from which is likely generated the pair of silicon dangling ( Si ) and oxygen dangling ( Si O ), where ( ) and ( ) indicate bonds with three oxygen atoms and an unpaired electron, respectively (see section 3.1). Fluorine is a Figure 2.3: Schematic illustration of fluorine incorporation in a SiO 2 matrix (from ref. [24]). silica network modifier, because it considerably affects the viscosity and softening temperature of silica glass [26,27]. Recently, thanks to high transparency in the vacuum ultraviolet (VUV) range without any increase of optical defects [28,29,30] (Figure 2.4), the F-doping has received a large attention through the application of silica glass as an optical material for projection photolithography at 157 nm of F 2 excimer laser [25, 31]. It is well known that the shape of the fundamental absorption edge in the exponential (Urbach) region can yield information on the disorder effects [32]. Skuja et al. [30] demonstrate that fluorine doping affects Urbach VUV absorption edge by increasing its steepness. It is evident from Figure 2.4 that when F was doped to 1 wt%, the transmittance at 157 nm increased to 80 % in comparison to non doped glass. This gain in VUV transparency by F-doping is mainly due to a reduction of concentration of strained bonds in the silica network. Because the Si F bond is stronger than the Si O bond, a monotonic increase in the optical bandgap of fluorine doped SiO 2 would be expected with increasing the F content [25, 31]. The replacement of a single bridging oxygen atom with a terminal fluorine, results in the formation of SiO 3\2 F tetrahedra, that produces a depolymerization of the silicate network and

29 2.3. Phosphorus doped optical fibers 21 Figure 2.4: VUV absorption spectra of SiO 2 :F glasses as a function of F content. From ref. [25]. consequently the lowering the viscosity of silica [33]. The positive effects of F-doping seems to have at least 4 mechanisms: (a) by quenching of distinct color centers absorbing in the edge region, (b) by reducing of the structural disorder by breaking up the strained bonds in glass network, (c) by increasing of the band gap due to the higher energy of Si-F bond as compared to Si-O bond, (d) by reducing the glass viscosity, thus allowing to achieve more easily a lower fictive temperature of the glass. Actually, the feasibility to exploit these properties in the fabrication of F-doped silica fibers is conditioned by several queries including the dependence on F-concentration also on the basis of its influence in modulating the silica refractive index and the role of F after drawing silica preform. Despite the great importance assumed by silica glasses doped with F, their structure is not yet well characterized. It is known, as above mentioned, that fluorine is incorporated in the glass matrix essentially as Si-F bonds, taking the place of a bonding oxygen. For all glasses containing 1 wt% fluorine, a small fraction of the fluorine is bonded to silicon atoms containing four bridging oxygen atoms, resulting in fivefold coordinated silicon of the type SiO 4\2 F [27]. 2.3 Phosphorus doped optical fibers Phosphorus-doped a SiO 2 is a material of fundamental importance in fiber optics communications and in microelectronics. First of all, the addition of P 2 O 5 allows to improve the refractive index of the glass (Figure 2.1), for this reason phosphorus doping is often used in optical fibers to achieve an optimal refractive index profile [6]. Phosphorus is indeed used in optical fibers to ameliorate the internal glass structure [34] thanks to its ability in modifying the viscosity of the core and cladding regions [34]. Phosphate glasses are potentially good ultraviolet (UV) transmitting materials allowing the fabrication of thin glass films for appli-

30 22 2. Dopants in optical fibers cation in microlithography and laser systems [34]. They are also good materials for highly effective optical amplifiers, especially via co-doping with rare-earth elements [7,35], and phosphosilicate glasses are promising candidates as radiation sensors due to their closely linear response to radiation dose [36]. It has been reported that strong photosensitive properties can be induced in P-doped silica by hydrogen loading or high temperature treatment in a hydrogen-oxygen flame [37, 38]. A thorough understanding of the microscopic arrangements by which P impurities are incorporated in silica, as well as of the properties of the resulting P-related point defects, would be useful to optimize the performance of P-doped SiO 2 in applications (see section 3.2.2). Due to the fact that phosphorus is first of all used as co-dopant in fiber technology, it is particularly difficult to separate its contribution in optical fiber attenuation in the UVvisible region [39]. Only a few papers in literature have investigated this issue, so that little information is available at the moment on P-related point defects and on their generation and transformation mechanisms [40, 34, 41]. The situation is very different from the case of other dopants, like germanium for which, as seen in section 2.1, a vast amount of knowledge deriving from experimental and theoretical work has been accumulated. 2.4 The optical fibers under irradiation exposure The appearance of new radiative environments integrating silica components, such as optical fibers [42, 10], requires their immunization under ionizing radiation. In particular two important applications of optical fibers in the nuclear industry are related to plasma diagnostics in fusion reactors [43] and transmission of signals from inaccessible parts of nuclear installations [44]: in this field the relevant doses are above 1 MGy [45]. When optical fibers are subjected to radiation, whether it consists of high energy light, X-rays, γ-rays, neutrons or high energy cosmic particles, their optical properties change due to the interaction of the radiation in the fiber core and in the cladding material. The main effects result from electronic processes: electrons are excited to leave their normal (bound) position, changing the physical and chemical properties of the silica glass. The most obvious among the optical effects is the radiation induced optical attenuation, that depends essentially from wavelength: this is in general a not desirable effect because it causes degradation of the performance of the optical fiber systems. Moreover, it is also possible to use the fiber response under radiation exposure as a detector for radiation. At the present time, there are several applications for the optical fibers under radiation environments. As an example, fiber diagnostic and imaging are new interesting fields for the development of the optical fibers technology under radiation exposure. Historically, the first interest in fiber response under ionizing radiation comes from the military sphere. Moreover,

31 2.4. The optical fibers under irradiation exposure 23 due to the confidential nature of these information, very few literature data exist on this subject [46]. In contrast the fiber applications in space environment or civil nuclear environment have been largely investigated. The interaction of the radiation with the fiber material is a complex process with quite a number of dependencies on parameters related to the fibre fabrication process, operating environment and radiation type. Exposure to radiation can induce stable alterations of the material [47, 48], often related to point defects generation and conversion processes [21] (see section 3).

32 24 2. Dopants in optical fibers

33 Chapter 3 Point defects in optical fibers In 1966 Charles K. Kao a and A. Hockham, two English researchers of the British Post Office, found and demonstrated that the high-loss, till then observed in the existing optical fibers, arose from impurities in the glass, rather than from an underlying problem with the technology itself [49]. The presence of defects in optical fibers often causes the appearance of new energy levels located inside the band gap of the dielectric [50, 51]. As a consequence, the glass absorbs a more important part of the transmitted signal giving rise to an attenuation of the light guided inside and consequently in a degradation of the fibers themselves. The fiber radiation response depends on many intrinsic parameters: core and cladding dopants, impurity content, strain [52, 22], which are generally not accessible for researchers. Usually classical optical fibers for telecommunications are used in the IR, from 835 to 1600 nm, where the optical transmission is the largest. Moreover new technological fields, like medical application and plasma diagnostic, need the use of light guides in the visible and UV region were optical transmission is affected by many losses [50, 53, 54, 55]. All these aspects motivate the necessity to investigate the exact nature of point defects, checking their origin and properties and so reducing degradation effects also in the UV-visible domain. Point defects and their precursors in the amorphous silica network are introduced during the fabrication process, through dopants and the interaction with ionizing radiation (high energy, photons including UV laser irradiation, particles). Numerous publications are related with defects in a SiO 2 [56, 57, 50, 58]. Most types of defects have optical absorption and luminescence bands and could be detected by optical absorption (OA) in the visible, UV, or infrared spectral range, Raman and photoluminescence (PL) spectroscopies. Detailed information and identification on the subset of paramagnetic defects is obtained by electron paramagnetic resonance spectroscopy (EPR). Often, the combined use of different techniques a Nobel price in physics 2009

34 26 3. Point defects in optical fibers allows to infer information not available by examining separately the results of single observations. The formation of paramagnetic point-defects in silica glass has been studied from two points of view: transformation of diamagnetic precursors (also point defects) and the breaking of intrinsic Si-O bonds. Defects can be distinguished in intrinsic, when they are due to a variation of the basic silica elements (silicon or oxygen) and extrinsic, if they are related to presence of impurities in the silica matrix (H, Ge, P, etc.). Extrinsic defects due to the presence of impurities (Cl, H and so on) are always present in variable concentrations in the material. On the other hand, how above explained (see section 2), selected impurities can be deliberately added by doping to induce many useful properties [59, 51, 60]. To provide a background for the presentation of the results, the following sections of this chapter are devoted to review in more detail the current understanding of defects in a SiO 2, particularly with regard to the generation and conversion of defects related to the optical fibers. 3.1 Intrinsic point-defects Amorphous silica is the principal building material for glassy fiber waveguides. The structure of the glass network and point defects has been the subject of extensive studies through a large variety of experimental techniques and theoretical modelling [61, 62, 63, 64]. An illustrative picture of an amorphous silicon dioxide network is shown in Figure 3.1. The SiO 2 network is Si O Figure 3.1: structure of the amorphous silica, with Si atoms in grey and O atoms in black. The angle α define the spatial configuration of two connected tethraedra. built with SiO 4 tetrahedra joined at the corners so that each Si-atom is bound to four O-atoms and each O atom is the bridge between two Si-atoms. The angles defining the relative spatial orientation of each pair of connected tethraedra are statistically distributed between 120 and 180 [65,66]. This description of the microscopic structure of amorphous silica is known as the

35 3.1. Intrinsic point-defects 27 Continuous Random Network (CRN) model, and is mainly based upon the evidences coming from X-ray and neutron diffraction [62, 61, 64, 67]. The structural order in glass can generally be divided into different stages or ranges [62]. The first stage is the tetrahedron structural unit SiO 4 followed by the interconnection of adjacent units. A third stage is the network topology for describing the intermediate range order in shortest path ring structures. Finally, the long range density fluctuations over several tens of are the fourth stage of structural order. A point defect in the intrinsically disordered structure of silica can be defined as any deviation from the perfect glass structure defined by the CRN model, provided that it is localized in a region whose dimensions are comparable to the interatomic distance [62]. In the following the characterization of the main intrinsic defects are summarized Oxygen Deficient Centers Oxygen deficient centers (ODC) are formed when an oxygen is missing or removed for instance by irradiation from its Si-bonding position. The silicon dangling bond, or E center, is the most widely investigated oxygen deficient defect in a SiO 2. It consist of a silicon atom with six electrons in three pairs and one unpaired electron: Si, the symbol ( ) represents three bonds to oxygen atoms, ( ) represents one unpaired electron (Figure 3.2 (B,C,D)). E defect was observed for the first time in 1956 by Figure 3.2: Oxygen deficient centers in silica (from [50]). (A): Relaxed oxygen vacancy (ODC(I) center). (B,C): silicon dangling bond (E center) relaxed into the plane of the neighboiing oxygens (B) or relaxed towards neighboring bridging oxygen atom (C). (D): Surface-type SiE center. (E): Twofotdcoordinated Si atom (ODC(II) center). R.A.Weeks in neutron-irradiated α-quartz (E 1) and in silica [68] (Figure 3.2(B)). Thus far, at least four different types of E centers have been observed in silica: E γ, E δ [69, 70, 71, 72], E α [73], E β [69,59] (Figure 3.2(B,C,D)). They differ from each other in the second coordination environment around the respective silicon atom and they are distinguishable, at least in principle, either by their spectroscopic properties or on the basis of their generation mechanism. Additionally there exists a surface type E center [74] due to isolated silicon dangling bonds (Figure 3.2(D)). E center is found almost in every specimen exposed to radiation, but it can

36 28 3. Point defects in optical fibers be also formed by the fibre drawing process [59]. The E family of centers are paramagnetic and give rise to strong EPR signals. The EPR spectrum of the E center consists of a single main resonance line (Figure 3.3) and of four hyperfine doublets with splitting of 40 mt (strong hyperfine), 0.8 mt and 0.9 mt (weak hyperfine) and 0.05 mt (very weak hyperfine) [58, 75, 76]. An optical absorption band centered at 5.8 ev with FWHM=0.8 ev and Figure 3.3: X-band electron paramagnetic resonance spectrum of E center (from [75]). oscillator strength f=0.14, well correlates with the growth of E EPR signature [77,78] and it is usually ascribed to E, even if the nature of the optical transition involved is still controversial [60, 79]. The formation efficiency of E defects strongly depends both on the content of the hydroxyl radicals (OH) in the glass and and on the irradiation energy. It was shown by Hanafusa [80] and Hibino [81] that E defects also exist in non-irradiated optical fibers. The strong tensions during the preform drawing process seem to be at the origin of the defect formation. Also the drawing conditions, like temperature or drawing speed, influence the E concentration in optical fibers: in particular it was shown that the E concentration growth as a function of the drawing temperature, follows the Arrhenius law [80, 81]. The neutral oxygen vacancy (ODC(I)) [82] consists of a bond between two Si atoms and it is indicated as Si Si (Figure 3.2(A)). This diamagnetic ODC is electrically neutral and intrinsic to oxygen deficient silica since it disappears by oxidation. It gives rise to strong optical absorption bands around 7.6 ev [82, 83]. By hydrogen loading of silica samples it was shown that the oxygen vacancy converts itself to Si-H groups according to the reaction [50]: Si Si +H 2 Si H + H Si (3.1) The twofold coordinated Si (ODC(II)) [84] consists of a Si coordinated with two O atoms and denoted by = Si where ( ) represents two paired electrons in the same orbital (Figure 3.2(E)). This defect is also called silicon lone pair center or divalent Si. It shows a relatively weak absorption band, called B 2 band, with peak at 5.03 ev and FWHM 0.4 ev and two photoluminescence emissions at 4.3 ev and 2.7 ev to singlet-singlet and triplet-singlet transitions occurring in the same defect [50].

37 3.1. Intrinsic point-defects Oxygen associated hole centers The oxygen dangling bond or Non Bridging Oxygen Hole Center (NBOHC) ( Si O ) [50,85] is shown in Figure 3.4(A). It is detectable by its characteristic EPR signal, as well as by Figure 3.4: Oxygen excess-related color centers in a SiO 2 (from [50]). (A): Non Bridging Oxygen Hole Center (NBOHC). (B): Peroxy radical (POR). (C): Peroxy bridge. (D): Interstitial oxygen molecule. (E):Interstitial ozone molecule. its optical activity, consisting in three absorption bands at 2.0 ev, 4.8 ev and ev, which excite a photoluminescence emission peaked at 1.9 ev. All these absorbtion bands, and in particular the intense bands at 4.8 ev and 6.8 ev, make NBOHC the defect that more influences the transmission of silica in the UV and VUV spectral ranges. There are several formation mechanisms for the NBOHC. In optical fibers these centers could be created during the drawing process [86] and their concentration grows in particular with the O 2 [86] flux and the drawing tension [87]. In bulk silica the NBOHC are usually created after energetic radiation exposure (X, γ, UV) [85, 88]. The peroxy radical (POR), independently from its formation mechanism, consists in a silicon atom linked to an oxygen molecule: Si O O [58] (Figure3.4(B)). It has an unpaired electron delocalized on two oxygen atoms that are not equivalent from a chemical point of view: the electron spends 75% of the time on the more distant oxygen atom from the silicon one. Several formation mechanisms were proposed in literature to explain the formation of these defects. Like NBOHC and E, the peroxy radical can be induced during the drawing process and it can be revealed by EPR measurements at low temperature [86]. The following reaction was proposed for the POR formation: Si O O Si Si O O + Si + e (3.2) Some authors supposed that the same mechanism could be responsible for the POR formation under ionizing radiation [59, 89]. Other oxygen excess related defects are the peroxy bridge, the interstitial oxygen molecule and the interstitial ozone molecules (Figure3.4(C, D, E)). The presence of O 2 [50] in silica has been inferred from out-gassing experiments, from reaction with H 2 forming Si-OH groups and from conversion from E centers to POR centers.

38 30 3. Point defects in optical fibers Finally, the self-trapped hole (STH) may be the first defect to form under the influence of ionizing radiations. Its principal characteristic is the capture of a hole on a 2p orbital from a doubly linked oxygen atom [90]. Two different STH species were identified: the STH1 and the STH2 [91], consisting on a self trapped hole on one or two oxygen atoms respectively. 3.2 Extrinsic point-defects Among the impurities present in silica fibers, hydrogen, germanium, phosphore and fluorine are the most diffuse. How above explained, Ge, P and F are very important dopants in fiber technologies. On the other hand several other impurities are often integrated in the silica matrix whether for the difficulty in eliminated them during the preparation procedure or for improving the fiber properties Ge-related defects Germanium may be arranged within a SiO 2 in many different configurations, each of which constitutes a specific point defect. Since Ge and Si are isoelectronic elements, it is qualitatively expected that many Ge-related point defects are structurally identical to Si-related centers apart from the substitution of Si with Ge [92]. Starting from the comparison between a Gedoped silica glass and a pure silica glass, it is possible to show that defects related to germanium are predominant on the intrinsic ones [19]. This property implies an UV absorption from two to three order od magnitude more intense in germanosilicate glasses, even before irradiation exposure [93], as compared to pure silica. Actually defects in germanium doped silica, and than in germanosilicate optical fibers, are the main subject of many experimental and theoretical works in order to know the origin of the photosensitivity in these glasses (see section 2.1). Several researchers have shown that a contribution to the photosensitivity is due to the variation of the UV optical absorption spectra associated with the so called Ge-lone pair center (GLPC ) [55,54,21]: a dicoordinate germanium atom with a lone pairs (= Ge ) [94, 95]. This defect is characterized by an OA band at 5.1 ev related to the optical transition S 0 S 1 [54], from the ground state to the first excited singlet state. It has been suggested that the bleaching after radiation exposure (UV laser, γ-rays) of this OA band, often referred as B 2β, associated with the growth of several new absorption signals (Figure 3.5), is at the basis of the permanent glass refractive index change. This observation clearly suggests that the defect responsible for the B 2β band are converted by UV radiation to other centers [55, 97, 96, 94, 95]. GLPC defect is also characterized by two photoluminescence (PL) bands at 4.2 and 3.1 ev, related to the transitions from the excited electronic states of singlet (S 1 ) and triplet (T 1 ), respectively, to the ground state (S 0 ) [92,98].

39 3.2. Extrinsic point-defects 31 Figure 3.5: Difference absorption spectrum, showing the bleaching of the 5.1 ev band and the growth of two components at 4.5 ev and 5.8 ev. From Fujimaki et al. [96]. Figure 3.6: Evolution of PL signal associated with GLPC after UV pulsed laser irradiation. The inset shows the OA spectra acquired before (dashed line) and after (solid line) the irradiation. Figure adapted from Cannas and Origlio [21]. Then the energetic level scheme, pictured in Figure 3.7 and associated with GLPCs, consists of a ground singlet S 0 and the excited S 1 and T 1 states. The radiative decay channels from S 1 and T 1 are described by the rates K S and K T respectively, while the ISC process linking S 1 and T 1 is characterized by K ISC b. Though the determination of the GLPC spatial distribution in optical fibers is crucial to probe the silica refractive index variation, convenient experiments have not been performed yet, the main obstacle being the small fiber dimensions. Experimental literature data on the defect spatial distribution in fiber, exist mainly for elements of intrinsic nature [99, 100]. b other non-radiative channels can be neglected [92].

40 32 3. Point defects in optical fibers S 1 K ISC OA 5.2 ev K S PL 4.2 ev PL 3.1 ev ~ 10-8 s T 1 S 0 O Figure 3.7: General scheme of GLPC diamagnetic defect. Solid arrows indicate the radiative transition in absorption and luminescence. Dashed arrows indicate the Inter System Crossing (ISC) non-radiative transition. The most common Ge-related paramagnetic defects that are detected by EPR in irradiated Ge-doped a SiO 2 are the GeE center and the Germanium Electron Centers (GECs) Ge(1) and Ge(2) (Figure 3.8). (a) (b) (c) Figure 3.8: Microscopic structures proposed by Neustrev [19] as models for (a): Ge(1), (b):ge(2) and (c): GeE defects. The microscopic structures of GeE and Ge(1) have been unambiguously identified by EPR studies, further supported by theoretical calculations. The GeE, which is observed also in pure GeO 2, is structurally identical to the E center apart from substitution of Si with Ge ( Ge ) [101,102,55] (Figure 3.8 (c)). This center was put in relationship with an absorption band at 6.2 ev-6.4 ev [19, 103]. The Ge(1) consists in an electron trapped at the site of a substitutional 4-fold coordinated Ge precursor (GeO 4) [19,104] (Figure 3.8 (a)). An absorption band at 4.4 ev-4.6 ev has been attributed to this center [19, 19, 105, 106]. Finally the structure of the defect responsible of the latter EPR signal, the Ge(2) center, is still debated. Indeed, its structural model was first ascribed to a trapped electron center at the site of a GeO 4 unit, such as the Ge(1), on the basis of the similarities of their 73 Ge hyperfine coupling constants, differing from Ge(1) for the number of Ge nearest neighbors

41 3.2. Extrinsic point-defects 33 ions [107]. According to this attribution, an absorption band at 5.8 ev was assigned to Ge(2) center [54, 105]. However, subsequent studies, based on the defect annihilation, suggested an alternative model for Ge(2): an ionized twofold coordinated Ge (= Ge ) [19, 96]. Even the circumstance that the g value of Ge(2) is smaller than does not permit its conclusive assignment to a trapped electron center, because this line of reasoning is rigorously valid only for very simple paramagnetic centers, and generally cannot be extended to point defects in silica [108]. The EPR signals related to GeE and GECs centers are reported in Figure 3.9. Figure 3.9: EPR signature of the GeE, Ge(1) and Ge(2) paramagnetic defects in germanosilicate irradiated silica. From Fujimaki et al. [106] P-related defects Several literature papers are focused on the study and characterization of paramagnetic P- related point defects generated by ionizing radiation [40,109]. In a defect-free SiO 2 glass each oxygen would bridge two SiO 4 tetrahedra. On the other hand, the ideal P 2 O 5 glass would be characterized by one nonbridging and three bridging oxygens per phosphorus: any deviation from this structure can be considered as a defect [40]. Most of the current understanding of P-related defects in SiO 2 derives from electron paramagnetic resonance (EPR) experiments on irradiated phosphosilicate glasses. Electron paramagnetic resonance allowed to identify 4 main P-related paramagnetic point defects, referred to as P1, P2, P4 and Phosphorus Oxygen Hole Center (POHC ) centers [110, 40] that were not observable before irradiation exposure [40]. Figure 3.10 shows the supposed structures for the above mentioned P-related defects, together with the precursors suggested by Griscom et al. in ref. [40] for all these defect structures.

42 34 3. Point defects in optical fibers Figure 3.10: Main phosphorus related paramagnetic defects induced by radiation in P-doped a SiO 2 silica. The supposed precursor structures are also showed (from ref. [40]). In P4, P1 and P2, the unpaired electron is localized on the central P atom, bonded to a different number of oxygen atoms, 2, 3, 4 respectively [40, 111, 112, 38, 109]. Hence, their structure can be represented as [(O ) 2 P ] 0, [(O ) 3 P ] +, and [(O ) 2 P ( O) 2 ] 0 respectively c. The paramagnetic signal of POHC is ubiquitous in P 2 O 5 -containing glasses. In the simplest model of this defect, the P atom is bonded to three bridging O atoms and to a fourth nonbridging O which hosts the unpaired electron: [(O ) 3 P O ] +. However, this structure (here referred to as l-pohc) has been argued to be stable only at low temperature, while the room-temperature stable form of POHC (here referred to as r-pohc) was proposed to feature an electron shared by two non-bridging oxygen atoms bonded to the same phosphorus [(O ) 2 P ( O) 2] 0 [40]. l-pohc and r-pohc supposedly feature two slightly different EPR signals. Figure 3.11 shows the EPR signature of the POHC center: arrows indicates the metastable PHOC form. Figure 3.11 also shows in the central part of the EPR spectrum a signal attributed by Griscom et al. [40] to the so called Si(E )(P), a species of Si(E ) center with phosphorus next-nearest-neighbors. After clarifying by EPR the microscopic structure of these defects, data obtained by optical absorption (OA) studies of irradiated P-doped silica were interpreted by proposing associations between some of the observed OA bands (Figure 3.12) and the paramagnetic centers [40, 38]. In particular Griscom proposed the attributions presented in Figure 3.13 for the OA bands of Figure 3.12, observed in the range 3 6 ev. According to these assignments, r-pohc centers absorb at 2.2, 2.5 and 5.3 ev, while l-pohc have an absorption band at about 3.1 ev. In contrast, much less is known about diamagnetic P-related centers in SiO 2. Based on the results obtained by several independent experimental techniques, including Raman and c In P4, the P atom hosts an additional lone pair, not represented.

43 3.2. Extrinsic point-defects 35 Figure 3.11: EPR signature detected at low temperature (77 K) of the stable form of POHC defect. Additional peaks indicated by the arrows are supposed to be due to the metastable POHC variant. In the central part of the spectrum the lineshape related to Si(E )(P) is also visible (from ref. [40]). Figure 3.12: Radiation induced optical absorption spectrum in a P-doped silica sample (from ref. [40]). Infrared measurements, phosphosilicate glass is generally believed to consist of an intermixed random network of [(O-) 2 Si(-O) 2 ] 0 and [(O-) 3 P=O] 0 tetrahedra randomly bonded by sharing O atoms, this being consistent with the fact that [(O-) 3 P=O] 0 is the basic building block of pure stochiometric (P 2 O 5 ) phosphate glass [113, 114, 115, 116]. In this model each P atom is bonded to 3 bridging O atoms and a single doubly-bond non-bridging O, and thus each site can be argued to be a potential precursor for l-pohc via ionization of the non-bridging oxygen. In contrast, r-pohc should be formed by ionization of a defective site [(O-) 2 P(=O) 2 ] where the P atom bonds two bridging oxygen atoms with single bonds and two more non-bridging oxygen with double bonds. Finally, P4, P1, P2 centers are supposedly formed by irradiation via hole or electron trapping on hypothetical diamagnetic precursor defects where P is 2-, 3-, and 4-fold

44 36 3. Point defects in optical fibers Figure 3.13: Attribution of the main P-related absorption bands showed in Figure 3.12 in irradiated phosphosilicate glasses. Also the average peak energy (E), the full with at half maximum (W ) and the oscillator strength (f) are reported. From [40]. coordinated respectively (Figure 3.10): d [(O-) 2 P:], [(O-) 3 P:] 0, [(O-) 2 P(-O) 2 ] + [40,111,112,38]. In the intermixed random network model, also these sites should be considered as randomly occurring point defects. d In the 2-fold coordinated precursor of P4 center, the P atom hosts an additional lone pair, not represented.

45 Part II Materials and methods

46

47 Chapter 4 The canonical samples This chapter is focused on the description of the particular specimens used to perform the experiments discussed in the rest of the work. An important part of the interest in our approach is based on the choice in the design of our samples that hereafter we will call canonical samples. These samples have to be representative of commercial fibers that have already been tested and will be used in future facilities [117]. They have also to offer an easier interpretation of their responses thanks to their custom designs. Previously, two different studies used a set of homemade samples to understand the influence of several process and composition parameters on the radiation response of single-mode germanosilicate optical fibers at 1.3 and 1.55 µm [118, 22]. Due to the good knowledge of their sample characteristics, E.J. Friebele et al. [118] were able to obtain statistically significant correlations between the γ-ray steady-state Radiation Induced Attenuation (RIA) and some of the fabrication parameters. As an example, they found that for doses of rad at -35 C, the RIA level at the end of the irradiation is correlated with the Ge content in the fiber core (for a Ge/F doped cladding). The second study was devoted to the transient X-ray radiation response of Ge-doped fibers and showed that, on lowering the standard preform deposition temperature from 2000 to 1600 C and the drawing tension from 140 to 20 g, the induced losses slightly decrease wavelengths [119] and the influence of various cladding co-dopants. However, these two studies were limited by the difficulty to obtain the samples with the characteristics needed to get unambiguous correlations. For the present work, we design the structures to overcome these difficulties. First, a quantitative analysis of the influence of a dopant can only be achieved if all the differently-doped glasses have been made with strictly identical processes. From a practical point of view, due to the non-negligible influence of MCVD process parameters [119], this can not be achieved by investigating several preforms and fibers. It must be done within a single sample. Secondly, new spectroscopic techniques are now accessible, thus allowing to spatially resolve the radiation-induced changes in the fiber with micrometer resolution. For example, we show the efficiency of the confocal microscopy of luminescence (CML) to characterize the

48 40 4. The canonical samples radiation-induced point defects in passive [100] or rare-earth doped [120] optical fibers. These spatially-resolved techniques enable the characterization of new fiber designs (such as the samples studied in this Thesis) that would have not been possible in the past. 4.1 Tested optical preforms and fibers The purpose of this Thesis is to study the role of the dopants in the properties and in the radiation response of the multi-mode optical fibers. To this aim the experiments were carried out on three types of silica optical fiber and preform samples, doped with germanium, fluorine and phosphorus respectively. These samples are prototype not commercialized yet: their future use can be planned for specific fields, like space or nuclear power plants, where expositions to high irradiation doses are expected. All preform and associated fiber samples were made through the Modified Chemical Vapor Deposition process (see section 1.1.4) by ixfiber SAS [121]. About 50 mm of each prototype preform has been kept for analysis whereas the other part of the preform has been drawn to obtain several hundreds meters for each fiber. The refractive index profiles for the three types of samples are presented in Figure 4.1. Standard conditions of fiber manufacturing (preform deposition and drawing process) have been used for these waveguides. To quantitatively investigate the influence of the dopants concentration, the structure of each preform, and then of each fiber, has been designed with several steps of concentration of one doping element, Ge or F or P, in the core. This particular sample structure was thought and realized ad hoc with the purpose of studying the dopant influence, thus maintaining the other fabrication parameters as fixed. The fiber and preform main characteristics are listed in Table 4.1. Using spatially-resolved techniques [100, 120, 122], we will then be able to study the fiber or preform properties and the radiation response to the dopant concentration for samples with strictly identical MCVD process parameters. Dramatic importance in the sample fabrication had the choice of the dopant concentration levels. Part of the dopant concentration values have been chosen to reproduce the classical range of concentrations measured on commercial fibers (e.g.,from 2 to 12 wt.% for the Gedoped fibers). The other ones have been defined in relation with our ab initio calculations conducted on a 108 atoms silica-based supercell [123]. At the fabrication stage, the obtained multi-step radial distribution of the dopant along the fiber diameter can be roughly estimated through measurements of the fiber and preform refractive- index profiles (Figure 4.1). A more accurate estimation of the concentration values of the dopants is obtained by electron microprobe analysis (EMPA) which also allows to inspect the impurities content inevitably present in the samples. Fiber and preforms are made up of four cylindrical layers (core part, zones 1-4) of high pure synthetic silica differently doped,

49 4.1. Tested optical preforms and fibers 41 Table 4.1: Parameters related to the canonical fiber and preform samples. n refers to the refractive index change at λ=633 nm with respect to a SiO2. Dopants and impurities average concentrations were evaluated by electron microprobe analysis. (a) Ge canonical samples P canonical samples Zone <Ge> <Cl> n Pref. <P> <Cl> n Pref. Fiber (wt%) (wt%) ( 10 3 ) diam. (wt%) (wt%) ( 10 3 ) diam. diam. (mm) mm (µm) Cladding Core (b) F canonical samples Zone <F> <Cl> n Pref. Fiber (wt%) (wt%) ( 10 3 ) diam. diam. (mm) (µm) Coating Cladding Core

50 42 4. The canonical samples G e n (x ) (a ) F n (x ) (b ) P n (x ) (c ) R a d ia l d is ta n c e (m m ) Figure 4.1: refractive index change ( n) measured at λ=633 nm with respect to a SiO 2 in: (a) germanium, (b) fluorine and (c) phosphorus canonical preform samples. following a multiple step distribution. The layers were deposited in a tube of undoped fused silica which forms the cladding (zone 0). The preform samples have a diameter of about 10 mm, with a 5 mm inner doped region, and they were cut and polished into plates of approximately 1.5 mm thickness. The fiber/preform length ratio is about and the fiber core diameter is 62.5 µm.

51 4.1. Tested optical preforms and fibers 43 Ge-doped canonical samples Ge-doping increases the refractive index of a SiO 2, so germanium doping profile grows from the boundaries to the center. The samples have been doped with several amounts of germanium Figure 4.2: Microscopic preform view obtained with an optical microscope. Numbers from 0 to 4 refer to the zones listed in Table 4.1 with different amounts of germanium. from 2 wt.% in the exterior part (zone 1) and up to 11 wt.% in the inner-center part (zone 4), as shown by the microscopic vision in Figure 4.2. Doped parts of this fiber contain typical levels of chlorine impurity ( 1200 part per million (ppm)) and OH-groups ( 60 part per billion (ppb)). Figure 4.3 shows the Ge and Cl trend inside the fiber and preform samples. The average Ge and Cl concentrations measured by EMPA for each zone are given in Table 4.1. The drawing speed was 40 m/min, the drawing tension 70 g and the temperature of the furnace 1600 C. These fibers exhibit pre-irradiation optical characteristics at 1.55 µm close to that of commercial fibers, that is 0.34 db/km. F-doped canonical samples In the fluorine doped samples the core consists in three zones (zones 2 to 4) with three different F-concentrations (see Figure 4.4). The F-incorporation inside a SiO 2 decreases its refractive index, so F-doping profile has an opposite trend with respect to Ge as shown in Figures 4.1 and 4.5. Optical cladding (zone 1) corresponds to the zone with the highest F-doping region ( 1.8 wt%). More details are presented in Table 4.1. The outer cladding (zone 0) is made of pure-silica. Only a little part of the signal is guided in this part of the waveguide. As a consequence, its contribution to the global transmission can be considered as negligible. Doped parts of this fiber contain very small amounts of chlorine impurity ( 1000 ppm) and OH-groups ( 200 ppb) typical of MCVD glasses. In this fiber the

52 44 4. The canonical samples 1 2 D is ta n c e fro m p re fo rm c e n te r (m m ) G e O 2 (% w t) C l (% w t) D is ta n c e fro m fib e r c e n te r (µm ) 0.0 Figure 4.3: Impurities trend inside canonical Ge-doped fibers (lower x scale) and preforms (upper x scale). Empty squares represent the Ge content, full circles the Cl content. The estimation of the concentration values of the elements is obtained by EMPA analysis. Numbers from 0 to 4 refer to the zones listed in Table Figure 4.4: Microscopic preform view of the F-doped canonical sample. Numbers from 0 to 4 refer to the zones listed in Table 4.1. The different sample zones are clearly visible. attenuation at 1.55 µm is about 1.9 db/km. P-doped canonical samples The P canonical samples are made up of an outer undoped high purity silica layer, and four internal cylindrical layers (core part, zones 1-4) of highly pure synthetic silica doped

53 4.1. Tested optical preforms and fibers 45 D is ta n c e fro m p re fo rm c e n te r (m m ) F C l F -c o n te n t (w t% ) C l-c o n te n t (w t% ) D is ta n c e fro m fib e r c e n te r (µm ) 0.0 Figure 4.5: Fluorine (empty squares) and chlorine content (full circles) in F-doped fibers (lower x scale) and preforms (upper x scale). Impurities concentrations were evaluated by EMPA analysis. with different P-amounts. Phosphorus doping profile grows from the boundaries to the center (a ) µm 0 Figure 4.6: Enlarged view of the P-doped fiber canonical sample. Numbers from 0 to 4 refer to the various sample zones listed in Table 4.1 with different P-amounts. The various zones are clearly visible. (Figure 4.1) following the desired multiple step distribution, as shown in the microscopic image of the fiber sample in Figure 4.7. The core-cladding part does not contain a relevant concentration of extrinsic impurities, except for chlorine that is present with a maximum concentration of 0.2 wt%. The preforms had an initial diameter of mm with a 5 mm doped region and they were subsequently

54 46 4. The canonical samples D is ta n c e fro m p re fo rm c e n te r (m m ) P -c o n te n t (w t% ) C l P C l-c o n te n t (w t% ) D is ta n c e fro m fib e r c e n te r (µm ) Figure 4.7: P-content (empty squares) and Cl-content (full circles) obtained by electron microprobe analysis at various distances from the fiber (lower x scale) and preform (upper x scale) center. cut into mm 3 samples and polished. Phosphorus and chlorine concentrations were checked by EMPA, giving the results shown in Table 4.1 and in Figure 4.7. The spectral attenuation at 1.5 µm, obtained with the cutback method, is about 0.5 db/km for this fiber. The OH content is evaluated as sensibly inferior to 10 ppb.

55 Chapter 5 Experimental set-ups This chapter is concerned with the description of the instruments and setups used to perform the experiments discussed in the rest of the work. Several irradiation sources were used to analyze the generation processes of defects both on fibers and on preforms. Additionally, various spectroscopic techniques have been employed to identify the precursor sites in non-irradiated samples or stable point defects in irradiated samples. Some of these techniques have been applied both on fibers and on preform samples allowing the study of the influence of the drawing effects on defects generation. Other techniques could be applied only on preform (time resolved luminescence, absorption) or on fiber samples (radiation induced attenuation), for the particular sample structure. 5.1 Irradiations UV laser irradiations UV exposures at 5 ev (248 nm) were carried out at room temperature with two distinct setups:a pulsed KrF laser and a continuum (CW) Ar-laser and. UV pulsed laser Irradiation exposures with the high power KrF pulsed laser were performed at a repetition rate of 10 Hz, a duration time of 30 ns for every pulse and a pulse energy varying from 100 to 400 mj. Irradiations on fibers were conducted by moving the samples at a constant speed, transversally to the UV laser. The energy of the laser pulses is measured with a pyroelectric detector; the accuracy, taking into consideration the laser fluctuations, is ±10%. Continuum UV laser The setup used for the continuum UV exposures was developed at the Hubert Curien Laboratory. Irradiations were performed using an UV Argon Laser, emitting at 244 nm (5.1 ev)

56 48 5. Experimental set-ups and having a gaussian intensity profile. For preform samples the beam was unfocused on the center of the preform thanks to a spherical lens. The laser energy was measured by a power meter with an accuracy of 5%. For the fiber samples the irradiation system is more complicated [117,124] and it is described in Figure 5.1. The fiber samples are mechanically uncoated Figure 5.1: Schematic representation of the experimental setup used for the CW ultraviolet (244 nm, 5.1 ev) exposures of optical fibers. to prevent the UV light absorption by their acrylic coatings. A variable fiber length (from few millimeters to several meters) filed past the CW laser beam. To this purpose, the fiber was maintained in a V-groove and was interdependent with a tended thread by a counterweight. This thread was pulled by a rotating motor. The 244 nm light was focused by a spherical lens, leading to a spot size of some mm diameter on the fiber. By an appropriate choice of the fiber translation speed ( cm/s) and of the laser power (5 100 mw), it was possible to vary the fluence value. The uncertainties in the fluence evaluation were mainly attributed to the mechanical part of this setup and they were estimated within ±20% γ-ray and X-10 kev irradiations The samples were exposed to γ-rays produced by a 60 Co source available in the IGS-3 irradiator of the Department of Nuclear Engineering of University of Palermo. γ-rays have energies between 1.17 MeV and 1.33 MeV. Irradiation was performed at room temperature, in ordinary atmosphere at a dose rate of 1.39 kgy/h. The 10-keV X-ray irradiations of fiber and preform samples were performed at room temperature using an ARACOR Semiconductor X-ray irradiator [125,126] at the French atomic energy center (CEA). The irradiated zone is homogeneous over a diameter size of 2.5 cm; the dose was varied from 50 Gy up to 2 MGy (two different dose rates: 10 Gy/s and 0.1 kgy/s) in fibers and from 1 kgy up to 2 MGy (dose rate = 0.1 kgy/s) in preform samples.

57 5.2. Absorption Absorption How shown in the introduction section, point defects in a SiO 2 can introduce new electronic levels inside the valence and the conduction band. So, using electromagnetic radiation at energy lower than the gap, it is possible to induce transitions corresponding to the absorption bands of these materials. The optical absorption experiments were conducted sending electromagnetic radiation on the sample, varying continuously the light energy during a fixed interval and than analyzing the transmitted light. According to the Lambert-Beer law, the transmitted radiation intensity I T is linked to the incident intensity I 0 by the relation 5.1 [127]: I T ( ω) = I 0 ( ω) exp{ α( ω) d} (5.1) where d is the sample thickness and α( ω) is the optical absorption coefficient, measured in cm 1. Knowing the absorbing species concentration, N, the absorption coefficient can be expressed by the following equation: α( ω) = Nσ( ω) (5.2) where σ( ω) is the absorption cross section [50]. Optical absorption spectra presented in the result section were performed with two different spectrometers: the JASCO V-560 and the AVANTES S2000 spectrophotometers. The JASCO spectrophotometer The JASCO V-560 spectrometer is a double beam spectrophotometer providing the measured absorption values in optical density (O.D.), defined as: O.D. = log 10 ( I0 (λ) I(λ) ) (5.3) The incident beam I 0 is split thanks to a beam splitter and it is alternatively sent to the sample and to the detector. In this way we can have the control, and than the correction, of every source fluctuation. The source is made of two lamps: a deuterium lamp, working in the ultraviolet (UV) region from 340 nm to 190 nm, and a Xenon lamp, operating in the visible and infrared range ( nm). The detector is a photomultiplier (PMT). The AVANTES spectrophotometer In the optical fiber AVANTES S2000 spectrophotometer the source is a deuterium lamp (D 2 ) that injects light into an optical fiber that splits up in two channels, referred to as Master

58 50 5. Experimental set-ups and Slave. The optical fibers are multimode pure silica core\f 2 -doped silica cladding with diameter of 200 µm. They are loaded with H 2 to better resist to the prolonged exposure to UV light without being deteriorated. The light carried by the master channel gets out of the fiber and is used as the probe beam (PB). The PB is collimated by two lens and it is coupled to another fiber that brings it to the detector. The two lenses are mounted on independent micrometric positioning controls (xyz), which permit both to control the alignment of the PB to the sample and to optimize the collection efficiency after the sample. The slave channel passes through a variable attenuator, after which it goes to the detector. Since the slave channel does not traverse the sample, it could be used to correct experimental data for the temporal drift of the lamp. The detector consists in a 1200 lines/mm grating with blaze at 300 nm, dispersing on a 2048 channels Charge Coupled Device (CCD) array. The instrument works in the 200 nm 500 nm range with a spectral resolution of 5 nm. Before acquiring a spectrum it is necessary to obtain a dark reference signal D(λ) from the master channel when the D 2 lamp is disconnected from the fibers. If I 0 (λ) and I(λ) are the signals acquired respectively without and with the sample, the absorption profile of the specimen is given by: ( ) I0 (λ) D(λ) O.D. = log 10 (5.4) I(λ) D(λ) Absorption in the VUV spectral range VUV absorption spectra in the ev range were obtained using an ACTON SP-150 single-beam spectrophotometer, equipped with a 30 W D 2 lamp and two 1200 lines/mm monochromators, and working in N 2 flux. The acquired spectra were corrected by subtracting the contribution due to surface reflectance. 5.3 Photoluminescence and Raman spectroscopy Photoluminescence An optical property of great importance in the study of point defects in a SiO 2 is the photoluminescence (PL), i.e. the process by which a system excited by light with wavelength λ ex emits light at λ em > λ ex while decaying back to its ground state [50, 128, 129]. To picture the physical processes determining the PL emission band, let us consider the two-level system of Figure 5.2. Due to the absorption process OA, a number N(λ ex ) of defects will be in the excited state during exposure of the sample to the excitation light. Some of these defects can decay radiatively with a rate k r, thus originating the photon emission or luminescence. The remaining excited defects relaxes back to the ground state by a temperature dependent nonradiative process, with a rate k nr, in which the energy is dissipated by emission of phonons.

59 5.3. Photoluminescence and Raman spectroscopy 51 ZPL OA k R k NR O Figure 5.2: Electronic-vibrational level scheme of a two levels point defect. The continuous arrow oriented upward represents the absorption transitions OA from the ground state to the exited state. The continuous arrow oriented downward represents spontaneous emission from the excited state to the ground state. The dotted arrow represents the non-radiative decay process. The grey arrow indicates the ZPL transition and the dashed arrow within the excited levels represents the internal relaxation process. In Figure 5.2 the transition at lowest energy (that is from (0, 0) to (1, 0)) is called the zero phonon line (ZPL). The PL intensity is given by: I P L (λ ex, λ em ) = k r N(λ ex )L P L (λ em ) (5.5) where L P L (λ em ) is the emission lineshape determined by homogeneous and inhomogeneous contributions [51,50,15]. The variation rate of the excited state population, N, depends on the absorption and decay processes, both radiative and non radiative, according to the following equation: dn(λ ex, T ) dt = I 0 (λ ex ) [1 exp { α(λ ex )d}] (k r + k nr (T ))N(λ ex, T ) (5.6) Steady state luminescence experiments are performed when the system undergoes a continuous excitation. In this case dn/dt = 0 and combining Equations 5.6 with 5.5 we get: where η = kr k r+k nr photons and absorbed photons. I P L (λ ex, λ em, T ) = ηi 0 (λ ex ) [1 exp { α(λ ex )d}] L P L (λ em ) (5.7) is the luminescence quantum yield, defined as the ratio between emitted Two basic types of measurements are possible: measuring the intensity I P L (λ ex, λ em ) as a function of λ em for fixed λ ex we obtain the shape and intensity of the band emitted by the center, that is the emission spectrum (PL);

60 52 5. Experimental set-ups acquiring I P L (λ ex, λ em ) as a function of λ ex for fixed λ em, we have the excitation spectrum (PLE), which represents a measurement of the efficiency of the emission process in dependence of the excitation wavelength. Differently from stationary PL measurements, in the time-resolved PL measurement the time decay of the emitted light after an exciting light pulse is studied. After the excitation of a point defect with a light pulse, that produces a population of N(0) of the excited state, the light source is switched off (I 0 = 0) and N(t) decays as: N(t) = N(0)e t/τ (5.8) with τ = 1/(k r + knr). From equations 5.5 and 5.8 we obtain the luminescence timedecay: I P L (λ ex, λ em, T, t) = k r I P L (λ ex, λ em )N(0)e t/τ (5.9) At sufficiently low temperatures, the non-radiative decay channels are usually quenched, i.e. k nr k r, so that the measurement directly yields the radiative decay time τ = 1/k r. The importance of the knowledge of τ relies in the possibility of calculating the oscillator strength f of the center [50]: f = 1 m 2 c 3 (5.10) E 2 τ 2e 2 Not all the point defects that feature a measurable absorption band decay by emitting luminescence, but when this occurs, their study by PL spectroscopy has some important advantages with respect to OA. In particular, PL is more selective, as it often allows to isolate a center whose absorption band overlaps to those arising from other defects, based on the different emission properties Stationary and time resolved luminescence setup In this section we are going to describe the experimental setup used for acquiring the luminescence data of preform samples. As shown in the schematic representation of Figure 5.3 the equipment is mainly constituted by a laser source, a sample chamber, a dispersion system and a detection one. The tunable laser (VIBRANT OPOTEK) [130] is an integrated system that emits pulses of 5 ns duration with a maximum repetition rate of 10 Hz in the range ( ) nm. The excitation beam with a spot size of 1mm 2 hits the sample mounted in the so-called 45 back-scattering geometry. The emitted light is collected by a lens and then arrives at the detection system. The energy of the laser pulses is measured with a pyroelectric detector capable of giving in output a short electric pulse for each laser shot (some mus), whose amplitude is read by a digital meter. It is positioned before the sample holder

61 5.3. Photoluminescence and Raman spectroscopy 53 Xe-lamp Tunable Laser OPOTEK Sample Spectrograph CCD+ Delay generator Figure 5.3: Schematic representation of the equipment used for the luminescence measurements. to measure the intensity of incident laser radiation. The accuracy of pulse energy measures, taking into consideration the laser fluctuations, is 10%. Time resolved luminescence spectra are performed with a detection system consisting of a Spectrograph and an Intensified CCD Camera. This latter amplifies the input luminescence signal: for each photon that strikes the photocathode surface many photons are produced. Moreover, the possibility of varying the photocathode voltage allows to enable or to disable the CCD: in the GATE ON mode the photocathode voltage is -200 V and the CCD sees the light, in the GATE OFF mode the photocathode voltage is 0 and the CCD does not see the light. This peculiarity permits the detection of time resolved luminescence spectra synchronized with the laser excitation pulses. In fact, the CCD is triggered by an electronic synchronization signal produced by the laser 60 ns before the pulse. The CCD can accumulate 4 ns (0.1-1)s τ Δt I PL I PL λ em λ em Figure 5.4: Diagram of the CCD timing. in a time window defined by the width parameter t (Gate Width) and by its delay τ (Gate Delay) from the origin of the time scale. So, as shown in the diagram of Figure 5.4, the Gate

62 54 5. Experimental set-ups Width determines the amplitude of the time window during which the CCD is enabled to reveal the luminescence light (GATE ON mode); while the Gate Delay regulates the temporal shift of the acquisition window with respect to the trigger signal. In the next chapters all luminescence spectra are presented as a function of the energy E instead of the wavelength λ, so they require a specific correction procedure: the CCD counts are directly proportional to the luminescence spectral density di/dλ that is the intensity collected with a constant spectral bandwidth dλ. Since energy and wave number are linked by the relationship E = hc/λ, the spectral density di/de with respect to E must be multiplied for spectral dispersion λ 2 because to a constant spectral bandwidth in λ corresponds a spectral bandwidth which depends from the emission energy E Photoluminescence under synchrotron radiation excitation Excitation (PLE) end emission spectra excited in the Vacuum UV (VUV) range on preform samples were carried out under excitation by pulsed synchrotron radiation at the SUPER- LUMI station on the I-beamline of HASYLAB, DESY (Hamburg) [131]. Measurements were performed in the spectral range ev, with a pulse width of 130 ps, an interpulse of 500 ns, and a spectral width of 0.3 nm. The excitation beam is directed into the primary monochromator with a 1200 lines/mm grating (Figure 5.5; the excitation wavelength can be Figure 5.5: Schematic representation of the experimental station used for PL, time resolved PL and PLE under synchrotron radiation (figure adapted from [131]). varied from 310 nm to 50 nm (4.0 to 24.8 ev). The emitted light was spectrally dispersed by a 300 grooves/mm grating blazed at 300 nm and acquired by a liquid nitrogen cooled CCD camera (1100 Princeton Instruments) for PL spectra. Luminescence spectra were corrected both for the spectral response and dispersion of the detecting system, while excitation spectra were corrected for the spectral efficiency of the

63 5.3. Photoluminescence and Raman spectroscopy 55 exciting light, using a sodium salicylate sample as a reference. 0.3 nm, while emission bandwidth was 20 nm. Excitation bandwidth was During laser-excited luminescence and measurements, we put the sample behind a properly built mask so as to allow spatial selection of the various preform zones (see Section 4.1). In time resolved PL and in PL under synchrotron radiation, temperature dependencies (from 10 to 300 K) were investigated using continuous flow helium cryostats Raman measurements The Raman effect is the anelastic scattering of light (by a molecule or a point defect) due to emission or absorption of a vibrational quantum [132, 127, 133]. If E i is the energy of the incident photons, scattering at E s < E i implies the excitation of a vibrational mode of energy ω = E i E s. A Raman spectrum consists of a plot of the scattered intensity as a function of ω. This spectroscopy allows to probe the vibrational modes of a molecule or a point defect, sometimes bearing some advantages with respect to common IR spectroscopy. For example, it can happen that a vibrational mode is Raman-active but non IR-active, or vice versa [132, 127]. Raman spectra presented in this PhD Thesis were all performed by the microraman spectrometer described here below Confocal Micro-spectroscopy setup Confocal microscopy luminescence (CML) measurements and microraman measurements were performed with the LabRAM Aramis (Jobin-Yvon) integrated confocal microraman system whose scheme is presented in Figure 5.6. The confocal microscope is coupled to a 460 mm focal length spectrograph equipped with a four interchangeable gratings turret. The different excitation wavelengths are supplied by up to one internal laser (He-Ne, 633 nm, 2.0 ev) and by two external lasers: a He-Cd laser working at 442 nm (2.8 ev) and 325 nm (3.8 ev) and an Argon laser working at 488 nm (2.5 ev). Briefly the principle of confocal microscopy consists of focusing the laser source through the objective of the microscope and carrying out a spatial filtering of the signal coming from the illuminated volume, by using a diaphragm of small diameter placed in the conjugated plane where the magnified image of the sample is formed by the objective (see Figure 5.7). On the incoming path, the laser beam is reflected towards the microscope by the means of a special filter (holographic notch filter or dielectric edge filter) used in injection/rejection mode. On the return path to the spectrograph, the Raman backscattered light is fully transmitted

64 56 5. Experimental set-ups CCD detector Grating turret Laser HeNe External Lasers Filter turret Autofocus Camera White light source Figure 5.6: Top view of the the LabRAM Aramis (Jobin-Yvon) integrated confocal microraman system with the optical path. Figure 5.7: Scheme of micro-luminescence and micro-raman analysis set-up. through the filter towards the confocal slit-hole located at the entrance of the spectrograph. The spectrograph disperses the multichromatic Raman signal onto the CCD multichannel detector. The optical drawer constitutes the coupling platform between the laser, the sampling chamber (microscope or macro chamber) and the spectrograph. In order to reduce the laser power at the sample, a density filters wheel driven by the software, can be used. During our measurements we make sure that the power level of the probe light was reduced to few

65 5.4. Electron Paramagnetic Resonance measurements 57 hundreds of µw to avoid photobleaching effects. For Raman measurements, the Raman signal is then collected by the same microscope objective (backscattered configuration) and follows the return path to the spectrograph. The raman filter filters out the backscattered laser light (Rayleigh scattering) whereas the Raman signal is transmitted to the spectrograph entrance slit. The sample can be translated, under computer control, with an accuracy of about 0.1 µm. The excitation beam penetrates of a few micrometers into the sample. The spot diameter focalized on the samples varies as a function of the microscope objective used. Using a 50 objective, the spot size is some µm. With this spot size it is possible to inspect in detail not only the preform samples, but also the fibers. This is a new tool of our study: previously the largest part of scientific investigation on defects in optical fibers consisted in the direct study of bulk samples and the subsequent transfer of information to the fibers [19,20,21]. However, this approach cannot take into account the peculiarities implied in the fiber preparation procedure, such as the drawing process, which can generate precursors and influence the defect generation [22, 134, 135]. Micro-Raman and micro-luminescence investigations, allowing inspection of defect directly inside optical fibers, permit us of overcoming this limit. 5.4 Electron Paramagnetic Resonance measurements The Electron Paramagnetic Resonance (EPR), also referred to as Electron Spin Resonance (ESR) spectroscopy measures the absorption of microwave radiation corresponding to the energy splitting of an unpaired electron when it is placed in a strong magnetic field. EPR is a spectroscopic technique that detects the presence of these unpaired electrons in a chemical system [136, 137, 138]. This can yield meaningful structural and dynamic information, even from ongoing chemical or physical processes (i.e. kinetics, etc.) without influencing the process itself. In the absence of an external magnetic field the two possible electron spin states (spin up and spin down) are degenerate. When an atom or molecule with an unpaired electron is placed in a magnetic field, the spin of the unpaired electron can become aligned either in the same direction (spin up) or in the opposite direction (spin down) of the applied field. These two possible electron alignments have different energies (i.e. are no longer degenerate) and are directly proportional to the applied magnetic field strength. This is called the Zeeman effect. The EPR measurements reported in this Thesis were performed by a Bruker EMX spectrometer working at ω 0 =9.8 GHz. In Figure 5.8 it is reported a simplified scheme of the instrument. The microwave radiation travels down a waveguide to the sample, which is held in place in a microwave cavity held between the poles of two magnets. A variable attenuator permits to regulate the actual power P i incident to the cavity from a maximum value 200 mw down to 200 nw. In this way, P i is usually chosen so as to avoid the saturation of the observed

66 58 5. Experimental set-ups Source Detector Attenuator A/D A/D Amplifier Lock-in RC Filter Magnet Modulator Integrator Resonant cavity EPR Signal Figure 5.8: Schematic representation of the Bruker EMX spectrometer. The main sections are visible: magnet, cavity, source, attenuator, modulation system and detector. magnetic resonance transition. The microwaves arriving on the entrance of the cavity are partially absorbed by the sample and partially reflected to join the revelation system. The reflected power P R is measured by a detector that gives a current signal I proportional to the square root of P R. Spectra are obtained by measuring the absorption of the microwave radiation while scanning the magnetic-field strength. EPR spectra are usually displayed in derivative form to improve the signal-to-noise ratio. It is important to underline that the doubly-integrated intensity of the EPR spectrum is proportional to the number N of paramagnetic centers. In this sense, if a reference sample is available, EPR may be used to provide a measurement of the absolute concentration ρ = N/V of every paramagnetic defect, where V is the volume of the sample. To this purpose, in this work it was used a specimen where the absolute number of E centers (purposely generated by γ irradiation) was known by spin-echo measurements [139, 140]. The accuracy of the absolute concentration measurements obtained by ESR, based on comparison with the spinecho reference sample, is estimated as 20% a. a this error never explicitly appears when reporting in the following the uncertainties on the concentration measurements. The reason for this choice is that the uncertainty on the concentration of E in the spinecho reference sample plays the role of a systematic error, affecting in the same way all the concentration measurements here reported.

67 Part III Ge-doped fibers and preforms

68

69 Chapter 6 Measurements on non-irradiated samples As discussed in detail in the introductory chapters (section 2.1), identification of defects responsible for the permanent change of the refractive index in Ge-doped glasses under radiation exposure is a clue to make clear the microscopic origin of transparency loss and photosensitivity. Consequently, the study and the characterization of point defects in Ge-doped a SiO 2 is a very effective subject to understand the fiber properties and to improve their performances. We have seen that germanium lone pair centers are the main responsible defects for photosensitivity property. The GLPC optical activity variation under UV-radiation exposure contributes to the permanent refractive index variation of the glass. So the exact knowledge of the GLPCs concentration before radiation exposure and the modifications induced by the drawing process, result to be a fundamental aspect in optical fibers technology. The study of GLPCs variations in optical fibers can so be adopted as probe for testing the refractive index modulation induced by external factors. To this aim, this section is devoted to introduce the optical properties of the as-grown materials performing direct investigation of GLPCs inside our germanosilicate step-index canonical fibers and preforms. First of all we performed absorption measurements on preforms samples before irradiation in order to investigate GLPC presence in the different sample zones. Figure 6.1 shows the OA spectra detected by the use of the AVANTES spectrometer in the spectral range ev; in the figure the variation of the OA intensity as a function of the different sample zones, that is as a function of the Ge-content, is also shown. For GeO 2 >5% the OA signal saturates, indicating defects concentration higher than our detection system ( 70 cm 1 ). As expected, the OA band centered at 5.2 ev and related to GLPCs, grows with the Ge-content. Though the determination of the GLPC spatial distribution in optical fibers is crucial to

70 62 6. Measurements on non-irradiated samples 9 0 ) -1 [G e O ] (w t% ) 2 A b s o rp tio n C o e ffic ie n t (c m G L P C E n e rg y (e V ) Figure 6.1: OA spectra detected in pristine Ge-doped canonical sample. The arrow specifies the GeO 2 content from the exterior to the inner sample part. The OA band at 5.2 ev is due to GLPC centers whose microscopic structure is showed in the inset. For [GeO 2 ]>5 wt% the OA signal saturates. probe the silica refractive index variation, convenient experiments have not been performed yet, the main obstacle being the small fiber dimensions. So far, studies dealing with GLPCs were extracted from bulk samples and than transferred to fibers [19, 20, 21], thus remaining affected by an intrinsic deficiency caused by the information lack related to the drawing process. Nevertheless the use of Confocal Microscopy Luminescence (see section 5.3.5) allows to examine the variation of the PL bands linked to GLPCs even on micrometric scale. Than we performed direct investigation of GLPCs inside germanosilicate step-index optical fibers by using CML technique with the purpose of examining their radial distribution along the core. The comparison with the corresponding preforms permits to recognize the actual role of the drawing process in modifying the defect formation, and consequently to control the fiber photosensitivity. The recognition of GLPC concentration in preform samples is not possible via OA B 2β band due to our measurement conditions: in the central sample region the absorption coefficient is too high and cause the saturation of the detected signal (see Figure 6.1), thus preventing a careful evaluation of defects concentration. There exists opinion that there could be two very closely spaced optical bands at about 5.2 ev due to Ge-related oxygen-deficient centers, and that only one of them could give rise for

71 63 PL [55]. Otherwise, using the linear correlation between B 2β band and the GLPC PL activity, we assumed that the GLPC absorption band was unique. This assumption could systematically affect the absolute evaluation of GLPC concentration, nonetheless, for our purposes, the relative concentration variation is still meaningful. In principle this problem could be overcome estimating concentration by the singlet-triplet GLPC absorption band at 3.8 ev. Nevertheless, this OA band is much weaker than the singlet-singlet one (about 10 4 times) and it is not detectable in our samples. GLPCs in fibers and preforms were revealed by the He-Cd ion laser excitation line of the LabRam Aramis spectrometer (photon energy 3.8 ev, power 0.5 mw). We used a 40 objective and a diaphragm diameter of 100 µm. The excitation beam penetrates of a few µm into the sample as to be sure that the collected signal was not due to abnormal surface defect concentration detection. GLPCs were checked monitoring the intensity variation of the 3.1 ev phosphorescence band by direct excitation S 0 T 1 (E ex = 3.8 ev) (see GLPCs levels scheme in the left inset of Figure 6.2). As an example, in Figure 6.2 OA at 5.2 ev and PL at 3.1 ev (right inset) detected in zone 2 of preform sample ([GeO 2 ] 5 wt%) are depicted. The use of a laser probe 5 0 ) -1 A b s o rp tio n C o e ffic ie n t (c m S 1 O A 5.2 e V S 0 E E X 3.8 e V P L T e V P L (a rb. u n its ) E n e rg y (e V ) E n e rg y (e V ) Figure 6.2: OA spectrum detected in zone 2 ([GeO 2 ] 5 wt%) of Ge-doped pristine preform. The right inset shows the PL spectrum al 3.1 ev obtained exciting at 3.8 ev in the same preform zone. In the left inset transitions giving rise to the observed bands are schematically depicted. at 3.8 ev has a double advantage:

72 64 6. Measurements on non-irradiated samples First of all we have seen that the germanosilicate exposition to UV laser induces the bleaching of the optical activity related to GLPCs [21] (see section 3.2.1). Consequently exciting the GLPC activity by the use of a laser beam at the energy of 5.2 ev, that is by the excitation S 0 S 1, could produce an intensity diminution of the 3.1 and 4.2 ev PL bands not related to an actual defect concentration decrease in pristine samples, but rather to an undesirable sample irradiation. The result will be an invasive measurement that alters the evaluation of defects concentration. in contrast, as shown in Figure 6.1, our samples are almost transparent in the excitation spectral range from 2.5 to 4 ev, making sure that the use of a laser probe at 3.8 ev results in a non invasive analysis. In the second place, it is worth to note that in our samples the absorption at 5 ev is very high: comparing Figure 6.1 we can see that at 5.2 ev the absorption coefficient exceeds 70 cm 1 in the central core region. As a consequence, the luminescence signal obtained exciting in the peak of the OA band, will not be proportional to GLPC concentration, due to the fact that the sample will be not uniformly illuminated during PL measurements. GLPC-related emissions at 3.1 ev detected in the fiber layers with different Ge-content are shown in Figure 6.3. CML spectra were recorded from 370 nm (3.4 ev) to 500 nm (2.5 ev) Figure 6.3: 3D plot of the CML band related to the T 1 S 0 GLPC transition, measured at different distances from the fiber core center and obtained under excitation at 3.8 ev in the fiber sample. The PL intensity decreases with the distance from the center, i.e. with the GeO 2 content.

73 6.1. Discussion: the drawing effect 65 with a spatial resolution of about 1 µm. Qualitatively, we observe that the PL intensity decreases with the distance from the center, i.e. with the Ge content. 6.1 Discussion: the drawing effect To perform a quantitative calculation of defects concentration, [GLPC], we convert the PL intensity PL(GLPC) to an absolute concentration measurement thanks to the evaluation of the oscillator strength, f, for the GLPC centers. Starting from Smakula s equation [50], the 5.2 ev absorption intensity in the preform is determined by its linear correlation with the PL bands and its oscillator strength, estimated to be f = 0.07 [21] from the radiative decay time of the 4.2 ev (τ 7.8 ns) transition, measured at T=10 K under synchrotron radiation [141]: [GLP C] = P L(GLP C) [cm 3 ] (6.1) The calculation giving rise to Equation 6.1 was performed for preform samples and than transferred to fibers a. GLPC-related emissions at 3.1 ev detected in the fiber layers with different Ge-content are shown in Figure 6.4. To draw the GLPC radial distribution in the fiber sample, the side (a) of Figure 6.5 shows their concentration measured at various distances from the fiber center. For sake of clarity in the lower part of Figure 6.5 (side (b)) the microscopic fiber view is sketched, so as to localize the various zones where these defects are detected. To point out the influence of drawing effects on the defect generation, Figure 6.5 shows the comparison of [GLPC] as a function of [Ge], as measured in preform and in fiber. In the preform such a dependence has an almost linear trend over the whole range with a coefficient β pref In contrast, in the fiber is observed a discontinuity at [Ge] cm 3 (8 wt%): over the lower range the experimental points overlap with those of the preform, whereas at larger Ge concentrations the ratio [GLPC]/[Ge]=β fiber exceeds β pref up to about five times (β fiber ) for [Ge] cm 3 (11 wt%). The interpretation of this result could be related to the differences between radial stress of the fiber and of the preform. Both materials are characterized by a thermoelastic stress that increases linearly with the Ge concentration [142, 143, 144]: as this kind of stress is linked to the temperature expansion coefficient, it results in a positive tensile stress [143]. This means that highly doped core regions, as those investigated in our experiment, are more likely to exhibit tensile core stresses. On the other side, the drawing process can introduce an additional negative compressive stress in the fiber samples [145]. The interplay between these two contributions influences therefore the net fiber stress, which could result to be either a Our experimental set-up does not allow to perform spatial resolved OA measurements on fibers.

74 66 6. Measurements on non-irradiated samples (a ) -3 [G L P C ] ( c m ) (b ) Figure 6.4: (a): GLPC concentration obtained by a transversal mapping of the fiber sample at various distances from the core center. (b): microscopic view of the fiber cross section. The different layers are distinguishable. The x scales in the upper and lower figures coincide. negative or positive in core zones with a different Ge content [145, 143], differently from the preform where the radial stress is always positive. We suggest that a sudden change of the stress in the core inner part causes the observed discontinuity of GLPC concentration, due to a new defect generation mechanism. This result is relevant for its connection with the refractive index change in the core zones. As the drawing process dramatically modifies the distribution of UV optically defects inside the fiber, the photosensitivity properties cannot be deduced simply by the fiber composition or by measurements on preform samples. From a handy point of view, it seems possible to

75 6.1. Discussion: the drawing effect P re fo rm F ib e r ) -3 [G L P C ] ( c m [G e O ] (w t% ) 2 Figure 6.5: GLPC concentration trend detected in the different sample layers as a function of Ge content both in fiber (empty squares) and preform (full circles). Figure 6.6: Left: Assial stress as a function of GeO 2 content in a graded index Ge-doped fiber. Right: Stress profile of a graded index Ge-doped fiber. Figures from Lee et al. [142]. modulate the fiber photosensitivity during the fabrication process by a systematic control of the fiber drawing conditions. In conclusion, our experimental results point out that the CML is a powerful and not invasive tool to probe the defect spatial distribution directly in optical fibers. We have demon-

76 68 6. Measurements on non-irradiated samples strated that the GLPC radial distribution is different in fibers and preforms. This difference can change the photosensitivity of the fiber regarding to the preform sensitivity. A possible explanation for this effect is the changes in the mechanical stress introduced by the drawing process. Such a stress can lead to an enhancement of the defect generation in some specific part of the fiber. Our study shows that the drawing process has to be considered to explain the defect generation mechanisms at the origin of the glass photosensitivity. More detailed analysis with fibers designed with variable process parameters and stress measurements have to be done to control this phenomena in order to enhance the photosensitivity of germanosilicate fibers or to reduce their degradation under radiative environments.

77 Chapter 7 Effects of the UV and X-ray irradiation As pointed out in section the identifications of centers causing the degradation of the fiber properties under radiation exposure is one of the main subject in the current study of optical fibers response. Many studies, both experimental and computational, have investigated the generation and conversion processes of Ge-related point defects under radiation exposure. The research interest is mainly motivated by the practical importance of these materials in photonic applications, from standard optical fibers as waveguides for telecommunications to non-linear optical devices. This multi-use of Ge-doped silica is strongly influenced by its response to radiation; for instance, it is known that radiation exposure induces Ge-related defects that are cause of attenuation for the waveguides, thus leading to detrimental losses of part of the transmitted signals. The identification of the specific defects, sensitive to radiation, is therefore of crucial importance to make clear the microscopic origin of technologically relevant processes. Nerveless the current understanding of many important aspects is not at all complete, and several relevant questions are still debated. In the following we present the results of a series of experiments on the effects of the radiation exposure on canonical Ge-doped fibers and preforms [126, 123, 146, 122] to get a deeper clarification of the microscopic origin of the photoinduced structural changes observed in these materials. 7.1 EPR results One of the main techniques employed to investigate the effects of irradiation on Ge-doped canonical samples is the electron paramagnetic resonance (see section 5.4). After X-ray exposure we observe the presence of paramagnetic centers, as an example Figure 7.1 shows the EPR spectra detected in fiber and preform samples after 20 kgy X-ray

78 70 7. Effects of the UV and X-ray irradiation irradiation. g = X d o s e 2 0 k G y P re fo rm F ib e r g = g = M a g n e tic F ie ld (m T ) Figure 7.1: EPR spectra in fiber (open circles) and preform (full line) Ge-doped samples after a X-dose of 20 kgy. We can note that in both samples the irradiation induces the same defect species but with a different efficiency: defect concentration in fiber is higher than in preform. It is also apparent that these EPR spectra are composite signals, resulting from the overlap of different contributions. As shown in Figure 7.1 we can distinguish the g values of three germanium related defects: 2.007, 1.999, To obtain information on the concentration of every defects specie contributing to the total EPR lineshape, we performed a deconvolution of the measured spectra. We found that the EPR spectra can be least-square-fitted, after any irradiation, by a linear combination of the lineshapes associated with the following Ge-related paramagnetic point defects, already described in section 3.2.1: (i) Ge(1) (GeO 4), (ii) GeE ( Ge ) and (iii) Ge(2), whose structure is still questioned between models consisting either in a trapped electron center or in a hole center (see section for details). As an example, Figure 7.2 shows the fitting result on fiber (panel (a)) and preform (panel(b)), after a X-ray dose of 20 kgy. Red solid line plots the best fitting function. The normalized EPR lineshapes of these three defects, used in the best fitting procedure, are shown separately in panel (c) of the Figure 7.2. Ge(1) and GeE lineshapes are experimental curves measured in γ-irradiated Ge-doped samples [147], where it has been possible to single out the two defects. The Ge(2)

79 7.1. EPR results (a ) (c ) 1 0 G e (1 ) E P R In te n s ity (a rb. u n its ) (b ) G e E 0 G e (2 ) M a g n e tic F ie ld (m T ) Figure 7.2: EPR signal detected in (a): fiber and (b): preform samples after 20 kgy X-ray exposure. Solid red lines plot the best fitting functions obtained as a superposition of the three lineshapes related to GECs and GeE centers. Panel (c) reports the three base lineshapes used for the best fitting procedure. lineshape is a curve taken from literature, obtained by Friebele et al. [101] by a simulation procedure aimed to fit EPR lineshapes observed in γ-irradiated multimode Ge-doped fibers. We point out that fitting an EPR signal with a linear combination of single-defect lineshapes is founded on the assumption that the centers are sufficiently far from each other that their lineshape is not influenced by mutual interactions. This condition is usually satisfied at defects concentrations up to cm 3. Figure 7.3 shows the changes in EPR spectra after different X doses in fibers and in preforms. Qualitatively we can deduce that the EPR signal intensity increases with increasing the X-ray dose both in fibers and in preforms, nevertheless the lineshape does not significantly change. We show that after UV pulsed laser, continuum UV laser irradiations or γ exposure the EPR spectra show the appearance of the same defect species with similar trend, even if with different intensity. As an example in Figure 7.4 the comparison between a spectrum obtained in preform sample after a X-dose of 2 MGy and the signal detected in gamma irradiated preform sample (total dose=91 kgy) is reported. In the rest of this section, only results

80 72 7. Effects of the UV and X-ray irradiation X d o s e (k G y ) (a ) P re fo rm X d o s e (k G y ) (b ) F ib e r M a g n e tic F ie ld (m T ) Figure 7.3: EPR signal intensity as a function of the total deposited X-dose in (a): preform and (b): fiber canonical Ge-doped samples. obtained after X-ray irradiations will be discussed. From the best fit coefficients appearing in the linear combination of the GEC defects (Figure 7.2(c)), we are able to evaluate the contribution of the three defect species to the overall signal measured upon exposure at each total deposited X-dose. We point out that EPR measurements give results on the whole sample, regardless the different sample layers with various Ge amount. The contribution of each paramagnetic specie, obtained with the above described procedure as a function of the X-ray dose is reported in Figure 7.5. It is evident that the GECs grow with the same trend on increasing the X-dose, Ge(2) being the defect induced in largest concentration over the considered dose range. Both in fiber and in preform Ge(1) and Ge(2) grow with the X irradiation dose till a saturation value, in contrast GeE centers do not reach a saturation value, at least at the considered doses. 7.2 Optical absorption Due to the reduced fiber dimensions, absorption measurements in the UV-visible region were performed only on preform samples. Results will be than extended to fibers. In our specific case, this experimental procedure is supported by results showed in section 7.3.3, in which the effects of the drawing process are illustrated: the results clearly show that only at low

81 7.2. Optical absorption 73 E P R In te n s ity (A rb. U n its ) X irra d ia te d (2 M G y ) γ irra d ia te d (9 1 k G y ) M a g n e tic fie ld (m T ) Figure 7.4: EPR spectra detected in preform samples after 10 kev-x-ray exposure (total deposited dose 2 MGy, full line) and after γ exposure (total dose=91 kgy, open circles). irradiation dose the drawing procedure can influence defect generation. The transmission losses from 2.5 ev to 5.5 ev of the preform sample after an X dose of 2 MGy, detected with the AVANTES spectrometer, are shown in the panel (a) of Figure 7.6. As the beam spot size diameter is smaller than the different zones dimensions, we are able to selectively investigate the distinct sample layers. The spatial resolution of this measurement is obviously limited by the diameter of the beam spot, that is about 0.5 mm. Through this procedure we can easily obtain the OA spectra as a function of the average GeO 2 content, as shown in Figure 7.6(a). The OA spectra referred to the different sample zones evidence the wing of an induced band centered at energies higher than 4.5 ev, more and more intense with increasing the Ge-content. We notice that the detection of the whole band in the heavy doped zones is limited by the presence of the OA band centered at 5.1 ev and related to the GLPCs, whose amplitude is higher than our detection limit ( 50 cm 1 ). To single out the different OA components, the spectra have been fitted by Gaussian curves. Panel (b) of Figure 7.6 shows the analysis of the spectrum detected in the preform zone containing 5 wt.% of GeO 2 ; it is accounted for by three bands: the first centered at (4.6±0.1) ev, FWHM of (1.8±0.2) ev, the second centered at (5.14±0.02) ev, FWHM of

82 74 7. Effects of the UV and X-ray irradiation (a ) E P R In te n s ity (a rb. u n its ) (b ) G e (1 ) G e E G e (2 ) X d o s e (k G y ) Figure 7.5: Integrated EPR intensity related to GECs and GeE centers as a function of the X-ray dose detected in (a): fiber and (b): preform samples. (0.50±0.02) ev and the last centered outside the investigated range ( 6 ev) to take into account higher energies contributions. In regard to the absorption tail centered at 4.5 ev, Figure 7.7 shows its dependence from the X total deposited dose. The graph illustrated the OA increasing, from 2.7 to 4.8 ev, in the central part of the preform sample, containing the higher GeO 2 -content (zone 4 in Figure 4.2 and in Table 4.1, GeO 2 11%). Similar trends are observed in the other sample zones. From the Figure 7.7 we can qualitatively conclude that, with respect to pristine sample, the radiation exposure causes an increasing in transmission losses at about 4.5 ev. 7.3 Discussion: radiation effects The use of several spectroscopic techniques allows from one side to identify defect species causing radiation losses in the considered spectral range and on the other side to localize

83 7.3. Discussion: radiation effects 75 ) -1 A b s. C o e ff. (c m (a ) [G e O ] (w t% ) (b ) [G e O 2 ] = 5 w t% E n e rg y (e V ) Figure 7.6: (a): Transmission losses detected in the various preform zones after 2 MGy X-ray exposure. The arrow specifies the GeO 2 content from the exterior to the inner sample part. (b): Absorption spectrum detected in the zone 2 (5 wt.% of GeO 2 ) superimposed to the best fit curve, dashed lines represent the Gaussian components. them inside the fiber sample in the considered spectral range. This result is a very original and interesting feature in the study of the optimization of fiber manufacture and it will be discussed in the following two points: localization of defect species generation of GECs defects Localization of defect species In the previous section we have shown the EPR results on preform and fiber Ge-doped canonical samples after radiation exposure. The results allow to identify various defect species present in the samples, nevertheless this measurement procedure does not permit to localize the defects in the various sample layers. We have seen in Section 4, that our specific canonical samples were made with concentric layers containing different amount of germanium. We can reasonably expect that the defect distribution along the sample layers is not uniform: Ge-related species will be concentrated in the central sample zone with the grater Ge-content (<[GeO 2 ]> 11%). In the case of absorption measurements, this assertion is immediately confirmed by the

84 76 7. Effects of the UV and X-ray irradiation Z o n e 4 [G e O 2 ] ~ 1 1 w t% ) -1 A b s. C o e ff. (c m X -d o s e (k G y ) E n e rg y (e V ) Figure 7.7: Transmission losses from 2.7 to 4.8 ev, detected in the central preform zone (zone 4, [GeO 2 ] 11%) as a function of the X-ray exposure. The arrow specifies the X total deposited dose. radiation induced losses trend illustrated in Figure 7.6. While the absorption band centered at 5.1 ev is undoubtedly attributed to GLPC centers, defects causing losses in the region ev are not unambiguously identified. Neustruev [19] and Fujimaki et al. [106] tentatively attributed the absorption at 4.5 ev to Ge(1) centers. As shown in section 7.1, our samples reveal the presence of the Ge(1) EPR signature after radiation exposure (see Figure 7.2). Plotting the absorption variation at 4 ev detected in the central sample layer with respect the Ge(1) EPR intensity, we obtain the graph illustrates in Figure 7.8 (lower x-scale). It is apparent that Ge(1) EPR signal is proportional to OA absorption around 4 ev. The linear correlation is confirmed by the best fit function represented as a full line in Figure 7.8. This confirms that Ge(1) is the main responsible defect for the transmission losses at 4 ev. So, while EPR measurements do not allow to spatially resolve defect concentration, by relating Ge(1) OA band to their EPR signal, we can locate Ge(1) presence in the central part of the sample (the more doped part). Ge(1) concentration can be calculated considering the central part of the sample (x-upper scale in Figure 7.8). It is also possible to measure the Ge(1) cross section, σ, at 4 ev, defined as [50]: σ 4eV = α 4eV /N (7.1) where α is the variation of the absorption coefficient compared to pristine sample and N

85 7.3. Discussion: radiation effects [G e (1 )] ( c m ) ) -1 α 4 e V (c m G e (1 ) E P R In te n s ity (A rb. U n its ) Figure 7.8: Correlation between the EPR intensity of the Ge(1) signal (lower x scale) or the Ge(1) concentration (upper x scale) and the OA absorption variation at 4 ev. Full line represents the linear best fit curve. represents the defect concentration. To get the σ 4eV value for Ge(1) centers, we can evaluate the slope of the correlation curve of Figure 7.8, between Ge(1) concentration and the OA absorption variation at 4 ev. By a linear fit of the type: [Ge(1)] = A + σ α 4eV (7.2) we find that the cross section is σ 4eV cm 2. Hence it so possible to demonstrate that Ge(1) centers are responsible for the observed transmission losses and that they are mainly localized in the central part of the samples where the larger germanium content favors the electron trapping on GeO 4 sites Generation processes of GECs defects In section we have seen that Ge(1) defects are concentrated in central sample zone, with the higher Ge-content. Hence, starting from the global EPR measurements, we are able to evaluate their effective concentration in the different preform zones. The result is illustrated in Figure 7.9. Comparing Figures 7.5 and 7.9, we can observe that the Ge(1) concentration,

86 Effects of the UV and X-ray irradiation F ib e r ) 1 0 D e fe c ts c o n c e n tra tio n ( s p in /c m P re fo rm G e (1 ) G e E G e (2 ) X d o s e (k G y ) Figure 7.9: Concentration of the GECs centers as a function of the X-ray dose detected in (a): fiber and (b): preform samples. Defect concentrations are evaluated as concentrated in the central sample zones (<[GeO 2 ]> 10%). both in fibers than in preforms, is larger than the value expected considering only the EPR intensity trend. In particular, after an X-ray exposure of 2 MGy the Ge(1) concentration is about cm 3 in the inner core fiber part, that is about two order of magnitude larger than the value obtained considering the defects as equally distributed on the whole sample. For Ge(2) the defect concentration in fibers at the saturation value is cm 3. For the considered total deposited doses, the GeE centers do not reach a saturation value: after a total X deposited dose of 2 MGy, [GeE ] cm 3, both in fibers and in preforms. We note that the Ge(1) localization and the calculation of their concentration can be indirectly applied to fiber samples. This outcome is crucial to find out the radiation effect in modulating the optical properties of step index fibers, both the light transmission and the refractive index change. Regarding GECs generation mechanism, many authors have supposed the following reac-

87 Discussion: radiation effects 79 tion as the basic creation process of Ge(1) and Ge(2) centers in Ge-doped silica after UV or gamma exposure: = Ge + GeO 4 + hν (= Ge ) + + (GeO 4 ) (7.3) A germanium lone pair center (= Ge ) and a 4-coordinate germanium (GeO 4 ) generate, after radiation exposure, an ionized GLPC ((= Ge ) + ), that is a Ge(1) center, and a hole center on a 4-coordinate germanium ((GeO 4 ) ). Evidently Equation 7.3, implies that the structure for Ge(2) center is a hole center and not a trapped electron center. This also means that, to confirm the generation mechanism proposed in Equation 7.3, a direct proportionality between Ge(1) and Ge(2) concentration has to exist. According to our measurements, the dependence of Ge(1) concentration from Ge(2), both evaluated from EPR measurements, has the trend shown in Figure We point out that, F ib e r P re fo rm ) [G e (2 )]( s p in /c m [G e (1 )]( s p in /c m 3 ) Figure 7.10: Correlation between Ge(1) and Ge(2) center in fiber (open circles) and preform (full squares) Ge-doped canonical sample after irradiation exposure. to obtain the graph in Figure 7.10, we have considered also the Ge(2) centers as concentrated in the central sample zone, even if rigorously we have demonstrated this assertion only for Ge(1) centers. We justify this hypothesis on the basis of Equation 7.3: if Ge(1) and Ge(2) generation mechanisms coincide and if we find that the Ge(1) centers are only present in the central part of our canonical samples, this must be true also for Ge(2).

88 80 7. Effects of the UV and X-ray irradiation The drawing effect Up to now, only few studies investigated the influence of the drawing process on the fiber radiation response by comparing the behaviors of fiber and preform. Most of them have characterized the effect of varying the drawing tension, speed or temperature on the radiation response of optical fibers [119, 118, 148]. In our study, all our canonical samples have been made with standard MCVD drawing conditions. By comparing the concentrations of roomtemperature stable paramagnetic defects at the same doses in a preform and its corresponding fiber, we can estimate the global influence of the drawing process on the glass sensitivity. The obtained results are illustrated in Figure 7.11 for the Ge canonical samples. 1 0 G e (1 ) G e E G e (2 ) [F ib e r]/[p re fo rm ] 1 n e g a tiv e in flu e n c e o f d ra w in g p ro c e s s p o s itiv e in flu e n c e X -D o s e (k G y ) Figure 7.11: Defect concentration ratio on fibers and preforms in Ge-doped samples at different doses. Our results clearly show that the influence of the drawing process on the generation of Gerelated centers is dose-dependent. The dose-dependence is observed separately for the three paramagnetic defects: Ge(1), Ge(2) and GeE. The impact of the drawing process is strongly negative at low doses and it becomes nearly negligible at higher doses: at 2 MGy this ratio is about 1 and it becomes 10 for a dose of 2 kgy. These results show that at lower doses (space, military applications) the drawing process governs the generation of paramagnetic defects in Ge-doped glass whereas the glass composition seems to be the most influential parameter for high-dose applications (nuclear power plants, high energy physics). A possible explanation for this dose dependence is that drawing strongly increases the

89 7.3. Discussion: radiation effects 81 number of defect precursors such as GeODC(II) and the so called neutral oxygen vacancy (NOV) ( Ge T ; T =Si or Ge) [19, 106] which can turn into Ge(1), Ge(2) or GeE under irradiation following the process described by Equation 7.3 and by the following relation: Ge T +hν = Ge + + T + e (7.4) where Ge T is the NOV, T is either Ge or Si, and = Ge is the GeE center. At low doses, the contribution of defects generated from these precursor sites to the total concentration of defects is predominant whereas it may become less important at higher doses due to defect generation via other mechanisms. As above shown, our CML measurements on the Ge-doped fiber with UV excitation provided evidence for the presence of GeODC(II) in both pristine (see Section 6) and irradiated samples. By the way, Ge(1) and Ge(2) defects may be created from the preexisting and X-ray radiation-induced GeODC(II). We have previously shown that the Ge(1) and Ge(2) concentrations saturate at higher doses (>20 kgy) whereas GeE concentrations continuously increase up to 2 MGy (Figure 7.9) providing evidence for several generation mechanisms for this defect. Additional tests have to be done to fully understand the drawing influence on the fiber radiation response. Complementary EPR measurements have to be done at lower doses to investigate this effect for low-dose environments. Furthermore, different fiber samples have to be drawn from the same Ge-doped preform to determine the most favorable drawing conditions for this kind of optical fiber.

90 82 7. Effects of the UV and X-ray irradiation

91 Part IV Influence of further dopants: fluorine and phosphorus

92

93 Chapter 8 F-doped fibers and preforms In this chapter we will analyze the fluorine role in changing the structure and the radiation effects in the optical fibers. As shown in Section 2.2, there are several plausible reasons for justifying the addition of fluorine to the a SiO 2 matrix: F substitutes to OH groups or Cl, always present in optical fiber structure, without inducing any optical absorption in the transparent region of a silica glass. Si F bond is stronger than Si O, resulting in widening the optical band gap. Formation of Si F bonds makes the glass network to be more stable, i.e., it decreases structural disorder, which shift the vacuum-ultraviolet silica fundamental absorption edge (Urbach edge) to shorter wavelengths [30]. F favours structural relaxations, and makes it easy to fabricate silica glasses with lower fictive temperatures (T f ), which is defined as "the temperature at which the glass would find itself in equilibrium if suddenly brought to that temperature from its given state" [149]. The first part of the Chapter is devoted to the materials characterization by a micro Raman analysis. Successively the fluorine ability of improving the material radiation hardness is investigated. Specifically, after fiber and preform exposition to ionizing radiations of different nature, the samples are spectroscopically analyzed by electron paramagnetic resonance (EPR) to reveal the possible presence of point defects thus evaluating the fiber radiation resistance.

94 86 8. F-doped fibers and preforms 8.1 Raman results It is known that fluorine is incorporated into the silica glass structure as SiF with the silicon atom bonded to three oxygen atoms in the network [150]. Raman spectroscopy is an useful tool to determine the presence of SiF linkages in silica glass, since they give rise to a spectroscopic line around 945 cm 1. In the following we take advantage of the spatial resolution of the confocal micro-spectrometer elsewhere described (Section 5.3.5) to check F-presence in our samples. In Figure 8.1 we show the Raman spectrum of the zone 1 (see Table 4.1(b)), obtained in the spectral region ( ) cm 1. It is clearly visible the peak centered at 945 cm 1 R a m a n In te n s ity (a rb. u n its ) S i-f p e a k R a m a n S h ift (c m -1 ) Figure 8.1: Preform cladding (zone 1) Raman spectrum in the spectral range ( ) cm 1. The peak at 950 cm 1 is related to Si F stretching mode. corresponding to the Si F stretching vibration mode of a SiO 3 F tetrahedron [151]. The variation of this F-related line in the other zones is plotted in Figure 8.2, all spectra being normalized to the intensity of the peak around 800 cm 1 which is ascribed to the stretching mode of Si O bond [152]. Table 8.1 gives the nominal average F-concentrations for each zone, as evaluated by electron microprobe analysis, and the ratio of the Raman 945 cm 1 band and the 800 cm 1 reference band, for the various layers. As known from literature [27], for low F content (inferior to 2 wt%) only Si F links are present; this agrees with our results that point out a correlation between F content and the Si F Raman signal.

95 8.1. Raman results 87 R a m a n In te n s ity (a rb. u n its ) D is ta n c e fro m c e n te r m m (z o n e 0 ) m m (z o n e 1 ) 1.6 m m (z o n e 2 ) 1 m m (z o n e 3 ) m m (z o n e 4 ) R a m a n S h ift (c m -1 ) Figure 8.2: Raman spectra detected in the different sample zones listed in Table 4.1. The variation of the Si F band (950 cm 1 ) at different distances from the preform center is clearly visible. For sake of comparison, spectra are shifted and they are normalized to the 800 cm 1 band. All the other peaks are intrinsic features associated with the glassy silica matrix [153]. The peaks at 495 cm 1 and 606 cm 1 in Figures 8.1 and 8.2, superposed to the larger band at 440 cm 1, are the so-called defect bands D1 and D2. The D2 and the D1 bands are attributed to the three- and the four-membered ring structures, respectively [154]. In these small ring structures the Si O Si bond angle (130.5 for the three-membered ring and for the four-membered ring) [25] is significantly different from the stablest angle (150 ) [154,155]. As a consequence, these ring structures are composed of heavily strained S O Si bonds. According to estimates by Galeener [154], the atomic fraction of these rings is of the order of 1%. It has been demonstrated that strained Si O Si bonds, which are created on densification of SiO 2 glass by high pressure [156], cause the absorption edge shift to a longer wavelength [25]. The densified SiO 2 glasses are extremely sensitive to defect formation by irradiation. The formation of strained three- and four-membered ring structures has been found to be suppressed effectively by doping of a small amount of F because fluorine is incorporated in the form of Si F bonds that terminate in a continuous silica network structure [25]. This is evident from the Raman spectra depicted in Figure 8.3, where a comparison (without vertical shift) is reported between the F-free zone (zone 0 in Figure 4.4) and the 11 wt% F-doped zone

96 88 8. F-doped fibers and preforms Table 8.1: Fluorine average concentration and ratio of the Raman bands at 945 cm 1 and at 800 cm 1 band for the different preform zones. N.D. stands for not detected. Zone <F> (wt%) I(945 cm 1 I(800 cm 1 Coating N.D. Cladding ± ±0.02 Core ± ±0.01 R a m a n In te n s ity (a rb. u n its ) D 1 D is ta n c e fro m c e n te r: m m (z o n e 1 ; [F ]= 1.8 w t% ) m m (z o n e 0 ; [F ]= 0 w t% ) D R a m a n S h ift (c m -1 ) Figure 8.3: Raman spectra of the F-free and the F-doped (11 wt%) zones of the pristine canonical preform sample. D1 and D2 defect bands are highlighted. (zone 4 in Figure 4.4). The main difference between the two spectra, apart from the presence of the 945 cm 1 Si F band, is the intensity reduction of the D1 and D2 bands. The role of F is similar to that of OH groups in this respect. However, laser damage is drastically enhanced for OH doping because OH groups have strong optical absorption at a wavelength of 157 nm. The breaking of the continuous silica network structure by F doping largely facilitates structural relaxation in the process of cooling from the melt: F doping may be regarded as chemical annealing [25].

97 8.2. EPR measurements on irradiated samples EPR measurements on irradiated samples As pointed out in Section 2.2, recent studies have shown that radiation toughness of silica samples is achieved by incorporating Si F groups whose positive effect is assumed to be the reduction of defect precursors [150,27,24], such as strained bonds (Si O Si) from which is likely generated the pair of silicon dangling ( Si ) and oxygen dangling ( Si O ), where ( ) and ( ) indicate bonds with three oxygen atoms and an unpaired electron, respectively. These two paramagnetic defects, that as shown in the introduction section are also named E center and non bridging oxygen hole center (NBOHC) (see Section 2.2), are indeed one of the main causes of transparency loss due to their absorption bands peaked in visible and UV spectral range [157]. To this aim the role of fluorine doping in the response to UV pulsed laser and γ radiation of silica preforms and fibers was studied using EPR Results Figure 8.4 shows the EPR spectra of preform samples after exposure to UV- ( J/cm 2 ; panel (a)) and γ-rays (91 kgy; panel (b)); we observe that neither preforms nor fibers show EPR signals before exposure. Regardless the kind of irradiation, two identical signals are observed in our samples. The central part of the spectra shows the typical line shape of the E centers, whose concentration is measured to be (5.5±0.8) cm 3 in the UV irradiated sample, and (1.8±0.4) cm 3 in the γ irradiated one. In the spectra is also observed a structured signal extended over 10 mt, which is ascribed to the [AlO 4 ] 0 center, whose structure is shown in Figure 8.5. This defect, consisting in a substitutional Al atom [159,158], has been extensively studied in many experimental [160, 161] and theoretical [162, 104] studies. The calculated concentrations of these defects are (3.9±0.7) cm 3 and (6.9±0.4) cm 3 after UV- and γ irradiation, respectively. The presence of [AlO 4 ] 0 defect indicates the existence of extrinsic impurities in preform samples before exposure. Though the EPR spectra do not allow to spatially resolve the distribution of paramagnetic defects, as the sample core-cladding part is made of high pure F-doped silica, we can suppose that [AlO 4 ] 0 center are concentrated in the external layer of non doped silica, containing 10 ppm of Al impurities. Similar irradiation treatments were performed on fiber samples; in this case, the E centers could be observed only by second harmonic mode, whereas detection of [AlO 4 ] 0 EPR signal is prevented by our detection limit. Figure 8.6 shows the E -related second-harmonic signal, both in UV (Figure 8.6(a)) and γ (Figure 8.6(b)) irradiated samples. UV irradiations were conducted at different amounts of energy fluence, from 10 J/cm 2 to 132 J/cm 2. As clearly shown in Figure 8.7, E concentration grows with the UV fluence up to 55 J/cm 2, after that

98 90 8. F-doped fibers and preforms (a ) E S R s ig n a l (a rb. u n its ) (b ) M a g n e tic F ie ld (m T ) Figure 8.4: EPR first harmonic spectra in UV ( J/cm 2 ) (a) and γ (91 kgy) (b) irradiated preforms. The central enlarged zone shows the typical signal of E centers. Parts (a) and (b) have the same scale. Figure 8.5: Microscopic structure of the [AlO 4 ] 0 center. Figure adapted from ref. [158]. it saturates at a value of cm 3. In γ irradiated fiber, after a dose of 91 kgy, E concentration becomes (2.3±0.2) cm 3.

99 8.2. EPR measurements on irradiated samples (a ) U V F lu e n c e (J /c m 2 ) E S R s ig n a l (a rb. u n its ) (b ) M a g n e tic F ie ld (m T ) Figure 8.6: EPR second harmonic spectra in fiber samples after irradiation. (a): variation of the signal due to E centers with the UV fluence on fibers.(b): E signal in fibers after a γ dose of 91 kgy Discussion: generation of E centers The above reported results point out two aspects, common both to preform and fibers and to UV- and γ-irradiation: 1. The generation of E centers is due to the conversion of pre-existing precursors such as oxygen vacancy or extrinsic Si-H or Si-F bonds. This is consistent with the growth curve observed under UV irradiation, that manifests a saturating tendency due to the exhaustion of precursors; hence, the larger E concentration induced by γ radiation suggests the presence of different kind of precursor activated by one or the other radiation source. On the other side, the absence of NBOHC, ensured by EPR spectra within a limit of cm 3, and by luminescence measurements on preform within cm 3 [157], indicates that the radiolysis of strained Si-O bonds is an unlikely process in our F-doped silica samples.

100 92 8. F-doped fibers and preforms 9-3 [E ] ( c m ) F lu e n c e (J /c m 2 ) Figure 8.7: E concentration as a function of the energy fluence on UV irradiated fibers. 2. The drawing process does not weaken significantly the radiation toughness of the fiber samples. In fact, after γ-irradiation the ratio between the number of E defects in fiber (Figure 8.4(b)) and in preform (Figure 8.6(b)), both irradiated with 91 kgy, is [E ] fiber /[E ] pref =1.2. An analogous comparison can be made for UV irradiated samples. Considering the saturation value of E concentration on UV irradiated fibers, we can evaluate the ratio between this concentration and that found in preforms after an UV fluence of J/cm 2 : [E ] fiber_satur /[E ] pref =1.5. Hence, we can conclude that about the same amount of E centers are generated in fiber and preforms both under UV and γ radiation. These points qualitatively evidence the role of F-doping in governing the response to radiation of preform and fiber. In fact, irradiation of undoped silica, bulk [24] or fiber [163,164], produces the generation of the pair NBOHC and E center by the rupture of strained Si- O bonds, whereas the formation of NBOHC is inhibited in F-doped samples. This finding suggests that Si-F groups reduce the presence of strained bonds, this role upholds both in bulk and fiber, thus being crucial to improve the quality of optical fibers designed for visible and UV transmission.

101 Chapter 9 P-doped fibers and preforms As shown in the introduction sections (see Sections 2.3 and 3.2.2), despite the importance of P-doping in fiber technology, its real influence in defect generation, and consequently in light propagation attenuation, is still unknown. Not only the consequence of the radiation exposure in terms of attenuation, but also the structure of P-related defects in pristine P-doped a SiO 2 is still subject of study and discussion. In this chapter we report experimental studies on phosphorous-related point defects in amorphous silica, based on photoluminescence, absorption, and electron spin resonance measurements carried out on P-doped canonical fibers and preforms (see Section 4.1). 9.1 Optical activity of P-related point defects From the experimental point of view, the optical properties of diamagnetic P-related centers in silica are scarcely known at the moment, since only very little data exist on optical absorption and luminescence of as-grown phosphosilicate glasses, expect for the basic evidence that no strong UV absorption bands are generally induced in these materials just by P- doping [38, 165]. The purpose of this section is to contribute to a better understanding of these topics, by reporting data obtained by absorption and photoluminescence measurements in the UV and vacuum UV spectral ranges on P-doped optical fiber preforms Absorption and photoluminescence analysis Figure 9.1 shows the OA spectrum of the central zone of the canonical sample (zone 4, maximum P-content 7 wt%, see Figure 4.6) in the UV and in the VUV spectral range, detected

102 94 9. P-doped fibers and preforms with the JASCO V-560 and the ACTON SP-150 spectrometers respectively ) -1 α (c m E n e rg y (e V ) Figure 9.1: Absorption spectrum of the P-doped preform in the UV-VUV spectral range ( ev). The inset shows the enlargement of the OA spectrum from 3 to 6 ev. It is evident the presence of a strong absorption band in the VUV range centered at about 6.9 ev, an absorption tail for E>7eV, and a shoulder in the UV region, due to a band centered at about 4.8 ev (inset of Figure 9.1). By performing a PL emission measurement at room temperature under synchrotron excitation at the energy corresponding to the UV band (4.8 ev), we detected a broad asymmetric luminescence signal centered at 3.0 ev, reported in Figure 9.2(a). The PLE spectrum of this signal, measured with emission at E em =3 ev, is reported in Figure 9.2(b) and features two components: the first one is centered at 4.7 ev (with FWHM 0.7 ev) and the second one at 6.4 ev (with FWHM 0.6 ev). The shape of the emission band turns out to be approximately the same when excited at 4.8 ev (full squares) or 6.4 ev (open circles) by synchrotron radiation, or upon laser excitation at 4.8 ev (solid line), as shown by the comparison of the three signals in Figure 9.2(a). We studied the dependence of the 3.0 ev luminescence signal intensity on P concentration, by moving the laser excitation spot across the different preform zones via a micropositioning stage. The spatial resolution of this measurement is limited by the diameter of the laser spot, partialized by an iris, that is about 0.5 mm. As shown in Figure 9.3(a), we found

103 9.1. Optical activity of P-related point defects 95 P L In te n s ity (a rb. u n its ) 4 5 (a ) (b ) E E X (e V ) 4.8 (la s e r e x c.) E n e rg y (e V ) E n e rg y (e V ) P L E In te n s ity (a rb. u n its ) Figure 9.2: (a): Emission spectra measured in the central sample zone (zone 4), containing 7 wt% of phosphorus at 300 K, obtained exciting at 4.8 ev (full squares) and 6.4 ev (open circles) under synchrotron radiation and at 4.8 ev under laser excitation with W T =20 ms and T D =5 ns (solid line). (b): Excitation spectrum monitored at E em =3 ev detected at room temperature under synchrotron radiation in the range ev. that the 3.0 ev PL signal is observed only in the P-doped region of the sample, and rapidly disappears when moving away from the center of the perform. By comparing with the spatial dependence of P concentration, we see that the luminescent region is somewhat narrower than the doped region. Indeed, the 3.0 ev PL is mainly localized on zones III and IV of the preform. Figure 9.3(b) shows the PL intensity on the peak of the 3.0 ev band as a function of P-content. It appears that the luminescence signal shows up only when P concentration overcomes a 4% threshold.

104 96 9. P-doped fibers and preforms P L In te n s ity a t 3 e V (a rb. u n its ) (a ) P -c o n te n t P L In te n s ity P c o n te n t (w t% ) D is ta n c e fro m c e n te r (m m ) P L In te n s ity a t 3 e V (a rb. u n its ) (b ) P c o n te n t (w t% ) Figure 9.3: (a): PL intensity at 3 ev measured at room temperature under laser excitation at 4.8 ev with W T = 20 ms and T D =5 ns (full circles, left vertical scale) and phosphorus content (open circles, right vertical scale) as a function of the distance from the preform center. Vertical lines refer to the different sample zones (from 1 to 4) as listed in Table 4.1 and in Figure 4.6. (b): PL intensity at 3.0 ev as a function of the phosphorus content. We performed time-resolved emission measurements (Figure 9.4) on the 3.0 ev band, by acquiring at room temperature several emission spectra upon laser excitation, with W T = 500 µs and T D going from 1 µs to 25 ms. These data allow to study the decay kinetics of the PL signal at several spectral positions within the emission band. Since the decay turns out to be single-exponential at any fixed emission energy, the lifetimes were obtained by a fitting procedure with a single exponential function of time-resolved PL data at several spectral positions, as shown in Figure 9.5, where full lines represent the best fit curves. We found a slight dispersion of the lifetime inside the emission band: τ varies from 6.9 ms at E em = 2.7 ev to

105 9.1. Optical activity of P-related point defects ms at E em = 3.2 ev. Such a dispersion of the lifetime corresponds to a progressive red shift of the emission peak of Figure 9.4 during the decay, due to the low energy tail of the signal decaying slower than the right tail. Figure 9.4: Time resolved PL spectra measured under laser excitation at 4.8 ev with W T = 500 µs and T D from 1 µs to 25 ms. We also studied the dependence on temperature of the luminescence signal measured by exciting at 4.8 ev in the central sample region. Figure 9.6 shows the PL spectra under laser excitation at 4.8 ev at different temperatures. Notwithstanding a certain degree of scattering of data points, from this investigation we can clearly see that the emission intensity excited at 4.8 ev is poorly dependent on temperature in the range K. (inset of Figure 9.6); also the peak position and width do not depend significantly on temperature. The lifetime τ measured on the peak, at 3 ev, is independent of temperature as well within experimental accuracy, as shown in Figure 9.7 (a), where several decay curves at various temperatures are reported. Performing the same investigation on the left tail of the band, a small variation of the τ value with the temperature appears (full circles and triangles in Figure 9.7(b)). This may suggest the existence of another small component centered at energies < 2.8 ev. This secondary effect needs a specific investigation but it will not be analyzed in this Thesis. It is worth noting, however, that the presence of this component may contribute to the observed asymmetry of the emission band.