Transparent glass-ceramics functionalized by dispersed crystals

Transparent glass-ceramics functionalized by dispersed crystals

Accepted Manuscript Transparent glass-ceramics functionalized by dispersed crystals Xiaofeng Liu, Jiajia Zhou, Shifeng Zhou, Yuanzheng Yue, Jianrong Q...

5MB Sizes 0 Downloads 3 Views

Accepted Manuscript Transparent glass-ceramics functionalized by dispersed crystals Xiaofeng Liu, Jiajia Zhou, Shifeng Zhou, Yuanzheng Yue, Jianrong Qiu PII: DOI: Reference:

S0079-6425(18)30019-7 https://doi.org/10.1016/j.pmatsci.2018.02.006 JPMS 499

To appear in:

Progress in Materials Science

Received Date: Revised Date: Accepted Date:

5 June 2017 9 February 2018 10 February 2018

Please cite this article as: Liu, X., Zhou, J., Zhou, S., Yue, Y., Qiu, J., Transparent glass-ceramics functionalized by dispersed crystals, Progress in Materials Science (2018), doi: https://doi.org/10.1016/j.pmatsci.2018.02.006

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Transparent glass-ceramics functionalized by dispersed crystals Xiaofeng Liu,a Jiajia Zhou,b Shifeng Zhou,c* Yuanzheng Yued and Jianrong Qiub*

a

School of Materials Science and Engineering, Zhejiang University, Hangzhou 310027, People’s Republic

of China. b

State Key Lab of Modern Optical Instrumentation, College of Optical Science and Engineering, Zhejiang

University, Hangzhou 310027, People’s Republic of China. Email: [email protected] c

School of Materials Science and Engineering, South China University of Technology, Guangzhou,

510640, People’s Republic of China. Email: [email protected] d

Department of Chemistry and Bioscience, Aalborg University, DK-9220 Aalborg, Denmark

1

Abstract: Transparent glass ceramics (TGCs) with minimized scattering loss offer the combined characteristics of both glasses and (transparent) ceramics. The functionalities of the dispered crystals make TGCs a new generation of tailorable optical materials with a wide range of applications from optics to photonics. Most of conventional glass ceramics (GCs), e.g., silicate glass ceramics, contain crystals involving both network formers and modifiers, and they are known for their superior mechanical/thermal performances. In this paper, we pay more attention to those TGCs containing crystalline phases composed of only network modifiers, including nanocrystals of noble metals, metal fluorides, oxides, chalcogenides, etc. We review recent advances in conventional fabrication methods as well as in emerging techniques for the production of TGCs, such as solid state reaction, sol–gel and laser–induced crystallization. We then discuss the applications of TGCs, particularly the TGCs functionalized by crystals that exhibit various optical functionalities, including photoluminescence, optical nonlinearity, plasmonic absorption, etc. Experimental advances in the use of TGCs for lasers, optical amplifiers and different spectral converters are highlighted. We also anticipate that TGCs will find new applications, and the investigations into TGCs will unravel the mechanism of crystal formation, and hence, lead to the discovery of novel TGC systems.

2

1. Introduction ............................................................................................................................. 4 2.Nucleation and crystallization of GCs ......................................................................................... 7 2.1 Brief overview of nucleation and growth theory ................................................................... 7 2.1.1 Classical nucleation theory (CNT) ..............................................................................7 2.1.2 Beyond the classical nucleation theory ..................................................................... 11 2.2 Simulation of the nucleation and crystallization ................................................................. 12 2.3 Crystallization habits and textures of GCs .......................................................................... 16 2.3.1 Surface and volume crystallization ........................................................................... 16 2.3.2 Texture of the crystalline phases ............................................................................... 17 3. Typical GC systems and a new classification ............................................................................... 18 3.1 TypeI GCs: Network formers participate in crystallization ............................................... 19 3.1.1 Silicate GCs ............................................................................................................. 19 3.1.2 Phosphate, germanate, borate and beyond ................................................................ 21 3.1.3 Congruent crystallization.......................................................................................... 23 3.2 TypeII GCs: Network formers do not participate in crystallization ................................... 25 3.2.1 Metal oxides, fluoride and other dielectric crystals ................................................... 26 3.2.2 Non-oxide Semiconductor NCs ................................................................................ 29 3.2.3 Metal nanoparticles .................................................................................................. 30 3.3 Beyond type–I and type–II: Elemental NCs formed by reducing the glass network............. 31 4.Properties and characterization methods .................................................................................... 33 4.1 General requirements for optical transparency in GCs ........................................................ 33 4.2 Mechanical, thermal, electrical and chemical properties ..................................................... 35 4.3 Characterization techniques for GCs .................................................................................. 36 4.3.1 Xray diffraction ...................................................................................................... 37 4.3.2 Electron microscopy, electron energy loss spectroscopy (EELS) and scanning probe microscopy (SPM) .................................................................................................................... 38 4.3.3 Smallangle neutron / X-ray scattering..................................................................... 40 4.3.4 Inelastic Light–scattering ......................................................................................... 41 4.3.5 Raman spectroscopy................................................................................................. 42 4.3.6 Solid state nuclear magnetic resonance ..................................................................... 43 4.3.7 EXAFS / XANES..................................................................................................... 44 4.3.8 Thermal analysis ...................................................................................................... 45 5. Synthesis of TGCs: Conventional and unconventional techniques ............................................... 46 5.1 Crystal precipitation by confined solid state reaction .......................................................... 46 5.1.1 Controlled nucleation and growth by isothermal treatment ....................................... 46 5.1.2. Controlled nucleation and growth by cooling .......................................................... 48 5.1.3 Differences between typeI and typeII GC formation by solid state reaction .......... 48 5.1.4 Predicting crystallization process ............................................................................. 50 5.1.5 Manipulation redox equilibrium for elemental particle precipitation ......................... 52 5.2 Powder sintering, co-melting and frozen sorbent method .................................................... 55 3

5.2.1 Powder sintering ...................................................................................................... 55 5.2.2 Co-melting ............................................................................................................... 58 5.2.3 Frozen sorbent method ............................................................................................. 58 5.3 Solgel process .................................................................................................................. 60 5.4 Laser induced crystallization techniques............................................................................. 63 5.4.1 Continuous wave (CW) laser induced crystallization ................................................ 64 5.4.2 Pulse laser crystallization ......................................................................................... 67 5.5 Electron beam (E-beam) irradiation .................................................................................... 73 5.6 Orientated crystallization by external physical fields .......................................................... 75 5.6.1 Electric and magnetic fields ..................................................................................... 75 5.6.2 Mechanical stress ..................................................................................................... 77 5.7 Surface crystallization and control...................................................................................... 79 5.8 Melt–in–tube (MiT) technique for TGC fibers .................................................................... 81 6. Emerging applications of TGCs................................................................................................... 83 6.1 Luminescent TGCs for optical and photonic applications ................................................... 83 6.1.1 Luminescent TGCs as spectral converters for solar cells ........................................... 83 6.1.2 Whitelight LED...................................................................................................... 87 6.1.3 Laser gain materials ................................................................................................. 90 6.1.4 Optical amplification at communication wavelength region ...................................... 92 6.1.5 Scintillator and related applications .......................................................................... 96 6.1.6 TGCs with persistent luminescence .......................................................................... 99 6.2 TGCs containing optical functional crystals for nonlinear optics and photonics .................. 99 6.2.1 TGCs for second harmonic generation (SHG) .......................................................... 99 6.2.2 Metal and QD doped TGCs for third–order optical nonlinearity ............................. 103 6.2.3 Saturable absorbers for pulse laser.......................................................................... 107 6.2.4 GC containing plasmonic oxide NCs for electrochromic window ........................... 109 6.2.5 Photocatalysis ........................................................................................................ 111 6.2.6 Microwave absorption ............................................................................................ 111 6.2.7 TGC for enhanced IR radiance ............................................................................... 112 6.3 Miscellaneous optical applications ................................................................................... 112 6.3.1 TGC as optical components .................................................................................... 112 6.3.2 Radome dome materials ......................................................................................... 114 7. Conclusion and future prospects ................................................................................................ 115 Acknowledgements ....................................................................................................................... 118 References .................................................................................................................................... 118

1. Introduction

Glassceramics (GCs) are composite multiphase materials with crystallites dispersed in super–cooled liquid and glassy matrices. From a general perspective, they can be considered as the 4

solid state version of colloidal dispersion. The crystallites dispersed in a GC can be metals [1], semiconductors, oxides and non–oxides, which are also dispersible in liquid colloidal system. Glasses doped with metal chalcogenide quantum dots (QDs) or metal nanocrystals (NCs) are usually not referred to as GCs, but they can be regarded as a special class of GC as they are often fabricated by a typical GC route (e.g., thermal treatment of the parent glass), and thus share the important characteristics with common GCs. Similar to wet–chemistry process, GCs are often synthesized by heating the glass “solvent” to elevated temperatures, where precipitation of solvated species occurs as long as the thermal energy is sufficiently high to overcome the energy barrier of both nucleation and crystal growth. The crystalline phases precipitated in the glassy matrix not only improve the mechanical strength, but also provide new functionalities that are not found in their parent glasses. The earliest man–made GC is a solid state version of colloidal dispersion of Au NCs [1], which has been widely used to decorate windows in cathedrals and vases due to the charming reddish color that stems from localized surface plasmon resonance. Noble metals also played an essential role in the development of the first modern GC in the 1950s by Stooky, in which UV–precipitated Ag nucleus induced the crystallization of a Li 2O–SiO2 glass [1],[2],[3]. This invention has stimulated the development of various types of silicate based GCs, which have found a broad spectrum of applications from domestic use to defence technology mainly due to their enhanced mechanical properties. Unlike these “conventional” GCs, the precipitation of the functional crystals in a glassy matrix, such as semiconductors, plasmonic particles and luminescent crystals, has made these “nonconventional” GCs highly attractive for optics and photonics communities. Most of GCs can be made transparent by minimizing light attenuation due to scattering and absorption. In comparison to polycrystalline ceramics that are usually opaque due to scattering by pores and grain boundaries, most GCs are pore-free and highly transparent GCs (TGCs) containing 5

nanocrystals (NCs) of less than, say 30 nm, can be usually made by a controlled crystallization process. Larger crystals up to micrometre scale can exist in TGCs without loss of transparency only when the refractive index difference between the crystalline and the glassy phase is very small (<0.01). TGCs possess most of the important features of glasses and polycrystalline ceramics, such as high optical transparency, good mechanical stability, and tailorable optical properties. They are of particular interest for applications as optical materials, such as lens and laser gain materials. Compared with glassy materials, TGCs offer balanced properties and are regarded as alternatives to glasses and crystalline materials for optical applications as they can be easily processed into large plates or even fibers of optical quality comparable to glass. Moreover, the functional crystals precipitated in glassy matrix provide TGCs with new functionalities, such as second harmonic generation (SHG), which does not exist in pure glasses. The development of both conventional GCs and functional TGCs has been reviewed in several papers and books [4][5][6][7][8][9][10][11]. However, in recent years there have been growing numbers of studies that were devoted to new GC systems, new fabrication processes and new applications, especially in the photonics area. This review discusses the recent advances with focus on the preparation and applications of TGCs as optical materials. It should be noted here that TGCs discussed in this review are characterized by minimized scattering attenuation, while such TGCs can exhibit absorption caused by the judiciously introduced ions or nanoparticles (i.e., QDs). First, we give a brief overview of the fundamental of the physics and chemistry of GCs, such as the nucleation process, mechanical and optical properties. We then present an overview of the existing TGCs together with a new classification method. Afterwards, we discuss the methods for characterizing GC, including both spectroscopic methods and electron microscopy. In the following section, we discuss the advances in the methods for the preparation of TGC including both conventional processes by solid state reaction and the recently developed laser induced crystallization techniques. We introduce diverse applications of TGCs in section 6, especially those 6

in the optics and photonics fields. In the closing section, we present a short summary of this review along with a brief description on future research directions for TGCs.

2.Nucleation and crystallization of GCs

2.1 Brief overview of nucleation and growth theory 2.1.1 Classical nucleation theory (CNT) Glasses are known as supercooled liquids frozen under ambient kinetic conditions. Therefore, theoretical description of nucleation for forming GCs is primarily based on classical nucleation theories (CNT) that is originally applied for the description of crystal precipitation in solutions. Nucleation in different systems has been classified into homogeneous and heterogeneous nucleation. In the case of homogeneous nucleation, nucleuses are formed in the absence of nucleation agent or interface between a second phase as a result of local composition or density fluctuation. The change in free Gibbs energy can be described as follows [12]: (1) where r is the radius of a spherical nucleus,  is the interfacial energy for the formation of new phase, gv is the free energy change per volume due to the formation of nuclei, and GE is the elastic distortion energy caused by the structural change during crystallization. GE can be omitted for nucleation from solution or gaseous phase, but should be taken into account for surface crystallization. The elastic strain is caused by the density difference between the crystalline and the glassy phase and it has a remarkable effect on the thermodynamics of nucleation in glass at temperatures around 1.2 Tg (Tg: glass transition temperature). In this case, the Stocks-Einstein relation no longer holds, and the nucleation cannot be regarded as an event occurring in a viscous 7

flow [14]. For instance, in lithium disilicate GC, the nucleation rate is greatly suppressed by two orders of magnitude due to the presence of the elastic stress [15]. Since the surface term (4r2) is positive, G is positive for smaller nucleus but negative for large ones. Therefore, in order for nucleus to be capable of further growth, G should be smaller than G*. In this regard, a critical nuclei radius (r*) can be found when dG/dr = 0 [13]: (2) Accordingly, for nucleus with size r* the change in free energy can be described as: (3) Particles with size greater than r* are capable of further growth, while smaller particles are not stable and would be dissolved. Unlike homogeneous nucleation, heterogeneous nucleation that occurs in the presence of interfaces with a second phase or phase separation is energetically favoured and practically more often observed. The presence of a nucleation agent can dramatically reduce the kinetic barrier for the formation of a new phase. Fig. 1 presents a model for heterogeneous nucleation at the interface. The critical free energy in heterogeneous nucleation can be described as a function of contact angle () by: (4) where (5) and, (6)

8

where HL, SL and SH are the interfacial energies corresponding to substrate/liquid, solid/substrate and solid/liquid interfaces, respectively. From the above equations, the free energy is determined by the wettability of the substrate. If the substrate is completely not wettable, f() = 1 (as  =180 ), the process is reduced to homogeneous nucleation. If the substrate is wettable ( < 180 ), heterogeneous nucleation prevails over homogeneous nucleation as the critical free energy is reduced. Moreover, heterogeneous nucleation prefers small HL and SH, and a small mismatch of the lattice constants between the nucleation agent and the formed crystals is also beneficial to initialization of nucleation by solid state epitaxy. However, this does not mean that all the heterogeneous nucleation process starts from the interface between the glass and a second phase. In practical glasses, this process can be much more complex and sometime occurs subsequent to a liquid phase separation process. Since the surface of the crucibles always provides nucleation sites for heterogeneous nucleation, the use of containerless levitation method developed recently for glass synthesis therefore can efficiently suppress heterogeneous nucleation, giving access to a glass melt of much larger super–cooling than conventional process [16][17].

L

S

θ

H

Fig. 1 Schematic representation of heterogeneous nucleation. L, S, and H represent parent phase (glass), nucleus (crystal) and the heterogeneous substrate (catalyst). The contact angle is represented by

.

9

Nucleation rate (I), Growth rate (V)

I

V

Ostwald-Miers Temp. range 2/3 Tg

1 Reduced temperature (T/Tl)

Fig. 2 Nucleation rate (I) and growth rate (V) as a function of reduced temperature (T/Tl). Tl is the liquidus temperature of the glass. OM (OstwaldMiers) range represents the metastable range of super-cooling. With the thermodynamic barriers for homogeneous and heterogeneous nucleation, we can derive a general expression for steady state nucleation rate based on CNT as: (7) where A can be considered as the steady state nucleation rate, G* is the critical free energy change, which is given by eq. (3) for homogeneous nucleation, and by eq. (4) for heterogeneous nucleation, GD represents the kinetic barrier for nucleation, and k is Boltzmann’s constant. Since the thermodynamic term (G*) is much smaller than GD, the rate of heterogeneous nucleation can be orders of magnitude faster than that of homogeneous nucleation. In the non-steady state case, the time-dependent nucleation rate can be expressed as [18]: (8) where I0 can be considered as the steady state nucleation rate, and the  is non-steady state time lag (incubation period).

10

Different models have been proposed to describe the thermodynamic and kinetic factors of crystal growth. Since the growth process requires the diffusion of atoms, the growth rate (U) is dependent on the diffusion coefficient (D) or viscosity () of the glass melt [19]: (9) where f is the interface area, a0 is the diffusion jump distance,  is the viscosity of the glass melt and can be connected with diffusion coefficient (D) by the Stocks-Einstein relation, Gv is the change in free energy in the transition from glass to crystalline phase, T is the absolute temperature. A more detailed mathematical treatment of the growth kinetics of the GC can be found in the revi ew paper of Fokin and coworkers [14]. Fig. 2 shows the dependence of homogeneous nucleation and growth rate on temperature. In general, the temperatures for the maximum growth rate are 50 – 100 C above the glass transition temperature (Tg). Crystal growth will take place at the growth temperature for a nucleus as long as the critical size r* is attained. As shown in Fig. 2, there is a large overlap in the temperature range for nucleation and growth, hence both process can occur simultaneously. In general, the temperature for the maximum nucleation rate is much lower than that for the maximum growth rate. However, at the high temperature side (Ostwald–Miers range), nucleation stops so that only nucleus with size lager than r* that form at low temperatures would grow further to larger crystals. Fig. 2 can provide important guidelines for the design of temperature profiles of thermal treatment for the fabrication of TGCs, as discussed in 5.1. For a more detailed discussion on the CNT, the readers are recommended to refer ref. [7] and references therein.

2.1.2 Beyond the classical nucleation theory CNT can provide a qualitative picture for the understanding of the major physical mechanisms behind nucleation in glass. However, this theory fails in certain conditions in the description of the 11

nucleation rate by several orders of magnitude due to the use of improper approximations. New theories have been developed by taking into account of the pathways to nuclei formation based on experimental observation. The generalized Gibbs’s approach (GGA), for instance, assumes that the nucleus change their chemical composition in the nucleation process, which is in line with recent experimental observations and Ostwald’s rule of stages [20]. It was proposed by Ostwald that the final stage (crystalline phase) is formed via a series of energetically favored metastable phases. In this sense, the GGA can be considered as a modification of the CNT by the introduction of the composition changes. According to GGA, the critical nuclei size and free energy can be described by the following equations [14][21]: (10)

(11) where C is the volume concentration of the new phase. Unlike CNT which assumes a constant composition and continuous size increase in nucleation, the GGA describes a quite different nucleation process in which the size and composition both change until they reach the final state. The critical free energy, as expressed by eq. 11, is much lower than that predicted by CNT, thus giving a much higher rate of nucleation that is closer to experimental observation. The discrepancy between the nucleation rate predicted by CNT and that observed experimentally has been studied by several other groups [22],[23],[24],[25],[26],[27] and new insights into the nucleation and the crystallization are provided.

2.2 Simulation of the nucleation and crystallization Experimental verification of the nucleation and crystallization theory requires the direct measurement of the crystal and nucleus size, fraction and their time-dependent evolution, which can be accessed by using different characterization techniques. However, currently even with the 12

most advanced tools, a clear observation of the crystallization process in glass at its very beginning, i. e., nucleation stage, remains a tremendous challenge [28][29]. Information on short range (bond length, coordination number, etc.) and medium range (polyhedron connectivity) order can be provided by diffraction, scattering techniques and NMR, while ring size statistics can be only obtained by molecular dynamics (MD) simulation, which can be used to predict phase transitions of large systems (over 10 4 atoms). Simulations based on MD have been extensively used to understand the structure, glass transition and structural relaxation of different glasses. The simple colloidal glass model containing spheres of very different characteristics is widely employed to simulate the structures well as phase transitions that occur in glass systems [30][31][33][34][35][36]. Simulations on the colloidal glass can reproduce and follow well the behaviours predicted by theory, including phase separation and transition, and nucleation and crystallization kinetics. Experimentally, the colloid system allows for the direct observation of dynamics process by optical method (such as confocal microscopy) [34][35], therefore enabling a simple and direct understanding of dynamics process such as nucleus formation, from a very different angle. More detailed discussions on the crystallization and dynamics of colloidal glass have been given in refs. [36][37]. Beside the simple colloid model, MD simulations based on a Lennard-Jones liquid (interacting particles under Lennard-Jones potential) can well reproduce the picture of homogeneous and heterogeneous nucleation processes in GCs [38]. It has been found that the energy barrier for heterogeneous nucleation is minimal when the lattice spacing of the new crystalline phase is close to that of the impurity phase. Specifically for oxide glasses like silicate glasses, MD simulation based on energy minimization has been widely used to study the composition-structure relationship and to analyze the tendency to crystallization [39]. In addition, simulation can provide information about the ring statistics and therefore a comparison with crystalline phase can give some hints on the nucleation mechanisms. Regarding oxyfluoride TGCs for optical applications, special attention 13

has been paid to the RE-doped TGCs and the role of RE ions in the crystallization process. Simulation by Dantelle and coworkers indicates that Er3+ ions can enhance the crystallization of PbF2, which does not take place directly around Er 3+ ions but it probably starts at short distances from Er3+ [40]. In an oxyfluoride GC, Qiao and coworkers clarified through MD simulation that, during crystallization process, silicon atoms are coordinated exclusively to oxygen, while AE (alkali earth) and RE ions are coordinated to both F and O anions [41]. Therefore, Al in the oxyfluoride glass serves as the bridge that links the fluoride and oxide regions. With the progress of crystallization, oxide region is gradually excluded as a result of phase separation (see Fig. 3), leaving only fluoride network to become crystallized. This phase separation is in agreement with the event predicted by the Hard Soft Acid Base (HSAB) theory [42], which has been widely used for understanding the precipitation process in solutions. However, MD simulation can only infer that the composition of the fluoride region might be close to that of the crystalline phase. It is still difficult to predict the accurate structure and exact composition of the crystallized phase by only MD simulation. Furthermore, energy minimization can be made with the help of density functional theory, by which a crystalline phase can be predicted.

14

Fig. 3 MD simulation for the structures of different glasses. (a , b, c) Overview of the MD simulated structures. The phase separation is evident in b and c. (d, e, f): Detailed local structures. (g, h, i): The structure of the possible crystalline phases. g: monoclinic BaAl1.94Si2.06O8; h, β-BaAlF5 ; i: cubic BaF2. The glass compositions are 50SiO2 - 20Al2O3 - 30BaO (a, d, g), 50SiO2 - 20Al2O3 - 30 BaF2 (b, e, h), and 50SiO2 -16Al2O3-34BaF2 (c, f, i), Yellow sphere Si; magenta Al; green Ba; Red O; cyan F. Reprinted with permission from ref. [41], copyright 2012, American Chemical Society.

15

2.3 Crystallization habits and textures of GCs 2.3.1 Surface and volume crystallization As shown in Fig. 4, two distinct crystallization mechanisms are involved in preparation of GCs, i.e., surface and volume crystallization. It has been widely accepted and experimentally confirmed that the devitrification of glasses with Trg < 0.58 (Trg = Tg/Tm, where Tm is melting temperature) is predominated by volume crystallization, while surface crystallization may occur as well [39][43]. In this case, the maximum nucleation temperature is lower than Tg and therefore a homogeneous nucleation is favored. If Trg > 0.6, the nucleation rate near Tg is low and the heterogeneous nucleation becomes dominating, which is often observed in the surface crystallized GCs. From a structural perspective, for volume crystallization there is structural similarity between the glass and the corresponding isochemical crystal phase at the scales of both short and medium range order. It has been found that GCs with a small density difference (< 10%) between the crystalline and the glassy phase usually undergoes homogeneous nucleation and therefore volume nucleation may dominate [44]. If the density difference is large, the high elastic distortion energy at the interface (GE, see eq. 1) may suppress bulk crystallization and therefore the GCs tend to surface crystallization [45]. Practically, volume crystallization is often desirable for obtaining a broad range of new TGC materials with favorable optical homogeneity. In this case, nuclei in glass undergo continued growth with a linear increase in volume until they are impeded by neighboring crystals. This growth process is followed by the secondary growth, which results in fusion of smaller crystals in favour of larger ones (see Fig. 4). In many GCs, nucleation agents, such as TiO 2, are usually added to initialize nucleation, and thereby to assist volume crystallization by reducing the energy barrier for nucleation. Although there is strong confinement by the rigid glassy phase, many crystals with

16

anisotropic structures still preserve their growth behaviours, e.g., orientated growth, which is often seen in wet–chemistry synthesis. In comparison, surface crystallization will dominate if volume crystallization cannot be initiated. This is usually a heterogeneous process and thus thermodynamically more favorable. It is generally accepted that surface crystallization relies on surface nucleation initialized by tribochemical activation or other methods. After nucleation, surface crystallization proceeds vertically towards the interior of the glass and the crystals usually exhibit a strong preferred growth orientation. Due to the highly preferred orientation of the crystallite, surface crystallized GCs are of particular interest for certain applications, such as surface waveguide and optical SHG (see Fig. 4). a

b

Fig. 4 Bulk versus surface crystallization. (a) Bulk nucleation and growth: crystals are all randomly oriented. (b) Surface crystallization without internal crystal precipitation: crystals all have a preferred crystallization direction towards the inside of the glass. Reprinted with permission from ref. 7.

2.3.2 Texture of the crystalline phases The texture of the precipitated crystals plays an important role in the performances, especially mechanical properties, of GCs. The formation of textures during devitrification depends not only on the crystallization behavior, but also on the processing method. In GCs prepared by thermal 17

treatment, the size and shape of the crystals are usually completely different compared to conventional ceramics prepared by sintering. In general, the crystal precipitation in GCs is similar to that in wetchemistry synthesis, where a preferred growth direction is often observed due to anisotropy of the crystal structure. As summarized in a review by Beall [4], eight types of textures have been found in different GCs: 1) dendritic, 2) ultra-small crystals, 3) cellular membrane, 4) relict, 5) coastandisland, 6) card house, 7) acicular interlocking, and 8) lamellar twinned. The formation of some of these textures (such as dendrite) also occurs in wet-chemistry process, whereas other types of structures rely on both the special GCs composition and heat treatment condition.

3. Typical GC systems and a new classification

Historically, GCs are classified primarily according to their physical properties, chemical composition and the specific applications. In most current classifications, there is no clear distinction between, e.g., silicate and fluorosilicate GCs, despite that we all know that these GCs are very different in chemistry. In this review, GCs are classified, according to a rather simple but chemically reasonable criterion, into two different distinct types: 1) glass network formers participate in the crystallization process, e.g., in formation of silicate crystals from a silicate glass matrix; 2) glass network formers do not participate in the crystal formation, e.g., precipitation of fluoride crystals from an fluorosilicate glass host (Fig. 5). This classification therefore is based on the chemistry of GCs. Most of conventional GCs containing crystals such as silicates belong to the first type, whereas GCs containing metal fluorides and oxides, as well as those “doped” with metal NCs, metal chalcogenide QDs, etc., belong to the second type. The second type is usually not regarded as GCs. Unlike the first type, the glass network formers in the second type serve as the inert “solvent” for the precipitation of crystals in the formation of type-II GCs. The first type of GCs 18

often exhibits enhanced mechanical properties, while the second type shows new optical and photonic functionalities compared to their parent glasses. In addition, these two types of GCs also differ in their fabrication as well as in composition design as discussed in section 5.1. a

b

Crystallization of Crystallization of network formers network modifiers Type-I

Type-II

Fig. 5 Schematic illustration of the typeI and typeII GC. (a) Glass network formers are involved in the crystallization of typeI GC, while only network modifiers are crystallized in typeII GC, i.e., the glass network only serves as a “solvent” for the crystal precipitation. (b) The relation between type-I and type-II glass: Type-I, both network formers and modifiers are crystallized; type-II, ONLY network modifiers are crystallized.

3.1 TypeI GCs: Network formers participate in crystallization In this type of GCs, the glass network formers, such as silica, participate in the crystallization. Therefore, the glass network not only serves as the “solvent”, but also the reagents that provide atoms to the crystals precipitated. Most conventional GCs, such as silicate and quartz based systems, belong to this type. Glasses that crystalize congruently are also parent glasses for type-I GCs.

3.1.1 Silicate GCs Silicate GCs are the most popular among GC materials. A thorough description and classification of these GCs have been given in ref. 7. As silica is the major network former the precipitation of metal silicate crystals from glass usually requires high annealing temperature. 19

Therefore, nucleation agent, such as TiO 2 and ZrO2, are usually added to facilitate the nucleation. Depending on the types of crystallized phase, GCs based on a silicate parent glass can be further divided into different sub–groups based on the type of crystal phase, such as quartz, mica and etc. The types of silicate GCs are listed in Table 1. The composition of the parent glasses and the condition of heat treatment both determine the type of silicate crystals to be precipitated. Multicomponent glasses are usually used in order to increase the glass forming ability and other properties of the glass [3],[46],[47]. Glasses based on aluminosilicate, for instance, have been widely examined for the preparation of GCs with desirable mechanical properties. The simplest aluminosilicate is mullite, which can be precipitated from the binary glass SiO 2–Al2O3. However, to ease the fabrication process, multi–component glasses containing over 10% of B 2O3 are more favorable due to reduced melting temperature [48]. Among oxide GCs, the ternary Li 2O–Al2O3–SiO2 system is one of the most important GCs which crystalize into –Quartz or –Spodumene related solid solution phases. Along with the high mechanical strength, this GC demonstrate high thermal shock resistance due to the near zero coefficient of thermal expansion (CTE) [4], [7]. Depending on the composition of the parent glass, a variety of crystalline phase can be precipitated. Some typical crystals formed in the silicate glasses are given in Table 1. Beside quartzsolid solution based GCs obtained from the aluminosilicate glasses, in SiO2AEO (AE: Mg, Ca, Ba) based systems, a number of GCs, which contain silicate crystals with complex chemical compositions, have been developed. Different types of metal silicate crystals, such as enstatite (MgSiO 3) and cordierite (Mg 2Al4Si5O18), can be precipitated from MgO-containing glass [49]. In many of such GCs, the nucleation agents like ZrO2 or TiO2 have to be added to initialize the nucleation upon heat treatment. Compared to other silicate glasses, mica based GCs are of particular interest due to its facile machinability [50]. Different types of mica crystals can be precipitated in fluorosilicate glasses [7]. 20

3.1.2 Phosphate, germanate, borate and beyond Phosphate exhibits good glass–forming ability, and phosphate glasses are of special interest for applications in high power lasers and biomaterials in spite of their poor chemical stability. In phosphate glasses, PO4 is the basic building block, while PO 4 tetrahedra can only link to three neighboring PO 4 units due to the pentavalence of P. So far, different types of phosphate GCs have been developed for different applications [51],[52],[53]. Due to their favorable bio–compatibility, large numbers of phosphate GCs have been developed in recent years aiming for applications in human medicine for bone replacement. For instance, in the trade–marked GC BIOVERIT ® III, different phosphates including apatite and AlPO4 (with different structure types) and complex phosphates like Na5Ca2Al(PO4)4 are precipitated [51]. The parent glass contains SiO 2 and P2O5 as the network formers (see Table 1). The presence of SiO 2 can improve the chemical stability and mechanical strength of GCs.

Table 1 Selected examples of typeI GCs formed by partial crystallization of network formers (i.e., SiO2, GeO2, B2O3) 21

Crystal type Silicate

Ref.

80SiO24Al2O310.5Li2O5.5K2O0.02CeO20.04AgCl

Mullite (2SiO23Al2O3)

48SiO211B2O329Al2O310ZnO2K2O Additives: 0.1wt%Cr2O3, 0.4wt%As 2O3 68.8SiO219.2Al2O32.7Li2O1.8MgO1.0ZnO0.2Na2O 0.1K2O0.8BaO2.7TiO21.8ZrO20.8As 2O30.1Fe2O3 (Vision  Corning USA) 72.5SiO222.5Al2O35.0Li2O (Cercor Corning USA)

750900 C

9001000 C

[7]

43.5SiO229.6Al2O313.9Na2O5.6BaO7.4TiO2 (Centura  Corning USA) 56.2SiO219.8Al2O314.7MgO0.1CaO8.9TiO20.3As 2 O5 Corning Code 9606 54SiO233MgO13ZrO2 (Corning E2)

8201140 C

[7]

820 C /2h; 1260 C /8h

[7]

800 C /2h, 1240 C /4h 700950 C

[49]

600 C /24h; 750 C 24h 600 C 20 h /730 C 12 h

[51]

400-650C/8h

[54]

MgSiO3 (enstatite)

Silicophos phate Phosphate

Annealing condition 600 C

Lithium metasilicate Li2SiO3 Na2OZnO2SiO2

quartz solid solution (Li2,R)OAl2O3nSiO2 (n: 2~10) spodumene solid solution Li2OAl2O3nSiO2 (n: 4~10) (Na,K)AlSiO4 (nepheline) Mg2Al4Si5O18 (cordierite)

Fluorosilic ate

Glass composition*

Mica K1−x Mg3 (Al1−x Si3+x O10)F2 SiO2–MgO–Na2O–K2 O–CaO–P2O5 (Apatite) Ca3(PO4)2 + LiTi2(PO4)3

R2OZnOxSiO2 (R: Li, Na; X = 2~10)

30~50SiO2  3~20 B2O3  10~20Al2O3  4~12 K2O  15~25 MgO  4~10 F 40~50SiO230~35CaO2.5~5.0MgO5~10Na 2O0.5~3 K2O10~15P2O5 (by weight) P2O5–CaO–TiO2–Li2O

Germanote Li2TeO3 + Li2GeO3 TeO2–xGeO2–8TiO2–8BaO–22Li2O llurite *Compositions of the listed glasses are given in molar fraction (mol.%) unless otherwise stated.

[3] [46], [47] [48], [4] [4]

[50]

[53]

In the CaO–P2O5 system, different types of calcium phosphates can crystallize, depending on composition, such as β–Ca(PO3)2 and 2CaO3P2O5. In addition, due to the strong anisotropic structure of calcium metaphosphate, the oriented GC rod was prepared by Abe et al. by controlling the direction of crystallization [52]. To improve the physical and chemical properties of the glass, multi-component phosphate systems are often used for the preparation of GCs. As shown in Table 1, several metal phosphates have precipitated in phosphate glasses [52][53]. Germanate as well as borate GC are of less popularity compared to silicate and phosphate GCs. The research into such GCs has been driven primarily by the functionality of crystal phases, such as SHG. Indeed, SHG crystals have precipitated in several glasses including germanate and borate glasses. For instance, Ba 2TiGe2O8 (BTG) can form in a germanate glass with the composition of the crystalline phase, as discussed in section 3.1.3. Moreover, scientists have 22

developed TGCs from multicomponent glasses containing GeO 2 and a second network former, e.g., B2O3 and TeO2, since such GCs possess low intrinsic vibrational energy and high chemical stability. Qiu and coworkers have recently succeeded in fabricating germanotellurite TGCs based on the composition of (59x)TeO2xGeO28TiO28BaO22Li 2O3Er2O3 (x = 0, 10, 20, 30, 40, 50, 59). The precipitated crystals include Li 2TeO3 and Li 2GeO3 [54]. GCs can be prepared from pure non-oxide parent glasses, e.g., from fluoride and chalcogenide glasses. Compared to oxide GCs, these non-oxide GCs have attracted less attention, however, they have been explored for potential applications in nonlinear optics as well as infrared optics. Similar to oxides, a variety of metal and non-metal chalcogenides can be melted to form stable glasses with low melting temperatures. It has been found that partial crystallization can exert a strong influence on the optical properties. A large number of chalcogenide GCs have been investigated in the recent decades. For instance, GeGa 4Se8 and GeSe2 crystals form in the infrared (IR)–transparent glass of 80GeSe2–20Ga2Se3 upon controlled thermal treatment. The asformed GC with 40% crystalline phase demonstrated a significantly reduced thermal expansion coefficient and improved IR transmittance [55]. In recent decades, the GCs with enhanced mechanical and chemical durability have been developed for applications [56],[57],[58]. Partial devitrification of a chalcogenide glass can also modify the nonlinear optical properties [59],[60], e.g., SHG.

3.1.3 Congruent crystallization Most inorganic compounds crystalize during cooling of their melts (in case t hey do not decompose before melting). If their melts are highly viscous and do not crystallize upon cooling (at a moderate rate), glasses can form. These glasses can crystallize congruently upon isothermal treatment at the crystallization temperature. This type of GCs experiences a minimal local structural change and a shortest ionic diffusion path during crystallization. Furthermore, the crystal and the parent glass in such GCs have usually similar refractive index, resulting in minimized scattering 23

loss and high transparency. Table 2 gives several selected examples of such TGCs, which are of particular interest for photonic applications. More examples can be found in ref. [9]. One-component glasses made up of pure glass network formers, such as SiO 2 and P2O5, can crystallize under certain thermal conditions. For instance, pure quartz glass can certainly crystallize congruently, but applications of pure SiO 2 GC are limited due to difficulty in fabrication. In binary Li2O–SiO2 system, Li 2Si2O5 melt incongruently at temperatures above 1033 C. The formation of Li2Si2O5 from a stoichiometric parent glass has been confirmed and the mechanisms of nucleation have been under debate for decades as intermediate phases have been involved in the nucleation and phase formation [61],[62]. Likewise, there are many other binary silicate systems that can congruently crystallize, such as 3Al 2O3–3SiO2, and BaO–2SiO2 (Table 2) [63]. Furthermore, glass can form from many ternary silicates consisting of alkali metal oxides upon melting and quenching, and can also recrystallize upon heat treatment. In these silicate GCs, the precipitated crystals exist in the form of solid solutions phases and thus their chemical compositions could be slightly different from that of the parent glasses (containing nucleation agents), while high optical homogeneity is attainable in these systems due to their similar composition. In the binary system of CaO–P2O5 at the nearly stoichiometric composition, the precipitation of Ca–metaphosphate –Ca(PO3)2 can occur, together with the formation of trace amount of whitlockite (2CaO3P2O5). Moreover, the Ca–metaphosphate could crystallize along one direction, leading to enhanced mechanical strength [64], [65]. Several multi–component silicate glasses can crystallize congruently to form TGCs. For instance, Ba2TiSi2O8 (BTS), a nonlinear optical crystal, can be melted to form glass at the stoichiometric composition, and congruent crystallization of this glass takes place at around 860 C [66],[67]. The crystallization temperature of its germanate counterpart is lowered by 60 C. Fujiwara and co-worker have investigated both the crystallization mechanism and the potential applications of TGCs containing BTS or STS (Sr 2TiSi2O8) crystals of the Fresnoite structure 24

[68][69][70]. In addition, they found that glasses with a stoichiometry slightly different from BTS or STS also allow the precipitation of the same Fresnoite phase, and the crystallization behaviour may change from surface to bulk crystallization [70]. Similarly, both LaBSiO 5 and the germanate LaBGeO5 can form glass by conventional process, and they are both well–known for SHG [71][72][73][74]. For borates, there are also crystal phases that do not have well–defined melting temperature. Upon heat treatment of the glass a partially crystallized GC can be obtained. For instance, –BBO (BaB2O4), which is an important SHG crystal, can be precipitated from a parent glass of BaO–B2O3 with the exact stoichiometry [75]. More examples of glass systems that can crystallize isochemically can be found in chapter 2 (p.75) in ref. [7] and Table 1 in ref. [9].

Table 2 Examples of type–I GCs formed by congruent crystallization (i.e., the precipitated crystals have the same composition as the parent glass.) Compounds/glass composition** 2SiO2–Li2O/Li2Si2O5*

Annealing temperature ~ 500 C

remarks Rod-shaped crystals, Solid solution phase Spherulite crystals Solid solution phase Solid solution phase Needle-shaped crystals* SHG SHG Pyroelectric

~ 700 C --560 – 600 C 860 – 1100 C 800 – 1000 C Powder sintering at 1100 – 1200 C LaBGeO5 SHG Thermal treatment 950 C SHG –BaB2O4 (BBO) crystallization max at 585 C * Slight off-stoichiometry or small amount of additive (for nucleation) can be allowed. **Compositions of the listed glasses are given in molar fraction (mol.%) unless otherwise stated. 2SiO2–BaO/BaSi2O5* (sanbornite) 2SiO2–3Al2O3* (mullite) Li2O–Al2O3–2SiO2* (–eucryptite) –Ca(PO3)2* Ba2TiSi2O8 Ba2TiGe2O8 LaBSiO5

ref [7], [62] [7], [63] [7] [7] [65] [66], [67] [66] [71], [72] [73], [74] [75]

3.2 TypeII GCs: Network formers do not participate in crystallization The formation of typeII GC is quite similar to the precipitation of crystals from solution in wet chemistry synthesis. Water or organic solvent is employed in wetchemistry synthesis of NCs and they do not participate in the formation of the final product. Likewise, in the formation of typeII GCs, the crystal precipitation process in glass also does not involve the glass network 25

formers (e.g. SiO 2), which serves as the solvent medium for crystal precipitation. The chemical bonding in the crystals precipitated from the type–II GCs is mainly of ionic (or metallic) nature, and hence, their constituent ions are more mobile than those in the glassy matrix of highly covalent nature. The reactants under ambient temperature are oversaturated, yet they are kinetically immobilized in glassy phase. Upon heating, the corresponding cations and anions (i.e., fluoride and metal ions) become mobile, and their reaction results in the crystal formation. Ionic crystals, such as fluoride NCs, are one of the typical crystalline phases in type–II GCs, in which the fluoride ions are network modifiers and the silicate network formers are stable against crystallization. Besides, a number of metal nanoparticles (NPs) and oxides doped silicate glasses can be also classified as type–II GCs.

3.2.1 Metal oxides, fluoride and other dielectric crystals Various metal oxides can be precipitated from silicate glasses by carefully selection of the composition and thermal treatment conditions. The fabrication of oxide–precipitated GCs has also been partly driven by the need for new bulk optical materials with enhanced performance s but lower fabrication cost. Unlike silica, most metal oxides are ionic and many of them always act as glass network modifiers or intermediates. Therefore, the precipitation of these oxides can take place without participation of the glass network formers as long as a suitable thermal condition is set, as exemplified here by the various GCs containing binary and multinary crystals [76]. A list of such GCs containing oxides crystals are listed in Table 3. ZnO has become one of the most important oxide semiconductors and its incorporation into a glass host could be attractive for many applications. Pinckney reported the first ZnO precipitated TGC based on a parent glass of SiO2Al2O3ZnOK2O [76]. The glass must be melted at temperatures over 1600 C and crystallization occurred at temperatures between 650  1100 C. Controlling the crystallization condition can allow the formation of highly transparent ZnOdoped 26

GCs. Modifications of the initial composition developed by Pinckney have been reported later by other researchers. For instance, Fujiwara and coworkers successfully precipitated ZnO NCs in borosilicate and boroaluminate parent glasses at temperatures higher than 590 C [77][78][79]. Unlike ZnO, Ga2O3 have precipitated from a silicate glass of a simple composition: K2O–Ga2O3–SiO2 [80]. The Ga2O3–precipitated TGCs have been explored for further applications by introducing dopant ions, e.g., Ni2+, for photonic applications (see section 6.1). Oxides with much more complex compositions, such as ternary oxides, have been precipitated from silicate glasses with carefully selected compositions. Generally, the parent glass must contain sufficient amount of the targeted metal oxide in addition to the glass formers like SiO 2, at the same time the glass forming ability should not be substantially affected. For instance, the spinel –type MgAl 2O4 and the gahnite ZnAl 2O4 crystals can be precipitated from the SiO 2–Al2O3–ZnO–MgO system, in which the sum of MgO and ZnO should be higher than 13wt% [81],[82]. The nucleation agent, such as TiO 2 and ZrO2, is often added in order to control the nucleation process. Besides the AB2O4 crystals, the precipitation of functional ternary functional oxides, such as the perovskite ABO3 and other oxides with more complex compositions and structures has been reported in recent years [83–101] (see also Table 3 for more examples). The formation of silicate crystals generally involves the reaction of network formers, whereas fluoride precipitation involves only fluorine and metal ions. For instance, Bhattacharyya et al. found that the growth of BaF2 NCs in a specific silicate system was limited by increased diffusion barriers around crystallites due to an abrupt increase of viscosity of the silica rich shell of crystallites [102]. This finding is in line with Rüssel’s explanation about the crystallization in this type of TGCs, i.e., the coarsening of the hosted NCs is not favorable from a kinetic point of view. It is likely that there is a strong link between structural heterogeneity and NC formation as the aluminosilcate glasses is intrinsically heterogeneous at sub–nanometer scale [103],[104]. In fact, a more recent study based on electron microscopy observation suggested that the precipitation of 27

fluoride NCs was controlled by the amorphous phase separation that exists in the as-melt glass [105]. For decades, the development of nano–fluoride precipitated GCs has been largely driven by the need for development of better glass based luminescent materials with low phonon energy. Unlike silicates which form the glass network, fluoride ions in most oxide glasses do not participate in the network formation. Due to their ionic nature, different types of fluoride crystals can be precipitated easily from silicate glass, including alkali earth fluoride and rare earth fluoride. For instance, the simplest oxyfluoride glass is SiO 2–Al2O3–AEF2 (REF3) (AE: alkali earth, RE: rare earth), in which the fluoride can be precipitated at a much lower temperature than Tg [106],[107],[108]. RE ions, upon crystallization, enter preferably into the fluoride host of AEF2, or REF3 because of their close ionic size and electronegativity, making the RE ions highly efficient for upconversion [106]. Ternary fluorides, such as NaYF 4, have also been crystallized in a silicate glass matrix through careful design of the glass composition [110][110]. By careful control of the reaction condition, the transformation of the fluoride to oxyfluoride has been observed in a fluorosilicate glass [111],[112]. Due to their balanced material properties compared with single crystal fluoride and ceramics, fluoride TGCs doped with RE ions with high transparency are of particular interest for optical and photonic applications. Similarly, fluoride NCs precipitate not only from parent fluorosilicate glasses but also from other oxyfluoride glasses such as fluorophosphate and fluorogermanate ones. These glasses are mostly developed by the scientists from the optical materials community to reduce the phonon energy of the crystal/amorphous phase doped with RE ions. For instance, an Er 3+ doped fluorogermanate GC was prepared by Hu et al. from a parent glass of 45GeO 2 – 20BaF2 – 10AlF3 – 10Na2O – 5NaF – 8ZnO – 1GdF3 – 1ErF3 [113]. Upon thermal treatment of the parent glass at 600 C, NaBaAlF6 crystals were observed. Due to reduced phonon energy, these GCs demonstrated superior upconversion emission from Er 3+. 28

The preferable precipitation of metal halide crystals also occurs in fluorophosphate and chlorophosphate glasses as metal halides are ionic and they do not form strong bond to the glass network. As reported by Burdaev et al. [114], RE fluoride crystals with high concentration were precipitated from a parent glass of 0.05Ba(PO 3)2·0.95MgPb(Ba)CaSrAl(Ln) 2F14. The precipitation of metal chloride in an oxidebased glass was achieved by Hatefi using a parent glass of 45P 2O5 – 14Na2HPO4 – 25CaCl 2 – 15NaCl – 1Eu2O3 [115]. The formation of CaCl 2 crystals led to notable enhancement in both Stocks and antiStocks emissions from Eu 2+ ions, which were believed to enter into CaCl 2 lattice upon crystallization [115]. Apart from parent oxide glass, the precipitation of alkali (and alkali earth) metal chloride and bromide occurs in pure fluoride as well as metal chalcogenide glasses. In general, most oxide glasses, such as silicate, are not suitable hosts for metal chloride crystals possibly due to their large difference in melting temperature. Among all–fluoride glasses, fluorozirconate glasses exhibit a wide–glass forming range and have been extensively explored for application as optical materials. These fluoride glasses often have a limited solubility for metal chlorides and bromides. For instance, barium chloride NCs can be precipitated from a parent glass of 48ZrF 4 – 10BaF2 – 10BaCl 2 – 20NaCl – 3.5LaF3 – 3AlF3 – 0.5InF3 – 5EuF2 at annealing temperatures between 260 C and 290 C. Likewise, BaBr2 NCs can be crystallized from the parent glass of 52ZrF 4 – 20BaF2 – 5NaF – 15NaBr – 3AlF3 – 1.5LaF3 – 1.5YF3 – 1InF3 – 1EuF2 [116]. In a similar fashion, metal chalcogenide glasses with low melting temperature can dissolve small amount of metal chloride, which, upon heat treatment, can be precipitated from the glass host, resulting in the improved mechanical strength [117],[118].

3.2.2 Non-oxide Semiconductor NCs Before the invention of colloidal QDs, semiconductor NCs doped glasses were already extensively employed for studying the quantum size effect in the 1980s. Unlike alkali metal halide, 29

metal halide semiconductor NPs, such as CuCl, can be precipitated and stabilized in mo st oxide glasses, e.g., silicates [119],[120]. From reported examples, the precipitation of such metal halide QDs occurs easily in most glass matrix; ordinary silicate glasses can serve as stable and robust hosts to accommodate up to 1 wt.% of Cu–halide. Generally, the crystallization of Cu–halide occurs at a moderate temperature range far below the bulk crystallization temperature of the matrix glass [120]. Since the 1980s, metal chalcogenide QD doped oxide glasses have been developed for the studying the optical properties of QDs as the control of QD size is fairly easy by changing the heat treatment condition [121],[122],[123],[124],[125],[126]. Similar to the Cu–halide TGCs, silicate glasses can be a stable matrix for metal chalcogenide QDs. The precipitation of QDs can be initiated at temperatures below Tg of the parent glass. For instance, PbS can be precipitated from silicate glasses by annealing at 500 – 575 C [125]. In comparison, the crystallization temperature of ZnS and ZnSe increases by around 100 C in spite of the smaller cation ionic size. The higher crystallization temperature is related to the stronger covalent bond strength in the Zn–chalcogenides, as compared to that of the lead compounds.

3.2.3 Metal nanoparticles Metal NP doped glasses represent one of the earliest forms of GCs used by human beings. Noble metals, as they are noble, cannot form stable bond to the glass network and they usually form clusters easily upon thermal treatment. A small amount of gold or silver salts (<1%) can easily dissolve in silicate glass melt without changing the glass forming ability. Depending on the concentration of noble metal, the as–melted glass can be colorless if the noble metals are in their ionic form. Unlike oxide precipitation from oxide glasses, there are almost unlimited choices of glass compositions for dissolution and precipitation of noble metal NPs. Less noble metal, such as copper, has also been precipitated from oxide glass with carefully selected compositions [128]. Noble metal NCs have also precipitated in chalcogenide glasses which normally have a low melting 30

temperature. The doping of noble metal NPs, such as Ag and Au, into a chalcogenide glass host can be conducted by heating. The presence of noble metal NCs in these glasses contributes to the dramatic change in both linear and nonlinear optical absorption [129], [130].

3.3 Beyond type–I and type–II: Elemental NCs formed by reducing the glass network Unlike noble metal NCs, the formation of non-metallic elemental (e.g., Si) NPs in an oxide glass host is difficult as they form strong covalent bonds in the glass network. Similar to the solution synthesis of metal NCs, the precipitation of non-metallic NPs such as Si requires the presence of strong reducing agent dissolved in the glass network. Lin et al. developed a general method for the preparation of elemental NCs doped oxide glass by a careful control over the redox chemistry of the glass using solvated atomic Al (added to glass melt) as the reduction agent (see Table 3) [131],[132]. Elemental particles are generated by thermal treatment of the as-made glass at high temperature, where the atomic Al is immobilized and serves as the reducing agent. This methodology can be applied for the precipitation of not only elemental metal, which serves as network modifier, but also Si and Ge, which are essential building element of the glass network [131],[132]. A detailed discussion about the mechanisms of the formation of crystalline elemental particles by this process is given in section 5.1.

Table 3 Selected examples of typeII GCs in which glass network formers do not participate in the crystallization process. Crystal composition Metal ZnO oxide Ga2O3 MgAl2O4 (spinel) LiGa5O8:TM (TM: Cr3+, Co2+, Ni2+) Y3Al5O12

Glass composition* 30~50SiO2–12~25% Al2O3–15~40%ZnO–12~25% K2O (by weight) 6.37(K2O)–12.47(Ga 2O3)–81.13SiO2 SiO2–Al2O3–ZnO–MgO –TiO2–ZrO2

Annealing condition 650–1100 °C

Ref. [76]

900 °C >900 °C

[80] [82]

Li2O–Ga2O3–SiO2

680 °C

[85],[86],[87]

CaO–Y2O3–Al2O3–SiO2

1300–1500 °C

[88],[89] 31

BaFe12O19 LiTaO3 NaNbO3 KNbO3 (co–exist with other phases) –BBO (BaB2O4)

Fluoride

Chloride/B romide

BaB4O7 BiBO3 KTiOPO4 BaTiO3 PbxCd1–xF2:Yb, Er CaF2 LaF3 NaYF4:Yb, Er BaYF5 YOF:Yb,Er; LaOF:Er BaCl2 BaBr2 CuCl

Chalcogeni de QDs

CuBr ZnS ZnSe ZnTe CdS1xSex CdTe

Elemental particle

PbS PbSe Au Cu / Ag Pb Si Ge Bi

or Y2O3–Al2O3–SiO2 40BaO– 27Fe2O3–33B2O3 10~77.5 SiO2–2.2~55Al2O3 –10~70LiTaO3 14SiO2 –70 Nb2O5 – 16Na 2O–0~31BaO (by weight) K2O–Nb2O5–SiO2 40BaO – 15TiO2 – 45B2O3

5BaO–15B2O3–70TeO2 Bi2O3–B2O3 K2O–TiO2–P2O5–SiO2 20~25SiO2–6~20Al2O3–45BaOTiO2 30SiO2–15AlO1.5–24PbF2– 20CdF2–10YbF3–1ErF3 50SiO2–20Al2O3–30CaF2 40SiO2–30Al2O3–18Na2O–12LaF3 64SiO2–14.5B2O3–9.5Na2O–6NaF–6YF3–0.5ErF3–1 YbF3 45SiO2–15Al2O3–10Na2O–24BaF2–6Y2O3 65SiO2–15B2O3–14Na2O–4.5YF3–1YbF3–0.5ErF3 63SiO2–15B2O3–16Na2O–5.5LaF3–0.5ErF3 b) 48ZrF4–10BaF2–10BaCl2–20NaCl– 3.5LaF3–3AlF3–0.5InF3–5EuF2 52ZrF4–20BaF2–5NaF–15NaBr–3AlF3–1.5LaF3–1.5 YF3–1InF3–1EuF2, 10Na2O–52.5B2O3–7.5Al2O3–30SiO2–0.5CuCl–0.25 SnO 0.5CuCl–0.25SnO 64SiO2–32B2O3–1.8K2O–0.6Sb2O3–0.6CuBr 60SiO2–10B2O3–20Na2O–8ZnO–2ZnS 50SiO2–20Na2O–5Al2O3–21ZnO–4ZnSe (60-x) ZnO–xTeO2–40P2O5(x =0, 10, 20, 30, 40, 50 mol%) 68.9SiO2–5.6Na2O–11.5B2O3–1.0Al2O3–11.3ZnO–0. 59CdO–1.1Sb2O3 CdO, CdS, S, Se 47.66SiO2–30.52Na 2O–16.55B2O3–5.22ZnO+Te+Cd O 66SiO2–8B2O3–18K2O–4BaO–4ZnO–1.5PbS P2O5–Ga2O3–ZnO–AlF3–Na2O–1.5wt%PbSe 70SiO2–20Na2O–10CaO–0.1Au2O 50BaO–50P2O5 or 50CaO–50P2O5 +(4–10)Cu2O/Ag2O 10Na2O–40B2O3–40SiO2–6PbO–4Zn Si: 10Na 2O–48B2O3–2Al–40SiO2 Ge: 10Na 2O–48B2O3–2Al–40SiO2–4GeO2 Bi: 10Na 2O–48B2O3–2Al–40SiO2–4Bi2O3

820 °C / 4–24 h 800 and 1050 °C 750 – 950 °C 800 °C

[90] [91],[92],[93] [94] [95]

850 °C (surface initiated crystallization with ultrasonic activation) 390 °C 454°C

[96],[97]

600–1100 °C 470 °C

[98] [99] [100] [101] [106]

570–660 °C 450 °C /2h

[107] [108] [109]

615–665 °C 610–670 °C 615–635 °C b)260–290 °C

[110] [111], [112]

290 °C

[116]

460–535 °C

[119]

ZnS: 570–610 °C ZnSe: 630–650 °C Formed after melting 575–750 °C

[116]

[120] [121] [122] [123]

450 °C / 3h

[124]

500–575 °C 395–430 °C 550 °C 490 °C

[125] [126] [127] [128]

450–530 °C 450–525 °C; 450–500 °C 460–600 °C

[131] [132]

*Compositions of the listed glasses are given in molar fraction (mol.%) unless otherwise stated.

32

4.Properties and characterization methods

4.1 General requirements for optical transparency in GCs Light attenuation, when passing through GCs, is caused by atomic absorption and scattering as described by: (12) where IT and I0 are the intensity of the transmitted and the incident light, respectively, α and σ are the absorption coefficient and the turbidity of the sample, respectively, and

is the sample

thickness. Practically, absorption loss can be easily reduced by using purified raw materials. Therefore, in order to preserve high transparency of the glass, the scattering loss (i. e., the turbidity) has to be minimized. For GCs containing crystals with sizes much smaller than the wavelength of light, the crystals are widely separated (for diluted systems). Rayleigh–Gans scattering model gives the turbidity as [133]: (13) where N is the number density of the crystallites, R and V are the radius and volume of the crystal phase, respectively, n is the refraction index of the crystal and n is the refractive index difference between the glassy and the crystal phase, and k = 2/, where  is the wavelength of light. From the equation above, it is clear that the turbidity is inversely proportional to 4, which is known as Rayleigh scattering. In addition, the turbidity is extremely sensitive to the size of the particles ( R6) as well as the difference in refractive index of the (n2). Therefore, there are two general strategies to reduce the scattering loss in GCs materials. First, both the nucleation and crystal growth rates should be carefully controlled to precipitate NCs and to avoid overgrowth and 33

coarsening of crystals. Second, the refractive index of the crystal phase should be very close to that of the glassy phase and the crystals with large birefringence should be avoided in TGCs. This condition is easy to meet for GCs in which glassy phase and crystalline phase have nearly same composition (see section 3.1.3). Practically, highly transparent TGCs (in the visible range) without noticeable scattering effect can be obtained for crystal radius R < 15 nm and refractive index difference n < 0.1. This condition does not have to be stringently obeyed for a small n. For instance, crystals of micrometer scale can exist in a TGC based on BTS glass (see section 3.1.3), in which n ~0.01. It is known that the Rayleigh–Gans model overestimates the light attenuation due to scattering for concentrated particle dispersions. A quasi–continuum model has been developed independently by Andreev [134] and Hopper [135] for the description of interfering particles that produce large compensation effect. In this model, in addition to the requirement that particles sizes are far smaller than wavelength of light, the distance between the particles should be comparable to the particle sizes. The turbidity under such condition is given by: (14) where  = a + W/2 is the mean phase width (W: the inter-particle distance). According to this model, high transparency can be still achieved for crystal size up to 30 nm and refractive index difference (

) up to 0.3. However, the above model neglects multiple scattering that may become significant when the

thickness of the sample is comparable to the inverse turbidity. According to a recent model, GC can be described as a later stage phase separated structure. In this case, the turbidity can be expressed as [136]: (15)

34

where  is the volume fraction of the crystalline phase and R is crystal radius. For an oxyfluoride GC containing 0.3 fraction of fluoride NCs with a size of 15 nm (n=1.7, and n=0.1), the above theory predicts a turbidity of 0.2 m 1, which is closest to experiment. In comparison, the Mie theory predicts a turbidity of 200 m 1 and eq. 16 finds a turbidity of 7 m 1. For scattering by particles with size comparable to the wavelength of light, all of the above models break down. High transparency can be only preserved for n~0 in congruent GCs. The theory for describing the scattering of spherical particles with any size has been developed by Mie [137]. Analytic expression for the turbidity cannot be derived for the general scattering process, while the size (as compared to the light wavelength) and n still determine the magnitude of turbidity.

4.2 Mechanical, thermal, electrical and chemical properties The most striking property of GCs is the greatly enhanced mechanical strength as compared to parent glass and this was already realized during invention of GCs in the 1950s. The crystals formed in the glassy matrix can serve as the impediment for stopping the spreading of cracks during fracture, thus the brittle nature of glass can be overcome. For silicate GCs, it is now possible to achieve a flexural strength up to 500 MPs, which is much higher than that of most oxide ceramics. In addition, the toughness of GC, which is characterized by the critical stress intensity (KIC), can be as high as 3 MPam0.5 [7]. Due to their brittleness, glass is not machinable by common techniques. In comparison, GCs with certain compositions (e.g., GC based on mica) are machinable. That is, they are amenable to techniques such as drilling, milling, grinding and sawing. The machinability of GCs endows them easy accessibility to different pre–defined shapes with controlled surface roughness. Glasses are usually vulnerable to thermal shock because of its larger coefficient of thermal expansion (CTE) and brittleness. Partial crystallization in GCs can be used to modulate the CTE, as 35

the crystallized phases usually have much lower CTEs compared to the parent glasses. By properly choosing the glass systems in which the precipitated crystals have small or even negative CTE, an overall CTE of the GCs can be reduced down to zero. Due to the reduced CTE, these types of GCs exhibit high resistance to thermal shock, which is attractive for applications in cooking ware s and fireretardant windows. Moreover, the small or nearzero CTE are especially important for large size lens used in observatories. Depending on where GCs are applied, special requirements on the electrical and magnetic properties have to be met for GC production and design. For instance, when applied as insulating materials used in electronics and microelectronics industries, the electrical properties of the GCs will be of primary importance. In general, oxide GCs that are made up of light elements are good insulators, which often demonstrate low dielectric constant. The small dielectric loss is particularly important for applications in Radomes and relevant fields. Chemical properties, such as chemical durability and resorbability, of GCs are strongly dependent on the composition, and therefore they can be carefully controlled by varying composition. SiO2 itself exhibits high resistance to corrosion by acid (except HF), while other components, such as metal oxides, can be easily leached by acidic solution [138]. In alkali solutions, SiO2 can also be dissolved and most silicate glasses are not chemically durable, but the alkali resistance can be drastically improved by introducing alkali earth ions [139][140]. Compared with the parent glasses, GCs do not always show enhanced chemical durability, which may be closely related to the composition as well as the specific type of precipitated crystals.

4.3 Characterization techniques for GCs To understand the GCs and the crystallization process, different techniques have been applied for the examination of the microstructures, chemical compositions and optical properties of GC.

36

Structural characterization tools that used in material science, such as diffraction methods and electron microscopy, are widely applied for the examination of GCs.

4.3.1 Xray diffraction Glass is usually characterized by X–ray or neutron diffraction techniques. Using these techniques, we always observe an amorphous hump, instead of the Bragg peaks, and the central position of the hump is determined by the average inter–atomic distance. Crystals precipitated in glass lead to the appearance of diffraction pattern in addition to the amorphous background. From the diffraction pattern, not only the type of the crystallized phase, but also its crystal size as well as its volume fraction can be determined. The crystalized phases can be identified using computer database of the powder diffraction files. Due to the multi–component nature of most inorganic glasses, multiple crystalline phases are often precipitated possibly as solid solutions, which may hinder distinguish different crystallized phases. For instance, in many silicate glass systems, the formation of different types of quartz–related solid solutions containing Al 2O3 and other metal oxide is often resulted from devitrification. In this case, the use of the multinary phase diagram combined with quantitative XRD analysis (e.g., Rietveld method) is highly recommended in revealing the type and fractions of different phases. The peak width of XRD is connected with instrument factors as well as the sample characteristics. By excluding the instrument contribution, two factors associated with sample character are known to contribute to the peak broadening: strain ( ) and crystal size (d). The strain related peak width can be expressed by [141]: (16)

37

where B(2) is peak width at 2 . The derivation of this equation is based on several assumptions related with the crystal shape and orientation. For size related broadening, we have the well-known Scherrer equation [142]. (17) where L is the crystal size and k is a constant (k = 0.94 for FWHM of spherical crystals with cubic symmetry). Note that the equation is only valid when the size of the crystals is in the range of 5–1000 nm; use of this equation out of this range can lead to significant mistake. Another important feature of the GC, which can be determined from its XRD pattern, is the presence of preferred orientation of the precipitated crystals, which is often observed in GCs containing highly polar crystalline phase or surfaces crystallized GCs. The preferred orientation of the crystals along certain [h k l] directions can be directly observed by XRD from the abrupt large peak intensity. In a semi–qualitative manner, the degree of preferred orientation can be defined by the intensity ratio as P = I h1 k1 l1 / I h2 k2 l2, where the I h1 k1 l1 is the peak intensity of the plane exhibiting preferred orientation and I h2 k2 l2 is the strongest peak intensity for samples without preferred orientation. 4.3.2 Electron microscopy, electron energy loss spectroscopy (EELS) and scanning probe microscopy (SPM) The microstructure of the GCs is often examined using different types of electron microscopes, such as scanning electron microscope (SEM) and transmission electron microscope (TEM). SEM, for instance, can image the texture of the GCs from micrometer to nanometer scale depending on the resolution of the instruments. When combined with EDX (energy dispersive X–ray spectroscopy), SEM can provide information on the composition difference between the crystal phase and amorphous phase of the GCs. TEM provides higher magnification, which enables observation of NCs with size down to a few nanometers. The combination of TEM with electron 38

diffraction (ED) allows for the structure analysis of precipitated crystals with nanometer resolution. Furthermore, electron energy loss spectroscopy (EELS) has been widely implemented in modern HRTEM systems to assist the analysis of chemical characteristics with high spatial resolution. In an EELS measurement, the samples are exposed to an electron beam of known kinetic energy and the electrons that experience inelastic scattering and energy loss are analysed. The energy loss is connected with the chemical character of the examined materials, such as elemental composition, oxidation state and the type of bonding. Compared with EDX, EELS is especially useful for analysing light elements (such as H and C), and for detecting the subtle difference in chemical bonding of specific element between the glass and the crystalline phase. Currently, direct observation of glass structure at the highest resolution down to the sub-angstrom level can be realized by modern aberration-corrected (ac-) HRTEM as well as SPM (scanning probe microscope) [143][144][145][146][147]. Recently, Huang and coworkers carried out a direct detection of the short- and medium range order structure of a two-dimensional silica glass deposited on graphene by the ac-HRTEM [145], which verified the Zachariasen’s random network model. Moreover, the same authors showed that ac-HRTEM could image the atomic rearrangements in a silica glass upon excitation and therefore allowed for direct observation of individual ring rearrangements associated with strain, the dynamics of vacancies and phase transitions. Beside ac-HRTEM, a scanning tunnelling microscopy (STM) observation performed by Lichtenstein et al. presents unprecedentedly high resolution images on the interface between the crystalline and the amorphous phase in a 2D silica glass (Fig. 6) [146]. Their observation revealed a clear decrease of the degree of crystalline order within a thickness of only 1.6 nm when crossing the crystalline-amorphous boundary [146]. If applied to the above multi-component GCs beyond silica, these state-of-the-art techniques are anticipated to provide a general but deeper understanding at the atomic level on the nucleation and formation of crystals in the glass phase.

39

Fig. 6 The crystalline-amorphous interface of a 2D silica film at atomic resolution. (a) STM image and (b) STM image with superimposed atomic model of the top layer. Green and red balls represent silicon and oxygen atoms, respectively. (c) Structure represented by different sized rings (from 4 to 9 Si atoms per ring). (d) Ring size distribution across the crystalline-amorphous interface. (e) The number of rings per slice along the lateral coordinates. (f) Evolution of crystallinity across the crystalline-amorphous interface along the lateral coordinates. The average crystallinity for amorphous region is 0.42. Reprinted with permission from ref. [146], copyright 2012, American Physical Society.

4.3.3 Smallangle neutron / X-ray scattering Small angle neutron scattering (SANS) uses elastic neutron scattering at small angles to examine the microstructure of different solids or liquid samples with a characteristic scale of 1  40

100 nm. Compared with small angle X–ray scattering (SAXS) technique, SANS in general show higher sensitivity to light elements because neutrons are scattered by the nucleus. However, the limited availability of neutron sources compared with Xray strongly restricts the wide application of SANS. The gyration radius of a sample that can be known from a SANS (or SAXS) measurement is not radius of particles, yet it can be considered as the characteristic length of the nanoscale phase present in the sample. The first SANS on a GC was conducted by Loshmanov et al. in the 1970s on a silica glass containing TiO 2 NCs [148],[149]. The fluctuation of the atomic density due to the negative nuclear scattering of Ti greatly enhanced the sensitivity of SANS for this GC. Compared with XRD which is sensitive to the presence of crystalline phase, SANS can probe different type of composition fluctuation, such as phase separation in nanoscale. In a typical SANS (SAXS) spectra, not only the size of the nanophase (can be either crystalline or amorphous), but also the inter–particle distance as well as its volume fraction can be derived. Since SANS can examine liquid samples, this technique have been used to track the structural evolution around the glass transition temperature (Tg) and the crystallization process. From an in–situ experiment, the dynamics of phase separation and nucleation and growth of the nanophase can be examined. For instance, the structural evolution with time and temperature can be studied by the changes of spectral shape and peak positions.

4.3.4 Inelastic Light–scattering Elastic light scattering techniques are widely used in the determination of particle size distribution for liquid colloidal systems. In comparison, inelastic light scattering (ILS) has been used to examine the nanoscale inhomogeneities in glassy materials. In the ILS spectra, an asymmetric broad band known as the boson peak can be often observed in the low frequency region. Although the exact origin of the Boson peak is still in debate [150], it is generally accepted that the boson peaks can be connected with the excess vibration density of state (DoS) arising from 41

the nanoscale structural fluctuation due to the density inhomogeneity [151]. The frequency of the Boson peak is described by: (18) where

is the tranverse sound velocity, c is the light velocity, and (G: shear modulus, d: density), we can easily find

is the correlation radius. As . At tempertures slight

higher than Tg, G drops rapidly due to the transition to a low viscous state. For GC materials, however, the formation of the crystalline phase enhances the elasticity due to the reduction in specific volume. That is, the crystallization of the GC leads to the increase in G and the sound velocity. Takahashi et al. performed an in–situ ILS measurement around Tg for three different types dropped at temperatures approaching 1.0 –

of oxide glasses [151], [152]. It was found that the

1.1 Tg as the glass changes to a supercooled liquid state accompanied by ralaxation. With a further increase in temperature (1.11.2 Tg),

increased rapidly due to crystal growth in the

glassy matrix. A recent work performed on various silicate glass by Fujiwara and coworkers suggested that the introduction of small/medium sized alkali (earth) ions led to an increase in

,

while the tendecncy was reversed for lager ions [153].

4.3.5 Raman spectroscopy Raman spectroscopy is a technique used to determine vibrational modes in a system, which provide structural information of the system at a molecular level. For glass materials, groups such as SiO42– anions, gives typical vibrational peaks in the Raman spectrum, and the Raman spectrum also enables the distinguishing crystalline phase from the amorphous phase due to their local structural differences. Practically, Raman spectroscopic measurement is always carried out with a microscope such that the structural information with resolution down to a few micrometres can be revealed. This technique therefore becomes one of the few tools for studyi ng sample 42

microstructures for which XRD and EM are not suitable. For instance, the laser structured pattern or spots on the surface or inside glasses of micrometer scale can be readily examined by Raman spectrum. With the assistance of the computer controlled scanning stage, Raman mapping using characteristic Raman frequencies of the sample can be measured, which provides structural information that cannot be achieved by other mapping tools, such as EDX [154].

4.3.6 Solid state nuclear magnetic resonance Solid state nuclear magnetic resonance (NMR) has been frequently used to characterize the structure of glasses as well as ceramic materials. Unlike XRD and EM, solid state NMR is able to examine local structure around specific atoms and therefore allows for finding the chemical and geometrical differences in short range order between the amorphous and crystalline phases. A number of elements can be detected by NMR, and

29

Si, 31P and 27Al are among the most widely

examined elements due to their large abundance in many oxide glasses. In silicate glasses, for instance, the chemical shift of 29Si signal changes by 10 ppm with the conversion of non–bridging oxygen to a bridging one in the crystallized region [155],[156]. In addition, the 29Si spectrum also enables the identification of the various Si–O units by the chemical shift: –Q4 units Si(–OSi)4 (~107 ppm) and Si(–OAl)1(–OSi)3 (~97 ppm);–Q3 units Si 2O5–2 (~91 ppm). Therefore, this technique allows for the clarification of structural transitions accompanied by crystallization process [157]. Moreover, NMR can probe small structural differences, e.g. regarding bond length and angle, in short range (0.3–0.5 nm) between crystalline and amorphous phases, which cannot be explored by most other tools. Generally, the disorder in glass phase causes the broadening of the NMR peak as compared to that of the crystalline materials [158]. This very small structural change at short range order often occurs at the beginning of the crystallization process, i.e., the nucleation, for which only NMR provide a clearer physical picture compared to other techniques [39].

43

Besides the standard (one–dimensional) technique that can provide information on short range order, two–dimensional (2D) NMR can be used to examine the local structural order as well as the connectivity of a specific atom or polyhedron. The sequences for 2D NMR are based on dipolar interactions as well as scalar interactions, which can provide the spatial proximity and chemical connectivity of polyhedra. Hetero- and homo-nuclear based sequences have been used to investigate the interactions between polyhedrons that are built up from same or different elements [159],[160],[161],[162],[163],[164],[165],[166]. 4.3.7 EXAFS / XANES X-ray absorption spectroscopy (XAS) has been regarded as a powerful tool in the determination of the local geometric and electronic structures of a specific atom in both organic and inorganic materials. The spectrum of XAS can be divided into three different regions: 1) pre -edge region, which is determined by the transitions of electrons to the lowest unoccupied levels; 2) the XANES (X-ray absorption near edge structure) region, which is primarily contributed from transitions involving core levels to quasi-bound states that occur within 50 eV from the absorption edge; 3) the EXAFS (extended X-ray absorption fine structure) region that occurs at energies > 50 eV (generally in the 150 – 2000 eV range) from the absorption edge. Both of the XANES and EXAFS are element specific and can be used to examine the chemical environment around certain atoms at a very low concentration. XANES provide information on oxidation state, coordination number and the small geometrical distortion of a specific element, while EXAFS usually tells local structural information of a specific atom, such as type and number of the coordination atom, bond length, pair distribution function, etc [167]. Both of these techniques are widely applied for the examination of the local structures of glasses and GCs as well as the transition from glass to GC; their applications are only hampered by the limited availability of synchrotron light sources.

44

4.3.8 Thermal analysis Thermal analysis, including differential thermal analysis (DTA) and differential scanning calorimetry (DSC), have been widely used for the examination of the glass transition as well as the crystallization in glassy materials. The glass transition in glass is a typical endothermic event, which is characterized by the onset temperature of the peak, i.e., the calorimetric glass transition temperature (Tg) depending on the cooling rate. The standard Tg is defined as the onset temperature of the endothermic peak determined during the second up-scan at the rate of 10 K/min equal to the prior cooling rate [168]. In contrast, the crystallization is an exothermic event, which is characterized by the onset temperature (Tc) of the exothermic peak. When temperature exceeds its Tg, a glass will gradually become a viscous supercooled liquid and thereby the mobility of the atomic or molecular species are greatly enhanced compared to that at the lower temperatures (below Tg). For poor glass formers, when temperature rises to a certain value above Tg, crystals will start to form, inducing an exothermic peak. The difference between Tg and Tc is a measure of glass stability against crystallization and hence of glass forming ability [169],[170]. The smaller the Tg and Tc difference of a glass has, the poorer its glass forming ability is. DSC is a powerful method to quantify the glass forming ability, and is helpful for designing the composition of GCs. For certain types of glass compositions, multiple crystallization peaks can be observed at temperatures above Tg due to the formation of different crystals and the reaction between different crystalline phases. For instance, in oxyfluoride glasses, fluoride crystals are usually precipitated first, followed by the formation of oxide crystals. By combining with phase analysis tools like XRD, it is possible to identify the crystallization temperatures for each phase formed at different stages. Some glass systems exhibit even multiple glass transition peaks, which are a clear signature of phase separations. For example, three Tg values and hence three phases were identified in the ionomer glass (a kind of fluoro–phospho–alumino–silicate system) [171][172] [173]. One of the phases is the Ca–F–P nanophase (an amorphous phase rich in Ca, F and P) at a scale of 45

approximately 5 nm, which could lead to the formation of NCs. In literature, the DSC proves to be an effective tool for determining and monitoring the phase separation and crystallization behavior, and for providing insight into the mechanism of crystal formation. Therefore it can also be used to assist the designing and tailoring of NCs in glass. For GC synthesis, DSC or DTA are a necessary tool to design of the temperature profiles for isothermal treatment or dynamic heating.

5. Synthesis of TGCs: Conventional and unconventional techniques

5.1 Crystal precipitation by confined solid state reaction GCs are the solid state version of colloidal systems, and hence the fabrication of GCs can be regarded as the synthesis of crystal in supercooled glass via reactions in a confined manner in the solid state. Similar to the crystal precipitation in solution system, the formation of crystals in the glassy matrix is the product of reaction between the glass components yet in the solid state. Analogously, the solid-state reaction in a super-cooled liquid (glass) can also be triggered by raising the temperature, that is, via heat treatment, which is the most popular fabrication technique for GCs. Moreover, the crystallization of glass in most cases does not follow a thermodynamically favourable route due to kinetic factors, such as diffusion. This is especially important in the fabrication of GC with high transparency, which has usually to be performed in a controllable manner to allow the precipitation of the desired crystals with proper size and fraction. 5.1.1 Controlled nucleation and growth by isothermal treatment The fabrication of TGCs with the precipitation of specific crystals is not easy, as the process of crystallization could involve the simultaneous formation of different crystal phases especially for multicomponent parent glasses. Therefore, in order to find the right conditions for TGC fabrication, it is strongly advised to perform a thermal analysis of the parent glass using DSC 46

measurement to locate the glass transition (Tg) and crystallization (Tc) temperatures. As shown in Fig. 2, nucleation and crystal growth are two separated processes that reach their maximum rates at different temperatures. For a better control of the crystal growth especially for TGCs, a two stage isothermal treatment at two different temperatures is more often applied so as to separate the nucleation and growth. Crystal growth with only few numbers of nucleuses can result in the formation of larger grains, leading to loss of transparency. TGCs can be only obtained in the presence of sufficient number of nucleus so that crystal growth ends up with crystallites of limited sizes, preferably less than 20 nm. This condition can be loosened only when crystals have close refractive index to the parent glass, as discussed in section 4.1. Nucleation cannot be initiated for certain glasses even at temperatures of very low viscosity. This occurs for most type–I GCs, in which glass network formers, are involved in crystallization. Then, we can take advantage of heterogeneous nucleation by addition of nucleation agent, such as TiO2, ZrO2 and noble metal NPs. The presence of nucleation agent can initiate the crystallization of the glass at a significantly lower temperature. For instance, the first modern GCs are made based on heterogeneous nucleation using photoreduced Ag NPs as substrate. However, recent studies by Rüssel have shown that the effect of the transition metal (TM) oxide nucleation agents on nuleation strongly relied on the careful selection of concentration and glass system [174],[175],[176]. For instance, ZrO 2 has been found to be very effective in promoting nucleation in various systems, however, the addition of ZrO 2 of over 2% can eliminate volume crystallization in lithium disilicate glass completely. The inhibition of nucleation has been found for many oxides (e.g., Ta2O5, Nb2O5) in several glass systems, and the mechanisms have been explained by the change in viscosity, interfacial energy in the crystal-glass interface and the thermodynamic driving force in the presence of these TM oxides [174].

47

5.1.2. Controlled nucleation and growth by cooling Recently, controlled cooling has been found to be an alternative, effective approach to fabricate TGCs from properly designed phosphosilicate compositions [177],[178]. For a specific phosphosilicate composition, the size and fraction of NCs in TGCs and therefore its transparency can be tuned by varying cooling rate. This approach makes the additional isothermal treatment step unnecessary, and hence saves thermal energy and simultaneously obtains the functionalities of TGCs. For instance, a RE-doped phosphosilicate TGCs was obtained by the melt–quenching followed by devitrification during cooling. This Yb3+/Er3+ doped TGC exhibits highly efficient upconversion luminescence of Er 3+ at 521, 545 and 655 nm under 980 nm excitation. It was found that the type of alkali oxides had a strong impact on the formation of NCs in phosphosilicate melts during cooling [179],[180]. The Yb3+/Er3+ ions promote the Zn 2SiO4 crystal formation, but suppress the Na3PO4 and AlPO 4 formation during cooling. The non–isothermal melt–crystallization kinetics can be well described by the Avrami model [181]. The activation energy of crystallization in both the undoped and the Yb3+/Er3+ codoped samples during cooling can be determined using the differential iso–conversional method of Friedman [182]. The results indicate that the activation energy decreases with crystallinity during cooling, and the additions of Yb2O3 and Er2O3 leads to an increase in energy barrier of crystallization upon cooling.

5.1.3 Differences between typeI and typeII GC formation by solid state reaction Compared with type-I GCs, typeII GCs can be more easily designed as the precipitation of the crystalline phase such as chalcogenide QDs and fluorides is less dependent on the composition of the parent glasses. In general, type-II GCs containing crystal phase A can be fabricated by melting A with a parent glass B and then performing thermal treatment to precipitate A from glass B. This is indeed a “dissolution” and “precipitation” process that is widely employed in wet-chemistry synthesis. For instance, QDs doped GCs are always fabricated by the co-melting of a “precursor” 48

(such as metal chalcogenide) with an oxide glass (such as silicate glass) and subsequently by the thermal treatment (see examples in 3.2). In comparison, such a simple strategy cannot be applied to type-I GCs, and the crystal precipitation in type-I GCs remains somewhat accidental. Use of the phase diagram and HSAB theory can aid the design of such GCs, while the prediction of the crystallized phase for a glass under certain conditions (thermal treatment) remains somewhat illusive. Upon thermal treatment, the crystallization behaviours of typeI and typeII GCs are apparently different mainly due to the difference in the chemical composition of the parent glasses. In typeI GCs where crystals are usually generated by partial crystallization of the network former, the formation of crystals are the result of solid state reaction between covalent network formers (i.e. silica) and network modifiers which are usually ionic compounds (i.e., alkali metal oxides). Some of the ionic metal oxides exhibit higher mobility and are highly reactive at certain conditions, leading to the formation of metal silicate crystals. Metal oxides that do not react with the silica network at the heat treatment condition do not contribute to devitrification. Conventional route for the fabrication of type-I GCs usually requires a relatively high thermal treatment temperature, while the use of nucleation agent like TiO 2 can reduce the crystallization temperature by facilitating heterogeneous nucleation. In comparison, in the formation of type-II GCs the network modifiers (mostly metal oxide and halide) themselves can react and form crystalline phases without disturbing the covalent glass network. Many simple metal oxides and metal halides based GCs belong to this family. The reason for such type of crystal precipitation is closely connected to the chemical nature of the glass component. Apparently, the glass network formers, such as silica, only serves as the “solvent” and it is chemically inert and stable against crystallization compared to the more mobile, reactive, and thermodynamically immiscible network modifiers (such as metal halides), which preferably precipitate under thermal treatment conditions. From the chemical point of view, this “solvent” is strongly covalent, and hence, is very different from the ionic salt melt 49

which can also serves as the solvent for high temperature liquid phase synthesis [183]. Finally, all glasses can be completely devitrified and therefore a transition from typeII to typeI GC can occur during continuous devitrification of a glass under suitable thermal treatment conditions.

5.1.4 Predicting crystallization process It is not easy to perform a precise prediction on the crystallization process as well as phase separation for a given glass with complex composition based on either empirical rules or simulations based on MD or DFT. In inorganic chemistry, the HASB theory has been widely applied for the understanding precipitation in solution phases as well as solid-state reactions between inorganic substances [42]. In the HSAB theory, the interactions follow a simple rule: ‘‘hard, non-polarizable prefers hard, while soft, polarizable prefers soft’’. Application this theory to the devitrification process of glasses is successful in several systems, especially for synthesizing type-II GCs. For instance, the predictions based on HASB theory are quite reasonable for the crystallization of fluoride NCs from an oxyfluoride glass hosts as well as the precipitation of metal oxide NCs (such as the spinel phase) from silicate glasses. The problem with this theory is that only thermodynamically favored phases are predicted without considering the kinetic factors. Phase diagrams can be often used to forecast the crystallization pathways and to assist designing and fabrication TGCs in a predictable manner especially for type-I GCs. In general, precipitation involving the glass network formers can occur as long as there are compounds formed at certain temperature and compositions. When applying phase diagram in predicting GC process, one has to bear in mind that each phase in the phase diagram are thermodynamically stable at that temperature, and the compounds predicted in the phase diagram do not necessary form in the crystallization of the glass of the same composition due to kinetic factors. Therefore, it could be sometimes misleading to fabricate a GC with targeted crystalline phase by choosing the composition and heat treatment temperature only according to the phase diagram. 50

Temperature

Tm

Tn

tn

Time

Fig. 7 Schematic representation of TTT (timetemperaturetransformation) diagram. The TTT diagram can be used to assist the design of the heat treatment temperature and duration for a desired crystalline phase. Samples hold at temperature of Tn for duration of tn can reach transformation defined by this curve. As discussed above, the major drawback of using phase diagrams to predict the crystallization of a glass is that only thermodynamically favored phases appear in a phase diagram. In the crystallization of a glass, the kinetic factors usually determine the type of formed crystals, which are metastable. The time–temperature–transformation diagram (TTT) can be used as an alternative to phase diagram for the prediction of the crystallization of a glass system (Fig. 7). A schematic TTT diagram is shown in Fig. 7. In the TTT diagram, each coordinate in the curve is defined by time and temperature. Therefore, TTT diagram can be used to assist the designing of the thermal treatment conditions for the precipitation of the desired crystal phase for a given glass sample, as long as a pre–determined TTT diagram is available. However, it should be noted that it is still a challenge to experimentally draw a full TTT diagram for many glass systems [184].

5.1.5 Manipulation redox equilibrium for elemental particle precipitation In wet chemistry synthesis, elemental particles, such as noble metals, can be formed by thermal decomposition of unstable precursors or reduction in the presence of reducing agent. The same methodology can be used to get elemental particles precipitated in a glass matrix. Noble 51

metal, such as Au and Ag, can be precipitated in most oxide glasses through thermal treatment, as they cannot form stable bond to the glass network. The formation of atomic noble metal species, from precursor to metallic particles, begins at decomposition of the precursor salt (e.g. AgCl) during melting. At the melting temperature, the metal species are completely solvated and the following heat treatment assists the diffusion and clustering of the metal atoms to formed larger crystallites. A recent report showed that the precipitation of Ag in a soda-lime silicate glass began at a temperature below 410 C in the form of Ag-dimers [185]. Clustering and nucleation of noble metals could be triggered by thermal treatment as well as by laser irradiation. However, unlike wet-chemistry process, the formation of Ag NPs is not controlled by Ostwald ripening, but proceeds through the addition of Ag-dimers that immigrate in the glassy matrix [185]. Similar “decomposition” reactions occur in silicate glasses containing heavy metal oxides. Most silicate glasses melt at temperatures above 1400 C, at which heavy metal oxides become unstable and decompose into atomic species. For instance, silicate glasses doped with Bi 2O3 usually end up with a wine color [186],[187]. It has been accepted that this coloration can be assigned to Bi–related species and the followed thermal treatment at high temperatures can mobilize these Bi species, which then form clusters or larger Bi particles, leading to further darkening of the glasses [188]. Analogously, TeO 2 doped silicate glass, which melts at temperatures above 1400 C also shows a strong dark brown coloration arising from the precipitation of Te particles, which were observed by TEM as well as Raman spectroscopy [189]. The precipitation of elemental Te also occurred in a quaternary phosphate glass system of P 2O5-Al2O3-ZnO-TeO2 [190]. In this phosphate glass, the precipitation of Te NPs occurred during quenching of the glass melt, without need of a post-annealing process. The precipitated Te particles are crystalline with a central size of approximately 8 nm. The glasses exhibit a clear pink color, and this coloration becomes stronger with an increase in Te NPs concentration (Fig. 8). Based on optical absorption, glasses with a lower concentration of ZnO demonstrate a higher tendency for Te precipitation. Interestingly, these 52

glasses containing Te NPs show strong NIR luminescence centred at around 1100 nm, thus opening a new gate for photonic applications in lasers and optical amplifiers.

Fig. 8 TEM characterizations of Te precipitated glass. (a) Typical TEM image, (b) SEAD pattern, (c) HRTEM image of a single Te NC for glass with the composition of 55P2O5-25ZnO-20TeO2. (d) TEM image, (e, f) HRTEM images and (g) SEAD pattern for glass 50P2O5-5Al 2O3-25ZnO-20TeO2. (h, i) Typical TEM images for the as-melt glass and (j, k) after annealing at 480 C for 2 h for glass 66P2O5-5Al 2O3-25ZnO-4TeO2. Reprinted with permission from ref. [190], copyright 2012, Wiley-VCH. Metallic NPs are always synthesized by the reduction of their salts in solution phase. In a quite similar fashion, such a redox reaction can be performed in a glassy matrix, where metal Al is added as the reducing agent to precipitate Si and Ge from an oxide matrix (Fig. 9a–d) [131], [132]. At the glass melting temperature, the metallic Al is dissolved completely in the glass network, resulting in a homogeneous “solution” at high temperature, which can be quenched to form a colorless transparent glass. During melting in air, care has to be taken to avoid the complete oxidation of Al 53

to oxides, usually by reducing the dwelling time for melting. The as–formed glass contains strong reducing species related to Al metal, for which the exact structural origin is still unclear. Upon thermal treatment, a redox reaction is activated, resulting in the reduction of glass network, i.e., Si–O, and the formation of elemental Si by Si–O + Al*  Si + Al–O (see Fig. 9). A possible explanation is that Al is present as few–atom clusters or monovalent Al+, which are both highly reducing. Another explanation is the formation of unsaturation Si–O bond, that is, Si is divalent in the presence of Al-related species. A disproportionation reaction (Si 2+  Si4+ +Si 0) can result in the formation of elemental Si. This method has been successfully applied to the precipitation of Ge, Bi and Te elemental particles in oxide glasses containing GeO 2, Bi2O3 and TeO2. Similar to wet chemistry process, a higher reaction temperature favours the formation of crystals of larger sizes (Fig. 9a–d), thus allowing for controlled crystal growth [132]. Besides thermal activation, the formation of these elemental NPs can be also triggered by laser irradiation of the same parent glasses.

54

a

b

c

d

e

Fig. 9 Precipitation of elemental semiconductor NPs in an oxide glass. (a–d) TEM images of the glass containing different semiconductor (a, b: Ge; c, d: Si) NPs at two different heat treatment temperatures. (e) Schematic illustration of the formation of elemental NPs through the reduction by Al. Reprinted with permission from ref. [132], copyright 2011, American Chemical Society.

5.2 Powder sintering, co-melting and frozen sorbent method The three methods discussed in this section share certain common characteristics that are not found in conventional GC fabrication. First, the starting materials normally do not form stable glasses but they can be processed into TGCs by these processes. Second, using these methods we can easily obtain GCs containing the desired crystalline phases.

5.2.1 Powder sintering GCs can also be prepared by sintering of glass powders. This technique is much more efficient and versatile than solid-state reaction as the larger amount of fresh particle surfaces provides enormous number of nucleation sites. Starting from heterogeneous nucleation, the crystal 55

growth during powder sintering therefore begins from the interface and then occurs in the bulk of the glass phase. The starting glass powders can be prepared either by pulverization of conventional melt-quenched glasses or by the sol–gel process (see section 5.3). This method takes advantage of heterogeneous nucleation process at surfaces and does not rely on a heat treatment process. Powder sintering provides a versatile way for the incorporation of crystalline phase into the glassy matrix, and this crystalline phase is otherwise immiscible and cannot be vitrified by conventional procedures, e.g., melt–quenching. Thus, GCs can be prepared by co–sintering of the mixture of low–melting glass powder and stable crystalline particles. The crystalline particles remain stable and do not dissolve in the glassy phase, and the glass powder becomes softened and eventually densified after sintering. Therefore, the products are not different from common GCs, but GCs with the same composition cannot be prepared by the melt–quenching–devitrification process. A variety of low–melting glasses can serve as the matrix with sufficient chemical stability, including borate and heavy metal oxide based glasses. For instance, Y 3Al5O12:Ce3+ (YAG:Ce3+) crystals (melting temperature > 2000 C) were incorporated into a borosilicate glass by co–sintering of the YAG:Ce3+ and the glass powder at around 800 C (the liquidus temperature for the glass is above 1400 C) (Fig. 10) [191]. The sintering temperature is well below the melting temperature, but it is sufficiently high to soften the glass powders and to ensure densification during sintering. Direct sintering powders of crystalline materials at a designated temperature can also finally result in a crystal–in–glass GC composite. This can occur when the components in the raw materials have a big difference in the melting temperature, and part of the components has good glass–forming ability. During heating, the particles of the high melting components are welded together by the low-melting glass-forming components driven by the capillary force, resulting in the densified monolith. For instance, Oxynitride GCs, which are not accessible by conventional meltquenching due to high melting temperature, have been synthesized by sintering crystalline powder mixtures in an inert atmosphere [192],[193]. Sainz et al. reported the synthesis of a 56

YAlSiON GC by sintering a powder mixture of metal oxide with Si 3N4 at 1800 C [192]. In addition to the glassy phase, the resultant GC contains a mixture of crystalline phases as a result of high temperature reaction, including Si3N4, Si 4Al2O2N6, Y2SiAlO5N and Y10Al2Si3O18N4. Generally, GCs prepared by powder sintering are inferior in transparency as compared to their counterparts that are obtained by meltingquenching and heat treatment. The loss of optical transparency is caused by scattering by pores and grain boundaries which are unavoidable in sintering. Use of fine glass and crystal powders in combination with hot pressing method could reduce porosity of the products, while attaining high transparency comparable to glass is not possible. GCs of this type can find applications, where high transparency is not demanded, e.g., luminescent layer in LEDs. Fig. 10 presents a typical PiG (phosphor in glass)–derived GC containing YAG:Ce3+ crystals for LED applications. The GCs is translucent due to the presence of phosphor particles of micrometer size, pores and grain boundaries. a

cps/eV

b

c

Energy (keV)

Energy (keV)

Fig. 10 A sintered YAG GC by the PiG technique. (a) Transmittance spectrum of the sintered GC (base glass SiO2B2O3BaO/ZnO). Insets: photographs of the GCs sintered at different temperatures. (b) Photograph of the GC (glass: phosphor = 9:1). (c) SEM image and the corresponding EDX spectra of the GC plate. Reprinted with permission from ref. [191], copyright 2012, Optical Society of America.

57

5.2.2 Co-melting A more direct method for producing TGC is the co-melting process, in which the glass components and the pre-synthesised NCs are co-melted and then quenched to a TGC. Unlike conventional annealing process, this method provides almost an unlimited choice of crystal -glass combinations. However, there are two factors to be considered when using this method: 1) to avoid “dissolution” of the NCs during melting, the matrix glasses with a low melting temperature are usually selected; 2) to improve optical transparency, the index difference between the crystal and glass phase should be minimized. In this regard, crystals of nanometer size are favorable to retain high optical transparency. In a recent report, Zhao et al. succeeded in fabricating a composite glass using this process [194], where tellurite glass and solution processed LiYF4 NCs are the matrix and crystalline phases, respectively. To preserve high optical transparency, the concentration of NCs in the glass was below 170 ppm. When Er3+ doped fluoride NCs are incorporated into the glass, the resulting glass composite exhibits strong UC emission, which is typical for fluoride crystals. As photonic materials, a major advantage of the co-melting method over conventional process (i.e., quenching-annealing) is that the RE ions are doped exclusively in the crystals surrounded by a “clean” glass matrix. In this work, a critical problem of this process is the possible dissolution of NCs, especially fluoride NCs, by the glass melt at high temperatures, and this problem has, however, not been convincingly excluded.

5.2.3 Frozen sorbent method The “frozen sorbent (FS) method” was developed by Tanabe and co-workers of Kyoto University (Japan) for the fabrication of GCs that contained crystalline phase of high melting temperature. Unlike the method based on solid-state reaction or co-melting introduced above, this process does not need a lengthy thermal treatment process for the crystals to precipitate. For this method, the GCs are prepared by quenching a melt with a composition outside the glass-forming 58

region [195][196][197][198][198]. This method has been successfully applied for the fabrication of several oxide GCs with high transparency. For instance, Ca2SiO4 containing TGCs crystals have been fabricated by quenching a melt of 60CaO-40SiO2 to ambient temperature from 1550 C, where the liquid phase (glass melt) coexists with crystalline phase (Ca2SiO4) according to phase diagram. It should be noted that above 1650 C, this system is a homogeneous melt and quenching results in a homogeneous crystal-free glass [196]. The obtained TGCs were highly transparent and the RE-doped ones can be used for high power white LEDs. In terms of the colloidal chemistry, the FS method relies on the formation of a high temperature colloid containing dispersed crystalline phase, followed by quenching it to ambient temperature to obtain the desired GCs. Therefore, the equilibrium phase diagrams have a strong predictive power in the design of the GC composition and melting temperature for the FS method. According to Nakanishi et al., the merits of this process include [196]: 1) the ability to prepare novel GCs containing metastable phases; 2) possibility to prepare GCs with designed glass composition and crystalline phases by referring to the phase diagram.

Fig. 11 TGC containing Eu3+/Dy3+ codoped SrAl 2O4 fabricated by the FS method. (a) Compositions of the five samples fabricated and their optical photographs. The glass forming region is indicated in the diagram. (b) XRD patterns of the samples with both original surface and

59

polished surfaces. (c, d) SEM images taken on the polished surfaces of samples No. 4 and No. 5. Reprinted with permission from ref. [197], copyright 2011, The Ceramic Society of Japan. Beside the Ca2SiO4 and the Ca3Si2O7 based GCs, one of the prominent examples is the Al2O3 based GC, which has so far been only accessible by the FS method. Precipitation of pure crystalline Al2O3 from an oxide glass by thermal treatment is almost impossible as Al 3+ strongly bonds to the network formers like SiO 2 [198]. Using the FS method, the raw mixtures are transformed into a melt containing dispersed Al 2O3 particles at 1550 C (far below the melting point of Al 2O3, 2050 C), and quenching of the high temperature “colloid” results in a composite with known fraction of crystalline phase (Al 2O3), according to the level rule. Later, the FS method has also been applied to the fabrication of the SrAl 2O4 based TGC from the SrO-Al2O3-B2O3 system (see Fig. 11) [198]. The resultant TGCs are highly transparent and doping of Eu2+ and Dy3+ into the crystalline phase gives uncompromised persistent luminescence compared to corresponding phosphors. To summarize, this method is especially suitable for the precipitation of oxide crystals with high melting point in the glass phase, and these crystals otherwise cannot precipitate by the conventional process.

5.3 Solgel process Solgel is a liquid phase process developed for the synthesis of inorganic bulk and nanomaterials conventionally from metal alky–oxide precursors. This process can lead to formation of a variety of glasses with special compositions (e.g., high silica glass) under mild synthetic conditions, and these glasses are not achievable via the common melt–quenching method. In the sol–gel process, the sol is created by the catalytic hydrolysis of alky–oxide (such as TEOS) in a mixture of water and ethanol. The as-formed SiOX intermediate undergoes further condensation and polymerization in the following aging process, resulting in an increase in the viscosity, and steadily the formation of a transparent gel. In many examples, the use of expensive and volatile metal alky–oxide or organic compounds is unnecessary, and these compounds can be replaced by 60

common inorganic metal salts, such as nitrate and chlorides, and an organic chelating agent. Monolithic sol–gel glasses are finally obtained after careful aging of the gel for a long period of time and the densification process by sintering at high temperatures. Without such a lengthy procedure, non–crystalline powder is often resulted due to pulverization in the rapid gelation and drying process. The fundamental chemistry of sol–gel process for different metal and non–metal precursors has been discussed in many reviews [200], [201]. Sol–gel has so far been applied for the synthesis of a diverse range of glasses, from silicate to phosphate systems that contain multiple metal components. Crystallization of the sol–gel derived glasses can also be induced by controlled heat treatment to obtain monoliths, thin films and compressed powder products. Particularly, the crystallization of the sol-gel derived silica into quartz can take place at an unprecedentedly low temperature (< 1000 C) in the presence of catalytic alkali earth ions [188]. So far, GCs with desired bioactivity or mechanical properties have been prepared using this method in numerous research laboratories. For instance, a celsian based GC with the composition of SrOBaOAl 2O3SiO2 was synthesized by solgel process using Sracetate, Baacetate, Albutoxide and tetraethoxysilane (TEOS) [203]. The monolithic product was prepared by the sintering of cold pressed gel powder, and the results indicated that the formation of celsian occurred by a surface crystallization process at a temperature as low as 1040 C, far below the crystallization temperature required by the conventional melt-quenching glasses. The homogeneity of the sample can be improved by increasing the sintering temperature. By using the hot-pressing method, the celsian GC, which has the microstructure and density comparable to those of the products prepared by conventional melting process, can be produced. The process can be also employed for the preparation of fluoride containing GCs. For instance, a solgel derived silicate GC containing CeF 3 crystals was synthesized by Zheng et al. by using Ce(CH3COO)3 and CF3COOH as the Ce and F source, respectively [204]. During heat treatment of 61

the monolithic gel, crystallization of CeF 3 began at around 400 C, and CeF3 is transformed gradually into CeO2:F at higher temperatures by oxidation. This glass composite was applied as a UV light shield for biological protection as well as photocatalyst for photodegradation of dye solution. Similarly, ternary fluoride LiGdF 4 was crystallized from a solgel silica matrix by Lepoutre et al. The first crystals formed at around 600 C in this system are not LiGdF 4, but GdF3. By raising the heat treatment temperature, LiGdF4 crystals are observed as a result of the reaction of GdF3 with Li+ and F dispersed in the glass network. In the crystallization process, the RE dopant Eu3+ enters exclusively into the fluoride lattice, giving characteristic strong reddish photoluminescence [205]. The solgel process also enables preparation of various GCs in a more direct manner without a crystallization process by heat treatment. This is realized by the dispersion of pre–synthesized NCs in the liquid sol, and the following gelation process leads to the formation of solid gel containing dispersed NCs. This process therefore allows for almost unlimited choice of crystals as long as they remain stable in the sol and gelation process. The incorporation of functional NCs into amorphous gel can greatly extend the application of conventional silica and glass materials. For instance, colloidal QD ([email protected]) were incorporated into a silica gel by a normal solgel procedure using N,N–dimethylformamide (DMF) as the drying control chemical additive, forming a transparent gel with strong photoluminescence (Fig. 12) [206]. To enhance the compatibility of the QDs with the silica sol and to passivate the QDs surface, it is necessary to cap the QDs by organic ligands. In another example, CdSe/CdS/ZnS QDs were capped with 6–Mercaptohexanol (6–MHOH) and mixed with the silica sol. The obtained monolithic QDs–silica composite was applied as a converting layer in a white light LED [207]. Besides QDs, different types of inorganic NPs have been incorporated into silica gels by a similar sol–gel process without a crystal precipitation process. For instance, magnetic –Fe2O3 NPs were doped into a silica gel by using tetramethoxysilane (TMOS) as the silica source and DMF [208], water and morpholine as the 62

solvent. Compared with common GCs obtained by conventional method, these sol–gel derived composites are inferior in mechanical strength.

a

b

c

d

Normalized intensity 525

b

575 625 675 Wavelength (nm)

725

Fig. 12 QD–silica monolith synthesized by the sol–gel process. (a) Schematic diagram of the surface modified QDs. (b) Photographs of red light–emitting QD–Silica monolith (SM) with various concentrations of QDs in the silica matrix (from left: 0.12, 0.6, 1.2, 4.8 and 12 vol.%) under UV light. (c) PL spectra of QD–SMs with different QD concentrations after thermal treatment at 100 C and the QD–toluene solution (0.12 vol.%). (d) TEM images of sliced QD–SM with 12 vol.% QD. Reprinted with permission from ref. [207], copyright 2012, American Chemical Society.

5.4 Laser induced crystallization techniques Upon UV irradiation, noble metal NPs can be easily generated in glasses containing dissolved metal species, which catalysed the development of first modern GCs in the 1950s. In comparison, laser can be used to precipitate crystals in a localized fashion either on the surface or the interior of the glasses. Different types of lasers have been used to craft crystalline patterns in/inside glasses of diverse compositions. These types of glasses with laser induced crystal patterns have been envisioned for applications in photonics and other related areas.

63

5.4.1 Continuous wave (CW) laser induced crystallization In the recent decade, different lasers have been used to modify glass structures and to fabricate optical waveguide by taking advantage of the induced change in refractive index [209]. However, the precipitation of crystals in glass is not easy, since most oxide glasses, e.g., silicate glasses, are stable against crystallization at temperatures up to 1000 C. In addition, optical glasses are transparent in the visible and NIR range, thermal effect induced by most continuous wave (CW) laser operating in the visible/NIR range is too weak to initialize crystallization of the glass. By doping the glass with active metal ions centers, their absorption induces significant localized heating, leading to crystal precipitation inside glass or on glass surface. The first experimental demonstration of crystal precipitation was given by Komatsu et al. in an oxide glass with the composition (mol%) of 10RO10Sm2O380TeO2 (R=Mg, or Ba) [210]. The high content of Sm2O3 in this glass ensures high absorption of the laser, and the tellurite glass is relatively easier to crystallize compared to silicate glasses. Upon irradiation of a CW YAG:Nd laser at 1064 nm for only 30 s, dot patterns with diameter of around 50–150 m were generated, which induced the change of refractive index. XRD and TEM results confirmed the presence of Sm 2Te6O15 crystals in the laser–irradiated area. Irradiation for longer than 60 s resulted in the decomposition of Sm2Te6O15 crystals, being in agreement with the phase evolution of glass under thermal treatment. The crystallization was ascribed to the presence of Sm 3+ ions that strongly absorbed laser near 1000 nm, and no crystallization was observed when Sm 2O3 is replaced by Er2O3 [210]. This laser crystallization method was later extended for the “writing” of different crystalline patterns on glass surfaces, in which RE ions, such as Sm 3+ and Dy3+, are used as the absorber for the NIR laser. Of particular interest is the writing of crystalline patterns of nonlinear optical crystals for applications such as SHG (second harmonic generation). For instance, Bi 0.7Sm0.3BO3 crystal patterns were fabricated by direct scanning of the YAG:Nd laser on the surface of the glass with the composition of Sm 2O3–Bi2O3–B2O3 [211]. Similarly, line patterns of SHG crystals, such as 64

Sm2(MoO4)3, were fabricated by choosing appropriate glass compositions [212]. Functional crystals beyond SHG materials, such as Bi 2GeO5 (monoclinic) and Bi 4Ge3O12, have also been successfully precipitated with highly preferred orientation [213]. Again, Sm 2O3 was used as the absorber for the 1064 nm laser to induce local heating required for crystallization. Beside RE ions, transition metal centers, such as Cu 2+ and Ni 2+, can be also used as the absorber for the NIR laser due to their strong and broadband absorption. Particularly, the use of TM ion based absorbers instead of RE ions like Sm 3+ is advantageous and necessary as they do not contaminate the precipitated crystals containing large sized cationic sites. For instance, using Ni 2+ as the absorber, highly oriented Ba1−xY2x/3Nb2O6 crystals form in an Y2O3-BaO-Nb2O5-B2O3 glass upon laser irradiation [214]. Similarly, precipitation of LiNbO 3 crystals was demonstrated in a CuO (0.5%)–doped oxide glass with the composition of 40Li 2O–32Nb2O5–28SiO2 [215].

Fig. 13 Crystalline patterns created by a CW Yb:YVO4 fiber laser in a borate glass (8Sm2O3–42BaO–50B2O3). (a, b) Confocal scanning laser microscope (CSLM) photographs of a discrete crystalline line (a) and planar crystalline pattern (b). (c, d) CSLM photograph (c) and birefringence image (d) obtained by the Abrio IM imaging system for a bending line. The color of the line reflects the orientation of the polarization plane with respect to the slow axis. (e) Polarized optical micrograph of a spiral line pattern, together with a schematic illustration of the relationship 65

between the crystal orientation and SHG. (f, g) SHG wave images (the intensity of green light at 532 nm) taken using linearly polarized incident laser (at 1064 nm) with different EF orientations. Reprinted with permission from ref. [217],[218], copyright 2012, 2013, Elsevier. In these laser fabricated crystal patterns, the crystals are randomly oriented and are of polycrystalline nature. From viewpoint of photonic application, direct writing of single crystalline SHG crystals are much more attractive than that of polycrystalline counterparts. Using the same CW YAG:Nd laser, single crystalline pattern of β–BaB2O4 (β–BBO) was fabricated on the surface of a pre–crystallized RE 2O3–BaO–B2O3 (RE=Sm, Dy) glass, where the β–BBO crystals in the precursor GC are oriented randomly [216]. In order to fabricate the single crystal lines, the surface of precursor glass (preheated to 150 C to avoid fracture during laser irradiation) was covered by a layer of β–BBO powder (5–30 m) prior to laser irradiation. A single crystal line of the β–BBO was then generated by scanning the laser focal point across the glass surface at a speed of 3–8 m/s. During the laser irradiation, the β–BBO crystal powder was first dissolved in the bulk GC and then single crystalline β–BBO with its c–axis oriented along the scanning direction was obtained. Recent improvements showed that bending curves and curved shapes could be patterned on the surface of a 8Sm2O3–42BaO–50B2O3 glass using a YVO 4:Yb laser (CW, @ 1080 nm, 0.8 w) [217],[218],[219]. Using a polarized optical microscope, it can be seen that the c–axis of the –BBO crystals are aligned along the laser scanning direction, and this orientation is not disturbed by bending (Fig. 13). At the bending point, the orientation of the –BBO crystals changes gradually and follows scanning direction. The orientation of the crystals is also confirmed by the SHG, as shown in Fig. 13f, g. In addition to solid state NIR lasers based on RE-activated crystals, diode as well as excimer laser have also been used for the localized crystallization of different oxide glasses. For instance, diode laser at 808 nm was used to precipitate fluoride crystals in an oxyfluoride glass matrix. It was found that the devitrification of glass began at the laser power of 2.3 W at which the local temperature was above 750 K [220],[221]. The crystallization of RE fluoride manifested an obvious 66

crystal splitting of the emission spectra of the dopant ion Er 3+, which also served as the absorber for the laser. In the same oxyfluoride glass, RE fluoride can also be precipitated by using a CW argon laser operating at 514 nm, and this laser can be efficiently absorbed by Er3+ ions [221]. Besides solid-state lasers, CW CO2 laser (at 10.6 m), which gives a strong heating effect, can be also used to precipitate SHG crystals. In a 40BaO–45B2O3–15TiO2 glass, for instance, surface crystallization of BaTi(BO 3)2 were generated by irradiation with a CO2 laser at only 0.72 W for 300 s [222]. Compared with most oxide glasses, chalcogenide glasses are much easier to crystallize by laser due to the much lower temperature of crystallization. In recent years, Jain, et al has carried out laser crystallization on several chalcogenide glasses by using CW laser [223],[224]. For instance, in an 82SbSI–18Sb2S3 glass, Sb2S3 single crystals can be fabricated using a diode laser with a wavelength of 520 nm that is above the bandgap of the glass. Care has to be taken during laser patterning in order to precipitate the crystals without local melting of the glass. For the SbSI glass, it was found that crystallization of the SbSI crystals occurred only in a very narrow range of laser power, implying the chemical vulnerability of the glass under laser heating.

5.4.2 Pulse laser crystallization Compared with CW laser, pulse lasers offer higher peak power and better spatial resolution in micromachining of glass. With the rapid advance in ultrafast lasers in the recent decades, nanosecond (ns), picosecond (ps) and femtosecond (fs) lasers are now extensively used for the micro–fabrications. These lasers, especially fs laser, enable direct fabrication of microstructure in a sub–micrometer scale, promoting broad application in diverse areas. In inorganic glasses, pulse lasers have been used to craft different crystalline structures with high spatial resolution in a three-dimensional (3D) fashion. For nanosecond laser irradiation induced crystallization, the use of absorber ions is still used to maximize the thermal effect. For a La 2O3–B2O3–2GeO2 glass, which crystallizes congruently into 67

LaBGeO5 crystal, the crystallization can be induced by irradiation with a Cu–vapour ns laser (operating at 510.6 and 578.2 nm) with a repetition rate of 10–18 kHz and peak power of 10 9–1012 W/cm2 [225]. Using Cr2O3 as the absorber, crystal dots were generated in the laser scanned area. To facilitate the crystallization and meanwhile avoid cracking during laser irradiation, the glasses were preheated to 600 C for laser crystallization. From SEM observations, the crystallization pattern formed inside the glass contained dispersed crystalline spots, rather than dense crystalline regions. Both the size and the density of the crystal spots were affected by the laser operation parameters such as power and scanning speed. Fs lasers offer unprecedented peak power as compared to other pulse laser systems, and the use of fs laser for microfabrication exerts minimal thermal effect on the materials, therefore offering high spatial resolution. More importantly, different SHG crystal patterns have been made by irradiation with fs laser without using metal ion as absorbers. Unlike the interaction of materials with ns or CW lasers, fs laser generally induces multiphoton absorption, leading to the ionization of glass components and sometimes micro–explosion or plasma generation. This is followed by the temperature rise, leading to melting, element re–distribution and finally crystallization process. Therefore, fs laser allows the precipitation of 3D crystalline patterns or microstructures in a transparent glass, which is not easily achievable by most other type of lasers. In recent years, fs lasers have been widely used for crafting microobjects as well as photonic devices in transparent materials, including single crystals and glass [226]. SHG crystals, metal NPs and semiconductors NPs have been precipitated by fs laser in different types of glasses in a 3D space–selective manner. Typically, to induce crystallization, a Tisapphire fs laser of high repetition rate of 250 KHz (as compared to 1 kHz) is often used due to its relatively larger thermal effect, as compared to the low repetition fs pulse (with much higher pulse energy). The first SHG crystal precipitated from glass using fs laser was made by Miura et al. in a parent glass of 47.5BaO5Al 2O347.5B2O3, which is different from the crystallization of βBBO 68

crystals induced by a CW laser [227]. A fs laser with a pulse duration of 130 fs, repetition rate of 200 kHz and pulse energy of 4 J (laser power density: 41012 W/cm2) was employed in their experiments. Upon irradiation for only 10 min, the region around the focal area experienced structural change and crystallization of β–BBO occurred after irradiation of over 20 min, as confirmed by both optical microscope and XRD. Furthermore, moving the focal point inside the glass created a line of polycrystalline β–BBO crystals. Using the similar laser systems, a number of SHG crystals were created from their parent oxide glasses [9]. Comparatively, the generation of single crystalline patterns requires special care to adjust the scanning speed as well as laser power. By careful control of the irradiation conditions, 3D single crystalline waveguide of LaBG eO5 can also be produced in the parent glass. As reported by Stone et al., uniform crystalline lines of LaBGeO5 were fabricated by high angular–resolution electron diffraction (Fig. 14) [228]. It can be seen that the c-axis of the crystal is orientated along the laser scanning direction. At the junction, a gradual change of the orientation along the scanning direction is observed, indicating that the crystal line remains intact around the junction. This crystal–in–glass waveguide demonstrates an optical loss of 2.64 dB/cm. It has been observed that the crystallization process induced by fs laser and CW laser is mechanistically different. For instance, for a LaBGeO 5 glass, the crystallization takes place at much higher scanning speed for fs laser compared to that of CW laser (at 1064 nm), being ascribed to the different temperature gradient and other factors, such as elemental redistribution. Especially, elemental redistribution during laser irradiation leads to the deviation from the ideal stoichiometry of the LaBGeO 5 crystals, which is confirmed by microRaman analysis [229]. Although fs laser induces crystallization without the use of an absorber, the crystallization process nevertheless can be greatly enhanced in the presence of active ion centers. For the crystallization of BTS, for instance, the crystal growth became much faster in the presence of Ag+ ions, which changed to elemental Ag and possibly served as the nucleation centers and the 69

plasmonic absorber for the fs laser [230]. The formation of Ag NPs led to the yellow coloration in the laser irradiated area. Compared with the previous work of some present authors [231],[232], the time required to crystallize BTS from a similar glass doped with RE ions is greatly shortened, and this is explained by the strong localized heating effect of Ag NPs.

a

b

c d e

f Fig. 14 Direct laser writing of ferroelectric single crystalline LaBGeO 5 waveguide in glass. (a) EBSD (electron backscatter diffraction) mapping for a single–crystal line. (b) The color correspondence of crystal orientation parallel to each reference axis. (c) The lattice orientation (represented by a hexagonal cell). (d) Mapping of the angular deviations from the average orientation, verifying the absence of low–angle grain boundaries. (e) A Y–shaped junction. In each branch the optic axis (slow axis) is oriented approximately parallel to the growth direction. (f) The two branches are merged back to a single line. Reprinted with permission from ref. [228], copyright 2015, Nature Publishing Group. In addition to different SHG crystals, luminescent crystals doped with TM or RE ions can be patterned inside glass in a space–selective manner in different oxide glasses. Upon crystallization by fs lasers, the dopant ions at the same time occupy the preferred lattice sites due to elemental redistribution in the short melting period. Binary fluorides are the simplest crystals precipitated 70

from various oxyfluoride glasses. As demonstrated by Liu et al. [233], during laser crystallization of CaF2 from an oxyfluoride parent glass, Er 3+ ions preferably entered the CaF2 lattice, resulting in strong UC luminescence from the crystallized area. Likewise, RE ions can be doped into a SHG crystal pattern, such as BTS, without much change in the operating conditions. Moreover, fs laser can be used to control the distribution of different precipitated crystals in glass. This strategy was demonstrated by Zhou et al. based on a SiO 2/Na2O/Ga2O3/LaF3 glass, in which Ga2O3 and LaF3 crystals can precipitate upon thermal treatment [234]. Upon controlled fs laser irradiation, the spatial distribution of Ga 2O3:Ni and LaF3:Er crystals could be easily modulated, resulting in the generation of crystallized patterns with quite different luminescence properties (see Fig. 15). These techniques enable the crafting of luminescent nano/micro–structures inside a transparent medium and are anticipated to have important implications for photonic applications. Among the different components in glass, noble metal atoms are the most unstable and their clustering can be more easily activated. Fs laser irradiation allows, in a space-selective manner, for a fast clustering of noble metal atoms, such as Ag and Au. Patterns of Ag or Au NPs would develop after further growth of clusters into NPs by the subsequent thermal treatment [235]. The presence of these plasmonic NPs results in strong coloration of the parent glass. Since metallic nanostructures can be generated in an insulating glass matrix, this technique has been envisioned for potential application in integrated optical devices as well as solid-state electronics. In addition to noble metal, laser precipitation of non–noble metal NPs has also been realized. For instance, dots of Cu NPs were formed by fs laser irradiation of a parent glass with a composition of 5K2O–5CaO–89B2O3–1CuO [236].The formation of Cu NPs of 2 – 5 nm around the focal area was ascribed to a thermal reduction process.

71

Raman shift (cm-1)

i

Intensity (a. u.)

a

ii

iii

Raman shift (cm-1)

b

i

Distance (m)

ii

Intensity (a. u.)

Intensity (a. u.)

c

ii

iii

Intensity (a. u.)

Raman shite (cm-1)

i

iii

Wavelength (nm)

Fig. 15 Writing of dual–mode fluorescent crystal patterns in glass. (a) Micro–Raman spectra for the glass and laser modified area (i), and the micro–Raman mapping at 775 cm –1 (Ga2O3) (i) and 370 cm–1 (LaF3) (ii) for the crystalline patterns. (b, c) Fluorescence from crystalline patterns of LaF3:Er (b) and Ga2O3 (c). Relative fluorescence intensity change, image and fluorescence spectra are given from left to right in i, ii and iii. Reprinted with permission from ref. [234], copyright 2010, American Chemical Society. The precipitation of non–noble metal NPs and elemental semiconductors (such as Si and Ge) in different glasses have also been realized by fs laser irradiation. As discussed earlier, Lin et al. [131],[132],[237] produced a special type of GCs by addition of small amount of metal Al that changed the redox balance of the oxide parent glasses. The as–made glasses are transparent as metal species are completely and homogeneously solvated during melting and stabilized by quenching the glass to ambient temperatures. Laser irradiation triggers the redox reaction in such glasses, leading to the formation of elemental NPs around the focal area. Scanning of the focal point in the glasses 72

generates crystalline patterns made up of semiconductors, like Ge, which have potential to be applied in optoelectronics.

5.5 Electron beam (E-beam) irradiation Similar to lasers that precipitate crystals in glass in a localized manner, E-beam irradiation also induces the precipitation of crystals in different oxide glasses. In a series of work reported by Jiang et al., it was found that oxide as well as element crystals could be precipitated at the focal area of the E-beam generated by a HR-TEM system [238]. For instance, in an oxide glass with the composition of 3.4Na2O–3.3K2O–13.3Ga2O3–80.0SiO2, Ga2O3, NCs were generated by E-beam irradiation at the intensity of 1.8410−2 pA/nm2 for 2 min, as observed from the electron diffraction patterns shown in Fig. 16. The same Ga2O3 crystals can be created by subjecting the parent glass to thermal treatment of at 900 C for 6 h. With continued irradiation, the Ga2O3 crystals grow further, as evidenced by the polycrystalline–like diffraction ring in the ED (electron diffraction) pattern. However, the dopant Ni 2+ (added in the form of NiO) ions can enter the Ga2O3 lattice only upon thermal treatment, but this is not the case for E-beam irradiation. A possible reason is that Ni 2+ doping into Ga2O3 occurred by diffusion, which is a slow process and can only occur during thermal treatment. The formation of Ga 2O3 crystals by E-beam irradiation is too fast to allow the migration of Ni2+ ions into the Ga2O3 NCs [238]. Based on a recent TEM observation on an oxyfluoride TGC containing LaF3 NCs, Jiang and co-workers revealed that E-beam irradiation led to the precipitation of LaF3 NCs and facilitated the incorporation of dopant ions (Eu3+) into the LaF3 NCs [239]. This result is important for understanding the luminescence behavior in similar oxyfluoride GC materials.

73

c

e

b

d

f

g Intensity (a. u.)

a

nm-1

Fig. 16 Crystallization of Ga 2O3 induced by E-beam irradiation. (a–d) ED patterns of the parent glass (a–c) and GC (d). (e, f) Dark field images of the glass (e) and GC (f). (g) Integrated intensity profiles of ED of the glass and GC. Reprinted with permission from ref. [238], copyright 2007, American Institute of Physics. Since electrons carry negative charge, irradiation by Ga2O3 would lead to the reduction of the glass component and the formation of element particles and oxygen–related defect centers. The precipitation of metal and non–metal elemental particles has been observed in several oxide glasses upon E-beam irradiation during TEM observation. For instance, Jiang et al. found that GeO 2 in a GeO2–SiO2 glass is extremely sensitive to E-beam irradiation. After only 1 min exposure, particle–like spots appeared in the TEM image, and this was confirmed by in–situ EELS to be Ge particles. Growth of Ge crystals was observed after continued exposure to E-beam, while the precipitation of Si was not observed, suggesting robustness of SiO 2 network against electron damage [240]. The precipitation of metal crystals by E–beam took place in a 60ZnO–20B2O3–20SiO2 glass, in which Zn2+ was reduced to Zn under 100 keV E-beam irradiation. From in–situ TEM observations, the formation of Zn crystals took less than 1s, suggesting the extreme vulnerability of ZnO to E–beam. In both cases, the formation of elemental particles is accompanied by the reconstruction of the glass network and the evolution of oxygen–related defect. The selectivity of the precipitated elements seems to be correlated with the electronegativity or the electrochemical reducing potentials, as well as the role of the specific element in the glass 74

structures. Si atoms in SiO4 tetrahedra are stable, while other elements bonded weakly to oxygen are easier to be reduced and precipitated under E–beam irradiation.

5.6 Orientated crystallization by external physical fields 5.6.1 Electric and magnetic fields For most inorganic glasses, crystals formed by thermal treatment do not have a preferred orientation except for a few polar crystals. The presence of a strong static electric field (EF) or magnetic field (MF) can exert an external force on the formation of polar crystals, preventing random orientation during their growth [241]. In general the crystallization of polar compounds with a permanent dipole can be affected by an external EF [242],[243],[244]. For instance, for a glass melt of BTS–75SiO2, crystallization of BTS does not occur at 1300 C [242]. When an EF was set up between the Pt crucible and Pt rod inserted into the center of the melt, highly orientated crystallization occurred along the c–axis in the presence of an EF of only 1.2 V, leading to the formation of needle–shaped BTS crystals with the c–axis oriented perpendicularly to the electrode surface [242]. A similar effect was found in a 9Li 2Si2O5BaSiO3 glass. In the presence of EF, Li2Si2O5–crystals were oriented along the fast growing c–axis, which was perpendicular to the electrode surface, while the growth rate was slow in the other two axes. The general mechanism behind this process is that the EF has facilitated nucleation on the electrode surface, and the following radial growth of crystals leads to the observed orientation. In comparison to the directed crystallization, a complex crystallization behaviour was revealed in several different glasses [245],[246]. As shown in Fig. 17, there is an obvious orientation of the BTS crystals, while a closer observation by SEM and TEM reveals an irregular fourfold hierarchy structure. In addition to the highly oriented main trunk with primary and secondary branches, a fourth hierarchy is obs erved with TEM, which is built up with fine, irregular honeycomb and lamellar structures. The space between these structures is filled with the glass phase. 75

a

b

c

Fig. 17 EF assisted nucleation and crystallization in a Ba2TiSi2.75O9.5 glass. (a, b) SEM images of a sample cut parallel (a) and perpendicular (b) to the main growth direction. The Pt wire is placed in the left (not shown here). (c) A magnified image for the area framed in b. Reprinted with permission from ref. [245], copyright 2010 American Chemical Society. MF exerts a similar influence on crystallization of glass like that of EF. Even for paramagnetic or diamagnetic substances, external EF of several Tesla can still have a substantial effect on the growth direction of the crystals [247]. Crystallization of glass occurs in the presence of MF with the aim to align the crystals to obtain superior performances compared to GCs containing randomly oriented crystals. The first GC prepared under MF was reported by Komatsu et al. in a Bi–based superconducting GC containing Bi 2Sr2CaCu2Ox (Bi–2212) phase. Crystallization of this glass takes place via several intermediate steps, and the formation of the Bi–2212 phase occurs at around 780 C. In the presence of a MF = 10 T, crystallized Bi–2212 platelets are oriented preferably along the MF direction, i.e., the c–axis is oriented along MF. Without MF, preferred orientation was not observed by XRD. The GC prepared in MF shows enhanced performances compared to GC obtained by normal heat treatment [248]. The same authors also realized the crystallization of a 30BaO–15TiO2–55GeO2 in the presence of 10 T MF. The c–axis orientation was enhanced when the MF was perpendicular to the surface of glass, resulting in the enhanced SHG 76

signal for the surface crystallized Ba 2TiGe2O8 GC. When the direction of MF was parallel to the glass surface, both the c–axis orientation and SHG were suppressed [249]. In comparison with the strong MF, the use of a weak MF (0.1 T) also has a subtle effect on the crystallization of the fluoride glass fibers. As reported by Tucker et al., an external MF of 0.1 T effectively suppressed the crystallization of the glass fibers, which was tentatively explained by the difference in the magnetic susceptibility of the glass and crystals [250],[251].

5.6.2 Mechanical stress Without the use of an electric or a magnetic field, GCs with oriented crystallization can be made by the application of mechanical stress to the viscous melt. According to thermodynamic of crystallization, it is not the stress on the melt, but the shear flow, that causes the change of the crystallization behavior of the GC. The external stress normally affects the crystal growth of compounds that have a preferred growth direction, such as uniaxial crystals. Experimentally, the mechanical stress is applied to the glass melt often by extrusion and uniaxial pulling. A variety of GCs have been made by the extrusion process to precipitate crystals with the desired orientation, which determines the mechanical and electric properties [252],[253],[254],[255],[256][257]. During extrusion, different crystals may have different crystallization behaviour depending on their structure and energetics. Crystals in such GCs often have a rod or planar (such as mica) shape due to the strongly anisotropic growth. Fig. 18 shows SEM images of a typical extruded GC, in which the crystal phase is fluoroapatite. Before extrusion, the glass was first held at 1200 C to precipitate the rod–shaped fluoroapatite crystals [258]. Extrusion was performed at the temperature range of 675 – 800 C through a graphite die at a velocity of 1 mm / min, where the randomly oriented crystals are forced to align along the extrusion direction. The orientation of the crystals was confirmed by XRD (Fig. 18c) and EBSD patterns (not

77

shown here). The alignment of apatite crystals results in the improved mechanical properties for practical applications.

b

c

Intensity (a. u.)

a

10

20

30 40 2θ (degree)

50

60

Fig. 18 Texture of an extruded mica–based GC. (a, b) SEM images of cut planes parallel (a) and perpendicular (b) to the extrusion direction. (c) Corresponding XRD patterns for a and b. The pattern of JCPDS 15–0876 is shown for comparison. Reprinted with permission from ref. [258], copyright 2015, The Royal Society of Chemistry. Directional crystallization for GC can be also realized by unidirectional pulling. The crystallization process is mechanistically very similar to the extrusion process. This unidirectional crystallization process has been conducted for the fabrication of the oxide GCs based on different crystals. For instance, Hosono et al. prepared a calcium phosphate GC containing oriented –Ca(PO3)2 crystals [259], in which the growth direction of the crystal was parallel to the c–axis and the direction of pulling. Furthermore, it was found that the unidirectional crystallization can be affected by the small amount of additives and temperature. Partial replacement of Ca for Sr lead to enhanced orientated crystallization, while the preferred orientation was completely disrupted by the addition of MgO. A similar technique was used for the precipitation of oriented metal rods in inorganic glasses. For instance, by drawing the melt at high temperature, aligned Ag nanorods of different sizes were precipitated in an oxide glass, resulting in the anisotropic third–order nonlinear

78

optical properties [260]. In fact, this technique based on pulling has been widely used for the fabrication of optical polarizers, where organic molecules are aligned along the pulling direction.

5.7 Surface crystallization and control Many glasses favour surface crystallization over bulk crystallization for reasons that have been discussed in section 2.3 [44]. Kinetic studies on surface crystallization revealed that the nucleation sites are rapidly saturated in early stage of crystallization, preventing the direct measurent of nucleation rate. The factors affecting surface nucleation include: 1) elastic strain; 2) presence of foreign particles; 3) atmosphere for heat treatment; and 4) surface roughness and the degree of surface mechanical damage. The growth of surface–nucleated crystallites has been examined for a number of silicate glasses. Surface crystallization has been utilized in two different ways. First, it has been used for the production of GCs by sintering of the glass powders. Due to the large amount of reactive surface s, the ceramization via this method is much faster than that via conventional process, e.g., heat treatment, and the sintering temperature can be lowered accordingly. It is believed that the reactive surfaces have a catalytic effect on the crystallization and the small particles themselves may catalyse the growth of large crystals. Second, surface crystallization is often applied for the fabrication of planar GCs covered by a layer of oriented crystallites. These GCs could be used as a planar waveguide and they demonstrate interesting optical properties, such as SHG. In general, the crystallization propagates into the interior of glass, with a preferred growth direction pointing to the interior of the glass. The microstructures of different surface-crystallized TGCs covered by crystalline layers such as BaTiO 3 and BTS (Ba2TiSi2O8), have been examined by different groups [68][261],[262]. Fig. 19 shows a “perfectly surface-crystallised” TGC covered by a layer of dense, highly oriented Fresnoite-type STS crystals prepared from a non-stoichiometric STS (Sr2TiSi2O8)

79

glass. It can be seen that the presence of additional glass forming oxide (SiO 2) leads to the formation of parasitic amorphous SiO 2 spheres near the STS crystals.

Fig. 19 Microstructure of a surface-crystallized STS (Sr 2TiSi 2O8) TGC prepared from a non-stoichiometric parent glass. (a) Photographs of the STS parent glass and TGC. (b) Polarized light micrograph for the cross-section of the STS GC. Highly oriented crystalline phase can be clearly observed. (c) HRTEM image of the STS GC, showing phase separated domains. (d) EDX spectra for the crystalline domain (A) and the amorphous domain (B). Reprinted with permission from ref. [68], copyright 2013, Nature Publishing Group. There are different types of bulk TGCs prepared by taking advantage of surface crystallization [264],[265],[266],[267]. TGCs, which form via devitrification only on their surfaces, are of special interest for their potential applications in photonics. Ding et al. made several surface crystallized GCs covered by a crystalized surface. For instance, BTS (Ba2TiSi2O8) surface crystallized GC was synthesized from a parent glass with the composition of 33.3BaO–16.7TiO2–50SiO2 [267]. It was found that an ultrasonic treatment in a suspension of BTS particles for 15 min was indispensable for the formation of highly oriented surface crystalized GCs, which give rise to strong SHG. The similar method was applied for the synthesis of surface crystallized GCs covered with crystals of –BBO, LiNbO 3 and Sr2TiSi2O8 with thickness of greater than 50 m [264]. Again, surface ultrasonic activation and seeding were found to promote the surface crystallization.

80

5.8 Melt–in–tube (MiT) technique for TGC fibers Glass-forming melts can be processed into fibers by drawing at a proper viscosity. To avoid crystallization, the fiber drawing temperature should be between the glass transition temperature (Tg) and the crystallization temperature (Tx). Similar to the GC formation, the fibers of GCs are also finally obtained by a heat treatment. So far, only limited number of GC fibers doped with RE and TM ions have been fabricated by several groups, and these fibers are believed to be promising gain medium especially in the IR range. The difficulty in the fabrication of GC fibers by this conventional process is that, at drawing temperature, the coarsening of crystals often occurs due to a fast crystal growth rate (see Fig. 2), leading to the loss of transparency. The Melt–in–Tube (MiT) technique developed by Qiu and coworkers appears as a rather simple yet versatile technique for the fabrication of luminescent TGC fibers [268],[269]. In fact, the same method was also proposed earlier in 2012 by Dianov and coworkers for the fabrication of a silica cladding fiber with a Ni-doped germanosilicate core glass that contained Ga 2O3:Ni 2+ nanocrystals [270]. In a typical procedure, a rod of the parent glass is inserted into a silica tube and heated together to the fiber drawing temperature (for silica glass, ~1830 C ), at which the core glass transforms to a melt. The MiT was then drawn at a high speed (15 m/min) to form the precursor fiber, forming a glass-in-glass composite fiber. Afterwards, TGC fibers can be obtained by heat treatment at the crystallization temperature of the core (glass phase), while the silica sheath of the fiber remains in the glassy state. Different types of GC fiber have been fabricated by the MiT process. For instance, Fang et al. recently fabricated ZnAl 2O4:Cr3+ based GC fibers, which exhibited intense NIR fluorescence [268]. Later, Ni-activated TGC fibers were fabricated by this method, which demonstrated strong NIR emission from the crystalline phase of Ni activated LiGa 5O8, LiTaO3 and LiAlSi 2O6 crystals. The splice joint between the silica fiber and the gain fiber (with a GC core) can be clearly observed by naked eyes (Fig. 20) [271]. From the observation of the cross section by SEM, the boundary between the silica clad and the partial crystallized core is distinct, suggesting 81

limited inter–diffusion between silica sheath and the GC core due to the high drawing speed and rapid cooling rate. By further minimizing optical loss of the GC fibers, Yu et al. demonstrated optical amplification at 1300 nm. As the emission bandwidth of these MiT fibers was much larger than that of RE doped fibers, the Ni doped GC fibers are favorable for optical communication application with a broad working bandwidth (Fig. 20). The MiT technique is also expected to be amenable to the fabrication of Bi–doped as well as QD–doped GC fibers for NIR photonic applications. Furthermore, this MiT technique can be a general process for the fabrication of composite fibers. A team from MIT recently succeeded in the conversion of aluminium core preforms to silicon core fibers [272]. The mechanism behind the transformation of Al to Si core is apparently the result of the redox reaction between SiO 2 and Al, which is similar to the precipitation of silicon NPs from silica glass, as discussed in section 5.1.5.

Fig. 20 TGC fibers containing Ni-doped NCs. (a, b) Photographs of the bulk glass and the as-drawn fibers. (c) SEM image of the joint section between silica and the gain fiber (Ni-doped TGC). (d) Optical micrograph of the cross-section of the TGC fiber. (e, f) Raman mapping for the LiTaO3 (593 cm−1) and LiAlSi 2O6 (486 cm−1) crystals distributed within the fiber core. (g) 82

Photograph of the fiber under illumination with green light. (h, i) EDX mapping for Ta and Al. Reprinted with permission from ref. [271], copyright 2017, Nature Publishing Group.

6. Emerging applications of TGCs

6.1 Luminescent TGCs for optical and photonic applications TGCs can provide luminescent ions, such as RE and TM ions, a crystalline host, which can greatly enhance the PL efficiency. Most of such TGCs belong to type–II GCs as the precipitated crystals, e.g., fluorides, are not made of glass network formers. Combined with the uncompromised transparency, these luminescent TGCs have been extensively explored for application in the areas from energy to photonics.

6.1.1 Luminescent TGCs as spectral converters for solar cells Conventional single junction solar cells are only able to use the photons with energies larger than the bandgap (Eg) of the semiconductors. The spectral convertors allows for the more efficient use of solar radiation in spectral region, where the cell exhibits small or no photo–response. As spectral converter layers have to be transparent, mechanically robust and stable, luminescent TGCs are strong candidate materials for this application due to their high emission yield. Here we introduce three types of luminescent spectral converters based respectively on normal Stocks (down-shifted) emission, anti–Stocks emission (upconversion, UC) and quantum cutting. The spectral converters based on Stocks emission are widely explored, because Si–solar cells (for instance) perform optimally at spectral edge slight above its absorption edge (1.12 eV). To be an efficient spectral converter for Si–based cells, the materials should have emission band located in the region of 800 nm–1000 nm with near–unity quantum yield (QY) and strong and broad absorption band covering the whole short wavelength region (<600 nm). Such a spectral converter 83

has been realized in TGC materials. Fang et al. developed a YAG-based GC codoped with Cr 3+ and Yb3+. The energy transfer from Cr3+ to Yb3+ enables a down–shifted spectral conversion, as seen in the energy diagram in Fig. 21. This GC demonstrates efficient emissions from both Cr3+ and Yb3+, and could be excited at a broad spectral bandwidth of 400 – 700 nm, thus allowing efficient conversion of short wavelength visible light to red and NIR emissions [273].

a

Energy (103 cm-1)

Absorbance (a. u.)

b

400 600 800 1000 1200 1400 Wavelength (nm)

c

Intensity (a. u.)

Absorbance (a. u.)

d

400

500 600 Wavelength (nm)

700

700

800 1000 1100 1200 Wavelength (nm)

Fig. 21 Down–shifted emission in a YAG:Cr3+, Yb3+ based GC for spectral conversion. (a) Absorption spectra of the GC and glass. The inset shows the photographs of the samples. (b) Energy transfer mechanism from Cr 3+ to Yb3+. (c) Excitation spectra of the GC monitoring the emission of Cr3+ at 800 nm and the emission of Yb 3+ at 1030 nm. The dotted line is the spectra for 0.1Ce 3+ –0.5Yb3+ codoped GC. (d) Emission spectra for the Cr 3+–Yb3+ codoped GC under excitation at 450 nm. The dotted line is the Yb 3+ single doped GC. Reprinted with permission from ref. [273], copyright 2015, American Ceramic Society. Compared with Cr3+, most of other TM ions as well as RE ions are not suitable as activators in spectral converters due to their weak absorption in the visible region. GCs containing chalcogenide QDs can also produce emission band in a similar NIR region and strong excitonic absorption 84

covering entire visible wavelength region. The QD–doped TGCs are good converts and have received wide attention [274]. Nevertheless, as a transparent spectral converter, the size of the precipitated crystals should be carefully controlled and the concentration of impurity ions like Fe 3+ should be minimized in order to reduce scattering and absorption loss.

a

d Intensity (a. u.)

b

c

450

500

600 650 900 1000 1100 550 Wavelength (nm)

Fig. 22 Oxyfluoride TGC for NIR quantum cutting. (a) XRD patterns of the oxyfluoride parent glass AG6 (60SiO 2–20Al2O3–20CaF2–0.3Tb3+–xYb3+, x=6) and the corresponding GCs (denoted as GC6, x=6, and GC26, x=26). Diffraction peaks of cubic CaF 2 can be observed. (b) The photographs of the glass and GCs. (c) Energy transfer mechanism from Tb 3+ to Yb3+, enabling NIR quantum cutting. (d) Emission and excitation spectra for the GC samples. Cooperative quantum cutting is confirmed by the overlap in the excitation spectra for the NIR emission of Yb 3+ and visible emission from Tb 3+. Reprinted with permission from ref. [276], copyright 2008, American Institute of Physics。 Compared with normal Stocks emission, down converters based on quantum cutting characterize a luminescent QY up limit of 200% by theory. They have attracted considerable interest of scientists as they provide a new possible route for breaking the Queissner–Schokley limit for conversion efficiency confronted by solar cells. TGCs with quantum cutting emission are mostly based on RE-doped fluorides due to their low phonon energy. For solar cell application, quantum cutting emission in the NIR is more desirable since most semiconductors used in solar 85

cells have band gaps in the NIR. Among different quantum cutting mechanism s, one of the most import ion pairs are the RE 3+–Yb3+ (RE: Pr, Tb, Tm), in which RE 3+ ions at their high excited state transfer energy to two Yb 3+ ions by either a two–step process or a cooperative mechanism [275]. The quantum cutting emission has been demonstrated by a plenty of TGC materials including both fluoride and oxide based TGCs [276],[277],[278],[279],[280]. For instance, Ye et al. developed a series of oxyfluoride GC containing CaF 2 and REF3 (RE: Y, La) codoped with RE 3+–Yb3+ ion pair. One of the fluoride-containing TGCs is prepared from a parent glass with the composition of 60SiO2–20Al 2O3–20CaF2: 0.3Tb, xYb [276]. In the heat treatment temperature range of 665–675 C, the precipitated CaF 2 crystals enriched with RE ions are less than 20 nm in size, thus giving high transparency for the GCs (see Fig. 22). Strong NIR emission from Yb 3+ is generated under excitation at 484 nm (corresponding to 5D4 level of Tb3+), resulting in an internal quantum efficiency approaching 200%. A similar LaF 3 based GC was developed later by Ye and coworkers with the composition of 45SiO 2–12Na2O–23Al 2O3–20LaF3–0.5Tm3+–xYb3+ [277]. Fluoride based GCs for quantum cutting have also been developed by Wang et al. based on different oxyfluoride parent glasses [278],[279],[280]. In these TGCs, based on the spectroscopic measurements, the NIR emission of Yb3+ is efficiently sensitized by RE ions of Tb 3+, Pr3+, and Tm 3+ upon excitation in the visible region between 455–485 nm, which is approximately twice the energy of the Yb3+: 2F5/2. The systems showed a calculated QY approaching 200%, while applications of these TGCs for solar cells are however still illusive primarily due to low absolute QE values, which is usually less than 50%. Spectral converters based on UC enable the harvest of photons with sub–bandgap energy in the NIR region, and they therefore offer another strategy to overcome the Queissner–Schokley limit. Fluoride based TGCs are again preferable because UC process favours hosts of low phonon energy. Up to now, a number of fluoride crystals, such as CaF 2, NaYF4 and REF3, have been precipitated in a variety of parent glasses [106],[281],[282],[283],[284],[285],[286],[287],[288]. By 86

doping with RE ions, such as Yb3+–RE3+ (RE: Er, Tm), strong visible emission can be produced by pumping the 2F5/2 level of Yb3+, which serves as the sensitizer for RE ions. The use of these fluoride based TGCs as UC spectral converters is expected to enhance the efficiency of solar cells by harvesting NIR photons. However, the low QY for the UC process (usually far below 10%) combined with the limited absorption cross-section of RE ions remains as a major technical challenge for practical applications of such GCs for UC spectral conversion in solar cells [289].

6.1.2 Whitelight LED In the widely used phosphor–converted white light LEDs (WLED), the organic resins used for fixing the powder phosphor often suffer from coloration due to heat induced degradation after a long period of service, which strongly limits the lifetime of such WLEDs. To circumvent this problem, GCs of pure inorganic nature have been identified as ideal materials that offer super stability against thermal degradation compared to organic resins (see Table 4). At present, YAG:Ce3+ is one of the most popular phosphors with the best performance used in commercial LEDs. The precipitation of YAG:Ce 3+ crystals in an oxide glass was first carried out by Tanabe et al. through the heat treatment of a SiO 2–Al2O3–Y2O3 parent glass at temperatures between 1300 C–1500 C. The resultant GC showed a similar emission spectra as that of powder YAG:Ce3+, and a QY of 30% was recorded [290]. This GCs is translucent and the luminescence efficacy was higher for GC containing larger YAG:Ce3+ crystals [291]. In a more recent step towards commercialization of the product, the highest value of luminous efficacy of the GC-based WLED of 73.5 lm/W was achieved [292]. Besides the YAG based GCs, the red-emitting Li2Ge4O9:Mn4+ could be also “synthesised” using the glass-ceramic route, i.e., by heat treatment of a parent glass with the composition of 20Li2O-80GeO2-xMnO2 [293]. The red-emission of this GCs can be used to modify the color-rendering index, as well as the color temperature of normal white-LEDs based on YAG:Ce3+. 87

c

a

Intensity (a. u.)

b

420 480 540 600 660 720 Wavelength (nm)

Fig. 23 W LED lamp using a PiG GC for spectral conversion. (a) Photographs of a WLED lamp caped with a PIG GC (left) and the lamp in operation (right). (b) Emission spectra (normalized at 450 nm) of the lamp using GC chips of various thickness. (c) The corresponding CIE color coordinates calculated from the spectra shown in b. Reprinted with permission from ref. [295], copyright 2014, WILEY–VCH. The YAG:Ce3+ based GCs require high processing temperature and a careful heat treatment, which, from viewpoint of commercialization, is not competitive in terms of cost. To avoid use of high processing temperature, the combination of a low melting glass with the YAG:Ce3+ phosphor appears to be a more economic choice. This technique for GC fabrication has been termed phosphor-in-glass (PiG), in which the phosphors are mixed with the powders of low–melting glass and sintered to form dense GCs [191][294][295][296][297][298][299][300]. The glass used in fabricating such GCs, like borosilicate glass, still have to be melted at temperatures above 1200 C, while the GCs can be formed by sintering at a much lower temperature, where the glass become softened, typically at 600 C  800 C. Unlike GCs obtained by direct crystallization, the sintered GCs can have a wide range of compositions, and therefore their spectral performance can be tuned by adjusting the type and concentration of the phosphors (see Table 4). As demonstrated in the work of Zhang et al. [295], the WLEDs fabricated by the combination of the GaN chip with the PiG GC 88

of various thicknesses generated tunable white light emission with a luminous efficacy up to 124 lm/W (Fig. 23). In addition, the WLEDs showed strong resistance to aging at 150 C. Due to the high stability, the PiG glass can be also used as a robust converting material for white laser light. When excited by a 450 nm laser diode, this laser light showed much enhanced brightness compared to phosphor-based ordinary WLEDs [298]. The PiG methods can be easily extended for the fixation of other phosphors in the glass matrix as long as there is no reaction between phosphor and the glass matrix. Therefore, this process offers an easy access to WLEDs with better color rendering index and lower color temperature [301][302]. In terms of optical properties, WLEDs based on YAG:Ce 3+ phosphors demonstrate a high color temperature. Warm white light LED are developed recently by replacing YAG based phosphor with, e.g., oxynitride phosphor, which emits at wavelength slightly longer than that of YAG:Ce 3+. Oxynitride based GC, where the matrix is a low melting borosilicate (4Na 2O–56B2O3–40SiO2) glass, was reported by Segawa et al. by using a similar sintering procedure [303]. The resultant translucent GC contained 4 mass% of the oxynitride phosphor and generated near white light under excitation by a blue LED at 450 nm. It was found that the quantum efficiency of the GC composite does not change with the concentration of the phosphor particles in glass. Besides borosilicate, metaphosphate, pure metal borate and tellurite can also serve as the low melting glass matrix for the incorporation of the oxynitride phosphor [304],[305],[306]. The fabricated GCs are all translucent and the optical properties, such as quantum efficiency, differ from sample to sample.

89

Table 4. Selected examples of (semi-transparent) GCs used for WLEDs. Glass composition

Crystalline phase

SiO2-Al2O3-Y2O3/Gd2O3 20Li2O-80GeO2-xMnO2 (x=0.1-0.7) 30-40SiO2 – 30-40B2O3 – 25-35 RO (R=Ba, Zn) (25SiO2-25B2O3-35ZnO)15Al2O3-5K2O 10–30 Sb2O3, 10–30 B2O3, 5–30 TeO2, 10–25 ZnO, 5–20Na2O, 0–10 La2O3, and 0–10 BaO PbO-B2O3-SiO2-ZnO 60(0.99CaO-0.01EuO)-40SiO

Ca2SiO4:Eu2+ Ca3Si2O7:Eu2+

2

Emission wavelength**

Ref.

YAG:Ce / GAG:Ce3+ Li2Ge4O9:Mn4+

Fabrication method Heat treatment Heat treatment

~543 -553 nm ~670 nm

[290] [293]

YAG:Ce3+ (30wt%)

PiG

~ 534 nm

[191]

LuAG:Ce3+ CaAlSiN3:Eu2+ YAG:Ce3+ (1-9wt%)

PiG

~540 nm

[294]

PiG

~550 nm

[295]

YAG:Ce3+ (6-30 wt %)

PiG + tape casting FS

~550 nm

[297]

~515 nm

[195]

*Compositions of the listed glasses are given in molar fraction (mol.%) unless otherwise stated. **Peak positions of the measured emission spectra

6.1.3 Laser gain materials TGCs have received particular interest for application as alternative lasing medium to transparent ceramics, as well as laser glasses, since they can be easily processed into large plates, preserving sufficiently high transparency. However, although strong emissions have been observed in different GCs, demonstrations of lasing operation are rare possibly due to the large scattering loss by the dispersed crystallites especially in the visible range. According to scattering law, the light loss can be drastically reduced in the IR region for GCs, and indeed lasing in IR has been realized in TM doped chalcogenide TGCs [307][308][309]. Martyshkin and co-workers [308] were the first to synthesize a As2S3:As2Se3 glass containing Cr2+:ZnSe micrometer crystals for IR lasing. This GC was prepared by co–melting the starting materials of As 2S3, As2Se3 and Cr2+:ZnSe crystals in vacuum and the subsequent heat treatment at 290 C. The obtained TGCs demonstrated a broad emission band due to the Cr2+: 5E  5T2 transition that covered the spectral range of 1900 nm – 2800 nm with a peak at around 2400 nm. By pumping at 1560 nm, laser oscillation was observed at 2350 nm with a threshold energy of 13 mJ. A similar chalcogenide TGC containing Cr2+:ZnSe crystals was prepared by Karaksina et al. by direct dissolution the Cr 2+:ZnSe crystal in the base 90

glass upon melting and the subsequent heat treatment [309]. Moreover, those authors produced fibers using this chalcogenide TGC, which have diameters ranging from 170 – 300 m and length from 2 – 20 m. The fibers demonstrated similar emission characteristics covering a spectral range from 1800 nm to 3000 nm as that of bulk GC and an optical loss of around 3 dB/m. Since the crystals in most GCs exhibit random orientation yet with homogeneous distribution, they can serve as random scatters as they usually have a higher refractive index than the glassy phase. GCs with proper control of the crystal size and luminescence management therefore can be an ideal solid-state platform to support random lasing. Xu et al. reported an oxyfluoride GC containing Ba2LaF7 NCs (doped with Yb 3+/Er3+) which enabled random upconverted lasing emission at the wavelengths of 520 nm and 540 nm (by the Er 3+: 2H11/2  4I15/2 and 4S3/2  4I15/2 transition) upon moderate pumping at 980 nm (Fig. 24) [310]. From the temperature dependence of the emission, the photon-assisted population inversion was believed to be responsible for the lasing. In addition, an optical waveguide of this GC (fabricated by etching) with a cross section of 2 m  1 m demonstrated a lasing threshold of 378 nJ/cm 2 and 536 nJ/cm 2 at temperatures of 200 K and 473 K, respectively.

Fig. 24 Lasing spectra recorded at different pumping energies for an oxyfluoride TGC at (a) 200 K and (b) 473K. The composition of the GC is 50SiO 2-10AlF3-5TiO2-30BaF2-4LaF3- xErF3 -1YbF3. A 980 nm CW laser was used as the pumping source. Reprinted with permission from ref. [310], copyright 2012, Wiley-VCH. 91

6.1.4 Optical amplification at communication wavelength region Optical amplifiers based on RE–doped materials have been widely used in optical communication across the world. The search for more efficient and broadband gain media has stimulated constant effort in the development of luminescent materials in the NIR window for optical communication. TGCs can be strong candidate materials as the host material for the optical gain medium used in the optical amplifiers due to their enhanced luminescence properties compared to common glasses (see Table 5). Particularly, oxyfluoride GCs, which contain fluoride NCs dispersed in the oxide matrix, have received growing attention. Fluorides as the hosts for RE ions are favorable for NIR emission due to the low phonon energy, while the oxide matrix offers at the same time high stability and mechanical strength (see, e. g., [283],[311],[312]). Such a combination has been accessed in plenty of glass systems, where the activators are NIR emitting ions such as Er3+ (emits at 1550 nm), or Pr3+ (emits around 1300 nm) [283]. For instance, A LaF3(:Er)–based TGC was prepared through the heat treatment of a parent glass SiO2–Al2O3–Na2O–LaF3 at temperatures around 750 C – 800 C. The GCs demonstrated enhanced gain spectrum in both the gain width and the gain coefficient as compared with the parent glasses in the spectral region from 1525 nm to 1575 nm [311]. Currently, the major obstacle for the commercialization of optical amplifiers based on these GC materials is primarily associated with the difficulties in the fabrication of GC fibers with minimal scattering loss and high luminescence efficiency [313].

92

Fig. 25 Local crystal field tuning for modulated NIR emission in Ni 2+–doped TGCs. (a) Absorption spectra (normalized) of the TGCs and the parent glasses containing NCs of Ga 2O3:Ni 2+, Ga2O3:Ni 2+/In3+, Ga2O3:Ni 2+/F– and Ga2O3:Ni 2+/F–/In3+. (b) Emission spectra (normalized) of the TGCs containing Ga2O3:Ni 2+ NCs doped with different ions (excitation: 800 nm). (c) The corresponding Tanabe–Sugano diagram. (d) Ideal [GaO 6] octahedron (left) and the distorted configuration by In 3+ (middle) and F– (right) substitution. (e) Crystal field splitting of the Ni 2+ (3d8) energy level in glass (tetrahedral) and TGCs (octahedral). Reprinted with permission from ref. [316], copyright 2013, WILEY–VCH. Compared to RE ions, TM ions offer a much broader luminescence spectral width in the NIR region, making them highly desirable as activators in broadband optical gain medium for optical communication as well as ultrafast lasers. Most of the 3d TM ions, such as Ni 2+, Fe3+ and Cr4+[315][316][317][318][319][321], are not luminescent or show only a very low emission yield in oxide glasses due to the unfavourable coordination and ligand field. For instance, the Ni 2+ ions are luminescent only in an octahedral coordination that is often found in oxide crystals but cannot be provided by common oxide glass, where Ni 2+ ions are usually four or fivecoordinated. By judicious selection of the glass composition, NIR emission of Ni 2+ in the GCs has been realized in 93

the several systems, where Ni 2+ ions found the six–coordination octahedral sites in the precipitated crystals. The first Ni–containing TGC was made by Samson et al. with a parent glass of SiO2–Ga2O3–Al2O3–K2O–Na2O–Li2O.[314] At 0.05 wt% Ni 2+ (in the form of NiO) doping, gallate spinel NCs precipitated in the glass after heat treatment gave broadband emission centred at 1200 nm with a FWHM of 250 nm from the transition of 3T2g(3F)3A2g(3F). Later, Ni 2+–doped oxide TGCs emitting in the NIR have been fabricated in different systems containing crystallites of LiGa5O8, –Ga2O3, MgAl 2O4 and etc.[314],[315],[316],[317],[318],[322],[323][324][325][326]. Unlike trivalent RE ions, the d–d transition of Ni 2+ is strongly dependent on its surrounding coordination environment, i.e., the type of crystalline host, thus enabling the tuning of emission spectrum by crystal field. For instance, in an oxide TGC containing –Ga2O3, the emission of Ni 2+ by partial substitution of Ga3+ and O2– with In3+ or F–. The effect of the substitute can be seen from the change in the absorption and emission spectra (Fig. 25) [316]. However, optical amplification has rarely been demonstrated in these materials possibly due to low emission efficiency and high scattering loss. The first successful optical amplification was reported by Zhou et al . in a –Ga2O3:Ni 2+ containing TGC, which recorded a gain coefficient of 0.283 cm –1 at 1300 nm (pumping: 980 nm, 1.12 W) for a plate sample [315]. Optical gain in TGC fiber was observed until recently based on a TGC containing LiTaO3:Ni 2+ NCs [271]. Due to minimized scattering loss, optical amplification with a gain of 1.4 dB was realized at 1300 nm. This TGC based gain fiber shows obvious advantage over RE-based counterparts due to its larger gain bandwidth (480 nm), which would be highly attractive for next-generation high-capacity optical communication systems.

94

Table 5. Selected examples of TGCs used for broadband optical amplification in the optical communication wavelength region.

Parent glass composition* 66.5SiO2-19.5Ga2O3-6.5Al2O3-7.5Na2O 25.53Li2O-21.53Ta2O5-35.29SiO2-17.65Al2O3 48SiO2–24Li2O–16ZnO–8Al2O3–3K2O–1P2O5–0.1Cr2O3 66SiO2–8B2O3–18K2O–4BaO–4ZnO–1.5PbS, 66SiO2–8B2O3–18K2O–4BaO–4ZnO–1PbS, 66SiO2–8B2O3–18K2O–6ZnO–2ZnS–1PbO

Crystalline phase Ga2O3:Ni2+

Gain wavelength 1272-1348 nm 1300 nm

LiTaO3:Ni2+ LiAlSi2O6:Ni2+ Li2ZnSiO4:Cr4+ 1275-1350 nm PbSe 1330 nm 1550 nm

Gain coefficient 0.283 cm-1

ref

1.4 dB**

[271]

1.27 cm-1

[329]

1.26 - 2.89 cm-1

[331]

[315]

*Compositions of the listed glasses are given in molar fraction (mol.%) unless otherwise stated. **Reported in a TGC fiber and the fiber length are not given in the publication.

Cr4+ is another TM ion that gives NIR emission in the optical communication window from the d–d transition of 3T2  3A2. NIR emission from TGCs with precipitated Cr 4+ doped crystals has been reported in serval systems, where Cr 4+ occupies the Si site in the SiO 4 unit in different silicate/germanate crystals, including silicates like Mg2SiO4 and Ca2Al2SiO7 [327], [328], [319], and germanate like Ca 2GeO4 [320]. Due to high rate of non–radiative process, the NIR emission from Cr4+ in these materials does not finally enable the observation of apparent optical gain. Until recently, optical amplification was realized in a Cr4+ doped Li2ZnSiO4–based TGC from 1275 nm to 1350 nm with a gain coefficient of 1.27 cm –1 in a pumping power of 0.8 W (808 nm) for a plate sample [319], [329]. Compared with RE or TM doped TGCs, QDs doped TGCs have been extensively studied and the emission of QDs can be finely tuned to cover different spectral region by choosing the right compositions (specific QDs) and heat treatment conditions (specific size). Lead based semiconductor QDs give NIR emission and the emission energy was tunable by adjusting only the size of QDs [125]. Wundke et al. first realized optical gain at 1.3 m band (1317 – 1352 nm) with a maxima gain of 80 cm-1 [330] using a QD-doped TGC. Dong et al. showed that in an oxide TGC containing 1 mol%PbS, the NIR emission emerged after the precipitation of PbS NCs by heat 95

treatment (Fig. 26). The peak of the emission spectra is tunable between 1300 nm to over 2000 nm depending on the size of the PbS QDs, thus enabling optical gain at ultra–broad region covering almost the entire optical communication window [331]. In their experiment, the optical gain b

Absorbance

a

400

c

800

1200 1600 2000 2400 Wavelength (nm)

1000

Optical gain, g(I/I0)

Intensity (a. u.)

d

2200 1400 1800 Wavelength (nm)

720

810 780 750 Pump Power (mW)

840

(defined as Io/Ii, output/input intensity) exceeded 2.0 at 1550 nm (Fig. 26d). Fig. 26 PbS–doped TGC for NIR photonics. (a) HR-TEM image of the TGC containing PbS NCs. The inset shows the lattice fringes of a single PbS crystal embedded in the glassy matrix. (b) Absorption spectra of the glass and TGC samples accessed by heat treatment at different conditions. (c) Emission spectra of the PbS doped TGCs prepared at different conditions. (d) Optical gain as a function of pumping power measured at three different wavelengths. The inset shows the power dependence of the optical gain. (Compositions for P1: 66SiO 2 – 8B2O3 – 18K2O – 4BaO – 4ZnO – 1.5PbS, P2: 66SiO 2 – 8B2O3 – 18K2O – 4BaO – 4ZnO – 1PbS and Z1: 66SiO 2 – 8B2O3 – 18K2O – 6ZnO – 2ZnS – 1PbO) Reprinted with permission from ref. [331], copyright 2011, Elsevier.

6.1.5 Scintillator and related applications Compared with single crystals, luminescent TGCs have been regarded as cost-effective materials with almost uncompromised optical properties for the detection of high energy photons 96

(i.e., X-ray, -ray), neutrons as well as charged particles (i. e., cathode ray) in nuclear facilities and different medical applications [332][333][335][338][335][339][333][336]. From the beginning of the 21st century, there has been growing interest in the use of oxyfluoride TGCs as scintillators due to their high photon yield associated with the low phonon energy of the precipitated fluoride crystals and their high stability [332]. For instance, a recent work of Guo and co-workers showed that the X-ray excited luminescence was greatly enhanced in an oxyfluoride glass after the precipitation of KLu2F7:Tb3+ crystals [333]. Besides X-ray detection, the addition of 6Li or 10B isotopes can render the TGC highly sensitive to thermal neutrons. Struebing et al. recently reported the use of a 6Li-containing CaF2 oxyfluoride TGC as a neutron scintillator. It was found that the scintillation light yield of the TGC was enhanced by over 46 folds compared to the parent glass [334], and was comparable to commercial products. The enhancement of emission yield was also observed for similar oxyfluoride TGCs in cathodoluminescence [332]. Due to the relatively high photon yield and the low fabrication cost for large panels, TGCs become of particular interest for application in medical radiography, for instance, as recoverable X-ray storage plates or X-ray imaging panels. Edgar et al. found that X-ray irradiation of a BaBr 2 containing TGC doped with Eu2+ and Ce3+ resulted in significant photo-stimulated luminescence peaking around 485 nm (for Eu 2+) and 425 nm (for Ce 3+) [335]. This result therefore makes it attractive as an X-ray storage phosphor that can be a potential candidate to replace the conventional silver-halide based films. In the conventional X-ray imaging panels, polymer composite containing inorganic phosphors (or pixelated cesium iodide (CsI) crystals) are used as scintillator materials, which are coupled to a Si-based photodiode that convert the visible emission to electric signals. Apparently, single crystals are expensive, while the polymer based composites become opaque at large thickness, hence requiring the use of higher dose of radiation to acquire an acceptable imaging resolution [336][337]. Therefore, luminescent TGCs appear to be an ideal, low-cost solution for such applications. 97

Johnson and co-workers reported a Eu2+-doped fluorochlorozirconate for use as X-ray imaging panel [337], which showed image quality comparable to commercial crystalline material (CaWO 4). As reported by Okada et al. [340], an oxyfluoride GC containing CaF2:Sm3+ (composition: SiO2–Al2O3–CaF2–CaO–SmF3) was used as a dosimeter to characterize the microstructured X-ray beams employed in micro-beam radiation therapy (MRT) (see Fig. 27). X-ray irradiation leads to the reduction of Sm 3+ to Sm2+, resulting in an emission band peaking at 720 nm owing to the 4f55d17F0 transition. This TGC offers a high spatial resolution down to the micrometer scale and the detection dose range can reach several thousands of grays. Furthermore, the TGC based dosimeter is reusable since the Sm 2+ can be re-oxidized to Sm 3+ by heating the X-ray irradiated GC to high temperature or by a UV treatment.

Fig. 27 An oxyfluoride TGC containing CaF 2:Sm3+ for X-ray doesmeter. (a) Normalized response as a function of dose. Inset is the spectrum of the X-ray used for irradiation. (b) 98

Distribution of X-ray does recorded using a confocal microscope. (c) PL spectra of Sm 2+ recorded during heating the GC to different temperatures (top), and during exposure to UV for different durations. (d) PL intensity of Sm 2+ (excited at 633 nm) as a function of X-ray does for the TGCs before and after erasure by the two different processes. Reprinted with permission from ref. [340], copyright 2012, Wiley-VCH.

6.1.6 TGCs with persistent luminescence Different from the photoluminescent TGCs, long persistent luminescence can also be generated by TGCs containing the delicately designed crystalline phase doped with RE or TM ions. For instance, Tanabe and co-workers developed several TGCs containing crystalline phases, which are capable of generating long lasting phosphorescence [341], such as Zn1+xGa2-2xGexO4:Mn2+ (0x1) and SrAl 2O4:Eu2+, Dy3+ (fabricated by FS, see section 5.2.3). Interestingly, the GCs containing Zn1+xGa2-2xGexO4:Mn2+ NCs demonstrate controllable, multi-color persistent luminescence due to the changing of crystal field around Mn 2+ in the devitrification process upon annealing.

6.2 TGCs containing optical functional crystals for nonlinear optics and photonics Besides luminescent TGCs, optically functional crystals in glasses include nonlinear optical crystals, semiconductors, plasmonic crystals and etc. The presence of these crystals has enabled the application of TGCs in nonlinear optics and ultrafast photonics.

6.2.1 TGCs for second harmonic generation (SHG) The precipitation of functional crystals in glasses has enabled novel applications of these TGCs in optics and relevant areas. Especially, the precipitation of highly oriented optical active 99

crystals can lead to optical functionalities that are not found in ordinary glasses. Nonlinear optical crystals, such as BBO, BTS and KTiOPO 4 (KTP), have been precipitated inside various glasses and used for the demonstration of SHG (see, e.g., refs. [71],[72],[73],[100],[342]). Due to the precipitation of the SHG crystals, photon upconversion can be generated by pulse laser irradiation of the TGCs. For instance, in a BTS based GC, in which the crystal phase have a size of approximately one micrometer, blue, green and red upconversion by SHG can be produced upon irradiation with a fs laser beam (see Fig. 28) [343]. Importantly, this TGC can generate any color with near monochromatic color purity based on SHG. This result implies that, despite that crystals are randomly orientated, SHG can still occur in the absence of strict phase matching under pulse laser excitation [344]. The TGCs containing SHG crystals are expected to find application in 3D volumetric display based on multi–color UC of NIR photons [345].

Fig. 28 SHG from a TGC containing BTS crystals. (a) Spectrum of the upconverted luminescence from the focal area upon irradiation by an 800 nm fs laser. Insets are the power dependence of the spectral intensity of the SHG signal, and photographs of the sample under irradiation by an 800 nm fs laser. (b) Temporal intensity profile of the SHG radiation. (c) SHG spectra recorded from the focal area upon excitation at 900 nm, 1080 nm and 1230 nm. The inset images shows clear blue, green and red light radiated from the focal area. (d) The CIE chromaticity diagram showing the 100

color coordinates calculated from the spectra shown in c. (e) Simulated electric field E (upper panel) and E2 (middle panel) intensity distribution along the light path, as well as E2 (bottom panel) intensity distribution. The simulation was made by assuming a Gaussian beam profile. Reprinted with permission from ref. [343], copyright 2017, Nature Publishing Group. Quantitative assessment of the SGH performance of the GCs is usually made with the Maker fringe technique and the powder technique based on the theories developed by Kurtz–Perry (also known as Kurtz–Perry method) [346],[347]. Both of the techniques employed a reference (–quartz powder for Kurtz–Perry method, and single crystal for Maker fringe method). The theories and the experimental setup of the measurement have been described elsewhere [346],[347],[348],[350],[348]. The powder technique measures directly the SHG intensities of the pulverized TGCs compared to that of –quartz powder, and it is usually employed as a preliminary but fast assessment of the SHG performance. In comparison, from the Maker fringe measurement, the optical second–order optical nonlinearity can be derived. Komatsu’s group used this method for the study of a number of TGCs containing SHG crystals [348],[350],[348]. For instance, in the TGC containing orientated Ba 2TiGe2O8 (BTG) crystals, the second order nonlinearity d33 was found to be ~10 pm/V [350], which was comparable to d22 and d31 of LiNbO3. In another investigation, the d33 of surface crystallized TGCs of Ba2TiSi2O8 (BTS), Sr2TiSi 2O8 (STS) and Ba2TiGe2O8 (BTG) were found to be 132 pm/V, 7.20.4 pm/V and 223 pm/V, respectively [348]. The difference in the microstructure of the samples is believed to account for the varied d33 values obtained in different experiments. Whereas the above techniques measure the SHG signal from an ensemble of crystals distributed in the glass matrix, the micro–SHG (–SHG) method allows for the mapping of SHG intensity distribution generated by a single crystalline particle. This technique was based on a micro–Raman system equipped with a pulse laser for SHG and CW laser for Raman mapping. This technique was applied by Rodriguez et al. to study the TGCs containing different SHG crystals 101

[351],[352]. In a recent report [351], a LaBGeO 5 based TGC containing spherulitic particles of micrometer scale with radial distribution of crystallite was examined. It was found from a polarized –SHG measurement that the SHG intensity maxima orientated perpendicularly to the c–axis of the crystals (see Fig. 29), which matched the radial anti-ferroelectric orientation along the c–axis of each crystallite contained in the spherulitic particle.

a

b

c

Fig. 29 SHG from a single LaBGeO 5 spherulitic particle dispersed in a crystallized glass. (a) –SHG mapping of a spherulite probed with the excitation and the collected SHG signal polarized vertically. (b) Mapping of the Raman signal, indicating the c–axis orientation of the crystallites. (c) Combination of the –SHG and Raman mapping. Reprinted with permission from ref. [351], copyright 2015, American Institute of Physics. Besides oxide TGCs, the precipitation of SHG crystals has been reported in a number of chalcogenide TGCs, which are believed to find applications in IR photonics [58],[59],[353],[354],[355],[356],[357],[358],[359],[360]. In the single–component GeS 2 system, –GeS2 crystals can be precipitated at temperatures below 497 C. This GC demonstrated a clear SHG with a second order optical nonlinear susceptibility (2) of around 7.3 pm/V [59]. In multicomponent systems, different chalcogenide crystals have been precipitated with much enhanced nonlinear susceptibility and glass stability. For instance, the same –GeS2 crystals were crystalized in a Ge23Sb11S65Cd1 glass at the temperature of 480 C [355]. As the nonlinear susceptibility of precipitated crystals is responsible for the observed SHG of the bulk GCs, there is 102

continued effort in the search for a suitable parent glass, in which a crystalline phase with complex chemical composition but better SHG could precipitate [356],[357],[358],[359],[360]. For instance, the IR nonlinear crystal Li 2Ga2GeS6 was directly precipitated from a parent glass of 40GeS2–30GaS1.5–30LiS0.5, demonstrating a SHG intensity of 0.35 of the reference Z–cut quartz. On the other hand, laser writing of different SHG crystal patterns has also been performed in several glass systems, as discussed in section 5.3. The SHG from such crystalline patterns is expected to find further applications in fields such as, waveguide, integrated optical circuit and relevant fields. Besides the SHG waveguide inscribed in glasses, TGC fibers containing SHG NCs of BTS (Ba2TiSi2O8) have been fabricated by Fang et al. using the MiT method (see section 5.9). A proof-of-principle demonstration showed that the TGCs fiber efficiently converted the 1030 nm fundamental excitation to the frequency-doubled 515 nm SHG emission [361].

6.2.2 Metal and QD doped TGCs for third–order optical nonlinearity Noble metal NPs as well as semiconductor QDs exhibit the attractive third order nonlinear optical properties, which lead to potential applications in ultrafast optical switches and modulators. The third-order nonlinear susceptibility ((3)) is also known as the Kerr susceptibility, which is the nonlinear part of the total refractive index. It should be noted here that optically isotropic materials naturally exhibit small (3) values, which can be greatly enhanced through the introduction of absorbers like metal NPs [362]. The noble metal NPs doped glasses can be accessed by a conventional GC technique for oxide and non–oxide glasses and the third order nonlinear property is usually measured by a degenerate four–wave mixing (DFWM) and Z-scan techniques. Similar to the linear absorption spectra, (3) also depends on the size of the particles, and the maxima (3) is in the order of 10–7 esu at the wavelength of plasmon resonance for Ag and Cu NPs dispersed in a silicate glass. Furthermore, it was found that the (3) values are almost independent of the linear absorption coefficient, but increase with the particles size. These observations can be understood by 103

considering the size dependence of the imaginary of the dielectric constant of metal NPs and the local field behavior [363]. From pump–probe experiment, the noble metal NPs doped TGCs exhibited fast nonlinear optical response in the order of several picoseconds due t o fast cooling of hot electrons. Glasses doped with different QDs are of particular interest due to the attractive third order nonlinear optical properties associated with the quantum confinement effect of electrons and holes. Furthermore, glass as an insulator itself provides additional confinement for the QDs in a 3D manner. Different metal chalcogenide QD doped oxide glasses have been investigated and the third order nonlinear coefficient susceptibility ((3)) varies between 10 –6 and 10–10 esu depending on crystal size and concentration. Compared with metal chalcogenide, the Bohr radius of cupper halides, such as CuBr and CuCl, are much smaller (CuBr 1.25 nm, CuCl 0.65 nm), and therefore they are predicted to show much higher value of nonlinear susceptibility [364], [365]. The size dependence of optical properties was confirmed in both linear absorption and non–linear absorption. Moreover, it was found by Li et al. based on DFWM that the figure of merit (FOM, defined as (3)/, where  is the linear absorption coefficient) increased with the size of the QDs, following a R0.4 dependence [366]. In addition, the spectral measurement also showed the R2.0 dependence of the oscillator strength, implying the huge oscillator strength effect on the confined exciton generated in the NC. Kondo et al. found that the FOM for a CuCl doped glass was one order of magnitude higher than that of the metal chalcogenide QD doped glass [365]. Concerning quantum confinement, the translational motion of the excitons is confined when the size of the QD is larger than 4ab (ab Bohr radius of the exciton) for Wannier excitons [364]. The confinement of the exciton wavefunction leads to the increase in (3), which has been experimentally verified in different QD–doped glasses [366],[367]. When the wavelength of light is far larger than the size of QD, the excitons are coherent in the crystal and therefore have a microscopic polarization. The radiative decay rate, which determines the response time, can be described by the reciprocal of 104

radiative lifetime 1/ r, which was predicted to be proportional to (R/ab)3. Experimentally, the R2.1 dependence of the radiative decay rate was found for both CuCl and CuBr doped glasses, which supports the theoretical prediction [119][364][368]. The nonlinear optical absorption of TGCs doped with semiconductors is often assessed with the open and close aperture Z-scan method, from which the third order nonlinear parameters can be derived. This method has been performed in different TGCs doped with plasmonic metal NPs and semiconductor QDs. The samples often exhibit optical limiting behavior and the effect becomes pronounced with the growth of semiconductor NPs [132]. From the calculation, the third order nonlinear susceptibility was increased by over 1000 times compared to the parent glasses. Moreover, these TGCs demonstrate response time in the order of sub–picosecond range, which belong to the resonant case due to electronic transitions. Such TGCs are considered to have potential applications in ultrafast optical switch and photonics applications. Since the NLO absorption strongly depends on size as well as concentration of the NPs dispersed in glass, it becomes possible to manipulate the NLO absorption by controlled heat treatment. According to a recent report by Xiang et al., a Cu-precipitated glass demonstrated optical limiting and saturable absorption depending on the heat treatment condition (see Fig. 30) [369]. For the three types of glasses examined, the “red” sample contains Cu0 NPs, the “blue” sample contains only Cu2+, and the “green” contains both of them. The results suggest that both Cu NPs and the unreduced Cu2+ ions contribute to the NLO absorption, while optical limiting was only observed for the samples, in which all Cu2+ ions are reduced to atomic Cu 0 (Fig. 30).

105

a

b

red d

c

e

red

Blue

Green f

Green

Blue

Fig. 30 Nonlinear optical properties of Cu NPs implanted glass. (a-c) Open-aperture Z-scan curves and (d-f) closed-aperture Z-scan curves. “Red”, “Green” and “Blue” are glass samples prepared by heat treatment in hydrogen atmosphere for 10 h, 5 h and 0 h, respectively. Reprinted with permission from [369]. Copyright 2015, American Chemical Society. Similarly, chalcogenide glasses can also be doped with noble metal (such as Ag) NPs to enhance the thirdorder nonlinear optical properties. Different types of chalcogenide glasses have been studied and the plasmonic absorption of noble metal NPs was found to be responsible for the enhanced nonlinear optical properties, same as that of oxide glasses. For instance, Liu et al. developed an Ag NP doped 56GeS 2–24Ga2S3–20KBr chalcogenide glasses by ion implantation process. They found a clear dependence of the linear and nonlinear properties on the concentration of Ag NPs [129]. The doping of gold NPs to these chalcogenide glasses exerted a similar effect on the nonlinear optical properties arising from the plasmonic response [130][370]. Furthermore, partial crystallization of chalcogenide crystals was also found to significantly enhance the third order nonlinear optical properties [371][372].

106

6.2.3 Saturable absorbers for pulse laser Saturable absorbers (SA) are widely used in pulse lasers operating at wavelength region from visible to the MIR region with pulse durations from ns to sub-10 fs [373][374][375]. The generation of laser pulse is based on the passive Q-switching or mode-locking mechanisms, in which the SA works as the passive pulse compressor due to its strong intensity-dependent photobleaching process. As listed in Table 6, TGCs containing NCs doped with TM ions (i.e., Co 2+) show broad absorption bands in the NIR and MIR region and exhibit strong nonlinear absorption. In a Co 2+ doped TGCs containing ZnAl 2O4 gahnite nanocrystals, for instance, the broadband NIR absorption at 1200 – 1600 nm is ascribed to the 4A2  4T1 (4F) transition of tetrahedrally coordinated Co 2+ and its bleaching relaxation time is in the range of 500 – 800 ns depending on Co 2+ concentration [376]. Since the absorption of 4-coordinated Co 2+ overlaps the NIR emission of Er 3+ (by the transition of 4I13/2  4I15/2), the Co2+ doped TGCs have been used for pulse compression for different Er3+ doped crystal and glass lasers. The first doped Q-switching pulse laser enabled by a Co2+ doped TGC was reported in 1998 by Boiko and coworkers [377]. In their work, the use of a Co2+ doped GC containing gahnit (ZnAl 2O4) crystals led to the generation of Q-switched pulse from both a Er:glass and a Nd:YAG laser. Later in 2001, Wang et al. reported also an Er:glass laser, which produced Q-switched pulse of 5.5 mJ energy with 80 ns pulse duration. Following this report, similar TGCs containing different Co 2+ doped crystals have been developed with comparable performance [378][379][380][381][382][383]. For instance, a recent report by Loiko et al. demonstrated saturable absorption of TGCs containing a mixture of Co 2+:β-Zn2SiO4 and Co2+:ZnO NCs. At 1.54 m, the TGC characterized a low saturation fluence of 0.75 J/cm 2, a short recovery time of 830 ns, and a high laser damage threshold of 14 J/cm 2 [379]. This TGC was used as a SA in the diode-pumped Er,Yb:glass laser, which generated Q-switched pulse of 0.77 mJ energy with a pulse duration of 45 ns. The same author reported a similar TGC containing Co 2+:γ-Ga2O3 NCs, which showed similar absorption characteristics and was used as SA for Q-switched lasers based on 107

a Er,Yb:glass [380]. Alternatively, the integration of SA with an optical gain medium is appealing as it would greatly simplify the laser cavity design, and TGC appears to be a possible material solution. Yu et al. developed a oxyfluoride TGC containing Er:YF 3 and Co2+:ZnAl 2O4 NCs, where the Er:YF3 NCs served as the gain medium and the Co 2+:ZnAl 2O4 NCs acted as the pulse compressor. This TGC was expected to find application in self-Q-switching laser, but laser output was not reported [383]. Since the lifetimes of the 3d level of TM ions like Co 2+ are in the range of micro- to nanoseconds, further pulse compression into ps or fs range based on the mode-locking mechanism is not possible by these GCs. To realize ultrafast mode-locking pulse generation, there has a constant pursuit for an efficient SA with fast response time and large modulation depth due to NLO absorption. As discussed in section 6.2.2, QD-doped glasses demonstrate strong NLO absorption (with a negative NLO absorption coefficient) and high third-order susceptibility in the NIR and MIR spectral range. The use of QD-based SAs for ultrafast pulse generation was started in the last century [384][385] (see also Table 6). Compared to polymer composite or QD-doped films, incorporation of QDs in an inorganic glassy matrix ensures high stability that is indispensable for the generation of high energy pulse with long term stability. Furthermore, the strong size-dependent absorption band of QDs offers a broad working bandwidth that covers the NIR and MIR range (for lead based QDs). The first experimental demonstration of a QD-glass based SA was reported by Guerreiro et al. in 1997 [386]. Based on a two level absorption saturation model, they found that the PbS doped glass demonstrated a saturation intensity of 0.18 MW/cm 2. Their laser was based on a Cr: Forsterite crystal (Cr4+:Mg2SiO4) and the generated mode-locked pulse characterized a pulse duration of 4.6 ps at 110 MHz and a tenability range of 1207 – 1307 nm. Two years later, Bilinsky et al. reported a mode-locking Ti: sapphire laser with a commercial CdTe-doped glass, which operated in the 780–860 nm range with a pulse duration down to 2 ps [387]. By using lead based QDs such as PbSe, saturable absorption can be easily extended to longer wavelength for MIR laser 108

pulse generation. For instance, Denisov et al. demonstrated a MIR pulse laser operating at 2.09 m based on a Cr3+,Tm3+,Ho3+:Y3Sc2Al3O12 crystal and a PbS doped glass as SA. The output pulse characterized duration of 290 ps with an energy up to 0.5 mJ [388]. Up to now, by using QD-glass as SA, dozens of mode-locked as well as Q-switched pulse lasers operating in the NIR and MIR range have been realized using laser crystals (or glasses) activated with RE and TM ions [389][390][391][392]. Table 6. Selected examples of TGC-based SAs for pulse generation in NIR and MIR lasers. Parent glass composition*

Crystalline phase

MgO-Al2O3-SiO2-TiO2-CoO

13Li2O–23Ga2O3–64SiO2 Not reported

MgAl2O4: Co2+ ZnO: Co2+ Zn2SiO4: Co2+ Ga2O3: Co2+ PbS

RG-780, RG-830, RG-850** SiO2–Al2O3–NaF–Na2O–ZnO P2O5–Na2O–ZnO–AlF3–Ga2O3

CdTe PbS PbS

12K2O – 28ZnO – 12Al2O3 – 48SiO2-CoO

Laser gain medium / operating mode Er:glass/ Q-S

Operating wavelength

Pulse parameters

Ref.

1.54 m

80 ns / 5.5 mJ

[378]

Er,Yb:glass / Q-S

1.54 m

45 ns / 0.77 mJ

[379]

Er,Yb:glass / Q-S Mg2SiO4:Cr4+ / M-L Al2O3:Ti3+ / M-L YAG:Cr4+ / M-L Y3Sc2Al3O12: Cr3+,Tm3+,Ho3+

1.54 m 1207 – 1307 nm

25 ns / 1.75 mJ 4.6 ps / 110 MHz 2 ps 10 ps / 235 MHz 290 ps / 0.5 mJ

[380] [386]

780 – 860 nm 1460 – 1550 nm 2.09 m

[387] [390] [388]

Q-S: Q-Switching pulse, M-L: Mode-Locking

*Compositions of the listed glasses are given in molar fraction (mol.%) unless otherwise stated. **From Schott Glass Technologies.

6.2.4 GC containing plasmonic oxide NCs for electrochromic window An electrochromic glass window, which can change color and block the infrared irradiation in the presence of an electric field, could potentially reduce the energy consumption in buildings [393]. This strategy was realized recently by using an NbOx glass doped with indium tin oxide (ITO) NCs. This nanocrystal–in–glass composite was synthesized by a sol–gel like process. In order to be electrically active, the organic surface capping species of the ITO NCs derived by a wet –chemistry process was replaced by inorganic niobium based polyoxometalate (POM) clusters. The following condensation process by solvent evaporation resulted in the densely packed ITO NCs embedded in a POM matrix (Fig. 31a), which can be converted into an amorphous NbOx by heating at 400 C. This process therefore allows for a facile control of the composition of the GC, which otherwise is 109

not possible by conventional GC procedure. The electrochromic behavior of the composite was demonstrated by using lithium metal as the anode and 0.1 M LiClO 4 in anhydrous propylene carbonate as the electrolyte. When electrochemically reduced, the composite film showed strong absorption from red to NIR spectra region (Fig. 31b,c), and the electrochromic behavior was completely reversible. The origin of the electrically controllable absorption was ascribed to the localized surface plasmon resonance of ITO NCs due to the tunable conduction electron density [394]. Although the prototype device used liquid electrode and lithium metal as anode, which might hinder practical application, this special type of GC was anticipated to find applications as smart windows to control the amount NIR and visible light passing them.

Fig. 31 Electrochromic window based on a crystal–in–glass composite. (a) ITO NCs covalently linked to amorphous NbOx. The [NbO 6] octahedral units are shown in green and the ITO NCs are blue. In the inset, the niobium atoms are located at the center of the green octahedra, oxygen atoms are red and indium atoms are blue. (b) Transmittance spectra of a typical ITO–in–NbOx film, under different applied electrochemical voltage (versus Li/Li). (c) Absorbance change (at 500 nm, between 1.5 V and 4 V) as a function of the ITO volume fraction. (d) The use of electrochromic glass as a smart window for controlling the transmittance of the NIR and visible light. Left: The window is transparent to NIR and visible light without an electric load. Middle: Under an intermediate voltage the window block most of the NIR light. Right: Under high voltages, 110

the window blocks both NIR and visible light. The bottom shows schematically the design of the electrochromic window. The crystal-in-glass composite is coated onto one side of the window and serves as the electrode. Reprinted with permission from ref. [393] and [395], copyright 2013, Nature Publishing Group. 6.2.5 Photocatalysis GCs containing semiconductor NCs that are active in photocatalysis are of special interest as the fixation of the semiconductor crystals in a glass matrix could solve the problem of photocatalyst recovery. TiO 2, as one of the most popular semiconductor photocatalysts, has been precipitated in different types of silicate and borate glasses for photocatalytic applications [396][397][398][399]. Yoshida et al. developed a TiO 2 crystallized glass based on a borosilicate parent glass with the composition (in mol%) of 14TiO 2–23ZnO–45B2O3–18Al 2O3–4.5SiO2. This TiO2 based GC was used for photocatalytic water splitting, and their results showed that the nanostructuring of the GC surface by chemical etching resulted in remarkable improvement in the photocatalytic performance [397]. A more recent work reported by the same group demonstrated that the introduction of Ti3+ lead to enhanced absorption in the visible region and boosted the catalytic performance for water splitting under visible light irradiation [398]. Besides water splitting, the GC containing photocatalytically active semiconductors including TiO2, –Bi2Mo2O9 and BiVO 4 have been reported by Gad–Allah and coworkers and used for the photocatalytic degradation of organic pollutants in aqueous solutions [399] [400]. 6.2.6 Microwave absorption Besides photoluminescence, aliovalent doping increases the carrier density of a crystal and consequently results in the shift of plasma frequency to higher energy, usually in the NIR and mid–IR range. As discussed earlier, the dopant can always enter into the designed crystalline phase 111

in case that a suitable lattice is available in the crystals. In a ZnO based GC, the addition of Sb 2O3 to the parent glass finally resulted in the precipitation of ZnO:Sb, causing strong plasmonic absorption at spectral region from NIR to midIR and far–IR [76]. In addition to the strong IR absorption, the GCs also exhibit high microwave susceptibilities. At microwave frequency (2.45 GHz), the GC demonstrate very high absorption due to the presence of conduction electrons in ZnO:Sb crystals. This type of GC may have potential applications for microwave optics and relevant fields.

6.2.7 TGC for enhanced IR radiance TGCs with high IR emissivity are of particular interest for applications in architecture for energy saving. One of such GCs was synthesized by Wang and coworkers using titanium slags and MgCO3 as raw materials, and the crystalline phase was an iron-ion substituted cordierite, (Mg,Fe)2Al4Si5O18 [401][402]. This GC demonstrated favorable IR iradiance and CTE. Later, by using pure oxide as the starting materials [403], it was found that the presence of iron oxide could improve the IR emissivity in the region of 2-5 m for the GC based on a parent MgO–Al 2O3–SiO2 glass. The possible mechanisms behind the improvement in the IR emissivity might be ascribed to the absorption caused by d-d transitions of Fe 2+ and Fe3+ ions in a tetrahedral coordination.

6.3 Miscellaneous optical applications 6.3.1 TGC as optical components In general, GCs exhibit superior mechanical strength and tailorable CTE, and therefore have the potential to replace glassy materials in optical components as long as high transparency is preserved. A nepheline based (Na 3(Na,K)[Al 4Si4O16]) TGC was developed for the application as a bandpass filter used in telecommunication systems by Asahi Glass Co., Ltd. [404]. In order to obtain a high CTE expansion required by the application, the original nepheline based TGC was modified to incorporate higher amount of K 2O, and the addition of nucleation agent (ZrO2 or TiO2) 112

in a moderate amount was found to be crucial to initiating the nucleation process. The obtained TGC containing nepheline and kalsilite crystals was highly transparent and demonstrated much improved mechanical properties compared to glassy materials for the same application. The details of the device performance were not discussed in their work. Luminescent planar waveguide made up of crystal–in–glass composite of high transparency have found wide applications, from integrated optical amplifiers, laser systems to spectral converting layers for photovoltaics. Therefore, there is a continued search for a suitable glass compositions and fabrication methods that enable low loss and high luminescence efficiency. The preference for TGC for use as planar waveguide over glass is associated with the superior spectroscopic properties (see, e.g., ref. [405]). With respect to fabrication, the most popular process to fabricate the planar waveguide is the sol–gel process, which is usually followed by a heat treatment for crystallization. For such luminescent waveguide, RE–doped systems again receive particular interest due to their favorable optical properties. For instance, Jestin et al. developed a silica–hafnia GC film (as a planar waveguide) doped with Er3+ [406]. This waveguide demonstrated strong NIR emission and enhanced lifetime for the Er 3+ 4I13/2 level. Combined with a low loss of 0.3 dB cm–1, this GC waveguide is expected to be applied as C–band amplifiers. In comparison, a more “direct” approach was reported by Fujiwara and coworkers for the fabrication of TGC -based waveguide. They showed that the highly oriented crystalline layer in the surface crystallized TGCs containing SHG crystals of BTS (or STS) demonstrated measuredan ultralow propagation loss down to 0.6 dB/cm [407]. As the CTE of TGCs can be carefully controlled by adjusting the glass composition, TGCs with zero or even negative expansion become achievable by the selection of the right composition. The low thermal expansion is one of the most important requirements for large–sized optical lens used by telescopes in observatories, which are often located in remote areas that experiences large temperature fluctuation. In addition to high transparency, materials used for lens should be stable 113

against corrosion by rain, frost and ice. Certain types of TGCs can meet the above requirement, and importantly they can be processed to the desired size and shape by modern technologies. Zerodur, for instance, is one of these TGCs based on –quartz solid solution, which exhibits almost zero thermal expansion. This GC is produced from a parent glass based on SiO 2– P2O5–Al2O3–Li2O combined with a small amount of network modifiers and nucleation promoter. Besides the adjustable thermal expansion, this TGC is characterized by Knoop hardness of 0.1/20 (ISO 9385) of 620, Young’s modulus of 90.3 GPa (20 °C), and bending strength of approximately 110 MPa [408]. Different from large sized TGCs, photo–structured TGCs have found applications as microlens and arrays. This type of lens with a thickness of 20 m and diameter of several millimeters are widely used as optical components in electric and optoelectronic devices. Photo–structurable TGCs, such as fotoceram based on lithium silicate, have been adapted by Borrelli and Morse for the fabrication of integrated lens arrays (see e.g., refs. [7], [46]). In their process, the parent glass was first prepared in a thickness of around 20 m, and it was subjected to UV irradiation through a pattern to induce the formation of metal colloid. After the following crystallization process, the cylindrical areas protruded on both side of the glass film since the density of crystal was higher than that of the glass phase. Additional shaping of the lens arrays could be applied during the process depending on the specific applications.

6.3.2 Radome dome materials Materials used for Radomes that protect Radar equipment in the nose of missiles and aircraft have to survive extreme conditions, such as high temperature and high speed, and they must be transparent in the radio frequency. Therefore, candidate materials must have a very homogeneous and low dielectric constant and a low CTE, low dielectric loss, high strength, and high abrasion resistance. TGCs have been regarded as alternative for ceramics in such applications considering their favorable performances and cost. In addition, the transparency of the TGC also facilitates 114

quality control as defected areas can be easily observed. In 1959, Stookey developed the first commercial GC for Radome based on the Corning 9606 cordierite GC with the major composition of SiO2–Al 2O3–MgO.[7] The nose was fabricated by a centrifugal casting technique from a base glass using processes from glass spinning technology. Unlike ceramic materials, GCs are free of porosity and, exhibit high resistance to climatic effect and seawater corrosion. Compared with conventional technologies, ceramic methods based on casting and sintering could offer significant advantages in Radome manufacturing [409]. Since the use of ceramic process allows for the combination of material sintering and crystallization of glass, the procedure for GC fabrication can be greatly simplified and the process yields semi-finished products with minimal variation in shape. In addition, different process can be applied to obtain green products, which can be made of the glassy powder synthesized from either conventional melt-quenching or sol–gel process.

7. Conclusion and future prospects

GCs have shown to be an unique class of composite materials, in which the incorporation of crystalline particles not only enhances mechanical strength, but also creates new functionalities that are not found in their parent glasses. The preservation of high transparency makes TGCs highly competitive in the substitution of conventional glass and transparent ceramics for applications, especially in optics and photonics. GCs by their very nature are crystalline particle dispersion in a glassy solvent, and their crystallization kinetics therefore can be described by the slightly modified CNT. Accordingly, we propose a new criterion for categorizing GCs depending on whether or not glass network participates in crystallization, and we find two distinct types of GCs. In the first type of GCs, the glass network formers crystallize, such as silicate–based GC, and in the second type of GC, the glass network formers only servers as the solvent for the precipitation of network modifiers, such as fluoride and metallic NPs. These two types of GCs demonstrate di stinct 115

crystallization behaviour and functionalities, leading to quite different properties and applications. In general, the first type of TGCs is featured by enhanced mechanical properties, while the second type is usually characterized by functionalities that are directly linked to the precipitated crystals, e.g., optical properties. Similar to the wet–chemistry process, different types of crystals can be precipitated in glassy phase through different processes. Direct reaction in the solvent phase, here in the solid-state glass, remains as the most popular approach for synthesis of NCs in glasses. In comparison, sol–gel and sintering of glass powder have been applied as a cheap yet easy and versatile alternative for TGCs, albeit at the cost of mechanical strength and transparency. Furthermore, space–selective heating by lasers as well as electron irradiation can be used to generate localized crystal spot, including metals and various non-metal compounds. In the glass phase, the normal crystallization behaviour found in liquid phase, e.g., preferred orientation, can also be preserved, as exemplified in many surface and bulk crystallized GCs. TGCs can serve as solid-state materials applied in diverse fields. In the present review, we focus our discussion on the recent advances in the optics and photonic application o f type–II GCs, in which the precipitated crystals play a pivotal role in the performance. Doping with RE or TM ions lead to photoluminescence with spectroscopic properties comparable to crystalline materials, which are appealing for applications such as optical amplifiers for telecommunication, and spectral converter for solar cells and white–light LEDs. In addition, semiconductor, metallic NPs and nonlinear optical crystals precipitated from glasses have extended the application of glass to nonlinear optics that is far beyond the reach of normal glasses. TGCs as a composite material have been explored for decades with its applications extended to diverse fields. One of the fundamental questions for GCs that need to be answered in a more precise manner is the mechanism of nucleation and its dependence on composition and environmental parameters, such as temperature, strain, etc. Scattering and diffraction techniques 116

have been widely used for examining these materials, while these averaging techniques are not well suited for detecting nucleation that begins with vanishing volume fraction. Recently, advanced aberration-corrected transmission electron microscopy has allowed the direct observation of glass structure at the unprecedentedly high resolution. Currently, it seems possible that such technology is applied to delineate the nucleation process as well as the formation of metastable phases in specific inorganic glasses (such as silicate glasses) (see section 4.3.2); while an in-situ experiment remains different. Another crucial question concerning the design of GCs is that whether it is possible to precipitate the desired crystals by designing the glass compositions like what they have achieved in wet-chemistry synthesis. This question, however, has to be asked from two different angles. For a giving glass composition, the prediction of the type of crystals to be precipitated at a certain temperature is admittedly inaccurate according to HSAB theory, phase diagram or TTT diagram. The accuracy of the predictions might be improved with the assistance of modern MD simulation and DFT calculations, while the complex nature of most glass compositions still poses a tremendous challenge to modelling with the highest precision. Currently, our knowledge on how to choose the right glass composition for the precipitation of the desired crystalline phase remains quite limited and largely dependent on experiences, and the types of crystalline phases that can be precipitated in glasses are, to a certain extent, serendipitous. For many crystalline materials, the material design is assisted by modelling and calculation, therefore applying analogously the concepts and principles to GC-based materials could be a potential future direction. Currently, there is still a constant pursuit of new functionalities as well as the optimization of current applications. TGCs, when applied as an optical material, still have to be precisely refined, for instance, to reduce scattering loss, and this is very important for photonic applications, such as laser and optical amplification. The optimization of TGCs still requires much effort of scientists from both experimental and theoretical sides. Besides serving as optical material, the functionalization of TGCs through the incorporation of crystalline phase with magnetic, electric 117

properties may potentially enables much broader applications in modern technologies. Such investigations remain in its infancy and deserve being explored in the near future.

Acknowledgements

This work is financially supported by the National Natural Science Foundation of China (Grants no. 51472091, 61475047, 11504323), and the Open Fund of the State Key Laboratory of High Field Laser Physics (Shanghai Institute of Optics and Fine Mechanics, Chinese Academy of Sciences).

References

[1] "Gold Ruby Glass from Gibraltar". Glass.co.nz. Retrieved 2016-11-29. [2] Stookey SD. Photosensitively opacifiable glass. 1954, U. S. Patent 2684911. [3] Stookey SD. Catalyzed crystallization of glass in theory and practice. Ind Eng Chem 1959; 51: 805–8. [4] Beall GH. Design and properties of glass-ceramics, Annu Rev Mater Sci 1992; 22: 91–119. [5] Karpukhina N, Hill RG, Law RV. Crystallisation in oxide glasses – a tutorial review. Chem Soc Rev 2014; 43: 217486. [6] de Pablos–Martin A, Ferrari M, Pascual MJ, Righini GC. Glass-ceramics: A class of nanostructured materials for photonics. Riv Novo Cimento 2015; 38: 31169. [7] Höland W, Beall GH. Glass ceramic technology, 2nd Ed. Hoboken, New Jersey: John Wiley & Sons, Inc.; 2012. [8] Zanotto ED. Crystals in Glass, A hidden beauty. Hoboken, New Jersey: John Wiley & Sons, Inc.; 2013. [9] Komatsu T. Design and control of crystallization in oxide glasses, J Non–Cryst Solids 2015; 428: 156–75. [10] Zanotto ED. A bright future for glass-ceramics. Am Ceram Soc Bull 2010; 89: 19–27. [11] Montazerian M, Singh SP, Zanotto ED. An analysis of glass-ceramic research and commercialization. Am Ceram Soc Bull 2015; 94: 30–5. [12] Meyer K. Physikalisch–Chemische Kristallographie. Dt Verlag für Grundstoffindustrie, Leipzig, 1968. [13] Kingery WD, Bowen HK, Uhlmann DR. Introduction to Ceramics. New York: John Wiley & Sons; 1976. [14] Fokin VM, Zanotto ED, Yuritsyn NS, Schmelzer JWP. Homogeneous crystal nucleation in silicate glasses: A 40 years perspective. J Non-Cryst Solids 2006; 352: 2681–714. 118

[15] Schmelzer JWP, Potapov OV, Fokin VM, Muller R, Reinsch S. The effect of elastic stress and relaxation on crystal nucleation in lithium disilicate glass. J Non-Cryst Solids 2004; 333: 150-64. [16] Paradis PF, Ishikawa T, Yu JD, Yoda S. Hybrid electrostatic-aerodynamic levitation furnace for the high-temperature processing of oxide materials on the ground. Rev Sci Instru 2001; 72: 2811–15. [17] Yu JD, Arai Y, Masaki T, Ishikawa T, Yoda S, Kohara S, et al. Fabrication of BaTi 2O5 glass-ceramics with unusual dielectric properties during crystallization. Chem Mater 2006; 18: 2169–73. [18] Weinberg MC, Bernie DP. III. Kinetics of Crystallization of highly anisotropic particles. Glastech Ber Glass Sci Technol 2000; 72C: 129–37. [19] Vogel W. Chemistry of Glass, John Wiley & Sons, Rev. ed; 2006. [20] Ostwald W. Studien über die bildung und umwandlung fester körper. Z. Phys. Chem. 1897; 22: 289–330. [21] Cormier L, Nucleation in glasses–new experimental findings and recent theories. Proc Mater Sci 2014; 7: 60–71. [22] Krüger S, Deubener J. Lag time to crystal nucleation of supercooled lithium disilicate melts: A test of the classical nucleation theory, J Non–Cryst Solids 2015; 426: 1–6. [23] Gutzow I, Kashchiev D, Avramov I. Nucleation and crystallization in glass-forming melts - old problems and new questions, J Non–Cryst Solids 1985; 73: 477–99. [24] Avramov I, Keding R, Rüssel C, Crystallization kinetics and rigidity percolation in glass-forming melts. J Non–Cryst Solids 2000; 272: 147–53. [25] Schmelzer JWP, Abyzov AS, Fokin VM, Schick C, Zanotto ED, Crystallization of glass-forming liquids: Maxima of nucleation, growth, and overall crystallization rates, J Non–Cryst Solids 2015; 429: 24–32. [26] Schmelzer JWP, Abyzov AS, Fokin VM, Schick C, Zanotto ED, Crystallization in glass-forming liquids: Effects of decoupling of diffusion and viscosity on crystal growth, J Non–Cryst Solids 2015; 429: 45–53. [27] Schmelzer JWP, Abyzov AS, Fokin VM, Schick C, Zanotto ED, Crystallization in glass-forming liquids: Effects of fragility and glass transition temperature, J Non–Cryst Solids 2015; 429: 68–74. [28] Gibson JM. Viewing the seeds of crystallization. Science 2009; 326: 942–3. [29] Lee BS, Burr GW, Shelby RM, Raoux S, Rettner CT, Bogle SN, et al. Observation of the role of subcritical nuclei in crystallization of a glassy solid. Science 2009; 326: 980–4. [30] Anderson VJ, Lekkerkerker JNW. Insights into phase transition kinetics from colloid science. Nature 2002; 416: 811–5. [31] Russo J, Tanaka H. Nonclassical pathways of crystallization in colloidal systems. MRS Bull 2016; 41: 369–74. [32] Arai S, Tanaka H. Surface-assisted single-crystal formation of charged colloids. Nat. Phys. 2017; 13: 503–9. [33] Watanabe K, Kawasaki T, Tanaka H. Structural origin of enhanced slow dynamics near a wall in glass-forming systems. Nat Mater 2011; 10: 512–20. [34] Weeks ER, Crocker JC, Levitt AC, Schofield A, Weitz DA. Three-dimensional direct imaging of structural relaxation near the colloidal glass transition. Science 2000; 287: 627–31. [35] Gasser U, Weeks ER, Schofield A, Pusey PN, Weitz DA. Real-space imaging of nucleation and growth in colloidal crystallization. Science 2001; 292: 258–62. [36] Sciortino F, Tartaglia P, Glassy colloidal systems. Adv Phys 2005; 54: 471–524. [37] Gokhale S, Sood AK, Ganapathy R. Deconstructing the glass transition through critical experiments on colloids. Adv Phys 2016; 65: 363–452. [38] Wang H, Gould H, Klein W. Homogeneous and heterogeneous nucleation of Lennard-Jones liquids, Phys Rev E 2007; 76: 031604. [39] Zanotto ED, Tsuchida JE, Schneider JF, Eckert H, Thirty-year quest for structure–nucleation relationships in oxide glasses. Int Mater Rev 2015; 60: 37691. 119

[40] Dantelle G, Mortier M, Monteil A, Chaussedent S, Silva MAP. Molecular dynamics simulation study of erbium induced devitrification in vitreous PbF 2. J Chem Phys 2007; 127: 094509. [41] Zhao JJ, Ma RH, Chen XK, Kang BB, Qiao XS, DU JC, et al. From phase separation to nanocrystallization in fluorosilicate glasses: structural design of highly luminescent glass-ceramics. J Phys Chem C 2016; 120: 17726−32. [42] Pearson RG, Chemical Hardness - Applications From Molecules to Solids, Wiley-VCH, Weinheim, 1997. [43] Zanotto ED, Glass Crystallization Research — A 36-Year Retrospective. Part I, Fundamental Studies. Int J Appl Glass Sci 2013, 4: 10516. [44] Schmelzer J, Möller J, Gutzow I. Ostwald’s rule of stages: the effect of elastic strains and external pressure. Z Phys Chem 1998; 204: 171–81. [45] Zanotto ED, E. Muller E. A simple method to predict the nucleation mechanism in glass. J. Non-Cryst. Solid 1991; 130: 220–1. [46] Beall GH. Glass-ceramics: Recent development and application. In: Nucleation and Crystallization in Glasses and Liquids. Ceram Trans 1993; 30: 241–66. [47] Echeverría LM. New lithium disilicate glass–ceramic. Bol. Soc. Esp. Ceram. 1992; 5: 183–8. [48] Beall GH, Pinckney LR. Nanophase glass–ceramics. J Am Ceram Soc 1999; 82: 5–16. [49] Beall GH. Chain silicate glass–ceramics. J Non-Cryst Solids 1991; 129: 163–73. [50] Chyung K, Beall GH, Grossman DG. In Proceedings Tenth Int. Cong. on Glass, Kunugi M, Tashiro M and Saga N (eds.), The Ceramic Society of Japan, Kyoto, Japan, Part II: 1974; 122–9. [51] Höland W, Wange P, Naumann K, Vogel J, Carl G, Jana C, et al. Control of phase formation processes in glass-ceramics for medicine and technology. J Non-Cryst Solids 1991; 129: 152–62. [52] Abe Y, Kasuga T, Hosono H, de Groot K. Preparation of high-strength calcium-phosphate glass-ceramics by unidirectional crystallization. J Am Ceram Soc 1984; 67: C142–4. [53] Hosono H, Tsuchitani F, Kazunari K, Abe Y. Porous glass-ceramic cation exchangers: Cation exchange properties of porous glass-ceramics with skeleton of fast Li ion-conducting LiTi2(PO4)3 crystals. J Mater Res 1994; 9: 755–61. [54] Kang SL, Xiao XD, Pan QW, Chen DD, Qiu JR, Dong GP. Spectroscopic properties in Er3+-doped germanotellurite glasses and glass ceramics for mid-infrared laser materials. Sci Rep 2017; 7:43186. [55] Rozé M, Calvez L, Ledemi Y, Allix M, Matzen G, Zhang XH, Optical and Mechanical Properties of Glasses and Glass–Ceramics Based on the Ge–Ga–Se System. J Am Ceram Soc 2008; 91: 3566–70. [56] Lin C, Calvez L, Bureau B, Ledemi Y, Xu Y, Tao H, Zhang X, Zhao X. Controllability study of crystallization on whole visible-transparent chalcogenide glasses of GeS2 -Ga2S3-CsCl system. J. Optelectron. Adv Mater 2010; 12: 1684–91. [57] Rozé M, Calvez L, Hubert M, Toupin P, Bureau B, Boussard–Pledel C, Zhang XH, Molded glass-ceramics for infrared applications. Int J Appl Glass Sci 2011; 2: 12936. [58] Calvez L, Ma HL, Lucas J, Zhang XH, Selenium-based glasses and glass ceramics transmitting light from the visible to the far-IR. Adv Mater 2007; 19: 12932. [59] Lin CG, Tao HZ, Zheng XL, Pan RK, Zang HC, Zhao XJ, Second-harmonic generation in IR-transparent -GeS2 crystallized glasses. Opt Lett 2009; 34: 4379. [60] Guignard M, Nazabal V, Zhang XH, Smektala F, Moreac A, Pechev S, et al. Crystalline phase responsible for the permanent second-harmonic generation in chalcogenide glass-ceramics. Opt Mater 2007; 30: 338–45. [61] McMillan PW. Glass–Ceramics, 2nd Edn. New York: Academic Press; 1979. [62] Zanotto ED. Metastable phases in lithium disilicate glasses. J Non–Cryst Solids 1997; 19: 42–8. [63] Zanotto ED, James PF. Experimental test of the general-theory of transformation kinetics–homogeneous nucleation in a BaO·2SiO2 glass, J NonCryst Solids 1988; 104, 70–72. [64] Abe Y. Abnormal characteristics in crystallization of Ca(PO 3)2 glass. Nature 1979; 282: 55–6. 120

[65] Abe Y, Hosoe M, Kasuga T, Ishikawa H, Shinkai N, Suzuki N, Nakayama J. High-strength Ca(PO3 )2 glass-ceramic prepared by unidirectional crystallization. J Am Ceram Soc 1982; 65: C189–90. [66] Halliyal A, Bhalla AS, Newnham RE, Cross LE. Ba2TiGe2O8 and Ba2 TiSi2O8 pyroelectric glass-ceramics. J Mater Sci 1981; 16: 10238. [67] Cabral AA, Fokin VM, Zanotto ED, Chinaglia CR. Nanocrystallization of fresnoite glass. I. Nucleation and growth kinetics. J Non–Cryst Solids 2003; 330: 174–86. [68] Takahashi Y, Yamazaki Y, Ihara R, Fujiwara T. Parasitic amorphous on single-domain crystal: Structural observations of silicate glass-ceramics. Sci Rep 2013; 3: 1147. [69] Wisniewski W, Takano K, Takahashi Y, Fujiwara T, Rüssel C. Microstructure of transparent strontium fresnoite glass-ceramics. Sci Rep 2015; 5: 9069. [70] Takahashi Y, Yamazaki Y, Ihara R, Fujiwara T. Perfect surface crystallization and parasitic structures in nonstoichiometric glass-ceramics: Micro-/nanoscopic aspects. Appl Phys Lett 2013; 102: 191903. [71] Sigaev VN, Mamonov AB, Pernice P, Aronne A, Stefanovich SY, et al. Pyroelectric composites based on LaBSiO5 stillwellite. J Euro Ceram Soc 2000; 20: 12259 [72] Sigaev VN, Dechev AV, Kadyshman SL, Altakh OL, Stefanovich SY, Molev VI. Glasses in the La2O3-B2O3-SiO2 system and crystallization of the ferroelectric LaBSiO5 phase. Glass Phys Chem 1996; 22: 18. [73] Sigaev VN, Stefanovich SY, Sarkisov PD, Lopatina EV. Lanthanum borogermanate glasses and crystallization of the stillwellite LaBGeO5: І. Specific features of synthesis and physicochemical properties of glasses. Glass Phys Chem 1994; 20: 392–7. [74] Sigaev VN, Stefanovich SY, Sarkisov PD, Lopatina EV. Lanthanum borogermanate glasses and crystallization of the stillwellite LaBGeO5: II. Formation of glass-crystalline structure, dielectric and non-linear optical properties. Glass Phys Chem 1994; 20: 398–404. [75] Kao YH, Hu Y, Zheng HX, Mackenzie JD, Perry K, Bourhill G, et al. Second harmonic generation in transparent barium borate glass-ceramics. J Non-Cryst Solids 1994; 167: 247–54. [76] Pinckney LR. Transparent glass-ceramics based on ZnO crystals. Phys Chem Glasses: Eur J Glass Sci Technol B 2006; 47: 127–30. [77] Masai H, Toda T, Ueno T, Takahashi Y, Fujiwara T. ZnO glass-ceramics: An alternative way to produce semiconductor materials. Appl Phys Lett 2009; 94: 151908. [78] Masai H, Yanagida T, Fujiwara T. Photo- and radiaoluminescence of ZnO-precipitated glass-ceramics. Sensors Mater 2015; 27: 237-45. [79] Takahashi Y, Kinoshita M, Miyazaki T, Osada M, Terakado N, Fujiwara T. Affects of supercooled liquid state on the properties of the crystallized phase: Microscopic and spectroscopic observations in ZnO-crystallized glass-ceramics. Appl Phys Lett 2015; 107: 061902. [80] Ceccato R, Dal Maschio R, Gialanella S, Mariotto G, Montagna M, Rossi F, et al. Nucleation of Ga2O3 nanocrystals in the K2O–Ga2O3–SiO2 glass system. J Appl Phys 2001; 90: 2522–7. [81] Beall GH, Pinckney LR, High modulus glass-ceramics containing fine grained spinel-type crystals. Europ. Patent, 1995; EP0710627. [82] Pinckney LR, Transparent, high strain point spinel glass-ceramic. J Non–Cryst Solids 1999; 255: 171–7. [83] Takahashi Y, Meguro K, Naganuma H, Terakado N, Fujiwara T. Multiferroic BiFeO3 glass-ceramics: Phase formation and physical property. Appl Phys Lett 2014; 104: 221901. [84] Takahashi Y, Fuji N, Fujiwara T. Nano-sized Ba2NaNb5O15–NaNbO3 co-crystallized glass-ceramics in phosphoniobate system. Appl Phys Lett 2012; 100: 201907. [85] Yamaguchi I, Tanaka K, Hirao K, Soga N. Preparation and optical properties of transparent glass-ceramics containing LiGa5O8:Cr3+. J Mater Sci, 1996; 31: 35417. 121

[86] Tanaka K, Mukai T, Ishihara T, Hirao K, Saga N, Sogo S, et al. Preparation and Optical Properties of Transparent Glass-Ceramics Containing Cobalt(II) Ions. J Am Ceram Soc, 1993; 76: 2839–45. [87] Suzuki T, Murugan GS, Ohishi Y. Optical properties of transparent Li2O–Ga2O3–SiO2 glass-ceramics embedding Ni-doped nanocrystals. Appl Phys Lett 2005; 86: 131903. [88] Nishia M, Tanabe S, Fujita K, Hirao K, Pezzotti G. Phase-selective cathodoluminescence spectroscopy of Er:YAG glass-ceramics. Solid State Commun 2004; 132: 19–23. [89] Fujita S, Tanabe S. Fabrication, microstructure and optical properties of Er3+:YAG glass-ceramics. Opt Mater 2010; 32: 886–90. [90] Taubert J, Hergt R, Müller R, Ulbrich C, Schlüppel W, Schmidt HG, et al. Phase separation in Ba-ferrite glass - ceramics investigated by Faraday microscopy. J Magn Magn Mater 1996; 168: 187–95. [91] Beall GH, Ta2O5 -nucleated glass - ceramic articles. US Patent 1971; 3573939. [92] Ito S, Kokubo T, Tashiro M. Transparency of LiTaO3–SiO2–TiO2 –Al2O3 glass–ceramics in relation to their microstructure. J Mater Sci 1978; 13: 930–8. [93] Grossman D, Machinable glass-ceramics based on tetrasilicic mica. J Am Ceram Soc 1972; 55: 446–9. [94] Allen RE, Herczog A. Transparent high dielectric constant material, method and electroluminescent device. U.S. Patent 3114066, 1962. [95] Sigaev VN, Ryzhenkova VS, Lotareva SV, Golubeva VN, Stefanovich SY, Okada A. Glasses and their crystallization in the (1−х)KNbO3 ·хSiO2 system at low glass-forming oxide contents, 0⩽х⩽0.35. J Non–Cryst Solids 2010; 356: 958–65. [96] Kao YH, Hu Y, Zheng HX, Mackenzie JD, Perry K, Bourhill G, et al. Second harmonic generation in transparent barium borate glass-ceramics. J Non–Cryst Solids 1994; 167: 247–54. [97] Ding Y, Osaka A, Miura Y. Enhanced surface crystallization of P-barium borate on glass due to ultrasonic treatment. J Am Ceram Soc 1994; 77: 749–52. [98] Murugan GS, Suzuki T, Ohishi Y, Takahashi Y, Benino Y, Fujiwara T, Komatsu T. Second-harmonic generation in transparent surface crystallized glasses in the BaO-B2O3 -TeO2 system. Appl Phys Lett 2004; 85: 3405–7. [99] Ihara R, Honma T, Benino Y, Fujiwara T, Komatsu T. Second-order optical nonlinearities phases in crystallized of metastable BiBO3 glasses. Opt Mater 2004; 27: 403–8. [100] Sigaev VN, Pernice P, Aronne A, Akimova OV, Stefanovich SY, Scaglione A. KTiOPO4 precipitation from potassium titanium phosphate glasses, producing second harmonic generation. J Non–Cryst Solids 2001, 292: 59–69. [101] Kokubo T. Crystallization of BaO·TiO2 -SiO2-Al2O3 glasses and dielectric properties of their crystallized products. Bull Inst Chem Res Kyoto Univ. 1969, 47: 572–83. [102] Bhattacharyya S, Bocker C, Heil T, Jinschek JR, Höche T, Rüssel C, et al. Experimental evidence of self-limited growth of nanocrystals in glass, Nano Lett 2009; 9: 2493–6. [103] Rüssel C. Nanocrystallization of CaF2 from Na2 O/K2O/CaO/CaF2/Al2O3/SiO2 glasses. Chem Mater 2005; 17: 5843–7. [104] Moesgaard M, Keding R, Skibsted J, Yue YZ. Evidence of intermediate-range order heterogeneity in calcium aluminosilicate glasses. Chem Mater 2010; 22: 4471–83. [105] Lin CG, Bocker C, Rüssel C. Nanocrystallization in oxyfluoride glasses controlled by amorphous phase separation. Nano Lett 2015; 15: 6764-9. [106] Wang Y, Ohwaki J. New transparent vitroceramics codoped with Er 3+ and Yb3+ for efficient frequency upconversion. Appl Phys Lett 1993; 63: 3268–70. [107] Xue Z, Edwards TG, Sen S. Structural evolution during precipitation of alkaline-earth fluoride nanocrystals in oxyfluoride glasses: a multinuclear nuclear magnetic resonance spectroscopic study. J Am Ceram Soc 2011; 94: 2092–8. 122

[108] Hemono N, Pierre G, Munoz F, de Pablos–Martin A, Pascual MJ, Duran A. Processing of transparent glass-ceramics by nanocrystallisation of LaF3. J Eur Ceram Soc 2009; 29: 2915–20. [109] Liu F, Ma E, Chen DQ, Yu YL, Wang YS. Tunable red-green upconversion luminescence in novel transparent glass ceramics containing Er: NaYF4 nanocrystals. Appl Phys Lett 2006; 110: 20843-6. [110] Huang SM, Gao QC, Gu M. Enhanced luminescence in transparent glass ceramics containing BaYF5: Ce3+ nanocrystals. J Lumin 2012; 132: 750–4. [111] Zheng F, Xu SQ, Zhao SL, Deng DG, Hua YJ, Wang HP. Spectroscopic investigations on Er3+/Yb3+-doped oxyfluoride glass ceramics containing YOF nanocrystals. J Rare Earths 2012; 30: 137–41. [112] Chen SL, Zhao SL, Zheng F, Zhang C, Huang LH, Xu SQ. Enhanced up-conversion luminescence of Er3+:LaOF oxyfluoride borosilicate glass ceramics. Ceram Int 2013; 39: 2909–13. [113] Hu YB, Qiu SW, Gao Y, Qiu JB. Crystallization and spectroscopic properties in Er3+ doped oxyfluorogermanate glass ceramics containing Na. Opt Mater 2015; 45: 82–6. [114] Burdaev PA, Aseev VA, Kolobkova EV, Nikonorov NV, Trofimov AO. Nanostructured glass ceramics based on fluorophosphate glass with a high content of rare earth ions. Glass Phys Chem 2015; 41: 132–6. [115] Hatefi Y, Shahtahmasebi N, Moghimi A, Attaran E. Frequency-conversion properties of Eu3+ doped chlorophosphate glass ceramics containing CaCl 2 nanocrystals. J Lumin 2011; 131: 114–8. [116] Pfau C, Bohley C, Miclea PT, Schweizer S. Structural phase transitions of barium halide nanocrystals in fluorozirconate glasses studied by Raman spectroscopy. J App Phys 2011; 109: 083545. [117] Yang AP, Lin H, Chen DQ, Yu YL, Wang YS. Phase-separation induced homogeneous nucleation and growth of Cs3LaCl6 nanoparticles in chalcohalide glass. Mater Res Bull 2014; 49: 193–8. [118] Zhang XH, Ma HL, Lucas J. A new class of infrared transmitting glass-ceramics based on controlled nucleation and growth of alkali halide in a sulphide based glass matrix. J Non–Cryst Solids 2004; 337: 130–5. [119] Kondo Y, Kuroiwa Y, Sugimoto N, Manabe T, Ito S, Tokizaki T, et al. Third-order optical non-linearities of CuCl-doped glasses in a near resonance region. J Non–Cryst Solids 1996; 196: 90–4. [120] Masumoto Y, Kawazoe T, Yamamoto T, Observation of persistent spectral hole burning in CuBr quantum dots. Phys Rev B 1995; 52: 4688–91. [121] Xia ML, Liu C, Zhao ZY, Ai B, Yin QY, Xie J, et al. Formation and optical properties of ZnSe and ZnS nanocrystals in glasses. J Non–Cryst Solids 2015; 429: 79–82. [122] Du X, Zhang LL, Dong GP, Sharafudeen K, Wen J, Chen DD, et al. Coloration and nonlinear optical properties of ZnTe quantum dots in ZnO–TeO2–P2O5 Glasses. J Am Ceram Soc 2014; 97: 185–8. [123] Borrelli NF, Hall DW, Holland HJ, Smith DW. Quantum confinement effects of semiconducting microcrystallites in glass. J Appl Phys 1987; 61: 5399–09. [124] Barbosa LC, Reynoso VCS, de Paula AM, de Oliveira CRM, Alves OL, Craievich AF, et al. CdTe quantum dots by melt heat treatment in borosilicate glasses. J Non–Cryst Solids 1997; 219: 205–11. [125] Dong GP, Wu GB, Fan SH, Zhang FT, Zhang YH, Wu BT, et al. Formation, near-infrared luminescence and multi-wavelength optical amplification of PbS quantum dot-embedded silicate glasses. J Non–Cryst Solids 2014; 383: 192–95. [126] Lipovskii A, Kolobkova E, Petrikov V, Kang I, Olkhovets A, Krauss T, et al. Synthesis and characterization of PbSe quantum dots in phosphate glass. Appl Phys Lett 1997; 71: 3406. [127] Qiu JR, Jiang XW, Zhu CS, Inouye H, Si JH, Hirao K. Optical properties of structurally modified glasses doped with gold ions. Opt Lett 2004; 29: 370–2. 123

[128] Uchida K, Kaneko S, Omi S, Hata C, Tanji H, Asahara Y, et al. Optical nonlinearities of a high concentration of small metal particles dispersed in glass: copper and silver particles. J Opt Soc Am B 1994; 11: 1236–43. [129] Liu QM, He X, Zhou X, Ren F, Xiao XH, Jiang CZ, et al. Third-order nonlinearity in Ag-nanoparticles embedded 56GeS2–24Ga2S3–20KBr chalcohalide glasses. J Non–Cryst Solids 2011; 357: 2320–3. [130] Chen FF, Yu QS, Qiao BJ, Xu TF, Dai SX, Ji W. Investigations of structure and nonlinear optical properties of gold doped germanium-gallium-sulfur chalcogenide glasses. J Non-Cryst Solids, 2015; 412: 304. [131] Lin G, Pan HH, Qiu JR, Zhao QZ. Nonlinear optical properties of lead nanocrystals embedding glass induced by thermal treatment and femtosecond laser irradiation. Chem Phys Lett 2011; 516: 186–91. [132] Lin G, Luo FF, Pan HH, Smedskjaer MM, Teng Y, Chen DP, et al. Universal preparation of novel metal and semiconductor nanoparticle glass composites with excellent nonlinear optical properties. J Phys Chem C 2011; 115: 24598–604. [133] Kerker K. The Scattering of Light. New York: Academic Press; 1969. [134] Andreev NS. Scattering of visible light by glasses undergoing phase separation and homogenization. J Non–Cryst Solids 1978; 30: 99–126. [135] Hopper RW. Stochastic theory of scattering from idealized spinodal structures: II. Scattering in general and for the basic late stage model. J Non–Cryst Solids 1985; 70: 111–42. [136] Hendy S. Light scattering in transparent glass ceramics. Appl Phys Lett 2002; 81: 1171–3. [137] Mie G. Beiträge zur Optik trüber Medien, speziell kolloidaler Metallösungen. Ann Physik 1908; 25: 377–445. [138] Tournie A, Ricciardi P, Colomban P. Glass corrosion mechanisms: A multiscale analysis. Solid State Ionics 2008; 179: 2142-54. [139] Scheffler C, Förster T, Mäder E, Heinrich G, Hempel S, Mechtcherine V. Aging of alkali-resistant glass and basalt fibers in alkaline solutions: Evaluation of the failure stress by Weibull distribution function. J Non-Cryst Solids 2009; 355: 2588–95. [140] Oka Y, Ricker S, Tomozawa M. Calcium deposition on glass-surface as an inhibitor to alkaline attack. J Am Ceram Soc 1979; 62: 631–2. [141] Langford JI, Wilson AJC. Scherrer after sixty years: a survey and some new results in the determination of crystallite size. J Appl Cryst 1978; 11: 102–13. [142] Scherrer P. Bestimmung der Grösse und der inneren Struktur von Kolloidteilchen mittels Röntgenstrahlen. Nachr Ges Wiss Göttingen 1918; 26: 98–100. [143] Freund HJ. Controlling silica in its crystalline and amorphous states: A problem in surface science. Acc Chem Res 2017; 50: 446−9. [144] Huang PY, Kurasch S, Alden JS, Shekhawat A, Alemi AA, McEuen PL, et al. Imaging atomic rearrangements in two-dimensional silica glass: watching silica’s dance. Science 2013; 342: 224–7. [145] Huang PY, Kurasch S, Srivastava A, Skakalova V, Kotakoski J, Krasheninnikov AV. Direct imaging of a two-dimensional silica glass on graphene. Nano Lett 2012; 12: 1081–6. [146] Lichtenstein L, Heyde M, Freund H-J. Crystalline-vitreous interface in two dimensional silica. Phys Rev Lett 2012; 109: 106101. [147] Büchner C, Wang ZJ, Burson KM, Willinger MG, Heyde M, Schl gl R, et al. Large-area transferable wide band gap 2D silicon dioxide layer. ACS Nano 2016, 10, 7982−9. [148] Loshmanov A, Sigaev VN, Khodakovskaya RY, Pavlushkin NM, Yamzin II. Small-angle neutron-scattering on silica glasses containing titania. J Appl Crystallogr 1974; 7: 207–10. [149] Sigaev VN, Golubev NV, Ignat’eva ES, Champagnon B, Vouagner D, Nardou E, et al. Native amorphous nanoheterogeneity in gallium germanosilicates as a tool for driving Ga 2O3 nanocrystal formation in glass for optical devices, Nanoscale 2013; 5: 299–306. 124

[150] Chumakov AI, Monaco G, Fontana A, Bosak A, Hermann RP, Bessas D, et al. The role of disorder in the thermodynamics and atomic dynamics of glasses. Phy Rev Lett 2014; 112: 025502. [151] Takahashi Y, Osada M, Masai H, Fujiwara T. Crystallization and nanometric heterogeneity in glass: In situ observation of the boson peak during crystallization. Phys Rev B, 2009; 79: 214204. [152] Takahashi Y, Nakamura K, Osada M. Fujiwara T. Structural relaxation and quasi-elastic light scattering in glass: Approach by ferroelectric and ion-conducting phases. Sci Rep 2012; 2: 714. [153] Nakamura K, Takahashi Y, Osada M, Fujiwara T. Low-frequency Raman scattering in binary silicate glass: Boson peak frequency and its general expression. J Ceram Soc Jpn 2013; 121: 1012-4. [154] Yadav AK, Singh P. A review of the structures of oxide glasses by Raman spectroscopy. RSC Adv 2015; 5: 67583–609. [155] Mudoch JB, Stebbins JF, Carmichael ISE. High-resolution Si-29 NMR-study of silicate and aluminosilicate glasses - the effect of network-modifying cations. Am Miner 1985; 70: 332–43. [156] Schneider J, Mastelaro VR, Panepucci H, Zanotto ED. Si-29 MAS-NMR studies of Q(n) structural units in metasilicate glasses and their nucleating ability. J Non–Cryst Solids 2000; 273: 8–18. [157] Łączka K, CholewaKowalska K, Borczuch–Łączka M. Thermal and spectroscopic characterization of glasses and glass–ceramics of Li2O–Al2O3–SiO2 (LAS) system. J Mol Struct 2014; 1068: 275–82. [158] Ananthanarayanan A, Tricot G, Prasad Kothiyal G, Montagne L. A Comparative Overview of Glass-Ceramic Characterization by MAS-NMR and XRD. Crit. Rev. Solid State Mater Sci 2011; 36: 229–41. [159] Bertmer M, Eckert H. Dephasing of spin echoes by multiple heteronuclear dipolar interactions in rotational echo double resonance NMR experiments. Solid State Nucl Mag 1999; 15: 139–52. [160] Eckert H. Structural characterization of noncrystalline solids and glasses using solid state NMR. Prog Nucl Mag Sp 1992; 24: 159–293. [161] Eckert H, Elbers S, Epping JD, Janssen M, Kalwei M, Strojek W, et al. Dipolar solid state NMR approaches towards medium-range structure in oxide glasses. Top Curr Chem 2004; 246: 195–233. [162] Eden M. Homonuclear dipolar recoupling of half-integer spin quadrupolar nuclei: Techniques and applications. Solid State Nucl Mag 2009; 36: 1–10. [163] Gullion T, Vega AJ. Measuring heteronuclear dipolar couplings for I=1/2, S > 1/2 spin pairs by REDOR and REAPDOR NMR. Prog Nucl Mag Sp 2005; 47: 123–36. [164] Tricot G, Delevoye L, Palavit G, Montagne L. Phase identification and quantification in a devitrified glass using homo-and heteronuclear solid-state NMR. Chem Commun 2005; 42: 5289–91. [165] Holland W, Ritzberger C, Apel E, Rheinberger V, Nesper R, Krumeich F, et al. Formation and crystal growth of needle-like fluorapatite in functional glass-ceramics. J Mater Chem 2008; 18: 1318–32. [166] Holland W, Rheinberger V, Apel E, Ritzberger C, Eckert H, Monster C. Mechanism of nucleation and crystallization in high strength glass-ceramics. Phys Chem Glasses: Eur J Glass Sci Technol B 2007; 48: 97–102. [167] Cormier L, Cochain B, Duguć A, Dargaud O. Transition elements and nucleation in glasses using X-ray absorption spectroscopy. Int J Appl Glass Sci 2014; 5: 12635. [168] Yue YZ, Characteristic temperatures of enthalpy relaxation in glass, J Non–Cryst Solids 2008; 354: 1112–8. [169] Hrubý A. Evaluation of glass-forming tendency by means of DTA. Czech J Phys B 1972; 22: 1187–93. [170] Moesgaard M, Yue YZ, Compositional dependence of fragility and glass forming ability of calcium aluminosilicate melts, J Non–Cryst Solids 2009; 355: 867–73 [171] Tian KV, Yang B, Yue YZ, Bowron DT, Mayers J, Donnan RS, et al. Atomic and vibrational origins of mechanical toughness in bioactive cement during setting, Nat Commun 2015; 6: 8631. 125

[172] Pedersen MT, Tian KV, Dobó–Nagy C, Chass GA, Greaves GN, Yue YZ. Phase separation in an ionomer glass: Insight from calorimetry and phase transitions. J Non–Cryst Solids 2015; 415: 24–9. [173] Zheng QJ, Smedskjaer MM, Youngman RE, Potuzak M, Mauro JC, Yue YZ. Influence of aluminum speciation on the stability of aluminosilicate glasses against crystallization. Appl Phys Lett 2012; 101: 041906. [174] Thieme K, Avramov I, Rüssel C. The mechanism of deceleration of nucleation and crystal growth by the small addition of transition metals to lithium disilicate glasses. Sci Rep 2016; 6: 25451. [175] Thieme K, Rüssel C. Nucleation inhibitors-The effect of small concentrations of Al2O3, La2O3 or TiO2 on nucleation and crystallization of lithium disilicate. J Euro Ceram Soc 2014; 34: 3969–79. [176] Thieme K, Rüssel C, CeO2 and Y2O3 as nucleation inhibitors in lithium disilicate glasses. J Mater Sci 2016; 51: 989–99. [177] Liu SJ, Zhang YF, He W, Yue YZ. Transparent phosphosilicate glasses containing crystals formed during cooling of melts. J Non–Cryst Solids 2011; 357: 3897–900. [178] Liu SJ, Zhang YF, Yue YZ. Effect of cooling rate on crystallization in an aluminophosphosilicate melt. Phys Chem Glasses: Eur J Glass Sci Technol 2011; 52: 231–5. [179] Liu SJ, Fu GZ, Shan ZT, Ren XR, Zhang YF, Zhu CF, et al. Nanocrystal formation and photoluminescence in the Yb3+/Er3+ codoped phosphosilicate glasses. J Non–Cryst Solids 2014; 383: 141–5. [180] Liu SJ, Shan ZT, Fu GZ, Yue YZ. Influence of rare earth oxides on the non-isothermal crystallization of phosphosilicate melts during cooling. J Non–Cryst Solids 2014; 385: 75–80. [181] Avrami M. Kinetics of phase change I - General theory. J Chem Phys 1939; 7: 1103–12. [182] Friedman HL. Kinetics of thermal degradation of char-forming plastics from thermogravimetry. Application to phenolic plastic. J Polym Sci C 1964; 6: 183–195. [183] Liu XF, Fechler N, Antonietti M. Salt melt synthesis of ceramics, semiconductors and carbon nanostructures. Chem Soc Rev 2013; 42: 8237–65. [184] Krüger S, Deubener J. The TTT Curves of the heterogeneous and homogeneous crystallization of lithium disilicate – A stochastic approach to crystal nucleation. Front Mater 2016; 3: 42. [185] Simo A, Polte J, Pf nder N, Vainio U, Emmerling F, Rademann K. Formation mechanism of silver nanoparticles stabilized in glassy matrices. J Am Chem Soc 2012; 134: 18824–33. [186] Sun HT, Zhou JJ, Qiu JR. Recent advances in bismuth activated photonic materials. Prog Mater Sci 2014; 64: 1–72. [187] Fujimoto Y, Nakatsuka M. Infrared luminescence from bismuth-doped silica glass. Jpn J Appl Phys 2001; 40: L27981. [188] Zhou SF, Guo QB, Inoue H, Ye Q, Masuno A, Zheng BB, et al. Topological engineering of glass for modulating chemical state of dopants. Adv Mater 2014; 26: 7966–72. [189] Xu ZS, Xu C, Guo QB, Liu XF, Qiu JR. Effect of TeO2 on near-infrared emission from bismuth centers in borogermanate glasses. J Am Ceram Soc 2016; 99: 760–4. [190] Tan LL, Kang SL, Pan ZW, Yang YF, Yue YZ, Xu SH, et al. Topo-chemical tailoring of tellurium quantum dot precipitation from supercooled polyphosphates for broadband optical amplification. Adv Optical Mater 2016; 4: 1624–34. [191] Lee YK, Lee JS, Heo J, Im WB, Chung WJ. Phosphor in glasses with Pb-free silicate glass powders as robust color-converting materials for white LED applications. Opt Lett 2012; 37: 3276–8. [192] Sainz MA, Miranzo P, Osendi MI. Sintering behaviour and properties of YAlSiO and YAlSiON glass-ceramics. Ceram Int 2011; 37: 1485–92. [193] Luo ZW, Liu WZ, Qu G, Lu AX, Han GR. Sintering behavior, microstructure and mechanical properties of various fluorine-containing Y-SiAlON glass-ceramics. J Non–Cryst Solids 2014; 388: 62–7. [194] Zhao JB, Zheng XL, Schartner EP, Ionescu P, Zhang R, Nguyen TL, et al. Upconversion Nanocrystal-Doped Glass: A new paradigm for photonic materials, Adv Opt Mater 2016; 4: 1507–17. 126

[195] Nakanishi T, Tanabe S. Preparation and luminescent properties of Eu2+-activated glass ceramic phosphor precipitated with β-Ca2SiO4 and Ca3Si2O7. Phys Status Solidi A 2009; 206: 919–22. [196] Nakanishi T, Tanabe S. Novel Eu2+-activated glass ceramics precipitated with green and red phosphors for high-power white LED. IEEE J Sel Top Quantum Electron. 2009; 15: 11711176. [197] Nakanishi T, Katayama Y, Ueda J, Honma T, Tanabe S, Komatsu T. Fabrication of Eu:SrAl2O4-based glass ceramics using Frozen sorbet method. J Ceram Soc Jpn 2011; 119: 609–15. [198] J. Ueda, Tanabe S. Preparation and optical property of glass ceramics containing ruby crystals. J Am Ceram Soc 2010; 93: 3084–7. [199] Nakanishi T, Watanabe K, Ueda J, Fushimi K, Tanabe S, Hasegawa Y. Enhanced light storage of SrAl2O4 glass-ceramics controlled by selective europium reduction. J Am Ceram Soc 2015; 98: 423–9. [200] Hench LH, West JK. The sol-gel process. Chem Rev 1990; 90: 3372. [201] Ciriminna R, Fidalgo A, Pandarus V, Beland F, Ilharco LM, Pagliaro M. The sol-gel route to advanced silica-based materials and recent applications. Chem Rev 2013; 113: 6592620. [202] Carretero-Genevrier A, Gich M, Picas L, Gazquez J, Drisko GL, Boissiere C, et al. Soft-chemistry–based routes to epitaxial a-quartz thin films with tunable textures. Science 2013; 340: 827–31. [203] Fréty N, Taylor A, Lewis MH. Microstructure and crystallisation behaviour of sol-gel derived 1/2SrO-1/2BaO-Al2O3 -2SiO2 glass-ceramic. J Non–Cryst Solids 1996; 195: 28–37. [204] Zheng BB, Wang ZW, Guo QB, Zhou SF. Glass composite as robust UV absorber for biological protection. Opt Mater Express 2016; 6: 531–9. [205] Lepoutre S, Boyer D, Fujihara S, Mahiou R. Structural and optical characterizations of sol–gel based composites constituted of LiGdF4: Eu3+ nanocrystallites dispersed into a silica matrix. J Mater Chem 2009; 19: 2784–8. [206] Wang QB, Iancu N, Seo DK. Preparation of large transparent silica monoliths with embedded photoluminescent [email protected] core/shell quantum dots. Chem Mater 2005; 17: 4762–4. [207] Jun S, Lee J, Jang E. Highly luminescent and photostable quantum dot silica monolith and its application to light-emitting diodes. ACS Nano 2013; 7: 1472–7. [208] Cui HT, Wang MM, Ren WZ, Liu Y, Zhao YN, Highly transparent silica monoliths embedded with high concentration oxide nanoparticles, J Sol-Gel Sci Technol 2013; 66:512–7. [209] Miura K, Qiu JR, Inouye H, T. Mitsuyu, Hirao K. Photowritten optical waveguides in various glasses with ultrashort pulse laser. Appl Phys Lett 1997; 71: 3329–31. [210] Sato R, Benino Y, Fujiwara T, Komatsu T. YAG laser-induced crystalline dot patterning in a samarium tellurite glasses. J Non–Cryst Solids 2001; 289: 228–32. [211] Honma T, Benino Y, Fujiwara T, Komatsu T, Sato R. Nonlinear optical crystal-line writing in glass by yttrium aluminum garnet laser irradiation. Appl Phys Lett 2003; 82: 892–4. [212] Abe M, Benino Y, Fujiwara T, Komatsu T, Sato R. Writing of nonlinear optical Sm2 (MoO4)3 crystal lines at the surface of glass by samarium atom heat processing. Appl Phys Lett 2005; 97: 123516. [213] Yang C, Shinozaki K, Honma T, Komatsu T. Nano-crystallization and highly oriented crystal line patterning of Sm3+-doped Bi2GeO5 and Bi4Ge3O12 in bismuth germanate-based glasses. J Non-Cryst Solids 2017; 459: 116–22. [214] Hasegawa S, Shinozaki K, Honma T, Dimitrov V, Kim HG, Komatsu T. Dielectric properties of glass-ceramics with Ba1−xY2x/3Nb2O6 nanocrystals and laser patterning of highly oriented crystal lines. J Non-Cryst Solids 2016; 452: 74–81. [215] Komatsu T, Honma T. Laser-induced line patterning of nonlinear optical crystals in glass. IEEE J Sel Topics Quant 2008; 14: 1289–97. [216] Honma T, Benino Y, Fujiwara T, Komatsu T, Sato R. Technique for writing of nonlinear optical single-crystal lines in glass. Appl Phys Lett 2003; 83: 2796–8. 127

[217] Suzuki F, Ogawa K, Honma T, Komatsu T. Laser patterning and preferential orientation of two-dimensional planar beta-BaB2O4 crystals on the glass surface, J Solid State Chem 2012; 185: 130–5. [218] Ogawa K, Honma T, Komatsu T. Birefringence imaging and orientation of laser patterned beta-BaB2O4 crystals with bending and curved shapes in glass. J Solid State Chem 2013; 207: 6–12. [219] Komatsu T, Honma T, Laser patterning and characterization of optical active crystals in glasses. J Asian Ceram Soc 2013; 1: 9–16. [220] González–Pérez S, Martín IR, Jaque D, Haro–González P, Capuj N. Control of the local devitrification on oxyfluoride glass doped with Er3+ ions under diode laser irradiation. J Appl Phys 2010; 108: 103103. [221] González–Pérez S, Martín IR, Lahoz F, Haro–González P, Herreros J. Local crystallization in an oxyfluoride glass doped with Er3+ ions using a continuous argon laser. Appl Phys A 2008; 93: 983–6. [222] Avansi W, Mastelaro VR, Andreeta MRB. Laser induced modification on 40BaO–45B2O3 –15TiO2 glass composition. J Non–Cryst Solids 2006; 352: 3398–403. [223] Savytskii D, Knorr B, Dierolf V, Jain H. Challenges of CW laser-induced crystallization in a chalcogenide glass, Opt Mater Express 2013; 3: 1026–38. [224] D. Savytskii D, Knorr B, Dierolf V, Jain H. Laser-induced growth of oriented Sb2S3 single crystal dots on the surface of 82SbSI-18Sb2 S3 glasses. J Non–Cryst Solids 2016; 431: 36–40. [225] Sigaev VN, Alieva EA, Lotarev SV, Lepekhin MM, Priseko YS, Rasstanaev AV. Local crystallization of glasses in the La2O3-B2O3-GeO2 system under laser irradiation. Glass Phys Chem 2009; 35, 13–20. [226] Tan DZ, Sharafudeen KN, Yue YZ, Qiu JR. Femtosecond laser induced phenomena in transparent solid materials: Fundamentals and applications. Prog Mater Sci 2016; 76: 154–228. [227] Miura K, Qiu JR, Mitsuyu T, Hirao K. Space-selective growth of frequency-conversion crystals in glasses with ultrashort infrared laser pulses. Opt Lett 2000; 25: 408–10. [228] Stone A, Jain H, Dierolf V, Sakakura M, Shimotsuma Y, Miura K, et al. Direct laser-writing of ferroelectric single-crystal waveguide architectures in glass for 3D integrated optics. Sci Rep 2015; 5: 10391. [229] Stone A, Sakakura M, Shimotsuma Y, Stone G, Gupta P, Miura K, et al. Formation of ferroelectric single-crystal architectures in LaBGeO5 glass by femtosecond vs. continuous-wave lasers. J Non–Cryst Solids 2010; 356: 3059–65. [230] Zhu B, Liu Y, Ye S, Qian B, Lin G, Dai Y, et al. Greatly enhanced effect of silver on femtosecond laser-induced precipitation of nonlinear optical crystals in glasses. Opt Lett 2009; 34: 1666–8. [231] Zhu B, Dai Y, Ma HL, Zhang SM, Lin G, Qiu JR. Femtosecond laser induced space-selective precipitation of nonlinear optical crystals in rare-earth-doped glasses. Opt Express 2007; 15: 6069–74. [232] Zhu B, Dai Y, Ma HL, Zhang SM, Qiu JR. Direct writing Eu3+ -doped Ba2TiSi2O8 crystalline pattern by femtosecond laser irradiation. J Alloy Compd 2008; 460: 590–3. [233] Liu Y, Zhu B, Dai Y, Qiao XS, Ye S, Teng Y, et al. Femtosecond laser writing of Er3+-doped CaF2 crystalline patterns in glass. Opt Lett 2009; 34: 3433–5. [234] Zhou SF, Jiang N, Miura K, Tanabe S, Shimizu M, Sakakura M, et al. Simultaneous Tailoring of Phase Evolution and Dopant Distribution in the Glassy Phase for Controllable Luminescence. J Am Chem Soc 2010; 132: 17945–52. [235] Qiu JR, Jiang XW, Zhu CS, Shirai M, Si JH, Jiang N, et al. Manipulation of gold nanoparticles inside transparent materials, Angew Chem Int Ed 2004; 43: 2230–4. [236] Teng Y, Qian B, Jiang N, Yu L, Luo FF, Ye S, et al. Light and heat driven precipitation of copper nanoparticles inside Cu2+-doped borate glasses, Chem Phys Lett 2010; 485: 91–4.

128

[237] Lin G, Pan HH, Luo FF, Chen QX, Chen DP, Qiu JR, et al. Glass-Ceramics with Embedded Gallium/Aluminum Nanoalloys Formed by Heat Treatment and Femtosecond Laser Irradiation. J Am Ceram Soc 2012; 95: 776–81. [238] Jiang N, Wu BT, Qiu JR, Spence JCH. Precipitation of nanocrystals in glasses by electron irradiation: An alternative path to form glass ceramics? Appl Phys Lett 2007; 90: 161909. [239] Jiang N, Zhou SF, Su D, Qiu JR. Do Eu dopants prefer the precipitated LaF3 nanocrystals in glass ceramics? Phys. Stat. Solidi 2012; 6: 487-9. [240] Jiang N, Qiu JR. Spence JCH, Precipitation of Ge nanoparticles from GeO2 glasses in transmission electron microscope. Appl Phys Lett 2005; 86: 143112. [241] Rüssel C. Oriented crystallization of glass. A review. J Non–Cryst Solids 1997; 219: 212–8. [242] Keding R, Rüssel C. Oriented crystallization of fresnoite in an electric field. Ber Bunsenges Phys Chem 1996; 100: 1515–8. [243] Anspach O, Keding R, Rüssel C. Oriented lithium disilicate glass–ceramics prepared by electrochemically induced nucleation. J Non–Cryst Solids 2005; 351: 656–62. [244] Keding R, Stachel D, Rüssel C. Oriented fluororichterite/diopsid glass-ceramics prepared by electrochemically induced nucleation. J Non–Cryst Solids 2001; 283: 137–43. [245] Wisniewski W, Nagel M, Völksch G, Rüssel C. Irregular fourfold hierarchy in fresnoite dendrites grown via electrochemically induced nucleation of a Ba2TiSi2.75O9.5 Glass. Cryst Growth Des 2010; 10: 4526–30. [246] Wisniewski W, Carl R, Rüssel C. Complex growth structures of mullite after electrochemically induced nucleation. CrystEngComm 2014; 16: 1192–200. [247] Freitas AMB, Landgraf FJG, Nývlt J, Giulietti M. Influence of magnetic field in the kinetics of crystallization of diamagnetic and paramagnetic inorganic salts. Cryst Res Technol 1999; 34: 1239–44. [248] Toyohara N, Benino Y, Fujiwara T, Tanaka S, Uematsu K, Komatsu T. Crystal orientation in Bi-based superconducting glass-ceramics prepared in high magnetic field. Physica C 2005; 420: 88–94. [249] Toyohara N, Benino Y, Fujiwara T, Tanaka S, Uematsu K, Komatsu T, Takahashi Y. Enhancement and depression in second-order optical nonlinearity of Ba2TiGe2O8 in crystallized glass prepared in a high magnetic field. J Appl Phys 2006; 99: 043515. [250] Tucker DS, Lapointe MR, Jia ZY. The effects of a magnetic field on the crystallization of a fluorozirconate glass. J Mater Res 2007; 22: 1431–4. [251] Worsch C, Schaaf P, Harizanova R, Rüssel C. Magnetisation effects of multicore magnetite nanoparticles crystallised from a silicate glass. J Mater Sci 2012; 47: 5886–90. [252] Durschang BR, Carl G, Rüssel C, Gutzow I. Influence of pressure and shear flow on the crystallization behaviour of a glass melt based on the Li 2Si2O5 -composition. Ber Bunsenges Phys Chem 1996; 100: 1456–8. [253] Habelitz S, Carl G, Rüssel C. Processing, microstructure and mechanical properties of extruded mica glass-ceramics. Mater Sci Eng A 2001; 307: 1–14. [254] Habelitz S, Höche T, Hergt R, Carl G, Rüssel C. Microstructural design through epitaxial growth in extruded mica glass-ceramics. Acta Mater 1999; 47: 2831–40. [255] Höche T, Habelitz S, Avramov I. Crystal morphology engineering in SiO2-Al2O3 MgO-K2O-Na2O-F-mica glass-ceramics. Acta Mater 1999; 47: 735–44. [256] Habelitz S, Carl G, Rüssel C, Thiel S, Gerth U, Schnapp JD, et al. Mechanical properties of oriented mica glass ceramic. J Non–Cryst Solids 1997; 220: 291–8. [257] Yue YZ, Moisescu C, Carl G and Rüssel C. Influence of the suspended iso- and anisometric crystals on the flow of fluroapatite glass ceramic during extrusion. Phys Chem Glass 1999; 40: 243–7. [258] Wisniewski W, de Kloe R, Rüssel C. EBSD-measurements of textured apatite glass-ceramics. CrystEngComm 2015; 17: 2969–73. 129

[259] Hosono H, Shimizu Y, Ohsato H, Abe Y. Effect of crystallization temperature and additives on the orientation degree of calcium phosphate glass-ceramics prepared by unidirectional crystallization. J Mater Sci Lett. 1987; 6: 394–6. [260] Wang QQ, Han JB, Gong HM, Chen DJ, Zhao XJ, Feng JY, et al. Linear and nonlinear optical properties of Ag nanowire polarizing glass. Adv Funct Mater 2006; 16: 2405–8. [261] Wisniewski W, Otto K, Rüssel C. Oriented nucleation of diopside crystals in glass. Cryst Growth Des 2012; 12: 5035−41. [262] Wisniewski W, Nagel M, G. lksch, Rüssel C. Electron Backscatter Diffraction of Fresnoite crystals grown from the surface of a 2BaO·TiO2 ·2.75SiO2 Glass. Cryst Growth Des 2010; 10: 1414−8. [263] Yamazaki Y, Takahashi Y, Ihara R, Fujiwara T. Surface crystallization of fresnoite-type crystallized glasses with large thickness. J. Ceram. Soc. Japan 2011; 119: 757–62. [264] Ding Y, Miura Y, Osaka A. Stimulated surface crystallization of –barium borate on glass due to ultrasonic treatment and second harmonic generation. J Mater Res 1996; 11: 495–502. [265] Ding Y, N. Masuda, Miura Y, Osaka A. Preparation of polar oriented Sr2TiSi2O8 films by surface crystallization of glass and second harmonic generation. J Non–Cryst Solids 1996; 203: 88–95. [266] Ding Y, Osaka A, Miura Y, Toratani H, Matsuoka Y. Second order optical nonlinearity of surface crystallized glass with lithium niobate. J Appl Phys 1995; 77: 2208–10. [267] Ding Y, Miura Y, Osaka A. Polar-oriented crystallization of fresnoite (Ba2TiSi2O8 ) on glass surface due to ultrasonic treatment with suspensions. J Am Ceram Soc 1994; 77: 2905–10. [268] Fang ZJ, Zheng SP, Peng WC, Zhang H, Ma ZJ, Zhou SF, et al. Fabrication and characterization of glass-ceramic fiber-containing Cr3+-doped ZnAl2O4 nanocrystals. J Am Ceram Soc 2015; 98: 2772–5. [269] Fang ZJ, Zheng SP, Peng WC, Zhang H, Ma ZJ, Dong GP, et al. Ni2+ doped glass ceramic fiber fabricated by melt-in-tube method and successive heat treatment. Opt Express 2015; 23: 28258–63. [270] Mashinsky VM, Karatun NM, Bogatyrev VA, Sigaev VN, Golubev NV, Ignat'eva ES, et al. Microfluorescence analysis of nanostructuring inhomogeneity in optical fibers with embedded gallium oxide nanocrystals. Microsc. Microanal. 2012; 18, 259–65. [271] Yu YZ, Fang ZJ, Ma CS, Inoue H, Yang G, Zheng SP, et al. Mesoscale engineering of photonic glass for tunable luminescence. NPG Asia Mater 2016; 8: e318. [272] Hou C, Jia XT, Wei L, Tan SC, Zhao X, Joannopoulos JD, et al. Crystalline silicon core fibres from aluminium core preforms. Nat Commun 2015; 6: 6248. [273] Fang ZJ, Li Y, Zhang FT, Ma ZJ, Dong GP, Qiu JR, Enhanced Sunlight Excited 1-m Emission in Cr3+-Yb3+ codoped transparent glass-ceramics containing Y3Al5O12 nanocrystals. J Am Ceram Soc 2015; 98: 1105–10. [274] Reda SM. Synthesis and optical properties of CdS quantum dots embedded in silica matrix thin films and their applications as luminescent solar concentrators. Acta Mater 2008;56: 259–64. [275] Zhang QY, Huang XY. Recent progress in quantum cutting phosphors. Prog Mater Sci 2010; 55: 353–427. [276] Ye S, Zhu B, Chen JX, Luo J, Qiu JR. Infrared quantum cutting in Tb3+,Yb3+ codoped transparent glass ceramics containing CaF 2 nanocrystals. Appl Phys Lett 2008; 92: 141112. [277] Ye S, Zhu B, Luo J, Chen JX, Lakshminarayana G, Qiu JR. Enhanced cooperative quantum cutting in Tm3+- Yb3+ codoped glass ceramics containing LaF3 nanocrystals. Opt Express 2008; 16: 8989–94. [278] Chen DQ, Wang YS, Yu YL, Huang P, Weng FY, Near-infrared quantum cutting in transparent nanostructured glass ceramics. Opt Lett 2008; 33: 1884–6. [279] Chen DQ, Yu YL, Wang YS, Huang P, Weng FY. Cooperative energy transfer up-conversion and quantum cutting down-conversion in Yb3+:TbF3 nanocrystals embedded glass ceramics. J Phys Chem C 2009; 113: 6406–10. 130

[280] Lin H, Chen DQ, Yu YL, Yang AP, Wang YS. Near-infrared quantum cutting in Ho3+/Yb3+ codoped nanostructured glass ceramic. Opt Lett 2011; 36: 876–8. [281] Fedorov PP, Luginina AA, Popov AI. Transparent oxyfluoride glass ceramics. J. Fluorine Chem. 2015; 172: 22–50. [282] de Pablos–Martín A, Durán A, Pascual MJ. Nanocrystallisation in oxyfluoride systems: mechanisms of crystallisation and photonic properties. Int. Mater. Rev. 2012; 57: 165–86. [283] Tick PA, Borrelli NF, Cornelius LK, Newhouse MA. Transparent glass-ceramics for 1300-nm amplifier applications. J Appl Phys 1995; 78: 6367–75. [284] Dejneka MJ. The luminescence and structure of novel transparent oxyfluoride glass-ceramics. J Non–Cryst Solids. 1998; 239: 149–55. [285] Suzuki T, Masaki S, Mizuno K, Ohishi Y. Preparation of novel transparent glass-ceramics containing fluoride crystals. Opt Mater 2011; 33: 1943–7. [286] Gugov I, Mueller M, Rüssel C. Transparent oxyfluoride glass ceramics co-doped with Er3+ and Yb3+ - Crystallization and upconversion spectroscopy. J. Solid State Chem. 2011; 184: 1001–7. [287] Stevenson AJ, Serier–Blrault H, Gredin P, Mortier M. Fluoride materials for optical applications: Single crystals, ceramics, glasses, and glass-ceramics. J. Fluorine Chem. 2011; 132: 1165–73. [288] Liu F, Ma E, Chen DQ, Yu YL, Wang YS. Tunable red-green upconversion luminescence in novel transparent glass ceramics containing Er: NaYF4 nanocrystals. J Phys Chem B 2006; 110: 20843–6. [289] Liu XF, Qiu JR. Recent advances in energy transfer in bulk and nanoscale luminescent materials: from spectroscopy to applications. Chem. Soc. Rev. 2015; 44: 8714–46. [290] Fujita S, Sakamoto A, Tanabe S. Luminescence Characteristics of YAG glass–ceramic phosphor for white LED. IEEE J. Sel. Top. Quant. 2008; 4: 1387–91. [291] Fujita S, Umayahara Y, Tanabe S. Influence of light scattering on luminous efficacy in Ca:YAG glass-ceramic phosphor. J Ceram Soc Jpn 2010; 118: 128–31. [292] Nishiura S, Tanabe S, Fujioka K, Fujimoto Y. Properties of transparent Ce:YAG ceramic phosphors for white LED. Opt Mater 2011; 33: 688–91. [293] Kunitomo J, Suzuki R, Takahashi Y, Miyazaki T, Terakado M, Fujiwara T. Red-emissive Mn-doped Li2Ge4O9 phase synthesized via glass-ceramic route. J Ceram Soc Jpn 2014; 122: 725-7. [294] Lee JS, Arunkumar P, Kim S, Lee IJ, Lee HG, Bin W. Smart design to resolve spectral overlapping of phosphor-in-glass for high-powered remote-type white light-emitting devices. Opt Lett 2014; 39: 762–5. [295] Zhang R, Lin H, Yu YL, Chen DQ, Xu J, Wang YS. A new-generation color converter for high-power white LED: transparent Ce3+:YAG phosphor-in-glass. Laser Photonics Rev 2014; 8: 158–64. [296] Chen DQ, Xiang WD, Liang XJ, Zhong JS, Yu H, Ding MY, et al. Advances in transparent glass–ceramic phosphors for white light-emitting diodes—A review. J Euro Ceram Soc 2015; 35: 859–69. [297] Wang FY, Lin Y, Shi HL, Wang WC, Deng ZH, Chen J, et al. Introduction on the fabrication technique of phosphor in glass by tape-casting and investigation on the chromaticity property. Opt Express 2014; 22: A1355–62. [298] Zhang XJ, Yu JB, Wang J, Lei BF, Liu YL, Cho YJ, et al. All-inorganic light convertor based on phosphor-in-glass engineering for next-generation modular high-brightness white LEDs/LDs. ACS Photonics 2017; 4: 986-95. [299] Deng JK, Li W, Zhang HR, Liu YL, Lei BF, Zhang HM, et al. Eu3+-doped phosphor-in-glass: a route toward tunable multicolor materials for near-UV high-power warm-white LEDs. Adv Opt Mater 2017; 5: 1600910.

131

[300] Zhang XJ, Yu JB, Wang J, Zhu CB, Zhang JH, Zou R, et al. Facile preparation and ultrastable performance of single component white-light-emitting phosphor-in-glass used for high power warm white LEDs. ACS Appl Mater 2015; 7: 28122–7. [301] Lin H, Wang B, Xu J, Zhang R, Chen H, Yu YL, et al. Phosphor-in-glass for high-powered remote-type white AC-LED. ACS Appl Mater Interfaces 2014; 6: 21264−9. [302] Fujita S, Tanabe S. Glass-ceramics and solid-state lighting. Int J Appl Glass Sci 2015; 6: 356–63. [303] Segawa H, Hirosaki N. Europium-doped SiAlON and borosilicate glass composites for white light-emitting diode. Appl Opt. 2015; 54: 8727–30. [304] Segawa H, Hirosaki N, Ohki S, Deguchi K, Shimizu T. Exploration of metaphosphate glasses dispersed with Eu-doped SiAlON for white LED applications. Opt Mater 2015; 42: 399–405. [305] Segawa H, Ogata S, Hirosaki N, Inoue S, Shimizu T, Tansho M, et al. Fabrication of glasses of dispersed yellow oxynitride phosphor for white light-emitting diodes. Opt Mater 2010; 33: 170–5. [306] Segawa H, Hirosaki H. Fabrication of sodium borosilicate glass and Eu-doped SiAlON composites for white LED. J Ceram Soc Jpn. 2015; 123: 452–5. [307] Mirov SB, Fedorov VV, Martyshkin DV, Moskalev IS, Mirov MS, Gapontsev VP. Progress in mid-IR Cr2+ and Fe2+ doped II-VI materials and lasers [Invited]. Opt Mater Express 2011; 1: 898–910. [308] Martyshkin DV, Goldstein JT, Fedorov VV, Mirov SB. Crystalline Cr2+: ZnSe/chalcogenide glass composites as active mid-IR materials. Opt Lett 2011; 36: 1530–2. [309] Karaksina EV, Shiryaev VS, Ketkova LA. Preparation of composite materials for fiber optics based on chalcogenide glasses containing ZnS(ZnSe):Cr 2+ crystals. J Non–Cryst Solids 2013; 377: 220–4. [310] Xu XH, Zhang WF, Yang DC, Lu W, Qiu JB, Yu SF. Phonon-assisted population inversion in lanthanide-doped upconversion Ba2LaF7 nanocrystals in glass-ceramics. Adv Mater 2016; 28: 8045–50. [311] Dejneka MJ. Transparent oxyfluoride glass ceramics. MRS Bull 1998; 23: 57–62. [312] Hayashi H, Tanabe S, Hanada T. 1.4 m band emission properties of Tm 3+ ions in transparent glass ceramics containing PbF2 nanocrystals for S-band amplifier. J Appl Phys 2001; 89: 10415. [313] Tick PA, Are low-loss glass-ceramic optical waveguides possible? Opt Lett 1998; 23: 1904–5. [314] Samson BN, Pinckney LR, Wang J, Beall GH, Borrelli NF. Nickel-doped nanocrystalline glass-ceramic fiber. Opt Lett 2002; 27: 1309–11. [315] Zhou SF, Dong HF, Feng GF, Wu BT, Zeng HP, Qiu JR. Broadband optical amplification in silicate glass ceramic containing β-Ga2O3:Ni2+ nanocrystals. Opt Express 2007; 15: 5477–81. [316] Zhou SF, Li CY, Yang, Bi G, Xu BB, Hong ZL, et al. Self-limited nanocrystallization-mediated activation of semiconductor nanocrystal in an amorphous solid. Adv Funct Mater 2013; 23: 5436–43. [317] Zhou SF, Jiang N, Wu BT, Hao JH, Qiu JR. Ligand-driven wavelength-tunable and ultrabroadband infrared luminescence in single-ion-doped transparent hybrid materials. Adv Funct Mater 2009; 19: 2081–8. [318] Wu BT, Jiang N, Zhou SF, Chen DP, Zhu CS, Qiu JR. Transparent Ni2+-doped silicate glass ceramics for broadband near-infrared emission. Opt Mater 2008; 30: 1900–4. [319] Zhuang YX, Tanabe S, Qiu JR. Wavelength tailorability of broadband near-infrared luminescence in Cr4+-activated transparent glass-ceramics. J Am Ceram Soc 2014; 97: 3519–23. [320] Ivanov VA, Simanovskiy DV, Marychev MO, Andreev PV, Koseva I, Tzvetkov P, et al. Ca2GeO4:Cr4+ transparent nano-glass ceramics. J Non-cryst Solids 2017; 456: 76–82. [321] Zhuang YX and Tanabe S, Conversion of Valence State and Coordination State of Fe in Transparent Glass-Ceramics Containing Li2ZnSiO4 Nanocrystals, J Am Ceram Soc 2013, 96: 2864–9. [322] Wu BT, Zhou SF, Ruan J, Qiao YB, Chen DP, Zhu CS, et al. Energy transfer between Cr3+ and Ni2+ in transparent silicate glass ceramics containing Cr3+/Ni2+ co-doped ZnAl2O4 nanocrystals. Opt Express 2008; 16: 2508–13. 132

[323] Wu BT, Zhou SF, Ruan J, Qiao YB, Chen DP, Zhu CS, et al. Enhanced broadband near-infrared luminescence from transparent Yb 3+/Ni2+ codoped silicate glass ceramics. Opt Express 2008; 16: 1879–84. [324] Sigaev VN, Golubev NV, Ignat'eva ES, Savinkov VI, Campione M, Lorenzi R, et al. Nickel-assisted growth and selective doping of spinel-like gallium oxide nanocrystals in germano-silicate glasses for infrared broadband light emission. Nanotechnology 2012; 23: 015708. [325] Suzuki T, Ohishi Y. Broadband 1400 nm emission from Ni 2+ in zinc-alumino-silicate glass. Appl. Phys. Lett. 2004; 84: 3804–6. [326] Suzuki T, Horibuchi K, Ohishi Y. Structural and optical properties of ZnO-Al2O3-SiO2 system glass-ceramics containing Ni2+-doped nanocrystals. J. Non-Cryst. Solids 2005; 351: 2304–9. [327] Sharonov MY, Bykov AB, Owen S, Petricevic V, Alfano RR, Beall GH, et al. Spectroscopic study of transparent forsterite nanocrystalline glass-ceramics doped with chromium. J Opt Soc Am B 2004; 21: 2046–52. [328] Tanabe S, Feng X. Temperature variation of near-infrared emission from Cr4+ in aluminate glass for broadband telecommunication. Appl Phys Lett 2000; 77: 818. [329] Zhuang YX, Teng Y, Luo J, Zhu B, Chi YZ, Wu E, et al. Broadband optical amplification in silicate glass ceramics containing Li2ZnSiO4:Cr4+ nanocrystals. Appl Phys Lett 2009; 95: 111913. [330] Wundke K, Auxier J, Schülzgen A, Peyghambarian N, Borelli NF. Room-temperature gain at 1.3 m in PbS-doped glasses. Appl Phys Lett 1999; 75: 3060–2. [331] Dong GP, Wu BT, Zhang FT, Zhang LL, Peng MY, Chen DD, et al. Broadband near-infrared luminescence and tunable optical amplification around 1.55 mu m and 1.33 mu m of PbS quantum dots in glasses. J Alloy Compd 2011; 509: 9335–9. [332] Rogulis U, Elsts E, Jansons J, Sarakovskis A, Doke G, Stunda A, et al. Rare earth activated oxyfluoride glasses and glass-ceramics for scintillation applications. IEEE T. Nuclear Sci. 2012; 59: 2201–6. [333] Cao JK, Wang XY, Li XM, Wei YL, Chen LP, Guo H. Enhanced emissions in Tb3+-doped oxyfluoride scintillating glass ceramics containing KLu2 F7 nano-crystals. J Lumin 2016; 170: 207–11. [334] Struebing C, Chong JH, Lee GH, Zavala M, Erickson A, Ding Y, et al. A neutron scintillator based on transparent nanocrystalline CaF2:Eu glass ceramic. Appl Phys Lett 2016; 108: 153106. [335] Edgar A, Secu M, Williams GVM, Schweizer S, Spaeth JM. Structural phase changes in barium bromide nano-crystals in a fluorobromozirconate glass-ceramic x-ray storage phosphor. J. Phys.: Condens. Matter 2001; 13: 6259–69. [336] Beckert MB, Gallego S, Elder E, Nadler J. Transitioning glass-ceramic scintillators for diagnostic x-ray imaging from the laboratory to commercial scale. SPIE 2005; 5745: 1351–8. [337] Johnson JA, Schweizer S, Henke B, Chen G, Woodford J, Newman PJ, MacFarlane DR. Eu-activated fluorochlorozirconate glass-ceramic scintillators. J Appl Phys 2006; 100: 034701. [338] Beckert MB, Gallego S, Ding D, Elder E, Nadler JH. Medical imaging scintillators from glass-ceramics using mixed rare-earth halides. Opt Mater 2016; 60: 51320. [339] Edgar A, Spaeth JM, Schweizer S, Assmann S, Newman PJ, MacFarlane DR. Photostimulated luminescence in a rare earth-doped fluorobromozirconate glass ceramic. Appl Phys Lett 1999; 75: 2386–8. [340] Okada G, Ueda J, Tanabe S, Belev G, Wysokinski T, Chapman D, et al. Samarium-doped oxyfluoride glass-ceramic as a new fast erasable dosimetric detector material for microbeam radiation cancer therapy applications at the canadian synchrotron. J Am Ceram Soc 2014: 97; 2147–53. [341] Zhuang YX, Ueda J, Tanabe S. Multi-color persistent luminescence in transparent glass ceramics containing spinel nano-crystals with Mn2+ ions. Appl Phys Lett 2014; 105: 191904. [342] Jain H. Transparent ferroelectric glass-ceramics. Ferroelectrics 2004; 306: 111–27. [343] Zhu B, Qian B, Liu Y, Xu C, Liu C, Chen QQ, et al. A volumetric full-color display realized by frequency upconversion of a transparent composite incorporating dispersed nonlinear optical crystals. NPG Asia Mater 2017; 9: e394. 133

[344] Baudrier–Raybaut M, Haȉdar R, Kupecek P, Lemasson P, Rosencher E. Random quasi-phase-matching in bulk polycrystalline isotropic nonlinear materials. Nature 2004; 432: 374–6. [345] Downing E, Hesselink L, Ralston J, Macfarlane R. A three-color, solid-state, three-dimensional display. Science 1996; 273: 1185–9. [346] Kurtz SK, Perry TT, A powder technique for evaluation of nonlinear optical materials. J Appl Phys 1968; 39: 3798–813. [347] Jerphagnon J, Kurtz SK. Maker Fringes - A detailed comparison of theory and experiment for isotropic and uniaxial crystals. J Appl Phys 1970; 41: 1667–81. [348] Takahashi Y, Kitamura K, Benino Y, Fujiwara T, Komatsu T. Second-order optical nonlinear and luminescent properties of Ba2TiSi2O8 nanocrystallized glass. Appl Phys Lett 2005; 86: 091110. [349] Takahashi Y, Benino Y, Fujiwara T, Komatsu T. Optical second order nonlinearity of transparent Ba2TiGe2O8 crystallized glasses. Appl Phys Lett 2002; 81: 223–5. [350] Takahashi Y, Benino Y, Fujiwara T, Komatsu T. Large second-order optical nonlinearities of fresnoite-type crystals in transparent surface-crystallized glasses. J Appl Phys 2004; 95: 3503–8. [351] Truong LN, Dussauze M, Fargin E, Santos L, Vigouroux H, Fargues A, et al. Isotropic octupolar second harmonic generation response in LaBGeO5 glass-ceramic with spherulitic precipitation. Appl Phys Lett 2015; 106: 161901. [352] Vigouroux H, Fargin E, Gomez S, Le Garrec B, Mountrichas G, Kamitsos E, et al. Synthesis and multiscale evaluation of LiNbO3-containing silicate glass-ceramics with efficient isotropic SHG response. Adv Funct Mater 2012; 22: 3985–93. [353] Calvez L, Roze M, Ledemi Y, Ma HL, Lucas J, Allix M, et al. Controlled crystallization in Ge-(Sb/Ga)-(S/Se)-MX glasses for infrared applications. J Ceram Soc Jpn 2008; 116: 107982. [354] Tao HZ, Lin CG, Gu SX, Jing CB, Zhao XJ. Optical second-order nonlinearity of the infrared transmitting 82GeS2·18CdGa2S4 nanocrystallized chalcogenide glass. Appl Phys Lett 2007; 91: 011904. [355] Zheng XL, Tao HZ, Lin CG, Gu SX, Dong GP, Zhao XJ. Second harmonic generation in surface crystallized 30GeS2·35Ga2S3·35AgCl chalcohalide glasses. Opt Mater 2009; 31: 1434–8. [356] Liu QM, Zhang P. IR Li2Ga2GeS6 nanocrystallized GeS2-Ga2S3 -Li2S electroconductive chalcogenide glass with good nonlinearity. Sci Rep 2014; 4: 5719. [357] Tao HZ, Zhao XJ, Liu QM. Optical non-linearity in nano- and micro-crystallized glasses. J Non–Cryst Solids 2013; 377: 146–50. [358] Lin CG, Calvez L, Bureau B, Tao HZ, Allix M, Hao ZX, et al. Second-order optical nonlinearity and ionic conductivity of nanocrystalline GeS2-Ga2S3 -LiI glass-ceramics with improved thermo-mechanical properties. Phys Chem Chem Phys 2010; 12: 3780–7. [359] Dong GP, Tao HZ, Xiao XD, Lin CG, Guo HT, Zhao XJ. Second harmonic generation in transparent microcrystalline alpha-CdGa2S4-containing chalcogenide glass ceramics. Opt Commun 2007; 274: 466–70. [360] Lin CG, Tao HZ, Pan RK, Zheng XL, Dong GP, Zang HC, et al. Permanent second-harmonic generation in AgGaGeS4 bulk-crystallized chalcogenide glasses. Chem Phys Lett 2008; 460: 125–8. [361] Fang ZJ, Xiao XS, Wang X, Ma ZJ, Lewis E, Farrell G, et al. Glass-ceramic optical fiber containing Ba2TiSi2O8 nanocrystals for frequency conversion of lasers. Sci Rep 2017; 7: 44456. [362] Chakraborty P. Metal nanoclusters in glasses as non-linear photonic materials. J Mater Sci 1998; 33: 2235–49. [363] Uchida K, Kaneko S, Omi S, Hata C, Tanji H, Asahara Y, et al. Optical nonlinearities of a high concentration of small metal particles dispersed in glass: copper and silver particles. J Opt Soc Am B 1994; 11: 1236–43. [364] Nakamura A, Yamada H, Tokizaki T. Size-dependent radiative decay of excitons in CuCl semiconducting quantum spheres embedded in glasses. Phys Rev B, 1989; 40: 8585–8.

134

[365] Kondo Y, Kuroiwa Y, Sugimoto N, Manabe T, Ito S, Yoko T, et al. Ultraviolet irradiation effect on the third-order optical nonlinearity of CuCl-microcrystallite-doped glass. J Opt Soc Am B 2000; 17: 548–54. [366] Li YL, M. Takata, Nakamura A. Size-dependent enhancement of nonlinear optical susceptibilities due to confined excitons in CuBr nanocrystals. Phys Rev B 1998; 57: 9193–200. [367] Y. Kuroiwa, N. Sugimoto, Y. Kondo, T. Manabe, S. Ito, Nakamura A. Quenching treatment for improvement of nonlinear optical response time in high-x CuCl-doped glasses. Opt Commun 1999; 170: 285–9. [368] Li YL, Ohata M, Sasaki S, Nakamura A. Size-Dependent Enhancement of Third-order Optical Nonlinearity in CuBrxCl1-x Nanocrystals Embedded in Glass. Jpn. J Appl Phys 1998; 37: L33–5. [369] Xiang WD, Cao HH, Ma L, Huang YY, Pei L, Liang XJ. Valence state control and third-order nonlinear optical properties of copper embedded in sodium borosilicate glass. ACS Appl. Mater. Interfaces 2015; 7: 10162-8. [370] Chen FF, Dai SX, Lin CG, Yu QS, Zhang QY, Performance improvement of transparent germanium–gallium–sulfur glass ceramic by gold doping for third-order optical nonlinearities. Opt Express 2013; 21: 24847–55. [371] X. Shen, Chen FF, Lv X, Dai SX, Wang XS, Zhang W, et al. Preparation and third-order optical nonlinearity of glass ceramics based on GeS2–Ga2S3–CsCl pseudo-ternary system. J Non–Cryst Solids 2011; 357: 2316–9. [372] Lin CG, Calvez L, Ying L, Chen FF, Song BA, X. Shen, et al. External influence on third-order optical nonlinearity of transparent chalcogenide glass-ceramics. Appl Phys A 2011; 104: 615–20. [373] Keller U. Recent developments in compact ultrafast lasers. Nature 2003; 424: 831–8. [374] Liu XF, Qiu JR. Emerging low dimensional materials for nonlinear optics and ultrafast photonics. Adv Mater 2017; 29: 1605886. [375] Dymshits O, Shepilov M, Zhilin A. Transparent glass-ceramics for optical applications. MRS Bull 2017; 42: 200–5 [376] Denisov IA, Volk YV, Malyarevich AM, Yumashev KV, Dymshits OS, Zhilin AA, et al. Linear and nonlinear optical properties of cobalt-doped zinc aluminum glass ceramics. J App Phys 2003; 93: 3827–31. [377] Boiko RM, Okhrimchuk AG, Shestakov AV. Glass ceramics Co2+ saturable absorber Q-switch for 1.3  1.6 m spectral region. OSA Trends in Optics & Photonics, 1998; 19: 185–8. [378] Malyarevich AM, Denisov IA, Yumashev KV, Dymshits OS, Zhilin AA, Kang U. Cobalt-doped transparent glass ceramic as a saturable absorber Q switch for Erbium:glass lasers. Appl Opt 2001; 40: 4322–25. [379] Loiko P, Dymshits OS, Vitkin VV, Skoptsov NA, Zhilin AA, Shemchuk DV, et al. Saturable absorber: transparent glass-ceramics based on a mixture of Co:β-Zn2SiO4 and Co:ZnO nanocrystals. Appl Opt 2016; 55: 5505–12. [380] Loiko P, Dymshits OS, Vitkin VV, Skoptsov NA, Kharitonov AA, Zhilin AA, et al. Glass-ceramics with γ-Ga2O3:Co2+ nanocrystals: saturable absorber for 1.5–1.7 μm Er lasers. Laser Phys Lett 2015; 12: 035803. [381] Goldstein A, Loiko P, Burshtein Z, Skoptsov N, Glazunov I, Galun E, et al. Development of Saturable absorbers for laser passive Q-switching near 1.5 m based on transparent ceramic Co2+:MgAl2O4. J Am Ceram Soc 2016; 99: 1324–31. [382] Vitkin V, Loiko P, Dymshits O, Zhilin A, Alekseeva I, Sabitova D, et al. Passive Q-switching of an Er, Yb:glass laser with Co:Mg(Al,Ga)2O4 -based glass-ceramics. Appl Opt 2017; 56: 2142–9 [383] Yu YL, Huang F, Chen DQ, Lin H, Zhang R, Yang AP, et al. Co2+/Er3+ co-doped transparent glass ceramic containing both spinel ZnAl2O4 and orthorhombic YF3 for self-Q-switched laser. Laser Phys 2014; 24: 025101. [384] Rafailov EU, Cataluna MA, Sibbett W. Mode-locked quantum-dot lasers. Nat Photonics 2007; 1: 395–401. 135

[385] Lagatsky AA, Leburn CG, Brown CTA, Sibbett W, Zolotovskaya SA, Rafailov EU. Ultrashort-pulse lasers passively mode-locked by quantum-dot-based saturable absorbers. Prog Quant Electron 2010; 34: 1–45. [386] Guerreiro PT, Ten S, Borreli NF, Butty J, Jabbour GE, Peyghambarian N. PbS quantum-dot doped glasses as saturable absorbers for mode locking of a Cr:forsterite laser. App Phys Lett 1997; 71: 1595–7. [387] Bilinsky IP, Prasankumar RP, Fujimoto JG. Self-starting mode locking and Kerr-lens mode locking of a Ti:Al2O3 laser by use of semiconductor-doped glass structures. J Opt Soc Am B 1999; 16: 546–9. [388] Denisov IA, Skoptsov NA, Gaponenko MS, Malyarevich AM, Yumashev KV, Lipovskii AA. Passive mode locking of 2.09 m Cr,Tm,Ho:Y3Sc2Al3O12 laser using PbS quantum-dot-doped glass. Opt Lett 2009; 43: 3403–5. [389] Malyarevich AM, Savitsky VG, Denisov IA, Prokoshin PV, Yumashev KV, Raaben E, et al. PbS(Se) quantum dot doped glass applications as laser passive Q-switches. Phys Stat Sol B 2001; 224: 253–6. [390] Lagatsky AA, Leburn CG, Brown CTA, Sibbett W, Malyarevich AM, Savitski VG, et al. Passive mode locking of a Cr4+:YAG laser by PbS quantum-dot-doped glass saturable absorber. Opt Commun 2004; 241: 449–54. [391] Malyarevich AM, Yumashev KV, Lipovskii AA. Semiconductor-doped glass saturable absorbers for near-infrared solidstate lasers. J Appl Phys 2008; 103: 081301. [392] Onushchenko AA, Zhilin AA, Petrovski GT, Raaben EL, Gaponenko MS, Malyarevich AM, et al. Nanostructured glass-crystal materials with lead sulfide for passive Q switching of IR lasers. J Opt Technol 2006; 73: 576–83. [393] Llordés A, Garcia G, Gazquez J, Milliron DJ. Tunable near-infrared and visible-light transmittance in nanocrystal-in-glass composites. Nature 2013; 500: 323–6. [394] Garcia G, Buonsanti R, Runnerstrom EL, Mendelsberg RJ, Llordes A, Anders A, et al. Dynamically modulating the surface plasmon resonance of doped semiconductor nanocrystals. Nano Lett 2011; 11: 4415–20. [395] Korgel B. Composite for smarter windows. Nature 2013; 500: 278–9. [396] Yoshida K, Masai H, Takahashi Y, Ihara R, Fujiwara T. Selective crystallization of anatase and rutile by control of heat-treatment conditions and SiO2 addition in TiO2-crystallized glass. J Ceram Soc Jpn 2013; 121: 999. [397] Yoshida K, Takahashi Y, Ihara R, Terakado N, Fujiwara T, Kato H, et al. Large enhancement of photocatalytic activity by chemical etching of TiO2 crystallized glass. APL Mater 2014; 2: 106103. [398] Yoshida K, Takahashi H, Masai H, Terakado N, Takahashi Y, Fujiwara T, et al. Observation of visible light-driven water splitting by TiO2 crystallized glass. Int J Hydrogen Energy 2016; 41: 22055–8. [399] Gad–Allah TA, Margha FH. Photoactive transparent nano-crystalline glass-ceramic for remazole red dye degradation. Mater Res Bull 2012; 47: 4096–100. [400] Margha FH, Abdel–Wahed MS, Gad–Allah TA. Nanocrystalline Bi2O3 –B2O3 – (MoO3 or V2O5 ) glass-ceramic systems for organic pollutants degradation. Ceram Int 2015; 41: 5670–6. [401] Wang S, Liang K. High infrared radiance glass-ceramics obtained from fly ash and titanium slag. Chemosphere 2007; 69: 1798–801. [402] Wang SM, Kuang FH, Yan QZ, Ge CC, Qi LH. Crystallization and infrared radiation properties of iron ion doped cordierite glass-ceramics. J Alloy Compd 2011; 509: 2819-23. [403] Wang S. Effects of Fe on crystallization and properties of a new high infrared radiance glass-ceramics. Environ Sci Technol 2010; 44: 4816–20. [404] Maeda K, Hirose M, Kobayashi T. High performance transparent glass-ceramics for optical components. J Ceram Soc Jpn 2015; 123: 949–54. 136

[405] Ferrari M, Righini GC. Glass-Ceramic Materials for Guided-Wave Optics. Int J Appl Glass Sci 2015; 6: 240–8. [406] Jestin Y, Armellini C, Chiasera A, Chiappini A, Ferrari M, Moser E, et al. Low-loss optical Er3+-activated glass-ceramics planar waveguides fabricated by bottom-up approach. Appl Phys Lett 2007; 91: 071909. [407] Yamazaki Y, Takahashi Y, Ihara R, Fujiwara T. Ultra-low propagation losses in fresnoite-type precipitated crystallized glasses. Appl Phys Lett 2014; 104: 031901. [408] Zerodur® Information, Zerodur–Präzision aus Glaskeramik. Information SCHOTT 1991, 1–31. [409] Suzdal’tsev EI. Radio-transparent ceramics: yesterday, today, tomorrow. Refract Ind Ceram 2015; 55: 377–90.

137