International Journal of Applied Glass Science, 5/1/65-81(2014) DOI: 10. 111Iijag 12053 IN TERNATIONAL JOURNAL OF A pplied gl ass sC丨ENCE High-Performance Glass Fiber Development for Composite Applications Hong Li, *'I Cheryl Richards, * and James Watson Fiber glass Science and Technology, PPG Industries, Inc, 400 Guys Run Road, Cheswick Pennsylvania 15204 5 uQ The article provides a review of historical commercial glass fiber development and recent developments of high-perfor- mance glass fibers with improved mechanical performance for glass fiber-reinforced polymer-matrix composite applications lass composition design is outlined in conjunction with theoretical and experimental modeling approaches. Challenges in glass melting and fiber forming are briefly discussed. fiber processing. The article concludes with a summary that points to the continuous advancements in glass Glass fiber and glass fiber-reinforced plastic melting technology that meet the growing challenges of (GFRP) composite industries have been enjoying con- the commercial production of high-performance glass tinuous growth globally, especially in the most recent fibers decade. This article is intended to provide a general review, with examples, of the history of fiber glass Overview of Glass Fibers development as well as recent development. The review will cover glass fiber chemistry and composition design, mechanical property characterizations, and topics rele- Market growth vant to glass melting and fiber forming. This article by As GFRP materials have gained broader usage, no means provides a comprehensive review on the sub- global glass fiber output has steadily increased, to meet ject; rather, it provides researchers and professionals that demand, over the past decade. Applications in with an update on the state of glass fiber technology automotive, consumer goods, and industrial tanks, and with a focus on fibers with improved mechanical prop- piping markets have seen rapid expansion. Brisk growth erties. The article is divided into four major sections: has also been seen in wind turbine blades and printed (i)overview of glass fibers, (i) chemical approach to circuit board (PCB) applications. The growing glass fiber mechanical performance, (ii) glass fiber demands of existing GFRP products and identification mechanical property characterizations, and (iv) glass of new applications are further fueled by the mega ends of energy efficiency (automobile and aerospace "Members. The American Ceramic Society hlieppg.com industries requiring lighter weight), a cleaner environ- e 2013 The American Ceramic Sociery and wiley Periodicals, Inc ment(tied to energy efficiency and lower emissions)
High-Performance Glass Fiber Development for Composite Applications Hong Li,*,† Cheryl Richards,* and James Watson Fiber Glass Science and Technology, PPG Industries, Inc., 400 Guys Run Road, Cheswick, Pennsylvania 15204 The article provides a review of historical commercial glass fiber development and recent developments of high-performance glass fibers with improved mechanical performance for glass fiber-reinforced polymer–matrix composite applications. Glass composition design is outlined in conjunction with theoretical and experimental modeling approaches. Challenges in glass melting and fiber forming are briefly discussed. Introduction Glass fiber and glass fiber-reinforced plastic (GFRP) composite industries have been enjoying continuous growth globally, especially in the most recent decade. This article is intended to provide a general review, with examples, of the history of fiber glass development as well as recent development. The review will cover glass fiber chemistry and composition design, mechanical property characterizations, and topics relevant to glass melting and fiber forming. This article by no means provides a comprehensive review on the subject; rather, it provides researchers and professionals with an update on the state of glass fiber technology with a focus on fibers with improved mechanical properties. The article is divided into four major sections: (i) overview of glass fibers, (ii) chemical approach to glass fiber mechanical performance, (iii) glass fiber mechanical property characterizations, and (iv) glass fiber processing. The article concludes with a summary that points to the continuous advancements in glassmelting technology that meet the growing challenges of the commercial production of high-performance glass fibers. Overview of Glass Fibers Market Growth As GFRP materials have gained broader usage, global glass fiber output has steadily increased, to meet that demand, over the past decade. Applications in automotive, consumer goods, and industrial tanks, and piping markets have seen rapid expansion. Brisk growth has also been seen in wind turbine blades and printed circuit board (PCB) applications. The growing demands of existing GFRP products and identification of new applications are further fueled by the megatrends of energy efficiency (automobile and aerospace industries requiring lighter weight), a cleaner environment (tied to energy efficiency and lower emissions), *Members, The American Ceramic Society. † hli@ppg.com © 2013 The American Ceramic Society and Wiley Periodicals, Inc International Journal of Applied Glass Science, 5 [1] 65–81 (2014) DOI:10.1111/ijag.12053
International Journal of Applied Glass Science--Li, Richards, and Watson Vol.5,No.1,2014 and renewable energy production (wind turbines) E-Glass is the most widely used glass fber for Other key factors impacting the GFRP and is primarily composed of CaO, Al2O3, and expansion of the GFRP market have been the maturity SiOz, conditionally with B2O3 from 0 to 10 wt% of GFRP composites in commercial applications matu- E-Glass offers suitable mechanical properties(tensile rity and secure supplies of glass fiber products with strength and modulus), electrical properties [dielectric known performance features at competitive cost points. constant (Dk); dielectric loss (Df); and dielectric Figure I illustrates global glass fiber annual production, strength], and chemical stability for most GFRP appli whose growth in the recent decade has been strongly cations including those for PCB electronics and nume ous general industrial applications. General-purpose E-Glass fibers are defined according to ASTM D578 Classification and History of Commercial Glass specifications. Historically, E-Glass compositions started with relatively high concentrations of boron (B2O3)and Auorine (F or F2), which enhanced batch Glass fibers are the most common reinforcement melting, glass fining, and fiber drawing. Over the years, used for polymer-matrix composites and are classified E-Glass compositions with low or zero B2O3, and based on required key properties for specific composite essentially no fluoride, were developed to address envi- applications as highlighted in Fig. 2. The time periods ronmental and legislative regulatory requirements shown represent significant activities occurring These changes are reflected in the general purpose defi- research and development based on our literature search nition of E-Glass as shown in ASTM Standard D 578 and projections Section 4.2.2 4 A more specific designation for boron- free modified E-Glass composition is called out in Legend as improved resistance to corrosion by most acids 400 E-CR fiber development and commercialization and bushing technology that enabled high furnace fiber glass surfaced in the mid-1980s, was broadly cepted and produced in the 1990s, and is widely accepted today. Representative commercial E-CR fiber products in the market today include Advantexfrom 0042005200620072008200920102011 Owens Corning(OC, Columbus, OH), INNOFIBER CR fber glass from PPG Industries, Inc.(PPG, Pitts- burgh, PA), and E6-CR from Jushi Group Co. Ltd Fig.I. Global glass fiber production history(in metric ton). (Tongxiang, Zhejiang, China),etc. 1930194019501960197019801990200020102020 Fig. 2. History of major commercial glass fiber development
and renewable energy production (wind turbines). Other key factors impacting the growth and global expansion of the GFRP market have been the maturity of GFRP composites in commercial applications maturity and secure supplies of glass fiber products with known performance features at competitive cost points. Figure 1 illustrates global glass fiber annual production, whose growth in the recent decade has been strongly supported by Chinese production.1 Classification and History of Commercial Glass Fibers Glass fibers are the most common reinforcement used for polymer–matrix composites and are classified based on required key properties for specific composite applications as highlighted in Fig. 2. The time periods shown represent significant activities occurring in research and development based on our literature search and projections. E-Glass is the most widely used glass fiber for GFRP and is primarily composed of CaO, Al2O3, and SiO2, conditionally with B2O3 from 0 to 10 wt.%. E-Glass offers suitable mechanical properties (tensile strength and modulus), electrical properties [dielectric constant (Dk); dielectric loss (Df); and dielectric strength], and chemical stability for most GFRP applications including those for PCB electronics and numerous general industrial applications.2,3 General-purpose E-Glass fibers are defined according to ASTM D578 specifications.4 Historically, E-Glass compositions started with relatively high concentrations of boron (B2O3) and fluorine (F or F2), which enhanced batch melting, glass fining, and fiber drawing. Over the years, E-Glass compositions with low or zero B2O3, and essentially no fluoride, were developed to address environmental and legislative regulatory requirements. These changes are reflected in the general purpose defi- nition of E-Glass as shown in ASTM Standard D 578 Section 4.2.2.4 A more specific designation for boronfree modified E-Glass composition is called out in Section 4.2.4 of the standard. Known as E-CR Glass, it has improved resistance to corrosion by most acids. E-CR fiber development and commercialization have been further aided by advancements in furnace and bushing technology that enabled high furnace throughput. Large-scale commercialization of E-CR fiber glass surfaced in the mid-1980s, was broadly accepted and produced in the 1990s, and is widely accepted today. Representative commercial E-CR fiber products in the market today include Advantex from Owens Corning (OC, Columbus, OH), INNOFIBER CR fiber glass from PPG Industries, Inc. (PPG, Pittsburgh, PA), and E6-CR from Jushi Group Co. Ltd. (Tongxiang, Zhejiang, China), etc. Year 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 Glass fiber production (10 4 MT) 100 200 300 400 500 Legend Global China Fig. 1. Global glass fiber production history (in metric ton).1 Fig. 2. History of major commercial glass fiber development. 66 International Journal of Applied Glass Science—Li, Richards, and Watson Vol. 5, No. 1, 2014
amics.org/lAGS Higb-Performance Glass Fiber De Although E-CR glass fibers are evolving to beco fiber glass industry. High cost and technology barriers he GFRP industry standard for most corrosion-resis- remain as factors limiting the growth of D-Glass in tant applications, improved chemical resistance was PCB applications. One of the D-Glass derivatives, SI lly obtained from boron-free C-Glass fiber fibers (Nitto Boseki Co. Ltd, Fukushima, Japan) (Na2O, CaO, Al2O3, and Sio2), which offered good introduced small amounts of alkaline earth oxides and chemical resistance against acid attack. Boron-contain- alumina at the expense of boron to improve glass ing C-Glass fiber was invented in 1943(UE. Bowes, melting and fiber-forming characteristics. However, the US 2, 308, 857, OC, 1943), which had limited use in melting and forming processes remain significantly building material insulation applications because of challenged because of its lower resistance to acidic environments due to boron T relative to E-Glass. In late 2010, PPG introduced presence in the glass. The mechanical performance of INNOFIBER LD fiber glass, a low dielectric glass fiber boron-free C-Glass fiber is inferior to E-CR Glass and that offered glass-melting and fiber-forming characteris- E-Glass, and these properties limit its use as a rein- tics that were compatible with a E-Glass manufacturing fCO ment. Further limiting the broad commercial use pla Glass is the fact that it has lower hydrolytic resis S-Glass is primarily composed of MgO, Al2O3, tance under high humidity environments at elevated and Sio2 and was first developed in the 1960s primar temperatures. Starting in the mid-1960s, boron-free ily for high-temperature and high-strength applications C-Glass fiber and E-Glass fber products served the and later in 1970s for military ballistic protection general industrial market. Because of their lower cost, applications. S-Glass is difficult to fberize due to its boron-free C-Glass fiber products are still used in com- high liquidus temperature(1470.C). Liquidus tempera- bination with E-Glass fibers in nonstructural corrosion ture, Ti, is defined as the maximum temperature above barrier applications. However, boron-free C-Glass which all crystals are dissolved in molten glass. As a volume is <10% of fibers used in GFRP today. result, S-Glass has Ti greater than TE, indicating that Continuing on the theme of corrosion resistance, the glass exhibits a negative delta T(AT-TF-TD force concrete structures in the mid-1960s. This gla- Commercial examples include the S-2 Glass products AR-Glass fibers were developed as a solution to re from AGY(Aiken, SC). S-Glass derivatives, such as HS primarily composed of Na2O, CaO, ZrO2(17-24%), glass from Sinoma Science and Technology(Nanjing, and SiOz with a small amount of Al2O3. AR-Glass Jiangsu, China), offer melting technology improve fibers offer the highest resistance against alkaline as ments over S-2 Glass . Overall, however,commercial ell as acid-attacks. High concentrations of ZrO2 in applications using S-Glass fiber input are limited due the glass and its resultant higher melting temperature to significantly higher manufacturing costs in both (TM as defined by 10 Pa-s viscosity in glass industry) melting and fiber formi lead to very high product costs, which restricts its broad In the mid-1960s, R-Glass was first developed for use in general-purpose applications military applications. The glass is primarily composed As mentioned earlier, E-Glass offers adequate elec- of MgO, Cao, Al2O3, and SiOz. In its original chem trical properties, primarily driven by the low total alkali istry, such as S-Glass, usage was limited because of its (<2%)content of the glass, for PCB applications. high melting temperature requirement. Although these When higher signal transmission speeds in electronic early chemistries created melting challenges when man devices are required, D-Glass fibers or pure silica ufacturing R-Glass fibers, newly engineered R-Glass (SiO2) fibers offer the best achievable electrical proper- compositions have overcome the melting and fiber ties as measured using dielectric constant (Dk) and forming obstacles, becoming commercially attractive for dissipation factor(Df) among all fiber glass classes large-scale production D-Glass is also easier than pure SiO2 to melt and fiber ize. However, D-Glass fiber production is limited to small commercial scale because of its very high TN Historic Higb-Modulus and Higb-Strengtb Glass Fiber Developmen (1650° C)and high forming temperature(1400°C which is approximately 200C higher than traditional As earlier stated, both S-Glass and R-Glass have boron-containing E-Glass fiber products. Fiber-forming been restricted to limited applications because of high- temperature, TF, is defined at 100 Pa s viscosity by the temperature processing challenges preventing their
Although E-CR glass fibers are evolving to become the GFRP industry standard for most corrosion-resistant applications, improved chemical resistance was originally obtained from boron-free C-Glass fibers (Na2O, CaO, Al2O3, and SiO2), which offered good chemical resistance against acid attack. Boron-containing C-Glass fiber was invented in 1943 (U.E. Bowes, US 2,308,857, OC, 1943), which had limited use in building material insulation applications because of lower resistance to acidic environments due to boron presence in the glass. The mechanical performance of boron-free C-Glass fiber is inferior to E-CR Glass and E-Glass, and these properties limit its use as a reinforcement. Further limiting the broad commercial use of C-Glass is the fact that it has lower hydrolytic resistance under high humidity environments at elevated temperatures. Starting in the mid-1960s, boron-free C-Glass fiber and E-Glass fiber products served the general industrial market. Because of their lower cost, boron-free C-Glass fiber products are still used in combination with E-Glass fibers in nonstructural corrosion barrier applications. However, boron-free C-Glass volume is <10% of fibers used in GFRP today. Continuing on the theme of corrosion resistance, AR-Glass fibers were developed as a solution to reinforce concrete structures in the mid-1960s. This glass is primarily composed of Na2O, CaO, ZrO2 (17–24%), and SiO2 with a small amount of Al2O3. AR-Glass fibers offer the highest resistance against alkaline — as well as acid — attacks. High concentrations of ZrO2 in the glass and its resultant higher melting temperature (TM as defined by 10 Pas viscosity in glass industry) lead to very high product costs, which restricts its broad use in general-purpose applications. As mentioned earlier, E-Glass offers adequate electrical properties, primarily driven by the low total alkali (<2%) content of the glass, for PCB applications. When higher signal transmission speeds in electronic devices are required, D-Glass fibers or pure silica (SiO2) fibers offer the best achievable electrical properties as measured using dielectric constant (Dk) and dissipation factor (Df) among all fiber glass classes.5 D-Glass is also easier than pure SiO2 to melt and fiberize. However, D-Glass fiber production is limited to a small commercial scale because of its very high TM (1650°C) and high forming temperature (1400°C) which is approximately 200°C higher than traditional boron-containing E-Glass fiber products. Fiber-forming temperature, TF, is defined at 100 Pas viscosity by the fiber glass industry. High cost and technology barriers remain as factors limiting the growth of D-Glass in PCB applications. One of the D-Glass derivatives, SI fibers (Nitto Boseki Co. Ltd., Fukushima, Japan), introduced small amounts of alkaline earth oxides and alumina at the expense of boron to improve glassmelting and fiber-forming characteristics. However, the melting and forming processes remain significantly challenged because of its significantly higher TM and TF relative to E-Glass. In late 2010, PPG introduced INNOFIBER LD fiber glass, a low dielectric glass fiber that offered glass-melting and fiber-forming characteristics that were compatible with a E-Glass manufacturing platform.6,7 S-Glass is primarily composed of MgO, Al2O3, and SiO2 and was first developed in the 1960s primarily for high-temperature and high-strength applications and later in 1970s for military ballistic protection applications. S-Glass is difficult to fiberize due to its high liquidus temperature (1470°C). Liquidus temperature, TL, is defined as the maximum temperature above which all crystals are dissolved in molten glass. As a result, S-Glass has TL greater than TF, indicating that the glass exhibits a negative delta T (DT = TFTL). Commercial examples include the S-2 Glass products from AGY (Aiken, SC). S-Glass derivatives, such as HS glass from Sinoma Science and Technology (Nanjing, Jiangsu, China), offer melting technology improvements over S-2 Glass. 8,9 Overall, however, commercial applications using S-Glass fiber input are limited due to significantly higher manufacturing costs in both melting and fiber forming. In the mid-1960s, R-Glass was first developed for military applications. The glass is primarily composed of MgO, CaO, Al2O3, and SiO2. In its original chemistry, such as S-Glass, usage was limited because of its high melting temperature requirement. Although these early chemistries created melting challenges when manufacturing R-Glass fibers, newly engineered R-Glass compositions have overcome the melting and fiberforming obstacles, becoming commercially attractive for large-scale production. Historic High-Modulus and High-Strength Glass Fiber Development As earlier stated, both S-Glass and R-Glass have been restricted to limited applications because of hightemperature processing challenges — preventing their www.ceramics.org/IJAGS High-Performance Glass Fiber Development 67
International Journal of Applied Glass Science--Li, Richards, and Watson Vol.5,No.1,2014 use on large-scale commercial platforms. While no ulus have also been evaluated. For example, low silica breakthrough has been achieved in S-Glass fiber-pro- calcium aluminate glasses(by wt % 4-18 SiO2, 39-50 cessing technology, advances in R-Glass fibers, specifi CaO, and 39-48 Al2O3)were shown to offer high cally new glass composition development, have Young,s modulus between 98 and 112 GPa. All of significantly progressed since 2000. The accelerated the compositions lie across primary phase fields of ge development has been primarily driven by the market hlenite(2CaO Al2O3 SiO2), CaO A 2O3, and needs for longer wind turbine blades, which require 12Ca0. 7Al2O3 in the CaO-Al2O3-SiO2 ternary phase higher composite modulus. Commercial, large-scale diagram. These types of glasses are difficult to fiberize oduction of new R-Glass fibers, including INNOFI- because of their extremely high Ti(1335-1500C) BER XM glass fibers(PPG)o, I and OCV-HTM fibers greater than their T defined at 100 Pa's. To draw (OC), are now available for wind turbine manufac fibers without risking glass devitrification, which dis tures. The combination of new glass composition devel- ruts the continuous fber-forming process, one must opments of R-Glass fiber offerings aligned with markets draw the fibers at a higher temperature than TF or at that require higher performance has resulted in significantly lower viscosity than 100 Pas. As a result, ncreases in wind turbine blade length by 10-20% ver- fber-forming stability is adversely impacted, which will sus blades made using E-Glass fiber. In turn, the longe be discussed further lar blades improve power output by 21-41% without An extensive study of the liquidus surface of CaO, nificant increases in the overall blade weight. 3 MgO, and Cao/MgO aluminosilicate glasses has been In terms of glass chemistry, the new R-Glass fibers recently reported in a composition space(by wt %) are defined within the following space(by wt %) 56-65 0-22 CaO, 0-19 MgO, and 2-17.5 CaO and 1.5-13.5 SiO2, 13-20 Al2O3, 6-12 MgO, and 8-16 CaO. In Mgo. The reported compositions are plotted in per- addition, Li,O(1465C. In the commercial fiber-forming process, the is also possible, but result in a penalty in glass batch cost actual fiber-forming temperature must be kept great without significant product performance improvements. than Ti by no <50C to avoid glass crystallization The new R-Glass composition development space dis- prior to exiting bushings. Should these glasses be pro- tinctively differs original duced in fber forms, their actual forming temperatures R-Glass composition space, defined as(by wt %): 50- would be between 1320 and 1515C- substantially 60 SiO2, 25-26 Al2O3, 6-15 MgO, 2-9 CaO, and no higher than any known commercially produced new alkali(H. Scheller, Glass Compositions, German Patent, R-Glass fibers P1596751.8(C38351), Saint-Gobain, France, 1965) The new R-Glass fibers developed since 2000 Besides the new glass chemistry development for (cf. Table D) exhibit TL 1230C, except for HiPer better fiber processability, it should be emphasized that tex. Comparing the literature database with the successful production of new R-Glass compositions has commercially made new R-Glass fibers, it follows that also benefited from newer manufacturing technologies the liquidus surface of the MgO-CaO-Al2O3-SiO2 is that have been implemented in fiber glass industry more highly nonlinear than the predictions based on since the 1990s. These technologies include oxyfuel the phase diagrams(cf. Fig 3)once minor oxides are combustion for better energy delivery and efficiency, introduced to the primary composition space, including higher-quality refractory materials for higher-tempera- alkalis (mostly Li2O and Na2O), iron(Fe2O3 and ture operation electrical boost melting for enhanced FeO), titanium(TiO2), etc. melting capacity of the melter, and newer bus materials that provide longer service life at higher ating temperatures.(Note: A detailed review of Chemical Approach to Fiber Glass Mechanical manufacturing technology is beyond the scope of this Performance eyond the new composition developments in the section,several known models of glass R-Glass positions that offer high mod- Youngs modulus and theory of glass fracture and
use on large-scale commercial platforms. While no breakthrough has been achieved in S-Glass fiber-processing technology, advances in R-Glass fibers, specifi- cally new glass composition development, have significantly progressed since 2000. The accelerated development has been primarily driven by the market needs for longer wind turbine blades, which require higher composite modulus. Commercial, large-scale production of new R-Glass fibers, including INNOFIBER XM glass fibers (PPG)10,11 and OCV-HTM fibers (OC),12 are now available for wind turbine manufactures. The combination of new glass composition developments of R-Glass fiber offerings aligned with markets that require higher performance has resulted in increases in wind turbine blade length by 10–20% versus blades made using E-Glass fiber. In turn, the longer blades improve power output by 21–41% without significant increases in the overall blade weight.13 In terms of glass chemistry, the new R-Glass fibers are defined within the following space (by wt.%): 56–65 SiO2, 13–20 Al2O3, 6–12 MgO, and 8–16 CaO. In addition, Li2O (1465°C. In the commercial fiber-forming process, the actual fiber-forming temperature must be kept greater than TL by no <50°C to avoid glass crystallization prior to exiting bushings. Should these glasses be produced in fiber forms, their actual forming temperatures would be between 1320 and 1515°C — substantially higher than any known commercially produced new R-Glass fibers. The new R-Glass fibers developed since 2000 (cf. Table I) exhibit TL < 1230°C, except for HiPertexTM. Comparing the literature database20 with the commercially made new R-Glass fibers, it follows that the liquidus surface of the MgO-CaO-Al2O3-SiO2 is more highly nonlinear than the predictions based on the phase diagrams (cf. Fig. 3) once minor oxides are introduced to the primary composition space, including alkalis (mostly Li2O and Na2O), iron (Fe2O3 and FeO), titanium (TiO2), etc. Chemical Approach to Fiber Glass Mechanical Performance In this section, several known models of glass Young’s modulus and theory of glass fracture and 68 International Journal of Applied Glass Science—Li, Richards, and Watson Vol. 5, No. 1, 2014
R-Glass System Liquidus Temp TRE(Diopside, Anorthite, Quartz 1220-1410°C S-Glass System E-Glass System Sone SiO2 Liquidus Temp Cordierite) A203 1470°c Liquidus Tem CeO 9o% (Wollastonite 1070-1220°c Fig. 3. Typical cor ed as Sio Al,O, and(Cao+ MgO), of E-Glasses with and without boron(wi large blue circle), R-Glass(magenta), S-Glass (red), Low-SiO2 calcium aluminate glasses(purple), and high-Sioz alkaline earth(cao or MgO or both in gray triangle) aluminosilicate glasses on Cao-A2O3-SiOz Mgo-Al2O3-SiOz, and Cao-10%6MgO-Al2O3-SiOz phase failure probability by Weibull statistical analysis o=1/,(dUm/dr) Then, a general approach to glass chemistry design is The change of o with respect to the distance between two ions under the applied force, do/dr, leads to o=dr/ra(d Um/dr)=dE/ro(d- Um/dr) Young's Modulus of Glass and Glass Fibers The elastic modulus(E) is therefore inversely pro- Elastic modulus or Young's modulus of ionic- or portional to the fourth power of atomic spacing covalent-bonded inorganic crystalline solids is theoreti between two ions or two times of the Mandelung ally related to the electrostatic energy of attraction Uc (equal to -212241ro) of two opposite charged ions E=do/de=1/ro(d Um /dr2)=-2(a122/r4) (41, 22) with a spacing of ro. To account for many body interactions between ions within the solid. madelung energy(Um =aUo is used instead. The force applied to the two ions is dU/dr, and the related stress within Ithough the model has been cor a unit volume of the crystal r can be expressed as testing many crystalline materials, the a-value is not
failure probability by Weibull statistical analysis in conjunction with glass chemistry are first presented. Then, a general approach to glass chemistry design is outlined. Young’s Modulus of Glass and Glass Fibers Elastic modulus or Young’s modulus of ionic- or covalent-bonded inorganic crystalline solids is theoretically related to the electrostatic energy of attraction Uc (equal to z1z2e 2 /ro) of two opposite charged ions (z1, z2) with a spacing of ro. To account for many body interactions between ions within the solid, Madelung energy (Um = aUc) is used instead. The force applied to the two ions is dUm/dr, and the related stress within a unit volume of the crystal r 3 o can be expressed as r ¼ 1=r 3 o ðdUm=drÞ ð1Þ The change of r with respect to the distance between two ions under the applied force, dr/dr, leads to dr ¼ dr=r 2 o ðd2 Um=dr 2 Þ ¼ de=roðd2 Um=dr 2 Þ ð2Þ The elastic modulus (E) is therefore inversely proportional to the fourth power of atomic spacing between two ions or two times of the Mandelung energy, aU, per cubic volume of the system as E ¼ dr=de ¼ 1=roðd2 Um=dr 2 Þ¼2aðz1z2e2 =r 4 o Þ ¼ 2aUc=r 3 o ð3Þ Although the model has been confirmed from testing many crystalline materials, the a-value is not RO MgO, RO CaO, RO Al2O3 Al2O3 Al2O3 SiO2 SiO2 SiO2 E-Glass System S-Glass System R-Glass System Liquidus Temp (Wallastonite) 1070-1220oC Liquidus Temp (Diopside, Anorthite, Quartz) 1220 - 1410oC Liquidus Temp (Cordierite) ~ 1470oC Fig. 3. Typical composition projections, expressed as SiO2, Al2O3, and (CaO+MgO), of E-Glasses with and without boron (within a large blue circle), R-Glass (magenta), S-Glass (red), low-SiO2 calcium aluminate glasses (purple), and high-SiO2 alkaline earth (CaO or MgO or both in gray triangle) aluminosilicate glasses on CaO-Al2O3-SiO2, MgO-Al2O3-SiO2, and CaO-10% MgO-Al2O3-SiO2 phase diagrams.20,21 www.ceramics.org/IJAGS High-Performance Glass Fiber Development 69
International Journal of Applied Glass Science--Li, Richards, and Watson Vol.5,No.1,2014 /图图 8a well established for glass because of their disordered structures, which possess a wide range of ion-pair bond distances. Estimating the a-values for oxide glasses can be found in Bridge's model 8R8三8器 For oxide glasses, Sun's binding energy approach has been adopted to calculate Youngs modulus by Makishima and Mackenzie. The binding energy represented by the product of dissociation energy per unit volume (Gi) of glass constituent (X) and glas packing density(V. The Youngs modulus model of E=2V∑(GX) The model is subsequently modified by Zou and Toratani, proposing that each glass constituent(i)con- tributes to modulus as E;=2V Gi where Vi is the pad factor of the i-constituent in an oxide form(M.oj which can be estimated from 4T/3N,(p/M(x RM+yRo ) where p is the specific gravity of glass, M the molecular weight of glass, Ny the Avogadro's number, and RM and Ro the ionic radii of cation and oxygen, respectively. Specific modulus, S, of the i-th glass constituent is defined by E/p For a given glass composition,assum- 飞E ents, the specific modulus(S)of the glass is defined by Y∑(S: Xi where y[=∑(p;X/ Pm] accounts for uncer tainty between the calculated glass density(2Ip; Xi) uncertainty in the calculated density comes from an assumption of fixed coordination numbers for the cations and oxygen of the interest. In this approach, ESEl& the glass elastic modulus is calculated according to2 E E=pS=∑(n2X)∑(S·x) The Makishima-Machenzie and Zou-Toratani 兰88R odels assume that the Youngs modulus of solids can 2才内内一 be approximated by a linear combination of each indi vidual constituent contribution 25,26 Similar approaches to glass Youngs modulus of mplex glass systems can be also found els A general representation of a linear composition model is illustrated in Fig. 4, which provides a simplified view of the listed oxide utions to the overall gl Youngs modulus. In practice, especially in complex 2到 multicomponent glass systems, significant deviations between the me g到 have been reported' and represent a gap between the
well established for glass because of their disordered structures, which possess a wide range of ion-pair bond distances. Estimating the a-values for oxide glasses can be found in Bridge’s model.23 For oxide glasses, Sun’s binding energy approach24 has been adopted to calculate Young’s modulus by Makishima and Mackenzie.25 The binding energy is represented by the product of dissociation energy per unit volume (Gi) of glass constituent (Xi) and glass packing density (Vt). The Young’s modulus model of glass is expressed as E ¼ 2Vt XðGi XiÞ ð4Þ The model is subsequently modified by Zou and Toratani,26 proposing that each glass constituent (i) contributes to modulus as Ei = 2ViGi where Vi is the packing factor of the i-constituent in an oxide form (MxOy), which can be estimated from 4p/3Nv(q/M)(x R3 M þ yR3 O ) where q is the specific gravity of glass, M the molecular weight of glass, Nv the Avogadro’s number, and RM and RO the ionic radii of cation and oxygen, respectively. Specific modulus, Si, of the i-th glass constituent is defined by Ei/qi. For a given glass composition, assuming the property is a linear additive of all glass constituents, the specific modulus (S) of the glass is defined by c Σ(Si·Xi) where c [=Σ(qi·Xi)/qm] accounts for uncertainty between the calculated glass density (Σ[qi·Xi]) and the measured value (qm). The main source of uncertainty in the calculated density comes from an assumption of fixed coordination numbers for the cations and oxygen of the interest. In this approach, the glass elastic modulus is calculated according to26 E ¼ qmS ¼ Xðqi XiÞ XðSi XiÞ ð5Þ The Makishima–Machenzie and Zou–Toratani models assume that the Young’s modulus of solids can be approximated by a linear combination of each individual constituent contribution.25,26 Similar approaches to glass Young’s modulus of complex glass systems can be also found elsewhere.27,28 A general representation of a linear composition model is illustrated in Fig. 4, which provides a simplified view of the listed oxide contributions to the overall glass Young’s modulus. In practice, especially in complex multicomponent glass systems, significant deviations between the measured and the model-derived values have been reported29 and represent a gap between theory and practice. Table I. Commercial Glass Fibers Used for Structural Composite Applications1,12–17,22 Fiber Glass SiO2 (wt%) Al O2 3 (wt%) MgO (wt%) CaO (wt%) B O2 3 (wt%) R O2 (wt%) q (g/cm3 ) rf (GPa) E† (GPa) TL (°C) TF (°C) E 52 –62 12 –16 0 –5 16 –25 0 –10 0 –2 2.60 –2.65 2.8 –3.7* 72 –82 1080 –1220 1180 –1280 S2 (AGY) 64 –66 24 –25 9.5 –10 0 –0.1 0 0 –0.3 2.46 –2.49 4.6 –5.0* 87 –91‡ 1470 1571 T (Nittobo) 64 –66 24 –26 9 –11 <0.01 0 <1 2.49 4.7 84 1465 1500 HS (Sinoma S&T) 55 –62 23 –26 11 –16 – 0–4 <1 2.53 –2.54 4.6 –4.8 86 –88 1345 –1400 1440 –1450 R (OC) 58 –60 23 –26 5 –6 9 –11 0 – 2.55 4.5* 87 1410 1330 HiPer-texTM (OC) 50 –65 12 –20 6 –12 13 –16 0 –3 0 –2 2.55 4.1 –4.6 85 1280 1351 OCVTM-H (OC) 60 15.7 8.4 13.7 0 1.3 2.61 4.1* 87 1198 1268 INNOFIBER XM (PPG) 60 –61 15 –17 6 –10 13 –16 0 <1.0 2.56 –2.58 3.7 –3.8 88 –89 1207 –1224 1275 –1290 TM (CPIC) 56 –64 13 –20 7 –12 8 –13 – 0–1 2.62 4.5 84 –86 1170 –1225 1265 –1300 ViProTM (Jushi) 57 –65 14 –20 7 –12 8 –13 0 –2 0.1 –2 2.63 4.0 –4.3 86 1203 –1228 1275 –1300 *Pristine fiber strength values determined by testing fibers in liquid nitrogen without any moisture effect, which yields higher strength. †Fiber sonic modulus method, ‡heat-treated fiber exhibiting higher modulus. 70 International Journal of Applied Glass Science—Li, Richards, and Watson Vol. 5, No. 1, 2014
amics.org/lAGS nce Glas fiber Dew Nayo n8bn号8z2 115 105 ◇A Thermal effe Na,O 3o025 AXi of given oxide in glass(mol%) Fig. 4. First-order local, linear model illustrating individual There are major factors contributing to the dis Calculated Young,s Modulus(GPa) crepancies. First, local structure or surrounding oxygen Fig. 5. Comparison between linear mixture model derived environments of glass network formers(SiO2, B2O Young's modulus and the measured Young 's modulus of glasses in etc. )and conditional network formers (Al203) vary fiber form depending on concentrations of alkalis(Li2O, Na2O, K2O, etc. ) alkaline earth (MgO, CaO, SrO, etc.),and The theory implies that a material with their relative proportions. -> The linear composition packaging or smaller ro can have higher intrinsic dels cannot account for the structural variations ength. However, in real-world applications especially for network formers, including SiO2, B2O3 and Al2O3 been observed that glass or glass fibers have measured failure strengths much lower than the theoretical expec Secondly, glass density or molar volume is affected tation by fictive temperature or thermal history of the samples in terms of glass structure relaxation 6-In turn,an One of the most detrimental factors on glass or glass fiber strength is surface contact-induced damage annealed glass has lower fictive temperature, higher or surface Raws. Another source of faws, often encoun- density, and higher Youngs modulus as tered in applications, is the attack of corrosive medium the quenched form that originates during the high-tem- on glass or glass fber surfaces, varying from acid to perature glass fiberization process. Figure 5 compares basic solutions or vapors, as well as water in the form the measured fiber glass modulus as obtained by a sonic of liquid or vapor. Surface flaws serve as a stress method(discussed later)with the calculated modulus. concentrator when the material is under an applied ten- In general, a parallel downshift correlation line can be sile load; the weakest spot(the location where the most drawn from the ideal line of 1:1 correlation, which severe surface flaw has its path perpendicular to the illustrates a primary thermal effect on glass module he thermally induced change of glass Young's modulus app lied tensile load) causes glass or glass fiber to fail at varies between 10% and up to 20%.37.3 a tensile stress level well below the theoretical expecta- tion ergy-balance criterion acture strength or apparent strength(Oapp) of a solid is Fracture of Glass and Glass Fibers defined by"> Theoretical fracture strength of solids, according to app=(2E7/rC)' /2 (plane tensile stress) (7a) Orowan,is proportional to Youngs modulus and sur face energy (yo) of the material as Gapp=[2Ey/(1-v2)C]4(plane tensile strain =(E/)
There are major factors contributing to the discrepancies. First, local structure or surrounding oxygen environments of glass network formers (SiO2, B2O3, etc.) and conditional network formers (Al2O3) vary depending on concentrations of alkalis (Li2O, Na2O, K2O, etc.), alkaline earth (MgO, CaO, SrO, etc.), and their relative proportions.30–35 The linear composition models cannot account for the structural variations, especially for network formers, including SiO2, B2O3, and Al2O3. Secondly, glass density or molar volume is affected by fictive temperature or thermal history of the samples in terms of glass structure relaxation.36–38 In turn, an annealed glass has lower fictive temperature, higher density, and higher Young’s modulus as compared to the quenched form that originates during the high-temperature glass fiberization process. Figure 5 compares the measured fiber glass modulus as obtained by a sonic method (discussed later) with the calculated modulus. In general, a parallel downshift correlation line can be drawn from the ideal line of 1:1 correlation, which illustrates a primary thermal effect on glass modulus. The thermally induced change of glass Young’s modulus varies between 10% and up to 20%.37,38 Fracture of Glass and Glass Fibers Theoretical fracture strength of solids, according to Orowan,39 is proportional to Young’s modulus and surface energy (co) of the material as rth ¼ ðEco=roÞ 1=2 ð6Þ The theory implies that a material with denser packaging or smaller ro can have higher intrinsic strength. However, in real-world applications, it has been observed that glass or glass fibers have measured failure strengths much lower than the theoretical expectation. One of the most detrimental factors on glass or glass fiber strength is surface contact-induced damage, or surface flaws. Another source of flaws, often encountered in applications, is the attack of corrosive medium on glass or glass fiber surfaces, varying from acid to basic solutions or vapors, as well as water in the form of liquid or vapor.40–44 Surface flaws serve as a stress concentrator when the material is under an applied tensile load; the weakest spot (the location where the most severe surface flaw has its path perpendicular to the applied tensile load) causes glass or glass fiber to fail at a tensile stress level well below the theoretical expectation. By the Griffith energy-balance criterion, fracture strength or apparent strength (rapp) of a solid is defined by45,46 rapp ¼ ð2Eco=pCÞ 1=2 ðplane tensile stressÞ ð7aÞ rapp ¼ ½2Eco=pð1 m2 ÞC 1=2 ðplane tensile strainÞ ð7bÞ ΔXi of given oxide in glass (mol%) -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 Young's Modulus of Glass (GPa) 88 90 92 94 96 98 ZrO2 ZrO2 Al2O3 TiO2 MgO Na2O K2O Li2O B2O3 CaO SiO2 B2O3 Li2O CaO SiO2 Al2O3 TiO2 MgO Na2O K2O Fig. 4. First-order local, linear model, illustrating individual oxide effects on glass Young’s modulus.27 Calculated Young's Modulus (GPa) 80 85 90 95 100 105 110 115 120 125 Measured Young's Modulus (GPa) 80 85 90 95 100 105 110 115 120 125 S-glass Thermal effect Fiber Annealed bulk glass Fig. 5. Comparison between linear mixture model derived Young’s modulus and the measured Young’s modulus of glasses in fiber form. www.ceramics.org/IJAGS High-Performance Glass Fiber Development 71
International Journal of Applied Glass Science--Li, Richards, and Watson Vol.5,No.1,2014 With the presence of surface Aaws, Inglis demon- sphere after the samples were aged in hydrofluoric acid Stips can significantly magnify the stress applied on the is, Stip C, the fiber inert failure stress or str 4b strated"that tip geometry of the flaw or tip radius, vapor. For fibers with very sharp surface Aaws, tha material, which further reduces the measured strength approximately 30% of its theoretically predicted Om, of the material according to strength (17 GPa) m=(E70/4R)/(ce/C) The effect demonstrates that the measured strength (8a) of glass or glass fiber reflects a combination of intrinsic or 0m=0t/2(5in /C)/2 and extrinsic properties. For a given glass, the measured (86) tensile strength is highly sensitive to sample preparation quation 8b implies that the maximum strength of processes and processing environment, as well as sample a solid will be approximately 50% of its theoretical test environment. "Pristine" fiber strength often comes strength, and the same size of a critical surface Aaw from testing fibers made under controlled, low humid- with sharper crack tip(or lower radius at the crack tip) ity environments and without any surface contact and will further reduce the material strength. 0 tested under the same low humidity environment Strength of glass is not an intrinsic property of the within 2 h of the sample being made. "Inert pristine glass, which represents a combined effect of glass com fiber strength means that the pristine fibers were tested position, thermal history, testing temperature, and en in liquid nitrogen to minimize any atmosphere interac ronmental impact when combined with strain rate. 48-51 tion with glass or glass fiber surface under an applied In practice, usually one cannot prevent the samples force 52 As expected, strength of the pristine fibers mea- from coming in contact with dust in the air; therefore, sured in liquid nitrogen is higher than those measured surface flaws can be generated on the glass surface prior in ambient conditions, and the strength is closer to to testing. It is expected that the surface defects(O glass intrinsic property. generated are substantially larger than glass structural Solid fracture, including glass and glass fibers, defects from thermal or local density Fluctuation by under an applied tensile load at a point of its weakest more than an order of magnitude. Therefore, spot or the weakest link employs the Weibull statistical entiate glasses in terms of chemistry based on method3 to characterize the fracture behavior and to is important to compare glass samples that are made compare performance under different environments. and tested under the same environment According to the Weibull method, an accumulative Figure 6 illustrates the effects of surface Aaw probability of fracture(P) of a solid under an appl imension and geometry on silica fiber inert failure tensile stress level or follows stresses that were tested under liquid nitrogen atmo- Pr=1-expl-or/oo]. 20 or In[1/(1-P)= Blnof -Bingo SiO glass theoretical strengh where B and o. are the statistical linear regression fit- 15 ting parameters, which are often called Weibull modr lus (or shape parameter) and characteristic stress, respectively. ⊙ Figure 7 illustrates a Weibull plot of pristine tensile strength distributions of S-Glass, R-Glass and E-Glass fibers. Under the same sample preparation and ⊙ test conditions, the average fiber tensile strength ranks in an order S-Glass (5255 MPa)> R-Glass(3750 MPa)>E-Glass(3000 MPa). The Weibull modulus (B-value) of the S-Glass is substantially higher than both R-glass and e-glass Literature studies have Fig. 6. SiO, glass fiber inert failure stress aged in HF vapor, shown that fiber strength of a given glass composition illustrating efect of surface faw. The solid line was derived decreases as the sample gage length increases, which fol- from linear regression lows that as the materials sampling volume increases
With the presence of surface flaws, Inglis demonstrated47 that tip geometry of the flaw or tip radius, ftip, can significantly magnify the stress applied on the material, which further reduces the measured strength rm, of the material according to rm ¼ ðEco=4RoÞ 1=2 ðftip=CÞ 1=2 ð8aÞ or rm ¼ rth=2ðftip=CÞ 1=2 ð8bÞ Equation 8b implies that the maximum strength of a solid will be approximately 50% of its theoretical strength, and the same size of a critical surface flaw with sharper crack tip (or lower radius at the crack tip) will further reduce the material strength.40 Strength of glass is not an intrinsic property of the glass, which represents a combined effect of glass composition, thermal history, testing temperature, and environmental impact when combined with strain rate.48–51 In practice, usually one cannot prevent the samples from coming in contact with dust in the air; therefore, surface flaws can be generated on the glass surface prior to testing. It is expected that the surface defects (C) generated are substantially larger than glass structural defects from thermal or local density fluctuation by more than an order of magnitude. Therefore, to differentiate glasses in terms of chemistry based on strength, it is important to compare glass samples that are made and tested under the same environment. Figure 6 illustrates the effects of surface flaw dimension and geometry on silica fiber inert failure stresses that were tested under liquid nitrogen atmosphere after the samples were aged in hydrofluoric acid vapor.52 For fibers with very sharp surface flaws, that is, ftip C, the fiber inert failure stress or strength is approximately 30% of its theoretically predicted strength (≥17 GPa). The effect demonstrates that the measured strength of glass or glass fiber reflects a combination of intrinsic and extrinsic properties. For a given glass, the measured tensile strength is highly sensitive to sample preparation processes and processing environment, as well as sample test environment. “Pristine” fiber strength often comes from testing fibers made under controlled, low humidity environments and without any surface contact and tested under the same low humidity environment within 2 h of the sample being made. “Inert pristine” fiber strength means that the pristine fibers were tested in liquid nitrogen to minimize any atmosphere interaction with glass or glass fiber surface under an applied force.52 As expected, strength of the pristine fibers measured in liquid nitrogen is higher than those measured in ambient conditions, and the strength is closer to the glass intrinsic property. Solid fracture, including glass and glass fibers, under an applied tensile load at a point of its weakest spot or the weakest link employs the Weibull statistical method53 to characterize the fracture behavior and to compare performance under different environments. According to the Weibull method, an accumulative probability of fracture (Pf) of a solid under an applied tensile stress level rf follows Pf ¼ 1 exp ½rf =ro b ð9aÞ or ln½1=ð1 PfÞ ¼ blnrf blnro ð9bÞ where b and ro are the statistical linear regression fitting parameters, which are often called Weibull modulus (or shape parameter) and characteristic stress, respectively. Figure 7 illustrates a Weibull plot of pristine tensile strength distributions of S-Glass, R-Glass and E-Glass fibers. Under the same sample preparation and test conditions, the average fiber tensile strength ranks in an order S-Glass (5255 MPa) ≫ R-Glass (3750 MPa) > E-Glass (3000 MPa). The Weibull modulus (b-value) of the S-Glass is substantially higher than both R-Glass and E-Glass. Literature studies have shown that fiber strength of a given glass composition decreases as the sample gage length increases, which follows that as the materials sampling volume increases, (ζtip/C)1/2 0 10 20 30 40 50 60 Fiber inert failure stress, σm (GPa) 0 5 10 15 20 ~ SiO2 glass theoretical strength Fig. 6. SiO2 glass fiber inert failure stress aged in HF vapor, illustrating effect of surface flaw.52 The solid line was derived from linear regression analysis. 72 International Journal of Applied Glass Science—Li, Richards, and Watson Vol. 5, No. 1, 2014
amics.org/lAGS High-Performance Glass Fiber De the chance of a surface faw with dimension greater model, MgO and Na2O appear to have little effect on than critical defect size increases.. In this case, to TF; Cao slightly increases TF; and SiO2 and Al2O minimize the size effect on the fiber strength, the fiber contribute to TF increase the most. In this example gage length of all samples was kept the same(I inch), using commercial software package JMPM v 9.0(SAS and the diameter of the fibers was controlled at Institute, Inc, Cary, NC), a baseline glass composition 10±0.5m (by wt %): 58 SiO2-21 Al2O3-13 Mg0-7.5Cao 0.5Na2O) target glass modul Fiber Glass Composition Des 100 GPa and TF at 1300C. In this case, the designed glass was shown to have 97% probability (or 0.97 Development of a glass fiber product must be opti- desirability) to achieve the target property require- mized in terms of required performance, ease of glass ments (E, TE). Based on the composition design for ng ane nd fiber forming, and cost of both raw mate- the target properties, a smaller set of glasses can be and processing. For example, as depicted in Fig. 8, made accordingly to validate the designed glass prop R-Glasses properties(E, TE, AT)over the composition erties(E, TF)as a final step to close the composition of interest are highly nonlinear with some composition design loop. The process is an integrative process until changes. It is important to identify a composition the optimized glass is identified for trial prior to com- region where the designed glass can be robust and meet mercial production. the required performance and process targets. The with a proper modeling tool, a composition space following example (cf. Fig. 9) illustrates the use of can be searched to identify new compositions that have modeling tools in the final stage of glass composition not been otherwise possible using time-consuming Figure 9 illustrates the glass design process that xperiments. Thousands of glass compositions and their predicted properties(E, TF)can be generated on paper, considers both performance (Youngs modulus)and with a few experiments carried out later to validate the process(in terms of fiber-forming temperature)require- selected potential candidate glasses. In the composition igures 9a and b show the linear mixture design, it would be desirable to include liquidus(TD models of composition- property (E, TE), which are model or delta T(AT)model along with E and TF constructed here for a demonstration purpose. The out- model. Very often, experimental uncertainty of the puts from the models are shown in Fig. 9c. According measured Ti is greater than the measured E and TE to the modulus model, within the composition space of which can increase uncertainty of the output of the interest, CaO and Na2O have little effect on modulus; composition design model if Ti model were included SiO2 decreases modulus; and Al2O3 and MgO increase To increase accuracy of model design outcome, that is modulus significantly. According to the fiber-forming E and TF in this case, a Ti model can be used as a sep arate model for the prediction purposes. One can vali- date the promising designed ositions : reliable models based on E and TF and subsequently make the section of the most preferred compositions ◆ E-Glass (bca=18) that have desirable AT as determined by experiments. After the final selection of glass composition, extensive property tests should be conducted for addi tional validation at both pristine glass level and composite level. Figure 10 illustrates the mechanical fiber properties of INNOFIBER XM fiber glass at the pristine glass level. Based on the sonic modulus tests, Youngs modulus and pristine fiber strength of the ne fiber were confirmed to offer 10% and 20% improve In(or, MPa) ment,respectively, over a baseline E-Glass fiber. In terms of the pristine fiber strength distribution Fig. 7. Weibull plot of pristine singde fber tensile failure stress (Fig. 10d), it can be seen that the Weibull distribution of S-Glass R-Glass, and E-Glass(solid line- Weibull analysis). function (green color curve) fits the strength distribu
the chance of a surface flaw with dimension greater than critical defect size increases.48,49 In this case, to minimize the size effect on the fiber strength, the fiber gage length of all samples was kept the same (1 inch), and the diameter of the fibers was controlled at 10 0.5 lm. Fiber Glass Composition Design Development of a glass fiber product must be optimized in terms of required performance, ease of glass melting and fiber forming, and cost of both raw materials and processing. For example, as depicted in Fig. 8, R-Glasses properties (E, TF, DT) over the composition of interest are highly nonlinear with some composition changes.10,11 It is important to identify a composition region where the designed glass can be robust and meet the required performance and process targets. The following example (cf. Fig. 9) illustrates the use of modeling tools in the final stage of glass composition design process. Figure 9 illustrates the glass design process that considers both performance (Young’s modulus) and process (in terms of fiber-forming temperature) requirements.19 Figures 9a and b show the linear mixture models of composition — property (E, TF), which are constructed here for a demonstration purpose. The outputs from the models are shown in Fig. 9c. According to the modulus model, within the composition space of interest, CaO and Na2O have little effect on modulus; SiO2 decreases modulus; and Al2O3 and MgO increase modulus significantly. According to the fiber-forming model, MgO and Na2O appear to have little effect on TF; CaO slightly increases TF; and SiO2 and Al2O3 contribute to TF increase the most. In this example, using commercial software package JMPTM v 9.0 (SAS Institute, Inc., Cary, NC), a baseline glass composition (by wt.%): 58 SiO2–21 Al2O3–13 MgO–7.5CaO– 0.5Na2O) was designed to target glass modulus at 100 GPa and TF at 1300°C. In this case, the designed glass was shown to have 97% probability (or 0.97 desirability) to achieve the target property requirements (E, TF). Based on the composition design for the target properties, a smaller set of glasses can be made accordingly to validate the designed glass properties (E, TF) as a final step to close the composition design loop. The process is an integrative process until the optimized glass is identified for trial prior to commercial production. With a proper modeling tool, a composition space can be searched to identify new compositions that have not been otherwise possible using time-consuming experiments. Thousands of glass compositions and their predicted properties (E, TF) can be generated on paper, with a few experiments carried out later to validate the selected potential candidate glasses. In the composition design, it would be desirable to include liquidus (TL) model or delta T (DT) model along with E and TF model. Very often, experimental uncertainty of the measured TL is greater than the measured E and TF, which can increase uncertainty of the output of the composition design model if TL model were included. To increase accuracy of model design outcome, that is, E and TF in this case, a TL model can be used as a separate model for the prediction purposes. One can validate the promising designed compositions using more reliable models based on E and TF and subsequently make the section of the most preferred compositions that have desirable DT as determined by experiments. After the final selection of glass composition, extensive property tests should be conducted for additional validation at both pristine glass level and the composite level. Figure 10 illustrates the mechanical fiber properties of INNOFIBER XM fiber glass at the pristine glass level. Based on the sonic modulus tests, Young’s modulus and pristine fiber strength of the new fiber were confirmed to offer 10% and 20% improvement, respectively, over a baseline E-Glass fiber. In terms of the pristine fiber strength distribution (Fig. 10d), it can be seen that the Weibull distribution function (green color curve) fits the strength distribuln(σf, MPa) 7.5 8.0 8.5 9.0 lnln[1/(1-Pf)] -4 -3 -2 -1 0 1 2 Legend S-Glass (beta = 49) R-Glass (beta = 16) E-Glass (beta = 18) Fig. 7. Weibull plot of pristine single fiber tensile failure stress of S-Glass, R-Glass, and E-Glass (solid line — Weibull analysis). www.ceramics.org/IJAGS High-Performance Glass Fiber Development 73
International Journal of Applied Glass Science--Li, Richards, and Watson Vol.5,No.1,2014 1310 Cao Fig. 8. Composition- property space of high-modulus glasses, illustrating complex relationship of fiber Youngs modulus by the sonic method(E), fiber-forming temperature(TE), and fiber-forming window (47) over the simplified composition space(Mgo, Cao) where the changes of Sioz and Al203 were kept within a narrow range. 2 tion better than the normal distribution function(red outcome can be used as a good indicator that the new color curve). On the other hand, the normal distribu- fiber product offers good compatibility with different tion function can adequately represent the distributions resins of glass fiber elastic properties(Figs. 10a and b) Very often, the tests described above need to be At the composite level, it is important to evaluate carried out at the customer site. Positive confirmations the effects of fiber volume fraction on the composite from extensive tests from customer(s) conclude the ini properties, such as tensile fiber strength and tensile tial stage of the new product development cycle. In modulus,in different resin systems to determine the principle, the new fiber product is ready for commercial new fiber broad applications. Figure 1l compares uni- production pending business arrangements with the directional (UD)composite tensile moduli as a func- prospective customer(s) tion of fiber volume fraction between E-Glass and high-modulus INNOFIBER XM glass fiber. Once the extensive tests confirm that new high-modu- Glass Fiber Mechanical Property Characterizations lus fibers consistently provide higher UI tensile modulus over E-Glass by approximately 10% As discussed earlier, glass fiber tensile strength is ndent of resin type. For a given resin system, the influenced by many factors that can lead to significar ive modulus improvement is expected based on the variations in the final performance capability for a rule of mixtures. Summarizing all composite test given glass composition, including variation of the results, covering a broad range of resin systems, the fiber-d 48-51 rocess o minimize the effects of
tion better than the normal distribution function (red color curve). On the other hand, the normal distribution function can adequately represent the distributions of glass fiber elastic properties (Figs. 10a and b). At the composite level, it is important to evaluate the effects of fiber volume fraction on the composite properties, such as tensile fiber strength and tensile modulus, in different resin systems to determine the new fiber broad applications. Figure 11 compares unidirectional (UD) composite tensile moduli as a function of fiber volume fraction between E-Glass and high-modulus INNOFIBER XM glass fiber. Once again, the extensive tests confirm that new high-modulus fibers consistently provide higher UD composite tensile modulus over E-Glass by approximately 10% independent of resin type. For a given resin system, the relative modulus improvement is expected based on the rule of mixtures. Summarizing all composite test results, covering a broad range of resin systems, the outcome can be used as a good indicator that the new fiber product offers good compatibility with different resins. Very often, the tests described above need to be carried out at the customer site. Positive confirmations from extensive tests from customer(s) conclude the initial stage of the new product development cycle. In principle, the new fiber product is ready for commercial production pending business arrangements with the prospective customer(s). Glass Fiber Mechanical Property Characterizations As discussed earlier, glass fiber tensile strength is influenced by many factors that can lead to significant variations in the final performance capability for a given glass composition, including variation of the fiber-drawing process.48–51 To minimize the effects of CaO MgO Fiber sonic modulus, E (G Pa) 4 8 12 16 20 4 5 6 7 8 9 10 85 87 89 91 93 95 CaO MgO Forming temperature (o C) 4 8 12 16 20 4 5 6 7 8 9 10 1250 1270 1290 1310 1330 1350 CaO MgO Forming window (o C) 4 8 12 16 20 4 5 6 7 8 9 10 20 40 60 80 100 Fig. 8. Composition — property space of high-modulus glasses, illustrating complex relationship of fiber Young’s modulus by the sonic method (E), fiber-forming temperature (TF), and fiber-forming window (DT) over the simplified composition space (MgO, CaO) where the changes of SiO2 and Al2O3 were kept within a narrow range.17,22 74 International Journal of Applied Glass Science—Li, Richards, and Watson Vol. 5, No. 1, 2014