Proceedings of the 42nd International Conference on Advanced Ceramics and Composites, Ceramic Engineering and Science Proceedings, Issue 2

MECHANICAL BEHAVIOR AND PERFORMANCE OF CERAMICS AND COMPOSITES

Fracture Toughness of Modern and Ancient Glasses and Glass Ceramics as Measured by the SEPB Method

G. D. Quinn,¹ J. J. Swab,² and P. Patel²

¹Guest Researcher, National Institute of Standards and Technology, Materials Measurement Science Division, Stop 8520, Gaithersburg, MD 20899

²U.S. Army Research Laboratory, Weapons and Materials Directorate, Aberdeen Proving Ground, MD 21005

ABSTRACT

The fracture toughness of eight materials was measured by single-edged precracked beam (SEPB) method. These include two modern soda lime silicas that have low iron content, two structural lithium alumino silicate glass ceramics, two dental lithium disilicate glass ceramics, one dental ceramic-resin matrix composite, and a historical soda lime silica glass made in an ancient Roman glass factory that operated in Palestine until 383 AD. This glass was exposed to the environment for almost sixteen centuries. A new definition for fracture toughness of glasses is proposed.

INTRODUCTION

Fracture toughness, KIc, is a fundamental property that characterizes the brittleness of a material. Over the last five decades, there has been confusion as to whether glasses actually have a specific fracture toughness, KIc. Some of the confusion stemmed from environmentally-assisted slow crack growth (SCG) which readily occurs in glasses at room temperature in normal ambient atmosphere, and uncertainty as to whether KIc was simply a point on a slow crack growth, K-V curve. Some of the confusion stemmed from test method interpretation issues and fundamental shortcomings of some test methods. For example, a lot of erroneous fracture toughness data have been generated with the Vickers indentation fracture method as discussed by Quinn and Bradt¹ and Rouxel and Yoshida.², The matter of what fracture toughness of glasses was discussed in detail in our earlier papers and presentations.³,⁴,⁵,⁶ In order to clarify this property, we propose a definition of KIc for glasses:a

Fracture toughness, KIc, the critical condition for the onset of rapid crack extension in an inert environment.

The first ten words were written by George Irwin in the 1950s.⁷, ⁸ In our definition, we add the four key qualifier words: in an inert environment. These were written by Wiederhorn et al. in 1974⁹ when he described fracture toughness of glasses. Irwin's words including critical condition and onset of rapid crack extension are at the core of what he felt was fracture toughness, and appeared in many of his publications over the years. He often included unstable in place of the word rapid. He mostly dealt with brittle metals and environmentally-assisted slow crack growth at room temperature was not an issue. Wiederhorn, a pioneer in the matter of SCG in glasses, recognized that measurements of fracture toughness must take SCG into account and influences of the environment must be eliminated.⁹

This work is part of a program to evaluate materials and test methods to evaluate the fracture toughness of brittle materials. We are using test methods such as single-edged precracked beam (SEPB) and surface crack in flexure (SCF) methods since these have been rigorously refined and standardized for ceramics.¹⁰, ¹¹, ¹², ¹³, ¹⁴ Knowledge gained from the ceramics work can be carried over to glasses and glass ceramics for which no standards exist. Our experiences have shown that there are important nuances in testing glasses and glass ceramics that must be kept in mind. We have made presentations about this at the ICACC conference every year since 2015,³-⁵ and published a major journal paper in 2017.⁶ Salem¹⁵, ¹⁶, ¹⁷ has also applied SEPB and chevron notch in bending (CNB) methods devised for advanced ceramics to glasses. Rouxel and colleagues², ¹⁸, ¹⁹ have used SEPB and CNB effectively with glasses as well. Rouxel's work, which is supporting a new theoretical model for fracture toughness of glasses based upon fundamental principles, relies upon having rigorous, high quality data to support the theoretical model.² We agree with this approach. Progress in science is facilitated by access to accurate and precise data. Our goal in the present work is to obtain accurate and precise data for KIc of glass and glass ceramics, by practical and easy-to-do methods.

MATERIALS

Table 1 lists the three glasses, two lithium alumino silicate (LAS) glass ceramics, two dental lithium disilicate (L2S) glass ceramics, and one dental ceramic-resin composite that was tested for comparison.b

Table 1. Elastic properties are from Ref. 20 or the manufacturer. Values have the same number of significant figures as reported. Uncertainties, if available, are ± one standard deviation.

* From Ref. 20 ** Manufacturer data NA….Not available

The first two are soda lime silicas with reduced iron content to improve optical transmission, an important factor for thick armor glass applications. Results for these two glasses may be compared to common soda lime silicas and to a similar low-iron soda lime silica (Starphire®) that we reported on last year.⁵, ⁶

The Zerodur® and Robax® LAS glass ceramics were produced by Schott and were aimed at structural applications. They have very low thermal expansions. They are transparent since the crystallized component is beta-quartz silica crystallites that are 70 nm or smaller. The material is as much as 70% crystalline by volume.

The two lithium dislicate L2S glass ceramics were optimized for strength, fracture toughness and use as dental restorations. They had larger crystallites and were translucent with aesthetics suitable for restorations. They have identical composition but differ in how they are made. The e.max® CAD form was cast into partially-crystallized cylindrical blanks suitable for computer aided design and manufacture (CAD/CAM). After milling to final shape, they are further crystallized forming lithium disilicate (L2S). The 1 to 2 µm long crystals comprise 70% of the volume, are slightly elongated, and random in orientation. The e.max® Press material was formed into crystallized blanks. These could then be pressed at high temperature into restoration molds. The material has a preferred orientation of 3 to 6 µm long needle-like crystallites. Specimens for both forms were obtained from the manufacturer in 2013 in the form of bend bars 3 mm × 4 mm × 25 mm in size. The difference in processing routes of the two forms leads to different microstructure sizes and preferred orientations.

The hybrid ceramic-resin composite material was made in 2013. It was first fabricated into a porous skeletal network of feldspathic porcelain enriched with alumina. This was then infiltrated with a monomer resin mix to eliminate porosity. The monomer was then polymerized. Thus, the microstructure has two interconnected phases. The material was cut out of short dental CAD/CAM blocks and machined into 2.8 mm × 4 mm × 17 mm long bars.

The soda lime silica from the Jalame excavations was available in the form of cullet and furnace waste and debris pieces (Figure 1) that were excavated in the mid-1960's from an archeological site in modern Jalame, Israel.²¹ This general area, in the vicinity of Mt. Caramel was famous in ancient times as a glass manufacturing site. Pliny the Elder in approximately 77 AD described a nearly location, the mouth of the Belus River (now the Na'aman River), as the site where glass was invented by the Phoenician.²² We now know that glass predated this era and location by many centuries, but the area near Mt. Carmel and the mouth of the Belus river was a major site of glass manufacture for many centuries in antiquity. The Jalame site, on the basis of coin and terracotta lamp fragments, was a Roman era glass factory that was active between 351 AD and 383 AD. Originally it was thought that the Jalame factory may have been a secondary manufacturing site (vessel manufacture using glass ingots that were made at another nearby primary factory), but new evidence suggests that that the Jalame site was both a primary and secondary glass manufacturing site.

Figure 1. Pieces from the Jalame excavation. Three green cullet pieces are shown on the left. The middle shows SEPB test specimens that were cut from other green cullet pieces. The four blocks on the right are edge chip testing blocks for future evaluation. The scale shows cm.

The Jalame glass used sands near the mouth of the Belus river that included quartz grains and particles rich in lime minerals from skeletal remains of marine fauna. It was only necessary to add soda in the form of natron to be able to make glass. The composition and general characteristics of the Jalame glasses have been thoroughly documented by Brill²³. who collected hundreds of fragments of cullet, vessels, furnace waste, and tank debris. A comprehensive chemical analysis showed the composition of the cullet and vessel fragments were statistically indistinguishable in their major and minor oxides. The typical composition was what one would expect for glass of that period. It had a soda lime silica composition (14.1 % Na2 O, 8.7 % CaO, 70.5 % SiO2) with the usual impurities at the usual concentrations for ancient glasses. This mix is similar to modern soda lime silicas. Low potassium oxide (0.76 %) and magnesium oxide (0.60%) levels were noted and it was evident that natronc was the primary alkali additive for glass making. Several green Jalame glass cullet fragments were large enough that SEPB test specimens could be prepared. We were not able to fabricate test specimens from the furnace debris, since although chunks were large (e.g., up to ~20 cm in breadth), they were heavily faulted with internal cracking, devitrification products, reactions with furnace insulation, and other flaws. Our interest in this glass was two-fold: was the fracture resistance of the ancient glass like that of modern glass, or had exposure to the elements over almost sixteen centuries altered the fracture toughness in some manner? Indeed, R. Brill noted that some pieces seemed to exhibit unusual fracture resistance when he excavated and collected the pieces in 1965.²⁴

EXPERIMENTAL PROCEEDURE

Our SEPB procedure was described in detail previously ³-⁶ and only a few specific details are included here. Procedures described in the documentary ceramic standards were used with modifications as needed for the glasses. A transverse scratch was placed in the middle of a flexure specimen which was then inserted into a bridge precracker. A crack was popped-in part way into the cross section of the specimen. It was then broken in four-point flexural loading. After fracture, the initial crack size was measured on the fracture surface. Fracture toughness was computed from the fracture force, the critical crack size, and a stress intensity shape factor. We learned that there are many nuances in each step of testing glass by this method: in the bridge precracking step, in the final fracture, and in the examination and measurement of the precrack size. A six mm wide bridge gap was used for precracking at a crosshead rate of 0.5 mm/min. Pop in forces were typically 1,500N to 3,000N. A stethoscope was used to listen for crack pop in. A scratch was made by hand using a tungsten carbide scribe with a force of about 5 to 10 N for the glasses, or about 10 to 15 N for the glass ceramics. There forces were measured by duplicating this procedure on a digital mass balance. Full-length bend bars, 3 mm × 4 mm × 48–50 mm were used in most cases, and the halves of fractured specimens were reused if necessary for additional experiments. Greater care was taken to align the scratch carefully and to insert the piece in the bridge precracker to obtain straight and even cracks. As a result, our reject rate was reduced as compared to our work in past years. The full-length specimens were tested on 20 mm × 40 mm semi-articulating fixtures at 0.5 mm/min which caused fracture in 2-4 seconds. Shorter length specimens were usually tested on 10 mm × 20 mmm fixtures at 0.25 mm/min, which also caused fractures in 2-4 seconds. In our previous work, a range of loading rates were used and tests were done both in in air and in dry nitrogen environment to detect rate and possible slow crack growth (SCG) effects. There was no rate effect for any of the glasses. Pronounced rate effects were often observed for the air testing. At the fastest loading rates, the air-test outcomes sometimes approached the inert dry nitrogen results, but sometimes not. Thus, environmentally-assisted slow crack growth (SCG) was a factor that interfered with our goal of measuring the fracture toughness, KIc, in the absence of SCG. Therefore, all testing in the present work was done in dry nitrogen (99.999% pure) gas after a 2 minute purge.

After fracture, one test piece half was examined in a stereo optical microscope and very careful illumination to obtain specular reflections to see the popped in precrack, and any arrest lines or evidence of stable crack extension. The largest precrack was used with the fracture force to compute fracture toughness. This procedure is described in detail with many illustrations elsewhere.⁵,⁶

The SEPB fracture toughness was computed from the break force, the critical crack size, and the stress intensity shape factor. The original Srawley and Gross formula²⁵ for a precracked beam in uniform bending was used, as specified in the ceramic standards.

In this latest round of testing, we only had two outlier outcomes, both on the UltraWhite® low-iron soda lime silica. Outliers are atypically high KIc results that are probably caused by crack healing after crack pop in as discussed in refs. 5 and 6. These were rejected. Specimens whose precracks were misaligned, curved too much, or uneven from side-to-side on the fracture surface were rejected.

There was no problem obtaining full length bend bars for the glasses and the two LAS structural glass ceramics, but only smaller pieces (~3 mm × 4 mm × 25 mm) were available for the two L2S dental glass ceramics. These were tested on 10 mm × 20 mm fixtures at 0.25 mm/min. For the ceramic-resin composite only 3 mm × 4 mm × 17 mm bars were available and these were tested on 8 mm × 16 mm fixtures at 0.25 mm/min.

There was sufficient material to do some corroborative surface crack in flexure (SCF) experiments for the two dental L2S disilicates and the dental ceramic-polymer composite. The SCF test method utilizes bend bar specimens with a shallow semi-elliptical precrack made with a Knoop indenter. An indentation force 27.6 N was used. The indentation and its residual stresses was removed by hand grinding with 400 grit SiC dry abrasive papers abrasive papers so that a clean, semielliptical precrack was created. The short specimens were broken in 10 mm × 20 mm fixture for the two L2S glass ceramics and a 8 mm × 16 mm fixture for the ceramic polymer composite. The dry nitrogen inert environment was used for the two LS glass ceramics, but a normal air environment (25°C, 27% RH) with the ceramic-polymer composite. Fracture occurred in 8 to 9 seconds. The precrack sizes were measured on the fracture surfaces after fracture. Fracture toughness was computed from the stress in the beam at fracture, ω, the depth and width of the precrack, a and 2c, and a specific stress intensity shape factor, Y, for each crack:

(1) numbered Display Equation

The Newman-Raju²⁶ Y factors were used for the new results reported here. They are suitable for semielliptical precracks in materials with a Poisson's ratio near 0.3. The reader is alerted to new Y solutions by Strobl et al.²⁷ which are more suitable for glasses³,⁴ with lower Poisson's ratios (e.g., 0.18 to 0.22) and for precracks which are truncated semiellipses. This matter was discussed in more detail in our earlier papers.⁵,⁶) Refs. 5 and 6 have more on the experimental procedures, particularly on how to photograph and measure the SCF precracks.

RESULTS AND DISCUSSION

Table 2 shows the results for all eight materials. The standard deviations ranged from 0.01 to 0.05 for most part. Space limitations preclude our showing many precracks and fracture surfaces, but Figs. 2 and 3 show some examples. Precracks for the two low-iron soda lime silicas were very similar to glass precracks shown in our earlier papers.⁵,⁶

Table 2. Fracture toughnesses outcomes in dry N2 gas. The standard deviation is in parenthesis. The number of valid outcomes is listed third. The second line in some cells (in grey font) shows the air-tested values at the fastest loading rates.

* The second Zerodur entry includes 6 outcomes for which the precrack size was slightly too large.

† Tested in air.

Figure 2. SEPB precracks in the Jalame glass with the precracks marked. (a) is a transmitted-light image. The greenish tint is real. The small bubble is beneath the surface and had no influence on crack propagation. Perturbations on the bottom are from the starter scratch. (b) is another piece with specular-reflected illumination.

Figure 3. SEPB precracks in two glass ceramics. (a) shows the LAS form which was transparent since the crystallites were tiny. It appeared like a glass and specular reflected light highlighted the precrack quite well. (b) shows the dental L2S which had a fine granular fracture surface and conventional reflected light illumination was effective. Green dye from a felt tip pen helps delineate the precrack.

The two low-iron soda lime silicas had KIc's that are statistically the same as that for common soda lime silica. They also match the 0.75 MPa√(± 0.03) obtained last year for Starphire® glass, another low-iron soda lime silica.⁵,⁶ It is concluded that the removal of trace amounts of iron to improve clarity has no effect on fracture toughness of soda lime silica.

The Jalame glass has a slightly lower fracture toughness (0.716 MPa√m) than contemporary glasses. It was very easy to test and we had a 100% success rate. If a population value for contemporary soda lime silica of 0.760 MPa√m is assumed, the difference is -0.044 MPa√m, a −5.5% change. If a hypothesis is made that the Jalame glass has the same fracture toughness as common soda lime silica with a fracture toughness of 0.760 MPa√m, the one-sided t statistic is a rather large −10.1, which means the difference is statistically significant at well over a 99 % probability. This is in part a consequence of the extremely small standard deviation for the Jalame outcomes: 0.018 MPa√m (2.5 % coefficient of variation expressed as a percentage). So, although there is some obvious variability in the appearance of the Jalame test specimens (Fig. 1), KIc is remarkably consistent. Fig. 4 shows an "eyeball comparison of all the individual outcomes for the glasses. They all overlap, but the Jalame glass is indeed slightly less. As noted in the introduction, Jalame glass has the same composition as contemporary soda lime silica. The Jalame bulk glass density, 2.460 ± 0.001 g/cm³, was measured on the four precision-machined (nominally 12 mm × 16 mm × 5 mm) blocks shown in Fig. 1. The mass was measured with a precision balance to a resolution of 0.00001 grams and the physical dimensions to within 0.002 m with a micrometer. The density was slightly less than the typical value of 2.50 g/cm³ for soda lime silica, although the latter certainly has some variability and values of 2.48 to 2.52 g/cm³ are commonly reported. The original cullet and the fabricated test specimens were examined in a polariscope prior to testing, but only slight traces of possible residual stresses were detected. It is likely that any residual stresses would have been relieved when the test pieces were cut and machined. The slightly lower KIc values (4.0 % to 7.7% compared to the other soda lime silicas in Table 2) might be due to exposure to the environment for over 16 centuries. It is also plausible that the KIc of the Jalame glass when originally made may have been slightly less than that for contemporary materials. Edge chipping resistance measurements²⁸, ²⁹ are planned with the blocks shown in Fig. 1 to obtain alternative fracture resistance evaluations for the Jalame glass.

Figure 4. Individual KIc outcomes for five soda lime silicas. All data is for inert dry N2 conditions. Average values are listed. The uncertainty bars are ± one standard deviation. Results for the conventional float soda lime silica and the low iron Starphire® are from Refs. 5 and 6.

The glass ceramic outcomes are interesting. The two transparent lithium alumino silicates (LAS) with tiny beta quartz crystals (70 nm) were slightly different, but there was not much toughening. Both were only 0.1 to 0.15 MPa√m tougher than conventional glasses. The fracture surfaces looked glassy. Both were designed to have extremely low thermal expansions, and enhanced toughness was not the primary consideration. On the other hand, the two dental lithium disilicates (L2S), had larger crystallites (3 to 6 µm) and were designed to have enhanced toughness. They were about 2.5 times tougher than conventional glass. There was little difference in KIc between the CAD and pressed forms. Space limitations preclude a detailed comparison of these results to other outcomes in the literature, but glass ceramic KIc values are often in the low to mid 2's (MPa√m).

The hybrid ceramic-resin composite material had a disappointingly low KIc. The fracture surfaces looked like fracture surfaces in fine-grain ceramics. The SEPB precracks were easy to detect with reflected light and to measure on the fracture surfaces since there usually was a slight change in plane from the initial precrack to the final fracture plane. There was some difficulty in discerning many of the small SCF precracks and multiple evaluations had to be made on both halves of each broken test piece.

This work has shown that SEPB flexure bar methods may be applied to glasses and glass ceramics. It reinforces our earlier work which reached the same conclusions.³-⁶ The earlier work and the new results showed that consistent results could be obtained with two refined test methods: SEPB and SCF. Each has its advantages and disadvantages. Space limitations preclude a detailed tabulation of results from other investigations., but our results are very harmonious with work by Salem¹⁵-¹⁷ and Rouxel et al.²,¹⁸,¹⁹- Some small differences remain between ours and Rouxel's, and this may be due to the fact his group has tested in laboratory ambient conditions at slow loading rates in order to emphasize fully stable crack extension. The ambient testing raises the possibility that slow crack growth causes lower apparent toughness, which we proved occurs in our earlier work.⁵,⁶ In those studies and in an earlier study with alumina, ³⁰ we learned that the SEPB method is less susceptible to slow crack growth effects during air testing, as compared to the SCF method. We also believe it is not necessary to grow cracks stabily in SEPB testing to obtain KIc. It is a requirement for CNB testing. Irwin was the first to define fracture toughness as the onset of rapid' or unstable crack extension, and we concur with this approach.

Another primary outcome from our work over the last few years is that tests on short length specimens match those from full length bars. This is reassuring and will expand the applicability of these methods to materials such as dental ceramics where it is impractical to make long specimens. It also is helpful to be able to retest the halves of broken full length test pieces. In other words, a normal length specimen can give three results.

We recommend that both transmitted and specular reflected illumination be used to examine the precracks on the fracture surfaces. There was no evidence of stable crack extension in the inert dry N2 tests, but it was detected in our earlier work with tests done in air.⁵,⁶ Multiple arrest lines occurred in some specimens, especially if the crack was somewhat hooked or was not very straight. The largest arrest crack line (and any stable extension from it) should be used as the critical crack size. Wallner lines should not be confused with arrest lines. The NIST Guide to Practice for Fractography³¹ may be consulted for guidance on this. SEPB crack sizes should be measured very carefully (1% or better) since KI is very sensitive to the crack length. Only 5 or 10 successful experiments are necessary, but more specimens should be prepared to replace rejects or outliers. Guidance in the ceramic standard test methods may be used for the acceptance or rejection of precracks.

It is our opinion that inert atmosphere testing is the best procedure