Source: Glass Structure and Engineering

Glass Structure Engineering 5, 445–487 (2020). https://doi.org/10.1007/s40940-020-00138-2

Telesilla Bristogianni Faidra Oikonomopoulou Rong Yu Fred A. Veer Rob Nijsse

Currently, due to pollutants hindering closed-loop recycling, tons of high-quality commercial glass are degraded for recycling or landfilled. However, this kind of glass is a potentially valuable resource for casting strong, beautiful and unique building components. To explore the potential of this idea, different types of non-recyclable silicate glass are kiln cast into beams of 30 × 30 × 240 mm at relatively low temperatures (820-1120°C). Defects in glass samples due to cullet contamination and high viscosity of the glass melt are recorded and correlated with casting parameters. Then, a four-point bending test was performed on the kiln-cast sample and the industrially manufactured reference beam, and the bending strength range of 9-72MPa was obtained.

The results are analyzed according to the chemical composition, pollution degree and subsequent casting parameters in determining the bending strength, Young's modulus and the main strength limiting defects. It highlights the chemical composition with good performance and is therefore a key defect that leads to a sharp drop in strength (up to 75%). However, the defects located in the glass body are tolerated by the glass network and have little effect on the bending strength and Young's modulus. The prerequisites for high-quality recycled cast glass building components are determined and an outline for future research is provided.

So far, structural engineers and architects have rarely explored the huge potential of glass casting technology in the construction industry, but it has gradually been discovered after the success of all cast glass load-bearing structures (such as the exterior wall of the Crystal House in Amsterdam) (Oikonomopoulou et al. 2018c). The 3D molding possibilities provided by casting can provide robust glass components with larger cross-sections and more forms and colors than other glass processing methods currently available. While recognizing the structural and aesthetic advantages of cast glass components, there are also issues related to their environmental impact and life cycle.

Using currently non-recyclable glass as a raw material for glass casting at lower temperatures is a promising idea. It can not only solve the urgent problem of glass waste, but also solve the urgency of reducing the carbon footprint of glass building components (Bristogianni et al. 2018 Years; Oikonomopoulou et al., 2018b). In order to clarify the term "currently non-recyclable glass", in addition to the successful recycling of soda lime glass food and beverage containers, the rest is usually high-quality commercial glass due to coatings and/or adhesives.

The lack of infrastructure for the collection, product disassembly, and cullet separation of these different types of glass stems from manufacturers' hesitation to accept this cullet, thereby restricting or preventing its recycling. Therefore, since this glass cannot flow back to the original product system (closed loop recycling), it will be recycled down for applications such as aggregates, ceramic-based products, foam insulation materials, abrasives (Silva et al., 2017), or in Landfill disposal. Due to the need to find alternative routes, markets and end users to recycle a large amount of high-quality waste glass, it is worth exploring to transfer these waste parts to the construction industry by casting structural glass parts.

The above developments reveal gaps in the literature regarding the mechanical properties of cast glass parts and the design strength recommendations for structural uses. This is related to the lack of established manufacturing procedures and quality control standards. Therefore, the strength of cast glass products varies greatly according to each manufacturer and the corresponding glass composition and casting process used. The use of scrap glass shards complicates the problem and creates a series of traditional and new defects (Bartuška 2008; Bristogianni et al. 2019), which may affect the strength of glass products.

This article discusses the bending strength of recycled cast glass-a characteristic related to engineering practice. The purpose is to deeply understand the influence of casting parameters on strength and to evaluate the rationality of using waste glass to produce safe structural components. Therefore, in this work, the ability of various commercial glass waste silicates to be kiln-casted into structural parts at relatively low temperatures (820-1120°C) was tested and evaluated. Defects that occur are recorded and correlated with the production stage in which they occurred.

After that, two series of four-point bending experiments were carried out on kiln-cast glass beams with a size of 30 × 30 × 240 mm. The results are analyzed according to the chemical composition, pollution degree and the role of the following casting parameters in determining the flexural strength and the origin of fracture. Tests on a limited number of industrially manufactured components can be used as reference points.

Table 1 Sample preparation, cullet classification and furnace casting settings-full size table

2.1 Classification of cullet and sample preparation

This work has studied a series of characteristic commercial glasses for the production of common glass products such as float glass, fiberglass, cookware and laboratory glassware, cast glass bricks, crystal ware and CRT TV screens. 1 The choice of glass is to combine the types of waste glass provided by various glass manufacturing and recycling companies to solve the recycling problem of ready-made waste glass resources and solve practical problems.

X-ray fluorescence (XRF) analysis is performed using a Panalytical Axios Max WDXRF spectrometer to determine the chemical composition of the selected glass. The provided broken glass is thoroughly cleaned with isopropanol, and foreign objects (metal, plastic, cork) are manually removed if possible. According to the following classification, the contaminants identified in a given cullet still exist in trace amounts after the cleaning process, which are listed in Table 1:

A generation. Coating (soft, hard, mirror, enamel, glass frit) ii. Composition changes of the same glass type (different manufacturer, color tone) iii. External pollutants in the sorting process: a. Organic matter (such as plastics, textiles), b. Non-glass inorganic substances (such as ceramics, stones, porcelain, glass ceramics), c. Metal, D. Different glass types (e.g. borosilicate, lead glass)

Then use a cullet kiln to cast the 30 × 30 × 240 mm glass beam required for the four-point bending test. This particular beam size was chosen because it provides a relatively thick casting material so that the effect of defects in the block can be evaluated while keeping the mass below 1 kg, thereby reducing annealing time. For each glass fragment, at least 3 samples are produced for statistical purposes. 2 The fragments are placed in a disposable silica/gypsum investment mold made of Crystalcast M248,3 in a structured or random manner. Then put the mold into the ROHDE ELS 200S or ELS 1000S electric kiln (Figure 1) and perform kiln casting, which means that only one furnace is used for the complete casting process (heating, forming, annealing and cooling).

The glass sample is formed at a viscosity between 106–103.5 dPa s and a maximum temperature range of 820 to 1120 °C, selected according to the chemical composition of each glass. The selected viscosity (η) range is higher than the 104-101.5 dPa s forming and melting range adopted by the glass industry, while taking into account the risk of unevenness of the final product. Choosing the method of forming glass at a lower temperature is on the one hand to reduce the required energy and corresponding CO2 emissions, on the other hand, to aggravate the occurrence of defects and assess whether their existence can be used in structural glass products.

Therefore, for several samples (such as float glass, borosilicate rods), 2-3 different maximum temperatures were tested, corresponding to the viscosity range of 105-6 dPa s and 103-4 dPa s, in order to further study the defect effect. The influence of bending. All samples were kept at the highest temperature for 10 hours, 4 quenched to their annealing point at a rate of -160 degrees Celsius/hour, hot dipped for 10 hours and cooled to their strain point with a ramp of -4 degrees Celsius/hour, and then at a faster rate Controlled cooling to room temperature. This conservative annealing scheme ensures that the sample is stress-free, as seen by cross-polarized light.

1 The production of cathode ray tube (CRT) screens has ceased, but there are still a large number of CRT glass fragments resulting from the separation of the processed screens (Andreola et al., 2005). 2 If fewer samples are reported, the available cullet is not enough to produce three samples. Nevertheless, these samples are provided in this study to show the failure mode of a specific type of glass, not to derive absolute flexural strength values. 3 Crystalcast M248 is an investment powder composed of 73% silica (cristobalite, quartz), 23% calcium sulfate (gypsum) and 1% organic matter (Goodwin Refractory Services 2003; Gold Star 2019). The choice of mold material is related to the kiln casting technology used in this work, rather than the commonly reported melt quenching casting. Due to the selection of a relatively low molding temperature, the corresponding high viscosity of the heated glass does not allow it to be immediately poured from a melting (platinum or high alumina) crucible into a preheated (steel or graphite) mold for annealing. Therefore, the entire casting process must be carried out in a mold that can withstand temperatures up to 1150°C, will not adhere to the glass, and will not cause the sample to break during the cooling process. 4 In view of the high viscosity at the highest temperature and the size of the sample, it has been found from experience that staying for 10 hours is suitable for removing large bubbles (> 1 mm) and incorporating the coating into the glass network.

The sample is produced at a component height of 40 mm, and then cut into a certain size with a water-cooled rotary diamond grinding wheel cutter to remove top surface components that usually contain a large number of defects (such as surface crystallization, bubbles, alkali loss), wrinkles, and cracks. Then, the samples were ground and polished in the order of 60, 120, 200, 400, and 6005 grit sizes using a Provetro surface grinder and diamond grinding disc, and their final dimensions were recorded.

The inhomogeneity of the glass sample was observed with the naked eye and observed with a Keyence VHX-5000 or VHX-7000 digital microscope. By using a cross-polarizing filter, the internal residual stress in the glass sample can be qualitatively evaluated. Finally, by using elastic white and black spray paint to create a speckle pattern on one of the longitudinal surfaces, the beam is ready for digital image correlation (DIC) measurement.

The preparation process required to produce the kiln casting test piece is shown in Table 1. In addition to the kiln casting test pieces, the following industrial manufacturing test pieces were prepared as reference:

The grinding and polishing procedures followed to prepare the above samples are the same as those described for the kiln cast samples. However, the bottom and top surfaces of the float glass samples (single and bonded) remained in their receiving state (optical fine polishing), and only the cut edges were processed.

2.2 Four-point bending test device

The first series of experiments (12 kiln castings, 6 reference samples)

The first series of experiments were conducted to provide a general overview of the bending behavior of different glass samples. A Zwick Z10 displacement control universal testing machine was used to test samples in a laboratory air environment at a speed of 0.2 mm/min. The four-point bending fixture has a load roller span of 110 mm and a support roller span of 220 mm, with a fixed load pin of 10 mm diameter, and is loosely connected to the testing machine to allow some articulation (Figure 2a).

The second series of experiments (53 kiln castings, 5 reference samples)

The second set of experiments involved repeated tests for each glass category and provided accurate displacement data. The number of test samples for each glass category is set to three, which is limited to testing brittle materials. Due to the randomness of defects in glass, the strength of this material defaults to statistical data (Quinn et al., 2009). However, this research aims to cover a wide range of glass types and compare them based on their bending behavior to explore which recycled glass products have further structural potential.

For these tests, a Schenck 100KN displacement control hydraulic universal testing machine was used to test the specimen in a laboratory air environment with a displacement rate of 0.3 mm/min, which approximately corresponds to a rate of 0.5 MPa/s. 8 The loading roller of the four-point bending fixture has a span of 100 mm, and the support roller has a span of 200 mm, with a fixed loading pin of 20 mm diameter (Figure 2b). To allow fine adjustment and rotational movement, the support fixture is placed on a semicircular pin, while the loading fixture is loosely connected to the testing machine. In addition, a 1 mm thick silicone rubber strip is placed between each loading pin and the specimen.

In order to measure the beam displacement caused by bending, two methods are used: 1) The linear variable differential transformer (LVDT) displacement sensor (Solartron AX 2.5 spring push probe calibrated to 0.5μm accuracy) is placed at the midpoint of the beam Below. The lower surface of the beam (measure the point of maximum displacement), and 2) 2D-DIC measurement, using a high-resolution (50.6MP) Canon EOS 5Ds camera to take a photo of the surface of the beam spot every second. Use GOM Correlate software to analyze the 2DDIC measurement pictures. One image pixel corresponds to 31.5μm, so when the software accuracy is 0.05 pixels, any displacement above 1.57μm can be captured. The bending strength and Young's modulus calculation The bending strength (σ) is calculated by the following equation:

Where F is the maximum load, L is the support span, Li is the load span, b is the width of the beam, and d is the height of the beam. 9

The calculation of Young's (E) modulus is performed by correlating the force data obtained from the Schenck machine with (1) the maximum displacement from the LVDT sensor and (2) the maximum displacement from the DIC analysis (Figure 3).

5 Set 600 grit finishing according to ASTM C1161-13. In addition, Quinn et al. (2005) In their research on processing cracks in ground ceramics, it was observed that the sintering reaction combined with the bending of silicon nitride specimens using 600 grit grinding failed due to material defects rather than processing damage. This observation can be extended to glass samples. 6 Poesia is a manufacturer of cast glass bricks for the exterior of the crystal house (Oikonomopoulou et al., 2018a). 7 This UV-curable acrylate was chosen because it forms a strong bond with the glass surface, resulting in the overall behavior of the glued sample (Oikonomopoulou et al., 2018a). Under four-point bending, the bonded glass samples are expected to exhibit cohesive failure in the substrate (glass layer) instead of delamination. 8 The second series chose a slightly faster replacement rate in order to reduce the total number of DIC images in each experiment, thereby limiting the file size generated by the image processing software GOM Correlate to a maximum of 25 GB. The displacement rates of the 1st and 2nd series are both lower than the 1.1±0.2 MPa/s stress increase rate indicated by ASTM C158-02. Displacement control rate is better than force control to avoid collision of specimens in the event of failure, and when the crack front interacts with the interface encountered in the glass, it also allows potential ejection under maximum force (slight crack stop) mesostructure . 9 It should be noted that due to the fixed loading pins, there is a friction constraint of μ·F/2 at each pin, and μ is the friction coefficient (Quinn et al., 2009), so a positive system error may occur. This force produces a reaction moment of μ·F·d/2, so the above equation should be rewritten as:

Assuming medium μ = 0, 3, the systematic error of the first group of experiments may be 8.2%, and that of the second group is 9%. However, due to insufficient data of μ value, this study has not corrected the bending strength, and readers should consider the possibility of errors that approximate the above range.

Given that the cross-section of the beam relative to the span of the fixture results in a relatively rigid structural element, the shear deflection should be included in the total vertical deflection. The bending and shearing deflection of the mid-span relative to the beam point above the support pin, for a four-point bending fixture ratio of 1:2, is defined by the following formula:

Adding the two vertical deflections to solve the Young's modulus, we get: 10

3.1 Defect evaluation of kiln cast samples

The defects in the surface and body of the produced glass samples are qualitatively recorded 11 according to the type and reason. The goal is to correlate the found defects with the glass source used and the subsequent casting and post-processing procedures, and then evaluate their contribution to the flexural strength of the specimen. Classification of defects related to casting 12:

1. Crystal inclusions 2. Glassy non-uniformity (rope/ream) 3. Gaseous non-uniformity (bubbles)

An overview of the defect categories and their causes can be found in Figure 4. Based on this, Table 2 lists the documents of the defects observed for each glass type.

Table 2 Evaluation of kiln cast samples-full size table

10 For the calculation of Young's modulus, use the Poisson's ratio of soda lime quartz glass v = 0.22. Although there may be a deviation of ± 0.02 from this value in the tested glass, the effect on the result is negligible. 11 Quantitative analysis of the non-uniformity level of cast glass samples with considerable cross-sections-therefore multilayer defects and thin-walled glass-is a complex process involving many different testing methods (such as computer tomography to detect and Measure the density difference, 3D imaging real-time optical rotation method to define the position and shape of the rope, etc.). This analysis is not within the scope of this study, because the main purpose is to first determine the type and location of key defects that require future attention, and thereby determine the quantitative documentation. 12 Based on the classification of Bartuška (2008).

In more detail, the causes of these defects are related to one or more of the following manufacturing stages 13, 14:

Due to insufficient melting of the coating, several "flat" defects were observed in kiln-cast samples of float glass fragments covered with enamel paint or ceramic frit (Figure 5). XRF analysis of the two characteristic coatings (Table 3) showed that the composition is rich in high melting point metal oxides, especially chromium(III) oxide (Cr2O3 has a melting point of 2435°C, NIH database). X-ray diffraction (XRD)15 analysis (Figure 6) of the furnace-cast glass sample showed the presence of eskolaite (the mineral name for chromium oxide) under these conditions.

Minor composition changes can cause glass-like inhomogeneities, such as ropes and color streaks. Some examples with heavy streaks were found in the "Float combination" and "Lead CRT" (Figure 7) samples.

In this category, the presence of glass ceramics or glass families with different chemical properties in cullet (which cannot be detected by the naked eye, such as aluminosilicate fragments in borosilicate or soda lime silica cullet) is the most critical , Causing the sample to rupture when cooled, due to the strain caused by the change in thermal expansion. This is reflected in the "Float Combination" (Figure 8a) and "Maltha Borosilicate Mix" samples. More specifically, the "float assembly" specimens were cast using a compilation of flat glass fragments (approximately 20-50 mm wide) provided by Maltha Recycling. Because the erroneous deposition of glass ceramic plates (such as cooktops) in the flat glass collection container—a phenomenon that is often encountered—makes the entire container unsuitable for recycling, the flat glass combination is excluded from the recycling stream.

The XRF and XRD analysis of the characteristic slices of the flat glass compiled samples (Table 4; Figures 8b, 9) places the contaminants in a commercially available lithium aluminosilicate glass ceramic system, which is characterized by a coefficient of thermal expansion close to zero (Höland and Beer) 2020). The extremely low coefficient of thermal expansion (CTE) contrasts sharply with the typical float glass of 9.5 × 10-6/K (at 20–300 ºC) (Shelby 2005), leading to inevitable cracking. However, reducing the particle size of the flat glass compiled sample (fine glass or powder) can minimize the strain in the final cast product, so this strategy needs further research.

Traces of metal, clay, or stone can cause crystalline inclusions up to 2 mm in size, but the glass network can tolerate these inclusions (Figure 10, 11). However, further research is needed to identify crystalline inclusions (using a scanning electron microscope) and test whether their effect when the glass is subjected to a temperature gradient remains neutral.

B. Broken glass size and shape.

Within the mentioned glass viscosity range, the geometry of the cullet is usually reflected in the stripes and/or three-dimensional bubble film in the final glass part. In the case of very fine broken glass (such as the "car windshield" sample), this geometry is indistinguishable and there is a fairly high content of tiny bubbles (Figure 12).

Table 3 Coating composition-full size table

Table 4 The chemical composition, crystal phase and CTE of typical lithium aluminosilicate glass ceramics, compared with tested glass ceramic samples, cast "Float Combo Maltha" samples and typical window glass-full size table

*Lithium is a light element that cannot be detected by XRF analysis, so the percentage corresponding to lithium oxide is reflected as a higher content of silica. According to the bibliography, the provided composition should have a lithium oxide content of 2-3% and a silica content of 2-3% **Lithium oxide content in the chemical composition is expected to be less than 3% [1] XRD measurement by Ruud Hendrikx (Delft University of Technology, 3me) Bruker D8 Advance diffractometer, Bragg-Brentano geometry and Lynxeye position sensitive detectors were used for XRF measurement by Ruud Hendrikx (Delft University of Technology, 3me) using Panalytical Axios Max WD-XRF spectrometer; [3] Montazerian et al. (2015); [4] Songhan Plastic Technology Co., Ltd.; [5] Schott (2015b); [6] Holland and Bill (2020); [7] Shelby (2005); [8] Zheng ( 1977); [9] Brennan (1979); [10] Campbell and Haji (1975)

13 In Table 2, the defects caused in stage III.B. And IV (post-processing and processing defects) are not mentioned because they have nothing to do with the material and its casting method, but rather random, only related to the fracture analysis of each specific specimen. 14 The following microscope images were taken using Keyence VHX-5000 or VHX-7000 digital microscopes. 15 All XRD analyses in this work were performed using Bruker D8 Advance diffractometer, Bragg-Brentano geometry and Lynxeye position sensitive detector.

A. The arrangement of broken glass in the mold.

This is related to the geometry of the cullet (I. B), the firing plan and the corresponding viscosity (II. B, C) of the formed glass. The defined cullet shape and high viscosity can result in an organized mesostructure composed of bubble films (Figure 13b, 14), ropes, or crystalline interfaces, leading to more predictable failure modes (Figure 13). This organized structure also helps to distinguish the effects of these defects on the glass surface or on the whole.

B. Molding temperature and corresponding viscosity related to residence time.

The top temperature affects the degree of homogenization and bubble content. Due to the relatively low molding temperature, all samples have tiny bubbles. In addition, the "cage" principle that describes the mixing of dense liquids applies to this situation, which means that most of the molecules corresponding to the initial cullet will remain in the same position relative to its neighboring molecular cluster (the cullet). When the viscosity reaches 103.5 At the order of dPa s, the diffusion level will increase, but under no circumstances will it result in a completely mixed glass within a given residence time (see Figures 12 and 13).

C. Combine the firing plan with the temperature difference in the kiln that promotes crystallization.

This applies particularly to float and borosilicate glass samples formed at 970°C. In these samples, the complete interface between each broken glass piece crystallized. According to XRD analysis (Figure 15), the borosilicate samples developed into b-cristobalite crystals, while the float glass samples were wollastonite 2M, b-cristobalite and perovskite (Figures 16, 17). Crystallization is advantageous because the sample is formed below its liquidus point (the TL of a specific float glass is about 1080 degrees Celsius, and the TL of a specific borosilicate is about 1200 degrees Celsius), but the viscosity is low enough to allow kinetics Nucleate. Nucleation starts at the interface, because there, local composition changes occur due to the volatilization of alkali and boron (in the case of borosilicate glass).

However, as observed in the crystalline layer of the borosilicate sample, the consumption of these elements may result in an unstable local composition, which proved to be porous and water-absorbing (Figure 17). In addition to the above-mentioned "engineered" crystalline structure, temperature conditions and fluctuations in the kiln can also cause local and random crystallization in the form of stones where the composition changes. Local changes in the composition may be caused by contaminants in the raw material, contact with the mold material, volatilization of the compound, and air bubbles. Therefore, such gemstones are not only found in samples made of obviously contaminated broken glass (such as "car windshield" samples), but also in purer samples (such as "full tempered (FT) 17 float). Method” sample).

D. Reaction with the mold surface.

During the kiln casting process (within the studied viscosity range), the glass in contact with the silica/gypsum investment mold forms a thin crystalline interface, which can be easily removed by the post-treatment method described (Figure 18). However, of particular interest are the defects caused by the interaction of the mold with the glass, which are deep enough to be retained during grinding (Figure 19). These can be, for example, about. ø 1–2 mm is made of loose mold material, which accidentally gets mixed into the glass melt. Another characteristic defect is caused by the friction on the surface of the mold preventing the complete fusion of the broken glass pieces. Therefore, there will be partial or intra-network wrinkles on the glass surface, which can also encapsulate the mold material. During grinding, the tips of these defects may remain on the glass surface and can be observed to a depth of 5 mm. In the end, only one case was observed in which the glass was bonded to the mold surface and cracked due to changes in thermal expansion during cooling (sample "Borosilicate Mixture Maltha").

E. Quenching rate to the annealing point.

In this study, compared with sudden quenching in industrial glass casting, a lower quenching rate of -160°C/h was used. 18 Experimental results show that this rate is sufficient to prevent crystallization. However, it has been noted that a slower cooling rate may increase the degree of polymerization of the glass network and result in a denser glass (Ito and Taniguchi 2004). Although this has not been experimentally proven in this study, it is still possible to be taken into account.

A conservative annealing scheme is used, so the residual stress detected in the sample using the cross-polarization filter is negligible and does not seem to affect the flexural strength. Regarding the samples cut from standard Poesia glass tiles, they do have a small residual stress, which can also be seen from the sequence of stripes in the isochromatic pattern obtained from the Ilis StrainScope Flex circular polarizer (Figure 20), and it also strongly shows This glass has a tendency to crack during post-processing.

16 The liquidus point of the glass TL is found near the viscosity of 104 dPa s and is estimated based on the chemical composition of a given glass according to Fluegel (2007a). 17 The name "fully tempered" refers to the cullet used for these samples, which comes from broken fully tempered float glass plates. The final kiln cast parts are annealed and therefore not tempered. 18 In this study, quenching may even last for 4 hours and be performed in a kiln, which is essentially different from quenching under atmospheric conditions that lasts only a few minutes during hot pouring of glass.

A. Insufficient removal of existing defects.

As discussed in point II.D, not all surface defects can be completely removed by post-processing (Figure 21a). Among such defects, the exposure of air bubbles trapped in the glass body during grinding should be included. This will cause the stress concentration on the sharp edges of the glass surface to intrude into the semicircular shape, thereby reducing the strength. In addition, since air bubbles can provide favorable conditions for the formation of crystals inside, such air pockets exposed on the surface expose the additional risk of stone exposure (see Figures 19b, 21b).

B. Introduce new defects.

The new scratches introduced by “rebellious” abrasive grains (Quinn 2016) were mainly observed in the glass with lower hardness, in this study, especially the “Leerdam Lead” sample. Such as II.F. All samples have the risk of microcracks during the rough grinding process, and they are not fully removed in the later stages of grinding and polishing.

A series of processing defects (defects, cleavage, impact cone, point contact) randomly appeared in some samples. Compared with industrially produced glass (such as float glass, extruded rods), the reaction of cast samples to processing damage needs further study, but it has not been observed that purer cast samples are more susceptible than standard glass products. However, attention should be paid to glass samples with more serious contamination, because occasional large defects (> 2 mm) on the surface will magnify the impact of impact.

The results of the two series of experiments are shown in Figure 1 and Figure 2. 22, 23, 24, 25 and 26 and Tables 5, 6, 7 and 8. The data of the first series is mainly used for the first general guidance and confirmation of the second series. This is the main focus of this research. It should be emphasized that the number of test samples for each category is limited, so the presented The results are only indicative and not enough to draw statistical conclusions.

Table 5 Results of four-point bending test of the first series of kiln cast beams-full size table

Table 6 Results of the first series of four-point bending experiments on the reference beam-full size table

Table 7 Results of four-point bending test of the second series of kiln cast beams-full size table

Table 8 Results of the second series of four-point bending experiments on the reference beam-full size table

Although the first and second series are in the fixture settings (span, roller radius, connection details with the universal testing machine) and the sensitivity of the testing machine (the maximum applied load of the machine used in the first series is 10KN and 2), both The results of the test are consistent. More specifically, the first series of samples ("FT Float", "Schott DURAN 24 mm Rod", "Oven Door Coolrec") cast at 1120°C scored within the same bending strength range (40-50MPa), The sample melted at 970°C is significantly weaker (10-20MPa), while the performance of the pure monolithic float sample is slightly better (average bending strength is 55MPa). The performance level and value range are consistent with the results of the second series, except for the case of melting float samples at 970°C, the reported bending strength is significantly lower (< 10 MPa). This is due to the one-time micro-crack network on the surface of these samples, which cannot be easily removed by post-treatment.

In Figure 24, the flexural strength of the second series of specimens is depicted, and three main areas can be observed: specimens with a flexural strength of less than 30 MPa, between 30 and 55 MPa-where most samples are located, and Between 55 and 75 MPa. In all specimens, crack initiation started on the bottom surface (or very close), in the area between the support pins (the zone of maximum tensile stress, see Figures 27 and 28). As a general trend, glass samples produced with lower viscosity and cleaner cullet were found in the top area of ​​the bending strength graph, while samples with obvious strength limiting defects exposed on the bottom surface failed at low values.

An overview of the main fracture origins is given in Figure 29, summarizing the most critical defect categories: stones, crystalline interfaces, surface bubbles, and processing damage. The size of the fractured mirror was measured in the selected sample (Figure 30), and plotted according to the "Ohre's equation" (Quinn 2016) and the bending strength σ (Figure 31):

Where R corresponds to the mirror radius (in this study, the size of the mirror extending to the haze boundary of the bottom surface of the maximum tension was measured), and A is the characteristic mirror constant of each glass component.

Generally, the greater the failure stress, the smaller the fracture mirror encountered. As expected, the increase in critical defect width is responsible for the decrease in bending strength (Figure 32). The higher-strength specimens seem to fail mainly due to processing defects, while the stone or crystalline interface is the cause of the fracture of the lower-strength specimens. However, the type, size, number, and location of defects alone cannot prove why some glass samples score lower than others. The structural properties of each glass type need to be reviewed according to the chemical composition of the glass and its inherent defects (see section 4). In addition, due to environmentally-assisted slow crack growth (Quinn et al., 2009), the uncertainty of the fracture load may apply because the rate of applied load is slower (approximately half) than the rate recommended by the ASTM C158-02 guideline. The effects of slow crack growth should be further studied experimentally within a wider range of test speeds.

Regarding the Young's modulus, the calculated value based on the LVDT data is about 15% lower than the value in the literature. This is considered a systematic error and is attributed to the quality of the sensor. However, in the triplet of each test glass type, a matching E value is reported. In addition, according to the literature (Corning 1979; Campbell and Hagy 1975), the stiffness relationship between different glass series was found (Figure 26), in particular: 

The bending strength of cast glass specimens is related to its chemical composition and inherent defects. In order to understand under which conditions defects are the strength limiting factors, and when the mechanical properties related to the composition play a decisive role, the interpretation of the results is divided into the following categories: a. Uncontaminated glass sample, b. Contaminated and uncontaminated glass, c. An uncontaminated homogeneous glass sample compared with a sample with a crystalline interface, and d. Reference specimen. In this way, defects are classified and isolated, so their effects can be studied more clearly, without obvious defects (in the case of pure samples) highlighting the influence of chemical composition.

Select the purest and most homogeneous samples in each glass series included in this work for comparison (Figure 33, 34). Since these examples contain fewer defects, they highlight the influence of their chemical composition on their bending strength. Table 9 lists the calculated and/or measured physical and mechanical properties of these glasses, as well as data on similar compositions found in the literature.

Therefore, although LVDT calculations cannot provide accurate values, it can be reliably used for comparative analysis between different glass types. DIC measurements are used to provide more accurate data on maximum deformation and perform more accurate E modulus calculations for selected glass samples (see section 4a). Nevertheless, it is recommended to combine DIC measurements during 4-point bending with non-destructive test methods (such as pulse excitation technology) used to determine the E modulus in future tests to verify the reliability of the results.

As shown in Figure 33, in lead silicate, borosilicate, barium silicate, and AGC dark blue float glass samples, as the Young's modulus increases, the flexural strength increases by 19%. The increase in strength is attributed to the increase in the average bonding strength and atomic packing density of the glass network. This is related to Young's modulus by the following equation (Makishima and Mackenzie 1973):

Where Cg is the atomic packing density (also called atomic packing factor, APF), and Gt is the total dissociation energy per unit volume. According to the chemical composition obtained by XRF analysis, APF20 and Gt 21 were calculated and listed in Table 9.

Therefore, by looking at Table 9, it is expected that the lead silicate glass sample with the lowest dissociation energy and bulk density will also have the lowest strength, 22 and soda lime silicate (SLS) glass will have the highest strength. Also according to the literature, silicate containing BaO The bending strength and Young's modulus of glass are lower than CaO silicate, but higher than lead silicate (Volf 1984; Corning 1979). However, the Young's modulus alone does not prove that the deviation of the Poesia, Wertheim, and FT float glass samples from the linear E/strength relationship is correct, and further explanations are needed for each glass type.

Table 9 Measurement and calculation characteristics of selected (pure) glass (bold) and reference glass of similar composition-full size table

Poesia glass is an improved soda lime glass. Compared with traditional float glass, its molding temperature is lower (TL is about 980°C, so it is 80-100°C lower than SLS). It contains a small amount (< 3 wt%) of K2O and B2O3, and has a higher Na2O/CaO ratio than typical SLS formulations. Although the E modulus 23 is slightly lower than AGC dark blue glass, it exhibits the highest flexural strength of all the tested samples. This is due to the low brittleness of this special glass. Sehgal and Ito (1998) pointed out that higher molar volume (Vm) plays a key role in reducing brittleness, because a more open structure allows more deformation before crack initiation. More specifically, increasing the soda/calcium oxide ratio will reduce brittleness and partial replacement of potassium by soda. This is consistent with the compositional change of Poesia glass with the typical SLS formulation, which helps to form a more open structure (Figure 35), allowing a slight increase in the adjustment of the stress around the point/defect where the crack will start. "The Poesia 1070 ◦C specimen failed due to machining damage (Figure 36).

Wertheim glass has the highest measured Young's modulus and the highest calculated total dissociation energy, and the calculated molar volume is similar to Poesia glass. The higher stiffness (compared to SLS glass) can be attributed to the partial replacement of silica with alumina (≈ 5%), which reduces the openness of the network Sehgal and Ito (1998). Similar to Poesia glass, its molding temperature is lower than SLS glass (TL is about 1015 degrees Celsius), which may be related to the mixed alkali effect 24 and the presence of a small amount of boron trioxide (Morey 1932). According to the E/Vm characteristics of this glass, higher bending strength should be expected.

The reason for the failure of this glass under lower stress is related to its kiln casting at temperatures far below its liquidus point (820°C, 900°C), which leads to obvious inhomogeneities. These inhomogeneities are concentrated in the interface formed between each glass particle and form a 3-dimensional network of flat areas composed of bubbles and loose crystal structures. In addition, due to the reaction between the hot glass and the mold, the molding temperature is conducive to the formation of stones, which are sufficiently subsurface and cannot be completely removed in a standard post-treatment process. These stones seem to weaken the glass surface and help to form deeper streaks during the grinding process, which is the source of failure. The aforementioned 3D network may not be the cause of the crack origin, but considering that the sample fails due to defects close to the network, it may cause stress concentration areas along the surface (Figure 37).

It is also interesting to compare the "AGC Dark Blue Float" with the "FT Float" specimen. The composition of these two glasses is very similar, and the calculated atomic packing density and total dissociation energy are almost the same. However, the measured "FT float" glass sample has a lower Young's modulus and lower bending strength. This may be related to the thermal history of these two types of glass. On the one hand, the liquidus point of "AGC Deep Blue" glass is slightly lower (TL is about 1046 degrees Celsius, while "FT Float" is 1063 degrees Celsius). On the other hand, the dark color of AGC glass seems to contribute to the quality of the casting. The dense dark blue absorbs more infrared rays during the heating process than the transparent light blue, so the body heats up faster. In a similar way, due to greater heat loss due to radiation, dark glass solidifies faster during cooling (Kita˘ıgorodski˘ı and Solomin 1934; Burch and Babcock 1938).

The faster solidification rate will affect the coordination state of the transition metal oxides contained in the composition, thereby affecting the total dissociation energy of the network bonds-this is not taken into account in the calculations. In addition, the lower liquidus point and increased heat absorption promote the full fusion of glass fragments and eliminate stone formation, thereby reducing defects on the glass surface and improving bending strength. On the contrary, the inward folding of the glass surface of the "FT float" sample is caused by insufficient fusion of the cullet near the mold wall, and crystalline inclusions due to mold contamination are the main reason for the failure of "FT float" The sample, based on the analysis of the broken mirror (Figure 38). Due to these defects, the failure value of the "FT Float" glass sample is lower than expected compared with the rest of the samples.

19 The graph in Figure 33 is based on Young's modulus calculated from DIC measurements. The reported E modulus is about 5% higher than that in the literature. This may be partly related to test errors and partly to the material itself and its casting procedure. 20 The atomic packing density is calculated using the following formula:

Where xi is the mole fraction, Vi is the ionic volume of the i-th oxide, and Vm is the molar volume of the glass, specifically:

Where NA is the Avogadro number, rA,B = the ionic radius of the MxOy oxide, M is the molecular mass, and ρ is the density of the glass. Vi is derived from Makishima and Mackenzie (1973) and Inaba et al. (1999) Based on Pauling's ion radius. A model developed by Fluegel (2007b) was used to calculate the density based on the chemical composition. 21 The total dissociation energy is calculated based on the dissociation energy of the oxide components listed in the work of Inaba et al. (1999). 22 PbO is one of the lowest Gi reported by Makishima and Mackenzie (1973) and has a relatively high molar atomic mass. The increase in the mass of lead ions slows down the chemical reaction in the quenching process and results in less network organization/accumulation. 23 It should be noted that it was found that EPoesia calculated using APF and Gt corresponding to the chemical composition is much lower than EAGC blue. This may be related to erroneous estimates of the B2O3 content that cannot be determined by XRF analysis and/or higher packing density attributable to the thermal history of the kiln cast parts. Therefore, EPoesia derived from LVDT data is considered more reliable and can be used for further analysis. 24 This term describes anomalies observed in glasses and melts containing a mixture of two or more alkali metal oxides. According to Shelby (2005), the viscosity of such melts is lower than those containing the same amount of a single alkali metal oxide. 25 XRF identified a series of transition metals as colorants in the glass: 0.76% Fe2O3, 0.065% TiO2, 0.029% MnO, 0.023% Cr2O3).

Regarding the fracture analysis of the glass studied in this category, it was found that the most common failure causes were processing damage and handling defects (see also Figure 28, most "pure" specimens failed at edge defects), which proved that the purity is reasonable fragmentation The glass and the relatively high forming temperature (compared to the other glass samples in this work) eliminated the number of strength limiting defects. As mentioned above, there are exceptions in the "FT float" series and "Schott DURAN tube" samples. These borosilicate glass samples are actually formed under high viscosity (≈ 104.5 dPa s <TL), and are characterized by an increase in the number of bubbles (mainly concentrated at the interface between each broken glass piece during the molding process). These bubbles form clusters of crystal growth, and if they are located on or near the glass surface, they will become strength limiting defects that cause fracture (Figure 39). The bending data of the two glasses obtained from the first set of four-point bending experiments match the results of the second test.

(b) Contaminated and uncontaminated glass samples

This category studies glass samples cast from contaminated glass cullet kilns and compares them with the purer samples mentioned above. All the specimens studied, "float combination", "oven door", "car windshield" and "AGC enamel black" are typical soda lime silicates, with a large number of obvious crystalline inclusions and/or thick ropes . Their bending strength is slightly lower than that observed for FT Float specimens, and their Young's modulus is comparable. The "AGC Enamel Black" series seems to have the highest flexural strength in this category, which is due to the fact that these samples only use one type of glass for casting (so no rope is observed due to minor changes in composition). In addition, the larger Small shards of glass can be cleaned thoroughly, which is not the case with the smaller shards of the "oven door" and "car windshield" samples.

The failure values ​​of all samples are lower than most of the purer glasses studied above, mainly due to the formation of crystals on the surface (Figure 40). These stones are produced by inherent pollution or by the further reaction of the pollutants with the mold material. During the 4-point bending process, multiple defects located in most of these specimens were not activated, and they did not seem to reduce the Young's modulus. On the contrary, these defects are tolerable in the glass network. However, the more defects in the block, the greater the chance of these defects being exposed on the surface, and therefore the higher the risk of failure.

(c) Uncontaminated homogeneous glass samples and samples with crystalline interfaces

In this category, samples of soda-lime and borosilicate glass ("float 1cm, 3 horizontal layers", "float 1cm, 24 vertical layers", and " Schott DURAN 10 vertical layers""), and a more homogeneous version formed by the kiln at 1070°C and/or 1120°C ("FT Float 1120°C", "Schott DURAN 10 vertical layers 1070°C, 1120° C”) Conduct comparative studies. The special aspect of this category is "defects" or areas with changes in composition/structure, which are deliberately designed at specific locations and geometric patterns. Therefore, in contrast to the random stones in the above categories, in this In this section, the size and distribution of the crystal structure can be predicted. Therefore, their influence on the structure performance can be directly related.

Therefore, it can be observed that the fused "Float 1cm, 3 horizontal layers" specimen and the more uniform "FT Float" specimen exhibit very similar flexural strength and Young's modulus (Figure 41). This is because the crystal interface is located in the body, in a layer parallel to the bottom surface, and therefore will not be exposed to the maximum tensile stress zone. Therefore, they behave similarly to homogeneous samples. However, this is not the case for the "floating 1 cm, 24 vertical layers" sample, where the crystalline interface is exposed on the bottom surface and is actually aligned perpendicular to the tensile force. Although the Young's modulus remains similar, the bending strength is reduced by more than 20%. The starting point of fracture for these samples is always at these crystal-glass interfaces and starts from the adjacent glass zone. Therefore, the crystalline structure appears to act as a stress-inducing element, and its fracture toughness may be higher than that of the surrounding glass matrix, thereby weakening the glass sample.

The type and thickness of the crystalline interface also play an important role. The thin b-cristobalite layer produced in the "Schott DURAN 10 vertical layer 970°C" caused a sharp drop in strength of 75% and a drop in Young's modulus. In a similar way to the float example, the starting point of fracture is always at the crystal-glass interface.

At this point, to understand the influence of these crystalline structures and their geometrical arrangement, one should pay attention to molten glass samples produced at a viscosity of about 106 dPa s and at a point much higher than the liquidus (where the diffusivity is much higher). The samples produced at a viscosity of 105 or even 104 dPa s seem to retain the traces of the interface between each piece of cullet during the heating process, in the form of fine bubble films, ropes and crystal formation points. For example, this is evident in the "Schott DURAN 10 vertical layer 1070°C", which is kiln cast at a viscosity of 105 dPa s (Figure 42).

These samples contain the above-mentioned bubble film and stones, and their geometric arrangement is the same as the "Schott DURAN 10 vertical layer 970°C". Although these specimens are stronger than the smelting type, their flexural strength is 30% lower than that of the specimens kiln cast at a higher temperature of 50 ºC. They all fail due to crystal defects or bubbles located in one of these veils (Figure 43). This is a very critical issue because the samples produced at 1070°C and 1120°C look the same and transparent, and cannot be compared with the contaminated samples described in the above categories. This highlights how important a temperature difference of 50 ºC is when casting at viscosities near and below the liquidus point.

Tests on industrially manufactured glass samples to provide reference points for comparison with kiln cast glass samples. The following describes their structural properties by type.

Beams cut from standard Poesia cast glass bricks (initially hot cast at around 1200°C) are more uniform than kiln cast samples produced in the laboratory. Except for some fine streaks and few bubbles, they do not contain key defects such as pebbles, because the purity of the raw materials, the above-mentioned liquidus point formation temperature, sudden quenching under atmospheric conditions, and the stainless steel molds used for casting prevent them from being damaged. form. Nevertheless, the flexural strength of these specimens at 1070°C is 10% lower than that of the less uniform recast specimens. This is due to the faster annealing scheme used for these components, which causes the residual stresses in the glass to freeze and makes it more susceptible to damage. Therefore, during the size cutting and grinding of parts, multiple chips and the resulting cleavage damage were not completely removed during the polishing process due to insufficient annealing. Increased stress and processing defects are the cause of fracture, and the strength is lower.

Considering the individual float glass plate samples, these samples are the most uniform of all the studied samples, with original polished bottom and top surfaces. Since these samples were cut from larger float plates, their edges were ground and polished as described in Section 3. 2.1. All single glass samples from the first set of four-point bending experiments failed due to processing defects at the edges, in the bottom area of ​​the maximum tensile stress. The average bending strength of the 10 mm glass plate is 55MPa, which is 20% higher than the "FT Float 1120 ◦C" sample, but 20% lower than the highest-scoring samples "AGC Dark Blue 1120 ◦C" and "Poesia 1070 ◦C" C". There is no doubt that the quality of the bottom edge will greatly affect the bending strength of the float glass sample during the bending process.

According to the sample size and polishing quality and test settings, various bending strengths of 4-point bending are reported in the literature, ranging from 35 to 170 MPa (Veer and Rodichev 2011), 51-71.5 MPa (Veer 2007), 53-129 MPa (Yankelevsky et al. al. 2016) Just to name a few. Considering that the edge finish is relatively rough, the 55MPa strength of the tested sample in this study is at the low end of this range, and is consistent with the literature. More fine polishing can expect higher strength. In this sense, and considering that the kiln-cast specimens exhibit higher tensile strength and also fail due to machining defects, it can be deduced that industrial fine polishing of the kiln-cast specimens may result in higher strength. .

A beam produced by bonding (Delo Photobond 4468) 8/10 mm thick float glass layer and tested with the layer parallel to the bottom surface. The average bending strength is 48MPa (the first and second four-point bending series) , Which is 10% higher than the "FT Float 1120 ◦C" sample formed by the kiln. None of the specimens failed due to edge defects; the cause of the failure was slight operational damage to the bottom surface. Due to the adhesive layer, the Young's modulus of the bonded beam is lower than that of the whole, kiln-cast SLS specimen.

In general, the flexural strength values ​​obtained from the industrially produced reference samples are at the top of the 30-55MPa (second) region, and do not exceed the performance of the purest kiln-cast samples (found in the first region). This is an encouraging result because all kiln cast samples produced in this study have a certain degree of non-uniformity.

The ability of various types of commercial glass waste materials to be kiln-casted into structural parts at relatively low temperatures (820-1120°C) was tested, and the bending strength of the kiln-cast samples was evaluated.

Kiln casting experiments show that the careful separation of cullet in the recycling facility ensures successful casting. The glass network can tolerate coatings and trace external contaminants, such as organics and metals, but can cause defects and low bending strength, while contamination of glass ceramics and glasses with significant composition changes can cause samples to break during cooling. A glass composition with a lower liquidus point is beneficial to low-temperature kiln casting, which leads to a more uniform glass surface, because the lower viscosity during the molding process can minimize sintering defects and surface bubbles caused by mold contamination And the occurrence of stone formation.

Regarding the four-point bending experiment, although the number of test samples for each glass type is not enough to derive statistical data, they do provide a good overview and reasonable estimation of the pollution of each type of sample based on chemical composition and grade. Follow the casting parameters.

In samples produced from purer cullet and higher molding temperature, the influence of chemical composition on strength can be clearly observed. In these samples, a significant increase in strength and Young's modulus was observed, from lead silicate to borosilicate, barium silicate, and then to the soda-lime silicate series. Cleaner and more uniform samples fail mainly due to external defects caused by processing and handling damage. However, in samples with more serious contamination, the influence of the composition becomes obscured, in which the crystalline structure formed on the bottom surface in the region of maximum tensile stress is the main cause of fracture that causes a significant decrease in strength.

Among the soda-calcium silica series, a slightly improved formula containing a small amount of K2O and B2O3 and a higher ratio of Na2O to CaO is particularly promising. The lower viscosity of these glass melts facilitates the casting process, and their more open structure (higher molar volume) provides a less fragile alternative for Young’s modulus similar to that of SLS glass. This ultimately leads to higher bending strength.

Glass series with lower liquidus points, such as the lead silicate and barium silicate samples studied, are attractive for low-energy manufacturing. However, for structural applications that require higher strength, the barium silicate option is more promising because of its higher E modulus and less scratch sensitivity. 26

For more uneven samples, made of contaminated cullet at a temperature near the liquidus point, they still have good bending strength and are suitable for structural applications that require lower tensile strength, such as bricks. During the four-point bending test, the defects in the body will not be activated, and the contribution to the strength and E modulus is small or even negligible. However, an increase in the density of defects in the volume also means an increase in the density of surface defects, which will result in a decrease in the average strength. A higher molding temperature (above the liquidus point) will significantly help reduce the number of defects, but considering the economic and environmental advantages of low-temperature processing, only when higher design strength is required in each specific situation, this This kind of behavior is meaningful.

Crystalline geometry is produced in soda wollastonite and borosilicate samples produced at higher viscosities (106–105 dPa s). If these structures are in a block, the flexural strength of the specimen is equal to the flexural strength of a more uniform casting at a higher temperature (close to the liquidus point). However, depending on the nature of the resulting crystalline structure, exposing this structure to a surface under tension can cause a sharp drop in strength of even 75%. In this case, the starting point of fracture always occurs at the glass/crystal interface. Special attention should be paid to castings formed at a viscosity of 105-104 dPa s, because the glass product may appear to be homogeneous, but the front interface formed between each broken glass fragment still retains significant tiny Inhomogeneous areas of bubbles and stones. This kind of stratum exposed on the tensile surface is critical to the strength of the specimen.

The glass samples manufactured by industrial SLS are post-processed in the laboratory facility to match the sample size under study, and have similar bending strength to float glass furnace cast samples (1120°C), but the lower limit in the strength value report Scoring literature. Processing defects from processing to size, as well as insufficient annealing of cast bricks, are factors that cause lower strength. Not only will finer polishing significantly increase the strength of these samples, it will also significantly increase the strength of purer kiln-cast samples. However, considering that the lowest strength samples are less affected by the finer polishing quality, the statistical strength will not increase that much because it is dominated by these lower outliers.

26 Yamane and Mackenzie (1974) proved the proportional relationship between Vickers hardness and Young's modulus and bond strength in their model. For reference, Ainsworth (1954) measured the Vickers hardness of 18Na20·10BaO·72SiO2 (mol%) glass at 522 kg/mm2, and measured the Vickers hardness of 18Na20·10PbO·72SiO2 (mol %) glass at 445 kg /mm2.

The results of this study show the potential for recycling waste glass into cast structural building components. However, in order to safely apply such products, further verification is required, and the number of test samples for each category (≥ 30, Quinn et al. 2009) needs to be increased to obtain statistical predictions. In particular, in the case of contaminated samples, repeated testing is critical, as a higher degree of variability in mechanical properties is expected. System testing of such samples should be associated with quantitative documentation of the type and level of non-uniformity in the glass before testing. It is also necessary to conduct a careful and extensive fracture analysis of the test samples to determine the most critical defects and the relationship between defect size and flexural strength.

Physical and chemical identification of the crystalline structure of the glass surface through scanning electron microscopy is needed to classify such critical defects. Further testing is needed to determine the effect of the scale factor and the effect of static fatigue in a humid environment (the effect of slow crack growth). In addition, studying the behavior of crystalline inclusions in bulk glass under thermal gradients associated with architectural applications is important to eliminate the risk of thermal cracking. In addition, it is also recommended to use pulse excitation technology for non-destructive testing to determine the Young's modulus and the degree of unevenness of the cast glass. It is worth exploring whether this fast and cheap non-destructive technology can be used as a quality control method for cast glass products.

For the more polluting ingredients, attention should be paid to improving the quality of the stone-containing surface of the recycled glass. Chemical strengthening of the surface through ion exchange may be a costly solution, although it is unlikely to help resolve deep defects. Another simpler solution for high-viscosity castings (low diffusivity) is to build two (compatible) cullet qualities in the mold: purer along the more demanding areas, and the degree of contamination in the block Lower quality (Figure 1). 44). This kind of composite glass can use contaminated, unwanted cullet without compromising the strength of the final product.

Finally, the engineering of the crystal or bubble cap geometry within the glass is worthy of further exploration, as they can cause fractures in predictable strength ranges and locations. They may also cause building components to have a non-standard appearance and thus have a higher architectural appeal.

Acknowledgements The author would like to thank Giorgos Stamoulis for his significant contributions in preparing the four-point bending experiment and DIC measurement, as well as Kees van Beek for his guidance. We are especially grateful to George Quinn and James Varner for their valuable comments on the experimental process and fracture analysis of the results. Thank you very much for the feedback received from Bert Sluijs, Mauro Overend, Christian Louter and Karl-Heinz Wolf. We would also like to thank Ruud Hendrikx for XRF and XRD analysis, Mariska van der Velden for assistance in sample preparation, Sander van Asperen and Wolfgang Gard for providing Keyence digital microscopes in their laboratory, and Henning Katte and Daniel Schreinert (Ilis) for sponsoring StrainScope Flex use. The author is also very grateful to Cor Wittekoek (Vlakglasrecycling), Danny Timmers (Maltha Glasrecycling Nederland), Marco Zaccaria and François Boland (AGC Belgium), Brian Wittekoek (Coolrec), Bettina Sommer (Royal Leerdam Crystal) and Klaas Roelfsema (Schott The contribution of debris, which is critical to this work. Finally, we would like to thank Erik Muijsenberg (Glass Service) for the book "Glass Defects" (Bartuška 2008), and Peter de Haan (AGR Delft) for the "Color Atlas of Container Defects" (Aldinger and de Haan 2019) and "Glass Service" The "Color Atlas of Stones" (Aldinger and Collins 2016) book is very helpful in the process of defect classification and identification.

Conflict of interest The corresponding author declares that there is no conflict of interest on behalf of all authors.

Abrisa Technologies: SCHOTT Duran® Laboratory Glass (Tube) (2014) Ainsworth, L.: The relationship between the hardness of the diamond pyramid of glass and the strength and structure of the glass. society. Glass technology. 38, 501–547 (1954) Aldinger, BS, Collins, BK: A collection of colors of gems in glass. American Glass Research, Butler (2016) Aldinger, BS, de Haan, PW: A color map of defects in glass containers. American Glass Research, Butler (2019) Andreola, F., Barbieri, L., Corradi, A., Lancellotti, I., Falcone, R., Hreglich, S.: Glass-ceramic ray tube glasses obtained by recycling scrap cathodes . Waste management. 25, 183–189 (2005). https://doi.org/10.1016/j.wasman.2004.12.007 ASM International: Application of the desktop version of the Glasses Engineering Materials Manual. ASM International (1995) Bartuška, M.: Glass defects. Glass Service Inc. and Práh, Prague (2008) Brennan, JJ: Research project of silicon carbide fiber reinforced glass matrix composites. Joint Technology Research Center, Connecticut (1979) Bristogianni, T., Oikonomopoulou, F., Justino de Lima, CL, Veer, FA, Nijsse, R.: Structural cast glass parts made from waste glass: from landfill Transfer the daily discarded glass to the construction industry. Heron 63 (1/2 Special issue: Structural Glass) (2018) Bristogianni, T., Oikonomopoulou, F., Veer, FA, Nijsse, R.: The effect of manufacturing defects in the microstructure of cast glass on the structure. In: Zingoni, A. (ed.) Advances in Engineering Materials, Structures, and Systems: Innovation, Mechanics, and Applications, pp. 1703-1708. CRC Press, Leiden (2019) Burch, OG, Babcock, CL: The effect of glass color on the solidification rate in glass bottle manufacturing. J. Morning ceramics. society. 21(10), 345–351 (1938) Campbell, DE, Hagy, HE: glass and glass-ceramics. In: Lynch, CT (Editor) CRC Handbook of Materials Science, Volume 1. Two: Material composite materials and refractory materials. CRC Press, USA (1975) Chyung, K.: Transparent β-quartz glass ceramics. Corning Glass Works, US Patent US4018612A (1977) Corning: Characteristics of Corning glass and glass ceramic series. In: Materials for design engineers. United States (1979) Fluegel, A.: Glass viscosity calculation based on global statistical modeling method. Glass technology. EUR. J. Glass Science. technology. A 48, 13–30 (2007a) Fluegel, A.: A global model for calculating room temperature glass density based on composition. J. Morning ceramics. society. 90, 2622–2625 (2007b). https://doi.org/10.1111/j.1551-2916.2007.01751.x Friedrich & Dimmock Inc.: Comparative value of borosilicate glass. In: Simax Glass Properties Gold Star: Investment Casting Powder Safety Data Sheet, www.goldstarpowders.com (2019) Goodwin Refractory Services Ltd: Crystalcast (M248). www.grscastingpowders.com (2003) Gregory, C.: Hollow fiber. See: Sanghera, JSA, ID (ed.) Infrared Fiber Optics. CRC Press LLC, USA (1998) Heimerl, W.: Chemical resistance and corrosion, and ion release. In: Bach, H., Krause, D. (eds.) Composition and structure analysis of glass and glass ceramics. Springer, Berlin (1999) Höland, W., Beall, GH: Glass-Ceramic Technology, 3rd edn. Wiley, Hoboken (2020) Inaba, S., Fujino, S., Morinaga, K.: Young's modulus and composition parameters of oxide glass. J. Morning ceramics. society. 82, 3501–3507 (1999). https://doi.org/10.1111/j.1151-2916.1999.tb02272.x Ito, S., Taniguchi, T.: The effect of cooling rate on glass structure and mechanical behavior, MD simulation. J. Amorphous. Solid 349, 173–179 (2004). https://doi.org/10.1016/j.jnoncrysol.2004.08.180 Kita˘ıgorodski˘ı, II, Solomin, NW: The freezing rate of glass during work. society. Journal of Glass Technology 18, 323–335 (1934) Makishima, A., Mackenzie, JD: Direct calculation of the Young's modulus of glass. J. Amorphous. Solid 12(1), 35–45 (1973). https://doi.org/10.1016/0022-3093(73)90053-7 Martienssen, W., Warlimont, H.: Springer Condensed Matter and Material Data Book (2005) MatWeb: C-glass fiber. www.matweb.com Montazerian, M., Singh, SP, Zanotto, E.: Glass ceramic research and commercialization analysis. Yes. ceramics. society. bull. 94, 30–35 (2015) Morey, GW: The effect of boron oxide on the devitrification of soda lime silica glass Na2O–CaO–B2O3–SiO21. Quaternary ammonium salt. system. 15(9), 457–475 (1932). https://doi.org/10.1111/j.1151-2916.1932.tb13959.x Mueller, J., Boehm, M., Drummond, C.: CRT waste glass processing direction: electronic recycling industry exchanges. Waste management. 32, 1560–1565 (2012). https://doi.org/10.1016/j.wasman.2012.03.004 National Institutes of Health (NIH): PubChem database. https://pubchem.ncbi.nlm.nih.gov/ Oikonomopoulou, F., Bristogianni, T., Veer, FA, Nijsse, R.: The construction of the outer wall of the Crystal House: challenges and innovations. Glass structure. Britain. 3(1), 87–108 (2018a). https://doi.org/10.1007/s40940-017-0039-4 Oikonomopoulou, F., Bristogianni, T., Barou, L., Jacobs, E., Frigo, G., Veer, FA, Nijsse, R. : Interlocking cast glass components, exploring a detachable dry assembly structure glass system. Heron 63, 103–138 (2018b) Oikonomopoulou, F., Bristogianni, T., Barou, L., Veer, FA, Nijsse, R.: The potential of cast glass in structural applications. Lessons learned from the most advanced load-bearing cast glass in large castings and buildings. J. Construction. Britain. 20, 213–234 (2018c). https://doi.org/10.1016/j.jobe.2018.07.014 Priven, AI: Evaluate the score of four-fold coordinated boron in oxide glass from its composition. Glass Physics Chemistry 26(5), 441–454 (2000). https://doi.org/10.1007/BF02732065 Quinn, GD: NIST Recommended Practice Guide: Fracture of Ceramics and Glass, Second Edition. (2016) Quinn, GD, Swab, JJ: Glass fracture toughness measured by SCF and SEPB methods. J. Euro. ceramics. society. 37(14), 4243–4257 (2017). https://doi.org/10.1016/j.jeurceramsoc.2017.05.012 Quinn, GD, Ives, LK, Jahanmir, S.: Regarding the nature of processing cracks in ground ceramics: Part 2-Comparing with other silicon nitride and Comparison of damage diagrams. Mach. science. technology. 9, 211–237 (2005). https://doi.org/10.1081/MST-200059051 Quinn, GD, Sparenberg, BT, Koshy, P., Ives, LK, Jahanmir, S., Arola, DD: The bending strength of ceramic and glass rods. J. Test. Evaluate. 37, 222–244 (2009). https://doi.org/10.1520/JTE101649 Schott: Tubular glass photobioreactor (2015a) Schott: NEXTREMA® (2015b) Schott: DURAN technical data (2017) Sehgal, J., Ito, S.: A new kind The low brittle glass belongs to the soda lime silica glass series. 81(9), 2485-2488 (1998). https://doi.org/10.1111/j.1151-2916.1998.tb02649.x Shelby, JE: Introduction to Glass Science and Technology. Royal Society of Chemistry (2005) Silva, RV, de Brito, J., Lye, CQ, Dhir, RK: The role of glass waste in ceramic production and other applications: a review. J. Clean. product. 167, 346–364 (2017). https://doi.org/10.1016/j.jclepro.2017.08.185 Songhan Plastic Technology Co., L.: Schott Nextrematextcircled 724-3, 712-3 Glass ceramic special glass products: STARPHIRE Ultra-ClearTM soda lime glass Thompson, DA: Low liquidus glass for TV tube panels. US Patent US4331770A (1980) Veer, FA: Glass strength, non-transparent value. Heron 52, 87–104 (2007) Veer, FA, Rodichev, Y.: Structural strength of glass: hidden damage. Strength alma mater. 43, 302–315 (2011). https://doi.org/10.1007/s11223-011-9298-5 Volf, MB: The chemical method of glass. Elsevier, Amsterdam (1984) Yamane, M., Mackenzie, JD: Vickers hardness of glass. J. Amorphous. Solid 15(2), 153–164 (1974). https://doi.org/10.1016/0022-3093(74)90044-1 Yankelevsky, D., Spiller, K., Packer, J., Seica, M.: Fracture of soda lime glass sample tested in laboratory characteristic. can. J. Wenmingying. (2016). https://doi.org/10.1139/cjce-2016-0374 Yun, YH, Bray, PJ: NMR study of glass in Na2O–B2O3–SiO2 system. J. Amorphous. Solid 27(3), 363–380 (1978). https://doi.org/10.1016/0022-3093(78)90020-0 Zhdanov, SP, Shmidel', G.: Coordination state of boron in sodium borosilicate glass from NMR data. Fitz. Chin. Stokla 1(5), 452–456 (1975)

No. 61, Haier Road, Laoshan District, Qingdao City, Shandong Province, Post Code 266061

19, rue du Puits Romain, L-8070 8070 Bertrange Luxembourg

carat. Km from Estacion. 15.8 44415 Rubielos de Mora Teruel Spain

Ismail Horse. D-100 Karayolu CAD. No:44A, 34947 Tuzla, /İstanbul Turkey

Im Vogelsang 18 3340 Waidhofen/Ybbs Austria

Log in or register to post a comment