Glass Property Predictions and the Design and Characterization of Leucite Glass-Ceramics

Abstract

The aim was to optimize a series of aluminosilicate glasses for the synthesis of leucite glass-ceramics (GC) for dental applications. Appen predictive models were used to design a series of aluminosilicate glasses to control optical, thermal and mechanical properties. Glasses were produced using melt quenching methods, annealed and processed into powders and further heat-treated and milled to produce GC powders. Glasses/GCs were characterized using dilatometry, HTXRD, SEM and ²⁷Al MAS-NMR and GCs tested using biaxial flexural strength (BFS) for comparison to commercial leucite GC products. The results indicated good prediction to experimental measurement correlations (for coefficient of thermal expansion, refractive index and density) and provided evidence of leucite GCs’ optimization compared to commercial products. This included significant BFS and Weibull m increases, improved microstructural control and designed translucency, meeting the demands for strong, esthetic and durable single-tooth restorations. The simple predictive approach, combined with complementary characterization techniques, allowed structure–property relations of aluminosilicate glasses/glass-ceramics to be understood, and may find similar applications in other glass systems beyond dentistry.

1. Introduction

Glass properties are generally divided into two principal categories: those determined by structure and those dependent on composition [1]. Predictive models offer a practical route to identify glasses with targeted properties [2]. According to the Functional Glass Manufacturing Innovation Consortium (FGMIC) roadmap, glasses with structures that are engineered or compositionally tailored are termed “functional”. Performance and technology enhancement are the main goals for innovation in functional glass manufacturing, so new compositions with improved process control and tailored properties are developed [3].

Since the pioneering work of Winkelmann and Schott [4], a series of additive methods has been devised [5,6,7,8] that relate oxide glass composition to a property α, where

$$a={\sum }_i{\alpha }_i{P}_{i } ,$$

(1)

where αᵢ is a numerical constant characteristic of each oxide constituent and Pᵢ is the weight or mole fraction of the oxide added. This method is built on the thermodynamic concept of a glass as a solution of components, each contributing to the property of interest in proportion to its molar fraction. The function of composition and the property of interest are not always linearly related, requiring corrections for multi-component systems [2,9,10]. Relatively accurate predictions can, however, be achieved [9] for coefficient of thermal expansion (CTE) and refractive index (R.I.) using this method. Modifications have extended the use of this method for R.I. in bioactive glass formulations [11]. Various predictive models have also been suggested for glass development applications in engineering and optical fields [12,13].

After designing new materials, it is also important to understand their structure–property relationships. Solid-state magic angle spinning nuclear magnetic resonance (MAS-NMR) is essential in the characterization and understanding of changes in glass/glass-ceramic structure during experimentation [14]. When used in conjunction with high-temperature X-ray diffraction (HTXRD) techniques, useful insights are provided to make further glass design changes. The effects of various oxide substitutions on the aluminosilicate glass network and their subsequent crystallization can then be carefully mapped and understood. Aluminosilicate glasses with a broad range oxide content are versatile for tailored designing of a wide range of engineering applications [15]. In dentistry, aluminosilicate glasses are used to produce leucite glass-ceramics (GC) for various dental prosthesis applications (e.g., monolithic crowns, veneering of metal alloys or polycrystalline substrates for crowns and bridges) [16,17,18]. For these applications, specific tailored glass properties are required to achieve glass-ceramics with high strength, translucency and low solubility. The study aims to apply a predictive model to optimize a series of aluminosilicate glasses/glass-ceramics, together with advanced characterization techniques to understand and control their structure–property relations.

2. Materials and Methods

2.1. Glass Design

Glass compositions (

Table 1. Aluminosilicate glass compositions (mole %).

	SiO2	Al2O3	K2O	CaO	TiO2	Na2O	Li2O	MgO
Glass A	72.6	10.7	7.9	2.1	0.3	4.7	1.1	0.5
Glass B	70.9	10.0	10.9	2.1	0.5	4.0	1.1	0.5
Glass C	69.3	10.1	12.5	2.1	0.5	4.0	1.1	0.5

) were formulated using Appen factors (Appendix A,

Table A1. Appen factors for glass properties prediction used in the study.

Oxides	Thermal Expansion Coefficient (α) Appen Factors (×10−6/K, 20–400 °C)
SiO2	0.5 to 3.8
Al2O3	−3.0
K2O	46.5 (42.0)
CaO	13.0
TiO2	−1.5 to −3.0
Na2O	39.5
Li2O	27.0
MgO	6.0
The following remarks apply for specific Appen Factors: SiO2: α SiO2 = 10.5 − 0.1· p SiO2 when 67 ≤ p SiO2 ≤ 100, α SiO2 = 3.8, K2O when p SiO2 ≤ 67 for glass compositions where p Na2O < 1%, only the values in parentheses apply for TiO2: α TiO2 = 10.5 − 0.15· p SiO2 when 50 ≤ p SiO2 ≤ 80.

) to estimate the CTE and R.I. Network connectivity and density were also calculated for each composition. The predicted properties were determined using the Appen additive approach [7] as expressed in Equation (1), where αi is the characteristic Appen factor of the i-th oxide and Pi is its mole fraction in the glass. Contributions of each oxide to the fusion temperature were calculated to estimate processing conditions [4]. Aluminosilicate glasses were designed by adjusting compositions around the stoichiometry of leucite. A range of glass CTE was designed to promote tangential compression of the leucite phase in the GC to enhance flexural strength. The R.I. was designed to match the R.I. of leucite (1.5, [19]) encouraging translucent GCs. The fusion temperature was estimated to allow GC processing in commercial dental furnaces (<1200 °C) used to produce dental restorations.

2.2. Glass Synthesis and Processing

Glass batch A (Table 1) was heated inside an electric kiln (Fredrickson Kiln Co., New York, NY, USA) at 10 °C/min to 1316 °C with a 7 h hold and the furnace cooled. The glass frit was crushed, wet ground and screened to 125 μm [17]. Glass batches B and C (Table 1) were heated in a Pt/Rh (90/10) crucible, inside an electric furnace (UAF1600 furnace, Lenton, Hope Valley, UK) at a rate of 10 °C/min to 1550 °C and held for 5 h. The crucible with the molten glass was then transferred to a preheated furnace (Tris Burnout furnace, Dentalfarm, Turin, Italy) and annealed at 500 °C for 2 h, followed by furnace cooling. Glass C was re-homogenized after crushing and ball milling for 2 h (Pascal Engineering Ltd., London, UK) using alumina grinding media. The powder was remelted for 2 h at 1550 °C in a UAF1600 furnace (Lenton, UK), followed by air quenching (1 min). The glass was then annealed again (as before). The glass frits were crushed, ball milled (2 h) and sieved (Endecotts Ltd., London, UK) to obtain a particle size of ≤125 μm.

2.3. High-Temperature X-Ray Diffraction

Experimental glass powders (A, B, C) were screened to 38 µm, and high-temperature X-ray diffraction (HTXRD) was carried out. Powders were placed on a platinum strip within a high-temperature furnace (Anton Parr HTK 16, Anton Parr, Graz, Austria), equipped with precise temperature control (Anton Parr TCU 2000) and fitted to an X-ray diffractometer (X’Pert Pro, Panalytical, Almelo, The Netherlands). Measurements were performed using flat plate θ/2θ geometry, together with Ni-filtered Cu-Kα radiation (λ₁ = 1.540598 Å, λ₂ = 1.544426 Å). An initial X-ray diffraction (XRD) pattern was recorded at 25 °C, followed by data collection from 500 °C to 1100 °C in 20 °C increments. For each temperature, a short scan of 15 min was conducted over a 2θ range of 10° to 80°, with a step size of 0.0167° and a step time of 26.67 s. A 5 min hold was applied at each temperature to allow stabilization and equilibration of the domain structure before data acquisition. A National Institute of Standards (NIST) standard reference material (SRM, 660a, LaB₆, lanthanum hexaboride) was used for system calibration. The structural models (International Centre for Diffraction PDF-4 database) of tetragonal leucite (ICDD: 00-038-1423), cubic leucite (ICDD: 01-076-2298) and sanidine (ICDD: 00-025-0618) were used for phase identification.

2.4. Glass-Ceramic Powder Production

Portions of the screened (to 125 μm) glass powders (A, B, C) were weighed (50.0 g), placed on custom-made refractory trays and crystallized using two-step heat treatment parameters listed in

Table 2. Two-step crystallization heat treatments of the glasses.

Experimental Glasses	Nucleation Temp. [°C]	Nucleation Hold [h]	Crystallization Temp. [°C]	Crystallization Hold [h]	Heating Rate [°C/min]
Glass A	650	1	1120	1	10
Glass B	620	1	920	1	10
Glass C	620	1	795	1	20

, followed by furnace withdrawal and air-quenching. Glass-ceramic frits (A, B and C) were then crushed, ball milled and screened to 125 µm.

2.5. Solid State MAS-NMR

Experimental glasses and glass-ceramic powders (A, B, C), along with commercial glass-ceramic powders; IPS Empress Esthetic (shade ETC2, Ivoclar-Vivadent, Schaan, Liechtenstein) and Ceramco 3 Dentine (Batch No. 02111576, Dentsply Ceramco, Burlington Township, NJ, USA) were characterized using solid-state ²⁷Al Magic Angle Spinning Nuclear Magnetic Resonance (MAS-NMR). The experiments were performed on Bruker Avance III HD 800 MHz and Avance NEO 600 MHz NMR spectrometers (Bruker, Ettlingen, Germany), corresponding to magnetic fields of 18.8 T and 14.1 T. A short pulse length (0.5 µs) and a recycle delay of 0.5 s were employed for the ²⁷Al MAS-NMR measurements. The powder samples were packed into zirconia rotors (2.5 mm outer diameter) and spun at 24 kHz during measurements at a resonance frequency of 208.5 MHz. For ²⁷Al triple quantum (3Q) MAS experiment, a three-pulse program with split-t1 full echo acquisition was used. A 4.5 µs pulse was used for the 3Q excitation, followed by a conversion pulse of 1.2 µs and 38 µs of a 180 degree selective pulse. In total, 69 t1 experiments were acquired with rotor synchronized increments, with 2400 scans and 1 µs recycle delay within each and 48 dummy scans. In addition, commercial GC powder samples were run versus an experimental GC at resonance frequency of 156.4 MHz under MAS condition of 12 kHz using 4 mm zirconia rotors. All ²⁷Al MAS-NMR spectra were referenced using a 1 M solution of Al(NO₃)₃.

2.6. Differential Thermal Expansion Analysis

The CTE of cut glass-frit blocks and sintered GC specimens (6 mm depth × 6 mm height × 25 mm length, n = 1) for all experimental glasses and GCs were measured using a differential dilatometer (DIL 402PC, Netzsch Instruments, Selb, Germany). The temperature range was between 25 °C and 1200 °C, at a heating rate of 3 °C/min. A dedicated, standard alumina rod specimen was run before each measurement for device calibration. The subsequent data correction was carried out using Netzsch Proteous thermal analyses software (ver. 4.8.2, 2006, Netzsch Instruments, Selb, Germany). The alumina rod was run in a nitrogen gas atmosphere and using identical temperature range/heating rate parameters as the test specimens. CTE calculations took place between 100 and 400 °C as the plots were linear at that range for all specimens tested. The agreement between the predicted and measured CTE values for all experimental glasses (A–C), was evaluated using a chi-square goodness-of-fit test and the root mean squared error (RMSE).

2.7. Glass Density and Refractive Index Measurement

The density of the glass-frit specimens was determined for all experimental glasses based on Archimedes’ principle, using an analytical balance (Mettler AE 200, Mettler Instruments Ltd., Leicester, UK) equipped with a density determination kit (Mettler ME-33360, Mettler Instruments Ltd., Leicester, UK). The reference liquid was distilled water. The mass of each glass-frit specimen (n = 10 per test group) was measured in air and then in distilled water, with a precision of 0.1 mg. Following the procedure specified by the manufacturer, the specimen density ($\rho$) was calculated using the following relation:

$$\rho ={\displaystyle \frac{A}{A-B}}\ast {\rho }_0,$$

(2)

where A is the specimen weight in air, B is its apparent weight (buoyant force) when immersed in distilled water and ${\rho }_0$ is the density of distilled water at the measurement temperature.

The Becke line method was used to measure the refractive index (R.I.) of the experimental glasses (n = 3 per test group). Glass powders were placed between glass slides after adding a series of refractive index oils and viewed using a light microscope (Olympus Bx60, Olympus Optical Co., Southend-on-Sea, UK). The R.I. was found when the particular refractive index oil was matched with the glass grains and the Becke line was not visible.

2.8. Glass-Ceramic Specimen Preparation

Experimental GC (A, B, C) powders (125 μm) were used to fabricate 30 disk specimens per test group. Specimens were fabricated by moistening 1.1 g of powder with 0.4 mL of modeling liquid (CH B:24066, Vita Zahnfabrik, Bad Sackingen, Germany), followed by tissue condensation and hydraulic GC slurry pressing (3 × 10⁵ Pa, 1 min). The experimental disk specimens were transferred to a pre-heated (538 °C) porcelain furnace (Multimatt MCII, Dentsply, Konstanz, Germany) and sintered under vacuum at a rate of 38 °C/min to 1040 °C with 2 min hold for GC A, to 1180 °C with 2 min hold for GC-B and to 1100 °C with 2 min hold for GC C. A 3 min cooling cycle was applied at the end of the firings. All Specimens were wet ground to P1000 grade SiC paper. GC-A specimens were prepared in a previous study [17].

2.9. Biaxial Flexural Strength Testing

Thirty disk specimens (14 mm diameter and 2 mm depth) per experimental GC group were tested using the biaxial flexural strength test (ball on ring). GC-A was identically tested in a previous study [17]. Specimens were centrally loaded on a 10 mm diameter knife-edge support via a 4 mm diameter spherical ball indenter. A thin plastic sheet (0.03 mm) was positioned between the specimen surface and the indenter to evenly distribute the load. The load was delivered by an Instron universal testing machine (5567/H1580, Instron, Buckinghamshire, UK) at a crosshead speed of 1 mm/min. The biaxial flexural strength was calculated using the Timoshenko and Woinowsky–Krieger Equation [20], and a Poisson’s ratio of 0.25. Strength data were compared using a one-way ANOVA (SigmaStat, version 2.03, SPSS Inc., Chicago, IL, USA) and Tukey’s multiple comparison test (p < 0.05). The strength data were plotted using Weibull regression analysis with accuracy of fit assessed using a 90% critical correlation coefficient [21]. Statistical differences between Weibull modulus and characteristic strength for each test group were compared by the overlap of double-sided 95% confidence intervals.

2.10. Secondary Electron Imaging

Fractured GC BFS specimens and GC specimens after crystallization heat treatments were polished and etched (0.1% hydrofluoric acid, 60 s) and gold-coated (Agar Auto Sputter Coater, Agar scientific Ltd., Stansted, UK) and viewed using secondary electron imaging (10 kV accelerating voltage, 10 mm working distance), with a field emission scanning electron microscope (SEM) (FEI Inspect F, Oxford Instruments, Oxfordshire, UK). SEM Photomicrographs used for quantitative crystal size analysis were scanned using a pen tablet (CTL-460, Wacom Co., Ltd., Beijing, China) in combination with image analysis software (Sigma Scan Pro 5.0, Systat Software, Inc., Chicago, IL, USA). The leucite crystal size and area fraction were measured by manually drawing around each leucite crystal. Care was taken to ensure that the photomicrographs selected for measurements were representative of the overall crystalline distribution in each GC specimen explored.

3. Results

3.1. Appen-Predicted and Measured Values

Appen-predicted and measured CTE, Density and R.I. values for the selected experimental glasses can be seen in

Table 3. Predicted and measured property values for the experimental glasses.

Glasses	CTE [10−6/K, 100–400 °C]		Density [g/cm3]		Refractive Index (R.I.)		Predicted Glass Network Connectivity
	Predicted	Measured	Predicted	Measured	Predicted	Measured
Glass A	8.6	8.5 *	2.381	2.408 (0.014)	1.50	1.50	3.90
Glass B	9.7	9.8	2.397	2.418 (0.006)	1.51	1.51	3.87
Glass C	10.4	10.1	2.410	2.434 (0.006)	1.51	1.50	3.88

* Data taken from [17].

. Predicted glass network connectivity values are also presented (Table 3). Predicted and measured experimental glass CTE values were correlated (r² = 0.95, Appendix A, Figure A1). A root mean squared error of 0.19 × 10⁻⁶/K was calculated. The χ² test indicated no significant difference between measured and predicted CTE values (p < 0.05), since the critical χ² value was larger than the calculated at the 95% significance level (Degrees of freedom: 2).

In summary, the Appen-predicted properties showed a good agreement to the experimentally measured values.

3.2. Dilatometry Results

The CTE values for the experimental glasses are summarized in Table 3. The corresponding CTE plots are in Appendix A (Figure A2). The measured CTE (10⁻⁶ K⁻¹, 100–400 °C) for the experimental glass-ceramics were 18.6 for GC B and 21.0 for GC C. The CTE of GC A was previously reported as 19.0 × 10⁻⁶ K⁻¹ [17].

3.3. BFS Results

The Mean BFS (SD) and Weibull regression analysis results for the sintered GCs are shown in

Table 4. BFS Results for the Glass-Ceramics.

Glass—Ceramics	BFS
	Mean (SD) BFS [MPa]	Weibull m	C.I. for m (95%)	Characteristic Strength σ0 [MPa]	C.I. for σ0 (95%)
A	153.2 (21.7) a*	8.5 a	6.9–10.4	162.0 a	156.0–168.2
B	131.7 (9.0) b	17.6 b	14.1–21.9	135.6 b	133.2–138.1
C	212.3 (28.2) c	8.5 a	6.7–10.9	224.4 c	216.2–232.9

Different superscript letters indicate significant differences (p < 0.05) between groups.* Data taken from [17].

. There was a significant difference in Mean BFS between all the test groups (p < 0.05). All the GC test groups had higher Mean BFS (p < 0.05) than a previously tested commercial GC, Ceramco 3 (Mean (SD) BFS: 75.7 (6.8) MPa) [22]). There was a significant difference in characteristic strength between all the test groups according to the overlap of their double-sided confidence intervals. GC-B had significantly (p < 0.05) higher Weibull modulus than GC-A and GC-C.

In summary, there was an increased Mean BFS for all the designed novel leucite GCs compared to a commercial leucite GC.

3.4. HTXRD Results

The HTXRD plots for glasses A and B with identified Sanidine and Leucite phases can be seen in Figure 1a and Figure 2a,b. Inset Tables in Figure 1b and Figure 2c summarize major (M) and minor (m) cubic leucite and sanidine phases and temperatures with the co-existence of both phases. Cubic leucite was the only phase identified in glass C HTXRD (Figure 2b).

3.5. SEM Results

SEM photomicrographs of commercial GC (Ceramco 3) and experimental GCs (A, B and C) can be seen in Figure 3a–f. The results of the quantitative image analysis of the SEM photomicrographs are in

Table 5. Quantitative image analysis results for the experimental glass-ceramics.

Glass—Ceramics	Mean (SD) Crystal Size [μm2]	Leucite Area Fraction (%)
A *	0.99 (0.59)	25.0
B	1.45 (2.22)	54.2
C	0.57 (0.55)	50.0

* Data taken from [17]

. These values contrast with the larger Mean (SD) leucite crystal size (10.15 (30.63) μm²) of the Ceramco-3 GC specimen, with a 30.2% area fraction [22]. Extensive crystal and glassy matrix microcracking were also present (Figure 3a). GC-A (Figure 3b) presents spherical leucite crystals with signs of within-crystal microcracking only. A mixed morphology of spherical and fiber-like leucite crystals (Figure 3c), with some orientated domains (Figure 3d) was found for GC-B. The GC-C specimen displayed a highly crystallized leucite microstructure, containing fine spheres and high aspect ratio rods in orientated domains (Figure 3e), and areas of increased interconnectivity (Figure 3f).

In summary, various leucite crystal sizes, morphologies and area fractions were demonstrated in the novel glass-ceramics, which may have differing technical/clinical applications.

3.6. Solid State MAS-NMR Results

For the glasses, the ²⁷Al MAS-NMR spectra exhibit a narrow line at the center of gravity at 57.4–57.9 ppm. The position of this line indicates that the specimens contain four-coordinated Al(IV) species in the glasses (Figure 4a). The GC spectra (Figure 4b) indicated resonances around 59.5, 67.2 and a third feature at 62 ppm, which was clearer in GC-B and C. These three features were thought to be related to T sites in the framework of the leucite phase [23]. The ²⁷Al 3Q MAS-NMR spectrum of the GC-C (Figure 4c) illustrates the presence of three sites as indicated. Leucite-based commercial controls showed similar ²⁷Al MAS-NMR spectra in Figure 4d. Based on the comparison with the experimental GC-C, the longer low-frequency tail of the ²⁷Al line shape seen in the commercial comparison materials (Figure 4d) indicates the presence of the tetrahedral amorphous aluminum species Al(IV) in the glassy matrix [24]. In summary, ²⁷Al MAS-NMR indicated the structural differences between glasses and GCs as well as the relevant T sites in the framework of the leucite phase.

4. Discussion

Appen glass predictions used in the current paper (Table 3) were within 0.3 × 10⁻⁶/K of the actual values for CTE, 0.02–0.03 g/cm³ for density and 0.01 for refractive index, making it useful for the design of aluminosilicate glasses for dentistry. This accuracy is in agreement with other studies [25]. In fact, the predicted and measured CTE values were highly correlated (r² = 0.95, Appendix A, Figure A1), with no significant differences (p > 0.05). The higher CTE found in the glass C (10.4 × 10⁻⁶/K) and GC-C (21.0 × 10⁻⁶/K) may be due to its higher K₂O content (Table 1). This oxide makes a bigger contribution to the CTE and is essential to the formation of a higher area fraction (50%) of the leucite phase (KAlSi₂O₆), which has a high CTE (28 × 10⁻⁶ °C⁻¹) [26]. The glasses were designed systematically using Appen factors to match the CTE with the leucite phase. This benefits the generation of tangential compressive stresses within the glassy matrix and reduces strength limiting matrix microcracking [27]. SEM photomicrographs of glass-ceramics A-C indicated a lack of matrix microcracking (Figure 3b–f) compared with Ceramco 3 GC (Figure 3a), and a statistically higher BFS for the experimental GCs. Commercial GCs have been linked to lower CTE (Ceramco 3 CTE = 12.8⁻⁶ K⁻¹, 100–400 °C), matrix microcracking and the relief of beneficial residual stresses [22]. The BFS fracture specimens (Appendix A, Figure A3a–c) presented increasingly rougher intergranular fracture surfaces (from GC-A to C). Vacuum sintering of GC-B and C did lead to some microstructural coarsening, but with mixed morphology retained in GC-C (Appendix A, Figure A6a). These microstructural features initiating crack pinning, deflection and diversion due to favorable matrix compressive stresses [28]. Appen factors were previously used following EDX of the residual glass to estimate matrix glass network connectivity and thermal relationships between the glass and crystal phase in high strength leucite GCs [24]. GC-C demonstrated local changes in crystal microstructure where residual glass interfacial chemistry may differ to that of GC-A. The resultant residual glass strains may therefore influence fracture progression. This is illustrated in the tortuous fracture interaction with the twinned leucite crystals and residual glass (Appendix A, Figure A6b). Analytical methods were used to estimate the magnitude of residual stresses in lithium-alumina-silicate GCs and their influence on the generation of high BFS [25].

Previous studies regarding the optimization of leucite glass-ceramics relied on powder processing to successfully increase the BFS and reliability [17,22]. Although highly effective, powder processing techniques add additional manufacturing cost and the potential for contamination due to milling duration and media used. The ability to design and predict glass properties and thermally match with the required crystal phase will reduce processing costs and increased the flexural strength (>200 MPa) in GC-C. All formulations gave a significant flexural strength increase compared to the commercial GC (Ceramco 3). This property is useful to resist brittle fracture in ceramic restorations [29]. Typical values of Weibull modulus for ceramic materials (m = 5–15) were exceeded for CG-B (m = 17.6), indicating greater structural reliability. This allows the possibility to design dental prostheses with the potential for reduced clinical failure [30]. Control of crystalline morphology, volume fraction, chemical composition of the residual glass and physical properties, will also affect GC wear properties. This has importance in the preservation of tooth structure during function [31]. Ceramic microstructures and mechanical properties can be adjusted to suit clinical usage. High volume fraction and high aspect ratio crystal morphology may better suit crown substrates to prevent fracture. Conversely, a fine spherical crystal morphology may be more applicable for crown veneering materials to reduce enamel wear. There may be additional advantages in some systems/engineering applications to use glass prediction models where fracture toughness is advantageous. This may particularly apply in glass-ceramics with lower crystal volume fraction (20%) requiring careful control of residual glass thermal properties to reach high K_Ic values [32].

SEM imaging of the experimental glasses revealed signs of amorphous phase separation which appeared to be spherical domains of low connectivity. This was not fully characterized due to the size of the domains but may be associated with the displacement of TiO₂ from the glass network to form Ca or Na-Ca titanates [24,33], providing increased nucleation [34]. The higher leucite area fraction (50–54.2%) and a change in crystal morphology and growth mechanism in GC-B and C may be linked to higher TiO₂ content in these glasses (0.5%). This was similarly demonstrated in an aluminosilicate glass series with systematic increases in Ca/Ti ratios [35]. Appen predictions allowed close matching of glass R.I. with that of the leucite phase (Table 3) to create translucent glass-ceramics. This allows the creation of esthetic dental restorations where tooth shade colors can be subsequently introduced into translucent GCs to mimic tooth shades. Estimated glass fusion temperature predictions ensured processing temperatures of between 1040 And 1180 °C following crystallization. Processing of the novel GCs in current commercial vacuum dental furnaces was therefore achieved (<1200 °C). This ensured a viable route for restoration fabrication available in Dental Technology laboratories.

The ²⁷Al MAS-NMR spectra of the glasses (A, B, C) revealed four-coordinated Al(IV) species in the glasses in similar structural surrounding within three formulations (Figure 4a). Following crystallization heat treatments, it was clear from the spectra that the differing sites for crystalline leucite framework silicate development were present (Figure 4b–d) and confirmed by HTXRD (Figure 2). The ²⁷Al MAS-NMR spectra (Figure 4b) were in close agreement with Kohn et al. [23], who indicated the T1 site was at 58.5 ppm, T2 site at 61.5 ppm and T3 site at 66.5 ppm for leucite. Although the sites overlapped, the spectrum of GC-C shows the most distinct features in the MAS-NMR (Figure 4b) and 3Q NMR (Figure 4c) spectra. This structural factor can be linked to the high BFS encountered in GC-C (Table 4), as there was maximum consumption of aluminum by the leucite phase and minimum residual aluminum in the glassy matrix as indicated by ²⁷Al MAS NMR data (Figure 4d). The longer tail shown in the commercial GCs (Figure 4d) reflects the lower area fraction of leucite in the glassy matrix (Ceramco 3 = 30.2%) and IPS Empress Esthetic GC = 31.2% [17]), and with significantly lower BFS than GC-C. This contrasts with the high leucite area fraction (50%, Figure 3e,f) of finer leucite crystallites indicated in GC-C. MAS-NMR is therefore a valuable tool in assessing the structure/property relations and design of glasses and glass-ceramics. Differences in the more refined and developed microstructural development of GC-C versus GC-B, may be due to its higher K₂O content (Table 1), available for leucite crystal development. Re-homogenization of glass C may also have provided a relatively more uniform distribution of elements across the glass network, effecting the local diffusion for crystallization.

In conjunction with NMR, HTXRD enabled the identification and coexistence of both leucite and sanidine crystal phases in the GCs (Figure 1a,b and Figure 2a,c). These phases exist in the same phase field according to the phase diagram [36]. Barreiro and Vicente (1993) [37] confirmed sanidine crystallization in leucite containing dental ceramics (800–950 °C). It was possible to crystallize sanidine from 760 to 1000 °C (GC-A) and 720–880 °C (GC-B). Glass specimens were crystallized at 650 °C/1120 °C and 620/920 °C with holds (Table 2) and ramped through the sanidine crystallization range to avoid its crystallization. Sintering of the GC compacts in a commercial vacuum furnace was >1000 °C (Table 3) to further discourage the coexistence of two different crystal phases. This was to prevent crystal phase/residual glass mismatches for the CTE (Sanidine = 4.1 × 10⁻⁶ °C⁻¹ and leucite = 28 × 10⁻⁶ °C⁻¹; 25–700 °C) [26], the refractive index (leucite =1.50, Sanidine = 1.52 [38]) and the consequent negative effects on the mechanical properties and translucency. Room temperature XRD of the crystallized GC verified a bulk leucite phase in all sample groups (Appendix A, Figure A4). The GCs processing schedules used in the study were therefore verified. Future work to complement the HTXRD would be Raman spectroscopy to gain further structural and molecular information on these materials [39].

The Appen model in this study yielded accurate data for the design and synthesis of aluminosilicate glasses of high network connectivity and silica content (>69%). Previous studies indicated silica content needed to be >50% for the accurate prediction of glass properties [40,41]. For dental ceramic crown prosthesis applications, glasses with low solubility and bioactivity are required. Estimation of network connectivity (3.87–3.90) and glass properties is therefore essential for safe GC design for Dentistry. The Appen model has also been modified for broader applications in estimating the R.I. of bioactive glasses [11], and applied to dental glass ionomer and light cured composite applications. Graph-based deep learning approaches have been developed to predict properties of chalcogenide glasses [42]. This is witnessed together with a growing interest in data-driven and machine learning (ML) models. Analytical approaches like additive or regression-based methods are however still valued due to accuracy [13]. In the future, incorporation of more structural features [42] into predictive models, together with experimentally verified data input from different glass systems and transfer learning [43], should enhance ML models. ML used in XRD is then the next step in establishing chemical–structural property relations [44].

The current study established predictive glass modeling in conjunction with advanced characterization techniques, producing tailored Dental GC materials. They benefited from a thermally and optically matched residual glass, microstructural control, and enhanced flexural strength and processability. This approach may therefore benefit broader engineering and biomedical applications requiring tailored properties, like optics, glazes and bioactive glasses. Future work is to expand these techniques to design strong diopside GCs, which are thermally compatible with YTZP substrates to prevent chipping of crown restoratives.

5. Conclusions

The predictive Appen model applied in this study was successful in optimizing a series of aluminosilicate glasses and glass-ceramics. Furthermore, the advanced characterization techniques (MAS-NMR and HTXRD) utilized enabled a deeper understanding and control over their structure–property relations. The formulations synthesized can serve as high strength, functional, esthetic and durable leucite glass-ceramic restoratives for dental prostheses.

6. Patents

• Theocharopoulos A., Chen XH., Karpukhina N., Hill, R.G., Cattell, M.J. Leucite glass ceramics. US9856165B2, 2018.
• Theocharopoulos A., Chen XH., Karpukhina N., Hill, R.G., Cattell, M.J. Leucite glass ceramics. EP3013762B1, 2022.