To determine the influence of cement thickness and ceramic/cement bonding on stresses and failure of CAD/CAM crowns, using both multi-physics finite element analysis and monotonic testing.
Axially symmetric FEA models were created for stress analysis of a stylized monolithic crown having resin cement thicknesses from 50 to 500 μm under occlusal loading. Ceramic–cement interface was modeled as bonded or not-bonded (cement–dentin as bonded). Cement polymerization shrinkage was simulated as a thermal contraction. Loads necessary to reach stresses for radial cracking from the intaglio surface were calculated by FEA. Experimentally, feldspathic CAD/CAM crowns based on the FEA model were machined having different occlusal cementation spaces, etched and cemented to dentin analogs. Non-bonding of etched ceramic was achieved using a thin layer of poly(dimethylsiloxane). Crowns were loaded to failure at 5 N/s, with radial cracks detected acoustically.
Failure loads depended on the bonding condition and the cement thickness for both FEA and physical testing. Average fracture loads for bonded crowns were: 673.5 N at 50 μm cement and 300.6 N at 500 μm. FEA stresses due to polymerization shrinkage increased with the cement thickness overwhelming the protective effect of bonding, as was also seen experimentally. At 50 μm cement thickness, bonded crowns withstood at least twice the load before failure than non-bonded crowns.
Occlusal “fit” can have structural implications for CAD/CAM crowns; pre-cementation spaces around 50–100 μm being recommended from this study. Bonding benefits were lost at thickness approaching 450–500 μm due to polymerization shrinkage stresses.
There are no well-established criteria for the evaluation of the fit of CAD/CAM ceramic crowns, despite their rapid development and increased use. This situation is becoming especially troublesome to those attempting to standardize CAD/CAM equipment, processes and their products within the International Standards Organization Technical Committee 106 (Dentistry). This present work attempts to inform those discussions as well as to examine the issue of “fit” from a structural perspective for the clinician.
Although marginal misfit of indirect restorations has been studied for many years, often being cited as an important parameter for clinical longevity, evidence for such a view seems insubstantial. For example, there seems not to be any clinical evidence of the relationship between marginal fit and secondary caries , or performance in vivo . In fact recent secondary caries literature treats such lesions as being primary caries adjacent to restorations with the etiology for both related to the bacterial composition of marginal plaque and to the oral hygiene . Only one clinical study of subgingival marginal miss-fits of crowns appears to exist with the only negative outcome being a mild gingivitis response; finding correlations between marginal gap and gingival index and crevicular fluid volume, but none with pocket depth . It appears that clinicians’ long-time focus on marginal fit likely derives from its being the easiest parameter to evaluate with its importance simply being unquestioned. Manufacturers of CAD/CAM equipment appear then to have optimized systems to deliver “marginal fit” to satisfy the expected clinician judgment. As will be seen, internal fits are often quite different than marginal adaptation suggests during simple clinical evaluation.
Internal occlusal misfits of milled crowns can vary enormously both within a single system and between different CAD/CAM systems. Misfits varying from 24 μm beneath cusp tips up to 634 μm below the central fossa have been reported for one system . Internal misfits as large as 1316 μm have been reported beneath crowns . Mean occlusal misfits differed depending on marginal preparation design for one ceramic milled on one popular CAD/CAM machine (215 μm, 244 μm and 300 μm; with standard deviations of ±50 μm to 100 μm) . Many studies indicate that an average occlusal space centered on 100–200 μm is quite common . While design software for many CAD/CAM systems allows the specification of the pre-cementation space, the final results seem not to be as well-controlled as would be expected . As a result of these literature reports, this project focused on an occlusal misfit range of 50–500 μm. Ranges of experimental variables are often extended slightly above and below expected values for completeness and to accentuate trends.
Our interest in changing the focus to the internal fit derives from clinical findings, in vitro simulations and theoretical modeling indicating the importance of “support” for all-ceramic crowns . Since large internal misfits would result in a thick layer of a low elastic modulus material (cement), misfit may be structurally important for brittle materials, such as ceramic restorations. Additionally, thick layers of resin-based cements could produce residual tensile stresses within bonded ceramic crowns due to polymerization shrinkage.
Analytical solutions do exist for the influence of adhesive layer thickness on the critical load for radial cracking of ceramic plates on a compliant base . While such solutions can provide guidance, they assume that the loaded contact area is infinitesimal compared to the ceramic thickness, a situation that does not hold for the dental situation where occlusal contacts (wear facets) can be much larger than the thickness of the crown. For example, intaglio surface stress profiles were shown to vary greatly for a 1 mm thick ceramic loaded with 1 mm, 2 mm and 3 mm diameter contacts . Available analytical solutions also do not account for stresses due to polymerization shrinkage, and even if they did their flat-plate geometry would not extend to cement enclosed within a crown.
Additionally, the role of cement thickness on resistance to fracture of ceramic restorations has received little attention in the literature. In one FEA analysis, Proos et al. reported a stress increase of only 2% when cement thickness was set at 100 μm versus 50 μm, and concluded that neither the elastic modulus nor the thickness played a significant role since stresses were efficiently “transferred” between ceramic and dentin. Rekow et al. reported a slight influence on FEA maximum stresses over a very narrow range of cement thickness: 240 MPa at 80 μm; 204 MPa at 100 μm. Scherrer et al. reported only a gradual decrease in failure loads with resin cement thickness that did not become statistically significant until 300 μm. However the failure event in Scherrer et al. was not radial cracking, but “crushing” of the ceramic with a small steel ball sufficient to create a load drop. Therefore, the influence of cement thickness (let alone shrinkage) remains unclear regarding the fracture resistance of ceramic crowns.
Another factor, which likely affects the fracture resistance of ceramic crowns, is the quality of the bond established between ceramic and cement. One clinical study indirectly showed the importance of bonding where acid etching of an early glass-ceramic decreased the annual failure risk of crowns by about 50% .
Finite element analysis (FEA), combined with clinically validated monotonic testing of major model predictions, can provide important insights into the possible clinical behavior influenced by factors such as cement thickness and bonding . Results from such research may inform the discussion regarding the development of standards for CAD/CAM ceramic crowns. Such guidelines, especially if confirmed with further clinical observations, serve the objective of optimizing the survival of these restorations.
Occlusal resin cement thickness was evaluated in this study by both modeling and physical testing both for bonded and non-bonded crowns. The purposes were: (a) to evaluate the influence of occlusal resin cement thickness on ceramic stresses and failure loads; (b) to evaluate the influence of ceramic–cement bonding on internal ceramic stresses and failure loads; (c) to explore the interaction between occlusal resin cement thickness and bonding on stresses and failure loads; (d) to develop and validate an FEA model for this and future work (including polymerization shrinkage) by comparing model predictions with failure load values found experimentally.
Materials and methods
Finite element analysis
Two-dimensional axially symmetric FEA models were developed that included components based on the dimensions and materials to be used with the physical specimens. The stylized prepared molar was designed with a 1 mm rounded shoulder margin, 5.3 mm axial walls converging at 6°, an occlusal surface diameter of 7.6 mm and both margin-axial wall and occlusal-axial fillet radii of 0.5 mm. The overall height of the “molar” die (including “root” base) was 10.5 mm. The occlusal resin cement thicknesses were represented as being 50 μm, 100 μm, 300 μm and 500 μm. Ceramic crowns were 1.5 mm thick on their flat occlusal portion and axial wall thickness diminishing to 1.0 mm at the margin, with a constant 11 mm outer diameter and a height varying from 6.87 to 7.32 mm, depending on the cement thickness. The indenter was modeled as having a 2 mm diameter contact area with the ceramic. COMSOL Multiphysics (COMSOL Group, Stockholm, Sweden) was the software used for the isotropic, linear elastic FEA. The cement–ceramic interface was set as being either “bonded” (nodes coupled for displacement) or “non-bonded” (independent nodes at the interface) and the cement–dentin interface was bonded for all models. The bottom boundary of the die was fixed in the three axis of displacement; the boundaries in the central axis of the geometry ( r = 0) were set to “axial symmetry”. Compressive distributed load was applied on the top of the indenter (500 N). The other object boundaries had free displacement. The elastic material properties were defined according to Table 1 .
|Material||Elastic modulus ( E , GPa)||Poisson’s ratio ( ν )|
|G-10 piston/die||14.9 a||0.31 a|
|Resin cement||6.3 b||0.35 a|
|Feldspathic ceramic||64 c||0.25 a|
Polymerization shrinkage was mimicked as a thermal contraction. In order to determine the temperature drop (Δ T ) necessary to provide an axial shrinkage-strain of 4.65% (as reported by Spinell et al. for Multilink Automix ® (Ivoclar Vivadent, Schaan, Liechtenstein), a cement disk was modeled and “cooled” within COMSOL. The dimensions of the cement disk were the same used by these authors ( = 8 mm; h = 1 mm). The maximum shrinkage strain of 4.65% was obtained when Δ T was equal to −327 °C, for a cement thermal expansion coefficient ( α ) of 70 × 10 −6 . The cement layer was set to “include thermal expansion” and the negative Δ T was applied. In the models analyzed for “non-shrinking” cement, the option thermal expansion was not activated.
Triangular second order elements were used. Meshing was controlled through free mesh parameters and the final mesh size was determined by convergence testing. Maximum element sizes for the cement varied from 0.025 mm to 0.20 mm, in order to obtain 3 layers of elements for each cement thickness. Maximum element size of 0.1 mm for ceramic were chosen since they presented reasonable solution times and in which all the models analyzed had reached convergence (<3% solution change with mesh step decrease).
First maximum principal stresses were analyzed: (1) on the cementation surface directly beneath the piston, where failure is expected from occlusal loading and (2) in the occlusal-axial angle of the cementation surface of the crowns where the stresses due to shrinkage showed maximum values. This second measurement point was standardized using a line crossing the axis of symmetry (at the height of 7.5 mm) and the lower point of the external occluso-axial fillet.
Crown specimen fabrication and testing
Dies having the dimensions formerly described for the FEA model were machined from a woven glass-filled epoxy (NEMA grade G10, International Paper, Hampton, SC, USA) that was shown to have blunt contact stress–strain behavior identical to hydrated dentin . Four crown patterns were machined from aluminum, having 50 μm, 100 μm, 300 μm and 500 μm of internal occlusal space, an occlusal thickness of 1.5 mm and 100 μm of internal axial-wall space with respect to the G10 dies. These aluminum patterns were then scanned and crowns were machined from Vita Mark II blocks (Vita Zahnfabrik, Bad Säckingen, Germany, batch # 20280 and # 23580) using the CEREC InLab (Sirona, Bensheim, Germany). The different batches of blocks were equally distributed among the different groups. Twelve crowns were machined for each cement thickness, in order to have n = 6 per group (bonded and non-bonded groups). After machining, the crowns were tried onto their respective dies, and when binding was encountered the axial walls of the dies were adjusted. This adjustment was fundamental so that every crown could reach the maximum seating at the margin so that the intended occlusal cement thickness was achieved. The occlusal space was initially checked in 2 crowns per group, using poly(vinyl siloxane) and the values found were: 40 μm and 50 μm for the 50 μm group; 110 μm and 160 μm for the 100 μm group; 310 μm and 320 μm for the 300 μm group; 420 μm and 470 μm for the 500 μm group.
Before cementation, the crowns were lightly finished on their occlusal surface (silicon carbide grit # 600). Crowns were ultrasonically cleaned in deionized water for 5 min and their internal surfaces were etched with 9% hydrofluoric acid (Porcelain Etch, Ultradent Products Inc., South Jordan, UT, USA, batch # 461M) for 60 s, rinsed with flowing deionized water for 15 s and ultra-sonic cleaned in deionized water for 1 min. The dies were ultrasonically cleaned twice in deionized water (3 min each) and after drying they were treated with a primer mixture (Primer A and Primer B, Ivoclar Vivadent, Schaan, Liechtenstein, batch # 58561).
For bonded groups, silane was applied to the etched internal surface of the crowns, according manufacturer’s instructions (Ultradent ® Silane, Ultradent Products Inc., South Jordan, UT, USA, batch # 45LM). For non-bonded groups, poly(dimethylsiloxane) (PDMS) was lightly applied with a microbrush in the internal surface of the etched crowns (30,000 cSt, Aldrich Chemical Company, Milwaukee, WI, USA) and they were heated for 5 min on a hot plate (80 °C), for better flow and penetration into the etched ceramic surface features. Pilot work evaluated different viscosities of PDMS applied to the etched ceramic surface onto which resin cement was flowed, photo cured and removed. The viscosity of 30,000 cSt was chosen after analyzing the treated ceramic surface by SEM and verifying that this viscosity obturated micromechanical etch features and avoided the presence of residual cement after its removal; i.e. the cement was well adapted to the etched ceramic but not bonded.
Resin cement was applied through an automix tip to the occlusal surface of the dies and the crowns were seated under a load of 4.6 N (Multilink Automix, Ivoclar Vivadent, Schaan, Liechtenstein, batch # 41970). Excess cement was removed following 2 s of light curing from two opposite directions at the margin. Light curing was applied to two opposite axial walls for 20 s each. Following that, the weight was removed and curing light was applied for 20 s to the occlusal surface of the crown. Primer from this cement system applied to the dies initiated additional chemical polymerization.
Specimens ( Fig. 1 a) were stored dry for 96 h under room temperature before testing. An acoustic emission (AE) transducer, utilized to detect the first “pop-in” of the crack during the monotonic test (Pico, Physical Acoustics Co., NJ, USA) was coupled to the ceramic crown using silicon cement (Silicone Rubber Sealant, DAP, Baltimore, MD, USA) ( Fig. 1 b). Loading was performed at 5 N/s in an electroforce machine (Elf 3300, EnduraTEC Systems Co., MN, USA) via a 2 mm diameter flat-end piston made from woven glass-filled epoxy material (NEMA grade G10). This polyethylene strip (110 μm) was placed between the ceramic and the indenter to help avoid cone cracking due to edge-loading by the piston. AE signals occurring during the loading were pre-amplified, filtered and acquired by an AE signal process unit (μDisp, Physical Acoustics Co., NJ, USA). All specimens were analyzed under trans-illumination to confirm the presence of radial cracking and absence of non-clinical surface damage such as Hertzian cone cracking. The load necessary to cause the radial crack was calculated by multiplying the time for the acoustic event by the loading rate (5 N/s).
After the loading test, all specimens were embedded in acrylic resin. Twenty-four hours later, they were sectioned in two halves and polished for optical and scanning electron microscopy (SEM). The thickness of the ceramic and the cement beneath the point of load application were determined at the failure origin with a measuring microscope (Gaertner Scientific Corporation, Chicago, USA). Each specimen was also viewed by SEM looking for the radial cracks, secondary cracking due to cement shrinkage, for gaps in the cement/ceramic interface or any other characteristic which could help to interpret results (TM1000, Hitachi, Tokyo, Japan).
Load-to-fracture data from the monotonic tests were normalized (Eq. (1) ) for a ceramic thickness of 1.65 mm, taking into account the real ceramic thickness, given that the loads to fracture are proportional to the square of the porcelain thickness under load .