Introduction
Polishing is a process of wear on the surface of one material by another material
to produce a smooth surface.[1] Polishing is required to restore a smooth surface after the final adjustment of
dental restorations. An inadequately polished surface leads to gingival inflammation,
increased dental plaque accumulation, wear of the opposing and adjacent teeth, and
reduces the restoration strength and esthetics.[2]
[3]
[4]
[5]
[6]
[7] The current polishing modalities used in dentistry are mechanically based. However,
there is a polishing process that combines a mechanical process and chemical reaction
together called chemical–mechanical polishing (CMP).
CMP removes material using chemical and abrasive action complement each other to achieve
a highly smooth surface. It is different from the purely mechanical process and purely
chemical removal process. The aim of CMP is to minimize or eliminate direct material
removal by either mechanical abrasion or by chemical etching. Mechanical removal,
such as scratching, can cause severe damage to the surface. Corrosion occurs with
chemical etching. The abrasives used in CMP must chemically react with the polished
surface.[8]
The CMP model originated from glass polishing in which cerium oxide (ceria [CeO2]) was used as an abrasive. It is considered the most effective abrasive for polishing
glass.[9] The CMP occurring between CeO2 and silica (SiO2) have been investigated. Previous studies reported that silica on the glass surface
and ceria abrasive particles can form surface functional groups that temporarily attach
to each other.[10]
[11]
[12] These studies demonstrated that the formed layer is removed by the abrasive plowing
of the slurry particles, exposing a new unreacted surface. Polishing is thought to
occur, as the ceria particles repeatedly remove the silica network at the molecular
scale.
All-ceramic restorations are widely used in dentistry because of their esthetic appearance.
One of these materials is lithium disilicate glass ceramic (Li2Si2O5 or 2SiO2–Li2O) that contains 57 to 80% silica as the main component.[13]
[14] Because the main component of glass and lithium disilicate glass ceramic are silica,
we hypothesized that CeO2 would react with the silica in lithium disilicate glass ceramic, and thus could be
used as a CMP agent. The aims of the present study were to evaluate the CMP effect
of CeO2 as an abrasive to polish lithium disilicate glass ceramic, and determine if silica
would adsorb to CeO2. The null hypotheses were that there would be no significant difference in the surface
roughness (Ra) of lithium disilicate before and after polishing with ceria polishing
paste and there would be no significant difference in silica concentration before
and after adding ceria particles in solution.
Materials and Methods
Polishing Experiment
Twenty-two samples were used in this study. The sample size calculation was performed
using the data from our pilot study. Lithium disilicate samples (7-mm long, 6-mm wide,
and 5-mm thick) were prepared by cutting the blocks (HT A1, IPS e.max CAD, Ivoclar
Vivadent, Schaan, Liechtenstein) using a low-speed precision cutting machine (IsoMet,
1000 No. 11–2180, Buehler, Illinois, United States), ultrasonically cleaned with deionized
water for 5 minutes (CP360 Powersonic, Crest Ultrasonics, New Jersey, United States),
rinsed with deionized water, and dried and fired in a furnace (Programat P300, Ivoclar
Vivadent, Schaan, Liechtenstein) per the manufacturer's directions. After firing,
the samples were cooled in the furnace. Each sample was embedded in a polyvinyl chloride
pipe with epoxy resin. The position of the sample was set at the center on the pipe's
surface. After the epoxy resin had completely hardened, a registration mark was made
at the bottom of the pipe (4-mm wide and 6-mm deep) to allow the specimen to be aligned
at the same position during multiple roughness measurements.
The specimens (six specimens/round) were polished for 5 minutes with 180 grit silicon
carbide sandpaper (3M Wetordry abrasive sheet, 3M, Minnesota, United States) by a
polishing machine with an automatic head (NANO 2000 grinder-polisher with FEMTO-1000
polishing head, Pace Technologies, Arizona, United States). During polishing, the
samples and sandpaper were rotated at 200 rpm in the opposite directions. The pressure
applied on the samples was set at 1 kg/cm2. New sandpaper was used each round. The polished specimens were ultrasonically cleaned
in deionized water for 5 minutes, rinsed with deionized water and dried.
The Ra of the specimen was measured using a profilometer (Talyscan 150, Taylor Hobson,
Leicester, United Kingdom) to determine the baseline roughness. Five 2-mm measurements
were taken at the center of the sample (cut-off value of 0.25 mm and stylus speed
of 0.5 mm/s). The vertical distance between each transverse measurement was 0.4 mm.
The sample was then rotated 90 degrees and remeasured using the same procedure. The
Ra values were averaged to generate a mean Ra value per sample.
After the baseline roughness evaluation, the samples were randomly divided into the
C0.5 and C1 polishing paste groups (n = 10).
Polishing Paste Preparation
Deionized water was used as a lubricant. Ceria (CeO2 powder, <5 µm diameter particles, 99% trace metals basis, Sigma-Aldrich, Merck KGaA,
Darmstadt, Germany) was used as the abrasive in the present study. The polishing pastes
were prepared using different ratios of deionized water:ceria by weight: 1:0.5 (C0.5)
and 1:1 (C1). The polishing pastes were prepared by weighing the components to within
0.0001 g using an analytical balance (GR 200, A&D, Tokyo, Japan) based on each group's
composition and mixed using a spatula for 5 minutes. The mixtures were loaded into
a syringe (0.1–1 mL scale; Slip-tip disposble tuberculin syringe, Medline Industries,
Illinois, United States). The polishing pastes were used within 12 hours.
Polishing Method
The C0.5 and C1 group samples were polished with their respective pastes. Each polishing
paste (0.05 mL) was injected on the center of the specimen and then polished using
a felt wheel (Felt wheel, Jota, Ruthi, Switzerland, 2.2-mm diameter) on a micromotor
(Micromotor and handpiece, Saeyang microtech, Daegu, Korea) for 30 seconds, ultrasonically
cleaned in deionized water for 5 minutes, rinsed with deionized water and dried. A
new felt wheel was used for each group. The micromotor speed was set at 6,000 rpm,
calibrated using a tachometer (Digital tachometer, RS components Ltd., Corby, United
Kingdom). The polishing force was 0.39 N (equal to 40 g hand force). The operator
was calibrated using a precision scale before and during the procedure. The calibration
was repeated for every 10 specimens.[15] All polishing procedures were performed by one operator. After polishing, the specimens
were ultrasonically cleaned in deionized water for 5 minutes, rinsed with deionized
water and dried.
After polishing, the Ra of the specimen was measured using a profilometer as described
for the baseline roughness measurement. After the measurement, the samples were ultrasonically
cleaned in deionized water for 5 minutes, rinsed with deionized water and dried.
The Ra measurement was repeated after an additional 30 seconds of polishing until
120 seconds, that is, measured after 30, 60, 90, and 120 seconds of polishing, the
polishing had been performed.
Surface Roughness Measurement
The Ra of the polished surface was measured using a profilometer using the same procedure
at the same position as at the baseline roughness measurement.
Scanning Electron Microscopy Analysis
Two samples from each postpolishing (120 seconds) group and two unpolished samples
with the baseline roughness were removed from the epoxy resin and ultrasonically cleaned
in deionized water for 5 minutes, rinsed with deionized water, and dried. The specimens
were mounted on adhesive coated aluminum stubs (1 sample/stub) and gold sputter-coated
(100 second, 50 mA) using a sputtering device (JFC-1200 Fine Coater, JEOL, Tokyo,
Japan). The surface images were taken using an electron microscope (Quanta 250 FEG
scanning electron microscope, FEI, Oregon, United States) with 20 kV accelerating
voltage and 500X magnification to evaluate the surface morphology.
Statistical Analysis
The data were statistically analyzed using two-way repeated analysis of variance (ANOVA)
followed by Bonferroni correction to compare the differences in mean Ra values between
groups (SPSS version 26.0 for Windows, SPSS, Chicago, Illinois, United States). A
p-value of <0.05 was considered statistically significant.
Adsorption Experiment
Instrument Preparation
To avoid the dissolution of silicate ions, no glassware was used in this study. Nitric
acid (10% by volume) was prepared from diluted nitric acid (70% by volume) (Ajax Finechem,
Thermo Fisher Scientific Inc., Massachusetts, United States) with deionized water.
The instruments used in the study were soaked in the acid solution for 24 hours to
remove the residual ions on the instruments' surface, rinsed with deionized water
thrice and dried in a hot air oven (60°C) for 4 hours.
Specimen Preparation
Ten samples were used in this study. The sample size calculation was performed using
data from our pilot study. Lithium disilicate blocks (HT A1, IPS e.max CAD, Ivoclar
Vivadent) were cut transversely into 2-mm thick specimens using a low-speed precision
cutting machine (IsoMet, 1000 No. 11–2180, Buehler). The blade rotation speed was
300 rpm. The specimens were polished with 600 grit silicon carbide sandpaper (3M Wetordry
abrasive sheet, 3M) by a polishing machine (NANO 2000 grinder-polisher, Pace Technologies,
Arizona, United States) for 2 minutes. During polishing, the sandpaper was rotated
at 200 rpm. The polished specimens were ultrasonically cleaned in deionized water
for 5 minutes (CP360 Powersonic, Crest Ultrasonics), rinsed with deionized water and
dried. The specimens were fired in a furnace (Programat P300, Ivoclar Vivadent) as
per the manufacturer's directions. After firing, the specimens were cooled in the
furnace.
Adsorption Analysis
The specimens were soaked in 50-mL deionized water (1 specimen/container, n = 10). The plastic containers were covered with a cap, sealed with parafilm (Parafilm
M laboratory wrapping film, Bemis Company, Inc., Wisconsin, United States) and placed
on a magnetic stirrer (Yellow MAG HS7, IKA, North Carolina, United States) for 24 hours
at room temperature (25°C).
After 24 hours, the specimens were removed. Each liquid sample was divided into two
halves by drawing 25 mL of the solution and transferred into a new container using
a syringe (50-mL syringe, NIPRO, Osaka, Japan). The first half was stored in a closed
container, capped and sealed with parafilm and placed in a refrigerator (4°C). Ceria
particles (5 g, CeO2 powder, <5-µm diameter particles, 99% trace metals basis, Sigma-Aldrich, Merck KGaA)
were added into the second half of each sample. The containers were capped, sealed
with parafilm and left to equilibrate on a magnetic stirrer for 24 hours at room temperature
(25°C).
After 24 hours, the solutions were filtered through a 0.22-μm membrane filter (tube
top vacuum filter system, 0.22-µm pore, 50 mL, Corning, New York, United States),
closed with a cap, sealed with parafilm and placed in the refrigerator (4°C).
Inductively Coupled Plasma-Optical Emission Spectrometry
In this assay, the blank test was deionized water. The concentration of silicon (Si)
in the liquid samples was analyzed using an inductively coupled plasma-optical emission
spectrometry analyzer (ICP-OES Optima 7300 DV, PerkinElmer, Inc., Massachusetts, United
States). The analysis was repeated thrice per sample. The silicon concentrations were
averaged to generate the mean value per sample.
Statistical Analysis
The data were statistically analyzed using the paired t-test to compare the difference in mean silicon concentrations before and after adding
ceria particles (SPSS version 26.0 for Windows, SPSS). A p-value of <0.05 was considered statistically significant.
Results
The results of the polishing experiment comprising the mean Ra values, standard deviations,
and significant differences between the groups are presented in [Table 1]. The mean Ra values were not significantly different between the groups at baseline
(p > 0.05). Within each group, the mean Ra values significantly decreased as the polishing
time increased (p < 0.05; [Fig. 1]). Comparing the groups, the C1 group demonstrated a significantly lower mean Ra
value than the C0.5 group (p < 0.05).
Table 1
Mean Ra values of the groups at baseline and after polishing for 30, 60, 90, and 120 seconds
Group
|
Ra value (μm) and standard deviation
|
Baseline
|
30 seconds
|
60 seconds
|
90 seconds
|
120 seconds
|
C0.5
|
0.07979 ± 0.00197A
|
0.07504 ± 0.00060
|
0.07122 ± 0.00073
|
0.06835 ± 0.00072
|
0.06515 ± 0.00065
|
C1
|
0.07922 ± 0.00150A
|
0.07383 ± 0.00042
|
0.06900 ± 0.00074
|
0.06503 ± 0.00090
|
0.06306 ± 0.00063
|
Abbreviations: C, ceria; Ra, surface roughness.
Note: The same superscript letters represent no significant difference between group
(p > 0.05) by two-way repeated analysis of variance and Bonferroni correction.
Fig. 1 The mean Ra value of the groups before and after polishing for 30, 60, 90 and 120 seconds.
Asterisks (*) represent no significant difference between groups (p > 0.05) by two-way repeated ANOVA and Bonferroni correction. ANOVA, analysis of variance;
Ra, surface roughness.
The Ra of the samples observed by scanning electron microscopy (SEM; [Fig. 2]) correlated with their Ra values. The surface of the samples at baseline was the
roughest. After polishing, the C1 group had a smoother surface compared with the C0.5
group.
Fig. 2 Scanning electron microscope images (×500) of the sample surfaces polished with the
different polishing pastes, (A) surface before polishing, (B) C0.5 and (C) C1. C, Ceria.
The adsorption experiment results revealed that silicon was not found in the blank
test (deionized water); however, it was found in the solution after soaking the specimens
in deionized water (0.1037 ± 0.0019 mg/L). After adding ceria particles, the silicon
concentration in the filtered solution decreased to 0.0036 ± 0.0013 mg/L ([Table 2]). There was a significant difference (p < 0.05) in mean silicon concentration between before and after adding the ceria particles.
Table 2
The mean silicon concentrations before and after adding cerium oxide particles
Solution
|
Silicon concentration (mg/L)
|
Before
|
After added cerium oxide
|
Blank test
|
0.00
|
0.00
|
Lithium disilicate
|
0.1037 ± 0.0019
|
0.0036 ± 0.0013
|
Discussion
In the polishing experiment, we evaluated whether a CeO2 polishing paste could polish lithium disilicate glass ceramic. The Ra of the C0.5
and C1 groups were significantly lower (p < 0.05) than the baseline roughness. The Ra was significantly decreased (p < 0.05) when the ceria ratio was increased from 0.5 to 1. The surface morphology observed
in the SEM analysis correlated with the Ra values. The baseline roughness that had
the highest Ra value, demonstrated the roughest surface. The C1 group that had the
lowest Ra value demonstrated the smoothest surface. Based on these results, the first
null hypothesis was rejected. The results indicated that lithium disilicate ceramic
can be polished by the ceria polishing pastes evaluated in this study. Suratwala et
al reported that increasing the slurry ceria concentration resulted in decreased atomic
force microscope roughness.[16] Moreover, Wang et al reported that increasing the ratio of ceria particles resulted
in an increased polishing rate. At low concentration, the chemical formation rate
was slow and limited the overall polishing rate. Increased ceria particles led to
an increased number of active particles and chemical reactions which is involved in
the polishing rate.[17]
A previous study reported that water is an important factor in glass polishing due
to the presence of hydroxyl groups. The polishing rate increased with increased hydroxyl
reactivity. The rate was nearly zero in hydrocarbon liquids. The highest polishing
rate was found when water was present.[10] Therefore, deionized water was used as a lubricant in this study. Moreover, Plakhova
et al reported that CeO2 reacted with water to form surface functional groups on its particules which were
important in the CMP process. The reaction between ceria and water is described in
Eq. (1).[18]
CeO2 + 2H2O → Ce(OH)4 (1)
Lithium disilicate is a particle-filled glass ceramic. Its structure is approximately
70 vol% lithium disilicate crystals embedded in a glass matrix. Lithium disilicate
is composed of 57.0 to 80.0% SiO2, 9.0 to 11.0% Li2O, 0.0 to 13.0% K2O, 0.0 to 11.0% P2O5, 0.0 to 8.0% ZrO2, 0.0 to 8.0% ZnO, 0.0 to 5.0% Al2O3, 0.0 to 5.0% MgO, and 0.0 to 8.0% coloring oxides[13]
[14] The adsorption experiment results demonstrated that silica was found in the solutions
that the lithium disilicate specimen was soaked in. These results indicate that there
was a chemical reaction between lithium disilicate and water resulting in its dissolution.
Corresponding to previous studies, dental ceramic was found to dissolve in aqueous
solutions.[19]
[20] The reaction between silica (in lithium disilicate ceramic) and water is described
in Eq. (2). In water, silica forms a surface functional group which is silanol or ≡Si–OH. The
rate of surface removal was controlled by the hydrolysis of the siloxane network and
the rate of the reaction below the surface is controlled by the diffusion of water
in silica.
SiO2 + 2H2O ⇄ Si(OH)4
or ≡Si–O–Si≡ + H2O ⇄ 2(≡Si–OH) (2)
The adsorption assay results revealed that the mean silicon concentration of the solution
after adding ceria particles and filtering was significantly lower (p < 0.05) than the soaked lithium disilicate specimen solution. Based on these results,
the second null hypothesis was rejected. The decreased silicon concentration in the
solution after adding ceria particles indicated that silicon was adsorbed by ceria
particles.
The adsorption of silicon onto the ceria surfaces was confirmed in our pilot study.
The silicon removed from the filtered solution was found in the filtered ceria particles
that were analyzed using X-ray fluorescence. However, there was no silicon in the
control solution (ceria particles soaked in deionized water). These findings indicate
that the silicon was adsorbed by ceria particles.
According to Eqs. (1) and (2), the surface functional groups of the ceria particle and silica in water are cerium
hydroxide (≡Ce–OH) and silanol (≡Si–OH), respectively. During polishing, ceria particles
are in contact with the glass surface. A ≡Si–O–Ce≡ bridging bond forms at the interface
as a complex. The reaction is described in Eq. (3).
≡Si–OH + ≡Ce–OH → ≡Si–O–Ce≡ + H2O
or ≡Si–O–Si–OH + ≡Ce–O–Ce–OH → ≡Si–O–Si–O–Ce–O–Ce≡ + H2O (3)
The bond strength of the ≡Si–O–Ce≡ complex is stronger than that of ≡Si–O–Si≡ (in
silica). When the ceria particles receive the mechanical polishing force, strain is
placed on the ≡Si–O–Ce≡ complex. If the strain is high enough, the bond between ≡Si–O–Si≡
(silica) and ≡Si–O–Ce≡ complex will break. The ≡Si–O–Ce≡ is removed from the glass
surface and then the new unreacted surface is exposed. The CMP process cycle then
repeats.[10]
[12]
[21]
[22]
[23] This reaction is described in Eq. (4).
≡Si–O–Si–O–Ce–O–Ce≡ + H2O → ≡Si–OH + HO–Si–O–Ce–O–Ce≡ (4)
The present study investigated the CMP ability of CeO2 as an abrasive to polish lithium disilicate glass ceramic. The mechanical factors
that affect material removal are abrasive type, size and shape, load during polishing,
and polishing speed.[12]
[24]
[25]
[26]
[27]
[28] However, CeO2 powder with particles <5-µm diameter was the only abrasive type used in this study.
Load during polishing and polishing speed were fixed as controlled variables. Moreover,
the pH of the polishing slurry, point-of-zero charge, and temperature also affect
material removal.[21]
[22]
[29] To improve the efficiency of the polishing paste, these factors require further
investigation.
In the last decade, the use of CeO2 has tremendously increased in many fields, including as a fuel additive, electronics,
medicine, ceramic application, polishing agent, and agriculture.[9] However, some reports demonstrated that CeO2 escapes into the environment from sludge leakage and wastewater discharge and enter
the food chain.[30]
[31] Donovan et al also reported that CeO2 particles were found in drinking water.[32] Therefore, humans may unintentionally eat food or drink contaminated with CeO2. The present study has demonstrated that CeO2 particles react with silica in ceramic materials resulting in material removal. A
long-term study is needed to investigate the effect of food or drink containing CeO2 on dental ceramic restorations in vivo.