1. Introduction
The traditional manufacturing methods are often limited in the complexity of the technologically achievable shape of ceramic items. When piezoceramic items of complex geometry are demanded, some of these methods are physically incapable of producing the necessary geometrical features, while other methods require costly molds and dyes [
1] or involve complex processing after the sintering [
2]. Furthermore, these processing machining methods often lead to microstructure distortion and surface damage, which post-machining annealing can often ameliorate. However, in some applications employing piezoelectric shear modes, the geometry demands that machining is carried out after poling, which rules out the possibility of annealing due to the high probability of the piezoelectrical properties degradation [
3]. The use of additive manufacturing (AM) methods for piezoceramics production could overcome the mentioned adverse effects. In particular, the complex shapes of items with a slight increase in production cost, traded for good physical properties, become possible. This is especially critical for small-batch manufacturing [
4]. The most commonly used AM methods of piezoceramic green parts manufacturing are Fused Deposition Modeling (FDM) [
5], Selective Laser Sintering (SLS) [
6], Ceramic Binder Jetting/Inkjet Printing (BJ/IJ) [
7], Ceramic paste extrusion (Robocasting) [
8], and Stereolithography (SLA)/Digital Light Processing (DLP) [
9].
Photopolymerization-based AM, such as SLA/DLP, is a well-designed widespread technology applied to manufacturing of technical ceramics [
10]. The physical principle of layer applications to the build platform and selective photopolymerization differs between SLA and DLP technologies. The SLA approach uses highly viscous paste as the base material, as well as UV laser, to selectively polymerize contours and shade the areas within the build platform. The DLP method uses relatively low viscosity suspensions as the base material and either a simple UV/Vis source of light projected to the entire build platform through a physical mask, or a Digital Mirror Device containing arrays of micromirror to selectively reflect the light from a light source to the build platform. Stereolithography-based ceramic AM often reach precision down to 30 µm within a layer and the layer thickness of 10–50 µm. The final resolution is even better due to the shrinkage of 3D-printed parts during sintering (on average, by 15–25%). The volume fraction of ceramic powder reaches 70–80 wt.% in pastes and 60–70 wt.% in suspensions. The higher volume fraction leads to lower shrinkage and higher final density of the ceramic product. A thermal processing stage is required for the 3D-printed green parts to burn off the organic binder and sinter the ceramic material.
Both technologies were applied to piezoceramics 3D printing with different success. The 3D printing of piezoceramics with photopolymerization-based technologies is, in general, similar to the 3D printing of regular ceramic materials [
11]. Piezoceramic materials, such as BT, feature greater refractive indices when compared to the traditional technical ceramics Al₂O₃, ZrO₂, or yttrium stabilized zirconia (YSZ), that are successfully used in SLA and DLP technologies. When mixed with an organic binder, the high refractive index materials increase photon scattering during the UV/Vis exposure stage of the 3D printing process, efficiently increasing the polymerized area and decreasing the polymerization depth [
12]. The polymerization depth decrease requires thinner layers (down to 10–20 μm) when 3D-printing the material [
13]. As a result, a weak interconnect between the adjacent layers might happen, leading to delamination and in-layer fracturing during subsequent thermal processing of the 3D-printed items. In SLA/DLP methods, the 2–3 layers overlap of polymerization depth is typically provided during the photopolymerization to relieve these effects. The proper photopolymerization becomes an issue with thin layers, although the scientific community demonstrated some successful examples of such an approach.
Jang et al. [
14] provided a fundamental understanding of BT-based ceramic suspension behavior under photopolymerizing conditions. They tested different incremental loadings (0–55 vol.%) of BT powder in a suspension. The suspensions behavior under UV curing was tested with an SLA 3D printer having a 350 nm laser. The authors mentioned that the cure depth of the suspensions decreases from 550 μm at 10 wt.% solid loading to below 250 μm at 30 wt.% solid loading when 100× repetitions of UV exposure were used. At a single exposure, the cure depth below 100 μm was observed for the best case, and the other cases were well below 50 μm. The authors attributed the poor polymerization of the BT suspensions to the significant differences in refractive indices of the ceramic powder and the organic binder that lead to excessive photon scattering.
Chen et al. [
15] demonstrated a 3D printing of BT piezoceramic samples using a hand-made digital mirrors-enabled DLP setup [
13], and 70 wt.% solid loading of 100 nm BT powder. The authors did not mention the wavelength and power of the light source used for photopolymerization, exposure time, and layer thickness. The sintered at 1330 °C material had density of 93.7% of the theoretical maximum for BT, the dielectric loss tangent of 0.012 (versus 0.1 for typical BT), the ε (1 kHz) of 1350 (versus 1700 for typical BT), and piezoelectric module d
33 160 pC/N (versus 190 pC/N for typical BT). Song et al. [
16] demonstrated another attempt to 3D print BT-based piezoceramic parts using digital-mirrors-enabled DLP technology. They formulated a few ceramic suspensions using 30–80 wt.% of 1 μm-sized BT powder. The cure depth of 180 µm was reached for 30 wt.% solid loaded suspension at 16 s of exposure time, while, for more practical 60–80 wt.% solid loaded suspensions, the best cure depth of around 50 µm was achieved at 8 s and 16 s of exposure time. The authors deduced from the preliminary test the best layer thickness of 20 µm at exposure time 2–8 s for 60–80 wt.% loaded suspensions. Measured properties of BT samples fabricated by digital-mirrors-enabled DLP process and sintered at 1330 °C were relatively low: piezoelectric module d
33 = 87 pC/N, dielectric constant ε (1 kHz) = 920, dielectric loss tangent tan δ = 0.07.
Chen et al. [
17] fabricated BT piezoelectric samples using the commercial DLP printer (Asiga max, NSW, Australia) with 385 nm wavelength. During the DLP process, the layer thickness was set as 10 μm for all the studied samples. To prepare stable and homogeneous BT suspensions, the 40 vol.% BT powder with the submicron powder particle size (200 nm, 500 nm, and 600 nm) and 60 vol.% photocurable resin. The maximum relative density of 0.98 was obtained in the samples made of 200 nm powder and sintered in the range 1300–1330 °C, while the best piezoelectric properties were recorded in the samples prepared from 600 nm powder: ε (1 kHz) = 4423, tan δ (1 kHz) = 0.019, d
33 = 206 pC/N, relative density of 0.96. In Reference [
18], LCD-SLA 3D printing method with 405-nm LEDs UV light source was used to create BT piezoceramic samples. Three types of BT powders were used: one with a particle size d
50 = 3.4 μm, another BT powder with a particle size of d
50 = 1.02 μm, and nanoscale BT powder with a particle size of d
50 = 50–70 nm. The study showed significant delamination and spalling for samples prepared from micron-sized BT powder with d
50 = 3.4 μm. No such defects were found for the ceramic slurries prepared from finer powders. The highest relative density (0.90) and piezoelectric properties (ε
r = 1965, d
33 = 200 pC/ N, tanδ = 0.017) were measured in the sample sintered at 1300 °C with a dwell time of 4 h.
In the work of Cheng et al. [
4], BT ceramics were manufactured based on the SLA 3D printing method. The authors do not mention the light source’s wavelength and power, exposure time, and layer thickness. The authors used a photosensitive resin with undisclosed components and a BT powder (particle size 500 nm) to formulate SLA-suitable slurry with powder loading: 70, 75, 80, 82, 84, and 86 wt.%. The BT ceramic samples sintered at 1290 °C exhibited acceptable piezoelectric properties: d
33 = 166 pC/N at 80 wt.% BT powder. To evaluate the performance of 3D-printed BT piezoceramic samples, a 1.4 MHz focused ultrasonic array was fabricated and characterized. The 6 dB bandwidth of the array was at 40%, and the insertion loss at the central frequency was 50 dB. The results showed that the 3D-printed BT piezoceramic array has good potential as an ultrasonic transducer. Another promising example of using the SLA 3D printing technique for BT piezoceramics was shown in the work by Wang et al. [
19]. The suspensions with 40 vol.% BT nanoparticles (mean particle size 500 nm) displayed shear thinning behavior and relatively low viscosity of 232 mPa·s at a 46.5 s
−1 shear rate. The green BT samples fabrication was performed using a 3D printing device based on the bottom-up mode SLA process. The UV light source has a 405 nm wavelength laser providing a photopolymerization depth of 65 μm. After debinding and sintering at 1320 °C, the 3D-printed ceramic specimens showed a nanometer-level grain size and about 95% of the theoretical density, excellent dielectric properties (ε
r = 2726 and tanδ = 0.016 at 1 kHz), and acceptable piezoelectric constant, d
33 = 163 pC/N.
In all the above mentioned studies, light sources with wavelengths of 350–405 nm were used, limiting the depth of photopolymerization [
11]. In our previous work [
20], it was shown that the change of the laser wavelength to 465 nm could significantly increase the efficiency and speed of piezoceramic paste photopolymerization. At the moment, no data could be found in the literature on the piezoelectric properties of BT ceramic samples obtained with SLA at a wavelength higher than 405 nm. In addition, there was no direct comparison of additive and conventional manufacturing results for the BT piezoceramics, when the same initial powder and sintering conditions were used. This study was aimed at conducting a comparative study of BT ceramics manufactured with SLA and traditional semi-dry pressing, to identify the correlations between powder particle sizes, piezoelectric properties, relative density, microhardness, and microstructure of the material.
2. Materials and Methods
Based on the previous results [
20], the 465-nm range for the laser source was selected for experimental testing of SLA-based BT ceramics manufacturing. A series of samples were prepared using SLA technique and the traditional method of semi-dry pressing. The samples were thermally treated in the high-temperature oven, all at once, using the same heating procedure. The characterization of the sintered samples was conducted in accordance with the workflow shown in
Figure 1. As can be seen from the presented workflow (
Figure 1), the number of steps in the additive route is one less than in the conventional one.
2.1. Materials
A commercially available BT powder was purchased from a local supplier (LLC Aril, Ufa, Russia). The powder was ball milled in a planetary mill Pulverisette 6 (Fritsch, Idar-Oberstein, Germany) for 0 to 3 h in the presence of 3 mm zirconia balls and isopropyl alcohol to study the powder size influence on 465 nm SLA processes and BT ceramics properties. The powder was then kept in a furnace at 70 °C for 3 h to dry off the isopropyl alcohol. There were made 7 powder batches: (“as is”, milling for 15 min, 30 min, 45 min, 60 min, 120 min, 180 min). The pre-processed powder was characterized and used for experimental testing following the workflow shown in
Figure 1.
2.2. SLA 465 nm 3D Printing
The experimental SLA 3D printing setup (
Figure 2) used an industrial 465 nm laser LDM450-3-12 (Purelogic R&D, Voronezh, Russia), originally designed for laser cutting and engraving operations. The laser was attached to FANUC M-1iA/0.5A (FANUC, Oshino, Japan) industrial robotic 6-axis manipulator to enable mechanical XY-motion of the laser beam. A vat and a printing platform were designed and printed from ABS plastic using an FDM 3D printer Ultimaker S5 (Ultimaker, Utrecht, Netherlands). The printing platform’s ascending and descending was implemented with the vertical mechanical drive based on a Nema17 Stepper motor KS42STH48-1684A (HANPOSE 3D Technology Co., Guangzhou, China). The printing operations with the experimental SLA setup were similar to those of the commercial SLA machine, except that the paste was applied and leveled by hands, while keeping the layer thickness of 100 μm. A detailed description of the experimental setup testing is given in our previous work [
20].
After milling, the powder was mixed in a mill with a dispersant DYSPERBYK-100 (BYK-Chemie GmbH, Wesel, Germany). Next, the prepared powder was mixed with an organic binder, designed for SLA-based 3D-printing, purchased from a local supplier (Moscow, Russia), in a planetary mill Retsch PM400 Planetary (Retsch GmbH, Haan, Germany), at a speed of 300 rpm, for 2 h to achieve a good homogeneity. The resulting compositions were also placed in light-blocking jars. There were 7 configurations of ceramic paste with the maximum volume filling with powder BT (
Table 1) prepared for testing the effect of powder sizes on process parameters and samples properties. The volumetric filling of the paste with BT powder depended on the particle size. The further increase of the solid content in the paste made it unspreadable. The samples were prepared in a shape of rectangular prisms with a side of 15 mm and height of 3 mm.
2.3. Semi-Dry Pressing
To compare the properties and microstructure of additive samples, disk-shaped samples with a diameter of 10–11 mm and a height of 2–3 mm were made by semi-dry pressing method. BT powders were used after milling, identical to those used for paste polymerization tests and obtaining 3D-printed samples. The composition of the press powder, the method of its preparation, and the molding modes during semi-dry pressing were selected based on the work results [
21]. A solution of 5 wt.% of paraffin in petroleum was used to prepare the press powder. After introducing the binder, the mixture was dried in a furnace at 70 °C to a constant mass for the petroleum evaporation and rubbed through a sieve with a cell of 300 microns for powder deagglomeration. Further, the obtained press powder was uniaxially pressed by the hydraulic press at 100 MPa in the stainless-steel die with an internal diameter of 12 mm.
2.4. Post-Processing: Thermal Treatment, Metallization, Polling
For a correct comparison of the properties and microstructure of conventional samples with 3D-printed ones, the binder burn-out and firing were carried out, together with additive samples, simultaneously, in a furnace. The temperature and time modes of binder burning and firing are shown in
Figure 3.
The polarization of BT ceramic samples was carried out in an air environmental oven using compressed air flow for samples and electrodes cooling, at a temperature of 120 °C, an electric voltage of 1500 V (pulse current), wherein exposure at temperature and voltage was 15 min, and then cooling to 60 °C under the electric voltage. All samples withstood the polarization voltage without damage.
2.5. Measurement of Powder and Samples Properties
The milled powder particle size distribution and specific surface area were characterized with laser diffraction method, using Analyzette 22 NanoTec (Fritsch GmbH, Idar-Oberstein, Germany). Powder morphology and ceramics fractured surfaces microstructure were studied by scanning electron microscopy (SEM), using JSM-6390LA (JEOL Ltd., Tokyo, Japan). The powders primary particle size, as well as ceramic samples grain size, were assessed by the line-intercept method from the SEM images. Phase analysis of the initial powder and crashed ceramic samples was conducted with Rigaku D/Max-2500 (Rigaku Corp., Tokyo, Japan) X-ray diffractometer. The patterns were recorded using CuKα radiation in a range of 10°< 2θ< 70° with a step of 0.02°. Phases were identified with the use of Crystallography Open Database (COD) [
22]. GSAS software was applied for fitting of diffraction pattern profiles by Le Bail method and for the refinement of unit cell parameters by Rietveld method [
23,
24].
The ceramic samples in a shape of cylinders or rectangular prisms manufactured by traditional and additive methods, respectively, were polished to achieve plane and parallel bases and height of 1.5 ± 0.05 mm. During the polishing, the height of samples was controlled with the use of a frame. Then, the bases of the samples were metallized with silver paste and heated up to 800 °C for 15 min. Measurements of resistance, capacitance, and tangent of dielectric losses were carried out with the Immitance Meter RLC-781 05G (Good Will Instrument Co., Ltd., New Taipei City, Taiwan), at 1 kHz of AC frequency. The dielectric permittivity was calculated from the capacitance and geometrical parameters of samples. The piezoelectric module d33 was measured by a quasi-static method using the D33 Test Meter (Sinoceramics, Inc., Shanghai, China).
Density of ceramic samples was measured by Archimedes method. To study microhardness, BT ceramic samples were framed with epoxy pellets and polished with subsequent chemical etching by HNO3 water solution. Samples microhardness was measured with Nanovea PB1000 (Nanovea Inc., Irvine, CA, USA) analytical setup. Berkovich micro-indentor was used at the maximum load of 10 N. The loading rate was 1 N/min.
4. Discussion
The results described above showed that SLA technique with a wavelength of 465 nm allows manufacturing of BT ceramics with adequate phase, microstructural, piezoelectric, and mechanical properties from a commercially available powder with a micron particle size, which is commonly used in large-scale industrial manufacturing of different BT ceramic products. A comparison with the materials prepared in parallel by the traditional ceramic route revealed a great potential of additive manufacturing approach. In both preparation methods, rather high density (over 95%), as well as tetragonality, of BT ceramics was achieved, promising for satisfactory functional characteristics. The 3D printing technique appeared advantageous over the traditional one in suppressing of the undesired AG growth in TiO
2-excessive matrix due to the difference in green body compaction. This feature resulted in predictable relation of microhardness and density, as well as in stable and comparatively low dielectric losses. Along with piezo module d
33 and dielectric constant ε, the dissipation factor tgδ is sensitive to microstructure of piezoceramics and determines their suitability for any application [
33,
34]. Low dielectric losses are the desired characteristics for most of piezoceramics applications. In the current work, the complex of piezoelectric properties performed by the 3D-printed materials with the relative density over 75% (d
33 of 36.5–148.0 pC/N, ε of 1695–2055, and tgδ of 0.018–0.028) appeared comparable to that of the conventionally-manufactured samples (d
33 of 50.0–105.0 pC/N, ε of 2355–2525, and tgδ of 0.018–0.144), if not taking into account conventional low-density samples, fabricated from non-milled powder. Comparatively high dielectric permittivity of the conventional samples could be attributed to the conductivity which generated the increased losses, as well.
The microstructure analysis showed that the conventionally-manufactured samples from the initial non-milled powder differ markedly in grain size and relative density from other samples. The distinctive features of these microstructures are small grains (on average 1–3 microns), tightly sintered into a porous frame. The reason for this is probably the morphology of the initial powder, which is dense sintered agglomerates of particles in sub-micron size (
Figure 6 and
Figure 7). After compaction by pressing, such agglomerates do not allow dense packing of grains, and grain growth during the sintering is limited (
Figure 15a). Grains of this size in BT ceramics [
35] provide maximum values of the piezo module d
33, exceeding 300 pC/N, and also may be caused by intensification of flex-tensional deformations [
32]. This can explain the significant difference in the piezo module of the samples studied in this work. In addition, unlike the microstructures of the BT-6-P1 sample (
Figure 15b), which has a piezo module of d
33 = 90 pC/N, there is no abnormal grain growth that reduces piezoelectric properties, despite the high density of ceramics. On the other hand, milling of the initial powder, which has an excess of TiO
2 in relation to BaO, leads to a denser packing of particles during compaction and intensive mass transfer during sintering, which provokes AG growth.
Comparison of microstructures of samples fractured surfaces from non-milled powder obtained with pressing BT-1-P1 (piezo module d
33 = 315 pC/N) (
Figure 15a) and 3D printing (sample BT-1-AM2), piezo module d
33 = 37 pC/N (
Figure 16a), shows a pronounced difference in density and grain sizes. It can be explained by the technological feature of the paste preparation for SLA 3D printing operations used in this study. The powder is mixed with an organic binder in a planetary mill. Due to the mixing, the initial agglomerates of the non-milled powder were partially broken (
Figure 16b), fine particles were tightly packed, and AG growth occurred, mainly oriented along with the photopolymerization layers. Probably, the approach of using a large particle size powder consisting of dense agglomerates with a size of about 1 micron could be successfully applied in the piezoceramics SLA 3D printing if the methods of paste preparation without high-energy mixing are applied. However, this assumption requires detailed research.
In addition to the difference in microstructure and properties between additive and conventionally-manufactured samples, there is a pronounced difference in the dependence of the relative density on the milling time and, accordingly, on the size and specific surface area of the powder particles (
Figure 11). For conventionally-pressed samples, the dependence is usual, i.e., with an increase in the milling time and a decrease in the size of powder particles, the relative density of sintered ceramic samples increases. For 3D-printed samples, a local maximum of relative density is observed at 45 min of powder milling, after which the relative density decreases. This regularity correlates with the local maximum photopolymerization depth achieved with the powder milled for 45 min. Probably, for the photopolymer composition used in this study, the characteristics of BT-4 powder (d
50 = 0.93 μm, specific surface area 110,538 cm
2/cm
3) (
Table 2) are the most suitable. A further decrease in particle size and an increase in the specific surface area leads to a deterioration of photopolymerization, probably due to intense light scattering. The use of BT powder in the SLA 3D printing at a wavelength of 465 nm requires a certain composition and concentration of the photoinitiator for particles of different sizes and different specific surface areas.
Comparison of this work results with the available data on the BT ceramics SLA 3D printing process features and the achieved piezoelectric properties (
Table 6) shows that using a 465 nm wavelength light source can significantly increase the photopolymerization depth of more than 50 vol.% BT paste without any adverse effects of the layers’ delamination during printing. In addition, the most considerable value of the piezo modulus, d
33 = 148 pC /N, was achieved using a micron BT powder with a mean particle size of 1.5 μm, which is more practically applicable in comparison with submicron and nanoscale powders used in the studies of other authors. Finally, measurements of the microhardness of additive and traditional samples made it possible to conduct the first preliminary assessment of the level of mechanical properties of BT piezoceramics, not previously presented in the literature.