A comprehensive review of the photopolymerization of ceramic resins used in stereolithography

In recent years, there have been rapid advances in our understanding of ceramic stereolithography (CSL) as a precise and high-resolution additive manufacturing (AM) technique to fabricate complex ceramic parts. This review highlights the theoretical background and engineering capabilities of CSL with an emphasis on photopolymerization of ceramic resins. We present certain constraints and characteristics designed to achieve optimal printability and photo-curability goals in ceramic resins and discuss in details about the parameters that can affect these properties. We then describe the current market status of CSL as well as its remaining challenges and promising future directions.


Introduction
Ceramic materials are used in a wide range of applications owing to their outstanding properties, such as high hardness and strength, good high-temperature performance, excellent thermal shock resistance, and high chemical stability in harsh conditions [1,2]. However, it has always been a challenge to shape and process ceramic parts because of their inherent hardness and brittleness [2]. Conventional shaping methods, such as dry pressing, tape casting, slip casting, gel casting, and injection molding, are not capable of producing highly complex structures with detailed features [3]. Moreover, most of these shaping methods require molds, and the sintered parts typically require mechanical processing [1].
In contrast, additive manufacturing (AM) offers new routes to overcome the aforementioned disadvantages of the conventional methods, and it fabricates highly complex ceramic components [4]. AM, also known as three-dimensional (3D) printing, is a rapid prototyping technology for manufacturing complex 3D parts without molds in a short time [5,6]. According to ASTM/ISO standardization, AM technologies can be categorized into seven distinct classes, including selective laser melting, selective laser sintering, fused deposition modeling, laminated object manufacturing, direct ink writing, and stereolithography (SL) [7]. These techniques use the same process to construct components: printing successive two-dimensional (2D) layers of the starting material on top of each other until the entire 3D part is built [8].
Among these potential techniques, SL fabricates high-quality parts with higher resolution and smoother surface finish [9,10]. In 1986, Charles Hull invented SL, which produces 3D plastic parts by selectively irradiating and solidifying (photopolymerizing) layers of liquid resins made of photocurable monomers [7]. In 1994, Griffith et al. combined SL and ceramic manufacturing processes to build 3D ceramic parts [1]. This adapted version of SL for the fabrication of ceramic parts, rather than plastic parts, is known as ceramic stereolithography (CSL). In CSL, conventional raw resins are loaded with ceramic powders acting as fillers, and the ceramic-loaded resins are typically called ceramic resins. The presence of ceramic fillers increases the resin viscosity and reduces resolution owing to light scattering. Although the challenges of CSL and SL are different, the concept of photopolymerization is the same for both raw and ceramic resins [11].
This review paper aims to provide fundamental knowledge and practical understanding of CSL. As both SL and CSL have the same principle of operation, this review first focuses on the basics of SL and later CSL and their related issues.
As shown in Fig. 1, SL can be categorized into two classes based on the mechanism of applying UV light to the resin layer and solidifying it: projection-based stereolithography (PSL) and scanning-based stereolithography (SSL).
The PSL approach cures the entire layer at once by using a light mask through which a flood lamp illuminates the resin surface ( Fig. 1  (a)). The light mask in this system is dynamically generated by a digital micromirror device (DMD), which produces an image of each layer. PSL is appropriate for higher resolution printing of small parts due to the limited size of the patterned laser light. Moreover, PSL requires less printing time because each layer is printed in a single shot. In contrast, the SSL approach uses a laser scanner to cure each layer ( Fig. 1(b)). A dynamic mirror system is used to focus and direct the laser beam over the resin surface to polymerize a set of elementary volumes known as strands. In general, SSL is suitable for large size printing, however, at the cost of resolution [13][14][15]. In general, SLA consists of five core components: vat, recoater, platform, optics, and control systems. In SL, the starting material is in the form of a liquid photocurable resin contained inside the vat, which is usually integrated with an automated refilling pump and a level adjustment device. Liquid resins used in SLA can be relatively viscous; therefore, the resin surface cannot become flat owing to gravity alone. Hence, the recoater blade is typically used to distribute the resin and flatten its surface uniformly. The platform system is composed of a building platform on which the part is fabricated, and an elevator, which raises and lowers the building platform [7,16]. The optics system consists of a laser source (Diode, He-Cd, Argon), an acoustic optical modulator to switch the laser beam on and off rapidly, a Z-focus lens to maintain the laser in focus on the resin surface, and two inertia galvanometers to scan the laser beam across the resin surface. The control system comprises three components: (1) a process controller to check the sequence of machine operations, (2) a beam controller to convert the operation descriptions into actions for adjusting the scanning speed and focus depth, and (3) an environment controller to adjust the temperature and humidity of the chamber that contains the vat [7,12].

Orientation approaches of stereolithography apparatus
SLA may have two orientation approaches, i.e., top-down and bottom-up, depending on the light projection technology used in the device. In the top-down approach, the laser source is positioned above the vat, and parts are fabricated facing upward (Fig. 2). The building platform is immersed inside the vat and positioned immediately below the resin surface. Only a thin layer of resin, with a specific thickness, is exposed to the laser source from above. Once a layer is cured, the building platform moves downward along the Z-axis by one-layer thickness to spread a fresh resin layer on top of the previously cured layer [17][18][19][20].
This set-up suffers from several problems. The downward movement of the platform disturbs the equilibrium of the resin level, consuming extra time for the recovery of the equilibrium state and reducing the efficiency of the entire procedure. Moreover, it is difficult to control the layer thickness in this set-up because only gravity is involved in leveling the resin surface after the deposition of a fresh layer on top of the previously cured layer. The resin surface can be leveled by using the recoater blade. However, this process is time-consuming. Hence, lowviscosity resins are recommended for use in this set-up [21,13,22]. Surface tension can also be a problem when leveling the resin surface around the already printed part. Furthermore, the fresh resin layer is  always in contact with the oxygen in the air, which inhibits the photopolymerization of certain monomers, resulting in incomplete curing and surface tackiness. Moreover, this approach requires a large vat and more resin [23,24].
The bottom-up approach provides excellent potential to overcome the problems encountered in the above set-up. Unlike in the top-down approach, photopolymerization is performed by an irradiation source beneath the vat (Fig. 3).
This orientation requires a vat with a transparent and non-sticky bottom for the gentle detachment of each cured layer [22,19,25]. In this set-up, the vat is shallow, and less material is required. The building platform is positioned inside the vat against the transparent bottom, leaving only a small gap with one-layer thickness. Once a layer is exposed to UV light from beneath, the cured layer is sandwiched between the vat bottom and the previously cured layer. The building platform rises each time a layer is cured to peel off the cured layer and allow fresh resin to flow in between the vat bottom and the last cured layer. Therefore, the fresh resin to be cured is less in contact with oxygen, resulting in a higher photopolymerization rate. The part height is not limited to the vat depth as the part is built facing downward. The layer thickness can be precisely achieved because it is accurately controlled by the elevator and not by the resin fluid properties. Thus, this technique offers higher vertical resolution and better surface quality [26,25,27,17]. The major drawback of this configuration is the cured layer adhesion to the vat bottom. Suitable coatings, such as Teflon and silicone films, can be applied on the transparent vat bottom to reduce the attachment force, resulting in successful detachment of the part from the vat bottom [13,28].

Fundamental parameters of stereolithography apparatus
The fundamental parameters of the SL process can be divided into two groups: (1) technical parameters, which can be adjusted using the SLA, such as laser power, laser scanning velocity, layer thickness, and hatch spacing, and (2) photosensitive parameters, which are the intrinsic properties of the resin, such as penetration depth and critical energy dose. These parameters have a substantial effect on the geometrical accuracy of the manufactured parts [29,30].
Conventional photocurable resin is considered a relatively transparent medium, allowing UV light to pass through it. The laser intensity distribution through this medium runs the photopolymerization process and determines the cured areas. The laser beam used in SL is Gaussian, which indicates that its intensity decreases from the beam center according to the Gaussian law. The Gaussian beam propagates along the Z-axis with a symmetry axis coinciding with the Z-axis. Fig. 4(a) shows the Gaussian distribution of light exposure, which is defined as the laser source radiant energy per surface unit [31]. Once the Gaussian laser beam scans a straight line at a constant velocity over the resin surface, a parabolic cylinder is solidified, which is often referred to as a strand ( Fig. 4(b)). The curing profile reflects the intensity distribution of the incident beam. Each strand has a specific depth and width known as the curing depth (C d ) and curing width (C w ), respectively [32,30]. These two parameters have a significant impact on the vertical and lateral resolutions [33,34].
In modeling the photocuring behavior, it is assumed that photocurable resins obey the Beer-Lambert law of exponential absorption. According to this law and as also depicted in Fig. 4 (b), the exposed light has its maximum energy value (E max ) at the resin surface (dark blue), and it exponentially attenuates as light penetrates through the resin absorbing medium (light blue). Therefore, it can be concluded that the UV light penetration through the resin is limited because of its dispersion and absorption. The extent of photopolymerization increases with UV light illumination until the resin reaches the gel point, at which it is converted from the liquid-state to solid-state. C d is defined as the depth at which light energy is sufficient to bring the resin to the gel point. The light energy at C d is called critical energy (E c ) and is required to initiate photopolymerization. As shown in Fig. 4(b), the resin reaches the gel point if the light energy is higher than E c ; otherwise, it remains a liquid [32,35]. C d is calculated using Jacob's version of the Beer-Lambert law: where D p is the penetration depth at which the beam intensity is reduced to 1/e 2 (37%) of its value at the resin surface. The photocuring behavior is characterized by plotting C d (μm) versus ln(E max ) (J/cm 2 ) using a semi-log plot. The result is a line known as the working curve, from which D p and E c can be calculated. D p is the slope of the line, and E c is its intercept with the X-axis.
Both D p and E c are purely resin parameters and are independent of the exposure. Therefore, both the slope and intercept of the working curve are independent of the technical parameters, such as laser power, laser spot size, and laser scanning speed [32,36].
In SLA systems, E max can be controlled using two adjustable parameters: light power and scanning speed, which determines the light exposure time. E max is derived using the following equation: In this equation, P is the laser power, w 0 is the beam radius, and v s is the scanning speed. As indicated by Eq. (2), E max is proportional to the laser power and inversely proportional to the product of the beam radius and the scanning speed. By combining Eqs. (1) and (2), it can be elucidated that C d can be controlled by changing the level of radiation applied to the resin surface [13,32].
In addition to C d , C w is a key parameter in the SL process, and it strongly affects the lateral resolution. As shown in Fig. 4 (b), the maximum cured width occurs at the resin surface where the width of the parabolic cylinder is largest. This parameter can be calculated using the following equation: As indicated by Eq. (3), C w is directly proportional to the laser spot size and the square root of the ratio of C d to D p . Thus, strands with greater C d are also wider. However, the relationship between them is not linear. Notably, when w 0 and C d are maintained constant, C w depends only on D p . This is useful to remember when the resin must be changed [32]. In addition to the aforementioned parameters, some other important parameters affect the resolution of SL, such as layer thickness, hatch spacing, and hatch overcure (Fig. 5). Layer thickness is one of the most important working parameters and is the depth of a layer solidified at the same elevation. This user-definable parameter is the thickness of the 2D sliced layers of the part CAD model. In the building process, the layer thickness is controlled by lowering/raising the building platform inside the resin vat for a predetermined, shallow thickness [30,38,37].
The hatch strands are located in the region within a layer sandwiched between the additional part layers [13]. The distance between the two adjacent hatch strands is known as "hatch spacing." If hatch spacing is very small, the solidified strands will overlap, resulting in a solid layer. In contrast, when hatch spacing is large, the liquid resin will be trapped inside the part, and it must be solidified in the following post-curing step [39,30]. Overcure is the depth through which each strand slightly penetrates the lower adjacent layer, and it increases the curing of that layer. Overcure is the main reason that keeps the individual layers connected together to construct a complete 3D part. To assist the bonding of the printed layers, C d should be equal to the layer thickness plus a certain amount of overcuring. Typically, overcure is suggested to be estimated as approximately 10%-35% of the layer thickness to avoid problems such as delamination and print-through.

Photopolymerization and photocrosslinking
Polymerization is defined as the generation of macromolecules with longer chain length through the continuous addition of monomers/ oligomers with shorter chain length. This addition can be induced by several stimuli, such as heat, electron beam, and light. The photo-induced polymerization process is a chain reaction in which one photon yields initiating species and induces the incorporation of thousands of monomer/oligomer units. This process is called photopolymerization, in which monomers in the form of a liquid are converted into a solid polymer in a few seconds [40][41][42]. Photopolymerization depends on not only light quantity but also light quality (wavelength). The wavelength range used in photopolymerization includes both UV (200-400 nm) and visible (400-700 nm) light. Most commonly, the wavelength required for the photopolymerization of resins is in the blue region. In some rare cases, infrared light (700-1000 nm) has also been used [42,43]. If monomers and oligomers are multifunctional, which indicates that they contain at least two functional groups in their structure, photopolymerization shows a complex behavior. In addition to the formation of a linear chain of monomers, a cross-link is formed between two linear macromolecular chains, forming a 3D network structure. This process is known as photocrosslinking, and it does not occur in the polymerization of monofunctional monomers [13,44,42]. Photopolymerization requires monomers/oligomers, which polymerize to form highly cross-linked polymer structures, and photoinitiators (PIs), which generate reactive initiating species upon UV exposure [45,13]. Hence, conventional raw resins used in SL consist of PIs (∼5 wt.%), monomers (∼25 wt.%), and oligomers (∼70 wt.%). Additional components, such as inert dyes, dispersion agents, inhibitors, fillers, and plasticizers, can be added into the resin in low concentrations to optimize the properties of the resin [42].

Photoinitiator
Most monomers/oligomers do not generate reactive species to initiate photopolymerization per se. Therefore, low-molecular-weight organic PIs are used in resins to generate reactive species to attack the functional groups of monomers/oligomers. Once the functional group is decomposed, each reactive double bond C=C can form a new bond with another carbon atom from a different monomer molecule. When the strong covalent bonds replace the weak van der Waals interactions between adjacent molecules, the liquid resin transforms into a solid structure with different bulk properties [47,13,46].
Free-radical PIs form free radicals after the absorption of the incident UV light, which immediately attack the double bonds of specific monomers, such as acrylates and methacrylates [13,48]. In contrast to free-radical PIs, cationic PIs produce acids when exposed to light. The generated acids readily react with a bond in specific monomers, such as vinyl ethers and epoxides, and induce polymerization [14].

Oligomers
Oligomers (prepolymers/macromonomers) are molecules with intermediate molecular weight possessing a larger chain structure consisting of a few monomer units. Oligomer is the main component of raw resins, and it determines the final desirable physical and chemical properties of the cured film. However, it is rarely used in concentrated ceramic resins as it is typically in the form of a viscous liquid, which makes it difficult to deal with [46,42].

Monomers
In raw resins, monomers are used as reactive diluents, and are mainly added to reduce the viscosity of oligomer and control the properties of the cured film [42,49]. In ceramic resins, they are the major component of the photocurable liquid resin due to their low viscosity. They can be categorized into two different groups depending on the type of reaction required to polymerize them.
Monomers based on free-radical reactions require free radicals to initiate polymerization. Therefore, only free-radical PIs can induce the photopolymerization of this type of monomers. Once a monomer accepts a free radical from PIs, it transfers that radical to another monomer to form a polymer. Acrylates and methacrylates are the most widely used monomers for free-radical photopolymerization [49].
Monomers based on cationic reactions require only cationic PIs to initiate photopolymerization. Several classes of monomers, such as epoxides, vinyl ethers, propenyl ethers, siloxanes, cyclic acetals, and furfural, polymerize under the cationic mechanism. Epoxide is the most famous monomer in this class [50].

Ceramic stereolithography
The addition of fillers, such as hollow glasses and ceramic microspheres, into a raw resin was introduced by Fan [51]. This idea was initially proposed to reduce the typically large C d of raw resins. Fan proposed that the dispersed filler particles in the resin can deflect the incoming radiation, resulting in a decrease in C d and an increase in resin opacity. The novelty of this work was later used in CSL, which is a promising technology for the fabrication of advanced ceramic parts with complex geometries because of its high resolution [52]. CSL has significantly advanced during the last two decades, making it possible to produce ceramic parts with demanding end-use applications rather than building only prototypes [53].
As shown in Fig. 6, the CSL process consists of the following main steps: preparing a suitable photocurable ceramic suspension, the building of the ceramic part, debinding, and sintering. The preparation of photocurable ceramic resins requires the homogenous dispersion of ceramic particles in conventional raw resin. The addition of ceramic powders significantly increases the resin viscosity; therefore, processing of ceramic resins becomes more difficult compared with that of raw resins [54,55]. The building of ceramic parts has similar steps as those in SL. During the photopolymerization of ceramic resin, monomers/ oligomers act as a binder and constitute a matrix around the ceramic particles by bonding them together. This matrix confers sufficient cohesion to the fabricated green body, which is a composite part consisting of polymer and ceramic. In general, the composite green body is stiffer and stronger than the pure polymeric part [56,19].
In this technique, obtaining a pure ceramic part requires that the composite green body be subjected to appropriate thermal treatments. First, the organic phase (binder) should be removed through a specific treatment step (debinding). Subsequently, a second thermal treatment (sintering) must be applied at higher temperatures to achieve acceptable mechanical properties. The resin thermal analysis should be performed before any thermal treatment. This step ensures that appropriate thermal treatment is selected for debinding, in which the green body is heated at slow rates, usually up to 500°C. Slow heating rates avoid typically high shrinkage rate, prevent fast gas formation, and maintain even temperature distribution during debinding. Therefore, the final properties of the end-product are not downgraded. Debinding is followed by sintering, in which the ceramic part should undergo heating treatment from 500°C to high temperatures depending on the sintering temperature of the ceramic powder. Sintering eliminates the voids in the ceramic part by shrinkage and thus increases its density [52].

Advanced ceramic materials used in ceramic stereolithography
The following requirements should be satisfied for the utilization of ceramic materials in SL [57,58,19,59]: 1 The ceramic powder should present a small variation of refractive index (RI) with one of the raw resins. In the UV wavelength range, ceramic materials have RI ranging from 1.56 (silica) to 2.6 (silicon carbide), and most monomers have RI of approximately 1.5. Therefore, it is recommended to utilize ceramic powders having low or medium RI, such as silica and alumina. 2 The ceramic powder should not strongly absorb light in the UV wavelength range because UV light should penetrate through the ceramic resin for the photopolymerization process. Thus, it is challenging in CSL to employ ceramic materials, such as titania and silicon carbide, which are highly absorbing and opaque at UV wavelengths. 3 The ceramic powder should have a median particle size (d 50 ) smaller than the layer thickness, which is typically in the range of 25 μm to 100 μm. This requirement improves the vertical resolution of the fabricated part. In general, the suitable particle size is suggested to be between 0.05 μm and 10 μm.

Light attenuation in ceramic resins
Conventional raw resins are homogeneous and relatively transparent to the incident laser beam used for photopolymerization. The C d of such resins is relatively large because light attenuation in such a medium relies solely on absorption, which occurs owing to the absorbing agents present, such as PIs and inert dyes. The absorption process is dependent on the concentration and extinction coefficients of the absorbing agents [66,67].
Homogeneous raw resins become heterogeneous and more complex when small ceramic particles are mixed with them [66,57]. The submicronic and micronic ceramic particles suspended in photocurable resins significantly disrupt the photopolymerization process because of light scattering. Scattering is an optical phenomenon in which the direction of light changes owing to the presence of non-uniformities in the medium. The main difference between ceramic resins and conventional resins is the scattering of the forward beam, which deviates some of the UV radiation from the forward direction [31,68,11].
As shown in Fig. 8, a photon is scattered by a suspended particle in a ceramic resin when it hits the particle. Light scattering occurs at interfaces with different RIs, and the scattering coefficient can quantify it. This parameter, defining the efficiency of particles to scatter light, is dependent on several variables, such as the median particle size, the number of particles, and RI contrast between the particles and medium. Ceramic powders with smaller particle sizes have more surface area and interfaces. Thus, more scattering occurs in ceramic resins composed of fine particles. In contrast, larger particles transmit more light. It is evident that, by increasing the number of particles per unit volume, more scattering events occur in the medium. Moreover, scattering increases when the RI contrast is increased, indicating that particles with higher RI scatter light into larger angles of deflection [69].
The scattered light is not lost from the exposed medium, unlike the absorbed light. It only changes its direction and contributes to other directions [38]. Therefore, both light scattering and absorption attenuate the energy intensity of the light traversing the resin medium with dispersed solid particles [68,70,32]. Tomeckova and Halloran [67] proposed a predictive model that describes the light attenuation in ceramic resins. According to their work, the attenuation coefficient (1/ D p ), which is the inverse of resin sensitivity or penetration depth, can be modeled by the following equation: where S describes the scattering effect of ceramic particles, A represents the light absorption by PIs and inert dyes, and ϕ represents the volume fraction of ceramic powder in the resin.
The term S is related to the inverse of the scattering length (l*), which is defined as the distance passed by a photon before its propagation direction becomes randomized. A small scattering length corresponds to a considerable degree of scattering, whereas a large scattering length indicates a minimal degree of scattering. A ceramic suspension with a large scattering length is expected to have a cured profile, which is similar to that of conventional SL resins [68]. Thus far, there has been no simple closed-form expression for the scattering length in highly loaded suspensions. However, it has been experimentally determined that S increases as the ceramic volume fraction increases up to its maximum amount ϕ max . Moreover, the term A is related to the properties of PIs and inert dyes as given by the following equation: As indicated by Eq. (5), the light absorption by PIs can be quantified by ε P c P , which is simply the product of the concentration (c P ) and extinction coefficient (ε P ) of the PI. The light absorption by inert dyes can be quantified by the product of the concentration (c D ) and extinction coefficient (ε D ) of the dye [71]. Therefore, Eq. (4) can be rewritten as [67] = + − + D S ϕ εc εc 1 (1 )( ) p P P D D As indicated by Eq. (6), the ceramic powder dispersed in resin plays two significant roles: it attenuates UV light by scattering through the term S, and it reduces the photopolymerization rate (conversion) by diluting the resin through the term (1 − ϕ) [71].

Effect of scattering on curing depth and curing width
When the laser beam with an ideal Gaussian lateral intensity distribution penetrates a raw resin, in which no scattering occurs, the curing profile reflects the intensity distribution of the incident beam. Therefore, the curing profile has a narrow bullet-shaped structure, whose depth is typically larger than its width ( Fig. 9(a)) [35,29].
In contrast, the curing profile of ceramic resins is significantly different from the lateral intensity distribution of the Gaussian beam [35]. The curing profile of a ceramic resin is shallower than that of a raw resin because of light scattering. Although a ceramic resin might be composed of UV-transparent ceramic powder, the radiation scattering process generates high turbidity in the medium, which reduces the C d [72]. Moreover, light scattering delivers more radiation in sideways directions and enlarges C w . Therefore, unlike raw resins with a narrow curing profile, ceramic resins exhibit profile broadening, which degrades the feature resolution. In ceramic resins, the curing profile is typically shallower and broader, and it has a mushroom-shaped structure ( Fig. 9(b)). Therefore, it can be concluded that the presence of solid particles in ceramic resins reduces the printing accuracy owing to light scattering [73,52,14]. It is necessary to control C d and C w to obtain dimensional accuracy because both are essential parameters to be considered for printing detailed structures.
As thoroughly explained in Section 2.3, the C d of raw resins can be expressed by Eq. (1). However, ceramic resins consisting of a high volume fraction of ceramic powder do not obey this equation [38] because light attenuation in ceramic resins is also controlled by scattering rather than merely absorption [73]. The C d of such resins can be expressed by the following equation [57]: where d is the mean particle size of the powder, ϕ is the volume fraction of the powder, β includes the particle size and laser wavelength, and Δn is the RI contrast, which is the difference between the RIs of the suspended ceramic particles (n c ) and the photocurable matrix (n 0 ). By comparing Eqs. (1) and (7), it can be concluded that D p can be expressed by Eq. (8), which includes the scattering effect [74]: where Q is a factor representing the capability of matter to diffuse radiation and is given by the following equation: where h is the interparticle distance, and λ is the wavelength of the irradiation. By combining Eqs. (8) and (9), it can be deduced that D p or resin sensitivity is affected by the properties of the ceramic resin, including the solid loading of ceramic powder, the size of the suspended particles, and the RI of both powder and liquid resin. D p depends on the properties of the ceramic resin because the light scattering phenomenon is affected by the same properties. If D p is large, then small C d can be achieved by decreasing E max . In contrast, when D p is small, small C d can be achieved by increasing E max [71].
Moreover, C w in the horizontal platform can be calculated using the following equation: As indicated by Eq. (10), C w is proportional to the beam spot size and is always larger than the laser beam diameter ( w 2 0 ), suggesting a significant influence of lateral scattering. Furthermore, C w always increases upon increasing C d [75].
It is evident from Eqs. (7) and (10) that both C d and C w are significantly affected by the working parameters of the SLA and the parameters related to the ceramic powder, such as its volume fraction, RI, and mean particle size. For instance, more UV light is scattered when Δn is large. Owing to this strong scattering, both D p and C d are reduced. Moreover, more scattering events cause more resin around the laser beam to become cured. Thus, C w increases, and consequently, the lateral resolution is reduced. Therefore, choosing an appropriate liquid resin for a specific ceramic powder to reduce Δn is a crucial step for the fabrication of detailed structures [14].

Suitable photocurable ceramic resins
An essential step in CSL is to prepare a suitable ceramic resin, in which a fine ceramic powder should be mixed with a monomer-containing solution. This system often requires the addition of a dispersant to maintain resin stability and prevent agglomerations. In general, current photocurable ceramic resins can be divided into two groups: aqueous (acrylamide-based) and non-aqueous (acrylate-and epoxybased) resins [57,52].
In 1994, Griffith et al. [59] introduced the first aqueous photocurable suspension used in CSL. The prepared aqueous acrylamide-based solutions had low RI (1.38-1.44). Thus, the Δn was large, resulting in a small C d . Therefore, the choice of ceramic powders to be mixed with the aqueous resins is limited. Although this group of ceramic resins has several advantages, such as low viscosity, freedom from volatile organic compound emissions, and the ease of dispersing oxide ceramic powders, it has one main disadvantage related to the low strength of the as-gelled green body. Moreover, a large amount of water present in the suspension introduces several problems, including cracking and drying of the green body [38]. This group of ceramic resins has not been widely used and studied in CSL owing to these problems [76].
In contrast, non-aqueous ceramic resins have the advantage of providing higher strength for the cured green body, compared with aqueous ceramic resins. Therefore, many researchers have investigated this group of ceramic resins [77]. The acrylate-and epoxy-based resins typically have larger RI (1.47-1.55) than aqueous resins, which is very beneficial because only a small increase in the RI of the monomer solution can dramatically increase the C d . Thus, even a low E max is sufficient to cure a given C d , resulting in faster operation time. Non-aqueous ceramic resins are favorable for the processing of non-oxide ceramics because of the problems faced owing to the hydrolysis and oxidation of non-oxide particle surfaces in aqueous resins [38]. However, this group of photocurable ceramic suspensions has the limitation of having higher viscosities than those of aqueous suspensions when the same volume fraction of ceramic powder is used in the resin formulations [77].
It has been proven that the same formulation cannot be used for the preparation of other ceramic resins with various ceramic powders or monomer solutions. Hence, it is recommended that, for each ceramic type or compound, a specific formulation should be individually examined [72]. Thus far, numerous studies related to the development of ceramic resins have been conducted. Some of these studies are summarized in Table 1, which includes only some of their partial or final successful results. Such developmental research is crucial as the final quality of the fabricated part is strongly dependent on controlling the rheological and photocuring characteristics of the ceramic resin. In general, a photocurable ceramic resin should satisfy a few requirements to be printable including having high solid loading, low viscosity, and adequate C d .

High solid loading
In powder processing and consolidation of ceramics, the aim is to achieve high green density to produce ceramic parts with excellent dimensional and structural integrity. For the fabrication of high-quality and dense ceramic parts, the ceramic powder content or loading in green bodies should be as high as possible [54,108]. This requirement should also be satisfied in CSL because the as-gelled green bodies with low ceramic loading suffer from excessive shrinkage and delamination during the debinding and sintering steps [48]. Hence, typically, 50 vol. % ceramic powder is sufficient, but even higher loadings are desired. It is highly recommended to reduce the organic concentration of the suspension because high solid loading improves the final properties of the end-product by [54]: • minimizing porosity, which limits the strength of the part. • accelerating the debinding process and reducing the risk of part disruption.
• reducing sintering shrinkage and avoiding deformation and cracking.
• providing dense and homogeneous sintered ceramic parts.
The sintered ceramic part can have a density up to 99.9% of its theoretical density if the prepared suspension has sufficiently high solid loading and contains no air bubbles, and laminar defects are avoided while building and processing the part [54].

Low viscosity
Ceramic resins used in CSL should be highly mobile to produce ceramic parts with complex geometries within a reasonable processing time [48]. In other words, the resin viscosity should be relatively low, preferably with no pronounced yield point. This condition is required to facilitate the recoating and self-leveling of the resin [73]. Ceramic resins should be as fluid as conventional raw resins, with viscosities less than 3000 mPa·s, to conduct the CSL process in an SL resin vat [72]. This situation corresponds to a ceramic loading of less than 50 vol.%.
However, as previously explained, ceramic resins should have high solid loading in the range of 50-65 vol.%, which typically increases the viscosity. This is in contrast to the good homogeneity and low viscosity of the resin [48,56].
Therefore, it is imperative to use a dispersant to manage particle-particle interactions and maintain a low viscosity if the solid loading of the ceramic powder is more than 35 vol.%. Dispersants are usually used for different purposes, such as stabilizing ceramic suspensions, deagglomerating, and preventing sedimentation [57]. A colloidal dispersant must be carefully selected to disperse the high solid loading of the ceramic powder effectively while maintaining the viscosity below the upper limit [72,54]. The viscosity of 3000 mPa·s at the shear rate of 10 s −1 is commonly accepted as an upper limit for a processable ceramic resin used in CSL [48]. Some researchers have claimed that the viscosity of 5000 mPa·s can be considered as the upper limit [56,38]. The upper limit of viscosity is commonly used for the determination of solid loading [48]. Ceramic resins used in CSL should exhibit shear-thinning behavior, which is necessary for the recoating process [72,109]. In general, the rheology of ceramic resins is affected by several parameters, such as, dispersant, solid loading, dilent, particle size and particle shape (BET value). Some of these parameters are in more depth investigated for the preparation of suitable ceramic resins.

Dispersant.
In a study [56], Hinczewski et al. prepared different slurries containing 53 vol.% alumina in diacrylate monomer solution with different amounts of dispersant. It was observed that the slurries with the dispersant concentration of less than 1.2 wt.% exhibited shear-thickening behavior, which is unfavorable for the CSL process. However, the slurries with dispersant concentration more than 1.2 wt.% exhibited shear-thinning behavior. The minimum viscosity was obtained when 2.2 wt.% dispersant was added to this particular resin. The same procedure has been used in many studies to achieve the minimum viscosity for different ceramic resins and dispersants. This should be noted that, the effect of the dispersant on the resin rheology depends on the ceramic powder properties and the composition of the dispersant and monomer. Therefore, the amount of the dispersant should be carefully selected for each resin.

Solid loading.
In a study performed by Griffith et al. [72], different solid loadings of silica powder were dispersed in aqueous solutions. It was observed that, as the solid loading was increased, viscosity also increased. The suspension with 50 vol.% solid loading was observed to have the highest solid content whereas it had a viscosity below 3000 mPa.s. The authors also stated that, when silica was dispersed in an acrylate-based resin, solid loadings even below 30 vol.% resulted in viscosities higher than 3000 mPa·s at low shear rates. Table 1 The main properties of some of the suitable ceramic resins for the CSL process and the energy and wavelength of the laser. In an experimental work by Liao [38], alumina and silica were dispersed in different non-aqueous resins. The same relationship between solid loading and viscosity was observed, which can be justified by the fact that the greater the particle packing in the resin, the more it hinders the resin flow. The prepared 45 vol.% alumina resin and 50 vol.% silica resin had the highest powder content while maintaining viscosities below 5000 mPa·s. The effect of solid loading on resin viscosity has extensively been investigated in other studies as well [98,61] 4.4.2.3. Diluent. Diluents are generally added to resins to reduce the viscosity of the monomer-containing solution, without modifying its reactivity toward UV light. Large addition of diluents (> 40 wt.% of the resin weight) results in the fabrication of soft parts owing to the nature of diluents. In a study by Hinczewski [56], the optimal amount of diluent in the formulation of resin was determined based on the viscosity and UV curability of the highly loaded resins. It was observed that the alumina resin with 80 wt.% solid loading and 30 wt.% diluent (with respect to the resin weight) had the viscosity of 18 Pa·s, which is three times less than the viscosity of the resin without any diluent.

Adequate curing depth
In CSL, ceramic parts are built from layers with a thickness of 10-200 μm. The C d of the resin must be slightly greater than this range of layer thickness to fabricate a ceramic part within a reasonable time. High spatial resolution is achieved when the resin has a small C d and C w . In CSL, it is recommended to have a minimum C d of 200 μm. However, some researchers have cured layers with a thickness of 150 μm to construct smoother surfaces and finer features at the cost of fast construction [72,33]. As previously discussed, C d is dependent on the ceramic properties and working parameters, some of which are explained in this section. 4.4.3.1. Refractive index contrast. As indicated by Eq. (7), C d can be varied by changing Δn as it is inversely proportional to Δn 2 . Therefore, the RIs of ceramic powder and resin both affect C d [59]. Many studies have been conducted with scrupulous attention to the effect of ceramic materials, with varying RI, on the photopolymerization process [110]. It has been proven that the photopolymerization of resins with ceramic materials with low RI, such as SiO 2 (n = 1.56), or medium RI, such as Al 2 O 3 (n = 1.70), can be adequately performed. In contrast, the photopolymerization of ceramic materials with high RI, such as Si 3 N 4 (n = 2.10) and lead zirconate titanate (PZT) (n = 2.4), is challenging because of their short scattering lengths; consequently, more scattering events occur in the resin medium. This strong scattering effect results in a shallow C d , which reduces the bonding between the layers, and a broad C w , which reduces the lateral resolution. The scattering effect can be minimized by using ceramic powders with an RI close to that of the liquid resin [70].
The effect of Δn on the curing profile of a cured line is shown in Fig. 10. This effect was also investigated in a study [111] in which silica suspensions with varying RI for monomer solutions were prepared. It was observed that, at the E max of 203 mJ/cm 2 , the ceramic resin with low Δn had larger C d and smaller C w , whereas the ceramic resin with high Δn had shallower C d and larger C w .
In a study [82], silica (n = 1.46), alumina (n = 1.76), and zirconia (n = 2.05) powders were dispersed in the same liquid resins with solid loading of 60, 60, and 52 vol.%, respectively. It was observed that the zirconia suspension had lower C d than the silica and alumina suspensions owing to its larger RI. Layer thickness must be chosen appropriately such that it is slightly smaller than C d to ensure proper bonding between the layers. Therefore, based on the obtained values of C d in this study, the layer thickness of the silica, alumina, and zirconia suspensions were set to 100, 75, and 25 μm, respectively.
Another study [107] focused on the stereolithographic fabrication of complex ceramic parts with silicon nitride (Si 3 N 4 ) powder, which has a large RI and, therefore, low C d . The authors claimed that the C d of the suspension with raw Si 3 N 4 powder was 34 μm at the E max of 500 mJ/ cm 2 . A surface oxidation approach of Si 3 N 4 was proposed to improve the C d . It was observed that, after oxidizing the powder at 1150 and 1200°C for 1 h, the C d was effectively increased to 42 and 51 μm, respectively. However, particular applications require only certain ceramic materials having high RI. In this case, it is possible to reduce the scattering effect by increasing the RI of the resin so that Δn is small [70,112]. A study [72] investigated the effect of different monomer systems with varying RI to achieve a desirable C d . In this study, 50 vol.% silica powder was loaded in different aqueous solutions with various RIs. It was shown that, as Δn 2 decreased from 0.0317 to 0.0142, C d increased from 250 to 700 μm at the E max of 1500 mJ/cm 2 . In other words, only a small change (4%) in the resin RI resulted in a three-fold increase in C d .

4.4.3.2.
Volume fraction of ceramic powder. As indicated by Eq. (7), C d is inversely proportional to the volume fraction of ceramic powder. Experimentally, it has also been demonstrated that, by increasing the volume fraction of ceramic powder into the resin, C d decreases. Griffith et al. [72] prepared photocurable resins by adding different amounts of silica powder into an aqueous solution with the RI of 1.3824. It was observed that, at the E max of 1500 mJ/cm 2 , the suspension with 50 vol. % silica powder had a C d of 360 μm, which is higher than the minimum requirement (200 μm). The C d decreased to 330 μm when the resin was loaded with 55 vol.% silica powder. The authors also investigated the photopolymerization of a silicon nitride suspension, which had high RI and shallow C d . The suspension loaded with 10 vol.% ceramic powder had the C d of 21 μm, whereas it decreased to 10 μm when the suspension was loaded with 20 vol.% ceramic powder. This example proves the challenge of using ceramic materials with high RI in CSL.

Particle size and size distribution.
According to the light scattering theory, both mean particle size and size distribution of the ceramic powder strongly affect light scattering. Therefore, the resolution of CSL is dependent on the median particle size and distribution of the ceramic powder [31]. Suspensions prepared with coarser particle size have greater C d compared with suspensions prepared with finer ceramic particles, owing to the lower degree of scattering [113]. Although nanosized powders require lower sintering temperatures and improve the final properties of the end-product, they are unfavorable in terms of C d [31].
In a study by Sun et al. [], the Monte Carlo simulation was used to model C d and C w as a function of mean particle size ranging from 0.3 to 1 μm. It was shown that C d decreases by reducing the mean particle size. The light scattering effect became significant when the particle Fig. 10. The schematic view of the curing profile for two ceramic photocurable resins with low and high RI contrast (Δn) being exposed to the same exposure (Inspired by [35]). size approached the light wavelength (0.364 μm). Moreover, it was observed that the strong scattering effect caused by broad particle size distribution results in poor spatial resolution. In contrast, the narrow size distribution of ceramic particles is more desired for the fabrication of fine and smooth surfaces [70].
In contrast, in an experimental study by Griffith et al. [72], different grades (0.46 and 0.61 μm) of alumina powder were used to determine the effect of particle size on C d . It was shown that both suspensions with 50 vol% solid loading had C d greater than 200 μm. However, the suspension with the smaller particle size had a larger C d , i.e., 400 versus 300 μm. This result could not be reconciled with simple scattering theories, such as the Rayleigh-Gans theory and Mie theory, which relate particle size to light scattering for very dilute suspensions. The authors empirically discovered that, for highly concentrated suspensions, C d is strongly affected by β in Eq. (7), which is related to the square of particle size (d 2 ). Therefore, the authors stated that C d is inversely proportional to particle size for highly concentrated resins.

4.4.3.4.
Photoinitiator. C d is often considered to be synonymous with monomer-to-polymer conversion. Literature shows that, as the PI concentration increases, conversion and C d increase. In the case of low PI concentrations, the conversion to polymer is too low to form a gel; therefore, C d approaches zero. At the other extreme of high PI concentrations, the penetration depth of photons decreases. Therefore, there must be an intermediate optimal concentration to achieve adequate C d [114]. Chartier et al. [74] evaluated the influence of the PI concentration on the C d of alumina, silica, and zirconia suspensions with different PI concentrations ranging from 0.3 to 1 wt.% (with respect to the monomer weight). The result of this study showed no significant effect of the PI concentration on the C d and C w of the prepared suspensions. However, the laser beam was highly absorbed by the quickly polymerized surface at high PI concentrations (> 2 wt.%).

Energy dose.
The effect of E max on the curing profile was investigated in a study by Gentry et al. [68]. It was shown that, by increasing the E max (54,203, and 608 mJ/cm 2 ), both the C d and C d of the silica suspension with low Δn were increased. In another study [115], three different acrylate monomers were loaded with 60 vol.% silica powder. It was concluded that one way to increase C d is to increase E max . Another study [29] investigated the same effect and showed that both C d and C w were decreased as E max was decreased by increasing the scanning speed. When the E max was decreased from 15387.7 to 76.9 mJ/cm 2 , C d decreased from 244.00 to 85.60 μm, and C w also decreased from 1051.31 to 153.58 μm. Another work [74] showed that the dimensional resolution of the printed parts can be increased by decreasing the E max of the applied laser.

Ceramic stereolithography market: its current status, challenges, and future
The market for additively manufactured high-performance ceramic parts is relatively new and rapidly growing. Companies developing ceramic resins and providing manufacturing services have realized the ability of the CSL technique to produce ceramic parts with complex geometries, which can be used in different industries, such as aerospace, jewelry, electronics, dental, and biomedical fields. Therefore, significant investments are being made by the major suppliers of SLA systems in the market of CSL, such as 3DCeram Sinto, Lithoz, Admatec, and Prodways. Table 2 summarizes the main features of some of the SLA systems manufactured by these companies. Moreover, these companies offer ready-to-use ceramic resins for sale, some of which are mentioned in this table.
Although CSL is considered to provide the highest resolution among ceramic AM techniques, it suffers from considerable limitations. Despite the increasing number of available ceramic resins in the market, this technique is still restricted to the use of one resin at a time, which ultimately limits its potential applications. In fact, there are activities to use multiple resins at a time, however, it requires complex sequential polymerization and cleaning steps between the material changes [19].
Moreover, the manufacturing time of the conventional SLA printers is relatively long. Several approaches have been proposed to solve this issue. The scanning process itself in SSL systems restricts the decrease in the manufacturing time. This could be overcome by using precise and narrow scanning patterns only for areas where high resolution is required while using a broader scanning pattern for bulk features. The combination of SSL and PSL techniques is currently being developed and explored. These hybrid systems can further reduce the manufacturing time to rates similar to PSL while maintaining the high precision of laser illumination [116].
The printing resolution of PSL is dependent on the size of the micromirrors in DMD and the optical system. Generally, a higher resolution can be more easily achieved in a smaller printing area by using suitable optics. Therefore, the size of the building platforms used in such systems is relatively small. Hybrid systems can lead to a compromise between the size of the building platform and resolution. The large-area projection micro-stereolithography (LAPμ SL) is an image projection micro-stereolithography system that produces very small features over large areas rapidly. LAPμSL combines the advantages of SSL (large area but poor resolution) and PSL (high resolution but only over a small area), enabling the rapid printing of fine details over large areas. This technology uses optical techniques to write images in parallel by moving the DMD, in contrast to other competing technologies, which either require the rastering of beams to expose pixels in series or mechanical stage moves. Thus, centimeter-sized components with a micrometer-range resolution can be fabricated at a low cost. It is believed that these hybrid systems can expand the application range of CSL in the future after thorough investigation and development [14,116].

Conclusion
Among AM techniques, CSL is most widely used for the fabrication of ceramic parts owing to its high accuracy in printing complex geometries. SLA fabricates a 3D ceramic structure by printing layers of ceramic resin successively. Each 2D layer of the liquid ceramic resin, with a defined thickness between 25 and 100 μm, is exposed to the laser irradiation source and then becomes solidified. The solidification process is based on the photopolymerization of the liquid resin, which consists of a ceramic powder, an organic monomer, such as acrylate and epoxy, PIs, and other additives to achieve desirable properties. Once each layer of resin is exposed to UV light, the laser beam supplies the energy required to induce photopolymerization and photocrosslinking. The fabricated 3D structure is referred to as a composite green body consisting of polymer and ceramic. The green body undergoes debinding to burn out the contained polymer and achieve a pure ceramic part. This process is followed by sintering to obtain a dense ceramic part.
A critical step in CSL is to prepare a suitable ceramic resin for printing. The fine ceramic powder should be homogenously mixed with a monomer-containing solution, which often requires the addition of a dispersant to maintain the resin stability. A suitable photocurable ceramic resin used in CSL should satisfy a few requirements. First, the solid loading of the ceramic powder should be more than 40 vol% to avoid excessive shrinkage and delamination during the debinding and sintering processes. Second, the ceramic resin must be highly mobile for the fabrication of ceramic parts via SLA within an optimum manufacturing time. Typically, it is recommended that the ceramic resin should have low viscosity below 3000 mPa. s to ensure the self-leveling and recoating of the resin. Third, it is highly recommended that the C d of a highly loaded suspension should be equal to or greater than 150-200 μm after being exposed to UV radiation. C d is one of the most important parameters in CSL because it strongly affects the vertical and lateral resolutions. This parameter is defined as the depth at which the gel point is achieved and is strongly affected by the scattering phenomenon owing to the presence of ceramic particles. The scattering phenomenon is dependent on the median particle size, volume fraction, and RI of the powder. Therefore, the same variables control C d .