Ceramic Stereolithography: Additive Manufacturing

MR46CH10-Halloran

ARI

V I E W

Review in Advance first posted online on March 17, 2016. (Changes may still occur before final publication online and in print.)

N

I N A

Annu. Rev. Mater. Res. 2016.46. Downloaded from www.annualreviews.org Access provided by Gazi Universities Main Library (D-S) on 03/22/16. For personal use only.

14:22

C E

S

R

E

8 March 2016

D V A

Ceramic Stereolithography: Additive Manufacturing for Ceramics by Photopolymerization John W. Halloran Department of Materials Science and Engineering, University of Michigan, Ann Arbor, Michigan 48109; email: [email protected]

Annu. Rev. Mater. Res. 2016. 46:10.1–10.22

Keywords

The Annual Review of Materials Research is online at matsci.annualreviews.org

photopolymerization, stereolithography

This article’s doi: 10.1146/annurev-matsci-070115-031841

Abstract

c 2016 by Annual Reviews. Copyright All rights reserved

Ceramic stereolithography and related additive manufacturing methods involving photopolymerization of ceramic powder suspensions are reviewed in terms of the capabilities of current devices. The practical fundamentals of the cure depth, cure width, and cure profile are related to the optical properties of the monomer, ceramic, and photo-active components. Postpolymerization steps, including harvesting and cleaning the objects, binder burnout, and sintering, are discussed and compared with conventional methods. The prospects for practical manufacturing are discussed.

10.1

Changes may still occur before final publication online and in print

MR46CH10-Halloran

ARI

8 March 2016

14:22

INTRODUCTION


Additive manufacturing (AM) of ceramics can be accomplished with a wide variety of methods, typically involving a layered manufacturing approach. Three-dimensional objects are produced by combining many layers, with each layer patterned individually according to the design of the object. Some of the prominent ceramic AM methods involve photopolymerization of ceramic suspensions to pattern the layers. There have been several generations of ceramic AM involving patterning using laser scanning and image projection. This review concerns AM by photopolymerization, which is often termed ceramic stereolithography. Stereolithography, strictly defined, is a process whereby each layer is patterned by laser scanning on photopolymerizable medium to convert liquid monomer into solid resin (1). The method has been adapted to form ceramics by using a photopolymerizable suspension of ceramic powders in place of the liquid monomer. I expand the term stereolithography to include all cases in which ceramics are made with multilayer photopolymerization, including cases involving layers patterned by laser scanning and by mask image projection. Photopolymerization is not a new ceramic process. Photopolymerized tape casting systems were described in 1986 (2), and photopolymerized dental ceramic composites are commonplace (3). Photoimageable tape cast products are commercial products (4). The technique was applied for AM of ceramics using commercial laser scanning stereolithography machines (5, 6) or novel variants of laser scan devices (7, 8). Patterning can also be done by mask image projection (9, 10), typically with a spatial light modulator based on a digital mirror device (DMD) (11, 12). Applications are varied and include biomedical implants (13, 14), ceramic prototypes (15), cellular ceramics (16), complex investment casting cores (8, 17, 18), and integrally cored investment casting molds (19). The process has been adapted for microscale ceramic forming as microstereolithography (20–22) for a variety of applications (23–25). Several excellent reviews on AM of ceramics (26–29) have been recently published. These reviews cover the broad range of AM techniques used for ceramics and comprehensively present the current status of the field. Because the capabilities of various AM methods are available, those aspects of AM are not repeated in detail here. Instead, this article focuses on methods that involve photopolymerization of ceramic suspensions. The emphasis is on the practical fundamentals of the polymerization process as well as on the other steps necessary to produce accurate ceramic objects by stereolithography. To provide the reader with some context, below are some early and recent examples of ceramics built with stereolithography, illustrating the range of designs that can be realized. Figure 1 is an early example of a ceramic sculpture. It is a digital sculpture by Michael Rees that was built in alumina (Al2 O3 ) by laser scanning stereolithography by Alan Brady in 1998. This piece is approximately 5 cm wide and consists of several lobes and appendages. This design could not have been executed in one piece without using an AM method. Producing it by conventional methods would have required assembly of separately formed pieces. Figure 2 shows a lattice structure built from Al2 O3 by Johannes Homa in 2014. The structure is patterned with a fixed DMD using a Lithoz device. The object is approximately 3 cm tall, with struts as fine as 500 µm. This design could have been built only by AM, as it could not have been cast or molded. Figure 3a is an example of an investment casting core, which was built by Brady from Al2 O3 by laser scanning. This object is approximately 6 cm tall. The design of this core is based on a commercial injection molded core, so although it is a complex shape, it is an object that can be molded. It illustrates the ability to achieve state-of-the-art features without using injection molding tools. Compare Figure 3a with Figure 3b, another airfoil core, which was built in refractory silica

10.2

Halloran


MR46CH10-Halloran

ARI


a

8 March 2016

14:22

b

Figure 1 Digital sculpture by Michael Rees that was built in Al2 O3 by Alan Brady in 1998 by laser scanning stereolithography. (a) Top view, approximately 5 cm wide. (b) Detail. Reproduced with permission from the artist.

Figure 2 Lattice structure in an Al2 O3 -fixed digital mirror device approximately 3 cm tall with 500-µm struts. Figure courtesy of Johannes Homa.

www.annualreviews.org • Ceramic Stereolithography


10.3

MR46CH10-Halloran

ARI

8 March 2016

14:22


a

b

Figure 3 (a) Investment casting core that was built from a design based on an injection molded core. It was built in 1998 by Brady in Al2 O3 using laser scanning and is approximately 5 cm tall. (b) Investment casting core built in 2014 by Suman Das in refractory SiO2 using a scrolling digital mirror device. The core illustrates features typical of a state-of-the-art core, including rods for film-cooling holes in leading-edge serpentines with turbulence-enhancing surface features and core holes and a flag for the trailing edge.

(SiO2 ) in 2014 by Suman Das using scrolling DMD. This object is approximately 10 cm tall and has several fine features that could not have been produced with a two-piece mold. Figure 4 shows a 30-GHz Luneburg lens designed by Brakora & Sarabandi (30) and built in Al2 O3 by laser scanning by Walter Zimbeck. In this design, focusing is accomplished because the 30-GHz refractive index gradually changes as the thickness of the Al2 O3 struts varies from the inside to the outside. This object could not have been fabricated without AM, because the function of the lens required the Al2 O3 volume fraction (strut size) to vary precisely from center to exterior. Figure 5 shows a ceramic box with triple-helix array. This structure was built from SiO2 , with patterning using a scrolling DMD. The object has 168 intertwined helices, with 56 columns of three helical rods; each SiO2 rod is 120 µm in diameter. This box illustrates both fine features and the realization of an unmoldable design that was achieved by ceramic stereolithography.

PROCESSING BEFORE PATTERNING Ceramic AM techniques accomplish only the shaping step, as they are simply forming methods for the ceramic. They make only the ceramic green shape, i.e., the part before sintering, and not the finished ceramic object. So such techniques accomplish approximately one-third of the complete task, with the powder and body formulation done before AM and with sintering, microstructure development, and finishing done after AM. From the ceramic perspective, ceramic stereolithography is just a forming process for a green ceramic. The valuable properties of the ceramic product 10.4

Halloran



MR46CH10-Halloran

ARI

8 March 2016

14:22

Figure 4 A millimeter-wave Luneburg lens built from Al2 O3 . The lens was designed by Karl Brakora and was built by TA&T with laser scanning stereolithography. The lens is approximately 6 cm in diameter, with Al2 O3 strut width ranging from 650 µm in the center to 340 µm at the surface, which varies the refractive index at 30 GHz to focus the millimeter-wave radiation.

depend on the composition and microstructure, which are determined by sintering behavior as determined by particle size and heat treatment. Stereolithography is thus the same as any binderrich ceramic forming technology and is comparable to ceramic injection molding or ceramic tape multilayer fabrication. Much of the ceramic technology in stereolithography is similar to conventional ceramic technology, although it is adapted to work in the stereolithography process. Photopolymerizable

Figure 5 An object built in refractory SiO2 ceramic by a scrolling digital mirror device. This 40-mm box has an array of 168 intertwined helices, with 56 columns of three helical rods each with 120-µm diameter. Figure courtesy of Michael Middlemass, DDM Systems. www.annualreviews.org • Ceramic Stereolithography


10.5

ARI

8 March 2016

14:22

ceramic suspension technology is essentially a type of gel casting (31). As with gel casting (32), the green density of the formed body is determined by the ceramic volume fraction in the suspension. For adequate sintering behavior, the green density (and therefore the suspension volume fraction) must be 50 vol% or higher. So a key feature of the ceramic technology is achieving the proper rheology of highly loaded suspension (33). Fluid suspensions are used with some stereolithography systems (5). Other systems use more viscous pastes (7). The forming behavior is the same, but there are important differences in the mechanical design of these systems. The key difference between conventional gel casting and patterning by photo-gel casting is largely the gelation itself, which occurs via photopolymerization. Photopolymerization of ceramic suspensions involves several unique features, as discussed below. The layer-by-layer application method is another key difference between conventional gel casting and stereolithography. However, the application of the very thin and uniform liquid layer for stereolithography is similar in some ways to conventional tape casting. Further steps are required after all the layers are complete. The nature of these steps depends on whether the part is self-supported and built from above (see, e.g., Figure 2) or whether it was immersed in uncured suspension (see, e.g., Figures 1 and 4). There may be temporary support structures that must be removed (see, e.g., Figure 3a). If the object has trapped volumes (see, e.g., Figure 5), the uncured liquid must be drained. If the design has a dense array of small passages (see, e.g., Figure 4), the small spaces must be cleared of uncured suspension. The object is often just one part in a multipart build (see, e.g., Figure 3b). Multipart builds are one advantage of stereolithography, as many different objects can be built during the same run. This approach can significantly increase productivity, as one run lasting several hours can yield dozens of different objects, each closely nested to make productive use of the space within the build volume. For these multipart builds, the individual objects must be liberated from the build volume. Next it is necessary to remove the uncured suspension from the exterior and interior surfaces. Removing uncured suspension from the exterior surfaces (see, e.g., Figure 1) or relatively open structures (see, e.g., Figure 2) involves little cleaning. But removing the uncured suspension from the interior of complex parts can be a challenge, particularly for parts with tight interior features, such as the cores in Figure 3, the thin gaps between the struts in Figure 5, or the small spaces between the helices in Figure 5. After building, the part is a ceramic green body. As with ceramic injection molding, the space between the ceramic particles becomes nearly saturated in a polymer gel binder. The binder removal must be done by a well-controlled pyrolysis, followed by high-temperature sintering. These steps are the same as for conventional ceramic injection molding (34, 35). Binder removal is easy for designs with thin sections (Figure 2) but is more challenging for designs with thick sections (Figure 1) or a combination of thick and thin sections (Figure 3b). The rheological behavior of the photopolymerizable suspension is particularly important for the deposition of the fresh layers, which are typically only approximately 100 µm thick. The fresh layers must be strictly uniform for high-accuracy builds. It is necessary to control the rheology over a very wide range of shear conditions. During application of the fresh layer, the shear rate can dramatically change as the coating device passes from a thick uncured area to a thin liquid layer on a solid cured area. The fluid dynamics associated with this sudden change in shear rate is a well-known issue for conventional stereolithography (36). Much of the fresh layer is applied over the previously cured solid layer, whereas in other regions it is applied over a portion that has no solid layer, causing perturbations in the fluid suspension. The fresh liquid thickness over a previously cured solid layer can be ∼0.1 mm or less, whereas the total liquid thickness where there is no submerged solid can be ∼100 mm. As the coating device moves from a deep liquid to a thin liquid, shear rate and therefore shear stress can change by a factor of 1,000 or more. This change causes fluid dynamic instabilities in the liquid suspension and can create damaging stresses to


MR46CH10-Halloran

10.6

Halloran


MR46CH10-Halloran

ARI

8 March 2016

14:22

small solid features. The well-known trapped volume problem also presents fluid dynamic issues that make it difficult to achieve an ideally smooth layer over a complex surface with liquid and solid regions. Build strategies are incorporated to mitigate the unfavorable fluid dynamic effects that occur when one is applying and leveling suspension over partially completed layers. The rheology of the suspension at rest is important for leveling and for preventing sedimentation of coarse particles. These rheological challenges are addressed by using the principles of colloidal processing familiar from conventional ceramic forming. Fortunately, for most ceramic powders suspended in acrylate monomers, the colloidal behavior is often simpler than aqueous suspensions, and acrylate suspensions can be simple fluids (33, 37).


BUILDING MULTILAYER OBJECTS As with most AM schemes, the macroscopic object is built from many thin patterned layers. The typical layer thicknesses range from 25 to 250 µm. Thinner layers enable more detail and more accurate resolution, but at a cost of build time, because each layer takes a certain amount of time for recoat and for exposure. For laser scanning stereolithography, the patterning to define the design of each layer (and its ancillary support structures) is done by controlling the motion of a laser beam, with galvanometer mirror scanners. The patterning is defined by the instructions that control the scanning of the laser beam and is similar to drawing with a pencil. The actual polymerized layer is the superposition of edge contours and interior hatching. The lateral (x, y) resolution is determined by the focal width of the laser, as broadened by scattering. Laser scanning devices build the object from long, thin rows, with areas defined by cross-hatching. The hatching style has a second-order influence on lateral resolution. The typical lateral (x, y) resolution is on the order of 25–50 µm. Very small features may appear in the original design files but may be too small to define with the hatching resolution (46). Patterning with projected images is quite different. The image is typically projected as a bitmap of the complete layer or a portion of the layer. The patterning image is projected from a fixed DMD or a scrolling DMD. The lateral resolution depends upon the projection optics and scales with the characteristic spatial resolution of DMDs, which usually have 17-µm pixels. Fixed DMDs focus an image with the spatial resolution of DMDs. The fixed DMDs are best suited for relatively small objects, as they pattern a region several centimeters wide without loss of resolution by expanding the image. The more complex scrolling DMDs can pattern areas as large as 60 cm without loss of resolution, as the exposure head moves over a wide surface with scrolling images. After patterning, a new fresh layer of photosuspension must be applied. This step can be done by applying a liquid coating on the top with a blade. On other devices, the fresh layer can be applied from the bottom. Another device dips the partially completed build into a suspension. The most common light sources are UV lasers, UV and blue LEDs, and visible light. Ceramic stereolithography can be accomplished in a conventional commercial SLA machine by replacing the polymer resin with a suitable photopolymerizable ceramic suspension. However, several investigators and companies have produced purpose-build devices for ceramics. These devices can be relatively compact machines aimed at prototype or small runs of small parts, such as the Lithoz CeraFab 7500 (Lithoz GmbH, Vienna) and the 3DCeram FCP (3DCeram, Limoge), or quite large robust machines aimed at industrial production, such as the CTC6060 (DDM Systems, Atlanta). The minimum feature size in the build direction (z) depends on the layer thickness h. To build a part with a height H in the build direction requires n layers, where n = H/h. Relatively small objects require hundreds or a few thousand layers. Large parts, such as 1-m-tall industrial gas www.annualreviews.org • Ceramic Stereolithography


10.7

MR46CH10-Halloran

ARI

8 March 2016

14:22

turbine cores, may have 10,000 layers. The build time scales with n because each layer requires application of fresh liquid in a recoat step, which requires a certain time trecoat , typically 10–60 s, and because patterning also takes a certain time twrite , which requires a similar amount of time: tbuild = ntrecoat +

n

twrite,i = n(trecoat + twrite,ave ),

1.

1


where twrite,ave is the average time for writing a layer. For patterning with laser scanning, the write time depends on the laser power and the energy dose needed for polymerization, which controls the scanning speed for the laser. Scan speed is usually very fast, but if the dose is high and the laser power is low, scan speed can be rather slow. The time to write each layer depends on the area to be polymerized, and polymerization can be quite different for various layers in the design (38). For patterning with a fixed DMD, the write time is typically the same for all layers and depends on the power of the illumination and on the dose of radiation required for polymerization.

PRACTICAL FUNDAMENTALS OF PHOTOPOLYMERIZATION OF CERAMIC SUSPENSIONS All these devices operate by solidifying selected areas of the liquid by polymerization, so the phenomena associated with photopolymerization of ceramic suspension are common to all of them. Of primary interest to this review are the following practical fundamentals: (a) what determines the width and depth of the exposed feature (line or pixel); (b) how the depth and width of the exposed feature after polymerization depend upon energy dose and on the photoproperties of the suspension; and (c) how the photoproperties of the suspension depend on the nature of the ceramic powder, the active agents, and the inert dyes. In most cases, the monomer in the cured layers is not fully polymerized. There is a variation in the extent of polymerization with vertical position in the layers, which can be considered to be a cure profile. The variation of the cure profile with exposure conditions is discussed below for the case of single layers. During patterning, photoinitiators are activated in the illuminated areas. Photoinitiators typically create free radicals, which initiate the polymerization of the monomers. The extent of polymerization (α) increases with illumination until the suspension reaches the gel point, at which it is a soft solid (α = α gel ). The intensity of the light attenuates as it propagates into the depth beneath the surface (depth z), so the energy dose received at each depth [E(z)] varies. The depth at which the gel point occurs is the cure depth (Cd ). The cure depth is the depth z at which the energy dose suffices to bring the degree of polymerization to the gel point (α gel ), and the energy dose at that depth is the critical energy dose Ec . The relationship between the cure depth Cd and the applied energy dose E is expressed by E Cd = Dp ln , 2. Ec where the parameter Dp , which has units of length, is related to an attenuation length. The practical behavior of the photosuspension is determined by the magnitude of the critical energy Ec , which is the minimum energy required to initiate the polymerization and the magnitude of the attenuation length parameter Dp . Equation 2 defines the working curve that is used to control the polymerization depth. Figure 6 has measurements for the cure depth Cd as a function of energy dose E, plotted according to Equation 2 so that the Dp parameter can be measured from the slope and the Ec parameter can be obtained from the energy dose intercept. These measurements were taken for suspensions containing 60-vol% SiO2 powder in a hexane diol diacrylate suspension containing variable amounts of ketone photoinitiator and triazole dye. Notice that the slopes, 10.8

Halloran


MR46CH10-Halloran

ARI

8 March 2016

14:22

3,000


Cure depth, Cd (μm)

2,500

0.5% PI 0% UV absorber Dp = 1,010 µm Ec = 19 mJ cm–2

2,000

1,500

3% PI 0.15% UV absorber Dp = 188 µm Ec = 25 mJ cm–2

1,000

500

0

1

10

100

1,000

Energy dose, E (mJ cm–2) Figure 6 Cure depth versus energy for suspensions containing 60-vol% SiO2 powder in a hexane diol diacrylate suspension containing variable amounts of ketone photoinitiator (PI) and a triazole dye.

which are the Dp values, are quite different for the two formulations, as are the Ec values from the intercepts. It is useful to understand how these important parameters depend on the formulation of the suspension, the photoinitiators and dyes, and the nature and amount of the ceramic powder. The cure depth parameters can be expressed in terms of the concentration of the photoinitiator (39, 40). The influence of other contributions on Dp can be accurately modeled by a simple attenuation model (40). The attenuation model is expressed in terms of the attenuation by absorption of UV photons and by scattering of photons, which redirects photons from the forward beam (with the scattered photons broadening the beam). The critical energy dose is modeled with an inhibitor exhaustion model (40), which assumes that the critical energy is the minimum dose needed to overcome native inhibitors so that free radicals can be available for polymerization.

Attenuation Length Parameter Dp The attenuation model describes Dp as an attenuation length. As the UV beam passes beneath the surface, it is attenuated by absorption from inert dyes (which simply absorb photons) and photoinitiators (which absorb photons and generated free radicals). Absorption depends on the concentrations and extinction coefficients of the absorbing species. The UV forward beam is also attenuated by scattering from the particles, which depends on the ceramic volume fraction, particle size, and the refractive index difference between the ceramic and the liquid. The attenuation coefficient is 1/Dp : 1 = S + (1 − )(εP c P + εD c D ), 3. Dp where is the ceramic volume fraction in the suspension; cP and εP are the concentrations of the photoinitiator; cD and εD are the concentration and molar extinction coefficient of the dye; and www.annualreviews.org • Ceramic Stereolithography


10.9

MR46CH10-Halloran

ARI

8 March 2016

14:22

S describes the effects of light scattering, which is related to the inverse of the scattering length. There is no simple closed-form expression for scattering length in concentrated suspensions. For ceramic suspensions, S is experimentally found (41, 42) to increase with ceramic volume fraction until it reaches a maximum at a volume fraction max . This behavior can be empirically fit to obtain an approximate expression for S(): β S() = β − 2 , 4. 2max


where β is a fitting parameter that depends strongly on the refractive index difference between the ceramic and the liquid. This approximation allows for an expression for the attenuation coefficient that includes the volume fraction ceramic powder: 1 β 2 . = εP c P + εD c D + (β − εP c P − εD c D ) − 5. Dp 2max Figure 7 compares the prediction of the attenuation model (Equation 3) with measured values for a number of SiO2 suspensions cured with two different UV sources. The predicted values use data for the extinction coefficients from spectrophotometry. The agreement is satisfactory. For a constant dose, the cure depth Cd varies as 1/(n)2 , where n is the refractive index 0 = ncern−n . contrast, the difference in refractive index between the powder and the liquid n n0 0 An example is shown in Figure 8 for the case of SiO2 in aqueous media of varying refractive indexes (40). Photopolymerization works well for ceramics with low refractive index (like SiO2 ) or medium refractive index (like Al2 O3 ) but is more difficult for materials with high refractive index (like lead zirconate titanate) (43, 44), for which scattering lengths are very short. The expressions above are for UV-transparent ceramics. Some ceramics, such as SiC or TiO2 , are opaque at 1,200

60-vol% UV laser 50-vol% UV laser 60-vol% UV lamp

Predicted Dp (μm)

1,000

800

600

400

200

0

0

200

400

600

800

1,000

1,200

Measured Dp (μm) Figure 7 Values of the attenuation parameter predicted from the attenuation model (Equation 3) compared with measured values for 60-vol% SiO2 cured with a scanning UV laser (red circles), 50-vol% SiO2 cured with a scanning UV laser (blue triangles), and 60-vol% SiO2 cured with a medium pressure mercury UV lamp ( yellow squares). 10.10

Halloran


MR46CH10-Halloran

ARI

8 March 2016

14:22

800

700


Cure depth, Cd (μm)

600

500

400 50-vol% SiO2 in aqueous solution 300

200

100

0

0

20

40

60

80

Refractive index contrast, (Δn)–2 Figure 8 Cure depth at constant dose for 50-vol% suspensions of SiO2 powders, measured in acrylamide solutions in water and ethylene glycol, with the refractive index modified with ethylene glycol concentration. Cure depth is inversely proportional to the square of the refractive index difference between the SiO2 and the solution.

UV wavelengths and so are strongly absorbing. Absorption lengths can be very short in opaque ceramics, so photopolymerization cannot be easily done.

Critical Energy Dose Parameter Ec The second parameter in the working curve is the critical energy dose Ec . Polymerization occurs when radicals created by the photoinitiator are present in sufficient quantity to initiate the polymerization reaction of the monomers. But commercial monomers are stabilized against premature polymerization by the addition of inhibitors, whose function is to remove free radicals. Oxygen often acts as an inhibitor, particularly near the surface, where oxygen is absorbed from the air. A model for Ec was based on the presumption that, for UV doses smaller than Ec , the smaller number of free radicals created when a photoinitiator molecule absorbs a photon at low doses are consumed by inhibitors so that polymerization does not occur. This model also considers absorption of photons by inert dyes. The hypothesis of this model for critical energy is that Ec is given by the dose of UV photons that are not absorbed by inert dyes and that produce a population of free radicals large enough to exhaust the population of inhibitors in the monomer solution to provide surplus free radicals to begin the polymerization reaction. The critical energy then depends upon the number and effectiveness of inhibitors, on the concentration and absorption coefficients of inert dyes that absorb photons, and on the concentration and absorption coefficients of photoinitiators that create free radicals when they absorb photons. The ceramic powder plays a passive role by diluting the active monomer and an active role by limiting photon penetration depth by scattering. The critical energy dose can be approximated by Ec = (1 − )

hν 1 (γQ Q + γO O + γD c D ) . εP c P

6.



10.11

MR46CH10-Halloran

ARI

8 March 2016

14:22

100 90 80 70

Ec (mJ cm–2)

0.2% UV absorber 60 50 40


30 20 0% UV absorber 10 0

0.0

0.5

1.0

1.5

2.0

2.5

cPI–2 Figure 9 Critical energy dose for Al2 O3 suspensions in acrylate monomers as a function of the inverse concentration of ketone photoinitiator (PI).

Here hν/ is the photon energy and quantum efficiency; is the ceramic volume fraction; cP is the concentration of the photoinitiator; cD is the inert dye concentration, with γ D describing the effectiveness of removing photons by dye absorption; and Q is the concentration of quinonetype inhibitors and O is the concentration of oxygen inhibitors, with γ Q and γ O the effectiveness of removing radicals by these two inhibitors. More details are available in the original paper (41). We expect from Equation 6 that the critical energy dose should be inversely proportional to the photoinitiator concentrations. This prediction agrees with the observations in Figure 9, which shows the critical energy dose for Al2 O3 suspensions in acrylate monomers, as a function of the inverse concentration of ketone photoinitiator, at two different levels of inert dye. The observations fit well enough to make Equation 6 a useful model.

LINE WIDTH AND BROADENING FROM SIDE-SCATTER In the absence of scattering, the profile of the cured line reflects the intensity distribution of the beam. A laser beam with an ideal Gaussian lateral intensity distribution has a bullet-shaped cured line. For ceramic suspensions, the cured line shape deviates significantly from the lateral intensity distribution of the beam. Cured lines in ceramic suspensions can be broader and mushroom shaped (17, 45). Some of these shapes are illustrated in Figure 10, which shows an experiment by Gentry (46), in which a square beam of UV illuminated a SiO2 suspension through a glass slide. After sufficient energy dose, the shape of the cured line profile was observed. The refractive index contrast of the monomer solution was modified. With low refractive index contrast, the cured line has the same width as the slit, but significant broadening occurs if the refractive index contrast is high. The resolution of small features during patterning can be no better than the resolution projected from the bit map, but it can be degraded by broadening. Ceramic suspensions can suffer excess broadening, which degrades feature resolution. For laser scanning, the excess broadening 10.12

Halloran


MR46CH10-Halloran

ARI

8 March 2016

14:22

a

b

c

d

e

f

Schematic of curing

Low RI contrast E0 = 54 mJ cm–2



High RI contrast E0 = 203 mJ cm–2

High RI contrast E0 = 608 mJ cm–2

Area illuminated Glass slide

500 µm


Cured polymer line adhering to slide

Figure 10 Shape of the cured line in SiO2 suspensions with varying refractive index (RI) values for monomer solution. Panel a illustrates how a square-wave beam from a slit propagated into the suspension through a glass slide cures a broadened line. The cured line shape is similar to the square wave with low RI contrast (panel b) but is increasingly broadened with higher RI contrast (panels c–f ). The scale bar (500 µm) applies to panels b–f. Adapted from Reference 46.

is related to the energy spread of a Gaussian beam (1), which is exacerbated by scattering (47). With DMD projection, the broadening phenomenon is different, and the broadening can depend upon the feature size and upon cure depth and dose (48). For square-wave illumination, the excess broadening is the product of a characteristic width (the width sensitivity) and the logarithm of the ratio of the energy dose and a width critical energy (49). Excess broadening is not important for shallow cure depths but becomes important for deeper cure depth. There is less broadening in highly absorbing suspensions, such as those with a high photoinitiator or inert dye concentration. Excess broadening is more severe for high-refractive-index ceramics. Jacobs (1) addressed the line width for laser-scanned stereolithography for the case of a beam with a Gaussian intensity distribution of width W0 . The width of the cured region measured at the surface, w cure , is related to energy dose Emax , to the critical dose Ed , and to the Gaussian beam width W0 by √ Emax . 7. wcure = 2W 0 ln Ec Ceramic suspensions differ because scattering of the forward beam directs some UV radiation away from the forward direction by side-scattering. Photopolymerization by the side-scattered beam causes additional broadening. Hinczewski et al. (47) have proposed an empirical modification of Equation 6 for the line width of ceramic suspensions, cured with Gaussian beams, which fit three ceramic suspensions containing Al2 O3 , zircon (ZrSiO4 ), or SiO2 . Gentry & Halloran (49) observed that the broadening illustrated in Figure 10 can be described with a Beer-Lambert absorption of the side-scattered beam. An excess width (w ex ) due to broadening is defined as one-half the difference between the measured line width (w) and the width due to the illumination (wbeam ): w − wbeam . 8. wex = 2 The relationship between the energy dose E0 and the excess broadening can be described with a quasi-Beer-Lambert relationship: E0 , 9. wex = Dw ln Ew www.annualreviews.org • Ceramic Stereolithography


10.13

MR46CH10-Halloran

ARI

8 March 2016

14:22

1,750

Cure distance (μm)

1,500 1,250 1,000

Db

750 Cd

Wex

500


250 0

1

10

100

1,000

Energy dose (mJ cm–2) Figure 11 Cure depth (Cd , teal ) and excess width (w ex , yellow) as a function of energy dose for a 40-vol% SiO2 suspension in hexane diol diacrylate with 0.05 mol/L of a ketone photoinitiator. The broadening depth (Db ) is also indicated.

where Dw is considered the width sensitivity, the resin sensitivity parameter for side-scattered light. The parameter Ew is the width critical energy dose (critical energy for polymerization in the horizontal direction). This behavior arises because the suspension is a Beer-Lambert absorber, so polymerization from the horizontal beams from scattering would be similar to polymerization by the forward beam. Thus, the broadening behavior would also fit a semilogarithmic model. Figure 11 shows the cure depth and excess width as a function of energy dose for a SiO2 suspension. The width cure parameters (Dw , Ew ), however, are not the same as the cure depth parameters (Dp , Ec ). The broadening parameters (Dw , Ew ) that describe excess width vary with the amount of scattering by ceramic particles and with the factors that affect the absorption and critical energy dose of the monomer solution. Scattering depends on the refractive index difference between the ceramic powder and monomer, the solid loading of the ceramic suspensions, and the size distribution. The photoinitiator and dye determine the depth sensitivity and critical energy dose of the monomer solution. Broadening by side-scatter can be described in terms of broadening depth Db , which is the magnitude of cure depth Cd at the onset of broadening (where E = Ew ). In terms of the suspension parameters, Ew . 10. Db = Dp ln Ed Db varies with the logarithm of the refractive index contrast (n/nq ) and can be fit to an expression of the form n/n0 , 11. Db = −B1 ln B2 where B1 is broadening strength and B2 is a broadening index. The broadening strength describes how much the broadening depth changes with refractive index difference. A small value of B1 indicates a small change in the broadening depth with refractive index difference, whereas a large value of B1 indicates a large change in the broadening depth with refractive index difference. The experimental values of B1 and B2 depend on, for example, the photoinitiator concentration and incident intensity. 10.14

Halloran


MR46CH10-Halloran

ARI

8 March 2016

14:22

1,400 1,200

SiO2

Db (μm)

1,000 800 600 400 Al6Si4O13


200 0 0.001

Al2O3 ZrSiO4 0.01

0.1

1

Δn/n0 Figure 12 Effect of refractive index contrast (n/n0 ) on the broadening depth (Db ), as measured for suspensions containing SiO2 , Al6 Si4 O13 , Al2 O3 , and ZrSiO4 powders.

Figure 12 shows measurements for the broadening depth, from suspensions with three powders of different refractive indexes in the same monomer solution, plotted as a function of the logarithm of the refractive index difference. The refractive index of SiO2 is close to the refractive indexes of the acrylate monomers, so the broadening depth is large for the SiO2 suspension (1,040 µm). It is possible to use cure lines up to 1 mm deep with SiO2 before excess broadening becomes a problem. The broadening depth is 285 µm for mullite (Al6 Si4 O13 ) and 143 µm for Al2 O3 . ZrSiO4 has a higher refractive index, so broadening occurs even at a rather shallow cure depth of 55 µm. Similar behavior is observed with SiO2 suspensions when the refractive index of the monomer solution is changed.

Cure Profile: Variation of Monomer Conversion with Depth Attenuation occurs below the surface, so the energy dose received by the suspensions varies with depth (z) as E(z). The degree of polymerization, or monomer conversion, depends on the energy dose [α(E)], so monomer conversion can vary with depth. This variation in monomer conversion leads to a nonuniform cure profile, α(z). Such uniformity can be concerning when a large, complex object is being built from hundreds of sequential layers. There is no simple expression for the variation of monomer conversion with dose α(E), but the relation between conversion and dose is typically sigmoidal (50–52), similar to what is shown in Figure 13. The suspension behaves as a liquid (50) for small doses, for which the monomer conversion is less than the gel point (α gel ), ∼20% monomer conversion. We can identify the dose at which the suspension gels as the familiar critical energy dose Ec , so α(Ec ) = α gel . The monomer conversion increases with dose, and the polymer gel stiffens as the degree of polymerization increases. Eventually, α reaches an asymptote at maximum monomer conversion α max , which is usually ∼80% monomer conversion. Physically, α max corresponds to the point at which the reaction freezes because the glass transition temperature of the gel reaches ambient temperature. The maximum monomer conversion occurs at some energy dose Emax , where α(Emax ) = α max . Further illumination for doses larger than Emax does not cause further reaction, because the kinetics are too slow (although the reaction is not complete). www.annualreviews.org • Ceramic Stereolithography


10.15

MR46CH10-Halloran

ARI

8 March 2016

14:22


Degree of monomer conversion (%)

100

80

αmax

60

40

20

0

αgel

Ec

Energy dose, E

Emax

Figure 13 Degree of monomer conversion versus energy dose.

The top surface of a cured layer has α = α max . The bottom of a layer (z = h, the layer thickness) cannot be polymerized much beyond α gel , or else unwanted solid is cured under the layer, a condition known as print-through, which degrades the fidelity of the build. Thus, a stereolithography build, including hundreds of layers, can have within each layer a nonuniform cure profile. This nonuniformity can cause difficulties, such as cracking before or during binder removal, during subsequent processes (53, 54, 56). As dimensional shrinkage accompanies polymerization, serious distortions or stresses can develop (36). Consider now the subsurface energy dose as it attenuates with depth beneath the surface, E(z), which for a Beer-Lambert absorbing system is −z , E(z) = E0 exp Dp where E0 is the surface dose and Dp is the sensitivity of the suspension. Figure 14a illustrates variation of dose with depth for sensitivities ranging from a low-value Dp = 50 µm, which attenuates over a short distance, to a large-value Dp = 100 µm, which has much less variation in dose through the thickness. Combining E(z) with α(E), fit to typical values for a SiO2 suspension, Figure 14b is the approximate cure profile for a suspension with Dp = 1,000 µm that has received four different doses ranging from 100 to 700 mJ/cm2 . This would be the case for curing a layer in which the solid lies over uncured liquid (such as an overhanging feature). Superimposed onto Figure 14b is a vertical dashed line for a particular depth z = 100 µm, which is the layer thickness in this case. Also shown are horizontal dashed lines for E = Ec , below which the suspension is a liquid (α < α gel ), and E = Emax , above which the reaction is frozen and the degree of monomer conversion is α max . Dose versus depth can be combined with monomer conversion versus dose to illustrate the trade-off associated with the choice of surface dose E0 and layer thickness for a given sensitivity Dp . Figure 15 shows the calculated degree of monomer conversion versus depth for a suspension with a Dp of 100 µm and three surface doses: 100 mJ/cm2 (E1 ), 300 mJ/cm2 (E2 ), and 500 mJ/cm2 (E3 ). At the highest dose (E3 ), the monomer is fully cured and therefore uniform to below the 100-µm layer. However, an additional ∼80 µm were partially cured below the layer, so this case 10.16

Halloran


MR46CH10-Halloran

8 March 2016

14:22

1.0 0.9

Dp = 1,000 µm

0.8 Dp = 500 µm

0.7 0.6 0.5

Dp = 200 µm

0.4 0.3 0.2

Dp = 50 µm

0.1 0.0

0

50

100

150

500

b

450

Energy at depth Z, E(z)

Fraction of initial dose, E(z)/E0

a

E4

400 350

250

Emax

250

E2

200 150 E1

100 50

200

Layer thickness

E3

300

0

300

Ec 0

50

Depth, Z (μm)

100

150

200

Depth, Z (μm)

Figure 14 (a) Normalized energy versus depth in micrometers for several different values of sensitivity Dp . (b) Energy versus depth for a given Dp for several values of surface dose E0 , for doses 100 mJ/cm2 (E1 , teal ), 300 mJ/cm2 (E2 , blue), 500 mJ/cm2 (E3 , red ), and 700 mJ/cm2 (E4 , yellow).

is badly overcured. This case would degrade resolution by print-through. Regarding the E2 dose, there are ∼30 µm of solid beneath the layer, so the monomer is slightly overcured, and there are also ∼30 µm of solid material within the layer that has excess residual monomer. There is some undercuring. For the lowest dose (E1 ), there is an ∼50-µm region of undercured solid near the surface, with a liquid layer beneath the solid.

POSTBUILDING OPERATIONS After the final layer is cured, a number of important processing steps remain. The parts must be collected from the build volume in a harvesting operation, and any uncured suspension must be removed in a cleaning operation. Harvesting is simple for single objects (like Figure 2),

100 Layer thickness

90 80

Degree of monomer conversion (%)


ARI

αmax

70 60 50 40

E2

E1

E3

30 20

αgel

10 0

0

50

100

150

200

Depth, Z (μm) Figure 15 Schematic of conversion versus depth for three surface doses E0 of 100 mJ/cm2 (E1 , blue), 300 mJ/cm2 (E2 , red ), and 500 mJ/cm2 (E3 , yellow) for a suspension with Dp = 100 µm. www.annualreviews.org • Ceramic Stereolithography


10.17

ARI

8 March 2016

14:22

which need only be removed from the build platform. A design-like lattice in Figure 2 is selfsupporting, so there are no ancillary support structures to remove. Cleaning is easy, particularly in this case, in which the fresh suspension is applied from the bottom. Designs such as the core in Figure 3b require a more elaborate harvesting procedure. This complex shape has several non-self-supporting features, and the ancillary support structures must be carefully removed. Also, for this case, this core was built along with dozens of other objects in a large build volume, whereby different objects were nested around others in the build. Harvesting this core involved disassembling the nest of sibling builds. The cleaning operation was also more difficult for the Figure 3 cores because uncured suspension had to be removed from the small holes and channels without harming the thin posts and sections. Clearing the uncured suspension from the many small passages between the struts of the Luneburg lens (Figure 4) also required care. After the object is harvested and cleaned, it is ready for binder burnout and sintering. The cured polymer is removed by pyrolysis during binder burnout. The binder burnout process is similar to other ceramic forming methods that use a high level of binder systems, like multilayer ceramic capacitors and injection molding (34, 35). The challenges of binder burnout depend on the section thickness. Binder burnout is relatively easy if the minimum section size is a few millimeters and is more of a challenge if the section size is several centimeters. With ceramic stereolithography, some additional concerns involve light curing strategy (cure profile) (56) and residual monomer content (53). When pyrolysis is complete and all the residue from the polymer is removed, the rest of the sintering process is the same as any other ceramic process.


MR46CH10-Halloran

INDUSTRIAL-SCALE ADDITIVE MANUFACTURING BY STEREOLITHOGRAPHY Stereolithography can produce complex and sophisticated ceramic objects, but can it be used to manufacture ceramic products? Ceramic stereolithography may be effective as a toolless forming method for small production runs. In this case, Ceramic stereolithography is accomplishing what could be done by ceramic injection molding, but without the cost and delay to acquire hard tooling. However, considering only tooling undervalues the technology because molding has significant shape limitations. The value of ceramic stereolithography, or any AM process, is enabling the production of designs that cannot be molded or otherwise fabricated. These unmoldable designs are often never considered, because traditional design practice penalizes shapes that cannot be easily molded. The ceramic core in Figure 3a was built by stereolithography, but its design was based on contemporary commercial cores that are manufactured by conventional injection molding. Many of the features of this core could have been achieved by molding, so the benefit of AM in this case is that it enables toolless manufacture of a part that could be produced with injection molding tooling. Toolless manufacturing is an attribute of AM that is important for early applications, but it is perhaps not as important as the ability of AM to produce designs that are impossible to manufacture with injection molding. An example of the latter is the demonstration part shown in Figure 5 with 168 intertwined helical coils that are coiled to a tight pitch. AM methods should be compared with ceramic injection molding, the traditional method used to manufacture elaborate shapes. On a per-part basis, ceramic injection molding is much faster. Forming cycle times for ceramic injection molding range from several seconds for small parts to several minutes for parts as large as industrial gas turbine cores. If productivity is measured on the basis of cycle time alone, ceramic stereolithography might seem to be impractical for mass production of identical products. However, although completing all the layers in a platform can 10.18

Halloran



MR46CH10-Halloran

ARI

8 March 2016

14:22

take hours, the platform can include a large number of parts, so the cycle time per part can be much shorter. Also, an injection molding tool with one set of dies can produce only one type of object without switching to a new set of dies. The platform of an AM tool can simultaneously produce many different products. Cycle time alone does not capture the value of one ceramic stereolithography tool simultaneously producing a variety of different products. Different concepts of productivity are needed to compare conventional hard-tooled methods, which can make only one part, with AM manufacturing tools, which are essentially flexible. AM manufacturing is only now emerging, so there are no hard data on practical manufacturing at the time of this writing. Every product category can be a different case. Consider only the manufacture of airfoil cores, which are usually injection molded. Conventional injection molding requires die sets for each product, and molding tooling die sets typically involve investments on the order of $100,000 and acquisition delays of several months.

AM for Parts Conventionally Manufactured by Injection Molding Stereolithography requires no tooling, which thus provides an advantage for relatively small production runs. But for very large production runs lasting many years, manufacturing with hardtooled conventional injection molding could be advantageous. However, ceramics are abrasive. Tool life can be much shorter for ceramic injection molding than for plastics injection molding. A full analysis of ceramic injection molding cost needs to consider replacement or refurbishing of tooling. Also, as the tools wear with use, there can be subtle but significant changes in quality. For products like ceramic cores, with elaborate designs and tight dimensional tolerances, the quality of cores molded in new tools can be different from that of cores molded with worn tools or cores molded with refurbished tools. This difference in quality may affect product yield, which is as important as cycle time in determining productivity. There is no tool wear with stereolithography, as every part is an identical digital copy of the master CAD design. However, for large production runs, replacing conventional injection molding with AM is a challenge because all currently commercial ceramic cores now being produced by injection molding come from the restricted set of designs that are moldable. As injection molding is currently the only manufacturing method, all existing designs are of course capable of being produced by injection molding.

AM for Designs that Cannot Be Manufactured by Injection Molding Not every product design can be molded. Although, in the context of ceramic forming, we often think of injection molding as having excellent shape capability, the actual shapes that can be molded are in fact rather limited. There must be sprues and runners, and the material must flow into every location of the part, which imposes certain geometrical restrictions. Part geometry is even more restricted if only two-part dies are considered. More difficult shapes require multipart molds, which are much more costly. But even multipart molds have limitations. Now manufacturers advise their customers against unmoldable designs or resort to assemblies of molded subcomponents. But there are sound technical reasons why airfoil designers would like to use unmoldable cores with “impossible” features. These difficult cores can enable more effective airfoils, and the performance of the ultimate product (the turbine engine) can be improved with ceramic core designs that are impossible to mold. Ceramic stereolithography has the potential to enable impossible airfoil designs. The value of what might be enabled will contribute to what will emerge as the true productivity of ceramic stereolithography.



10.19

MR46CH10-Halloran

ARI

8 March 2016

14:22

CONCLUSIONS


Ceramic stereolithography refers to a number of AM methods involving patterning by photopolymerization of ceramic powders suspended in a polymerizable medium. A broad range of ceramic stereolithography techniques have been reported, and several effective devices are commercially available. A diverse range of ceramic prototypes and functional parts have been reported, particularly for ceramics with low refraction index. The practical fundamentals of suspension photopolymerization can be understood by using relatively simple physical principles. The relation between the desired cure depth and the necessary energy dose depends on the attenuation of the beam in the suspension and on the nature of the monomer and added polymerization inhibitors. Attenuation in the suspension can be understood in terms of absorption, which involves the concentration and extinction coefficients of dyes and photoinitiators, and in terms of scattering by ceramic particles, which involves the volume fraction and refractive index of the ceramic. The width of a cured feature depends on the lateral intensity distribution, with excess broadening caused by scattering, and so is related to the refractive index of the ceramic. The cure profile, or the fractional monomer conversion within each layer, can be described in terms of both the attenuated dose at each depth and the relation between dose and monomer conversion.

DISCLOSURE STATEMENT The author is a cofounder of DDM Systems, Inc.

LITERATURE CITED 1. Jacobs PF. 1992. Rapid Prototyping & Manufacturing: Fundamentals of Stereolithography. Dearborn, MI: Soc. Manuf. Eng. 2. Lee HD, Pober RL, Calvert PD, Bowen HK. 1986. Photopolymerizable binders for ceramics. J. Mater. Sci. Lett. 5:81–81 3. Lovell LG, Lu H, Elliot JE, Stansbury JW, Bowman CN. 2001. Effect of curing rate on the mechanical properties of dental resins. Dent. Res. 17:504–11 4. Taylor BE, Bidwell L, Lawrence A. 2001. New photoimageable LTCC technology for making a wide range of ceramic architectures and circuits. SPIE Proc. 4428:89–92 5. Griffith ML, Halloran JW. 1996. Freeform fabrication of ceramics by stereolithography. J. Am. Ceram. Soc. 79:2601–8 6. Zimbeck WR, Jang JH, Schultze W, Rice RW. 2000. Automated fabrication of ceramic electronic packages by stereo-photolithography. In Solid Freeform and Additive Fabrication, 2000, Vol. 625, ed. SC Danforth, DB Dimos, F Prinz, pp. 173–78. Warrendale, PA: Mater. Res. Soc. 7. Hinczewski C, Corbel S, Chartier T. 1998. Stereolithography for the fabrication of three dimensional ceramic parts. Rapid Prototyp. J. 4:104–11 8. Corcione CE, Greco A, Montagna F, Lucciulli A, Maffezzoli A. 2005. Silica moulds built by stereolithography. J. Mater. Sci. 40:4899–904 9. Ventura SC, Narang SC, Sharma S, Stotts J, Liu CL, et al. 1996. A new SFF process for functional ceramic components. In Solid Freeform Fabrication Symposium Proceedings, ed. DL Bourell, JJ Beaman, H Marcus, J Barlow, pp. 327–34. Austin: Univ. Tex. 10. Yuan D, Kambly K, Shao P, Rudrarju A, Cilio P, et al. 2009. Experimental investigations on photocurable ceramic materials systems for large area maskless photopolymerization (LAMP). In Solid Freeform Fabrication Symposium Proceedings, ed. D Bourell, RH Crawford, CC Seepersad, JJ Beaman, H Marcus. Austin: Univ. Tex. 11. Schwentenwein M, Homa J. 2015. Additive manufacturing of dense alumina ceramics. Int. J. Appl. Ceram. Technol. 12:1–7 10.20

Halloran



MR46CH10-Halloran

ARI

8 March 2016

14:22

12. Song X, Chen Y, Lee TW, Wu SH, Chen LX. 2015. Ceramic fabrication using mask-image-projectionbased stereolithography integrated with tape-casting. J. Manuf. Process. 20:456–64 13. Bian WG, Li DC, Lian Q, Zhang WJ, Zhu LH, et al. 2011. Design and fabrication of novel porous implant with pre-set channels based on ceramic stereolithography for vascular implantation. Biofabrication 3:034103 14. Gmeiner R, Deisinger U, Schonherr J, Lechner B, Detsch R, et al. 2015. Additive manufacturing of bioactive glasses and silicate bioceramics. J. Ceram. Sci. Technol. 6:75–86 15. Doreau F, Chaput C, Chartier T. 2000. Stereolithography for manufacturing ceramic parts. Adv. Eng. Mater. 2:493–94 16. Feltzmann R, Gruber S, Mitteramskogler G, Tesavibul P, Boccaccini AR, et al. 2012. Lithography-based additive manufacturing of cellular ceramic structures. Adv. Eng. Mater. 14:1052–58 17. Brady GA, Chu TM, Halloran JW. 1996. Curing behavior of ceramic resin for stereolithography. In Solid Free Form Fabrication Proceedings, ed. JJ Beaman, JW Barlow, DL Bourell, RH Crawford, pp. 403–10. Austin: Univ. Tex. 18. Halloran JW, Tomeckova V, Gentry S, Das S, Cilio P, et al. 2011. Photopolymerization of powder suspensions for shaping ceramics. J. Eur. Ceram. Soc. 31:2613–19 19. Bae C-J, Halloran JW. 2010. Integrally cored ceramic mold fabricated by ceramic stereolithography. Int. J. Appl. Ceram. Technol. 8:1255–62 20. Sun C, Zhang X. 2001. Experimental and numerical investigations on microstereolithography of ceramics. J. Appl. Phys. 92(8):4796–97 21. Sun C, Zhang X. 2002. The influences of the materials properties on ceramic micro-stereolithography. Sens. Actuators A 101:354–70 22. Bertsch A, Jiguet S, Renauld P. 2004. Microfabrication of ceramic components by microstereolithography. J. Micromech. Microeng. 14:197–203 23. Provin C, Monneret S, Gall HL, Corbel S. 2003. Three-dimensional ceramic microcomponents made using microstereolithography. Adv. Mater. 15:994–97 24. Adake CV, Gandi P, Bhargava P. 2015. Fabrication of ceramic component using constrained surface microstereolithography. Proc. Mater. Sci. 5:355–61 25. Kiriahar S, Niki T. 2015. Three-dimensional stereolithography of alumina photonic crystals for terahertz wave localization. Int. J. Appl. Ceram. Technol. 12:32–37 26. Travitzky M, Bonet A, Dermeik B, Fey T, Filbert-Demut I, et al. 2014. Additive manufacturing of ceramic-based materials. Adv. Eng. Mater. 16:729–54 27. Deckers J, Vleugels J, Kruth J-P. 2014. Additive manufacturing of ceramics: a review. J. Ceram. Sci. Technol. 5:245–60 28. Zocca A, Columbo P, Gomes CM, Gunster J. 2015. Additive manufacturing of ceramics: issues, poten¨ tialities, and opportunities. J. Am. Ceram. Soc. 98:1983–2001 29. Bose S, Vahabzedeh S, Ke DX, Bandyopadhyay A. 2015. Additive manufacturing of ceramics. In Additive Manufacturing, ed. A Bandyopadhyay, S Bose, pp. 143–84. Boca Raton, FL: CRC 30. Brakora KF, Sarabandi K. 2009. Integration of single-mode photonic crystal clad waveguides with monolithicly constructed ceramic subsystems. IEEE Antennas Propag. Lett. 8:433–36 31. de Hazen Y, Heinecke J, Weber A, Graule T. 2009. High solids loading ceramic colloidal dispersion in UV curable media via comb-polyelectrolyte surfactants. J. Colloid Interface Sci. 337:66–74 32. Tallon C, Franks GV. 2011. Recent trends in shape forming from colloidal processing: a review. J. Ceram. Soc. Jpn. 119:147–60 33. Chu T-MG, Halloran JW. 2000. High-temperature flow behavior of ceramic suspensions. J. Am. Ceram. Soc. 83:2189–95 34. Lombardo SJ. 2015. Minimum time heating cycles for diffusion-controlled binder removal from ceramic green bodies. J. Am. Ceram. Soc. 98:57–65 35. Song JH, Evans JRG, Edirisinghe MJ. 2000. Modeling the effect of gas transport on the formation of defects during thermolysis of powder moldings. J. Mater. Res. 15:449–57 36. Wu T, Das S. 2012. Theoretical modeling and experimental characterization of stress development in parts manufactured through large area maskless photopolymerization. In Solid Freeform Fabrication Symposium Proceedings, ed. D Bourell, RH Crawford, CC Seepersad, JJ Beaman, H Marcus, pp. 748–60. Austin: Univ. Tex. www.annualreviews.org • Ceramic Stereolithography


10.21

ARI

8 March 2016

14:22

37. Tomeckova V, Halloran JW. 2011. Flow behavior of polymerizable ceramic suspensions as a function of ceramic volume fraction and temperature. J. Eur. Ceram. Soc. 31:2535–42 38. Brady GA, Halloran JW. 1998. Solid freeform fabrication of ceramics by stereolithography. Nav. Res. Rev. 50:39–43 39. Lee JH, Prud’homme RK, Aksay IA. 2001. Cure depth in photopolymerization: experiments and theory. J. Mater. Res. 16:2536–3544 40. Tomeckova V, Halloran JW. 2010. Predictive models for photopolymerization of ceramic suspensions. J. Eur. Ceram. Soc. 30:2833–40 41. Garg R, Prud’homme RK, Aksay IA. 1998. Optical transmission in highly concentrated dispersions. J. Opt. Soc. Am. A 15:932–35 42. Wu KC, Seefeldt KF, Solomon MJ, Halloran JW. 2005. Prediction of ceramic stereolithography resin sensitivity from theory and measurement of diffusive photon transport. J. Appl. Phys. 98:024902 43. Singh P, Smith LS, Bezdecny M, Cheverton M, Brewer JA, Venkataramani V. 2011. Additive manufacturing of PZT-5H piezoceramic for ultrasound transducers. In IEEE Int. Ultrasonics Symp. Proc., pp. 1111–14. doi:10.1109/ULTRSYM.2011.0273 44. Woodward DI, Purssell CP, Bilson DR, Hutchins DA, Leigh SJ. 2015. Additively-manufactured piezoelectric devices. Phys. Status Solid. A 212:2107–13 45. Chartier T, Badev A, Abouliatim Y, Lebaudy P, Lecamp L. 2012. Stereolithography process: influence of the rheology of silica suspensions and of the medium on polymerization kinetics—cured depth and width. J. Eur. Ceram. Soc. 32:1625–34 46. Gentry SP. 2012. Improving the resolution of manufacturing methods using photopolymerizable ceramic suspensions. PhD Diss., Univ. Mich. 47. Hinczewski C, Corbel S, Chartier T. 1998. Stereolithography for the fabrication of three dimensional ceramic parts. Rapid Prototyp. J. 4:104–11 48. Rudraraju A. 2013. Digital data processing and computational design for large area maskless photopolymerization. PhD. Diss., Georgia Inst. Technol. 49. Gentry SP, Halloran JW. 2013. Depth and width of cured lines in photopolymizable ceramic suspensions. J. Eur. Ceram. Soc. 33:1981–88 50. Tomeckova V, Teyssandier F, Norton SJ, Love BJ, Halloran JW. 2012. Photopolymerization of acrylate suspensions. J. Photochem. Photobiol. A 247:74–81 51. Wu KC, Halloran JW. 2005. Photopolymerization monitoring of ceramic stereolithography resins by FTIR methods. J. Mater. Sci. 40:71–76 52. Chu T-MG, Halloran JW. 2000. Curing of highly loaded ceramic suspension in acrylates. J. Am. Ceram. Soc. 83:2375–80 53. Bae C-J, Halloran JW. 2011. Influence of residual monomer on cracking of ceramics fabricated by stereolithography. Int. J. Appl. Ceram. Technol. 8:1289–95 54. Pfaffinger M, Mitteramskogler G, Gmeiner R, Stampf J. 2015. Thermal debinding of ceramic-filled photopolymer. Mater. Sci. Forum 825–826:75–81 55. Renap K, Kruth JP. 1995. Recoating issues in stereolithography. Rapid Prototyp. J. 1:4–16 56. Mitteramskogler G, Gmeiner R, Feltzmnn R, Gruber S, Hofsteller C, et al. 2014. Light curing strategies for lithography-based additive manufacturing of customized ceramics. Addit. Manuf. 1–4:110–18


MR46CH10-Halloran

10.22

Halloran
