change to the GATOR2 binding site via movementof helix C3 (Fig. 5D). Alternatively, a segment ofthe partially disordered linker domain, whichcontacts the leucine-binding pocket via Leu261in helix L1 (fig. S3A), is also in close proximityto the GATOR2 binding site in our structure(Fig. 5C). Therefore, changes in the leucine-binding state of Sestrin2 could potentially alterthe position of the linker domain, thereby affect-ing the availability of the GATOR2 binding site.Despite these insights, several important ques-

tions remain. Fully understanding how leucinebinding causes dissociation of Sestrin2 fromGATOR2 will probably require ascertaining thestructure of either apo-Sestrin2 or the Sestrin2-GATOR2 complex. Furthermore, understandingthe exact mechanism by which Sestrin2 inhibitsthe mTORC1 pathway awaits the elucidation ofthe molecular function of GATOR2.Finally, as a critical regulator of cell growth,

mTORC1 is misregulated in various human dis-eases, including cancer and diabetes, as well asin aging (1, 29). By revealing the mechanismthrough which a natural small molecule regulatesthis pathway, our results may enable the identi-fication of compounds to pharmacologically tar-get the nutrient-sensing pathway upstream ofmTORC1 in vivo.

REFERENCES AND NOTES

1. R. Zoncu, A. Efeyan, D. M. Sabatini, Nat. Rev. Mol. Cell Biol. 12,21–35 (2011).

2. C. C. Dibble, B. D. Manning, Nat. Cell Biol. 15, 555–564 (2013).3. J. L. Jewell, R. C. Russell, K. L. Guan, Nat. Rev. Mol. Cell Biol.

14, 133–139 (2013).4. M. Potier, N. Darcel, D. Tomé, Curr. Opin. Clin. Nutr. Metab.

Care 12, 54–58 (2009).5. J. S. Greiwe, G. Kwon, M. L. McDaniel, C. F. Semenkovich,

Am. J. Physiol. Endocrinol. Metab. 281, E466–E471(2001).

6. K. S. Nair, R. G. Schwartz, S. Welle, Am. J. Physiol. 263,E928–E934 (1992).

7. H. L. Fox, P. T. Pham, S. R. Kimball, L. S. Jefferson, C. J. Lynch,Am. J. Physiol. 275, C1232–C1238 (1998).

8. C. J. Lynch, H. L. Fox, T. C. Vary, L. S. Jefferson, S. R. Kimball,J. Cell. Biochem. 77, 234–251 (2000).

9. A. Efeyan, D. M. Sabatini, Biochem. Soc. Trans. 41, 902–905(2013).

10. C. Buerger, B. DeVries, V. Stambolic, Biochem. Biophys. Res.Commun. 344, 869–880 (2006).

11. K. Saito, Y. Araki, K. Kontani, H. Nishina, T. Katada, J. Biochem.137, 423–430 (2005).

12. L. Bar-Peled, L. D. Schweitzer, R. Zoncu, D. M. Sabatini, Cell150, 1196–1208 (2012).

13. Y. Sancak et al., Science 320, 1496–1501 (2008).14. Y. Sancak et al., Cell 141, 290–303 (2010).15. R. V. Durán, M. N. Hall, EMBO Rep. 13, 121–128 (2012).16. R. Zoncu et al., Science 334, 678–683 (2011).17. S. Wang et al., Science 347, 188–194 (2015).18. M. Rebsamen et al., Nature 519, 477–481 (2015).19. L. Bar-Peled et al., Science 340, 1100–1106 (2013).20. R. L. Wolfson et al., Science 351, 43–48 (2016).21. L. Chantranupong et al., Cell Rep. 9, 1–8 (2014).22. A. Parmigiani et al., Cell Rep. 9, 1281–1291 (2014).23. F. H. Niesen, H. Berglund, M. Vedadi, Nat. Protoc. 2, 2212–2221

(2007).24. A. V. Budanov, A. A. Sablina, E. Feinstein, E. V. Koonin,

P. M. Chumakov, Science 304, 596–600 (2004).25. J. F. Gibrat, T. Madej, S. H. Bryant, Curr. Opin. Struct. Biol. 6,

377–385 (1996).26. H. A. Woo, S. H. Bae, S. Park, S. G. Rhee, Antioxid. Redox

Signal. 11, 739–745 (2009).27. A. Koshkin, C. M. Nunn, S. Djordjevic, P. R. Ortiz de Montellano,

J. Biol. Chem. 278, 29502–29508 (2003).28. M. Peng, N. Yin, M. O. Li, Cell 159, 122–133 (2014).29. M. Laplante, D. M. Sabatini, Cell 149, 274–293 (2012).

ACKNOWLEDGMENTS

D.M.S. is a founder, a member of the Scientific Advisory Board, a paidconsultant, and a shareholder of Navitor Pharmaceuticals, which istargeting for therapeutic benefit the amino acid sensing pathwayupstream of mTORC1. We thank all members of the Sabatini andSchwartz laboratories for helpful insights. We also thank Cell SignalingTechnologies for providing many antibodies. This work is based onresearch conducted at the Northeastern Collaborative Access Teambeamlines, which are funded by the National Institute of GeneralMedical Sciences from the National Institutes of Health (P41GM103403). The Pilatus 6M detector on 24-ID-C beam line is funded bya NIH-ORIP HEI grant (S10 RR029205). This research used resourcesof the Advanced Photon Source, a U.S. Department of Energy (DOE)Office of Science User Facility operated for the DOE Office of Science byArgonne National Laboratory under contract no. DE-AC02-06CH11357.This work was supported in part by the NIH Predoctoral Training GrantT32GM007287. This work has also been supported by grants from NIH(R01CA103866 and AI47389) and the U.S. Department of Defense(W81XWH-07- 0448) to D.M.S. Fellowship support was provided by NIHto R.L.W. (awards T32 GM007753 and F30 CA189333), L.C. (F31

CA180271), and T.W. (F31 CA189437). T.W. is also supported by anaward from the MIT Whitaker Health Sciences Fund. M.E.P. issupported by the Sally Gordon Fellowship of the Damon Runyon CancerResearch Foundation (award DRG-112-12) and a Department ofDefense Breast Cancer Research Program Postdoctoral Fellowship(award BC120208). D.M.S. is an investigator of the Howard HughesMedical Institute. Coordinates and structure factors for the x-ray crystalstructures of Sestrin2 have been deposited in the Protein Data Bank(PDB) with accession code 5DJ4.

SUPPLEMENTARY MATERIALS

www.sciencemag.org/content/351/6268/53/suppl/DC1Materials and MethodsFigs. S1 to S6Table S1References (30–44)

11 August 2015; accepted 5 November 2015Published online 19 November 201510.1126/science.aad2087

REPORTS◥

3D PRINTING

Additive manufacturing ofpolymer-derived ceramicsZak C. Eckel, Chaoyin Zhou, John H. Martin, Alan J. Jacobsen,William B. Carter, Tobias A. Schaedler*

Theextremelyhighmeltingpoint ofmanyceramics adds challenges toadditivemanufacturingascompared with metals and polymers. Because ceramics cannot be cast or machined easily,three-dimensional (3D) printingenables abig leap in geometrical flexibility.We report preceramicmonomers that are cured with ultraviolet light in a stereolithography 3D printer or through apatterned mask, forming 3D polymer structures that can have complex shape and cellulararchitecture.Thesepolymer structures canbepyrolyzed to aceramicwith uniformshrinkage andvirtually noporosity. Siliconoxycarbidemicrolattice andhoneycombcellularmaterials fabricatedwith this approach exhibit higher strength than ceramic foams of similar density. Additivemanufacturing of such materials is of interest for propulsion components, thermal protectionsystems, porous burners, microelectromechanical systems, and electronic device packaging.

In comparison with metals and polymers,ceramics are difficult to process, particularlyinto complex shapes. Because they cannot becast ormachined easily, ceramics are typicallyconsolidated from powders by sintering or

deposited in thin films. Flaws, such as porosityand inhomogeneity introduced during processing,govern the strength because they initiate cracks,and—in contrast to metals—brittle ceramics havelittle ability to resist fracture. This processing chal-lenge has limited our ability to take advantage ofceramics’ impressive properties, including high-temperature capability, environmental resistance,and high strength. Recent advances in additivemanufacturing have led to a multitude of differenttechniques, but all additive manufacturing tech-niques developed for ceramicmaterials are powder-based layer-by-layer processes that are restricted

to a small number of compositions (1, 2). Only afew of the commercially available three-dimensional(3D) printing systems offer printing of ceramics,either by selective curing of a photosensitiveresin that contains ceramic particles, selectivedeposition of a liquid binder agent onto ceramicparticles (binder jetting), or selective fusion of apowder bed with a laser (3, 4). All these tech-niques are limited by slow fabrication rates, andin many cases, a time-consuming binder removalprocess. By starting with powders, consolidationto a dense part is an almost insurmountablechallenge, and residual porosity is typically un-avoidable. Furthermore, many additive processesintroduce large thermal gradients that tend tocause cracks in ceramics. Pores, cracks, and in-homogeneities are responsible for the low strengthand poor reliability of additively manufacturedceramic parts.Polymer-derived ceramics were discovered in

the 1960s (5). Uponheat treatment (typically underinert atmosphere), they pyrolyze into SiC, Si3N4,

58 1 JANUARY 2016 • VOL 351 ISSUE 6268 sciencemag.org SCIENCE

HRL Laboratories, LLC, 3011 Malibu Canyon Road, Malibu,CA 90265, USA.*Corresponding author. E-mail: taschaedler@hrl.com

RESEARCH

BN, AlN, SiOC, SiCN, BCN, or other compositions,whereas volatile species (CH4, H2, CO2, H2O, andhydrocarbons) leave thematerial. Preceramicpoly-mers are currently used to synthesize ceramic fibersand to densify ceramic matrix composites by in-filtration. Two-dimensional photolithography andsoft lithography have been demonstrated (6, 7).The absence of a sintering step enables lower syn-thesis temperatures without the need for pressure,as comparedwith classical ceramicpowderprocess-ing, and the absence of sintering additives resultsin improved thermomechanical properties (8).By attaching thiol, vinyl, acrylate,methacrylate,

or epoxy groups to an inorganic backbone such asa siloxane, silazane, or carbosilane, ultraviolet(UV)–active preceramicmonomers can be obtained(7, 9). Two different additivemanufacturing tech-niques based on photopolymerization can be usedto achieve spatial control. For conventional stereo-lithography (SLA), sufficient polymerization in-hibitor and UV absorber are added to the resinformulation to confine the polymerization to thelaser exposure point and to minimize scatter tomaintain fidelity in the features of the printedpart.UV light is then scanned across the resin surfaceto expose a cross section and build up a thin slice(30 to 100 mm) of the part to be manufactured. Al-though almost any geometry can be fabricatedwiththis approach, the process is slow, because every 30-to 100-mm thin layer has to be exposed separately.Structureswith linear features extending from theexposure surface, such as lattices and honey-combs, can be formed 100 to 1000 times as rapidlywith the self-propagating photopolymer wave-guide technology (SPPW) (10, 11). Monomers areselected to promote a change in the index of re-fraction upon polymerization, which causes inter-nal reflection of the UV light, trapping it in thealready-formed polymer. This exploits a self-

focusing effect that forms a polymer waveguide,tunneling the light toward the tip of the wave-guide and causing it to polymerize further. Thisreduces the need for additives that control scat-ter and UV absorption. The architecture of thematerial or structure can then be defined by apatterned mask that defines the areas exposedto a collimated UV light source (10).Bothmethods produce parts consisting of cross-

linkedpolymer (Fig. 1),where the cross-linkdensitydepends on exposure parameters and can be in-creased by thermal treatments or additional UVexposure. Unpolymerized resin can be recycled andreused.The configuration and microstructure of the

preceramic polymer determine the composition,microstructure, and yield of ceramic after pyrol-ysis. A high cross-link density is necessary to pre-vent the loss of low–molecular mass species andfragmentation during pyrolysis. Siloxane-basedpolymers with their Si-O-Si backbone result insilicon oxycarbides, whereas silazanes introducenitrogen due to their Si-N-Si backbone. Combin-ing siloxanes with silazanes results in a SiOCNceramic after pyrolysis. The addition of silanecompounds typically reduces the amount of oxy-gen and pushes the ceramic composition towardSiC (8). The ratio of carbon in the final ceramiccan be tailored by adding phenyl groups on theside chain of the polymer or using a carbon-basedcross-linking agent such as divinyl benzene. Theprecursor chemistry can also be changed to in-corporate other elements—for example, B or Zrto enhance temperature capability (12); Fe, Co, orNi to introduce magnetic properties; or Cu, Pd,or Pt for catalytic properties (13). To fabricate thestructures shown in Fig. 1 a UV-curable siloxaneresin system was formulated by mixing (mercap-topropyl) methylsiloxane with vinylmethoxysi-

loxane and adding UV free-radical photo initiator,free-radical inhibitor, andUV absorber. The result-ing liquid resin was used in a benchtop stereo-lithography 3D printer (Formlabs Form 1+). Tofabricate the larger microlattice and honeycombstructures via SPPW for mechanical testing, theresin was reformulated without UV absorber,poured into a DELRIN reservoir, and exposedwith UV light through a patternedmask (see thesupplementary materials for details).Pyrolysis at 1000°C in argon was accompanied

by 42% mass loss and 30% linear shrinkage. Theresultant ceramic is amorphous, as ascertained byx-ray diffraction (XRD) and transmission electronmicroscopy (TEM), and has a composition of 26.7atomic percent (at %) Si, 33.4 at % C, 4.1 at % S,and 35.8 at % O, or SiO1.34C1.25S0.15, as measuredby inductively coupled plasma mass spectrom-etry. The ceramic structures fabricated are fullydense, with no porosity or surface cracks ob-served by scanning electronmicroscopy and TEM(Fig. 2). Ceramic parts fabricated with the self-propagating photopolymer waveguide processexhibit a very smooth surface (Fig. 2A), whereasparts fabricated by stereolithography show thetypical steps at the surface from the layer-by-layerprinting process (Fig. 2C). As the undulationscould act as stress concentrators and negativelyaffect the mechanical properties, all mechanicaltests were performed on parts fabricated bySPPW. The SiOC ceramic fractures in a conchoi-dal manner typical for brittle amorphous mate-rials, with curved breakage surfaces and ripples(Fig. 2B). To avoid shattering on pyrolysis, theprinted polymer structure is typically limitedto features with less than ~3mm in thickness inone dimension and the heating rate to less than20°C/min, so that evolving gases can escape. Byselecting appropriate cellular architectures, large

SCIENCE sciencemag.org 1 JANUARY 2016 • VOL 351 ISSUE 6268 59

Fig. 1. Additive manufacturing of polymer-derived ceramics. (A) UV-curable preceramic monomers are mixed with photoinitiator. (B) The resin is exposedwith UV light in a SLA 3D printer or through a patterned mask. (C) A preceramic polymer part is obtained. (D) Pyrolysis converts the polymer into a ceramic.Examples: (E) SLA 3D printed cork screw. (F and G) SPPW formed microlattices. (H) Honeycomb.

RESEARCH | REPORTS

ceramic structures can be fabricated, with thesize only limited by the equipment. This fabrica-tion process introduces no noticeable gradientsin composition, and temperature gradients canbe mitigated by the cellular architecture, re-sulting in remarkably uniform shrinkage duringpyrolysis. The shape of the polymer structureis therefore maintained well and the shrink-age can be predicted, as long as any surfaces incontact with the structure during pyrolysisare lubricated to prevent sticking. Various cell-ular architectures have been demonstrated withthe self-propagating photopolymer waveguide pro-cess, including microlattices with densities of 0.22to 0.35 g/cm3 (Fig. 1G), honeycombswith densitiesof 0.3 to 0.8 g/cm3 (Fig. 1H), and pyramidaltruss cores with graded density (fig. S4C).Compression and shear testing was performed

on as-pyrolyzed silicon oxycarbide structures, andthe results are summarized in Fig. 3 and table S1.A compressive failure strength of 163 MPa wasmeasured on a honeycomb structurewith a densityof 0.8 g/cm3 using a prescribed displacement rateof 10 mm/s. Shear testing was performed on fourmicrolattices with densities of 0.22 to 0.35 g/cm3

according to ASTM C273, using a single lapshear test fixture, and resulted in ultimate shearstrengths in the range of 3.7 to 4.9 MPa andmodulus values of 830 to 1570 MPa. Failure incompression was catastrophic by sudden brittlefracture, whereas failure in shear was gradual bysuccessive brittle fracture of single struts. Themechanical properties of silicon oxycarbide mic-rolattice structures are compared to those ofceramic foams of similar density (Fig. 3). Note-worthy is the ~10 times higher compressivestrength as compared with commercially avail-able SiC foams (Duocel) and aluminosilicate foam(ceramic insulation) (table S2), as well as siliconoxycarbide foams (14). The improvement in shearstrength does not appear as large in Fig. 3B becausethe values reported for SiC and aluminosilicatefoams are flexural strength, which is measuredby a bending test and is generally higher. Evenin comparison to state-of-the-art cellular sand-wich core materials, aluminum alloy honeycomb(HexWeb) and closed-cell polymer foam (Diviny-cell), the polymer-derived ceramic cellularmaterialslook favorable. Sample details and measurementresults are summarized in table S1.Themechanical properties of a cellularmaterial

depend on the mechanical properties of the solidconstituent material, the relative density of the cel-lularmaterial, and the cellular architecture (i.e., thespatial configuration of voids and solid). Two factorscontribute to the observed high strength. First, theordered, periodic architectures are inherently moremechanically efficient than a random foam archi-tecture. Gibson and Ashby (15) have described thegeneral relationships for the elastic modulus (E)and failure strength (s) of a cellular material as

E ≈ C1 (Es) (r/rs)n1 (1)

s ≈ C2 (ss) (r/rs)n2 (2)

The terms Es and ss are the elastic modulus andrepresentative failure strengthof the solidmaterial,

respectively. The term r/rs is the relative density ofthe cellularmaterial, which is defined as its density(r) divided by the density of the solid constituentmaterial (rs). rs of SiOC is 2.05 g/cm3. The pro-portionality constants C1 and C2 are related tothe geometric configuration of the cellular mate-rial with respect to the loading direction. Theexponents n1 and n2 are 2 and 1.5, respectively,for foams, where the cell struts exhibit bending-dominated deformation during elastic loading (15).Conversely, a latticematerial can exhibit stretching-dominated deformation, when the latticemembersare configured so that they are loaded either in

tension or compression, which results in much-improved mechanical properties that decreaselinearly with density (n1 = 1 and n2 = 1). Theceramic microlattices exhibit a scaling n2 = 1.06(R2 = 0.88), and the honeycombs show n2 = 1.18(R2 = 0.92), demonstrating stretching-dominatedmechanical performance. The difference in com-pressive strength arising from the different scalingof stretch-dominated versus bending-dominatedarchitecture should be a factor of 3.2 at a relativedensity of 10% and increases to 5.8 at 3%. Theproportionality constant C2 for a brittle foam is~0.2 (15), whereas the constant is estimated to be

60 1 JANUARY 2016 • VOL 351 ISSUE 6268 sciencemag.org SCIENCE

Fig. 3. Strength ofpolymer-derived SiOCmaterials comparedto ceramic foams.(A) Compressive strength.(B) Shear strength.

Fig. 2. Electronmicroscopy characterization of SiOCmicrolattice and cork screw. (A) SPPW-formedlattice node showing smooth surface. (B) Fracture surface of a strut. (C) SLA printed corkscrew showingundulations on the surface. (D) 3D printing step size is 50 mm. (E) Bright-field TEM image showing noporosity. (F) TEM diffraction indicating amorphous structure.

RESEARCH | REPORTS

1/3 formicrolattices (16), 0.87 for pyramidal trussstructures (16), and 1 for honeycombs, accountingfor additional increases in strength.This high intrinsic strength is the second factor

besides the architecture contributing to the muchhigher strength of these cellular materials as com-paredwith previously reported ceramic foams. Thehigh intrinsic strength of polymer-derived siliconoxycarbide is attributed to a low distribution offlaws in the material, as ascertained by the ab-sence of porosity and cracks in electron micros-copy. The fracture strength of brittle materials isgiven by

sfracture ¼ffiffiffiffiffiffiffiffiffiffiEgr4ad0

rð3Þ

where g is the surface energy and d0 is the equi-librium distance between atomic centers, whichtogether withEdetermine the theoretical strength,whereas cracks with half-length a and crack-tipradius r introduce stress concentrations that re-duce the fracture strength. Because the flaw pop-ulation in the polymer-derived ceramic materialcan be controlled well through the high purityof the starting resin and the development ofsmooth and pore-free surfaces, higher fracture-strength values with a tighter distribution aremeasured as compared with other ceramicmate-rials, especially when derived from conventionalpowder routes. Ultimate strength values are re-ported, but due to the brittle nature of the mate-

rial, they coincide with the yield and fracturestrength.To calculate themodulus of the solid constitutive

SiOCmaterial, the equation for shearmodulus ofmicrolattices (17) is used

G ¼ E

8sin2ð2qÞ r

rsð4Þ

Because accurate modulus measurements couldonly be performed in shear testing. An averageYoung’s modulus of 102 ± 26 GPa is obtained,which is in the range reported for similar com-positions (5).The silicon oxycarbide family of polymer-

derived ceramics has demonstrated excellenthigh-temperature properties, including remark-able resistance to crystallization, oxidation, andcreep (8, 18). These properties have been ascribedto the amorphous material exhibiting nano-domains of silica tetrahedra that are encasedin a network of graphene (19). The heart of the1- to 3-nm domains is formed by silicon-oxygentetrahedra, and the interdomain boundariesconsist of layers of sp2 carbons. Silicon atomsbonded to one or two carbons substituted foroxygen make up the interface between silicadomains and graphene walls (19).The silicon oxycarbide microlattice structures

showed excellent stability at high temperaturesin air. At 1300°C, the structures gained ~0.15%mass over 10 hours, and most of this mass gainoccurred within the first 2 hours. It is hypothe-

sized that this is associated with a replacementreaction at the SiOC surface, creating an amor-phous SiO2 oxide layer and releasing CO or CO2.This oxide growthwasqualitatively observed as ashift in interference coloration at the micro-lattice surface. After each subsequent heat treat-ment, there was a shift in iridescent colorationassociated with increased thickness of the clearand thin (100 to 1000 nm) oxide scale, consistentwith thin-film interference coloration. At 1400°C,the samples showed a slow but steady mass de-cline of ~1% after 10 hours. This mass loss wasattributed to the “burn off” of free carbon in theSiOC structure (20). After 10 hours at 1400°C, ahazy surface oxide was observed. This oxidationproduct was characterized to be cristobalite byXRD. A similar behavior was observed at 1500°C,1600°C, and 1700°C, albeit with an increasingmass loss rate and more pronounced cristobaliteoxidation products (Fig. 4C). The highest temper-ature to which SiOC samples were exposed was1700°C, and no degradation other than surfaceoxidation was observed. Themass loss is normal-ized by the surface area (Fig. 4A). The change inoxide structure is attributed to the oxidation pro-duct being amorphous at or below 1300°C, where-as above 1400°C it crystalizes to cristobalite. O2

diffusion into the bulk oxidizes available free car-bon, andat evenhigher temperatures, carbothermalreduction of SiO2 from free carbon in the struc-ture begins (20). The oxide shell on the surface ap-pears to slow these reactions by limiting diffusion

SCIENCE sciencemag.org 1 JANUARY 2016 • VOL 351 ISSUE 6268 61

Fig. 4. High-temperature oxidation of silicon oxycarbidemicrolattice. (A) Mass changemeasured after consecutive heat treatments at different temperaturesnormalized by surface area. (B) Mass change compared with other materials. [Data from (24–30)] (C) Fracture surface of a SiOC microlattice heat-treated1300°C/10hours + 1500°C/10hours selected for extraction of (D) Focused ion beam lamella. (E) TEM image of theSiOC region. (F) TEM image of theSiO2 region.

RESEARCH | REPORTS

of the O2 into the structure and COX products outof the bulk. Upon cooling, there is a phasechange with 7% volume change in the cristobalite(21), as well as a large shift in coefficient ofthermal expansion (22), which leads to a crackedsurface oxide. Upon reheating, the oxidation ap-pears to restart underneath the cracked oxidelayer, leading to a multilayer oxide scale afterseveral heat treatments. X-ray diffraction did notdetect phases other than cristobalite, indicatingthat bulk crystallizationproducts, specifically b-SiC,were not present or were below the detectionlimit due to their small size and volume fraction(fig. S3). TEMof a sample heat-treated for 10 hoursat 1300°C followed by 10 hours at 1500°C re-vealed the onset of bulk crystallization with scat-tered b-SiC crystals <10 nm inside the amorphousmatrix. A lamella was milled out of a fracturedsurface of amicrolattice strut, as indicated by therectangle in Fig. 4C, so that oxide and SiOC basematerial could be analyzed (Fig. 4D). Bright-fieldimages showed small crystallites of a few nano-meters in size in both the oxide and SiOC region.High-resolution imaging could identify the crys-tallites as graphite and b-SiC, based on the latticespacing and diffraction pattern (Fig. 4E). Thesmall size of 5 to 10 nm of the crystals and thehigh fraction of remaining amorphous matrixindicate that crystallization had just started. Thecrystallites in the silicon oxide region are evensmaller (Fig. 4F), consistent with the recent for-mation of this oxide region. Larger crystals areprobably present in older oxide layers further fromthe interface, contributing to the cristobalite dif-fraction pattern recorded by XRD below. Note-worthy were small pores in the SiOC region thatwere not observed before the heat treatmentsand presumably developed due to carbon leavingas CO or CO2 gas.This indicates that theamorphousSiO1.34C1.25S0.15

is more stable than other silicon oxycarbide com-positions, which crystallize sooner (23). The high-temperature stability with respect to mass changein air is compared with other materials in Fig. 4B(mass change was extrapolated from reportedmass versus time curves after 1 hour exposure inair). The silicon-oxycarbide structures show bet-ter oxidation performance than silicon oxycarbidematerials from previous studies, which used dif-ferent starting precursors, compositions, and py-rolysis temperatures (20, 24, 25). Silicon oxycarbideis more resistant to oxidation than SiC and Si3N4

and has been investigated as oxidation protectioncoating for these materials (8).Various ceramic compositions can be processed

with our approach, including materials that aredifficult to form via sintering of powders, such asSiOC, Si3N4, and SiC ceramics. In this demonstra-tion, we focused on structures out of silicon oxy-carbide, and our cellular SiOC materials exhibitstrength 10 times as high as commercially avail-able ceramic foams of similar density and survivetemperatures of 1700°C in air with surface oxi-dation. Such cellular ceramic materials are ofinterest for the core of lightweight, load-bearingceramic sandwich panels for high-temperatureapplications—for example, in hypersonic vehicles

and jet engines. Stereolithography of ceramics willopenopportunities forcomplex-shaped, temperature-and environment-resistant ceramic structures fromthe microscale—e.g., in microelectromechanicalsystems (MEMS) or device packaging—to themacroscale—e.g., in propulsion or thermal protectionsystems.

REFERENCES AND NOTES

1. J. Deckers, J. Vleugels, J.-P. Kruth, J. Ceram. Sci. Technol. 5,245–260 (2014).

2. N. Travitzky et al., Adv. Eng. Mater. 16, 729–754 (2014).3. A. Zocca, P. Colombo, C. M. Gomes, J. Günster, J. Am. Ceram.

Soc. 98, 1983–2001 (2015).4. T. T. Wohlers, T. Caffrey, Wohlers Report (Wohlers Associates,

Fort Collins, CO, 2013).5. P. Colombo, G. Mera, R. Riedel, G. D. Sorarù, J. Am. Ceram.

Soc. 93, 1805–1837 (2010).6. S. Martínez-Crespiera et al., Sens. Actuators A Phys. 169,

242–249 (2011).7. L.-A. Liew et al., Sens. Act. A 95, 120–134 (2002).8. P. Colombo, R. Riedel, G. D. Sorarù, H. J. Kleebe, Eds., Polymer

Derived Ceramics (DEStech Publications, Lancaster, PA, 2010).9. M. Schulz et al., Adv. Eng. Mater. 6, 676–680 (2004).10. A. J. Jacobsen, W. Barvosa-Carter, S. Nutt, Adv. Mater. 19,

3892–3896 (2007).11. T. A. Schaedler et al., Science 334, 962–965 (2011).12. R. Riedel et al., Nature 382, 796–798 (1996).13. M. Zaheer, T. Schmalz, G. Motz, R. Kempe, Chem. Soc. Rev. 41,

5102–5116 (2012).14. P. Colombo, J. R. Hellmann, D. L. Shelleman, J. Am. Ceram.

Soc. 84, 2245–2251 (2001).15. L. J. Gibson, M. F. Ashby, Cellular Solids: Structure and

Properties (Cambridge Univ. Press, Cambridge, 1997).16. A. J. Jacobsen, W. Barvosa-Carter, S. Nutt, Acta Mater. 55,

6724–6733 (2007).17. A. J. Jacobsen, W. Barvosa-Carter, S. Nutt, Acta Mater. 56,

2540–2548 (2008).18. T. Varga et al., J. Am. Ceram. Soc. 90, 3213–3219 (2007).

19. A. Saha, R. Raj, D. L. Williamson, J. Am. Ceram. Soc. 89,2188–2195 (2006).

20. T. Xu, Q. Ma, Z. Chen, Ceram. Int. 37, 2555–2559 (2011).21. M. D. Beals, S. Zerfoss, J. Am. Ceram. Soc. 27, 285–292 (1944).22. L. Huang, J. Kieffer, J. Chem. Phys. 118, 1487–1498 (2003).23. A. Saha, R. Raj, J. Am. Ceram. Soc. 90, 578–583 (2007).24. S. Modena, G. D. Soraru, Y. Blum, R. Raj, J. Am. Ceram. Soc.

88, 339–345 (2005).25. G. Chollon, J. Eur. Ceram. Soc. 20, 1959–1974 (2000).26. W. C. Tripp, H. C. Graham, J. Am. Ceram. Soc. 59, 399–403 (1976).27. E. Opila, S. Levine, J. Lorincz, J. Mater. Sci. 39, 5969–5977 (2004).28. W. C. Tripp, H. C. Graham, J. Electrochem. Soc. 118, 1195–1199 (1971).29. J. A. Coppala, M. Srinivasan, K. T. Faber, R. H. Smoak (The

Carborundum Company, USA)., in Proceedings of InternationalSymposium on Factors in Densification and Sintering of Oxideand Non-oxide Ceramics, October 3 to 5, 1978, Hakone, Japan(Gakujutsu Bunken Fukyu-kai, Tokyo, 1979), pp. 400–417.

30. N. M. Geyer, Aeronautical Systems Division Technical Report61-322, from the published Proceedings for the MaterialsSymposium, 13 to 15 September 1961, Phoenix, AZ.

ACKNOWLEDGMENTS

The authors gratefully acknowledge financial support byHRL Laboratories, LLC, and the Defense Advanced ResearchProjects Agency under the Materials with ControlledMicrostructural Architecture program managed by J. Goldwasser(contract no. W91CRB-10-0305) and thank N. Verma (Universityof California, Santa Barbara) for TEM analysis and C. G. Leviand C. S. Roper for useful discussions. Patent applications havebeen filed under serial numbers 62/183580, 62/128410, and62/092733 with the U.S. Patent and Trademark office.

SUPPLEMENTARY MATERIALS

www.sciencemag.org/content/351/6268/58/suppl/DC1Materials and MethodsFigs. S1 to S4Tables S1 to S3References

18 August 2015; accepted 13 November 201510.1126/science.aad2688

BLACK HOLE PHYSICS

A radio jet from the optical and x-raybright stellar tidal disruption flareASASSN-14liS. van Velzen,1* G. E. Anderson,2,3 N. C. Stone,4 M. Fraser,5 T. Wevers,6 B. D. Metzger,4

P. G. Jonker,6,7 A. J. van der Horst,8 T. D. Staley,2 A. J. Mendez,1 J. C. A. Miller-Jones,3

S. T. Hodgkin,5 H. C. Campbell,5 R. P. Fender2

The tidal disruption of a star by a supermassive black hole leads to a short-lived thermalflare. Despite extensive searches, radio follow-up observations of known thermal stellar tidaldisruption flares (TDFs) have not yet produced a conclusive detection.We present adetection of variable radio emission from a thermal TDF, which we interpret as originatingfrom a newly launched jet. The multiwavelength properties of the source present a naturalanalogy with accretion-state changes of stellar mass black holes, which suggests that allTDFs could be accompanied by a jet. In the rest frame of the TDF, our radio observations arean order of magnitude more sensitive than nearly all previous upper limits, explaining howthese jets, if common, could thus far have escaped detection.

Although radio jets are a ubiquitous andwell-studied feature of accreting compact ob-jects, it remains unclear why only a subsetof active galactic nuclei (AGNs) are radio-loud. A stellar tidal disruption flare (TDF)

presents a novel method with which to study jet

production in accreting supermassive black holes.These flares occur after perturbations to a star’sorbit have brought it to within a few tens ofSchwarzschild radii of the central supermassiveblack hole and the star gets torn apart by the blackhole’s tidal force. A large amount of gas is suddenly

62 1 JANUARY 2016 • VOL 351 ISSUE 6268 sciencemag.org SCIENCE

RESEARCH | REPORTS