Nanotwinned Diamond with Unprecedented Hardness and Stability

LETTER doi:10.1038/nature13381

Nanotwinned diamond with unprecedented hardness and stability

Quan Huang1*, Dongli Yu1*,BoXu1*, Wentao Hu1*, Yanming Ma2, Yanbin Wang3, Zhisheng Zhao1, Bin Wen1, Julong He1, Zhongyuan Liu1 & Yongjun Tian1

Although diamond is the hardest material for cutting tools, poor synthesized from C60, amorphous carbon and glassy carbon, but Knoop thermal stability has limited its applications, especially at high tem- hardness decreases significantly to 70286 GPa (ref. 6). The observed peratures. Simultaneous improvement of the hardness and thermal hardness deficiency seems to originate from intergranular fracturing stability of diamond has long been desirable. According to the Hall2 along poorly sintered grain boundaries, rather than the reverse Hall2 Petch effect1,2, the hardness of diamond can be enhanced by nanostruc- Petch effect resulting from grain-boundary sliding6. Technically, the turing(bymeansof nanograinedand nanotwinned microstructures), synthesis of well-sintered nanograined diamond while maintaining a as shown in previous studies3–7. However, for well-sintered nano- smaller grain size remains a challenge. grained diamonds, the grain sizes are technically limited to 10230 nm Nanotwinning is an effectivemechanism for acquiring a smaller char- (ref. 3), with degraded thermal stability4 compared with that of natural acteristic size of microstructure, because twin boundaries possess lower diamond. Recent success in synthesizing nanotwinned cubic boron excess energy than grain boundaries. It has been verified experimentally nitride (nt-cBN) with a twin thickness down to ,3.8 nm makes it that, at nanoscale, twin boundaries show a hardening effect identical to feasible to simultaneously achieve smaller nanosize, ultrahardness those of grain boundaries for metals11,12. Ubiquitously nanotwinned struc- and superior thermal stability5. At present, nanotwinned diamond tures have been introduced into superhard materials through the success- (nt-diamond) has not been fabricated successfully through direct con- ful synthesis of nt-cBN with an average twin thickness of ,3.8 nm at versions of various carbon precursors3,6,7 (such as graphite, amorphous HPHT5. These nt-cBN bulk samples have a superior combination of 5 carbon, glassy carbon and C60). Here we report the direct synthesis high hardness, high toughness and high thermal stability . The syn- of nt-diamond with an average twin thickness of ,5 nm, using a thesis of nt-diamond has not yet been reported but is highly desirable precursor of onion carbon nanoparticles at high pressure and high in view of the excellent performance of nt-cBN. temperature, and the observation of a new monoclinic crystalline Experience in the synthesis of nt-cBN through an onion-like BN pre- form of diamond coexisting with nt-diamond. The pure synthetic cursor suggested the use of onion carbon as precursor in the fabrication bulk nt-diamond material shows unprecedented hardness and ther- of nt-diamond. Onion carbon, a high-energy metastable carbon consist- mal stability, with Vickers hardness up to ,200 GPa and an in-air ing of concentric graphite-like shells (Extended Data Fig. 1), is structur- oxidization temperature more than 200 6C higher than that of nat- ally similar to onion-like BN and can be produced in large amounts13.A ural diamond. The creation of nanotwinned microstructures offers a high concentration of puckered layers and stacking faults in onion car- general pathway for manufacturing new advanced carbon-based ma- bon may provide the key for the nucleation of nt-diamond at HPHT, as terials with exceptional thermal stability and mechanical properties. for nt-cBN5. In fact, isolated onion carbon particles have been observed Diamond is the hardest, stiffest and least compressible crystalline toconvert into diamond nanocrystalsunder intense electron irradiation material with exceptionally high thermal conductivity. Tools made of even at ambient pressure14. diamond are widely used for cutting and shaping hard substances such The onion carbon nanoparticles (,20250 nm in diameter) used in as stones, glasses and ceramics. However, diamond is energetically unsta- our study were characterized by transmission electron microscopy (TEM) ble relative to graphite under ambient conditions, with an inherent draw- to contain numerous puckering and stacking faults (Fig. 1a). X-ray dif- back of poor thermal stability. In air, the onset oxidation temperature fraction (XRD) characterization of the onion precursors and recovered is ,800 uC for natural diamond8,9, resulting in the severe wear of dia- samples after HPHT treatments is presented in Extended Data Figs 2 mond tools at high temperatures. and 3. The inter-shell spacings of untreated onion carbon were centred The synthesisof materialsharder than natural diamond haslong been on 0.3485 nm. When treated below 10 GPa and 2,000 uC, onion carbon sought10. The Hall2Petch relation1,2 offers a general pathway to enhan- retained the original nested crystal structure. Samples recovered from cing hardness by decreasing characteristic size of microstructures (for 10215 GPa and 1,400–1,850 uC were black and opaque (Fig. 1b inset), example grain size or twin thickness). Nanograined diamond has been and contained cubic diamond and an unidentified carbon phase. This successfully synthesized through direct conversions of certain carbon latter phase has not been observed before and seems to be inherently precursorsat high pressureand hightemperature (HPHT)3,6,7. The pres- related to the specific structural transformation of onion carbon precur- sure and temperature conditions6 needed to synthesize nanograined dia- sors at HPHT. Transparent samples were recovered from 18225 GPa monds are much higher than those for growing single-crystal diamonds and 1,85022,000 uC, with pure cubic diamond as indicated by the XRD in the industry. High pressure is necessary to control grain size effec- patterns. The synthetic temperature of cubic diamond from onion car- tively by suppressing atomic diffusion, which promotes growth. Nano- bon was ,450 uC lower than that from graphite3,6, allowing easier indus- grained diamonds synthesized from pure graphite at 2,30022,500 uC trial fabrication. and 12225 GPa reach a grain size of 10–30 nm, with a high Knoop Typical TEM and high-resolution TEM (HRTEM) images of a black hardness of 110–140 GPa (ref. 3) but a reduced onset oxidation tem- opaque sample (synthesized at 10 GPa and 1,850 uC) are shown in perature of ,680 uC in air4. At lower temperatures (,1,800 uC), nano- Fig. 1b, c and Extended Data Fig. 4a, b. Cubic diamond was the dom- grained diamonds with a smaller grain size (5210 nm) have been inant phase, with lamellar {111} nanotwins. The new secondary carbon

1State Key Laboratory of Metastable Materials Science and Technology, Yanshan University, Qinhuangdao 066004, China. 2State Key Laboratory for Superhard Materials, Jilin University, Changchun 130012, China. 3Center for Advanced Radiation Sources, University of Chicago, Chicago, Illinois 60439, USA. *These authors contributed equally to this work.

a b c M

0.5 mm

C M

C 10 nm 100 nm 5 nm

Figure 1 | Onion carbon nanoparticles and a bulk sample synthesized at (C) domains form a {111} twin boundary (TB). Several M-diamond (M) 10 GPa and 1,850 6C. a, HRTEM image of onion carbon nanoparticles. domains are associated with cubic diamond twins containing stacking faults b, TEM image of the sample showing nanotwinned microstructure. Inset: (SFs). Fast Fourier transforms of M-diamond and cubic diamond, shown in the photograph of the black opaque sample (,2 mm in diameter). c, HRTEM upper and lower insets, respectively, indicate that lattices of M-diamond and image of the area marked with the red box in b. Two adjacent cubic diamond cubic diamond are coherent. phase was clearly seen with HRTEM. The d spacings deduced from interrupted by interlocked areas where adjacent nanocrystals inter- selected-area electron diffraction (SAED) patterns (Extended Data Fig. 4d–f) sected and merged, making it difficult to determine individual nano- and XRD data (Extended Data Table 1) of this new phase do not match grains unambiguously. The nanotwins were predominantly thinner any reported carbon phase. The new phase (denoted M-diamond) had than 10 nm. Figure 2c shows a twin thickness distribution derived from a monoclinic structure with lattice parameters of a 5 0.436 nm, b 5 444 nanotwins on the basis of HRTEM measurements. The average 0.251 nm, c 5 1.248 nm and b 5 90.9u. All the C2C bonds were sp3 thickness, ,5 nm, is the smallest microstructural size so far achieved in hybridized, as indicated by the electron energy loss spectrum measure- diamonds. In our transparent nt-diamond samples, stacking faults were ments (Extended Data Fig. 4c), similar to those in cubic diamond. In also observed in nanotwins (Fig. 2b and Extended Data Fig. 5). These the TEM images, thin, elongated (and occasionally polygonal) M- stacking faults, due to extensive twinning, altered the stacking sequence diamond domains intersected adjacent nanotwinned cubic diamond of (111) planes in diamond15 and produced weak shoulders of the strong (C-diamond) domains, forming coherent boundaries parallel to the dia- (111) reflection in the XRD patterns (Extended Data Fig. 2). These ob- mond (111) planes. The orientation relations between M-diamond (M) served planar faults together with the secondary phase of M-diamond and C-diamond (C) as determined from SAED were M(001)//C(111) also caused the asymmetries in both the (111) and the (220) peaks of and M[010]//C[011] (Extended Data Fig. 4d–f). diamond (Extended Data Figs 2 and 3). The HRTEM images of a transparent pure nt-diamond sample (syn- A hardness value should be determined by the asymptotic region of thesized at 20 GPa and 2,000 uC; Fig. 2a inset) revealed that C-diamond the hardness–load curve10,16. We found that our samples reached asymp- contained a high density of lamellar {111} nanotwins (Fig. 2a, b). Unlike totic hardness at a load of 4.9 N (Fig. 3a). Vickers and Knoop hardnesses nt-cBN, in which individual nanograins can be clearly characterized5, measured at 4.9 N for six different transparent pure nt-diamond sam- high-angle grain boundaries in nt-diamond (Fig. 2b) were frequently ples (Fig. 3b and Extended Data Table 2) showed unprecedentedly high

a b c GB

4 0.5 mm Interlocked area M 3

Percentage of total 1

SFs

SFs 0 0 2 4 6 8 10 12 14 16 18 20 Twin thickness (nm) 30 nm 5 nm

Figure 2 | A nt-diamond bulk sample synthesized at 20 GPa and 2,000 6C. boundaries (GB) are interrupted by interlocked twins. Inset: SAED pattern a, TEM image of nanotwinned microstructure. Inset: photograph of the corresponding to the central area of a. The four-fold-like pattern is from the transparent sample (,1 mm in diameter). b, HRTEM image of intersecting twin domains with four different orientations. c, Thickness distribution of the nanotwins (marked with the red box in a), viewed along the [101] zone axis of nanotwins measured from HRTEM images. The average twin thickness is diamond. Lamellar {111} nanotwins, stacking faults and residual M-diamond ,5nm. (trace) are present. Twin boundaries are marked with red arrows. Grain

12 JUNE 2014 | VOL 510 | NATURE | 251 ©2014 Macmillan Publishers Limited. All rights reserved RESEARCH LETTER values: 1752203 and 1682196 GPa, respectively. Two high loads of 9.8 the fracture toughness—can be broken through processes of controlled and 19.6 N were applied to create cracks for fracture toughness deter- nanotwinning in covalent materials. mination. The determined fracture toughness values ranged from 9.7 The thermal stability of different pure nt-diamond samples was char- to 14.8 MPa m0.5 (Fig. 3c and Extended Data Table 2). Meaningful in- acterized by thermogravimetry curves measured in air. At a heating rate dentation hardness can be measured reliably as long as the shear strength of 5 uC min21, the onset oxidation temperatures of nt-diamond and of the sample is smaller than the compressive strength of the diamond natural diamond were ,980 and ,770 uC (Fig. 4a), respectively. Ex- indenter16; this requirement was satisfied because no visible plastic de- tended Data Fig. 7 compares the thermal stability of nt-diamond with formation of indenter diamond tip was observed after measurements other tool materials measured at a heating rate of 10 uC min21. The of hardness and fracture toughness (Extended Data Fig. 6). Both the onset oxidation temperature of nt-diamond (,1,056 uC) was again much achieved hardness and the trade-off between hardness and toughness higher than those of natural diamond (,805 uC), synthetic diamond of our nt-diamond samples are significantly superior to those of other powders (,725 uC), nanograined diamond (,680 uC)4 and Co-WC popular tool materials, such as cobalt-bonded tungsten carbide (Co-WC)17 (80021,000 uC)17 (Fig. 4b), and even rivalled that of ng-cBN (,1,187 uC)4. and previously reported diamond-related materials18–21 (Fig.3b,d),yield- The oxidation of diamond generally has two simultaneous processes9, ing diamonds with unsurpassed mechanical properties. The simultaneous namely the oxidation of graphitized diamond and the oxidation of dia- improvement in hardness and fracture toughness in our nt-diamond mond itself. Previous experiments have shown that the oxidation tem- is intimately related to the ubiquitous nanotwinning microstructure. perature of graphite in air is ,50 uC lower than that of diamond23. The presence of ultrafine nanotwins introduces extra hardening, which According to the size-dependent pressure2temperature phase diagram is probably due to both the Hall–Petch and quantum confinement ef- derived from nanothermodynamic theory24, diamondbecomes energet- fects at nanoscale5, while gliding of dislocations along densely distributed ically stable over graphite at deep nanometre scale (,5 nm). This would twin boundaries enhances fracture toughness22. Our results demonstrate certainly delay the graphitization of nt-diamond and would result in a that the old paradigm—the higher the hardness of a material, the lower higher oxidation temperature. Moreover, compressive stress introduces additional resistance to the oxidation of diamond. Given that the internal stress induced by nanotwinning boundaries increases with reduced 25 a b twin thickness , the oxidation process of nt-diamond may be retarded 500 250 450 because of the presence of ultrafine nanotwins. Differential scanning 400 nt-D ng-D

SC-D calorimetry (DSC) measurements provided further evidence that thin- 350 200

400 Co-WC (GPa) 300 Co-PCD , K ner nanotwins result in an even higher oxidation temperature of 1,300 uC H 250 (Extended Data Fig. 7a), consistent with the aforementioned specu- 300 200 150 3 4 5 6 7 8 F lation. Thus, both mechanical properties and thermal stability depend

(GPa) (N) (GPa) V V

H 200 H 100 primarily on the achieved average twin thickness. The successful syntheses of nt-diamond and nt-cBN show that nano- 100 50 twinning microstructure is an effective route for simultaneously enhan- cing the hardness, fracture toughness and thermal stability of superhard 0 2 3 4 5 6 7 8 0 materials. Our experimental results on nt-diamond further confirm F (N) c d that there is continuous hardening at nanotwinning sizes down to ,5nm, 20 250 which agrees with previous results on nt-cBN5 but is in stark contrast with the sharp softening of metals at these nanometre scales. We therefore 200 ) 15 predict that pursuing microstructure with thinner nanotwin sizes may 0.5 150 lead to findings of covalent materials with even superior properties. (GPa) 10 V Here it may be instructive to estimate the lower limit of nanotwin H (MPa m 100 Ic thickness and the corresponding ultimately achievable hardness (Hua) K of diamonds. If we take {111} twins in nt-diamond as the model sys- 5 50 tem, the estimated minimal twin thickness, lmin,is3d111 5 0.618 nm ng-D nt-D SC-D Co-PCD Co-WC 0 0 0 5 10 15 20 K (MPa m0.5) a b Ic 770 °C ° 100 980 C nt-D

Figure 3 | Typical mechanical properties of nt-diamond and its comparison 1,200 ng-D SC-D Co-WC

with other tool materials. a, H of nt-diamond and natural diamond crystal as Co-PCD V 50 a function of applied load (F). Beyond 4.9 N, HV decreases to the asymptotic nt-D 1,000 values of ,200 GPa for nt-diamond (red line). For natural diamond crystals, TG (mass%) Natural diamond C) ) 0 ° ° ° our measured H values are ,110 GPa on the {110} face (blue line) and 720 C 960 C ( V –1 a ,62 GPa on the {111} face (pink line). Error bars indicate 1 s.d. (n 5 5). Inset: 0 T 800 plot of HK against F for nt-diamond. b, c, HV (b) and KIc (c) for different –8 tool materials, including nt-diamond (nt-D), nanograined diamond (ng-D; grain size 10230 nm)3, single-crystal diamond (SC-D)18, cobalt-bonded –16 600 Exothermic 19 17 polycrystalline diamond (Co-PCD) and Co-WC . d, Plot of H against K Heat flow (W g –24 V Ic 200 400 600 800 1,000 1,200 for nt-diamond in comparison with available data on other forms of diamond. T (°C) The data for nt-diamond are shown as solid red circles above the shaded envelope. The published data are from representative diamond materials, Figure 4 | Typical thermal stability of a nt-diamond sample. a, Comparison including type Ia natural SC-D (open upward triangles18), IIa natural SC-D of the onset oxidation temperatures of a nt-diamond bulk sample (red) with a (open squares18), HPHT-grown SC-D (open downward triangles18), CVD- natural diamond crystal (cyan). Both thermogravimetry (TG; top) and DSC grown SC-D (open hexagons18), annealed IIa natural SC-D (filled squares18), (bottom) curves were measured in air at a heating rate of 5 uC min21. The onset CVD-grown SC-D annealed at HPHT (filled hexagons18), Co-PCD (large grey oxidation temperature of the nt-diamond (980 uC from thermogravimetry or circle19), CVD-grown PCD (large pink oval20) and aggregated diamond rod 960 uC from DSC) was more than 200 uC higher than that of the natural (Knoop hardness, filled upward triangle21). The hardness of nanograined diamond (770 uC from thermogravimetry or 720 uC from DSC). b, Comparison diamond reaches 1102140 GPa (ref. 3), but no fracture toughness data of working temperatures (Ta) in air of nt-diamond with other tool materials, were reported. Those data are therefore not included in the figure. including ng-D4,SC-D21, Co-PCD3 and Co-WC17.

252 | NATURE | VOL 510 | 12 JUNE 2014 ©2014 Macmillan Publishers Limited. All rights reserved LETTER RESEARCH because the atomic stacking sequence along the Æ111æ direction of dia- 5. Tian, Y. et al. Ultrahard nanotwinned cubic boron nitride. Nature 493, 385–388 (2013). mond (lattice parameter a 5 0.3568 nm) is …ABCABC…(Extended 6. Sumiya, H. & Irifune, T. Hardness and deformation microstructures of nano- Data Fig. 8). Assuming that the Hall2Petch effect is no longer applicable polycrystalline diamonds synthesized from various carbons under high pressure 26,27 and high temperature. J. Mater. Res. 22, 2345–2351 (2007). at such a scale , Hua for nt-diamond is estimated with the fol- 5,28 7. Sumiya, H. & Harano, K. Distinctive mechanical properties of nano-polycrystalline lowing formula according to our hardness model : Hua 5 H0 1 kqc/ diamond synthesized by direct conversion sintering under HPHT. Diamond Relat. lmin,whereH0 is the hardness of single-crystal diamond (,90 GPa) Mater. 24, 44–48 (2012). 1=3 8. Sun, Q. & Alam, M. Relative oxidation behavior of chemical vapor deposited and and kqc 5 211Ne 5 187.7 GPa nm is the quantum confinement hardening coefficient for a covalent crystal28, which is proportional to the type IIa natural diamonds. J. Electrochem. Soc. 139, 933–936 (1992). 9. Johnson, C. E., Bennett, J. M. & Nadler, M. P. Oxidation of diamond windows. valence electron density Ne (0.705, ref. 29). Thus, Hua for nt-diamond J. Mater. Res. 10, 2555–2563 (1995). is 394 GPa. This presents a technical challenge to synthesize nanotwinned 10. Chaudhri, M. M. & Lim, Y. Y. Harder than diamond? Just fiction. Nature Mater. 4, 4 microstructures with the required twin thickness to achieve such an (2005). 11. Lu, L., Shen, Y. F., Chen, X. H. & Lu, K. Ultrahigh strength and high electrical exceptional hardness property. conductivity in copper. Science 304, 422–426 (2004). Finally, the experimental HPHT conditions for synthesizing nt- 12. Lu, L., Chen, X., Huang, X. & Lu, K. Revealing the maximum strength in nanotwinned diamond and nt-cBN are essentially identical. This opens up the possi- copper. Science 323, 607–610 (2009). 13. Choucair, M. & Stride, J. A. The gram-scale synthesis of carbon onions. Carbon 50, bility of manufacturing nt-diamond/nt-cBN composites. Such nanotwinned 1109–1115 (2012). composites are expected to possess intermediate oxidation temper- 14. Banhart, F. & Ajayan, P. M. Carbon onions as nanoscopic pressure cells for ature and hardness between those of nt-diamond and nt-cBN but with diamond formation. Nature 382, 433–435 (1996). greater fracture toughness as a result of the combined contributions 15. Silva, F., Be´ne´dic, F., Bruno, P. & Gicquel, A. Formation of Æ110æ texture during nanocrystalline diamond growth: an X-ray diffraction study. Diamond Relat. Mater. from nanotwinning and composite effects. 14, 398–403 (2005). 16. Tian, Y. et al. Controversy about ultrahard nanotwinned cBN reply. Nature 502, METHODS SUMMARY E2–E3 (2013). 17. Upadhyaya, G. S. Materials science of cemented carbides—an overview. Mater. We fabricated onion carbon particles with diametersof ,20250 nm by usingblack Des. 22, 483–489 (2001). 30 carbon powders through an impinging-streams technology . HPHT experiments 18. Yan, C. S. et al. Ultrahard diamond single crystals from chemical vapor deposition. were performed with a 10-MN double-stage large-volume multi-anvil system with Phys. Status Solidi A 201, R25–R27 (2004). the standard COMPRES 10/5 sample assembly consisting of a 10-mm spinel (MgAl2O4) 19. Lammer, A. Mechanical properties of polycrystalline diamonds. Mater. Sci. Technol. 4, 949–955 (1988). 1 MgO octahedron with a Re heater and a LaCrO3 thermal insulator. Temperature was measured with type C W–Re thermocouples, and pressure was estimated from 20. Sussmann, R. S. et al. CVD diamond windows for infrared synchrotron applications. Nuovo Cimento D 20, 503–525 (1998). previously obtained calibration curves at different temperatures for the multi-anvil 21. Dubrovinskaia, N., Dub, S. & Dubrovinsky, L. Superior wear resistance of apparatus5. Recovered samples were ,122 mm in diameter and 0.220.5 mm in aggregated diamond nanorods. Nano Lett. 6, 824–826 (2006). height. Microstructures were investigated with a transmission electron microscope 22. Lu, K., Lu, L. & Suresh, S. Strengthening materials by engineering coherent internal (JEM-2010) with an accelerating voltage of 200 kV. Component phases were iden- boundaries at the nanoscale. Science 324, 349–352 (2009). tified by TEM and XRD (Cu K ; D8 Discover). A microhardness tester (KB 5 BVZ) 23. Joshi, A., Nimmagadda, R. & Herrington, J. Oxidation kinetics of diamond, graphite, a and chemical vapor deposited diamond films by thermal gravimetry. J. Vac. Sci. was used to measure HV and KIc with a diamond Vickers indenter as well as HK with Technol. A 8, 2137–2142 (1990). ~ 2 a diamond Knoop indenter. HV was determined from HV 1,854:4F=L1,whereF 24. Yang, C. C. & Li, S. Size-dependent temperature2pressure phase diagram of (in newtons) is the applied load and L1 (in micrometres) is the arithmetic mean of carbon. J. Phys. Chem. C 112, 1423–1426 (2008). the two diagonals of the Vickers indentation. HK was determined from HK~ 25. Weissmu¨ller, J. & Cahn, J. W. Mean stresses in microstructures due to interface 2 stresses: a generalization of a capillary equation for solids. Acta Mater. 45, 14,228:9F=L2, where L2 (in micrometres) is the longer diagonal of the Knoop indentation. Five hardness data points were obtained at each load, and the hardness 1899–1906 (1997). 26. Yip, S. Nanocrystals: the strongest size. Nature 391, 532–533 (1998). values were determined from the asymptotic-hardness region. KIc was calculated 27. Li, X., Wei, Y., Lu, L., Lu, K. & Gao, H. Dislocation nucleation governed softening 0.5 1.5 from KIc 5 0.016(E/HV) F/C for radial cracks formed in the bulk nt-diamond and maximum strength in nano-twinned metals. Nature 464, 877–880 sample18, where C (in micrometres) is the average length of the radial cracks mea- (2010). sured from the indent centre, and E 5 1,000 GPa is Young’s modulus of diamond18. 28. Tian, Y., Xu, B. & Zhao, Z. Microscopic theory of hardness and design of novel The presented K values were averaged over three data points determined at loads superhard crystals. Int. J. Refract. Met. Hard Mater. 33, 93–106 (2012). Ic 29. Gao, F. M. et al. Hardness of covalent crystals. Phys. Rev. Lett. 91, 015502 (2003). of 9.8 and 19.6 N. Oxidation resistance was studied by measuring thermogravime- 30. Tamir, A. Impinging-stream Reactors: Fundamentals and Applications (Elsevier, try and DSC curves in air, using NETZSCH STA 449 C over the temperature range 1994). 2021,500 uC. Acknowledgements This work was supported by the National Natural Science Online Content Any additional Methods, Extended Data display items and Source Foundation of China (51121061), the Ministry of Science and Technology of China Data are available in the online version of the paper; references unique to these (2011CB808205 and 2010CB731605), the National Natural Science Foundation of sections appear only in the online paper. China (51332005, 51172197, 11025418 and 91022029) and the US National Science Foundation (EAR-0968456).

Received 17 November 2013; accepted 15 April 2014. Author Contributions Y.J.T. conceived the project. Y.J.T., D.L.Y., B.X. and Y.B.W. designed the experiments. Q.H. synthesized onion carbon precursors. Q.H., D.L.Y., B.X., 1. Hall, E. O. The deformation and ageing of mild steel. III. Discussion of results. Proc. Y.J.T., Y.B.W. and Z.S.Z. performed the HPHT experiments, W.T.H. performed TEM Phys. Soc. Lond. B 64, 747–753 (1951). observations, and B.W. performed molecular dynamics simulations. Y.J.T., B.X., D.L.Y., 2. Petch, N. J. The cleavage strength of polycrystals. J. Iron Steel Inst. 174, 25–28 Y.M.M., Y.B.W., J.L.H. and Z.Y.L. analysed the data. Y.J.T., B.X., Y.M.M. and Y.B.W. co-wrote (1953). the paper. All authors discussed the results and commented on the manuscript. 3. Irifune, T., Kurio, A., Sakamoto, S., Inoue, T. & Sumiya, H. Materials—ultrahard polycrystalline diamond from graphite. Nature 421, 599–600 (2003). Author Information Reprints and permissions information is available at 4. Solozhenko, V. L., Kurakevych, O. O. & Le Godec, Y. Creation of nanostuctures by www.nature.com/reprints. The authors declare no competing financial interests. extreme conditions: high-pressure synthesis of ultrahard nanocrystalline cubic Readers are welcome to comment on the online version of the paper. Correspondence boron nitride. Adv. Mater. 24, 1540–1544 (2012). and requests for materials should be addressed to Y.J.T. ([email protected]).

Extended Data Figure 1 | Schematic icosahedral model of a ten-shell onion carbon. The icosahedral-quasicrystal-like model of an onion carbon particle was relaxed from a nested buckyonion of C60,C240,C540,C960,C1,500,C2,160, C2,940,C3,840,C4,860 and C6,000. This model was constructed with the same classical molecular dynamics technique as that used in our previous work5.The spacings between adjacent shells in the model vary from ,0.300 nm to ,0.340 nm.

Extended Data Figure 2 | Phase transformation of onion carbon compacts at HPHT. XRD patterns of onion carbon precursor (Raw) and seven samples recovered from different conditions indicated by P (in GPa)–T (in uC) pairs. The inter-shell spacing of the starting onion carbon nanoparticles is ,0.3485 nm. For the two samples recovered from 8 GPa/2,000 uC and 15 GPa/ 1,200 uC, the onion carbon structure does not show significant alteration except that the inter-shell spacing decreases to 0.3305 and 0.3361 nm, respectively. Cubic diamond appears when the applied pressure is more than 10 GPa and temperature is more than 1,400 uC, with an accompanying new carbon phase recognized in the black opaque samples synthesized at 1,850 uC or below. A small amount of residual onion carbon can be detected in the sample recovered from 15 GPa/1,400 uC. At pressures of 18–25 GPa and temperatures of 1,850– 2,000 uC, the recovered samples changed from translucent to transparent, and only the diffraction peaks of cubic diamond can be seen in XRD patterns. Weak shoulders of the (111) peaks of diamond (red arrows) appear in three samples synthesized at pressures of 18220 GPa and temperatures of 1,85021,950 uC. Asymmetry in the (111) and (220) peaks of diamond was often observed in the samples synthesized at pressures below 20 GPa and temperatures below 1,950 uC.

Extended Data Figure 3 | XRD patterns of a sample recovered from 10 GPa and 1,850 6C. All the recorded d spacings of visible diffraction peaks are listed in Extended Data Table 1. Insets: two peaks overlapping the cubic diamond reflections. Most of these extra reflections can be indexed with a monoclinic structure (M-diamond) as shown in Extended Data Table 1.

Extended Data Figure 4 | TEM images, electron energy loss spectrum phases. d2f, SAED patterns along the [010], [150] and [130] zone axes of M, (EELS) and SAED measurements on a sample recovered from 10 GPa and respectively, recorded by rotating an M crystal. (111) and (200) spots of the 1,850 6C. a, TEM image showing interlaced twins. b, HRTEM image twinned C phase, overlapping with some spots of the M phase as a result of corresponding to the area in the red box in a. A monoclinic M-diamond (M) coherent growth, are marked by red circles and boxes, respectively. The domain is observed between two cubic diamond (C) domains. c, EELS spectra determined orientation relations between M and C phases are M(001)//C(111) of M and C phases. All the C–C bonds are sp3 hybridized in both M and C and M[010]//C[011].

Extended Data Figure 5 | HRTEM observations of three nt-diamond bulk transformation of onion carbon to diamond. The probability of stacking faults samples synthesized at different HPHT conditions. a–c, HRTEM and and the volume fraction of M-diamond decrease with elevated synthesis corresponding TEM (inset) images of three representative samples, O-366 temperature and pressure, as confirmed by our HRTEM observation. The (a), P-368 (b) and M-363 (c) as listed in Extended Data Table 2. TBs are marked abundant stacking faults in the nanotwins result in the appearance of a shoulder with red arrows. The measured average twin thicknesses are ,5.2 nm for near the (111) peak (Extended Data Fig. 2), for example in the XRD pattern of sample P-368, ,5.4 nm for sample O-366 and ,7.9 nm for sample M-363; the sample O-366. The asymmetries of the (111) and (220) peaks of diamond smaller the average twin thickness, the higher the hardness. The full width at shown in Extended Data Fig. 2 can be attributed to planar faults and the half-maximum (FWHM) of the (111) peak is mainly related to the nanograin secondary phase in microstructure. On the one hand, a twin fault can itself size: samples O-366 and P-368 have a larger FWHM as a result of their smaller produce peak asymmetry; on the other, M-diamond also contributes to peak nanograin size. Both pressure and temperature can promote the phase asymmetry because of peak overlap, as demonstrated in Extended Data Fig. 3.

Extended Data Figure 6 | Comparison of Vickers indenter tip before and after hardness and fracture toughness tests of nt-diamond. a, b, Scanning electron microscopy images of the square pyramid diamond tip before (a) and after (b) the tests of nt-diamond. A load of 9.8 N was used during the hardness and toughness tests. As shown in b, the indenter, with a dark imprint of ,6.9 mm 3 ,6.9 mm on the tip matching the permanent indentation on the tested nt-diamond, shows no visible plastic deformation. c, d, Photographs of indentations on the standard calibration block equipped by microhardness tester KB 5 BVZ. The indentations were formed at a load of 1.96 N before (c) and after (d) the tests, with the same tip as shown in a and b. The indenter tip produced an almost identical indentation (or standard hardness value) on the calibration block after the nt-diamond tests. These calibration results ensured the accuracy, repeatability and reliability of the unprecedented hardness and exceptional toughness values of nt-diamond reported in the present study.

Extended Data Figure 7 | Comparison of in-air oxidation resistance of bulk nt-diamond with other diamonds measured at a heating rate of 10 6Cmin21.a, Comparison of the onset oxidation temperatures determined from measured thermogravimetry curves. The onset temperature was ,1,056 uC for a bulk nt-diamond, ,805 uC for a natural diamond crystal, ,725 uC for synthetic diamond powders and ,680 uC for a nanograined diamond4. b, Comparison of the onset oxidation temperatures determined from the exothermic trough in the measured heat flow curves of DSC. The onset temperature was ,1,035 uC for the nt-diamond, ,750 uC for the natural diamond and ,705 uC for the synthetic diamond. The exothermic peaks located at 1,280 uC and 1,320 uC for the nt-diamond were probably due to the presence of finer nanotwins. The above-measured oxidation temperatures are consistent with those determined from the corresponding thermogravimetry curves.

P Extended Data Figure 8 | Atomic arrangements of a {111} 5 3 twin boundary in cubic diamond. The twin boundary is projected along the Æ011æ direction. Because of the stacking sequence of ABC for diamond structure, the minimum twin thickness is 3d111, where d111 is the planar distance along the direction of Æ111æ in the unit cell of cubic diamond.

Extended Data Table 1 | Comparison of d spacings (dobs) observed from XRD and SAED with those of proposed M-diamond structure and cubic diamond

The sample was synthesized at 10 GPa and 1,850 uC. The dcal values were calculated with the monoclinic structural parameters a 5 0.436 nm, b 5 0.251 nm, c 5 1.248 nm and b 5 90.9u. Asterisks indicate unknown peaks.

0.5 Extended Data Table 2 | Vickers hardness HV (GPa), Knoop hardness HK (GPa) and fracture toughness KIc (MPa m ) for six transparent pure (XRD standard) nt-diamond bulk samples

HV and HK values were measured at a fixed load of 4.9 N. The KIc values were measured at loads of 9.8 and 19.6 N. Error bars indicate 1 s.d. (n 5 5 for HV and HK, and n 5 3forKIc). The FWHMs of (111) peaks in the XRD patterns of nt-diamond samples are also listed.