This is the accepted version of the journal article: Hong, Seokmo; Lee, Chang-Seok (Samsung Advanced Institute of Technol- ogy. Inorganic Material Lab.); Lee, Min-Hyun (Samsung Advanced Institute of Technology. Inorganic Material Lab.); [et al.]. «Ultralow-dielectric-constant amorphous boron nitride». Nature, Vol. 582, issue 7813 (June 2020), p. 511-514. DOI 10.1038/s41586-020-2375-9 This version is available at https://ddd.uab.cat/record/236030 under the terms of the license 1 Ultra-low dielectric constant amorphous boron nitride 2 Seokmo Hong1, Chang-Seok Lee5, Min-Hyun Lee5, Yeongdong Lee3,9, Kyung Yeol Ma2,9, 3 Gwangwoo Kim1, Seong In Yoon2,9, Kyuwook Ihm6, Ki-Jeong Kim6, Tae Joo Shin7, Sang 4 Won Kim5, Eun-Chae Jeon10, Hansol Jeon3, Ju-Young Kim3, Hyung-Ik Lee8, Zonghoon 5 Lee3,9, Aleandro Antidormi11, Stephan Roche11,12, Manish Chhowalla13, Hyeon-Jin Shin5, and 6 Hyeon Suk Shin1,2,4,9 7 1 8 Department of Chemistry, 2Department of Energy Engineering, 3 School of Materials Science 9 and Engineering and 4Low-Dimensional Carbon Materials Center, Ulsan National Institute of 10 Science and Technology (UNIST), Ulsan 44919, Republic of Korea 5 11 Inorganic Material Lab., Samsung Advanced Institute of Technology (SAIT), Suwon 433-803, 12 Republic of Korea 6 13 Pohang Accelerator Laboratory, Kyungbuk 37673, Republic of Korea 7 14 UNIST Central Research Facilities & School of Natural Science, Ulsan 44919, Republic of 15 Korea 8 16 Analytical Engineering Group, Samsung Advanced Institute of Technology (SAIT), Suwon 17 433-803, Republic of Korea 9 18 Center for Multidimensional Carbon Materials, Institute for Basic Science (IBS), Ulsan 44919, 19 Republic of Korea 10 20 School of Materials Science and Engineering, University of Ulsan, Ulsan 44610, Republic of 21 Korea 11 22 Catalan Institute of Nanoscience and Nanotechnology (ICN2), CSIC and BIST, Campus UAB, 23 Bellaterra, Barcelona, Spain 12 24 ICREA, Institucio Catalana de Recerca i Estudis Avancats, Barcelona, Spain 13 25 Department of Materials Science and Metallurgy, University of Cambridge, Cambridge, UK 26 27 Corresponding authors: HSS (shin@unist.ac.kr), HJS (hyeonjin.shin@samsung.com) and MC 28 (mc209@cam.ac.uk) 29 1 1 The decrease in signal processing speed due to increased resistance and capacitance delay 2 resulting from aggressive miniaturisation of logic and memory devices is a major obstacle 3 for continued down scaling of electronics.1-3 In particular, minimizing the dimensions of 4 interconnects – metal wires that connect different device components on the chip – is 5 crucial for device scaling. The interconnects are isolated from each other by non- 6 conducting or dielectric layers. Much of the recent research has focused on decreasing 7 the resistance of scaled interconnects because integration of dielectrics using 8 complementary metal oxide semiconductor (CMOS) compatible processes has proven to 9 be exceptionally challenging. The key requirements for interconnect isolation materials 10 are that they should possess low relative dielectric constants (referred to as values), 11 serve as diffusion barriers against migration of interconnect metals such as cobalt into 12 semiconductors and be thermally, chemically and mechanically stable. In 2005, the 13 International Roadmap for Devices and Systems (IRDS) recommended dielectrics with 14 -values of < 2.2 and the most recent report recommends dielectric values of ≤ 2 by 15 2028.4 Despite this, state-of-the-art low- materials, such as silicon oxide derivatives, 16 organic compounds, and aerogels exhibit  values > 2 and possess poor thermo- 17 mechanical properties.5 Here, we report a dielectric thin film with ultra-low  values 18 of1.78 and 1.16 – close to that of air ( = 1) – at 100 kHz and 1 MHz, respectively, in 19 amorphous boron nitride (a-BN) obtained using CMOS compatible low temperature 20 process. We demonstrate that 3 nm thin a-BN is mechanically and electrically robust with 21 breakdown strength of 7.3 MV/cm – exceeding requirements. Cross-sectional 22 transmission electron microscopy reveals that a-BN is able to prevent diffusion of cobalt 23 interconnect atoms into silicon under very harsh accelerated conditions – in contrast with 2 1 reference barriers. Our results demonstrate that the amorphous counterpart of two- 2 dimensional hexagonal BN possesses the ideal low- dielectric characteristics for high- 3 performance electronics. 4 5 Modern high-performance logic and memory devices used in multifunctional electronics are 6 obtained by materials and design innovations that have enabled aggressive reduction of 7 transistor size and packing of more circuits in smaller areas.1-3,6-11 However, the reduction in 8 dimensions of metal interconnects and increased packing density has led to an increase in 9 resistance – capacitance (RC) delay that is becoming comparable to operation speed of the 10 devices. Ideally, both R and C should be simultaneously reduced for continuous scaling of 11 devices. However, development of electrically, mechanically and thermally robust low- 12 materials (using CMOS compatible processes that are good inter-metal and inter-layer 13 dielectrics as well as diffusion barriers against electro-migration of metal atoms from 14 interconnects has been challenging. 15 State-of-the-art strategies for achieving low- dielectrics have involved reducing the 16 polarisation strength and density of SiO2 ( = 4) by incorporation of fluorine ( = 3.7 for SiOF) 17 or CH3 ( = 2.8 for SiCOH) and introducing porosity (porous SiCOH or pSiCOH possess  = 18 2.4).1,12 The realisation of IRDS recommendations for 2028 calls for urgent development of 19 ultra-low- dielectrics with -values of less than 2.13,14 IRDS has also indicated that the greatest 20 challenge concerning interconnect development is the introduction of new materials with 21 reduced dielectric permittivity. Boron-based compounds such as BCN and amorphous BN (a- 22 BN) have been investigated as potential low- dielectrics – showing promising results.5 23 However, recent reports on BN dielectrics show films with turbostratic structure or high 3 1 dielectric constants that indicate crystalline structure.15,16 In this study, we utilized low 2 temperature inductively coupled plasma–chemical vapour deposition (ICP-CVD, see Methods 3 and Extended Data Fig. 1) to obtain 3 nm thick BN layers on Si substrates (See Extended Data 4 Fig. 2 for growth on Cu and SiO2 substrates). Transmission electron microscopy imaging and 5 diffraction results shown in Fig. 1 reveal that the films are amorphous with no discernible long- 6 range order – hence we refer to them as amorphous BN (a-BN). Cross-sectional chemical 7 mapping confirms that the films consist of B and N (Extended Data Fig. 3). X-ray photoelectron 8 spectroscopy (XPS) was used to obtain chemical information. The B/N atomic ratio was found 9 to be ~ 1:1.08 (Fig. 2a and b) with B 1s and N 1s peaks at 190.4 eV and 397.9 eV, respectively, 10 – indicating that the films are sp2-bonded B and N.17,18 Molecular dynamics simulations shown 11 in Extended Data Fig. 4 confirm the amorphous structure of BN films which is consistent with 12 the result in Fig. 1. 13 14 Fig. 1 | Atomic structure of amorphous boron nitride. (a) Low-magnification TEM image; 15 (b) Selective area electron diffraction showing diffuse pattern with no discernible crystalline 16 rings; (c) High-resolution TEM image; (d) Magnified image of red box in (c) demonstrating 17 disordered atomic arrangement; (e) Fast Fourier Transform results for area depicted in (d) 4 1 demonstrating diffuse diffraction pattern that is typical of an amorphous film. 2 3 Fig. 2 | Chemical structure of a-BN. XPS profiles for (a) B 1s and (b) N 1s peaks; (c) Raman 4 spectra of a-BN and epitaxially grown tri-layer h-BN (used as reference) on SiO2/Si. The 5 Raman spectrum of bare SiO2/Si substrate is identical to that of a-BN – suggesting that no 6 distinct crystalline h-BN modes are present in a-BN.; (d) FT–IR spectrum measured using s- 7 polarised radiation at an incident angle of 60°; (e) PEY-NEXAFS spectra for the B K-edge of 8 a-BN, measured at incident angles of 30°, 55°, and 70° - showing no dependence on orientation. 9 Raman spectra of a-BN and crystalline tri-layer hexagonal-BN (for comparison) reveal that 10 the h-BN E2g mode at 1373 cm-1 is absent in a-BN (Fig. 2c).17,18 Fourier transform infrared 11 spectroscopy (FTIR) spectrum in Fig. 2d shows that there is an absorption peak near 1370 cm- 1 12 that is attributed to the transverse optical mode of BN in a-BN. Another IR mode located near 13 1570 cm-1 confirms the amorphous nature of sp2-bonded BN19. We do not observe any N–H or 14 B–H bonds with FTIR (Extended Data Fig. 5). Detailed chemical and density analysis were 5 1 conducted with Rutherford Backscattering Spectroscopy (RBS) and Elastic Recoil Detection 2 Analysis (ERDA) – the results of which are shown in Extended Data Fig. 6. 3 Angle-dependent near-edge X-ray absorption fine structure (NEXAFS) measured in partial 4 electron-yield (PEY) mode at Pohang Light Source-II 4D beam line was used to investigate the 5 chemical and electronic structures of a-BN. In NEXAFS, X-ray absorption excites core 6 electrons of B and N to unoccupied states— that is, 1s electrons are excited to empty π* and/or 7 to σ* states. In the 1s  π* transition, the spatial orientation of π orbitals strongly impacts the 8 transition probability. Thus, information pertaining to the relative orientation of orbitals in h- 9 BN layers can be obtained by varying the incidence angle of X-rays.20 NEXAFS spectra 10 obtained for a-BN sample at incident angles of 30°, 55°, and 70° are shown in Figs. 2(e). The 11 observed resonance at 192 eV corresponds to the 1s  π* transition in boron.20 The resonance 12 intensity of the 1s  π* transition in a-BN demonstrates negligible variation with X-ray 13 incidence angle [Fig. 2(e)] – strongly indicating that BN planes are randomly oriented 14 throughout the material. Similar conclusions can be drawn from NEXAFS spectra of N K-edge 15 (Extended Data Fig. 7). Additionally, NEXAFS confirms that a-BN is completely sp2- 16 hybridised.20,21 For completeness, we also deposited BN films at different ICP – CVD 17 parameters such as power, temperature and pressure. We found that temperature was the most 18 important parameter with ideal amorphous BN film deposition occurring at 400oC at 30W 19 plasma power. Above this temperature we obtained nanocrystalline BN (nc-BN) as shown in 20 Extended Data Fig. 8 21 We now discuss the dielectric properties of a-BN. The dielectric constant is a physical 22 measure of how easily electric dipoles can be induced in materials by application of an 23 electrical field. The  value of air or vacuum is 1, but electric polarizability in solid state matter 6 1 arises from dipolar, atomic and electronic components that are most relevant for high 2 performance electronics. The contributions from these can be measured as a function of 3 frequencies ranging from 10-kHz–30-MHz. The relative dielectric constants () for a-BN and 4 h-BN, for comparison, at different frequencies are shown in Fig. 3a. It can be seen that - 5 values at 100 kHz are 3.28 and 1.78 for h-BN and a-BN, respectively. The values are average 6 of measurements on > 50 devices. The distribution of measured values and the corresponding 7 error bars at 100kHz are provided in Fig. 3b and Table I. Remarkably, at 1 MHz frequency, the 8 observed value for a-BN further reduces to 1.16, which is close to the value of air or vacuum. 9 The low values of a-BN are attributed to nonpolar bonds between BN and also absence of 10 order that prevents dipole alignment even at high-frequencies. The values for a-BN compare 11 extremely favourably to other reports in the literature, as shown in Extended Data Table 1. We 12 have confirmed the electrical measurements of values with those obtained by measuring the 13 refractive index of a-BN with spectroscopic ellipsometry and using the relationship: n2 = .22 14 The refractive indices of h-BN and a-BN at 633 nm wavelength were found to be 2.16 and 1.37, 15 respectively, as indicated by the green stars in Fig. 3(b). Thus, -values for h-BN and a-BN 16 from ellipsometry are 4.67 and 1.88, respectively – closely matching the values obtained with 17 electrical measurements at 100kHz. Low- dielectric materials are sometimes made porous to 18 exploit the low  value of air but this decreases the density of the material, which in turn results 19 in poor mechanical strength. It can be seen from Fig. 3c that a-BN possesses the lowest 20 dielectric constant at the highest density in comparison with well-known low  materials 21 reported in the literature [Extended Data Table 1]. We have also measured the mechanical 22 properties of a-BN films to confirm their strength. The results of nanoindentation 23 measurements shown in Extended Data Fig. 9 indicates that the a-BN films possess hardness 7 1 and stiffness values that are equal to or greater than those of silicon (>11 GPa). The nanoscratch 2 test results also shown in Extended Data Fig. 9 suggest that the films are very well adhered to 3 the substrates. 4 The electrical breakdown strength of a-BN was extracted by measuring the current density 5 with applied bias (Fig. 3d) on vertical sandwich type devices. The data in Fig. 3d reveal that 6 there is a slight increase in current density due to Poole–Frenkel (P–F) tunnelling at low 7 voltages and above 2.2 V, the leakage current sharply increases leading to electrical breakdown. 8 As the thickness of a-BN is 3 nm, the breakdown field is extracted to be 7.3 MV-cm-1 – this is 9 nearly twice that of h-BN (see Table I) and the highest reported for materials with dielectric 10 constants of less than 2 as shown in Fig. 3e. The a-BN film also exhibits exceptionally low 11 leakage current density of 6.27 µA/cm2 at 0.3 V – thus, demonstrating its potential for 3 nm 12 node devices. The key dielectric properties of a-BN and h-BN are summarised in Table 1. 13 Table 1. Electrical characteristics of a-BN and h-BN Electrical properties Film properties Dielectric constant Breakdown Field Reflective index (n) Density @100 kHz / @1 MHz (MV-cm-1) @ 633 nm (g-cm-3) h-BN 3.28 / 2.87 4.0 2.16 2.1 a-BN 1.78 / 1.16 7.3 1.37 2.1~2.3 14 15 A key step in back end of line (BEOL) CMOS fabrication of logic and memory devices is 16 the deposition of a diffusion barrier between the low- dielectric material and the metal wire 17 interconnects to prevent metal atom migration into the insulator. Ideally, this step can be 18 eliminated if the low- dielectric material can also serve as the diffusion barrier. We have 19 therefore tested the diffusion barrier properties of a-BN by depositing 80 nm of cobalt film on 8 1 a-BN and annealing the Co/a-BN/Si devices in vacuum for 1 h at 600 °C. This annealing 2 condition is extremely harsh and under similar conditions severe diffusion of cobalt in Si occurs 3 when industry standard TiN is used as the barrier layer (Extended Data Fig. 10). In contrast, no 4 diffusion of Co or silicide formation was observed with a-BN in cross-sectional TEM results 5 shown in Fig. 3f (additional data in the form of energy dispersive spectroscopy composition 6 maps are shown in Extended Data Fig. 11) – suggesting that a-BN can serve as both the low- 7 dielectric and the diffusion barrier. The comparison of breakdown bias in Co/a-BN/Si and 8 Co/TiN/Si devices measured at various temperatures suggest that films are stable at high 9 temperatures (Extended Data Fig. 12). Our results suggest that a-BN is an excellent low- 10 material for high performance CMOS electronics. 11 12 Fig. 3 | Dielectric properties of a-BN. (a) Dielectric constant determined using capacitance– 13 frequency measurements on metal–insulator–metal (MIM) structures (thick blue and red lines 14 denote averages; inset illustrates optical image of MIM structure); (b) Distribution of dielectric 9 1 constant values at 100 kHz and refractive indices (green stars) calculated via ellipsometry 2 measurements; (c) Density versus dielectric constant of low- materials reported in literature 3 (blue circles) with red circle denoting a-BN reported in this study; (d) Typical current–voltage 4 (J−V) curves for h-BN (approximately 1.2 nm thick; blue curve) and a-BN (3 nm thick; red 5 curve) films; (e) Breakdown field versus dielectric constant for low- materials reported in 6 literature (blue circles) with red circle denoting a-BN; (f) Cross-sectional TEM images of a- 7 BN after thermal-diffusion test performed for 1 h at 600 °C. The bottom image shows 8 magnified view of red box marked in upper image. 9 10 1 References 2 1 Shamiryan, D., Abell, T., Iacopi, F. & Maex, K. Low-k dielectric materials. Mater Today 7, 34- 3 39, (2004). 4 2 Moore's deviation. Nature Nanotechnology 12, 1105, (2017). 5 3 Akinwande, D. et al. Graphene and two-dimensional materials for silicon technology. Nature 6 573, 507-518, (2019). 7 4 The International Roadmap for Devices and Systems: 2017, 4.5. Interconnect Technology 8 Requirements. (2017). 9 5 King, S. W. Dielectric Barrier, Etch Stop, and Metal Capping Materials for State of the Art 10 and beyond Metal Interconnects. Ecs J Solid State Sc 4, N3029-N3047, (2015). 11 6 del Alamo, J. A. Nanometre-scale electronics with III–V compound semiconductors. Nature 12 479, 317-323, (2011). 13 7 Venema, L. Silicon electronics and beyond. Nature 479, 309-309, (2011). 14 8 Franklin, A. D. Nanomaterials in transistors: From high-performance to thin-film applications. 15 Science 349, aab2750, (2015). 16 9 Koenderink, A. F., Alù, A. & Polman, A. Nanophotonics: Shrinking light-based technology. 17 Science 348, 516-521, (2015). 18 10 Liu, C. et al. Small footprint transistor architecture for photoswitching logic and in situ 19 memory. Nature Nanotechnology 14, 662-667, (2019). 20 11 Xiang, D., Liu, T. & Chen, W. Fused computing and storage in a 2D transistor. Nature 21 Nanotechnology 14, 642-643, (2019). 22 12 Grill, A. PECVD low and ultralow dielectric constant materials: From invention and research 23 to products. Journal of Vacuum Science & Technology B 34, 020801, (2016). 24 13 More Moore Roadmap - The International Technology Roadmap for Semiconductors 2.0 25 White Paper: 2015, 7. Interconnect Scaling, 15-16 page., (2015). 26 14 Joo, W. J. et al. Realization of continuous Zachariasen carbon monolayer. Sci Adv 3, e1601821, 27 (2017). 28 15 Liu, J. et al. Plasma deposition of low dielectric constant (k=2.2∼2.4) Boron Nitride on 29 methylsilsesquioxane-based nanoporous films. Journal of Applied Physics 96, 6679-6684, 30 (2004). 31 16 Glavin, N. R. et al. Amorphous Boron Nitride: A Universal, Ultrathin Dielectric For 2D 32 Nanoelectronics. Adv Funct Mater 26, 2640-2647, (2016). 33 17 Kim, G. et al. Growth of high-crystalline, single-layer hexagonal boron nitride on recyclable 34 platinum foil. Nano Lett 13, 1834-1839, (2013). 35 18 Jang, A. R. et al. Wafer-Scale and Wrinkle-Free Epitaxial Growth of Single-Orientated 36 Multilayer Hexagonal Boron Nitride on Sapphire. Nano Lett 16, 3360-3366, (2016). 37 19 Plass, M. F., Fukarek, W., Mändl, S. & Möller, W. Phase identification of boron nitride thin 11 1 films by polarized infrared reflection spectroscopy. Appl Phys Lett 69, 46-48, (1996). 2 20 Kim, D. Y. et al. Role of hydrogen carrier gas on the growth of few layer hexagonal boron 3 nitrides by metal-organic chemical vapor deposition. AIP Advances 7, 045116, (2017). 4 21 Jiménez, I. et al. Core-level photoabsorption study of defects and metastable bonding 5 configurations in boron nitride. Physical Review B 55, 12025-12037, (1997). 6 22 Dharma-wardana, M. W. C. Relation of the refractive index to the dielectric constant 7 containing Doppler-like spatial dispersion. Journal of Physics A: Mathematical and General 8 9, L93-L97, (1976). 9 23 Ling, X. & Zhang, J. Interference Phenomenon in Graphene-Enhanced Raman Scattering. The 10 Journal of Physical Chemistry C 115, 2835-2840, (2011). 11 24 Plimpton, S., Crozier, P. & Thompson, A. LAMMPS-large-scale atomic/molecular massively 12 parallel simulator. Sandia National Laboratories 18, 43, (2007). 13 25 Los, J. et al. Extended Tersoff potential for boron nitride: Energetics and elastic properties of 14 pristine and defective h-BN. Phys Rev B 96, 184108, (2017). 15 26 Tersoff, J. Empirical interatomic potential for silicon with improved elastic properties. Phys 16 Rev B 38, 9902, (1988). 17 27 Matsunaga, K. & Iwamoto, Y. Molecular dynamics study of atomic structure and diffusion 18 behavior in amorphous silicon nitride containing boron. J Am Ceram Soc 84, 2213-2219, 19 (2001). 20 28 Al-Ghalith, J., Dasmahapatra, A., Kroll, P., Meletis, E. & Dumitricǎ, T. Compositional and 21 Structural Atomistic Study of Amorphous Si–B–N Networks of Interest for High-Performance 22 Coatings. The Journal of Physical Chemistry C 120, 24346-24353, (2016). 23 29 Kimura, K., Joumori, S., Oota, Y., Nakajima, K. & Suzuki, M. High-resolution RBS: a powerful 24 tool for atomic level characterization. Nuclear Instruments and Methods in Physics Research 25 Section B: Beam Interactions with Materials and Atoms 219-220, 351-357, (2004). 26 30 Kitahara, A., Yasuno, S. & Fujikawa, K. Study of thin-film thickness and density by high- 27 resolution Rutherford backscattering spectrometry and X-ray reflectivity. Transactions of the 28 Materials Research Society of Japan 34, 613-615, (2009). 29 31 Hatton, B. D. et al. Materials chemistry for low-k materials. Materials Today 9, 22-31, (2006). 30 32 Cheng, Y.-L. & Lee, C.-Y. in Nanoporous Materials Porous Low-Dielectric-Constant Material 31 for Semiconductor Microelectronics (2018). 32 33 Zhao, L. et al. Role of copper in time dependent dielectric breakdown of porous organo- 33 silicate glass low-k materials. Appl Phys Lett 99, (2011). 34 34 Volinsky, A. A., Palacio, M. L. B., & Gerberich, W. W. in Materials Research Society Symposium 35 - Proceedings "Incompressible" pore effect on the mechanical behavior of low-K dielectric 36 films. 567-572. 37 35 Das, A. et al. Characterisation and integration feasibility of JSR's low-k dielectric LKD-5109. 12 1 Microelectron Eng 64, 25-33, (2002). 2 36 Chang, S. Y. et al. Mechanical property analyses of porous low-dielectric-constant films for 3 stability evaluation of multilevel-interconnect structures. Thin Solid Films 460, 167-174, 4 (2004). 5 37 Li, H. Y. et al. Process improvement of 0.13 mu m Cu/Low K (Black Diamond (TM)) dual 6 damascene interconnection. Microelectron Reliab 45, 1134-1143, (2005). 7 38 Pang. B., Y. W., Lee. P., Naik. M. in Semiconductor Fabtech 285-289 (London : ICG 8 Publishing, 1994-, 1999). 9 39 Nagendra, V. in Nanoindentation in Materials Science Mechanical Characterization of Black 10 Diamond (Low-k) Structures for 3D Integrated Circuit and Packaging Applications Ch. 10, 11 (2012). 12 40 Tyberg, C. et al. in Polymers for Microelectronics and Nanoelectronics Vol. 874 ACS 13 Symposium Series Ch. 12, 161-172 (American Chemical Society, 2004). 14 41 Im, J., Townsend, P. H., Curphy, J., Karas, C. & Shaffer, E. O. in Metallization of Polymers 2 15 (ed Edward Sacher) 53-60 (Springer US, 2002). 16 42 Grill, A. Plasma enhanced chemical vapor deposited SiCOH dielectrics: from low-k to extreme 17 low-k interconnect materials. Journal of Applied Physics 93, 1785-1790, (2003). 18 43 Brandrup, J., Immergut, E. H. & Grulke, E. A. Polymer handbook. 4th edn, (Wiley- 19 Interscience, 2004). 20 44 Lee, H. J. et al. Structural characterization of porous low-k thin films prepared by different 21 techniques using x-ray porosimetry. Journal of Applied Physics 95, 2355-2359, (2004). 22 45 Laturia, A., Van de Put, M. L. & Vandenberghe, W. G. Dielectric properties of hexagonal 23 boron nitride and transition metal dichalcogenides: from monolayer to bulk. npj 2D 24 Materials and Applications 2, 6, (2018). 25 13 1 Acknowledgements We thank Y. K. Kim from UNIST Central Research Facilities (UCRF) for 2 cross sectional HR-TEM images. This work was supported by Samsung Electronics (Samsung- 3 SKKU Graphene/2D Center), the research fund (NRF-2017R1E1A1A01074493 and NRF- 4 2019R1A4A1027934) and the grant (CASE-2013M3A6A5073173) from the centre for 5 Advanced Soft Electronics under the Global Frontier Research Program through the National 6 Research Foundation by the Ministry of Science and ICT, Korea. The NEXAFS experiments 7 were performed at 4D, 6D and 10A2 beamline of the Pohang Accelerator Laboratory (PAL) 8 were supported in part by the Ministry of Science and ICT, POSTECH, and UNIST. 9 Author contributions H.S.S. and H.J.S. planned and supervised this project; S.H., K.Y.M., 10 G.K, S.I.Y and H.S.S. performed experiments for growth and characterization.; C.-S.L., M.- 11 H.L. and H.-J.S. fabricated electrical devices; S.W.K measured ellipsometry and analyzed; H.L. 12 measured RBS and analyzed; Y.L. and Z.L. measured TEM data. K.I, K.K. and T.J.S. measured 13 NEXAFS data; E.-C.J, H, H.J., and J.-Y.K measured mechanical property and adhesion; A.A. 14 and S.R. performed MD simulations; MC advised on the project, wrote and edited the 15 manuscript with H.S.S. All authors contributed to writing the manuscript and agreed on the 16 content of the paper. 17 Competing interests The authors declare no competing interests. 18 Additional information 19 Supplementary information is available for this paper at www.nature.com/nature. 20 Correspondence and requests for materials should be addressed to H.S.S. or H.J.S. 21 Reprints and permissions information is available at www.nature.com/reprints. 22 23 Methods 24 Si-substrate cleaning: 25 Si substrates were ultrasonicated in acetone for 10 min and subsequently cleaned using iso- 26 propenyl alcohol (IPA) and N2 gas. Upon completion of the organic-solvent cleaning process 27 that lasted 5 min, the Si substrates were subjected to O2 plasma treatment to not only remove 28 any carbon impurities remaining on the surface but also to make the surface hydrophilic and 14 1 enhance its wettability. Subsequently, the substrates were immersed in 10% dilute HF solution 2 for 10 min to remove native oxide. Finally, anhydrous ethanol and N2 gas were used to remove 3 residual HF solution from the surface. 4 Growth of a-BN: 5 The clean Si substrates were placed inside the inductively coupled plasma chemical vapour 6 deposition (ICP-CVD, Extended Data Fig. 1) system at centre of the furnace. Borazine 7 (purchased from Gelest) precursor flask was placed in a water bath at -15 °C. The temperature 8 during use was ramped up to 25 °C. For uniform growth, the substrate was tilted by ~ 30° using 9 a Cu support. Prior to ramping up the furnace temperature, pressure inside the CVD system 10 was reduced to its base value of 1 × 10-4 torr and 20 sccm of H2 gas was introduced. 11 Subsequently, the furnace temperature was increased at the rate of 10 °C/min to a set target 12 value (400 °C for a-BN), which was maintained for 20 min before starting the deposition. 13 During the growth period, plasma generation was performed at 30-W power by activating the 14 ICP (inductively coupled plasma) unit under a flow of borazine gas at 0.05 standard cubic 15 centimetre per minute (sccm, controlled by mass flow controller). Growth was conducted for 16 90 min. At the end of the deposition, the borazine flow and plasma generation were terminated, 17 and the furnace was cooled to room temperature using 20 sccm of H2 gas. 18 Transfer technics of samples: 19 To transfer a-BN films, hydrofluoric acid transfer technique described in Ref 18 employed here. 20 Characterisation: 21 Scanning electron microscopy (SEM; Verios 460; FEI) and atomic force microscopy (AFM; 22 Dimension Icon; Bruker) were used to determine sample surface morphologies. XPS (K-Alpha; 23 Thermo Fisher) observations were performed to determine chemical composition. Raman 24 spectra were measured using a micro Raman spectrometer (alpha 300; WITec GmBH) 25 equipped with a 532 nm laser. To obtain accurate Raman spectra, samples were transferred on 26 SiO2 (300 nm)/Si substrates to amplify the Raman signal by multiple reflections.23 FT–IR 27 spectra were acquired using a Varian FT-IR 670 spectrometer equipped with a Seagull variable 28 angle reflection accessory. Incident light was polarised using a wire-grid polariser. NEXAFS 29 analysis was performed using the 4D PES beamline at the Pohang accelerator laboratory in 15 1 Korea. During NEXAFS analysis, samples were attached to a molybdenum holder and loaded 2 into a vacuum chamber. The analysis chamber, maintained at a base pressure of 5 × 10-10 Torr, 3 was equipped with an electron analyser (R3000; Scienta) and X-ray absorption spectroscopy 4 detector having a retarding filter for facilitating operation in the partial electron-yield mode. 5 For high-resolution imaging and selected area electron diffraction (SAED), low-voltage Cs 6 aberration-corrected transmission electron microscope (Titan Cube G2 60-300; FEI) was 7 performed at 80 kV using a monochromatic electron beam. To facilitate sample observation 8 using TEM, a-BN was transferred on to SiN TEM grids (hole diameter: 1 µm). High-resolution 9 cross-sectional transmission electron microscopy (JEM-2100F; JEOL) was performed to 10 confirm the barrier performance. 11 Molecular dynamic simulations and computations: 12 We modelled the Si substrate using a 6-layer diamond cubic rectangular slab whose free surface 13 is perpendicular to the z-axis for BN nucleation and growth. The slab was periodic in the xy- 14 plane with 18x18 repetition of the unit cell, with a total of 15552 Si atoms. The top five layers 15 were completely unrestrained during the simulations, while the bottom layer was kept fixed. 16 The system contains 8211 atoms of Boron and Nitrogen in a 1:1 ratio with an additional amount 17 of 400 H atoms (~5%) to be consistent with the experimental observations. All the simulations 18 were performed using LAMMPS.24 Throughout the simulation the temperature of the substrate 19 is held constant using a Nosé-Hoover thermostat in an NVT ensemble at T=673 K. A wall with 20 a harmonic repulsive potential is placed 25 Å above the substrate. The film is grown using the 21 following method: all the atoms (Boron, Nitrogen and Hydrogen) are initialized with random 22 velocities in a region of 20 Å above the substrate. They are constantly thermalized at the growth 23 temperature and allowed to settle and cool on the Si substrate. To prevent premature B-N bond 24 formation, the minimal distance between the initial B and N sources was set to be 1.90 Å, larger 25 than the B-N bond length of 1.44 Å in the h-BN lattice. The equation of motion was numerically 26 solved using the velocity-Verlet integration scheme. Each simulation was run for more than 15 27 ns at a time step of 0.25 fs. After the growth process, the systems were further relaxed in a NPT 28 ensemble at T= 300 K. The extended Tersoff potential for BN was employed to describe the 29 chemical processes (e.g. bond formation and dissociation) among the atomic species 30 involved.25 This model potential has been specifically designed to correctly describe the 31 dependence on coordination and chemical environment of the bonding occurring in B, N and 16 1 B-N systems. Thanks to its versatility, it allows to describe large scale atomistic simulations 2 whose number of atoms exceeds a few thousands. To describe the interaction within the Silicon 3 substrate we used the Tersoff model potential which has been proved to faithfully reproduce 4 both mechanical and morphological properties of silicon-based systems.26 We treated the Si-N 5 and Si-B interactions via the Tersoff potential parameterized,27 which has been already 6 employed to study the compositional and structural features of Si-B-N networks.28 Finally, we 7 modelled all the interactions involving Hydrogen through a Lennard-Jones potential. 8 Ellipsometry: 9 An automated angle M-2000F rotating compensator ellipsometer equipped with an X–Y 10 mapping stage, focusing probes, and accompanying software—Complete-EASE 6.39 from J.A. 11 Woollam Co., Inc. was used in this study. Ellipsometric data were acquired in the wavelength 12 range 250–1000 nm wavelength range with a 1.6 nm resolution at incidence angles of 65°, 70°, 13 and 75°. Optical properties of both films were determined using the Kramers–Kronig consistent 14 dispersion model involving use of 3 Lorentz oscillators. 15 High-resolution Rutherford backscattering and high-resolution elastic recoil detection 16 analysis: 17 To investigate element compositions of the two ultra-thin films considered in this study, high- 18 resolution Rutherford backscattering spectrometry (HR-RBS) was employed.29 The said HR- 19 RBS analysis was performed by irradiating samples using a 450-keV He+ beam generated by 20 HRBS-V5000 (KOVELCO). Employing the same system, high-resolution elastic recoil 21 detection analysis (HR-ERDA) was simultaneously performed for hydrogen using 500-keV 22 N+ ions. Typical beam-current values during HR-RBS and HR-ERDA analyses equalled 40 nA 23 and 6 nA, respectively. 24 Density measurement: 25 Peaks corresponding to elements—B, N, O and Si—were observed individually in the HR- 26 RBS spectra. Areas covered by the peaks reflect both the thickness and density 27 simultaneously. The areal density (atoms/cm2) was measured,30 which in turn, facilitated 28 calculation of the a-BN film density by considering their thicknesses. Note that the presence of 29 oxygen was observed owing to surface contamination. 17 1 Breakdown-voltage and dielectric-constant measurement: 2 J–V and C–f (capacitance–frequency) characteristics of the films with metal/a-BN/n-Si stacks 3 were measured using a Tektronics K4200A-SCS parameter analyser system and Karl Suss PA- 4 200DS semi-auto probe station. a-BN-based capacitors were fabricated on directly-deposited 5 or transferred BN films on n-Si substrates. To prevent polymer contamination, a shadow mask 6 with 200-μm diameter pattern was used, and a 100 nm Cu electrode was deposited over the a- 7 BN–Si stack. Post device fabrication, capacitance–voltage units (CVU) within K4200A-SCS 8 were first employed to facilitate C–f measurement. We carried out capacitance-frequency 9 measurements in a frequency range of 1 kHz to 10 MHz with a 0.5 V hold bias and ± 30 mV 10 AC drive. Measured capacitance values do not demonstrate a significant change as a function 11 of the applied voltage of 0.5 V. Therefore, the relative dielectric constant was evaluated using 12 the relation 𝜅 = 𝐶𝑡⁄𝐴𝜀0 , where C denotes capacitance, t denotes a-BN-film thickness, A 13 denotes area, and 𝜀0 denotes the dielectric constant. At high frequencies exceeding 5 MHz, 14 significant noise levels were observed in the capacitance owing probably to low impedance of 15 the a-BN capacitor. Subsequently, J–V characteristics of both film samples were determined 16 using source-measurement units (SMUs) within K4200A-SCS. The applied-voltage value was 17 swept from 0 to 10 with 1 pA resolution and 10-mA compliance current. Additionally, 18 measurements were in 50 mV current steps and 10 NPLC (number of power line cycles) to 19 prevent degradation due to bias stresses. 20 Diffusion-barrier performance estimation: 21 To evaluate the diffusion-barrier performance of films considered in this study, approximately 22 3 nm thick samples of a-BN and TiN (RF sputtering) were deposited on a Si substrate. 23 Subsequently, the samples were coated with an 80 nm Co layer using DC sputtering. Post 24 deposition, samples were placed inside a furnace for annealing. The furnace temperature was 25 ramped up at a high rate of 40 °C/min. During annealing, thermally activated diffusion was 26 observed to occur at the interface between Co and dielectric barrier materials. 27 Data availability 28 The datasets generated and/or analysed during the current study are available from the 29 corresponding author on reasonable request. 30 18 1 2 3 Extended Data Fig. 1 | (a) ICP-CVD system with borazine mass flow controller (MFC) for 4 precise control of borazine flow. The a-BN films were grown on Si substrates at 400 °C. 5 19 1 2 3 Extended Data Fig. 2 | (a-c) Raman, XPS, and EDS mapping image of a film grown on copper 4 foil (30W plasma power, Growth Temp = 300 oC) and transferred onto SiO2 substrates so that 5 Raman could be measured. The Raman spectrum on the aBN film is similar to those of bare 6 amorphous SiO2 – similar to spectra on Si reported in our original paper. (d-f) Raman, XPS, 7 and EDS image of a film grown directly on SiO2 (10W plasma power, Growth Temp = 200 oC). 8 The spectra are largely the same for all substrates. The dielectric properties obtained from 9 spectroscopic ellipsometry reveal no influence of substrate since the structure and composition 10 of the films is similar in all cases. Scale bar is 40 nm. 11 20 1 2 Extended Data Fig. 3 | High angle annular dark field (HAADF)-STEM (left) image overlaid 3 with energy dispersive spectroscopy (EDS) mapping of carbon (red), nitrogen (green), and 4 silicon (blue). EDS overlapped image (right) for all elements is shown on the right. Scale bars 5 = 20 nm. 6 7 8 21 1 2 Extended Data Fig. 4 | Image of a -BN sample grown on a Si substrate at T= 673K: a) side 3 view, b) top view. Atomic species are shown in different colours: Si (yellow), Blue (N), Pink 4 (B). (c) Mass density profile along the transverse direction (z-direction) in the system of (a) 5 and (b). Different colours denote densities of different chemical species. The simulated density 6 of a-BN is consistent with the experimental result obtained in this study. The black dashed line 7 corresponds to the BN mass density from experimental measurements. 8 9 22 1 2 Extended Data Fig. 5 | FTIR spectra for a-BN showing the absence of B-H and N-H. 3 4 5 23 1 2 Extended Data Fig. 6 | (a) HR-RBS and (b) HR-ERDA spectra for a-BN film within energy 3 ranges of 240–400 keV and 52–68 keV, respectively; (c) Element composition calculated using 4 HR-RBS and HR-ERDA spectra. 5 6 24 1 2 Extended Data Fig. 7 | Partial electron yield near-edge X-ray absorption fine structure (PEY- 3 NEXAFS) spectra for N K-edge of a-BN, demonstrating trivial angle dependency of N K-edge. 4 25 1 2 Extended Data Fig. 8 | Comparison of a-BN and nanocrystalline BN (nc-BN) films. (a) 3 Schematic of the nc-BN film structure deposited at 700 oC. (b) Low magnification TEM images 4 of nc-BN. (Selected area electron diffraction pattern in the inset shows a typical polycrystalline 5 ring pattern) (c) High resolution TEM images of nc-BN clearly shows small crystallites of h- 6 BN. (Cross sectional TEM image in the inset indicates a layered structure) (d) Magnified image 7 of (c) blue box area and (e) FFT image show hexagonal superstructure of multilayer h-BN. (f) 8 XPS profiles for B 1s and N 1s peaks observed in 3-nm-thick a-BN and nc-BN samples. (g) 9 Raman spectra for a-BN, nc-BN, and epitaxially grown tri-layer h-BN (used as reference and 10 measuring 1.2-nm thick) samples transferred onto SiO2/Si substrates. (h) FT–IR spectra for a- 11 BN (red line) and nc-BN (blue line) measured using s-polarised radiation at an incident angle 12 of 60°. (E1u LO mode related with amorphous of BN, see ref. 19) 26 1 2 Extended Data Fig. 9 | (a-b) Nanoindentation results show that deposition aBN on Si 3 substrates leads to enhancement in surface hardness and stiffness. (c) Nanoscratch test results 4 reveal that scratch depth of 40 nm (>10 times the film thickness) is required to delaminate the 5 film – suggesting excellent adhesion with Si substrate. SEM observations show that the scratch 6 regions are clean and no evidence for delamination of aBN can be found for scratch depth 7 shallower than 40 nm. 27 1 2 Extended Data Fig. 10 | Cross-sectional TEM images of Co (80 nm)/TiN (3 nm)/Si films after 3 thermal diffusion tests at different temperatures: (a) 600 °C for 60 min, (b) 600 °C for 30 min, 4 (c) 400 °C for 30 min, and (d) enlarged cross-sectional TEM image and EDS line profile in (a). 5 28 1 2 Extended Data Fig. 11 | (a) Large-area cross-sectional TEM image and EDS line profiles of 3 Co/a-BN(3 nm)/Si sample after annealing at 600 oC for 60 min. (b) EDS maps of Co and Si 4 showing that Co is isolated above the aBN film and does not diffuse into the Si. (c) EDS maps 5 of a magnified area in (b). 6 29 1 2 Extended Data Fig. 12 | Comparison of breakdown bias at different temperatures for a-BN 3 and TiN barriers. 4 30 1 Extended Data Table 1 | Comparison of dielectric constants of various dielectric materials Breakdown Dielectric Density Modulus Hardness field constant (g/cm3) (GPa) (GPa) (MV/cm) Ref. SiO2 4 2.2 55~70 3.5 >10 31,32 FSG Ref. 3.5~3.8 2.2 ~50 3.36 >10 31,32 (Fluorinated silicon glass) OSG (organosilicate glass or 3.0 Ref. 2.8~3.2 1.8~1.2 6.6~8.4 1.2~1.7 carbon-doped silicon Ref. 33 31,32,34 glass) HSQ ~3.0 3.3 MSQ ~2.5 2.7~12.5 0.19~0.936 Ref. 36 Ref. 35 4.75 Black Diamond (SiCOH) 2.7~3.3 <2.26 10~20 1.5~3.0 Ref.38,39 Ref. 37 SiwCxOyHz 2.7~3.0 9~15 1.3~2.4 6~10 Ref. 40 SiLK 2.65 2.7 0.38 4 Ref.40,41 SiCOH 2.8 1.32 16.2 1.69 Ref. 42 pSiCOH 2.4 1.06 4.2 0.28 Ref. 42 (pore <1.5 nm) pSiCOH 2.05 0.87 3.3 0.28 Ref. 42 (pore <2.5 nm) a-CHpolymeric 2.2~2.3 0.92~0.94 Ref.34,43 Porous HSQ (hydrogen 2.2 0.98 Ref. 44 silsesquioxane) (porosity 46%) Porous MSQ (methylsilsesquioxane) 1.85 0.89 Ref. 44 (porosity 34%) BCN 3.7~4.6 Ref. 5 (boron carbon nitride) h-BN 3.29~3.76 2.1 19.5~100 0.6~3 Ref. 45 a-BN or 2.2~2.4 Ref.15 amorphous h-BN 5.9 Ref.16 This a-BN 1.16~1.78 2.1~2.3 7.3 work 2 31