Skip to content

Effect of phase and interfacial structure on thermal transport in BN/diamond heterostructures

FieldValue
JournalAPL Materials
Published2026-02-01
AuthorsSamreen Khan, Saurabh Vishwakarma, Avani Patel, Frank Angeles, R. J. Nemanich
DOI10.1063/5.0311083

An interfacial thermal conductance of 300 ± 140/100 MW m⁻² K⁻¹ was measured for c-BN on (111) diamond at 300 K. This value was about three times the 85 ± 8.5 MW m⁻² K⁻¹ measured for t-BN on (001) diamond. The c-BN interface maintained this conductance despite a 1 to 2 nm disordered region. Its film conductivity varied from about 5 to 20 W m⁻¹ K⁻¹ with depth and was roughly 40 times lower than bulk c-BN.

BN grown on (001) boron-doped diamond was columnar turbostratic BN, or t-BN. Its TEM fast Fourier transform showed diffuse, broad rings without sharp spots, consistent with stacking disorder. BN grown on (111) diamond was cubic BN, or c-BN. Sharp diffraction spots and cubic periodicity showed long-range crystalline order, although dense twinning and stacking faults remained. The t-BN film also contained c-BN nanograins with an average diameter of about 7 nm.

At the buried interfaces, the t-BN remained sp2 bonded, while EELS identified the c-BN interface as sp3 bonded. The c-BN interface contained a 1 to 2 nm region with more disorder than the film interior. The authors attributed this region to initial BN nucleation during PECVD. This structural contrast formed the basis for their comparison of interface conductance.

Both substrates were single-crystal diamond doped with boron at 10²⁰ cm⁻³. Hydrogen plasma cleaning proceeded at 760 °C and 9 × 10⁻⁵ Torr for 90 min. ECR-PECVD then used fluorine chemistry, with hydrogen-to-fluorine flow ratios controlling the BN phase.

The t-BN film on (001) diamond used H2 at 4 SCCM and BF3 at 4 SCCM, giving an H/F ratio of 1. The c-BN film on (111) diamond began with H2 at 1.5 SCCM and BF3 at 2 SCCM, giving an H/F ratio of 0.75. Later flows changed to H2 at 2 SCCM and BF3 at 2 SCCM, giving an H/F ratio of 1.

Fluorine preferentially etched sp2 BN over sp3 BN. High hydrogen flow reduced etching efficiency and favored sp2 phases, while limited hydrogen promoted sp3 c-BN growth. TEM gave film thicknesses of about 150 to 160 nm. Hydrogen absorption during plasma cleaning also darkened the diamond near the interface.

Each sample carried two transducer regions. TDTR used an approximately 80 nm Al layer over a 2 nm Ti layer. TR-MOKE used an approximately 8 nm Ti/Co/Pt multilayer with four Co/Pt bilayers and a Pt cap. Their thermal response times were about 1000 ps and 100 ps, respectively. The shorter TR-MOKE response confined heat closer to the surface and improved sensitivity to buried interfaces.

The TDTR pump was modulated at 10.7 MHz. Its in-phase and out-of-phase signals were fitted with a 3D heat-diffusion model for the multilayer structure. TR-MOKE used a model that included optical absorption throughout the stack because its semitransparent transducer absorbed pump energy within the diamond.

Diamond thermal conductivity was calibrated independently at 500 W m⁻¹ K⁻¹. Measurements at 6.5, 4, 2.5, and 1.25 MHz gave an average of about 540 W m⁻¹ K⁻¹ for the (111) boron-doped substrate.

The Al/Ti-to-c-BN interface conductance was 300 MW m⁻² K⁻¹. The corresponding conductance for the t-BN film was about 180 MW m⁻² K⁻¹. These values describe the metal transducer interfaces rather than the buried BN/diamond interfaces.

TDTR could not separately resolve BN conductivity and buried-interface resistance. The fitted lumped resistance was 24 m² K GW⁻¹ for t-BN and its interface, compared with 11 m² K GW⁻¹ for c-BN and its interface. Assuming zero BN/diamond interface resistance, these values corresponded to effective conductivities of 7 and 13 W m⁻¹ K⁻¹.

The best t-BN fit gave a Ti/t-BN conductance of 400 ± 150/80 MW m⁻² K⁻¹. It gave a t-BN conductivity of 17 ± 2.5 W m⁻¹ K⁻¹ and a t-BN/diamond conductance of 85 ± 8.5 MW m⁻² K⁻¹.

No constant-conductivity model fitted the c-BN data well. A two-layer model used a 40 nm surface region with a conductivity of about 5 W m⁻¹ K⁻¹. A deeper 110 nm region had a conductivity of about 19 W m⁻¹ K⁻¹. This model produced a 3.7% root mean square error.

Contour analysis gave a deeper-region conductivity of 17 to 30 W m⁻¹ K⁻¹ and a c-BN/diamond conductance of 200 to 440 MW m⁻² K⁻¹. Their central values were 20 ± 10/3 W m⁻¹ K⁻¹ and 300 ± 140/100 MW m⁻² K⁻¹. The deeper layer and buried interface had a combined resistance of about 8.4 m² K GW⁻¹. TR-MOKE could not determine these two properties independently because they produced similar signal decays.

The measured c-BN conductivity of about 20 W m⁻¹ K⁻¹ was far below the 900 W m⁻¹ K⁻¹ bulk value used in the analysis. The authors described this as roughly a 40-fold reduction.

Bulk c-BN carries heat through phonons from about 2 to 12 THz. Their mean free paths range from 2 µm at low frequency to about 50 nm at high frequency. The film results indicate mean free paths between 1 and 50 nm. Point defects from fluorine incorporation could scatter phonons above 5 THz. However, those phonons carry less than 100 W m⁻¹ K⁻¹ in bulk c-BN, so point-defect scattering alone cannot explain the measured reduction.

Extended defects appear more important. The authors calculated conductivities of 40 and 180 W m⁻¹ K⁻¹ for mean free paths limited to 10 and 50 nm. An extended-defect spacing of 10 to 50 nm, combined with point defects near 10²⁰ cm⁻³, could produce the observed conductivity near 20 W m⁻¹ K⁻¹. Twinning, stacking faults, strain, and small sp2-bonded regions remain possible contributors. Raman spectra from the same films also indicated a high defect density.

The t-BN conductivity fell within the previously reported 8 to 30 W m⁻¹ K⁻¹ range for turbostratic BN with varying disorder. It was about four times the 2 to 5 W m⁻¹ K⁻¹ through-plane conductivity of h-BN. The authors related this result to the vertical orientation of the t-BN basal planes.

Cubic BN remains of interest for high-power and high-frequency devices because its bandgap is 6.4 eV. The measured film properties show that bulk c-BN conductivity does not describe a defective PECVD film. Device thermal analysis should use film conductivity and buried-interface conductance measured under the intended growth conditions.

The experiment compared one t-BN film on (001) diamond with one c-BN film on (111) diamond. Phase and substrate orientation therefore changed together, leaving their separate effects unresolved. The proposed defect mechanisms were not measured directly.

The depth-dependent c-BN conductivity also remained coupled to its interface conductance. The authors suggest that changing surface chemistry during growth may reduce this problem. They also note that sample bias can weaken as the film thickens because the doped diamond screens the bias. Thermal conductance measurements may therefore help guide c-BN process development.

The study needs single-crystal, boron-doped diamond substrates for ECR-PECVD growth and thermal transport measurements. The diamond materials catalog from 6CCVD lists single-crystal and polycrystalline diamond grown by MPCVD, boron-doped diamond, and metallization with Ti or Pt. Custom fabrication requirements for orientation-specific substrates and compatible metal layers can be discussed through engineering support.

View Original Abstract

Cubic BN (c-BN) is attractive for high-power and extreme-environment electronics. However, thin-film synthesis is challenging due to its propensity to form mixed-phase microstructures of sp2-and sp3-bonded polytypes. We combine time-domain thermoreflectance and time-resolved magneto-optic Kerr effect measurements to determine the effect phase composition, bond disorder, and interfacial structure have on thermal transport of BN thin-films grown on boron-doped diamond by plasma-enhanced chemical vapor deposition. High-resolution transmission electron microscopy reveals BN film grown on (001) diamond is turbostratic (t-BN), while the film grown on (111) diamond is c-BN. At 300 K, the t-BN film has a thermal conductivity of 17 ± 2.5 W m−1 K−1, while the c-BN film has a graded conductivity that varies between 5 and 20 W m−1 K−1 as a function of depth. The c-BN film’s thermal conductivity is ≈ 40× lower than that of bulk c-BN, indicating that defects limit phonon mean-free-paths to ∼10 nm. The c-BN/diamond interface conductance is ≈300 +/− 140/100 MW m−2 K−1, about threefold higher than t-BN/diamond. We credit this factor of three difference to interfacial bonding caused in the sp2 vs sp3 bonded polytypes.

  1. 2018 - Ultrawide-Bandgap semiconductors: Research opportunities and challenges [Crossref]
  2. 2021 - Ultrawide-bandgap semiconductors: An overview [Crossref]
  3. 1974 - Ultraviolet and infrared spectra of cubic boron nitride [Crossref]
  4. 2020 - Ultrahigh thermal conductivity in isotope-enriched cubic boron nitride [Crossref]
  5. 2017 - Recent development of boron nitride towards electronic applications [Crossref]
  6. 2002 - Electrical characteristics of thin film cubic boron nitride [Crossref]
  7. 2023 - Nucleation of cubic boron nitride on boron-doped diamond via plasma enhanced chemical vapor deposition [Crossref]
  8. 2004 - Thick and adherent cubic boron nitride films grown on diamond interlayers by fluorine-assisted chemical vapor deposition [Crossref]
  9. 2000 - The roles of hydrogen and fluorine in the deposition of cubic boron nitride films in the Ar-N2-BF3-H2 system [Crossref]
  10. 2003 - Recent advances in cubic boron nitride deposition [Crossref]