Skip to content

Accurate estimation of thermal properties for AlGaN/GaN HEMTs on diamond

MetadataDetails
Publication Date2022-07-01
JournalAIP Advances
AuthorsYao Li, Yulong Chen, Chao Zhang, Xi Wang, Hongbin Pu
InstitutionsXi’an University of Technology, Xidian University
Citations5
AnalysisFull AI Review Included

Accurate Estimation of Thermal Properties for AlGaN/GaN HEMTs on Diamond: Implications for Advanced Thermal Management

Section titled “Accurate Estimation of Thermal Properties for AlGaN/GaN HEMTs on Diamond: Implications for Advanced Thermal Management”
  • Application Focus: Thermal modeling of AlGaN/GaN High Electron Mobility Transistors (HEMTs) integrated with polycrystalline diamond (PCD) heat spreaders for RF and high-power electronics.
  • Core Problem Addressed: Inaccurate channel temperature prediction due to oversimplified thermal conductivity assumptions in GaN-on-diamond stacks.
  • Methodology: Finite Element Method (FEM) simulation comparing four thermal conductivity models across varying Thermal Boundary Resistance (TBR), GaN thickness, gate pitch, heat source width, and power dissipation.
  • Key Finding 1 (Material Anisotropy): PCD thermal conductivity is depth-dependent and anisotropic (cross-plane vs. in-plane) due to columnar grain morphology, ranging from 454 W/m·K (near nucleation) to 2569 W/m·K (bulk).
  • Key Finding 2 (GaN Scaling): GaN thermal conductivity is thickness-dependent, scaling from 74 W/m·K at 0.5 µm to 214 W/m·K at 5 µm.
  • Engineering Insight: Using constant PCD and constant GaN thermal conductivity yields a closer match to the fully anisotropic model than using mixed anisotropic/constant combinations.
  • Practical Impact: Up to 11.19 K maximum temperature rise errors occur when improperly accounting for anisotropic PCD properties, directly impacting device reliability projections.
ParameterValueUnitContext
Device Length (L)670µmTotal chip dimension
Device Width (W)530µmTotal chip dimension
Heat Source Length (Lg)0.5µmGate footprint
Heat Source Width (Wg)240µmGate footprint
Heat Source Thickness (Hg)0.2µmGate footprint
Gate Pitch (S)20µmInter-gate spacing
GaN Buffer Thickness (hbuf)1.4µmActive region thickness
Diamond Substrate Thickness (hsub)100µmPCD heat spreader
Embedding Depth (De)0.2µmHeat source depth below surface
Thermal Boundary Resistance (TBR)6.5m2·K/GWGaN/Diamond interface
Gate Fingers12countMulti-finger device structure
Au Thermal Conductivity (kAu)317W/m·KMetallization layer
Constant GaN k (kCGaN)160 × (T/300)−1.4W/m·KTemperature-dependent reference
Constant PCD k (kCPCD)2000 × (T/300)−0.55W/m·KTemperature-dependent reference
Max Channel Temperature Error (TBR variation)−7.72 to +11.19KModel deviation from benchmark
Power Dissipation Range1-10W/mmOperating condition sweep
  • Simulation Approach: Three-dimensional Finite Element Method (FEM) thermal simulation with quarter-symmetry approximation.
  • PCD Thermal Conductivity Model: Depth-discretized anisotropic model across eight subsections (0-1, 1-5, 5-15, 15-30, 30-45, 45-60, 60-75, 75-100 µm) accounting for columnar grain evolution.
  • GaN Thermal Conductivity Model: Approximate solution of the phonon Boltzmann transport equation with thickness-dependent cross-plane and in-plane values.
  • Temperature Scaling: PCD conductivity scaled as kPCD × (T/300)−0.55; GaN conductivity scaled as (T/300)−1.4.
  • Comparative Analysis: Four model cases evaluated: (1) Anisotropic GaN + Anisotropic PCD (benchmark), (2) Anisotropic GaN + Constant PCD, (3) Constant GaN + Anisotropic PCD, (4) Constant GaN + Constant PCD.
  • Parameter Sweeps: Channel temperature rise evaluated against TBR (6.5-38 m2·K/GW), GaN thickness (0.5-5 µm), gate pitch (5-50 µm), heat source width (50-300 µm), and power dissipation (1-10 W/mm).

For replicating or extending GaN-on-diamond HEMT thermal management research, High Thermal Conductivity Polycrystalline Diamond (PCD) substrates are essential. The paper’s depth-dependent PCD model (454-2569 W/m·K cross-plane) requires material with verified columnar grain morphology and thickness control. 6CCVD provides:

  • Optical Grade SCD and Heavy Boron Doped PCD for applications requiring controlled thermal properties.
  • PCD substrates up to 10 mm thickness and 0.1 µm-500 µm film thickness to accommodate the 100 µm diamond layer specified in this study.
  • Plates/wafers up to 125 mm for full-wafer GaN-on-diamond integration.

The 100 µm PCD substrate with specific surface roughness requirements for GaN epitaxy is within standard capability. For advanced thermal interface engineering:

  • Custom Metalization: Ti/Pt/Au or Au metallization schemes for gate, drain, and source contacts can be deposited in-house to match device fabrication requirements.
  • Laser Cutting Services: Precision dicing to device dimensions (670 µm × 530 µm chip footprint) with minimal kerf loss.
  • Surface Polishing: Ra < 5 nm on inch-size PCD ensures low thermal boundary resistance (TBR), critical for the 6.5 m2·K/GW target specified in this work.

6CCVD’s in-house PhD engineering team can assist with material selection for GaN-on-diamond thermal management projects. Key consultation areas include:

  • Optimizing PCD thickness for target thermal resistance and mechanical stability.
  • Balancing columnar grain size for desired cross-plane vs. in-plane conductivity ratios.
  • Matching PCD surface properties to minimize GaN nucleation challenges and TBR.
  • Selecting appropriate metalization stacks for high-temperature, high-power device operation.

For custom specifications, material consultation, or technical collaboration on GaN-on-diamond HEMT integration, visit 6ccvd.com or contact our engineering team directly.

View Original Abstract

For accurate estimation of the thermal characteristics of AlGaN/GaN high electron mobility transistors on diamond, the anisotropic thermal conductivities of polycrystalline diamond (PCD) and GaN with their thickness dependences are included in the finite element simulation model, of which the PCD thermal conductivity kPCD is calculated through detailed analyses of the grain sizes in the directions parallel and vertical to the interface and relative phonon-grain boundary scatterings. In addition, the GaN thermal conductivity kGaN is the result of an approximate solution of the phonon Boltzmann transport equation. To investigate the effects of anisotropic GaN and PCD thermal conductivity, four cases of thermal conductivity models are compared: anisotropic kGaN and kPCD, anisotropic kGaN with constant kPCD, constant kGaN with anisotropic kPCD, and constant kGaN and kPCD. Then the maximum channel temperature simulated from the four above-mentioned models and the errors relative to the case of anisotropic kGaN and kPCD are investigated with respect to thermal boundary resistance, GaN layer thickness, gate pitch, heat source width, and power dissipation. The results show that the combination of anisotropic kPCD and constant kGaN overestimates the maximum channel temperature rise, and using constant kPCD and anisotropic kGaN underestimates the maximum channel temperature rise for almost all the simulations in the text. However, the channel temperature stemming from constant kPCD and kGaN is generally closer to the case with anisotropic PCD and GaN thermal conductivity.