QuantumATK - New Release Notes

Machine-Learned (ML) Force Fields for Realistic Structures and Thermal Properties

• 1000-10,000x shorter computational time vs DFT enable ab initio accuracy for large system sizes and time-scales greatly exceeding those accessible to DFT.
• Use ML Force Fields - Moment Tensor Potentials (MTPs) with molecular dynamics (MD) to:
  • Generate realistic complex structures of novel crystal and amorphous bulk materials, alloys, interfaces, and multilayer stacks
    • Example applications: structure generation of GST materials for PCRAM, high-k metal gate stacks using the Multilayer Builder GUI (watch a video to learn more)
  • Simulate mechanical and thermal properties, e.g., for 2D materials
  • Model surface processes (thermal ALD & ALE)
  • Use in other cases where conventional Force Fields are not available/difficult to develop
• Available MTP library with pre-trained MTPs for a range of bulk materials and interfaces.

Automated Generation of New Machine-Learned Force Fields

• Automatic training tools and GUI templates for crystal and amorphous bulk materials, interfaces and molecules 
• More efficient active-learning based generation of DFT training data during MD by starting from several different initial configurations in parallel 
• Improved MTP training framework, including tools to finding most different training configurations to reduce MTP training datasets

Machine Learning-Based Surface Process Modeling

• Efficiently simulate thermal ALD/ALE processes using specifically trained ML Force Fields, MTPs, with ab initio accuracy
• Obtain parameters for feature scale models to optimize yield
• Pre-trained MTP is provided for HfCl₄ deposition on HfO₂ surfaces (ALD)
• Use a special MTP training protocol to generate ML Force Fields for new processes/materials

Realistic Physics of Complex Materials, Interfaces and Multilayer Stacks

• Hybrid DFT HSE06-DDH method with LCAO basis sets for accurate electronic property simulations of realistic 1000+ atom systems
  
  • Extension to metals and interfaces/stacks containing metals (in addition to semi and insulators)
  • 2x speed-up for 1000+ atom systems and up to 20X speed-up for smaller systems
• 10x more efficient electron-phonon coupling simulations; benefit for mobility simulations of systems with many k- and q-points
• > 100x faster Hamiltonian Derivatives for systems with large unit cells and more accurate and faster Dynamical Matrix simulations 
  • Due to Wigner-Seitz method, enabling accurate simulations with smaller unit cell dimensions
  • Important for electron-phonon coupling, mobility, phonon bandstructure and DOS, Raman, dielectric tensor, and electrooptical tensor

Realistic Nanoelectronic IV Characteristics

• Improved inelastic transport in systems with strong electron-phonon coupling, such as bulk-like devices, using the newly implemented One-Shot Self-Consistent Born Approximation method
• Faster IV calculations and more accurate transport bandgaps with HSE06-DDH-NEGF
• More accurate on-state calculations using Neumann boundary conditions in the transport direction compared to Dirichlet at the DFT level

Multiscale QuantumATK-Sentaurus Device Workflow for 2D FET Engineering

• QuantumATK - Sentaurus Device QTX - Sentaurus Device workflow to investigate the impact of various parameters on the 2D material-based FET performance (Id-Vg, Id-Vd, and C-V characteristics)
  • Different 2D materials and number of layers for channel
  • Source/drain materials and orientations
  • Gate stack material parameters
  • Device architecture and dimensions
  • Doping concentrations and interface trap distribution
• Interactive GUI for setting up and analyzing the workflow results

Novel STT-MRAM Memory Design

• Model magnetization switching ability of different materials for MTJs in STT-MRAM devices by efficiently computing Spin Transfer Torque at finite bias

Battery Materials Modeling Improvements

• New ionic conductivity and self-diffusion analysis for battery materials
• Possibility to include long-range electrostatic interactions estimated from DFT in Force Fields when modeling liquid battery electrolytes

Polymer Simulation Improvements

• Added Crosslink Builder templates for alcohol-isocyanate and sulfur vulcanization reactions
• Faster crosslink reaction simulations 
• Possibility to constrain bond lengths and angles in MD and optimization of molecules

NanoLab GUI Improvements

• New NanoLab GUI layout, enabling to work efficiently with data intensive projects based on multiple simulations (watch a video to learn more)
• More stable and efficient Job Manager to submit and monitor jobs
• Improved plotting framework, including possibility to have dual axes: one logarithmic and another one-linear scale, and color code the data to match the particular axis