Official Resources
- Source Repository: https://github.com/DeePTB-Lab/dpnegf
- Documentation: Included in repository
- License: Open source
Overview
DPNEGF (DeePTB-NEGF) is a Python package that integrates the Deep Learning Tight-Binding (DeePTB) approach with the Non-Equilibrium Green's Function (NEGF) method, establishing an efficient quantum transport simulation framework with first-principles accuracy. It enables fast quantum transport calculations using ML-trained tight-binding Hamiltonians.
Scientific domain: ML-accelerated quantum transport, DeePTB-NEGF
Target user community: Researchers needing fast quantum transport simulations with DFT accuracy using machine learning
Theoretical Methods
- Non-equilibrium Green's function (NEGF)
- Deep Learning Tight-Binding (DeePTB)
- Slater-Koster parameterization
- LCAO Kohn-Sham Hamiltonian
- Machine learning Hamiltonian
- Quantum transport in open-boundary systems
Capabilities (CRITICAL)
- Quantum transport with ML-trained Hamiltonians
- DeePTB-SK (Slater-Koster) transport
- DeePTB-E3 (LCAO) transport
- Open-boundary system simulation
- Transmission function calculation
- I-V characteristics
- First-principles accuracy at TB speed
Sources: GitHub repository, npj Comput. Mater.
Key Strengths
ML-Accelerated Transport:
- DFT accuracy at TB speed
- Orders of magnitude faster than DFT+NEGF
- Systematic improvement with training
- Generalizable to new structures
DeePTB Integration:
- Well-established ML-TB framework
- Trained on DFT data
- Environment-corrected Hamiltonians
- Multiple basis options
NEGF Framework:
- Open-boundary conditions
- Self-energy calculation
- Transmission and conductance
- Bias-dependent transport
Inputs & Outputs
-
Input formats:
- DeePTB model files
- Device structure
- Transport configuration
-
Output data types:
- Transmission vs energy
- I-V characteristics
- Local density of states
- Current density
Interfaces & Ecosystem
- DeePTB: ML tight-binding framework
- PyTorch: ML backend
- Python: Scripting
- NumPy: Numerical computation
Performance Characteristics
- Speed: Much faster than DFT+NEGF
- Accuracy: Near DFT quality
- System size: Thousands of atoms
- Memory: Moderate
Computational Cost
- ML prediction: Fast (milliseconds per Hamiltonian)
- NEGF calculation: Minutes
- Full I-V: Hours
- vs DFT+NEGF: 10-100x speedup
Limitations & Known Constraints
- Training data dependent: Quality limited by DeePTB training
- DeePTB dependency: Requires trained model
- Newer code: Less established than traditional NEGF
- Documentation: Limited
Comparison with Other Codes
- vs Transiesta: DPNEGF is ML-fast, Transiesta is DFT-accurate but slow
- vs Nanodcal: DPNEGF is open source ML, Nanodcal is commercial LCAO
- vs NanoNet: DPNEGF uses ML Hamiltonians, NanoNet uses manual TB
- Unique strength: ML-accelerated quantum transport with DFT accuracy, DeePTB-NEGF integration
Application Areas
Nanoscale Electronics:
- FET device simulation
- Nanowire transport
- 2D material devices
- Contact resistance
Break Junctions:
- Atomic-scale contacts
- Conductance quantization
- Structure-dependent transport
- Comparison with experiment
High-Throughput Transport:
- Materials screening for devices
- Transport property databases
- Structure-property mapping
- Device optimization
Best Practices
DeePTB Training:
- Use diverse training structures
- Validate Hamiltonian accuracy
- Test extrapolation carefully
- Monitor energy convergence
NEGF Calculation:
- Use sufficient energy grid
- Include adequate lead layers
- Check self-energy convergence
- Validate against DFT+NEGF
Community and Support
- Open source on GitHub
- Developed by DeePTB Lab
- Published in npj Comput. Mater.
- Active development
Verification & Sources
Primary sources:
- GitHub: https://github.com/DeePTB-Lab/dpnegf
- J. Zou et al., npj Comput. Mater. (2024)
Confidence: VERIFIED
Verification status: ✅ VERIFIED
- Source code: ACCESSIBLE (GitHub)
- Documentation: Included in repository
- Published methodology: npj Comput. Mater.
- Active development: Ongoing
- Specialized strength: ML-accelerated quantum transport with DFT accuracy, DeePTB-NEGF integration