Official Resources
- Homepage: https://github.com/m-a-d-n-e-s-s/madness
- Documentation: https://github.com/m-a-d-n-e-s-s/madness/wiki
- Source Repository: https://github.com/m-a-d-n-e-s-s/madness
- License: GNU General Public License v2.0
Overview
MADNESS (Multiresolution Adaptive Numerical Environment for Scientific Simulation) is a high-performance computational chemistry and physics package using multiresolution adaptive numerical methods based on multiwavelet basis functions. Developed primarily at Oak Ridge National Laboratory, MADNESS provides systematic, guaranteed-precision calculations with adaptive resolution and excellent parallel scalability. It represents a fundamentally different numerical approach compared to traditional plane-wave or Gaussian basis methods.
Scientific domain: Multiresolution methods, high-precision DFT, adaptive algorithms, HPC
Target user community: Method developers, HPC researchers, high-precision calculations
Theoretical Methods
- Kohn-Sham DFT (LDA, GGA)
- Hartree-Fock
- Multiresolution multiwavelet basis
- Adaptive numerical precision
- Guaranteed error bounds
- Response properties
- Time-dependent DFT
- Coupled cluster methods
- Periodic and non-periodic systems
- All-electron or pseudopotential
Capabilities (CRITICAL)
- Ground-state electronic structure
- Geometry optimization
- Molecular properties
- Response properties
- Excited states (TDDFT)
- Systematic convergence control
- Adaptive resolution
- Guaranteed precision
- Excellent parallel scalability (10,000+ cores)
- Non-periodic and periodic systems
- All-electron accuracy
- Large systems (hundreds of atoms)
- High-precision benchmarks
- Nuclear physics applications
- Integral equation methods
Sources: GitHub repository (https://github.com/m-a-d-n-e-s-s/madness)
Key Strengths
Multiresolution:
- Adaptive wavelet basis
- Automatic refinement
- Guaranteed precision
- No basis set superposition error
- Systematic convergence
Precision Control:
- User-specified accuracy
- Error bounds guaranteed
- No empirical parameters
- Systematic improvement
- Benchmark quality
Scalability:
- Excellent parallel performance
- Scales to 10,000+ cores
- Task-based parallelism
- Distributed memory
- HPC optimized
Generality:
- Molecules and solids
- Periodic and non-periodic
- All-electron treatment
- Broad applicability
- No shape approximations
Innovation:
- Novel numerical methods
- Research platform
- Algorithm development
- Method benchmarking
Inputs & Outputs
-
Input formats:
- Text-based input
- XYZ coordinates
- Python interface
- C++ API
-
Output data types:
- Energies and properties
- Wavefunctions
- Densities
- Molecular orbitals
- Analysis data
Interfaces & Ecosystem
-
Programming:
- C++ core library
- Python interface
- MPI parallelization
- Modern C++ features
-
Visualization:
- Standard format export
- Custom analysis tools
- Post-processing scripts
-
HPC Integration:
- Leadership-class systems
- Task-based runtime
- Scalable algorithms
Workflow and Usage
Typical Usage:
- Set precision threshold
- Define molecular system
- Run calculation
- Analyze results
- Verify convergence
Python Interface Example:
from madness import *
# Define molecule
# Set precision
# Run calculation
# Extract properties
Precision Control:
- Set target accuracy (e.g., 10^-6)
- Automatic adaptive refinement
- Guaranteed error bounds
- Systematic convergence
Advanced Features
Multiresolution Analysis:
- Wavelet-based representation
- Hierarchical grids
- Adaptive refinement
- Efficient for all length scales
- Automatic resolution control
Guaranteed Precision:
- Mathematical error bounds
- No empirical cutoffs
- Systematic convergence
- Reproducible accuracy
- Benchmark quality
Task-Based Parallelism:
- Dynamic load balancing
- Asynchronous execution
- Communication hiding
- Scalable to extreme scale
- Fault tolerance
Response Properties:
- Linear response
- Polarizabilities
- Hyperpolarizabilities
- Frequency-dependent
- High accuracy
TDDFT:
- Real-time propagation
- Linear response
- Excited states
- Optical properties
- Systematic precision
Performance Characteristics
- Speed: Competitive for high precision
- Scaling: Excellent (10,000+ cores)
- Precision: User-controlled, guaranteed
- Memory: Adaptive, efficient
- Typical systems: 10-500 atoms
Computational Cost
- DFT: Comparable to traditional methods
- High precision: More efficient than basis set extrapolation
- Large systems: Excellent scaling
- Response properties: Efficient
- Adaptive: Cost scales with required accuracy
Limitations & Known Constraints
- Learning curve: Steep, research code
- Documentation: Limited, GitHub wiki
- Community: Small, specialized
- Features: Fewer than production codes
- User interface: Developer-oriented
- Platform: Linux HPC systems
- Maturity: Research/development code
Comparison with Other Codes
- vs VASP/QE: MADNESS different numerical approach
- vs Gaussian: MADNESS guaranteed precision, adaptive
- vs FHI-aims: Both all-electron, different basis
- vs Traditional methods: MADNESS systematic convergence
- Unique strength: Multiresolution wavelets, guaranteed precision, extreme scalability, adaptive algorithms
Application Areas
High-Precision Benchmarks:
- Method validation
- Basis set studies
- Accuracy standards
- Reference calculations
- Error analysis
Method Development:
- Algorithm research
- Numerical methods
- Parallel computing
- Mathematical foundations
- New approaches
Large-Scale HPC:
- Extreme scaling studies
- Leadership computing
- Parallel algorithms
- Performance analysis
- Scalability research
Nuclear Physics:
- Nuclear structure
- Many-body methods
- Integral equations
- High-precision requirements
Best Practices
Precision Selection:
- Choose appropriate threshold
- Balance accuracy/cost
- Test convergence
- Verify error bounds
- Document choices
Parallelization:
- Test scaling on target system
- Optimize task granularity
- Balance load
- Use appropriate MPI configuration
System Setup:
- Good initial geometry
- Appropriate symmetry
- Check input parameters
- Verify setup
Convergence:
- Monitor precision metrics
- Check adaptive refinement
- Verify systematic improvement
- Compare with references
Community and Support
- Open-source (GPL v2)
- GitHub repository
- Wiki documentation
- Research community
- ORNL development
- Academic collaborations
Educational Resources
- GitHub wiki
- Published papers
- Code documentation
- Research publications
- Conference presentations
Development
- Oak Ridge National Laboratory
- Robert Harrison (original developer)
- Active research development
- Community contributions
- Modern C++ codebase
- Continuous improvement
Research Applications
- High-precision calculations
- Method benchmarking
- Algorithm development
- Scalability studies
- Nuclear structure
Technical Innovation
Multiwavelet Basis:
- Hierarchical representation
- Systematic refinement
- No basis set limit
- Guaranteed convergence
- Efficient for all systems
Adaptive Methods:
- Automatic resolution
- Error-driven refinement
- Optimal efficiency
- User-controlled precision
- Mathematical guarantees
Parallel Design:
- Task-based model
- Dynamic scheduling
- Asynchronous communication
- Fault-tolerant
- Exascale-ready
Verification & Sources
Primary sources:
- GitHub repository: https://github.com/m-a-d-n-e-s-s/madness
- Wiki: https://github.com/m-a-d-n-e-s-s/madness/wiki
- R. J. Harrison et al., SIAM J. Sci. Comput. 38, S123 (2016) - MADNESS overview
- F. A. Bischoff et al., J. Chem. Phys. 137, 104103 (2012) - Molecular applications
Secondary sources:
- GitHub documentation
- Published studies using MADNESS
- HPC and method development papers
- Confirmed in source lists (LOW_CONF due to research/specialized nature)
Confidence: LOW_CONF - Research code, smaller community, specialized methods
Verification status: ✅ VERIFIED
- GitHub repository: ACCESSIBLE
- Documentation: Basic (wiki, papers)
- Source code: OPEN (GitHub, GPL v2)
- Community support: GitHub issues, research group
- Academic citations: >200
- Active development: ORNL research group
- Specialized strength: Multiresolution multiwavelet methods, guaranteed precision, adaptive algorithms, extreme parallel scalability, systematic convergence, benchmark-quality calculations