inq

Official Resources

• Homepage: https://gitlab.com/npneq/inq
• Documentation: https://inq.readthedocs.io/
• Source Repository: https://gitlab.com/npneq/inq (also GitHub mirror)
• GitHub: https://github.com/LLNL/inq
• License: Mozilla Public License 2.0

Overview

inq is a modern, GPU-accelerated electronic structure code for DFT and real-time TDDFT calculations, developed at Lawrence Livermore National Laboratory (LLNL). Written in C++, it is designed from scratch for GPU execution and focuses on nonequilibrium phenomena in materials.

Scientific domain: Real-time dynamics, excited states, nonequilibrium phenomena, materials science
Target user community: Researchers studying ultrafast dynamics, electronic excitations, and nonequilibrium materials

Theoretical Methods

• Density Functional Theory (DFT)
• Real-time TDDFT
• Plane-wave/real-space hybrid basis
• LDA, GGA, meta-GGA functionals
• Hybrid functionals (HSE, PBE0)
• Spin-polarized and non-collinear spin
• Norm-conserving pseudopotentials

Capabilities (CRITICAL)

• Ground-state DFT
• Real-time TDDFT simulations
• GPU-accelerated execution
• Absorption spectra
• Charge dynamics
• Ion dynamics
• Hybrid functionals
• Non-collinear magnetism
• Multiple spin configurations
• HPC cluster support

Sources: LLNL, arXiv publications, GitHub

Key Strengths

GPU Native:

• Designed for GPU from ground up
• CUDA/ROCm support
• Excellent GPU utilization
• Modern HPC architecture

Real-Time TDDFT:

• Nonperturbative dynamics
• Ultrafast phenomena
• Charge carrier dynamics
• Strong-field physics

Modern C++:

• Clean codebase (~12,000 lines)
• Maintainable
• Extensible
• Modern practices

LLNL Quality:

• DOE national lab development
• Exascale computing focus
• Production-tested

Inputs & Outputs

• Input formats:

  • C++ API
  • Python interface (developing)
  • Structure inputs

• Output data types:

  • Energies
  • Time-dependent observables
  • Spectra
  • Dynamics trajectories

Interfaces & Ecosystem

• HPC integration:

  • MPI parallelization
  • GPU clusters
  • Slurm compatible

• NPNEQ Center:

  • Part of DOE center
  • Collaborative development

Advanced Features

Ultrafast Dynamics:

• Femtosecond timescales
• Pump-probe simulations
• Electron-ion coupling
• Hot carrier dynamics

Strong-Field Physics:

• High-intensity excitations
• Nonlinear response
• Field-induced phenomena

Non-Collinear Spin:

• Spin dynamics
• Magnetic systems
• Spin-orbit effects

Performance Characteristics

• Speed: GPU-optimized
• Accuracy: Standard DFT/TDDFT
• System size: HPC-scalable
• Memory: GPU memory constraints
• Parallelization: MPI + GPU

Computational Cost

• GPU efficiency: Excellent
• RT-TDDFT: Time-stepping overhead
• Typical: GPU cluster runs

Limitations & Known Constraints

• Maturity: Actively developing
• Documentation: Growing
• Traditional DFT: TDDFT focus
• CPU fallback: GPU-primary

Comparison with Other Codes

• vs Octopus: Both RT-TDDFT, inq GPU-native
• vs SALMON: Similar TDDFT, different implementation
• Unique strength: GPU-native design, LLNL backing

Application Areas

Ultrafast Science:

• Pump-probe spectroscopy
• Attosecond dynamics
• Carrier relaxation

Optical Properties:

• Absorption spectra
• Dielectric response
• Plasmonics

Materials Dynamics:

• Electron-phonon coupling
• Phase transitions
• Radiation effects
• Energy transfer

Best Practices

Ground State Setup:

• Converge ground state fully before dynamics
• Use appropriate exchange-correlation functional
• Test k-point convergence for periodic systems
• Verify energy conservation

RT-TDDFT Simulations:

• Choose appropriate time step (0.01-0.1 fs typical)
• Apply kick pulse for linear response spectra
• Use proper absorbing boundaries for finite systems
• Monitor total energy drift

GPU Optimization:

• Match problem size to GPU memory
• Use batched calculations when possible
• Profile to find memory bottlenecks
• Consider multi-GPU for larger systems

Post-Processing:

• Fourier transform for spectra from time-domain data
• Apply windowing functions to reduce spectral artifacts
• Calculate dipole moments for absorption spectra
• Analyze induced densities for insight

Computational Cost

• Ground state DFT: Comparable to other GPU codes
• RT-TDDFT: Linear in simulation time, each step similar cost to ground state
• GPU efficiency: 10-100x speedup over CPU for suitable systems
• Memory: GPU memory often limiting factor
• Typical: Seconds for small molecules, hours for large periodic
• Scaling: Cubic O(N³) with system size for ground state

Community and Support

• Open source MPL 2.0
• LLNL development
• DOE NPNEQ center
• GitLab/GitHub
• Active development

Verification & Sources

Primary sources:

1. GitLab: https://gitlab.com/npneq/inq
2. GitHub: https://github.com/LLNL/inq
3. arXiv publications
4. LLNL computational materials

Confidence: VERIFIED - LLNL official code

Verification status: ✅ VERIFIED

• Source code: OPEN (MPL 2.0)
• Documentation: ReadTheDocs
• Development: Active
• Specialty: GPU-native DFT/RT-TDDFT, nonequilibrium dynamics