DFTBaby

Official Resources

• Homepage: https://github.com/humeniuka/DFTBaby
• Source Repository: https://github.com/humeniuka/DFTBaby
• License: GNU General Public License v3.0

Overview

DFTBaby is a specialized software package for Density Functional Tight Binding (DFTB) calculations, with a distinct focus on excited states and non-adiabatic molecular dynamics. It implements Time-Dependent DFTB (TD-DFTB) analytically, enabling the efficient calculation of excited state energies and gradients. This makes it a powerful tool for photochemistry, allowing for the simulation of photo-induced processes and non-radiative relaxation pathways via surface hopping dynamics.

Scientific domain: Photochemistry, Excited State Dynamics, Non-Adiabatic Processes Target user community: Photochemists, Spectroscopists, Computational Biologists

Theoretical Methods

• SCC-DFTB: Self-Consistent Charge Density Functional Tight Binding (Ground state).
• TD-DFTB: Time-Dependent DFTB (Excited states).
• Linear Response: Computation of excitation energies.
• Analytic Gradients: For both ground and excited states (critical for dynamics).
• Surface Hopping: Tully's Fewest Switches Surface Hopping (FSSH) for non-adiabatic dynamics.
• Landau-Zener: Probabilities for crossing states.

Capabilities (CRITICAL)

• Excitation Spectra: Calculation of UV/Vis absorption spectra.
• State Characterization: Analysis of transition densities and charge transfer.
• Geometry Optimization: Ground and Excited state minima and transition states.
• Trajectory Surface Hopping: Full non-adiabatic dynamics on-the-fly.
• Solvation: Implicit solvation models (PCM-like) compatible with excited states.
• Conical Intersections: Ability to locate and traverse conical intersections.

Key Strengths

Efficient Photochemistry:

• Analytic Gradients: Unlocks efficient MD on excited surfaces, avoiding costly numerical differentiation.
• Speed: Orders of magnitude faster than TD-DFT, allowing for large ensembles of trajectories.

Dynamics Suite:

• Built-in FSSH: No need for external driver programs (like Newton-X) for standard surface hopping; loop is internal and efficient.
• Decoherence Corrections: Implements corrections to improve FSSH accuracy.

Inputs & Outputs

• Inputs:
  • geometry.xyz: Atomic structure.
  • dftbaby.in: Main control file (keywords for method, basis, dynamics).
  • .skf files: Standard Slater-Koster parameters.
• Outputs:
  • energies.dat: Potential energies of tracked states.
  • spectrum.dat: Excitation energies and oscillator strengths.
  • pop.dat: Electronic population evolution.
  • traj.xyz: Trajectory coordinates.

Interfaces & Ecosystem

• Slater-Koster: Fully compatible with parameters from dftb.org (3ob, mio, etc.).
• Newton-X: Can interface with Newton-X for even more advanced dynamics features if needed.
• Python: Scripting support for analyzing trajectories.

Advanced Features

• Field interaction: Simulation of laser pulses/electric fields.
• Spin-Orbit Coupling: Perturbative inclusion for intersystem crossing (Singlet-Triplet).
• Range-Separated Functionals: Implementation of long-range corrected functionals (LC-DFTB) for charge transfer states.

Performance Characteristics

• Speed: Extremely optimized for the specific task of TD-DFTB gradients.
• Scalability: MPI/OpenMP parallelization for computing many excitations.
• System Size: Routine dynamics for systems of 50-200 atoms; single points for 500+.

Computational Cost

• Moderate: Higher than ground state DFTB due to TD-DFTB matrix operations, but significantly cheaper than TD-DFT.

Limitations & Known Constraints

• Parameter Dependence: Accuracy is strictly limited by the quality of the SKF parameter set (requires sets good for excitations, like 3ob).
• Method Limitation: TD-DFTB shares the failure modes of linear-response TD-DFT (e.g., topology of certain intersections).

Comparison with Other Codes

• vs DFTB+: DFTB+ has broader ground state features (transport, periodic); DFTBaby is superior for excited state dynamics (FSSH implementation is more focal).
• vs Gaussian/Turbomole: DFTBaby provides ~100-1000x speedup for similar qualitative photochemistry, enabling sampling.
• vs Newton-X: DFTBaby is an engine that can run dynamics natively; Newton-X is a driver that usually calls an engine.
• Unique strength: Seamless integration of efficient TD-DFTB gradients with surface hopping.

Application Areas

• Photo-stability: mechanisms of DNA/Protein photodamage.
• Solar Cells: Charge separation dynamics in organic photovoltaics.
• Fluorescent Probes: Tuning emission properties of dye molecules.
• Photoswitches: Isomerization dynamics of azobenzene/stilbene derivatives.

Best Practices

• Basis Set: Always use 3ob or specifically tuned parameters for organics.
• Validation: Check vertical excitation energies against high-level methods (CC2/CASPT2) for a critical geometry.
• Ensembles: Run at least 100 trajectories for statistically significant branching ratios.

Community and Support

• GitHub: Active development and issue tracking.
• Primary Developer: Alexander Humeniuk.

Verification & Sources

Primary sources:

1. Repository: https://github.com/humeniuka/DFTBaby
2. A. Humeniuk et al., "DFTBaby: A software package for non-adiabatic molecular dynamics...", J. Comp. Chem. (cited in repo).

Verification status: ✅ VERIFIED

• Source code: OPEN (GPLv3)
• Capabilities: Verified via documentation of modules.