User Guide to Semiempirical Tight Binding
This user guide focuses on the semiempirical quantum mechanical methods GFNn-xTB, their descendants, and corresponding composite schemes as implemented in the xtb (extended tight binding) program package.
We provide a number of detailed guides dealing with common task that can
be performed easily with the
All guides are usually structured the same way, starting with some simple
examples using only the commandline and the default settings followed
by a trouble shooting section.
Detailed inputs are provided in a ready to use fashion to solve some
more special but still common tasks with
xtb together with some
insights into the theory used behind the scences.
Recent developments, news and publications
We are happy to release a new version of
xtbwith exciting new features. First of all, we improved the user-friendliness of the error messages in the geometry reader by adopting our IO-library, which is already in wide use in dftd4, gcp and other projects. No more invalid input provided obscure error messages, but actual pointers on what went wrong with the input.
We also improved the capability of existing geometry readers to keep up to date with the parent programs, for example we now support the
$eht charge=0 unpaired=0line to set the system charge and number of unpaired electrons, which was added in a Turbomole 7.5. Furthermore, we are happy to have now support for QChem molecule files (
.qchem), FHI-aims geometry inputs (
geometry.in) and QCSchema formatted JSON (
A new version of DFTB+ is now available with support for the xTB methods. Due to the integration in DFTB+ most of the features available for the DFTB Hamiltonian can be readily used with the xTB Hamiltonians as well, including periodic calculations with k-point sampling, geometry optimizations, molecular dynamics and frequency calculations. Both GFN1-xTB and GFN2-xTB are available for the DFTB+ version 21.2 at the moment.
Furthermore, we integrated parts the battle-proven rational function optimizer from
xtb --optinto DFTB+ to allow fast and robust geometry optimizations both for molecular and periodic systems. Preliminary tests show an order of magnitude improvements in the convergence compared with the previous default (c.f. dftbplus#862).
This new update of the QCxMS program is linked against the AVX2 processor extension for increased performance. CID and EI runmode changes and bugfixes are implemented. The local run-script pqcxms was re-written and now performs better for parallel calculations.
For a detailed description of all changes, check out the GitHub repository.
This new update of the QCxMS program changes the way xtb is used inside the code. Instead of a standalone implementation of xtb version 5.8.1 in the source code, we switched to using the tblite library, which allows updates to the latest version of xtb. This, in return, leads to a significant increase in the computational speed of calculations done with the GFNn-xTB methods, while keeping the code independent from third party software.
Furthermore, the PlotMS program has been updated as well.
We released a new version of
xtbwith a significantly improved memory footprint for large scale calculations and improved parallelisation for frequency calculations. The parallel evaluation of hessians with GFN-FF is now possible, overall we improved the stablility of the parallelisation which was slightly degraded in version 6.4.0. For xTB calculations the required
OMP_STACKSIZEhas been significantly reduced by restructuring the integral evaluation slightly.
Also, this version of
xtbnow supports the COSMO/CPCM solvation model using the ddPCM library.
See the news archive for all posts.
xTB in Other Quantum Chemistry Programs
The xTB-methods are now officially available in other quantum chemistry programs!
in Orca 4.2 an IO-based interface to the
xtbbinary is available
AMS 2019 implements GFN1-xTB in their DFTB module
the entos program implements GFN1-xTB (also available in the webinterface)
the computational chemistry framework cuby4 supports
Turbomole does support GFN1-xTB and GFN2-xTB since version 7.4
QCEngine supports calculations with the
CP2K has an GFN1-xTB implementation since version 7.1
DFTB+ support GFN1-xTB and GFN2-xTB since version 21.2
We missed your project here? No problem, just give us hint at the mailing list or open an issue at github.
- Setup and Installation
- Quickstart into Production
- Commandline Usage
- Geometry Input
- Detailed Input
- Compiling from Source
- Singlepoint Calculations
- Geometry Optimization
- Exploration of the potential energy surface (PES)
- Implicit Solvation
- Calculation of Vibrational Frequencies
- Molecular Dynamics Simulations
- Meta-Dynamics Simulations
- Reaction Path Methods
- Growing String Method
- Periodic Boundary Conditions
- External Potentials and Embedding
- GFN-Force-Field (GFN-FF)
- Introducing GFN-FF
- Theoretical background
- How to use it
- GFN-FF specific settings
- GFN-FF specific input
- Periodic Boundary Conditions
- 2D to 3D structure converter
- GNF-FF specific output
- C API to the extended tight binding program
- Python Integration for the xtb API
- Community resources
- Introduction to ENSO
- Setting up ENSO
- Manually restarting calculations and modifying enso.json
- Sorting out rotamers at DFT level detected by CREST
- Running on a cluster
- Spectra Plotting
- Introduction to CENSO
- Setting up CENSO
- Censorc keyword definitions
- Trouble shooting
- Usage examples
- Calculate fast DFT(B97-D3(0)/def2-SV(P)+gcp) single-point energies on GFNn-xTB input geometries
- Calculate free energies in solution phase (CHCl3) on GFNn-xTB geometries
- Calculate free energies on populated, DFT optimized conformers
- Calculation of NMR spectra
- Calculation of optical rotation
- Restarting calculations
- Introduction to QCxMS
- Setting up QCxMS
- Running QCxMS
- Workflow of QCxMS
- Input keywords in qcxms.in
- Input Details
- Command line Options
- QCxMS and QCEIMS related publications