TBMaLT, a flexible toolkit for combining tight-binding and machine learning

Tight-binding approaches, especially the Density Functional Tight-Binding (DFTB) and the extended tight-binding schemes, allow for efficient quantum mechanical simulations of large systems and long-time scales. They are derived from ab initio density functional theory using pragmatic approximations and some empirical terms, ensuring a fine balance between speed and accuracy. Their accuracy can be improved by tuning the empirical parameters using machine learning techniques, especially when information about the local environment of the atoms is incorporated. As the significant quantum mechanical contributions are still provided by the tight-binding models, and only short-ranged corrections are fitted, the learning procedure is typically shorter and more transferable as it were with predicting the quantum mechanical properties directly with machine learning without an underlying physically motivated model. As a further advantage, derived quantum mechanical quantities can be calculated based on the tight-binding model without the need for additional learning. We have developed the open-source framework-Tight-Binding Machine Learning Toolkit-which allows the easy implementation of such combined approaches. The toolkit currently contains layers for the DFTB method and an interface to the GFN1-xTB Hamiltonian, but due to its modular structure and its well-defined interfaces, additional atom-based schemes can be implemented easily. We are discussing the general structure of the framework, some essential implementation details, and several proof-of-concept applications demonstrating the perspectives of the combined methods and the functionality of the toolkit.

DFTB+, a software package for efficient approximate density functional theory based atomistic simulations

DFTB+ is a versatile community developed open source software package offering fast and efficient methods for carrying out atomistic quantum mechanical simulations. By implementing various methods approximating density functional theory (DFT), such as the density functional based tight binding (DFTB) and the extended tight binding method, it enables simulations of large systems and long timescales with reasonable accuracy while being considerably faster for typical simulations than the respective ab initio methods. Based on the DFTB framework, it additionally offers approximated versions of various DFT extensions including hybrid functionals, time dependent formalism for treating excited systems, electron transport using non-equilibrium Green's functions, and many more. DFTB+ can be used as a user-friendly standalone application in addition to being embedded into other software packages as a library or acting as a calculation-server accessed by socket communication. We give an overview of the recently developed capabilities of the DFTB+ code, demonstrating with a few use case examples, discuss the strengths and weaknesses of the various features, and also discuss on-going developments and possible future perspectives.

Implementation and benchmark of a long-range corrected functional in the density functional based tight-binding method

Bridging the gap between first principles methods and empirical schemes, the density functional based tight-binding method (DFTB) has become a versatile tool in predictive atomistic simulations over the past years. One of the major restrictions of this method is the limitation to local or gradient corrected exchange-correlation functionals. This excludes the important class of hybrid or long-range corrected functionals, which are advantageous in thermochemistry, as well as in the computation of vibrational, photoelectron, and optical spectra. The present work provides a detailed account of the implementation of DFTB for a long-range corrected functional in generalized Kohn-Sham theory. We apply the method to a set of organic molecules and compare ionization potentials and electron affinities with the original DFTB method and higher level theory. The new scheme cures the significant overpolarization in electric fields found for local DFTB, which parallels the functional dependence in first principles density functional theory (DFT). At the same time, the computational savings with respect to full DFT calculations are not compromised as evidenced by numerical benchmark data.

Extensions of the Time-Dependent Density Functional Based Tight-Binding Approach

The time-dependent density functional based tight-binding (TD-DFTB) approach is generalized to account for fractional occupations. In addition, an on-site correction leads to marked qualitative and quantitative improvements over the original method. Especially, the known failure of TD-DFTB for the description of σ → π* and n → π* excitations is overcome. Benchmark calculations on a large set of organic molecules also indicate a better description of triplet states. The accuracy of the revised TD-DFTB method is found to be similar to first principles TD-DFT calculations at a highly reduced computational cost. As a side issue, we also discuss the generalization of the TD-DFTB method to spin-polarized systems. In contrast to an earlier study, we obtain a formalism that is fully consistent with the use of local exchange-correlation functionals in the ground state DFTB method.

Infection and autoimmunity: Lessons of animal models

While the key initiating processes that trigger human autoimmune diseases remain enigmatic, increasing evidences support the concept that microbial stimuli are among major environmental factors eliciting autoimmune diseases in genetically susceptible individuals. Here, we present an overview of evidences obtained through various experimental models of autoimmunity for the role of microbial stimuli in disease development. Disease onset and severity have been compared in numerous models under conventional, specific-pathogen-free and germ-free conditions. The results of these experiments suggest that there is no uniform scheme that could describe the role played by infectious agents in the experimental models of autoimmunity. While some models are dependent, others prove to be completely independent of microbial stimuli. In line with the threshold hypothesis of autoimmune diseases, highly relevant genetic factors or microbial stimuli induce autoimmunity on their own, without requiring further factors. Importantly, recent evidences show that colonization of germ-free animals with certain members of the commensal flora [such as segmented filamentous bacteria (SFB)] may lead to autoimmunity. These data drive attention to the importance of the complex composition of gut flora in maintaining immune homeostasis. The intriguing observation obtained in autoimmune animal models that parasites often confer protection against autoimmune disease development may suggest new therapeutic perspectives of infectious agents in autoimmunity.

DFTB+, a Sparse Matrix-Based Implementation of the DFTB Method†

A new Fortran 95 implementation of the DFTB (density functional-based tight binding) method has been developed, where the sparsity of the DFTB system of equations has been exploited. Conventional dense algebra is used only to evaluate the eigenproblems of the system and long-range Coulombic terms, but drop-in O(N) or O(N2) modules are planned to replace the small code sections that these entail. The developed sparse storage structure is discussed in detail, and a short overview of other features of the new code is given.

Self-Interaction and Strong Correlation in DFTB

The density functional based tight-binding (DFTB) method can benefit substantially from a number of developments in density functional theory (DFT) while also providing a simple analytical proving ground for new extensions. This contribution begins by demonstrating the variational nature of charge-self-consistent DFTB (SCC-DFTB), proving the presence of a defined ground-state in this class of methods. Because the ground state of the SCC-DFTB method itself can be qualitatively incorrect for some systems, suitable forms of the recent LDA+U functionals for SCC-DFTB are also presented. This leads to both a new semilocal self-interaction correction scheme and a new physical argument for the choice of parameters in the LDA+U method. The locality of these corrections can only partly repair the HOMO-LUMO gap and chemical potential discontinuity, hence a novel method for introducing this further physics into the method is also presented, leading to exact derivative discontinuities in this theory at low computational cost. The prototypical system NiO is used as an illustration for these developments.

Overcoordinated hydrogens in the carbon vacancy: donor centers of SiC

Epitaxial silicon carbide is likely to contain hydrogen and vacancies ( V); therefore, V+nH complexes are likely to influence its electronic properties. Using ab initio calculations we show that neutral and positive H atoms are trapped by carbon vacancies ( V(C)) in three-center bonds with two Si neighbors. The double positive charge state of V(C)+H is not stable in equilibrium and in the triply positive state H binds only to one of the Si neighbors. At most two H atoms can be accommodated by a single V(C). The V(C)+nH complexes have donor character and exhibit rather atypical vibration modes for Si-H bonds. Occupation levels and spin distributions were calculated and compared for V(C)+H and V(C).