10.3 Density functional theory

Density functional theory (DFT) is a powerful computational method for studying the electronic structure of materials. It simplifies complex quantum systems by focusing on electron density rather than many-body wavefunctions. This approach balances accuracy and efficiency, making it widely used in physics, chemistry, and materials science.

DFT's foundation lies in the Hohenberg-Kohn theorems and Kohn-Sham equations. These principles allow researchers to calculate properties like ground-state energies, molecular geometries, and band structures. Various approximations and extensions have been developed to improve DFT's accuracy and broaden its applications across different fields.

Fundamentals of density functional theory

• Density functional theory (DFT) is a quantum mechanical modeling method used to investigate the electronic structure of many-body systems
• DFT has become a popular tool in computational physics, chemistry, and materials science due to its balance between accuracy and computational efficiency
• The fundamental idea behind DFT is to describe a many-electron system by its electron density rather than its many-body wavefunction

Electron density as central variable

• Electron density $\rho(r)$ is the central variable in DFT
• $\rho(r)$ represents the probability of finding an electron at a specific position $r$ in space
• All properties of the system can be determined from the electron density, including the total energy, which is a functional of the electron density $E[\rho(r)]$
• Using electron density simplifies the many-body problem by reducing the number of variables from 3N (where N is the number of electrons) to just 3 spatial coordinates

Hohenberg-Kohn theorems

• The Hohenberg-Kohn theorems form the basis of DFT
• The first theorem states that the ground-state electron density uniquely determines the external potential (up to a constant) and hence the Hamiltonian of the system
• The second theorem establishes a variational principle for the energy functional $E[\rho(r)]$, stating that the electron density minimizing the energy functional is the true ground-state density
• These theorems prove that the electron density contains all the information needed to describe the system and that the ground-state density can be found by minimizing the energy functional

Kohn-Sham equations

• The Kohn-Sham equations are a set of single-particle equations that allow for the practical application of DFT
• They replace the original many-body problem with an auxiliary system of non-interacting electrons moving in an effective potential $v_{eff}(r)$
• The effective potential includes the external potential, the Hartree potential (describing the classical Coulomb interaction between electrons), and the exchange-correlation potential (accounting for all quantum mechanical effects)
• Solving the Kohn-Sham equations self-consistently yields the ground-state electron density and energy of the system

Exchange-correlation functionals

• The exchange-correlation (XC) functional is a key component of DFT, capturing the complex quantum mechanical interactions between electrons
• The exact form of the XC functional is unknown, so approximations are used in practical calculations
• Different levels of approximation lead to various classes of XC functionals with varying accuracy and computational cost

Local density approximation (LDA)

• LDA is the simplest approximation for the XC functional
• It assumes that the XC energy at each point in space depends only on the local electron density at that point
• The XC energy is approximated using the XC energy of a homogeneous electron gas with the same density
• LDA works well for systems with slowly varying electron densities (simple metals) but may struggle with more complex systems (molecules)

Generalized gradient approximations (GGAs)

• GGAs improve upon LDA by incorporating information about the gradient of the electron density $\nabla \rho(r)$ in addition to the local density
• This allows GGAs to better describe systems with rapidly varying electron densities (covalent bonds, surfaces)
• Popular GGA functionals include PBE (Perdew-Burke-Ernzerhof) and BLYP (Becke-Lee-Yang-Parr)
• GGAs generally provide a good balance between accuracy and computational efficiency for many systems

Hybrid functionals

• Hybrid functionals combine a portion of exact exchange from Hartree-Fock theory with exchange and correlation from DFT
• The mixing of exact exchange helps to mitigate the self-interaction error present in pure DFT functionals
• Popular hybrid functionals include B3LYP (Becke, 3-parameter, Lee-Yang-Parr) and PBE0 (Perdew-Burke-Ernzerhof hybrid)
• Hybrid functionals often improve the accuracy for properties like band gaps and reaction barriers but are more computationally expensive than pure DFT functionals

Meta-GGA functionals

• Meta-GGA functionals extend GGAs by including the second derivative of the electron density $\nabla^2 \rho(r)$ or the kinetic energy density $\tau(r)$
• This additional information allows meta-GGAs to capture more complex electronic effects and further improve accuracy
• Examples of meta-GGA functionals include TPSS (Tao-Perdew-Staroverov-Scuseria) and SCAN (Strongly Constrained and Appropriately Normed)
• Meta-GGAs can provide high accuracy for a wide range of systems but are more computationally demanding than GGAs

Basis sets in DFT

• Basis sets are used to represent the Kohn-Sham orbitals in DFT calculations
• The choice of basis set affects the accuracy and computational cost of the calculation
• Different types of basis sets are suited for different systems and properties

Plane wave basis sets

• Plane waves are a natural choice for periodic systems (crystals, surfaces)
• They are orthonormal, complete, and easy to implement with fast Fourier transforms (FFTs)
• The accuracy of plane wave calculations can be systematically improved by increasing the kinetic energy cutoff
• Plane waves are less efficient for systems with localized electrons (molecules, defects) due to the large number of basis functions needed

Localized basis sets

• Localized basis sets (atomic orbitals, Gaussian functions) are well-suited for molecular systems
• They provide a compact representation of the electronic structure and allow for efficient calculations
• Popular localized basis sets include Gaussian-type orbitals (GTOs) and numerical atomic orbitals (NAOs)
• The accuracy of localized basis sets depends on the size and quality of the basis set used (double-zeta, triple-zeta, polarization functions)

Projector augmented wave (PAW) method

• The PAW method is an all-electron approach that combines the efficiency of pseudopotentials with the accuracy of full-potential methods
• It divides the electron density into a smooth part (described by plane waves) and a rapidly varying part near the nuclei (described by localized functions)
• The PAW method allows for the efficient treatment of core electrons while retaining the accuracy of all-electron calculations
• It has become a popular choice for accurate DFT calculations of both periodic and molecular systems

Self-consistent field (SCF) procedure

• The SCF procedure is used to solve the Kohn-Sham equations iteratively until self-consistency is reached
• It involves repeatedly updating the effective potential based on the current electron density until the input and output densities converge

Solving Kohn-Sham equations iteratively

• The SCF procedure begins with an initial guess for the electron density (superposition of atomic densities, previous solution)
• The Kohn-Sham equations are solved using this density to obtain the Kohn-Sham orbitals and eigenvalues
• A new electron density is constructed from the occupied Kohn-Sham orbitals
• The process is repeated using the updated density until self-consistency is achieved

Convergence criteria

• Convergence criteria are used to determine when the SCF procedure has reached self-consistency
• Common criteria include changes in the total energy, electron density, or Kohn-Sham eigenvalues between iterations
• Tighter convergence criteria lead to more accurate results but may require more iterations and computational time
• A balance must be struck between accuracy and efficiency based on the specific system and properties of interest

Mixing schemes for charge density

• Mixing schemes are used to combine the input and output electron densities during the SCF procedure to improve convergence
• Simple mixing (linear combination of input and output densities) can be unstable for some systems
• More advanced mixing schemes (Pulay mixing, Broyden mixing) use information from previous iterations to accelerate convergence
• The choice of mixing scheme and parameters can significantly affect the convergence behavior and computational cost of the SCF procedure

Applications of DFT

• DFT has become a widely used tool for studying the properties and behavior of materials at the atomic scale
• It finds applications in diverse fields, including physics, chemistry, materials science, and biochemistry

Electronic structure calculations

• DFT can be used to calculate the electronic structure of molecules, clusters, and solids
• This includes the determination of band structures, density of states, and orbital energies
• Electronic structure calculations provide insights into the bonding, stability, and optical properties of materials
• Examples include the study of semiconductor band gaps, molecular orbital energies, and surface states

Geometry optimization

• DFT can be used to find the lowest-energy configuration of a system by optimizing its atomic positions
• Geometry optimization is essential for predicting stable structures, reaction pathways, and transition states
• It involves the minimization of the total energy with respect to the atomic coordinates using gradient-based methods (conjugate gradient, quasi-Newton)
• Applications include the determination of molecular geometries, crystal structures, and adsorption sites on surfaces

Molecular dynamics simulations

• DFT can be combined with molecular dynamics (MD) to study the time evolution of systems at the atomic level
• In DFT-based MD, the forces on the atoms are calculated using DFT at each time step, allowing for the simulation of chemical reactions and phase transitions
• DFT-MD can be used to investigate the dynamical properties of materials, such as diffusion coefficients, vibrational spectra, and thermal conductivity
• Examples include the study of ion transport in battery materials, water splitting on catalyst surfaces, and the folding of proteins

Spectroscopic properties

• DFT can be used to simulate various spectroscopic techniques, providing a direct link between theory and experiment
• This includes the calculation of infrared (IR), Raman, and X-ray absorption spectra
• By comparing simulated and experimental spectra, DFT can aid in the interpretation of spectroscopic data and the identification of chemical species
• Examples include the assignment of vibrational modes in IR spectra, the analysis of X-ray absorption near-edge structure (XANES), and the prediction of NMR chemical shifts

Limitations and challenges

• Despite its success, DFT has several limitations and challenges that must be considered when applying the method to real-world problems
• These limitations arise from the approximations made in the exchange-correlation functionals and the treatment of certain types of systems

Accuracy vs computational cost

• There is a trade-off between the accuracy of DFT calculations and their computational cost
• More accurate exchange-correlation functionals (hybrid, meta-GGA) are generally more computationally expensive than simpler approximations (LDA, GGA)
• The choice of basis set also affects the balance between accuracy and cost, with larger basis sets providing better results but requiring more computational resources
• Finding the optimal balance for a given problem requires careful consideration of the system size, desired properties, and available computational resources

Treatment of strongly correlated systems

• DFT struggles to accurately describe systems with strong electron-electron correlations, such as transition metal oxides and rare-earth compounds
• The local nature of most exchange-correlation functionals fails to capture the complex many-body effects in these systems
• Specialized methods, such as DFT+U and dynamical mean-field theory (DMFT), have been developed to address this issue by introducing additional parameters or explicit many-body correlations
• However, these methods often require careful tuning and validation against experimental data

Describing long-range interactions

• Standard DFT functionals have difficulty describing long-range interactions, such as van der Waals (dispersion) forces
• These interactions play a crucial role in the stability and properties of many systems, including molecular crystals, layered materials, and biomolecules
• Several approaches have been proposed to include dispersion effects in DFT, such as empirical corrections (DFT-D), nonlocal functionals (vdW-DF), and many-body dispersion (MBD) methods
• The choice of dispersion correction depends on the specific system and the required accuracy, with more advanced methods being computationally more demanding

Overcoming self-interaction error

• The self-interaction error (SIE) is a common problem in DFT, arising from the incomplete cancellation of an electron's Coulomb interaction with itself
• SIE leads to incorrect descriptions of charge transfer, band gaps, and reaction barriers, particularly in systems with localized electrons
• Hybrid functionals and self-interaction correction (SIC) schemes can help mitigate SIE by introducing exact exchange or removing the self-interaction term
• However, these approaches often come with increased computational cost and may not fully eliminate SIE in all cases

Advanced topics in DFT

• As DFT continues to evolve, new methods and approaches are being developed to address its limitations and extend its applicability to more complex systems
• These advanced topics represent the cutting edge of DFT research and offer exciting possibilities for future applications

Time-dependent density functional theory (TDDFT)

• TDDFT is an extension of DFT that allows for the study of time-dependent phenomena, such as electronic excitations and charge transfer processes
• It is based on the Runge-Gross theorem, which establishes a one-to-one mapping between the time-dependent external potential and the time-dependent electron density
• TDDFT can be used to calculate absorption spectra, excitation energies, and electron dynamics in molecules and materials
• Applications include the study of photochemical reactions, light-harvesting systems, and ultrafast electron transfer

Density functional perturbation theory (DFPT)

• DFPT is a powerful tool for calculating the response of a system to perturbations, such as atomic displacements or electric fields
• It allows for the efficient computation of phonon frequencies, dielectric constants, and Born effective charges
• DFPT is based on the idea of expressing the perturbation as a functional of the electron density and solving the coupled perturbation equations self-consistently
• Applications include the study of lattice dynamics, ferroelectricity, and piezoelectricity in materials

Combining DFT with many-body methods

• DFT can be combined with various many-body methods to improve the description of electron correlations and excited states
• Examples include the GW approximation for quasiparticle energies, the Bethe-Salpeter equation (BSE) for optical spectra, and the dynamical mean-field theory (DMFT) for strongly correlated systems
• These hybrid approaches leverage the strengths of both DFT (efficient description of ground-state properties) and many-body methods (accurate treatment of correlations and excitations)
• Applications include the study of photoemission spectra, exciton binding energies, and metal-insulator transitions

Machine learning approaches to DFT

• Machine learning (ML) techniques are being increasingly applied to DFT to accelerate calculations and discover new materials
• ML can be used to develop accurate and transferable interatomic potentials based on DFT data, enabling large-scale molecular dynamics simulations
• Neural networks can be trained to predict the electronic structure and properties of materials directly from their atomic configuration, bypassing the need for expensive DFT calculations
• Active learning strategies can guide the exploration of vast chemical spaces to identify promising materials for specific applications
• The combination of DFT and ML offers the potential for accelerated materials discovery and optimization, with applications ranging from catalysis to energy storage