Calculating Energy Bands Alt reference: Ashcroft Ch.10-11 Sommerfeld (1927) Drude (1900) Schrdinger equation (1925) Wannier function (1937) Tight Binding (1954) GW approx. DFT (1965) (1964) Fireball (ab-initio DFT)

kp (1955) LCAO Recently, more of a Muffin-tin Hubbard (1928) (1963): focus on computational approx. Simplest (1947) implementation and interacting Kroniq-Penney accuracy particles (1931) (e-) Approximation & Sch. Equation 2 2 ( U (r )) k (r ) (k ) k (r ) 2m An independent SE for each e- is simplification

The ind. e- approx. however doesnt ignore all U(r)=periodic potential + periodic interactions To know U(r) with interactions, you need To know , you need U(r) What to do? Guess U(r), use to solve Then what? Use to get better guess of U(r), repeat same Generalizations to all Methods Except with the simplest 1D examples, the S. E. cannot be solved exactly All methods require approximations And high speed computing! Thus the type of approximations people have tried has been limited by computing techniques and computing power Focus on higher energy bands as tight binding is pretty good for lower bands Hexagonal lattice

The Cellular Method (1934) First iterative approach was the cellular method by Wigner and Seitz (know that name?) Since we have periodicity, it is enough to solve the S.E. within a single primitive cell Co The wavefunction in other cell is then ik R over R ) e harmonics, k (r ) is a sum need BCs k ( r spherical Computationally challenging to solve B.C.s Results in potential with discontinuous derivative at cell boundary Muffin-tin potential Solves both complaints of the last method: One way to make sure continuous is set to 0! U(r)=V(|r-R|), when |r-R| ro (the core region)

=V(ro)=0, when |r-R| ro (interstitial region) ro is less than half of the Of nearest neighbor distance course you are just moving your discontinuous derivative problem, but maybe to better spot? How to deal with a discontinuous derivative (but continuous ) Best to use variational principle rather than S.E. 2 2 2 ( 2m (r ) U (r )) (r ) dr E[ k ] 2 ( r )

dr E[k] is the energy of (k) of the level k 2 Drawback: 2 ( Different U (r )) k (starting r ) (kpotentials ) k (r ) 2m can give different results Two methods use the muffin tin potential Augmented Plane-Wave method (APW) In the interstitial region =eikr k, In the atomic region, k, satisfies S.E. 2 2

( k , (r ) V ( r R )k , (r )) k , (r ) 2m Only k dependence is in the interstitial region 2 2 k In interstitial region: Hk , (r ) k , (r ) 2m Thousands of APWs can be used Another approach using Muffin tin The other method is called the Greens function approach or the Korringa, Kohn, and Rostoker (KKR) method Formulation seems very different, but it has been established that the methods yield the same results using the same potential k (r ) G ( k ) (r r ' )U (r ' ) k (r ' )dr '

iK r r ' 2m e G (r r ' ) 2 4 r r ' Orthogonalized plane wave method (OPW) Good if dont want a doctored potential Orthogonalized plane waves defined as: k e ik r c k bc (r ) c Core levels needed (generally tight binding) Constants b determined by orthogonality c c

k * 0 dr (r ) k (r ) c * ik r This implies bc dr k (r ) e Second term small in interstitial region So close to a plane wave in interstitial region Pseudopotential Method Began as an extension of OPW If we act H on e ik r k b c c k (r )

c In the outer region, this gives ~ free energy What goes on in core is largely irrelevant to the energy, so lets just ignore it U(r)=0 , when r Re (the core region) =-e2/r, when r Re (interstitial region) Pseudopotential Method Calculation of band structure depends only on the Fourier components of the pseudopotential at the reciprocal lattice vectors (edges of the BZ). Usually, only a few values of U are needed Constants from models or fits to optical measurements of reflectance and absorption Great predictive value for new compounds Often possible to calculate band structures, cohesive energy, lattice constants and bulk moduli from first principles Nearly free e-s Tight-binding/LCMO

Large overlap Wave functions electrons indep. ~ plane waves of each other Assume energy Linear combination of is unchanged and solve for

Assume some kp Theory Useful for understanding interactions between bands Critical points of BZ have specific properties. If critical point energies are Wannier functions known, treat nearby points 1st order = unperturbed as critical energy plus correction atomic orbital perturbation Pseudopotential Method includes Coulomb repulsion & Pauli exclusion. No exact

way to calculate V(r), guess and iterate. Valence bands->charge density=*->V Density functional theory (DFT) takes into account Coulomb, exchange and correlation energies of electrons. Guess and iterate. Gives good bandstructure. kp Theory Most holes (electrons) spend most of their time near the top (bottom) of the valence (conduction) band so properties nearby these points important kp Theory Most holes (electrons) spend most of their time near the top (bottom) of the valence (conduction) band so properties nearby these points important Based on perturbation theory V~ is the periodic potential (of the lattice), and VU is the confinement potential

V0 and x0 are some arbitrary positive constants. If VU is small, then the solutions to the S.E. are of the Bloch form: Essense of kp Theory Plug Bloch nk ( x) into unkS.E. ( x )e Reference in notes ikx After lots of manipulation: When we plug in the Bloch wavefunction, we can write the Schrodinger equation in this form. Ek Nearly free e-s Tight-binding/LCMO

Large overlap Wave functions electrons indep. ~ plane waves of each other Assume energy Linear combination of is unchanged and solve for

Assume some kp Theory Useful for understanding interactions between bands Critical points of BZ have specific properties. If critical point energies are Wannier functions known, treat nearby points 1st order = unperturbed as critical energy plus correction atomic orbital perturbation

Pseudopotential Method includes Coulomb repulsion & Pauli exclusion. No exact way to calculate V(r), guess and iterate. Valence bands->charge density=*->V Density functional theory (DFT) takes into account Coulomb, exchange and correlation energies of electrons. Guess and iterate. Gives good bandstructure. Basics of DFT-LDA (+U) DFTs failures: Materials containing localized d and f electrons whose contributions to exchange and correlation are not accurately computed in the commonly used local density approximation. Most basic fix is LDA+U, where to the DFT functional is added an orbital-dependent interaction term characterized by an energy scale U, the screened Coulomb interaction between the correlated orbitals. Success = the high temperature superconducting cuprates. LDA+Us notable failures. While there are exceptions, it is not expected to adequately describe a system which is not a good insulator. Dynamical Mean Field Theory (Tudor)

DMFT (or LDA+DMFT) goes beyond LDA+U by allowing the interaction potential of the correlated orbitals to be energy (frequency) dependent. This frequency dependent potential, or self-energy, is computed for the correlated orbitals only using many-body techniques within an accurate impurity solver. This calculation can be done as accurately as one desires, and it is significantly cheaper in CPU time than solving a full many-body problem