Electronic Structure Calculations, Density Functional
Transcription
Electronic Structure Calculations, Density Functional
Tutoriel BigDFT G RENOBLE Review Introduction Atomistic Methods DFT Kohn-Sham Pseudopotential Electronic Structure Calculations, Density Functional Theory and its Modern Implementations Solving Basis sets Plane waves Thierry Deutsch Direct Minimisation Diagonalisation Gaussians Real space L_Sim - CEA Grenoble Conclusion 19 October 2011 Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Outline 1 Review of Atomistic calculations Introduction Atomistic Methods 2 Density Functional Theory (quick view) Kohn-Sham formalism Pseudopotential Solving Kohn-Sham equations 3 Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians Real space basis sets 4 Conclusion Review Introduction Atomistic Methods DFT Kohn-Sham Pseudopotential Solving Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians Real space Conclusion Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Review Introduction Atomistic Methods Review of Atomistic simulations DFT Kohn-Sham Pseudopotential Solving Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians Real space Conclusion Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Review of Atomistic Simulations Theory – Experiment – Simulation Hardware – Computers Algorithms Review Introduction Atomistic Methods DFT Kohn-Sham Pseudopotential Solving Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians Real space Conclusion Atomistic Simulations Force fields (interatomic potentials) Tight Binding Methods Hartree-Fock Density Functional Theory Configuration interactions Quantum Monte-Carlo Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Review of Atomistic Simulations Theory – Experiment – Simulation Hardware – Computers Algorithms Review Introduction Atomistic Methods DFT Kohn-Sham Pseudopotential Solving Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians Real space Conclusion Atomistic Simulations Force fields (interatomic potentials) Tight Binding Methods Hartree-Fock Density Functional Theory Configuration interactions Quantum Monte-Carlo Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Algorithms Search of efficient algorithms: Fast Fourier Transform (FFT) : TCPU ∝ n log (n) ; Order Natoms methods Review Introduction Atomistic Methods Electrostatic (Hartree) potential of an electronic density ρ(r ): DFT Kohn-Sham Pseudopotential ρ(r 0 ) 0 dr V |r − r 0 | Z V (r ) = Solving Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians Real space Conclusion Mesh: n = 100 × 100 × 100 = 106 Direct: n2 = 1012 : ∝ 100 s FFT: n log (n) ∝ 106 : ∝ 0.1 ms Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Algorithms Search of efficient algorithms: Fast Fourier Transform (FFT) : TCPU ∝ n log (n) ; Order Natoms methods Review Introduction Atomistic Methods Electrostatic (Hartree) potential of an electronic density ρ(r ): DFT Kohn-Sham Pseudopotential ρ(r 0 ) 0 dr V |r − r 0 | Z V (r ) = Solving Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians Real space Conclusion Mesh: n = 100 × 100 × 100 = 106 Direct: n2 = 1012 : ∝ 100 s FFT: n log (n) ∝ 106 : ∝ 0.1 ms Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Atomistic Simulations Two intrinsic difficulties for numerical atomistic simulations, related to complexity: Interactions The way that atoms interact is known: i } ∂Ψ ∂t = H Ψ Review Introduction Atomistic Methods H ψ = E0 ψ Exploration of the configuration space (ART, minima hopping, spline search) DFT Kohn-Sham Pseudopotential Solving Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians Real space Conclusion Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Atomistic Simulations Two intrinsic difficulties for numerical atomistic simulations, related to complexity: Interactions The way that atoms interact is known: i } ∂Ψ ∂t = H Ψ Review Introduction Atomistic Methods H ψ = E0 ψ Exploration of the configuration space (ART, minima hopping, spline search) DFT Kohn-Sham E pot Pseudopotential Solving Basis sets Plane waves Direct Minimisation Diagonalisation R1 Gaussians Real space Conclusion Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Atomistic Simulations Two intrinsic difficulties for numerical atomistic simulations, related to complexity: Interactions The way that atoms interact is known: i } ∂Ψ ∂t = H Ψ Review Introduction Atomistic Methods H ψ = E0 ψ Exploration of the configuration space (ART, minima hopping, spline search) DFT Kohn-Sham E pot Pseudopotential Solving Basis sets Plane waves Direct Minimisation Diagonalisation R1 Gaussians Real space Conclusion R2 Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Atomistic Simulations Two intrinsic difficulties for numerical atomistic simulations, related to complexity: Interactions The way that atoms interact is known: i } ∂Ψ ∂t = H Ψ Review Introduction Atomistic Methods H ψ = E0 ψ Exploration of the configuration space (ART, minima hopping, spline search) DFT Kohn-Sham E pot Pseudopotential Solving Basis sets Plane waves Direct Minimisation Diagonalisation R1 Gaussians Real space R3 Conclusion Laboratoire de Simulation Atomistique R2 http://inac.cea.fr/L_Sim T. Deutsch Atomistic Simulations Two intrinsic difficulties for numerical atomistic simulations, related to complexity: Interactions The way that atoms interact is known: i } ∂Ψ ∂t = H Ψ Review Introduction Atomistic Methods H ψ = E0 ψ Exploration of the configuration space (ART, minima hopping, spline search) DFT Kohn-Sham E pot Pseudopotential R1000 Solving Basis sets Plane waves Direct Minimisation Diagonalisation R1 Gaussians Real space Conclusion R41 Laboratoire de Simulation Atomistique R3 Rn R2 http://inac.cea.fr/L_Sim T. Deutsch Goals of Atomistic Methods Review Introduction Atomistic Methods DFT Kohn-Sham Pseudopotential Solving Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians Real space Conclusion Criteria Goals General (Generality) all atoms of the periodic table all kinds of bonds (transition state) Precision geometry atomic positions lengths of bonds angles of bonds energy of bonds vibration frequencies dipolar moments UV and visible spectra photo-emission magnetic moments System size Time scale Laboratoire de Simulation Atomistique ±0.001 angström ±0.001 angström ±1 degree 0.001 eV/atom (0.1 kJ/mol) 1 GHz 0.01 Debye (1 Debye = 3.336 × 10−30 C.m 0.01 eV 0.01 eV 0.01 µB 104 atoms for ab initio methods 106 atoms for parametrized methods 1s http://inac.cea.fr/L_Sim T. Deutsch Criteria for classification 3 criteria Review Introduction Atomistic Methods P G 1→ Generality (elements, alloys) DFT Kohn-Sham 2→ Precision (∆r , ∆E) Pseudopotential Solving 3→ System size (N, ∆t) Basis sets Plane waves S Direct Minimisation Diagonalisation Gaussians Real space Conclusion Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Quantum Mechanics Schrödinger equation for N electrons in an external potential Vext (r ): antisymmetric Ψ (r1 , r2 , . . . , rN ) = −Ψ (r2 , r1 , . . . , rN ) Review ~2 Introduction Atomistic Methods DFT H = − 2m Kohn-Sham N ∑ ∇2ri + i 1 ∑ 1 4πε0 i = 6 j |ri − rj | N + ∑ Vext (ri ) i Pseudopotential Solving Basis sets H Ψ (r1 , r2 , . . . , rN ) = E Ψ (r1 , r2 , . . . , rN ) Plane waves Direct Minimisation Diagonalisation Ion-electron interactions: Gaussians Real space Conclusion Vext (r ) = − 1 ∑ Zα 4πε0 α |Rα − r | Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Hartree-Fock and Post Hartree-Fock We build the most simple antisymmetric wavefunction for N electrons from N wavefunctions at 1 electron (Slater determinant): φ1 (x1 ) φ2 (x1 ) Φ (x1 , . . . , xN ) = .. . φN (x1 ) Review Introduction ... ... .. . ... .. . φN (xN ) φ1 (xN ) φ2 (xN ) The Hamiltonian are exact. Atomistic Methods DFT Kohn-Sham Pseudopotential Solving Basis sets Many Slater determinants: Configuration interaction, MP2, coupled clusters, . . . P G P G Plane waves Direct Minimisation Diagonalisation Gaussians Real space Conclusion S Laboratoire de Simulation Atomistique S http://inac.cea.fr/L_Sim T. Deutsch Hartree-Fock and Post Hartree-Fock We build the most simple antisymmetric wavefunction for N electrons from N wavefunctions at 1 electron (Slater determinant): φ1 (x1 ) φ2 (x1 ) Φ (x1 , . . . , xN ) = .. . φN (x1 ) Review Introduction ... ... .. . ... .. . φN (xN ) φ1 (xN ) φ2 (xN ) The Hamiltonian are exact. Atomistic Methods DFT Kohn-Sham Pseudopotential Solving Basis sets Many Slater determinants: Configuration interaction, MP2, coupled clusters, . . . P G P G Plane waves Direct Minimisation Diagonalisation Gaussians Real space Conclusion S Laboratoire de Simulation Atomistique S http://inac.cea.fr/L_Sim T. Deutsch Quantum Monte Carlo Very complicated wavefunctions for N electrons (Jastrow factor): φ1 (x1 ) .. Ψ (x1 , . . . , xN ) = . φN (x1 ) Review ... .. . ... φ1 (xN ) .. ∏ J (rij ) . φN (xN ) i 6=j Introduction Atomistic Methods DFT Integration of the total energy: E = Kohn-Sham < Ψ|H |Ψ > < Ψ|Ψ > Use of Monte-Carlo methods for the integration. Pseudopotential Solving Basis sets P G Plane waves Direct Minimisation Diagonalisation Gaussians Real space Conclusion S Laboratoire de Simulation Atomistique Diffusive Monte Carlo (fixed node, sign problem) http://inac.cea.fr/L_Sim T. Deutsch Density Functional Theory (DFT) Theorem: The total energy of the ground state in an external potential Vext is a functional of the electronic density ρ(r ): Z Vext (r )ρ(r )dr E [ρ(r )] = F [ρ(r )] + r Review Variational principles: E [ρ0 (r )] ≤ E [ρ(r )] ∀ ρ(r ) Introduction Atomistic Methods DFT Kohn-Sham Pseudopotential Solving Basis sets Plane waves Kohn-Sham Equation: (approximation) Having a one-electron hamiltonian in a mean field. P G 1 2 − 2 ∇ + Veff (r ) ψi = εi ψi Direct Minimisation Diagonalisation Gaussians Real space Conclusion occupied ρ(r ) = ∑ |ψi (r )|2 S i Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Density Functional Theory (DFT) Theorem: The total energy of the ground state in an external potential Vext is a functional of the electronic density ρ(r ): Z Vext (r )ρ(r )dr E [ρ(r )] = F [ρ(r )] + r Review Variational principles: E [ρ0 (r )] ≤ E [ρ(r )] ∀ ρ(r ) Introduction Atomistic Methods DFT Kohn-Sham Pseudopotential Solving Basis sets Plane waves Kohn-Sham Equation: (approximation) Having a one-electron hamiltonian in a mean field. P G 1 2 − 2 ∇ + Veff (r ) ψi = εi ψi Direct Minimisation Diagonalisation Gaussians Real space Conclusion occupied ρ(r ) = ∑ |ψi (r )|2 S i Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Tight Binding Methods Semi-empirical quantum methods We solve a one-electron parametrized hamiltonian (Slater-Koster parameter): H | ψi >= εi | ψi > Review Introduction Atomistic Methods DFT Kohn-Sham Pseudopotential Solving occupied Band term: Eband = εi P G ions Plane waves Direct Minimisation ∑ i Basis sets Repulsive term: Erep = Diagonalisation ∑ f (RI,J ) I ,J Gaussians Real space Conclusion Etotal = Erep + Eband S Mainly used for electronic properties (106 atoms) or transport Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Choice of Atomistic Methods Chemistry P G Review Introduction Atomistic Methods Compromise between accuracy, system size and CPU time. DFT Kohn-Sham Physics S S P G P G P G Force Fields Pseudopotential Solving Basis sets Plane waves Tight Binding S Hartree-Fock S Direct Minimisation Diagonalisation Ab initio (DFT) Real space Conclusion Configuration Interactions Quantum Monte-Carlo Laboratoire de Simulation Atomistique P G Gaussians http://inac.cea.fr/L_Sim S P G S T. Deutsch Choice of Atomistic Methods Chemistry P G Review Introduction Atomistic Methods Compromise between accuracy, system size and CPU time. DFT Kohn-Sham Physics S S P G P G P G Force Fields Pseudopotential Solving Basis sets Plane waves Tight Binding S Hartree-Fock S Direct Minimisation Diagonalisation Ab initio (DFT) Real space Conclusion Configuration Interactions Quantum Monte-Carlo Laboratoire de Simulation Atomistique P G Gaussians http://inac.cea.fr/L_Sim S P G S T. Deutsch Choice of Atomistic Methods Chemistry P G Review Introduction Atomistic Methods Compromise between accuracy, system size and CPU time. DFT Kohn-Sham Physics S S P G P G P G Force Fields Pseudopotential Solving Basis sets Plane waves Tight Binding S Hartree-Fock S Direct Minimisation Diagonalisation Ab initio (DFT) Real space Conclusion Configuration Interactions Quantum Monte-Carlo Laboratoire de Simulation Atomistique P G Gaussians http://inac.cea.fr/L_Sim S P G S T. Deutsch Choice of Atomistic Methods Chemistry P G Review Introduction Atomistic Methods Compromise between accuracy, system size and CPU time. DFT Kohn-Sham Physics S S P G P G P G Force Fields Pseudopotential Solving Basis sets Plane waves Tight Binding S Hartree-Fock S Direct Minimisation Diagonalisation Ab initio (DFT) Real space Conclusion Configuration Interactions Quantum Monte-Carlo Laboratoire de Simulation Atomistique P G Gaussians http://inac.cea.fr/L_Sim S P G S T. Deutsch Choice of Atomistic Methods Chemistry P G Review Introduction Atomistic Methods Compromise between accuracy, system size and CPU time. DFT Kohn-Sham Physics S S P G P G P G Force Fields Pseudopotential Solving Basis sets Plane waves Tight Binding S Hartree-Fock S Direct Minimisation Diagonalisation Ab initio (DFT) Real space Conclusion Configuration Interactions Quantum Monte-Carlo Laboratoire de Simulation Atomistique P G Gaussians http://inac.cea.fr/L_Sim S P G S T. Deutsch Choice of Atomistic Methods Chemistry P G Review Introduction Atomistic Methods Compromise between accuracy, system size and CPU time. DFT Kohn-Sham Physics S S P G P G P G Force Fields Pseudopotential Solving Basis sets Plane waves Tight Binding S Hartree-Fock S Direct Minimisation Diagonalisation Ab initio (DFT) Real space Conclusion Configuration Interactions Quantum Monte-Carlo Laboratoire de Simulation Atomistique P G Gaussians http://inac.cea.fr/L_Sim S P G S T. Deutsch Review Introduction Density Functional Theory Atomistic Methods DFT Kohn-Sham equations Kohn-Sham Pseudopotential Solving Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians Real space Conclusion Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Ab initio calculations with DFT Several advantages 4 Ab initio: No adjustable parameters 4 DFT: Quantum Review Introduction Atomistic Methods DFT Kohn-Sham mechanical (fundamental) treatment Main limitations 8 Approximated approach 8 Requires high computer power, limited to few hundreds atoms in most cases Wide range of applications: nanoscience, biology, materials Pseudopotential Solving Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians Real space Conclusion Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Quantum Mechanics Schrödinger equation for N electrons in an external potential Vext (r ): antisymmetric Ψ (r1 , r2 , . . . , rN ) = −Ψ (r2 , r1 , . . . , rN ) Review ~2 Introduction Atomistic Methods DFT H = − 2m Kohn-Sham N ∑ ∇2ri + i 1 ∑ 1 4πε0 i = 6 j |ri − rj | N + ∑ Vext (ri ) i Pseudopotential Solving Basis sets H Ψ (r1 , r2 , . . . , rN ) = E Ψ (r1 , r2 , . . . , rN ) Plane waves Direct Minimisation Diagonalisation Ion-electron interactions: Gaussians Real space Conclusion Vext (r ) = − 1 ∑ Zα 4πε0 α |Rα − r | Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch The Hohenberg-Kohn theorem Schrödinger equation N 1 2 1 H = ∑ − ∇ri + Vext (ri , {R }) + ∑ i =1 1 2 i 6=j |ri − rj | 2 Very difficult to solve for more than two electrons! Review Introduction Atomistic Methods DFT Kohn-Sham Pseudopotential Solving Basis sets The fundamental variable of the problem is however not the wavefunction, but the electronic density Z ρ(r ) = N dr2 · · · drN ψ∗ (r , r2 , · · · , rN )ψ(r , r2 , · · · , rN ) Plane waves Direct Minimisation Diagonalisation Gaussians Hohenberg-Kohn theorem (1964) Real space Conclusion The ground state density ρ(r ) of a many-electron system uniquely determines (up to a constant) the external potential . The external potential is a functional of ρ: Vext = Vext [ρ] Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch The Hohenberg-Kohn theorem Schrödinger equation N 1 2 1 H = ∑ − ∇ri + Vext (ri , {R }) + ∑ i =1 1 2 i 6=j |ri − rj | 2 Very difficult to solve for more than two electrons! Review Introduction Atomistic Methods DFT Kohn-Sham Pseudopotential Solving Basis sets The fundamental variable of the problem is however not the wavefunction, but the electronic density Z ρ(r ) = N dr2 · · · drN ψ∗ (r , r2 , · · · , rN )ψ(r , r2 , · · · , rN ) Plane waves Direct Minimisation Diagonalisation Gaussians Hohenberg-Kohn theorem (1964) Real space Conclusion The ground state density ρ(r ) of a many-electron system uniquely determines (up to a constant) the external potential . The external potential is a functional of ρ: Vext = Vext [ρ] Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Kohn-Sham formalism H-K theorem: E is an unknown functional of the density E = E [ρ] → Density Functional Theory Kohn-Sham approach Review Introduction Mapping of an interacting many-electron system into a system with independent particles moving into an effective potential. Atomistic Methods DFT Kohn-Sham Find a set of orthonormal orbitals Ψi (r) that minimizes: Pseudopotential Solving Basis sets Plane waves Direct Minimisation E =− 1 N Z ∗ 2 Ψi (r)∇ Ψi (r)dr + ∑ 2 i =1 1 Z 2 ρ(r)VH (r)dr Diagonalisation Z Gaussians + Exc [ρ(r)] + Real space Conclusion Vext (r)ρ(r)dr N ρ(r) = ∑ Ψ∗i (r)Ψi (r) ∇2 VH (r) = −4πρ(r) i =1 Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Kohn-Sham formalism H-K theorem: E is an unknown functional of the density E = E [ρ] → Density Functional Theory Kohn-Sham approach Review Introduction Mapping of an interacting many-electron system into a system with independent particles moving into an effective potential. Atomistic Methods DFT Kohn-Sham Find a set of orthonormal orbitals Ψi (r) that minimizes: Pseudopotential Solving Basis sets Plane waves Direct Minimisation E =− 1 N Z ∗ 2 Ψi (r)∇ Ψi (r)dr + ∑ 2 i =1 1 Z 2 ρ(r)VH (r)dr Diagonalisation Z Gaussians + Exc [ρ(r)] + Real space Conclusion Vext (r)ρ(r)dr N ρ(r) = ∑ Ψ∗i (r)Ψi (r) ∇2 VH (r) = −4πρ(r) i =1 Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Kohn-Sham formalism H-K theorem: E is an unknown functional of the density E = E [ρ] → Density Functional Theory Kohn-Sham approach Review Introduction Mapping of an interacting many-electron system into a system with independent particles moving into an effective potential. Atomistic Methods DFT Kohn-Sham Find a set of orthonormal orbitals Ψi (r) that minimizes: Pseudopotential Solving Basis sets Plane waves Direct Minimisation E =− 1 N Z ∗ 2 Ψi (r)∇ Ψi (r)dr + ∑ 2 i =1 1 Z 2 ρ(r)VH (r)dr Diagonalisation Z Gaussians + Exc [ρ(r)] + Real space Conclusion Vext (r)ρ(r)dr N ρ(r) = ∑ Ψ∗i (r)Ψi (r) ∇2 VH (r) = −4πρ(r) i =1 Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Performing a DFT calculation A self-consistent equation ρ(r) = ∑ Ψ∗i (r)Ψi (r), where |ψi i satisfies i 1 2 − ∇ + VH [ρ] + Vxc [ρ] + Vext + Vpseudo |ψi i = Ei |ψi i , 2 Review Introduction Atomistic Methods (Kohn-Sham) DFT “Ingredients” DFT Kohn-Sham Pseudopotential Solving Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians Real space Conclusion An XC potential, functional of the density several approximations exists (LDA,GGA,. . . ) A choice of the pseudopotential (if not all-electrons) (norm conserving, ultrasoft, PAW,. . . ) A basis set for expressing the |ψi i An (iterative) algorithm for finding the wavefunctions |ψi i A (good) computer. . . Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Exchange-Correlation Energy Exc [ρ] = Kexact [ρ] − KKS [ρ] + Ve−e exact [ρ] − VHartree [ρ] Review Introduction Atomistic Methods DFT Kohn-Sham Pseudopotential KKS [ρ]: Kinetic energy for a non-interaction electron gas DFT is an exact reformulation of the many-body problem Kohn-Sham formalism is exact (mapping of an interacting electron system into a non-interacting system Solving Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians Real space Conclusion Many types: Local Density Approximation (LDA) Generalized gradient approximation (GGA), meta-GGA, Hybrid functional (a part of exchange energy) Random Phase Approximation (RPA), . . . Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Exchange-Correlation Energy Exc [ρ] = Kexact [ρ] − KKS [ρ] + Ve−e exact [ρ] − VHartree [ρ] Review Introduction Atomistic Methods DFT Kohn-Sham Pseudopotential KKS [ρ]: Kinetic energy for a non-interaction electron gas DFT is an exact reformulation of the many-body problem Kohn-Sham formalism is exact (mapping of an interacting electron system into a non-interacting system Solving Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians Real space Conclusion Many types: Local Density Approximation (LDA) Generalized gradient approximation (GGA), meta-GGA, Hybrid functional (a part of exchange energy) Random Phase Approximation (RPA), . . . Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Local Density Approximation (LDA) Assumption: For each point r in space with a density ρ(r ), the exchange and correlation energy at that point is the same as for an uniform electron gas with that density: Z LDA ρ(r )εxc (ρ)dr = Exc V Review LDA (r ) Vxc = LDA ∂Exc ∂ρ(r ) = εxc (ρ) + ρ(r ) Introduction Atomistic Methods DFT Kohn-Sham ∂εxc (ρ) ∂ρ(r ) Pseudopotential Solving Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians Real space Conclusion εxc (ρ): Exchange-correlation energy for a uniform electron gas of 1/3 R LDA density (analytic): Exc = − 43 π3 ρ(r )4/3 dr Correlation energy (interpolated analytically): Based on Quantum Monte Carlo calculations (Ceperley and Alder (1980)) for the exchange and correlation of the electron gas. Based on uniform gas of electron: Good cancellation of error Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Generalized Gradient Approximations (GGA) Review Introduction Atomistic Methods Use the gradient of the density to have a more accurate exchange energy Many generalised gradient approximations: Perdew (1985), Lee, Yang and Parr (1988), Perdew and Wang (1991), PBE (1996),HCTH (1998), ... Perdew and Yue (1986) functional (which is relatively simple): 1/3 Z 3 3 GGA ρ(r )4/3 F (s)dr [ρ] = − Exc 4 π DFT Kohn-Sham Pseudopotential Solving s= |∇ρ(r )| 2kF (ρ(r )) kF = ((3π2 ρ(r ))1/3 Basis sets Plane waves Direct Minimisation F (s) = (1 + 1.296s2 + 14s4 + 0.2s6 )1/15 Diagonalisation Gaussians Real space Conclusion LDA: Overbind (energies too high) underestimate bond lengths GGA: In general better energies specially PBE Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Generalized Gradient Approximations (GGA) Review Introduction Atomistic Methods Use the gradient of the density to have a more accurate exchange energy Many generalised gradient approximations: Perdew (1985), Lee, Yang and Parr (1988), Perdew and Wang (1991), PBE (1996),HCTH (1998), ... Perdew and Yue (1986) functional (which is relatively simple): 1/3 Z 3 3 GGA ρ(r )4/3 F (s)dr [ρ] = − Exc 4 π DFT Kohn-Sham Pseudopotential Solving s= |∇ρ(r )| 2kF (ρ(r )) kF = ((3π2 ρ(r ))1/3 Basis sets Plane waves Direct Minimisation F (s) = (1 + 1.296s2 + 14s4 + 0.2s6 )1/15 Diagonalisation Gaussians Real space Conclusion LDA: Overbind (energies too high) underestimate bond lengths GGA: In general better energies specially PBE Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Hybrid functionals Add a small part of exchange interaction (25% to 50%) Review Introduction EExchange = − Atomistic Methods DFT Kohn-Sham φ∗i (r)φj (r)φi (r0 )φ∗j (r0 ) drdr0 |r − r0 | V Z ∑ i <j Many hybrid functionals: B3LYP, HSE, . . . Pseudopotential Solving The gap is (really) better and also cohesion energies Basis sets Plane waves Direct Minimisation Very time-consuming (10) Diagonalisation Gaussians Not the right physics and sometimes not so accurate. . . Real space Conclusion Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Hybrid functionals Add a small part of exchange interaction (25% to 50%) Review Introduction EExchange = − Atomistic Methods DFT Kohn-Sham φ∗i (r)φj (r)φi (r0 )φ∗j (r0 ) drdr0 |r − r0 | V Z ∑ i <j Many hybrid functionals: B3LYP, HSE, . . . Pseudopotential Solving The gap is (really) better and also cohesion energies Basis sets Plane waves Direct Minimisation Very time-consuming (10) Diagonalisation Gaussians Not the right physics and sometimes not so accurate. . . Real space Conclusion Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Performing a DFT calculation A self-consistent equation ρ(r) = ∑ Ψ∗i (r)Ψi (r), where |ψi i satisfies i 1 2 − ∇ + VH [ρ] + Vxc [ρ] + Vext + Vpseudo |ψi i = Ei |ψi i , 2 Review Introduction Atomistic Methods (Kohn-Sham) DFT “Ingredients” DFT Kohn-Sham Pseudopotential Solving Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians Real space Conclusion An XC potential, functional of the density several approximations exists (LDA,GGA,. . . ) A choice of the pseudopotential (if not all-electrons) (norm conserving, ultrasoft, PAW,. . . ) A basis set for expressing the |ψi i An (iterative) algorithm for finding the wavefunctions |ψi i A (good) computer. . . Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Pseudopotentials → smoothening of wavefunctions For chemical properties only the valence electrons are relevant: Eliminate the chemically inactive core electrons Review Vext (r ) = − 1 ∑ Zα 4πε0 α |Rα − r | Reduce the number of electron orbitals Introduction Atomistic Methods DFT Kohn-Sham Pseudopotential Solving Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians The pseudo-wavefunctions of the valence electrons are smooth Eliminate the rapid variations of the valence wavefunction in the core region Real space Conclusion A reasonable approximation The pseudopotential approximation is less severe than the approximate nature of the exchange correlation functional Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Non-Local Pseudopotentials V PP (r , r 0 ) = ∑ Ylm (r )Vl (r )δr ,r Ylm (r 0 ) 0 l ,m where Yl ,m are spherical harmonics. Review VNL = ∑ | φIlm > Vl <φIlm | I ,l ,m Introduction Atomistic Methods DFT where I is the index of atom, l and m and quantum numbers. Kohn-Sham Pseudopotential Solving Basis sets Plane waves Use ab initio atomic wavefunctions to obtain pseudopotentials. Direct Minimisation Diagonalisation Gaussians Real space Conclusion The (potential) energy contribution due to the valence charge density must be substracted away to give an ionic pseudopotential: Vlion ,ps (r ) = Vl ,ps (r ) + (VHartree [ρ(r )] + Vxc [ρ(r )])free atom Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Gaussian type separable Pseudopotentials (HGH) Local part Vloc (r ) −Zion = r r " 1 erf √ + exp − 2 2rloc ( C1 + C2 Review r rloc 2 + C3 r 2 # rloc 4 r + C4 rloc r 6 ) rloc Introduction Atomistic Methods DFT Nonlocal (separable) part H (~r ,~r 0 ) Kohn-Sham 2 Pseudopotential Solving H sep (~r ,~r 0 ) = Basis sets ∑ ∑ Ys,m (r̂ ) pis (r ) his pis (r 0 ) Ys∗,m (rˆ0 ) i =1 m Plane waves + Direct Minimisation Diagonalisation ∑ Yp,m (r̂ ) p1p (r ) h1p p1p (r 0 ) Yp∗,m (rˆ0 ) m Gaussians Real space Conclusion p1l (r ) = √ 2 rle l + 23 rl − 12 ( rr l )2 q Γ(l + 32 ) Laboratoire de Simulation Atomistique p2l (r ) http://inac.cea.fr/L_Sim = √ 2 r l +2 e l+ 7 rl 2 − 12 ( rr )2 l q Γ(l + 72 ) T. Deutsch Performing a DFT calculation A self-consistent equation ρ(r) = ∑ Ψ∗i (r)Ψi (r), where |ψi i satisfies i 1 2 − ∇ + VH [ρ] + Vxc [ρ] + Vext + Vpseudo |ψi i = Ei |ψi i , 2 Review Introduction Atomistic Methods (Kohn-Sham) DFT “Ingredients” DFT Kohn-Sham Pseudopotential Solving Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians Real space Conclusion An XC potential, functional of the density several approximations exists (LDA,GGA,. . . ) A choice of the pseudopotential (if not all-electrons) (norm conserving, ultrasoft, PAW,. . . ) A basis set for expressing the |ψi i An (iterative) algorithm for finding the wavefunctions |ψi i A (good) computer. . . Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Basis sets for electronic structure calculation How can we express the Kohn-Sham wavefunctions? Plane Waves 4 Localization in Fourier space, efficient preconditioning 4 Systematic convergence properties Review Introduction Atomistic Methods 8 No localization in real space. Empty regions must be “filled” with PW. Non adaptive DFT Kohn-Sham Pseudopotential Solving Basis sets Plane waves Gaussians, Slater type Orbitals 4 Real space localized, well suited for molecules and other open structures Direct Minimisation Diagonalisation Gaussians Real space Conclusion 4 Small number of basis functions for moderate accuracy 8 Many different recipes for generating basis sets 8 Over-completeness before convergence. Non systematic basis set. Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Basis sets for electronic structure? Wavelets A basis set both adaptive and systematic, real space based Review Introduction Atomistic Methods DFT Kohn-Sham Pseudopotential Wavelet basis sets Localized both in real and in Fourier space Solving Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians Allow for adaptivity (for internal electrons) Systematic basis set Real space Conclusion Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Performing a DFT calculation A self-consistent equation ρ(r) = ∑ Ψ∗i (r)Ψi (r), where |ψi i satisfies i 1 2 − ∇ + VH [ρ] + Vxc [ρ] + Vext + Vpseudo |ψi i = Ei |ψi i , 2 Review Introduction Atomistic Methods (Kohn-Sham) DFT “Ingredients” DFT Kohn-Sham Pseudopotential Solving Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians Real space Conclusion An XC potential, functional of the density several approximations exists (LDA,GGA,. . . ) A choice of the pseudopotential (if not all-electrons) (norm conserving, ultrasoft, PAW,. . . ) A basis set for expressing the |ψi i An (iterative) algorithm for finding the wavefunctions |ψi i A (good) computer. . . Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Kohn-Sham Equations: Operators Apply different operators Having a one-electron hamiltonian in a mean field. Review Introduction 1 2 − ∇ + Veff (r )(r ) ψi = εi ψi 2 Atomistic Methods Kohn-Sham Veff (r ) = Vext (r ) + Pseudopotential Solving Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians ρ(r 0 ) 0 dr + µxc (r ) V |r − r 0 | Z DFT E [ρ] can be expressed by the orthonormalized states of one particule: ψi (r ) with the fractional occupancy number fi ( 0 ≤ f i ≤ 1) : Real space Conclusion ρ(r ) = ∑ fi |ψi (r )|2 i Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Kohn-Sham Equations: Computing Energies Calculate different integrals E [ρ] = K [ρ] + U [ρ] Review K [ρ] = − Introduction Atomistic Methods 1 ~2 2 me Z ∑ i V dr ψ∗i ∇2 ψi DFT Kohn-Sham Pseudopotential Solving Basis sets Z U [ρ] = V dr Vext (r ) ρ(r )+ Direct Minimisation Gaussians Z 2 V | Plane waves Diagonalisation 1 ρ(r )ρ(r 0 ) + E [ρ] | xc{z } |r − r 0 | {z } exchange−correlation dr dr 0 Hartree We minimise with the variables ψi (r ) and fi Real space Conclusion Z with the constraint V Laboratoire de Simulation Atomistique dr ρ(r ) = Nel . http://inac.cea.fr/L_Sim T. Deutsch KS Equations: Self-Consistent Field − Set of self-consistent equations: 1 ~2 2 me 2 ∇ + Veff ψi = εi ψi with an effective potential: Review Introduction Atomistic Methods Veff (r ) = Vext (r ) + Z V DFT | Kohn-Sham ρ(r 0 ) |r − r 0 | {z } dr 0 Hartree + δExc δρ(r ) | {z } exchange−correlation Pseudopotential Solving Basis sets and: Plane waves Direct Minimisation ρ(r ) = ∑i fi |ψi (r )|2 Diagonalisation Gaussians Real space Conclusion Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch KS Equations: Self-Consistent Field − Set of self-consistent equations: 1 ~2 2 me 2 ∇ + Veff ψi = εi ψi with an effective potential: Review Introduction Atomistic Methods Veff (r ) = Vext (r ) + Z V DFT | Kohn-Sham ρ(r 0 ) |r − r 0 | {z } dr 0 Hartree + δExc δρ(r ) | {z } exchange−correlation Pseudopotential Solving Basis sets and: Plane waves Direct Minimisation ρ(r ) = ∑i fi |ψi (r )|2 Diagonalisation Gaussians Real space Conclusion Poisson Equation: ∆VHartree = ρ 2 2 2 (Laplacian: ∆ = ∂∂x 2 + ∂∂y 2 + ∂∂z 2 ) Real Mesh (1003 = 106 ): 106 × 106 = 1012 evaluations ! Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch KS Equations: Self-Consistent Field − Set of self-consistent equations: 1 ~2 2 me 2 ∇ + Veff ψi = εi ψi with an effective potential: Review Introduction Atomistic Methods Veff (r ) = Vext (r ) + Z V DFT | Kohn-Sham ρ(r 0 ) |r − r 0 | {z } dr 0 Hartree + δExc δρ(r ) | {z } exchange−correlation Pseudopotential Solving Basis sets and: Plane waves Direct Minimisation ρ(r ) = ∑i fi |ψi (r )|2 Diagonalisation Gaussians Real space Conclusion Poisson Equation: ∆VHartree = ρ 2 2 2 (Laplacian: ∆ = ∂∂x 2 + ∂∂y 2 + ∂∂z 2 ) Real Mesh (1003 = 106 ): 106 × 106 = 1012 evaluations ! Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch KS Equations: Self-Consistent Field − Set of self-consistent equations: 1 ~2 2 me 2 ∇ + Veff ψi = εi ψi with an effective potential: Review Introduction Atomistic Methods Veff (r ) = Vext (r ) + Z V DFT | Kohn-Sham ρ(r 0 ) |r − r 0 | {z } dr 0 Hartree + δExc δρ(r ) | {z } exchange−correlation Pseudopotential Solving Basis sets and: Plane waves Direct Minimisation ρ(r ) = ∑i fi |ψi (r )|2 Diagonalisation Gaussians Real space Conclusion Poisson Equation: ∆VHartree = ρ 2 2 2 (Laplacian: ∆ = ∂∂x 2 + ∂∂y 2 + ∂∂z 2 ) Real Mesh (1003 = 106 ): 106 × 106 = 1012 evaluations ! Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch KS Equations: Self-Consistent Field − Set of self-consistent equations: 1 ~2 2 me 2 ∇ + Veff ψi = εi ψi with an effective potential: Review Introduction Atomistic Methods Veff (r ) = Vext (r ) + Z V DFT | Kohn-Sham ρ(r 0 ) |r − r 0 | {z } dr 0 Hartree + δExc δρ(r ) | {z } exchange−correlation Pseudopotential Solving Basis sets and: Plane waves Direct Minimisation ρ(r ) = ∑i fi |ψi (r )|2 Diagonalisation Gaussians Real space Conclusion Poisson Equation: ∆VHartree = ρ 2 2 2 (Laplacian: ∆ = ∂∂x 2 + ∂∂y 2 + ∂∂z 2 ) Real Mesh (1003 = 106 ): 106 × 106 = 1012 evaluations ! Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch KS Equations: Self-Consistent Field − Set of self-consistent equations: 1 ~2 2 me 2 ∇ + Veff ψi = εi ψi with an effective potential: Review Introduction Atomistic Methods Veff (r ) = Vext (r ) + Z V DFT | Kohn-Sham ρ(r 0 ) |r − r 0 | {z } dr 0 Hartree + δExc δρ(r ) | {z } exchange−correlation Pseudopotential Solving Basis sets and: Plane waves Direct Minimisation ρ(r ) = ∑i fi |ψi (r )|2 Diagonalisation Gaussians Real space Conclusion Poisson Equation: ∆VHartree = ρ 2 2 2 (Laplacian: ∆ = ∂∂x 2 + ∂∂y 2 + ∂∂z 2 ) Real Mesh (1003 = 106 ): 106 × 106 = 1012 evaluations ! Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Ab Initio Methods: DFT Review Introduction Atomistic Methods DFT h − 1 2 ∇ +V (r ) 2 i +µxc (r ) ψki= εki ψki Kohn-Sham Pseudopotential Solving Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians Real space Conclusion Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Ab Initio Methods: DFT Review Introduction Atomistic Methods DFT h − 1 2 ∇ +V (r ) 2 i +µxc (r ) ψki= εki ψki Kohn-Sham Pseudopotential Solving Basis sets atomic orbitals Gaussians plane waves Slater augmented numerical Plane waves Direct Minimisation Diagonalisation Gaussians Real space Conclusion real space finite difference Wavelet Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Ab Initio Methods: DFT pseudopotential self-consistent (SCF) norm-conserving PAW all electrons Harris-Foulkes functional Review Introduction Atomistic Methods DFT h − 1 2 ∇ +V (r ) 2 i +µxc (r ) ψki= εki ψki Kohn-Sham Pseudopotential Solving Basis sets atomic orbitals Gaussians plane waves Slater augmented numerical Plane waves Direct Minimisation Diagonalisation Gaussians Real space Conclusion real space finite difference Wavelet Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Ab Initio Methods: DFT pseudopotential self-consistent (SCF) norm-conserving PAW all electrons Harris-Foulkes functional Review Introduction Atomistic Methods DFT Kohn-Sham Pseudopotential h − 1 2 ∇ +V (r ) 2 i +µxc (r ) ψki= εki ψki periodic Solving Basis sets non-periodic atomic orbitals Gaussians plane waves Slater augmented numerical Plane waves Direct Minimisation Diagonalisation Gaussians Real space Conclusion real space finite difference Wavelet Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Ab Initio Methods: DFT pseudopotential self-consistent (SCF) norm-conserving PAW all electrons Hybrid functionals Harris-Foulkes functional beyond LDA GW method LDA,GGA LDA+U Review Introduction Atomistic Methods DFT Kohn-Sham Pseudopotential h − 1 2 ∇ +V (r ) 2 i +µxc (r ) ψki= εki ψki periodic Solving Basis sets non-periodic atomic orbitals Gaussians plane waves Slater augmented numerical Plane waves Direct Minimisation Diagonalisation Gaussians Real space Conclusion real space finite difference Wavelet Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Ab Initio Methods: DFT pseudopotential self-consistent (SCF) norm-conserving PAW all electrons Hybrid functionals Harris-Foulkes functional beyond LDA GW LDA,GGA LDA+U Review Introduction Atomistic Methods DFT Kohn-Sham Pseudopotential h − 1 2 ∇ +V (r ) 2 i +µxc (r ) ψki= εki ψki periodic Solving Basis sets non-periodic atomic orbitals Gaussians plane waves Slater augmented numerical Plane waves Direct Minimisation Diagonalisation Gaussians Real space non-spin polarized Conclusion real space non-collinear finite difference spin polarized Wavelet collinear Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Ab Initio Methods: DFT pseudopotential self-consistent (SCF) norm-conserving PAW all electrons Hybrid functionals Harris-Foulkes functional relativistic non-relativistic beyond LDA GW LDA,GGA LDA+U Review Introduction Atomistic Methods DFT Kohn-Sham Pseudopotential h − 1 2 ∇ +V (r ) 2 i +µxc (r ) ψki= εki ψki periodic Solving Basis sets non-periodic atomic orbitals Gaussians plane waves Slater augmented numerical Plane waves Direct Minimisation Diagonalisation Gaussians Real space non-spin polarized Conclusion real space non-collinear finite difference spin polarized Wavelet collinear Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Ab Initio Methods: DFT pseudopotential self-consistent (SCF) norm-conserving PAW all electrons Hybrid functionals Harris-Foulkes functional relativistic non-relativistic beyond LDA GW LDA,GGA LDA+U Review Introduction Atomistic Methods DFT h − 1 2 ∇ +V (r ) 2 Kohn-Sham i +µxc (r ) ψki= εki ψki periodic Pseudopotential Solving Basis sets Plane waves Direct Minimisation non-periodic N 3 scaling atomic orbitals Gaussians plane waves Slater augmented numerical Diagonalisation Gaussians Real space O (N ) methods non-spin polarized Conclusion real space non-collinear finite difference spin polarized Wavelet collinear Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch List of ab initio Codes Plane Waves ABINIT — Louvain-la-Neuve — http://www.abinit.org CPMD — Zurich, Lugano — http://www.cpmd.org PWSCF — Italy — http://www.pwscf.org VASP — Vienna — http://cms.mpi.univie.ac.at/vasp Gaussian Review Introduction Atomistic Methods DFT Gaussian — http://www.gaussian.com DeMon — http://www.demon-software.com CP2K — http://cp2k.berlios.de Kohn-Sham Pseudopotential Solving Basis sets Plane waves Siesta — Madrid — http://www.uam.es/departamentos/ciencias/fismateriac/siesta (numerical basis sets) Direct Minimisation Diagonalisation Gaussians Real space Conclusion Wien — Vienna — http://www.wien2k.at (FPLAPW, all electrons) Real space basis set ONETEP — http://www.onetep.soton.ac.uk BigDFT — http://inac.cea.fr/L_Sim/BigDFT Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Basis sets Review Introduction Atomistic Methods DFT Minimisation of the electronic density Kohn-Sham Pseudopotential Solving Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians Real space Conclusion Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Plane Wave basis sets Natural basis set for electronic structure calculations in periodic solids. Review 1 Introduction √ exp(iG · r) Atomistic Methods V DFT Kohn-Sham Pseudopotential Solving Basis sets Plane waves Direct Minimisation Diagonalisation where V is the volume of the periodicity volume and G is a multiple of the reciprocal lattice vectors, i.e. there are integers l1 , l2 , l3 such that G = l1 b1 + l2 b2 + l3 b3 . Gaussians Real space Conclusion Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Plane Wave basis sets Matrix elements of the kinetic energy part of the Hamiltonian calculated analytically Z 1 2 exp(−iG · r) − ∇ exp(iG0 · r) = G2 δG,G0 2 Review Introduction Atomistic Methods DFT The potential energy part obtained numerically by calculating the wave-functions in real space Kohn-Sham Pseudopotential Ψ(r) = ∑ exp(iG · r) Solving G Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians Real space Conclusion and then integrating numerically in real space. The transformation on a grid in real space is done with the help of the Fast Fourier transformation at a cost of N log2 (N ) operations for a basis set of N plane waves. Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Basis sets: Planes Waves Operator approach: no need to set up Hamiltonian matrix FFT allows for quasi linear M log(M ) scaling where M is number of plane waves Review Introduction Atomistic Methods DFT Kohn-Sham Pseudopotential Solving Basis sets Plane waves Localization in Fourier space allows for efficient preconditioning techniques. Hence number of iterations independent of M Systematic convergence propertie No localization in real space. Empty regions have to be filled with plane waves. Only supercell approach. Direct Minimisation Diagonalisation Gaussians Real space Conclusion O(N) scaling hard to achieve Resolution cannot adaptively be refined around the nucleus. All electron calculations impossible, hard pseudopotentials difficult. Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Basis sets: Planes Waves Operator approach: no need to set up Hamiltonian matrix FFT allows for quasi linear M log(M ) scaling where M is number of plane waves Review Introduction Atomistic Methods DFT Kohn-Sham Pseudopotential Solving Basis sets Plane waves Localization in Fourier space allows for efficient preconditioning techniques. Hence number of iterations independent of M Systematic convergence propertie No localization in real space. Empty regions have to be filled with plane waves. Only supercell approach. Direct Minimisation Diagonalisation Gaussians Real space Conclusion O(N) scaling hard to achieve Resolution cannot adaptively be refined around the nucleus. All electron calculations impossible, hard pseudopotentials difficult. Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Basis sets: Planes Waves pseudopotential self-consistent (SCF) norm-conserving PAW all electrons Hybrid functionals Harris-Foulkes functional relativistic non-relativistic Review Introduction Atomistic Methods DFT h beyond LDA GW LDA,GGA LDA+U i 1 2 − 2 ∇ +V (r ) +µxc (r ) ψki= εki ψki periodic Kohn-Sham Pseudopotential Solving Basis sets Plane waves non-periodic N 3 scaling Direct Minimisation Diagonalisation Gaussians O (N ) methods Real space atomic orbitals Gaussians plane waves Slater augmented numerical non-spin polarized Conclusion real space non-collinear finite difference spin polarized Wavelet collinear Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Plane waves: List of Codes ABINIT — Louvain-la-Neuve — http://www.abinit.org PAW, linear response, GW, many features, good // Review Introduction Atomistic Methods DFT Kohn-Sham Pseudopotential Solving Basis sets Plane waves Direct Minimisation CPMD — Zurich, Lugano — http://www.cpmd.org USP, good //, QM/MM with Gromacs PWSCF — Italy — http://www.pwscf.org USP, linear response, // VASP — Vienna — http://cms.mpi.univie.ac.at/vasp robust, fast, hybrid functional, GW, // Diagonalisation Gaussians Real space Conclusion Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Direct Minimisation: Flowchart ( ) ψj = ∑ j cG eiG.r G Basis (NG ): 0, . . . , G Orthonormalized Review Introduction Atomistic Methods DFT Kohn-Sham Pseudopotential Solving Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians Real space Conclusion Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Direct Minimisation: Flowchart ( ) ψj = ∑ j cG eiG.r Basis (NG ): 0, . . . , G G Orthonormalized inv FFT ρ(r ) = ∑ fj |ψj (r )|2 Review j Basis (23 NG ): 0, . . . , 2G Introduction Atomistic Methods DFT Kohn-Sham Pseudopotential Solving Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians Real space Conclusion Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Direct Minimisation: Flowchart ( ) ψj = ∑ j cG eiG.r Basis (NG ): 0, . . . , G G Orthonormalized inv FFT ρ(r ) = ∑ fj |ψj (r )|2 j Review Basis (23 NG ): 0, . . . , 2G Introduction Atomistic Methods FFT DFT Kohn-Sham Pseudopotential Solving Vxc [ρ(r )] −G2 VH (G) = ρ(G) VNL ({ψj }) Veffective Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians Real space Conclusion Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Direct Minimisation: Flowchart ( ) ψj = ∑ j cG eiG.r Basis (NG ): 0, . . . , G G Orthonormalized inv FFT ρ(r ) = ∑ fj |ψj (r )|2 j Review Basis (23 NG ): 0, . . . , 2G Introduction Atomistic Methods FFT DFT Kohn-Sham Pseudopotential Vxc [ρ(r )] Solving −G2 VH (G) = ρ(G) VNL ({ψj }) Veffective Basis sets Plane waves Direct Minimisation Diagonalisation 1 2 j 2 G cG Kinetic Term Gaussians Real space Conclusion Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Direct Minimisation: Flowchart ( ) ψj = ∑ j cG eiG.r Basis (NG ): 0, . . . , G G Orthonormalized inv FFT ρ(r ) = ∑ fj |ψj (r )|2 j Review Basis (23 NG ): 0, . . . , 2G Introduction Atomistic Methods FFT DFT Kohn-Sham Pseudopotential Vxc [ρ(r )] Solving −G2 VH (G) = ρ(G) VNL ({ψj }) Veffective Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians Real space 1 2 j 2 G cG FFT δcGj = − ∂Etotal + Λjl cGl ∂cj∗ (G) ∑ l Conclusion Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim Kinetic Term Λjl =< ψj |H |ψl > T. Deutsch Direct Minimisation: Flowchart ( ) ψj = ∑ j cG eiG.r Basis (NG ): 0, . . . , G G Orthonormalized inv FFT ρ(r ) = ∑ fj |ψj (r )|2 j Review Basis (23 NG ): 0, . . . , 2G Introduction Atomistic Methods FFT DFT Kohn-Sham Pseudopotential Vxc [ρ(r )] Solving −G2 VH (G) = ρ(G) VNL ({ψj }) Veffective Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians Real space 1 2 j 2 G cG FFT δcGj = − Kinetic Term ∂Etotal + Λjl cGl ∂cj∗ (G) ∑ l Λjl =< ψj |H |ψl > Conclusion new ,j cG = j j cG + hstep δcG Laboratoire de Simulation Atomistique Steepest Descent, Conjugate Gradient, Direct Inversion of the Iterative Subspace http://inac.cea.fr/L_Sim T. Deutsch Direct Minimisation: Flowchart ( ) ψj = ∑ j cG eiG.r Basis (NG ): 0, . . . , G G Orthonormalized inv FFT ρ(r ) = ∑ fj |ψj (r )|2 j Review Basis (23 NG ): 0, . . . , 2G Introduction Atomistic Methods FFT DFT Kohn-Sham Pseudopotential −G2 VH (G) = ρ(G) Vxc [ρ(r )] Solving VNL ({ψj }) Veffective Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians Real space Conclusion 1 2 j 2 G cG FFT δcGj = − Kinetic Term ∂Etotal + Λjl cGl ∂cj∗ (G) ∑ l Λjl =< ψj |H |ψl > j Stop when δcG small new ,j cG = j j cG + hstep δcG Laboratoire de Simulation Atomistique Steepest Descent, Conjugate Gradient, Direct Inversion of the Iterative Subspace http://inac.cea.fr/L_Sim T. Deutsch Direct Minimization: List of Codes Plane Waves CPMD — Zurich, Lugano — http://www.cpmd.org Review Introduction Atomistic Methods DFT Kohn-Sham Pseudopotential Solving Basis sets Plane waves Gaussian CP2K — http://cp2k.berlios.de Real space basis set ONETEP — http://www.onetep.soton.ac.uk PARATEC — http://www.nersc.gov/projects/paratec BigDFT — http://inac.cea.fr/L_Sim/BigDFT Direct Minimisation Diagonalisation Gaussians Real space Conclusion Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Diagonalisation Scheme: Flowchart ) ( ψj = ∑ j cG eiG.r , {fj } G Basis (NG ): 0, . . . , G Orthonormalized Review Introduction Atomistic Methods DFT Kohn-Sham Pseudopotential Solving Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians Real space Conclusion Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Diagonalisation Scheme: Flowchart ) ( ψj = ∑ j cG eiG.r , {fj } G Basis (NG ): 0, . . . , G Orthonormalized inv FFT ρ(r ) = ∑ fj |ψj (r )|2 j Review Basis (23 NG ): 0, . . . , 2G Introduction Atomistic Methods DFT Kohn-Sham Pseudopotential Solving Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians Real space Conclusion Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Diagonalisation Scheme: Flowchart ) ( ψj = ∑ j cG eiG.r , {fj } Basis (NG ): 0, . . . , G G Orthonormalized inv FFT ρ(r ) = ∑ fj |ψj (r )|2 j Review Basis (23 NG ): 0, . . . , 2G Introduction Atomistic Methods FFT DFT Kohn-Sham Pseudopotential Solving Vxc [ρ(r )] −G2 VH (G) = ρ(G) VNL ({ψj }) Veffective Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians Real space Conclusion Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Diagonalisation Scheme: Flowchart ) ( ψj = ∑ j cG eiG.r , {fj } Basis (NG ): 0, . . . , G G Orthonormalized inv FFT ρ(r ) = ∑ fj |ψj (r )|2 j Review Basis (23 NG ): 0, . . . , 2G Introduction Atomistic Methods FFT DFT Kohn-Sham Pseudopotential Vxc [ρ(r )] Solving −G2 VH (G) = ρ(G) VNL ({ψj }) Veffective Basis sets Plane waves Direct Minimisation Diagonalisation 1 2 j 2 G cG Kinetic Term Gaussians Real space Conclusion Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Diagonalisation Scheme: Flowchart ) ( ψj = ∑ j cG eiG.r , {fj } Basis (NG ): 0, . . . , G G Orthonormalized inv FFT ρ(r ) = ∑ fj |ψj (r )|2 j Review Basis (23 NG ): 0, . . . , 2G Introduction Atomistic Methods FFT DFT Kohn-Sham Pseudopotential Vxc [ρ(r )] Solving −G2 VH (G) = ρ(G) VNL ({ψj }) Veffective Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians Real space 1 2 j 2 G cG Kinetic Term FFT 1 2 − 2 ∇ + Veff (r ) ψj = εj ψj Conclusion Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Diagonalisation Scheme: Flowchart ) ( ψj = ∑ j cG eiG.r , {fj } Basis (NG ): 0, . . . , G G Orthonormalized inv FFT ρ(r ) = ∑ fj |ψj (r )|2 Basis (23 NG ): 0, . . . , 2G j Review Introduction Atomistic Methods FFT DFT Kohn-Sham Pseudopotential Vxc [ρ(r )] Solving −G2 VH (G) = ρ(G) VNL ({ψj }) Veffective Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians Real space 1 2 j 2 G cG Kinetic Term FFT 1 2 − 2 ∇ + Veff (r ) ψj = εj ψj Conclusion ρnew (r ) = αρout + (1 − α)ρin Laboratoire de Simulation Atomistique ρout (r ) = ∑ fj |ψj (r )|2 http://inac.cea.fr/L_Sim Mixing Density: Anderson, Broyden,DIIS j T. Deutsch Diagonalisation Scheme: Flowchart ) ( ψj = ∑ j cG eiG.r , {fj } Basis (NG ): 0, . . . , G G Orthonormalized inv FFT ρ(r ) = ∑ fj |ψj (r )|2 Basis (23 NG ): 0, . . . , 2G j Review Introduction Atomistic Methods FFT DFT Kohn-Sham Pseudopotential Vxc [ρ(r )] Solving −G2 VH (G) = ρ(G) VNL ({ψj }) Veffective Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians Real space 1 2 j 2 G cG Kinetic Term FFT 1 2 − 2 ∇ + Veff (r ) ψj = εj ψj Conclusion Stop when |ρnew − ρin | small ρnew (r ) = αρout + (1 − α)ρin Laboratoire de Simulation Atomistique ρout (r ) = ∑ fj |ψj (r )|2 http://inac.cea.fr/L_Sim Mixing Density: Anderson, Broyden,DIIS j T. Deutsch Iterative diagonalization Standard diagonalization routines, such as found in software packages like LAPACK, are designed to find all the eigenvalues and eigenvectors of a matrix. Review Scale like n3 , where n is the dimension of the matrix Introduction Atomistic Methods DFT Kohn-Sham Pseudopotential Solving Basis sets Plane waves We need to calculate only a small number of eigenvalues and eigenvectors of matrices of a very large dimension, another type of diagonalization, namely iterative diagonalization is used: Direct Minimisation Diagonalisation Gaussians Real space Conclusion Lanczos, Davidson, RM-DIIS, Conjugate Gradient, LOBPCG, ... Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Diagonalization scheme: List of Codes ABINIT — Louvain-la-Neuve — http://www.abinit.org Diagonalization: Conjugate-Gradient, LOBPCG Mixing: Broyden, Pulay, Anderson, others based on dielectric constant Review Introduction CPMD — Zurich, Lugano — http://www.cpmd.org Atomistic Methods DFT Kohn-Sham Diagonalization: Lanczos, Davidson Mixing: Broyden, Pulay, Anderson Pseudopotential Solving PWSCF — Italy — http://www.pwscf.org Basis sets Plane waves VASP — Vienna — http://cms.mpi.univie.ac.at/vasp Direct Minimisation Diagonalisation Gaussians Diagonalization: RM-DIIS Real space Conclusion BigDFT — http://inac.cea.fr/L_Sim/BigDFT Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Gaussians Atom centered Gaussians are the most popular basis functions for electronic structure calculations in the quantum chemistry community. Review Introduction Atomistic Methods DFT (x − Xi )l1 (y − Yi )l2 (z − Zi )l3 × exp −α((x − Xi )2 + (y − Yi )2 + (z − Zi )2 ) Kohn-Sham Pseudopotential Solving Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians Real space Conclusion The different functions associated to Ri = (Xi , Yi , Zi ) differ by their extent, specified by α and by the exponents l1 , l2 , l3 . Because the Gaussians have qualitatively the shape of atomic orbitals one needs for moderate accuracy only a small number of basis function per atom (typically 10 to 20 per atom). Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Basis sets: Gaussians Real space localization makes them well suited for molecules and other open structures. Review Introduction Atomistic Methods DFT Kohn-Sham Pseudopotential Kinetic and overlap matrix elements can be calculated analytically. Good description of core electrons. Small number of basis functions necessary for moderate accuracy Solving Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians No systematic convergence. Over-completeness before convergence. This leads also to problems with O(N). Many different recipes for generating basis sets. Real space Conclusion Pulay forces have been implemented. Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Basis sets: Gaussians Real space localization makes them well suited for molecules and other open structures. Review Introduction Atomistic Methods DFT Kohn-Sham Pseudopotential Kinetic and overlap matrix elements can be calculated analytically. Good description of core electrons. Small number of basis functions necessary for moderate accuracy Solving Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians No systematic convergence. Over-completeness before convergence. This leads also to problems with O(N). Many different recipes for generating basis sets. Real space Conclusion Pulay forces have been implemented. Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Basis sets: Gaussians pseudopotential self-consistent (SCF) norm-conserving PAW all electrons Hybrid functionals Harris-Foulkes functional relativistic non-relativistic Review Introduction Atomistic Methods DFT h beyond LDA GW LDA,GGA LDA+U i 1 2 k k k − 2 ∇ +V (r ) +µxc (r ) ψ= i εi S ψi periodic Kohn-Sham Pseudopotential Solving Basis sets Plane waves non-periodic N 3 scaling Direct Minimisation Diagonalisation Gaussians O (N ) methods Real space atomic orbitals Gaussians plane waves Slater augmented numerical non-spin polarized Conclusion real space non-collinear finite difference spin polarized Wavelet collinear Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Real space basis set: wavelets (BigDFT) pseudopotential self-consistent (SCF) norm-conserving PAW all electrons Hybrid functionals Harris-Foulkes functional relativistic non-relativistic Review Introduction Atomistic Methods DFT h beyond LDA GW method LDA,GGA LDA+U i 1 2 − 2 ∇ +V (r ) +Vxc (r ) ψki= εki ψki periodic Kohn-Sham Pseudopotential Solving Basis sets Plane waves non-periodic N 3 scaling Direct Minimisation Diagonalisation Gaussians O (N ) methods Real space atomic orbitals Gaussians plane waves Slater augmented numerical non-spin polarized Conclusion real space non-collinear finite difference spin polarized Wavelet collinear Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Real space: List of Codes ONETEP — http://www.onetep.soton.ac.uk sinc function, O(N) method, inside Material Studio, good // Review Introduction BigDFT — http://inac.cea.fr/L_Sim/BigDFT wavelets, good // Atomistic Methods DFT Kohn-Sham Pseudopotential Solving Basis sets Plane waves PARATEC — http://www.nersc.gov/projects/paratec finite difference, good // OCTOPUS —http://www.tddft.org finite difference, TD-DFT Direct Minimisation Diagonalisation Gaussians Real space Conclusion GPAW — https://wiki.fysik.dtu.dk/gpaw Grid-based projector-augmented wave method, finite difference Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Real space: List of Codes ONETEP — http://www.onetep.soton.ac.uk sinc function, O(N) method, inside Material Studio, good // Review Introduction BigDFT — http://inac.cea.fr/L_Sim/BigDFT wavelets, good // Atomistic Methods DFT Kohn-Sham Pseudopotential Solving Basis sets Plane waves PARATEC — http://www.nersc.gov/projects/paratec finite difference, good // OCTOPUS —http://www.tddft.org finite difference, TD-DFT Direct Minimisation Diagonalisation Gaussians Real space Conclusion GPAW — https://wiki.fysik.dtu.dk/gpaw Grid-based projector-augmented wave method, finite difference Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Real space: List of Codes ONETEP — http://www.onetep.soton.ac.uk sinc function, O(N) method, inside Material Studio, good // Review Introduction BigDFT — http://inac.cea.fr/L_Sim/BigDFT wavelets, good // Atomistic Methods DFT Kohn-Sham Pseudopotential Solving Basis sets Plane waves PARATEC — http://www.nersc.gov/projects/paratec finite difference, good // OCTOPUS —http://www.tddft.org finite difference, TD-DFT Direct Minimisation Diagonalisation Gaussians Real space Conclusion GPAW — https://wiki.fysik.dtu.dk/gpaw Grid-based projector-augmented wave method, finite difference Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Real space: List of Codes ONETEP — http://www.onetep.soton.ac.uk sinc function, O(N) method, inside Material Studio, good // Review Introduction BigDFT — http://inac.cea.fr/L_Sim/BigDFT wavelets, good // Atomistic Methods DFT Kohn-Sham Pseudopotential Solving Basis sets Plane waves PARATEC — http://www.nersc.gov/projects/paratec finite difference, good // OCTOPUS —http://www.tddft.org finite difference, TD-DFT Direct Minimisation Diagonalisation Gaussians Real space Conclusion GPAW — https://wiki.fysik.dtu.dk/gpaw Grid-based projector-augmented wave method, finite difference Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Real space: List of Codes ONETEP — http://www.onetep.soton.ac.uk sinc function, O(N) method, inside Material Studio, good // Review Introduction BigDFT — http://inac.cea.fr/L_Sim/BigDFT wavelets, good // Atomistic Methods DFT Kohn-Sham Pseudopotential Solving Basis sets Plane waves PARATEC — http://www.nersc.gov/projects/paratec finite difference, good // OCTOPUS —http://www.tddft.org finite difference, TD-DFT Direct Minimisation Diagonalisation Gaussians Real space Conclusion GPAW — https://wiki.fysik.dtu.dk/gpaw Grid-based projector-augmented wave method, finite difference Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Review Introduction Conclusion Atomistic Methods DFT Kohn-Sham Pseudopotential Solving Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians Real space Conclusion Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch Conclusion Functionals: Mixed between physics and chemistry approach QM/MM (Quantum Mechanics/Molecular Modeling) Review Introduction Atomistic Methods DFT Kohn-Sham Pseudopotential Solving Basis sets Plane waves Direct Minimisation Diagonalisation Gaussians Real space Multi-scale approach (more than 1000 atoms feasible for a better parametrization) Order N (real space basis set as wavelets) Numerical experience (high performance computing): One-day simulation Better exploration of atomic configurations Molecular Dynamics (4s per step for 32 water molecules) Conclusion Laboratoire de Simulation Atomistique http://inac.cea.fr/L_Sim T. Deutsch
Similar documents
Milking Shorthorn Lots - Ayrshire Breeders Association
res. intermediate champion, 2015 EASTERN NATIONAL AT HARRISBURG selling dry BRED 7/30/15, due 5/6/16 TO MUDSLINGER Breeder: ROWE, JOHN STUART Current Owner: HALPIN, M&G AND M. MACKINSON OF HALPIN F...
More information