Linear response theory and TDDFT Claudio Attaccalite http://abineel.grenoble.cnrs.fr/ CECAM Yambo School 2013 (Lausanne) Motivations: +- hν Absorption Spectroscopy Many Body Effects!!! Motivations(II):Absorption Spectroscopy Absorption linearly related to the Imaginary part of the MACROSCOPIC dielectric constant (frequency dependent) Outline Response of the system to a perturbation → Linear Response Regime How can we calculate the response of the system? IP, local field effects and Time Dependent DFT Some applications and recent steps forward Conclusions Spectroscopy From Maxwell equation to the response function Materials equations: From Gauss's law: D(r ,t )=ϵ 0 E (r , t)+ P (r , t) ∇⋅E (r , t)=4 π ρtot (r , t) ∇⋅D(r ,t )=4 π ρext (r , t) Electric Field Electric Displacement Polarization In general: P(r ,t )=∫ χ (t−t ' ,r , r ') E (t ' r ' )dt ' dr ' +∫ dt 1 dt 2 χ 2 (...) E (t 1 ) E (t 2)+O ( E 3 ) For a small perturbation we consider only the first term, the linear response regime P(r ,t )=∫ χ (t −t ' ,r , r ') E (t ' r ') dt ' dr ' In Fourier space: P(ω)=ϵ0 χ (ω) E (ω)=ϵ0 (ϵ(ω)−1) E (ω) D(ω)=ϵ0 ϵ(ω) E (ω) Response Functions Moving from Maxwell equation to linear response theory we define D(ω) δ V ext (ω) ϵ(ω)= = ϵ0 E (ω) δ V tot (ω) δ V tot (ω) ϵ (ω)= δ V ext (ω) −1 where V tot (⃗r , t )=V ext (⃗r ,t )+V ind (⃗r ,t ' ) The induced charge density results in a total potential via the Poisson equation. V tot (⃗r t )=V ext (⃗r t )+∫ dt ' ∫ d ⃗r v (⃗r −⃗r )ρind (⃗r t ) ' δ ρind ϵ (ω)=1+ v δV ext δρind ϵ(ω)=1−v δ V tot −1 ' ' ' Our goal is to calculate the derivatives of the induced density respect to the external potential The Kubo formula 1/2 H =H 0 + H ext (t)=H 0 +∫ d r ρ(r) V ext (r , t) We star from the time-dependent Schroedinger equation: ∂ψ i =[ H 0 + H ext (t)] ψ(t) ∂t ...and search for a solution as product of the solution for Ho plus an another function (interaction representation)... iH t ̃ ψ(t)=e ψ(t ) 0 ̃ ) iH t ∂ ψ(t −iH t ̃ ext (t) ψ(t ̃ )= H ̃ ) i =e H ext (t)e ψ(t ∂t 0 0 ...and we can write a formal solution as: t −i ∫t H̃ext (t)dt ̃ ψ(t)=e 0 ̃ 0) ψ(t The Kubo formula 2/2 For a weak perturbation we can expand: t −i ∫t H̃ext (t)dt ̃ ψ(t)=e 0 1 t ̃ 0 )=[1+ ∫t dt ' H̃ext (t ' )+O ( H̃2ext )] ψ(t ̃ 0) ψ(t i 0 And now we can calculate the induced density: t ̃ )∣ρ(t) ̃ )〉≈〈ρ〉0 −i ∫t 〈[ρ(t), H ext (t ' )]〉+O( H 2ext ) ρ(t )=〈 ψ(t ̃ ∣ψ(t 0 ...and finally... t ρind (t )=−i ∫t 0 ρind (t )=ρ(t)−ρ0 ∫ dr 〈[ρ(r , t ),ρ(r ' t ')]〉 V ext (r ' , t ') Kubo Formula (1957) ind δ ρ (r , t ) ' ' χ (⃗r t ,⃗r t )= =−i 〈[δρ(r , t)δρ(r ' t ')]〉 δ V ext (r ' , t ' ) The linear response to a perturbation is independent on the perturbation and depends only on the properties of the sample How to calculated the dielectric constant We want to calculate: ind δ ρ (r , t ) ' ' χ (⃗r t ,⃗r t )= =−i 〈[δ ρ(r ,t )δ ρ(r ' t ' )]〉 δ V ext (r ' , t ' ) We expand X in an independent particle basis set χ (⃗r t ,⃗r ' t ' )= ∑ i , j,l ,m k χ i , j , l , m , k ϕi , k (r )ϕ∗j , k (r ) ϕl , k (r ' )ϕ∗m , k (r ') ∂ ρ̂ i , j , k χi , j , l , m , k= ∂ V l ,m ,k Quantum Theory of the Dielectric Constant in Real Solids Adler Phys. Rev. 126, 413–420 (1962) The Von Neumann equation (see Wiki http://en.wikipedia.org/wiki/Density_matrix) ∂ ρ̂ k (t ) eff i =[ H k +V , ρ̂ k ] ∂t ρ̂ k (t )=∑i f (ϵk , i )∣ψi , k 〉 〈 ψi , k∣ What is Veff ? Independent Particle Independent Particle ∂ ρ̂ i , j , k χi , j , l , m , k= ∂ V l,m ,k ∂ ρ̂ k (t ) eff i =[ H k +V , ρ̂ k ] ∂t Using: Veff = Vext ∂ i ∂ρi , j , k = ∂ [ H +V eff , ρ̂ ] k k i, j,k eff eff ∂t ∂ V l ,m,k ∂ V l ,m, k { H i, j ,k = ρ̂ i , j , k δi , j ϵi (k) ∂ ρ̂ k eff = δi , j f (ϵi , k )+ ⋅V +.... eff ∂V And Fourier transform respect to t-t', we get: f (ϵi , k )−f (ϵ j , k ) χ i , j , l , m , k (ω)= δ j ,l δi , m ℏ ω−ϵ j , k + ϵi , k +i η Optical Absorption: IP δ ρNI =χ 0 δ V tot 0 χ =∑ ij ϕi (r) ϕ*j (r) ϕ*i (r ' ) ϕj (r ' ) ω−(ϵi −ϵ j )+ i η Hartree, Hartree-Fock, dft... Non Interacting System Absorption by independent Kohn-Sham particles =ℑ χ 0 =∑ ∣〈 j∣D∣i〉∣2 δ(ω−(ϵ j − ϵi )) ij 2 8 π '' 2 ϵ (ω)= 2 ∑ ∣〈ϕi∣e⋅v̂ ∣ϕ j 〉∣ δ(ϵi −ϵ j −ℏ ω) ω i, j Particles are interacting! Time-dependent Hartree (local fields) Time-dependent Hartree (local fields effects) The induced charge density results in a total potential via the Veff = Vext + VH V tot r t =V ext r t ∫ dt ' ∫ d r ' v r −r ' ind r ' t ' Poisson equation. r ,t r ,t V tot r ' ' ,t ' ' r , r ' , t−t ' = = V ext r ' ,t ' V tot r ' ' ,t ' ' V ext r ' , t ' χ (⃗r t ,⃗r ' t ' )=χ 0 (⃗r t ,⃗r ' t ' )+∫ dt 1 dt 2∫ d r⃗1 d r⃗2 χ 0 (⃗r t , r⃗1 t 1 ) v ( r⃗1 −r⃗2) χ ( r⃗2 t 2 ,⃗r ' t ' ) ind Screening of the V ind V tot external perturbation ' 0 r ,r = ind r , t V tot r ' t ' Time-dependent Hartree (local fields) PRB 72 153310(2005) Macroscopic Perturbation.... δ ρind ϵ (ω)=1+ v δV ext δρind ϵ(ω)=1−v δ V tot −1 Which is the right equation? ...microscopic observables Not correct!! Macroscopic averages 1/3 In a periodic medium every function V(r) can be represented by the Fourier series: V (r)=∫ dq ∑G V (q+G) ei(q+G)r or V (r)=∫ dq V (q , r) eiqr =∫ dq ∑G V (q+G) ei(q+G )r iGr V (q ,r )= V (q+G)e ∑G Where: The G components describe the oscillation in the cell while the q components the oscillation larger then L Macroscopic averages 2/4 Macroscopic averages 3/4 Macroscopic averages 4/4 The external fields is macroscopic, only components G=0 Macroscopic averages and local fields If you want the macroscopic δ ρind ϵ (ω)=1+ v δV ext then invert the dielectric constant δρind ϵ(ω)=1−v δ V tot response use the first equation and Local fields are not enough.... −1 What is missing? Two particle excitations, what is missing? electron-hole interaction, exchange, higher order effects...... The DFT and TDDFT way DFT versus TDDFT DFT versus TDDFT Time Dependent DFT [ ] −1 2 ∇ +V eff (r , t ) ψi (r , t)=i ∂ ψi (r ,t ) 2 ∂t V eff (r ,t )=V H (r , t)+ V xc (r , t)+ V ext (r , t) Interacting System Petersilka et al. Int. J. Quantum Chem. 80, 584 (1996) I = V ext NI 0= V eff ... by using ... I = NI V ext = 0 V ext V H V xc V H V xc = 1 V ext V ext 0 v Non Interacting System TDDFT is an exact f xc theory for neutral excitations! q ,= 0 q , 0 q , vf xc q , q , Time Dependent DFT Choice of the xc-functional ...with a good xc-functional you can get the right spectra!!! Summary ● How to calculate linear response in solids and molecules ● The local fields effects: time-dependent Hartree ● Correlation problem: TD-Hartree is not enough! ● Correlation effects can be included by mean of TDDFT 29 References!!! On the web: ● ● ● ● http://yambo-code.org/lectures.php http://freescience.info/manybody.php http://freescience.info/tddft.php http://freescience.info/spectroscopy.php Electronic excitations: density-functional versus many-body Green's-function approaches RMP, vol 74, pg 601, (2002 ) G. Onida, L. Reining, and A. Rubio Books: