Real-time approaches to x-ray and optical spectra
J. J. Rehr, F. Vila, J. Kas
Department of Physics University of Washington
Seattle, WA USA
Nordita School on Photon-Matter Interaction Stockholm, Sweden Oct 3-7, 2016
Goal:
Real-time linear and non-linear responseTalk:
•
I. Linear and Nonlinear Optical Response RT-TDDFT•
II. Real-time core-level XAS•
III. Many-body Effects4
Real-time Approaches for Optical and X-ray Spectra
I. Real-Space & Real-Time Linear and Non-linear Optical Response
● Difficulty: frequency-space
is computationally demanding too-many excited states
•
Solution: RT-TDDFT - extension of SIESTA**Sanchez-Portal, Tsolakidis, and Martin, Phys. Rev. B66, 235416 (2002)
5
Real-Time-TDDFT*
6
*K. Yabana and G. F. Bertsch, Phys. Rev. B 54, 4484 1996.
Real time Linear Response
Induced Dipole Moment
Linear Response Function
Optical Absorption
Linear Dielectric Function
11
RT-TDDFT Formalism
•
Yabana and Bertsch Phys. Rev. B54, 4484 (1996)• Direct numerical integration of TD Kohn-Sham equations
• The response to external field is determined by applying a time-dependent electric field ΔH(t) = −E(t)·x.
• Optical properties determined from total dipole moment:
Sometimes more EFFICIENT than Frequency space methods 8
Numerical Real-time Evolution
•
Ground state density ρ0, overlap matrix S, and H(t) at each time-step evaluated with SIESTA•
Crank-Nicholson time-evolution: unitary, time-reversible Stable for long time-steps !•
Adiabatic GGA exchange-correlation (PBE) functionalCoefficients of Orbitals
10 _
_
, t = t + _ Δ t/2
Time (fs) Dipole p z(t) (a.u.)
•
Delta Function(Unit Impulse at t=0)
•
Step Function(Turn-off Constant E at t=0)
Example: CO Linear Response
pz(t) response due to applied Ez(t)
Time (fs) Energy (eV)
Im α(ω) Re α(ω)
12 E(t)
E(t)
0 0
Ground state without field
Ground state with constant field
Evolution for t>0
Evolution for t>0
Real time Nonlinear Response
•
The nonlinear expansion in field strength•
Accounting for time lag in system responseHow can we invert the equation to get
nonlinear response function? 15
?
Extraction of Dynamic Nonlinear Polarizabilities
• Set E
j(t) = F(t)E
j, and define expansion
pi(E)where p(1) yields linear response, p(2) first non-linear quadratic response, ….
• Quadratic response χ
(2)is then
17
Time (fs) Time (fs)
p(1
) ij
p(2
) ijk
Time (fs)
Frequency (eV) Frequency (eV)
Re F(ω) Im F(ω)
Time (fs)
Dynamic Nonlinear Response with Quasi-monochromatic Field F
δ(t)
•
Sine wave enveloped by another sine wave or GaussianSHG
OR
F(t)
Linear and Nonlinear response of CO
18
Example pNA: Nonlinear SHG
•
Comparison with other methodsEnergy (eV) β k(-2ω,ω,ω) (au)
Expt.
25
PBE
II. Real-time core-level XAS
jÃ(0)i = djbi
*
¹(!) = 1
¼Re
Z 1
0
dt ei!tGc(t)hÃ(t)jÃ(0)iµ(! + ²c ¡ EF): (1)
Time-correlation function approach
XANES with time-dependent DFT
Goal:
Time-dependent x-ray response Include core hole dynamics
Why use a real-time approach?
New experimental pulsed sources (XFEL, LCLS) Pump-probe experiments
Increased interest in time-dependent (TD) response
RTXS: The cartoon view
Atom GS
CH PP Screened CH
SCF Init TD
PAW
RTXS equations
XAS Absorption (FGR, ΔSCF,
FSR)
Core Hole Green’s Function Autocorrelation Function FT
Crank-Nicolson
Physical interpretation
Projected density of states p-DOS
given by autocorrelation function for Seed state of p-symmetry:
½
Ã(!) = ¡
¼jÃj1 2jIm R
10
dt e
i!thÃ(t)jÃ(0)i
Connection with Fermi golden rule agrees in limit t ∞
G(!) = R
10
dt e
i!tU (t; 0)
¹(!) = ¡ ¼1 ImhcjdyG(E)djciµ(E ¡ EF )
G(E) = [E ¡ H + i¡]¡1
¹(!) = P
k jhcjdjkij2±¡(! + ²c ¡ ²k)µ(E ¡ EF )
Check:
C K-edge of CO
Example: C K-edge XES of Benzene
Expt: Fister et al., Phys. Rev. B 75, 174106 (2007)
C K-Edge XAS of Diamond (C47H60 cluster)
G
+c(t) = e
i²cte
C(t)µ(t)
Cumulant expansion for core-hole Green’s function*
III. Many-body effects:
Intrinsic losses
*D. Langreth Phys Rev B 1, 471 (1970)
Particle-hole cumulant for XAS*
* cf. L. Campbell, L. Hedin, J. J. Rehr, and W.
Bardyszewski, Phys. Rev. B 65, 064107 (2002)
All losses in particle-hole spectral function AK
NiO
Intrinsic losses: real-time TDDFT cumulant satellites
Langreth cumulant in time-domain*
TiO2
*D. C. Langreth, Phys. Rev. B 1, 471 (1970)
Real-space interpretation: RT-TDDFT cumulant explains intrinsic excitations in TiO2
RT TDDFT Cumulant
Theory vs XPS
Interpretation: satellites arise from oscillatory charge density fluctuations between ligand and metal at frequency ~ ωCT due to turned-on core-hole
Charge transfer fluctuations
ωct
XPS
F. Fossard, K. Gilmore, G. Hug, J J. Kas, J J Rehr, E L Shirley and F D Vila
NIST Preprint; submitted to PRB 2016
RT-TDDFT cumulant Particle-hole cumulant
X-ray Edge Singularities
Low energy particle-hole excitations in cumulant explain edge singularities in XPS and XAS of metals
Excitation spectrum
Question: Does the cumulant method
work for correlated systems ?
Hedin’s answer * MAYBE
“Calculation similar to core case … but with more complicated fluctuation potentials …
… not question of principle, but of computational work...”
* L. Hedin, J. Phys.: Condens. Matter 11, R489 (1999)
Vn → -Im ε-1(ωn,qn)
Ce L3 XAS of CeO2
Spectral function
Spectral weights
Particle-hole cumulant for CeO
2Ce 5s XPS of CeO2
Real-time approach for x-ray Debye-Waller factors
Conclusions
Efficient RT-TDDFT approach for frequency dependent nonlinear optical response –
Accuracy comparable to frequency-domain methods for small systems; also applicable to large systems
Similar real-time approaches can be applied to dynamic structure, Debye-Waller factors , etc.
35
Acknowledgments:
Supported by DOE BSE DE-FG02-97ER45623
Thanks to
J.J. Kas L. Reining G. Bertsch
J. Vinson K. Gilmore L. Campbell T. Fujikawa F. Vila E. Shirley
S. Story S. Biermann M Guzzo M. Verstraete J. Sky Zhou C. Draxl et al.
& especially the ETSF
Conclusions
Particle-hole cumulant theory yields reasonable approximation for inelastic losses in XPS & XAS
All losses (intrinsic, extrinsic and interference) in
spectral function AK(ω) – can be added ex post facto
Interference terms explain mysteries in amplitudes
and energy dependence: adiabatic- sudden transition Theory also applicable to some d- and f-systems.
Many-body amplitudes S
02( ω) in XAS
•
Many-body XAS ≈ Convolution•
Explains crossover: adiabatic S02(ω) = 1to
sudden
transition S02(ω) ≈ 0.9|gq |2= |gqext |2 + | gqintrin |2 - 2 gqext gqintrin
≈ μ
qp( ω) S
02( ω)
Interference reduces loss!
Dynamic core-hole screening algorithm
Also: Grebennikov, Babanov and Sololov, Phys. Stat. Sol. 79, 423 (1977) and Privalov, Gel’mukhanov & Agren: Phys Rev. B 64, 165115 (2001)
alá Nozieres & De Dominicis
Extrinsic losses and Interference
Satellite strengths
XAS of Al
Particle-hole cumulant explains cancellation of extrinsic and intrinsic losses at threshold and
crossover: adiabatic
to sudden approximation
Extension to:
Extrinsic and Intrinsic losses
Question: How to extend theory to real-time approach?
Energy Dependent Spectral Function A(k,ω) Quasi-boson Model
Interference: Quasi-Boson Approach*
Excitations - plasmons, electron-hole pairs ... are bosons
*W. Bardyszewski and L. Hedin, Physica Scripta 32, 439 (1985)
“
GW++” Same ingredients as GW self-energy Vn → -Im ε-1(ωn,qn) fluctuation potentialsMany-body Model: |e- , h , bosons
>
Many-pole Self-energy Algorithm
*Plasmon-pole model many-pole model
-Im ε-1(ω) Many-pole Dielectric Function
~
Σ i gi δ(ω - ω i)Many-pole GW self-energy Σ(E)
* J. Kas et al. PRB 76, 195116(2008)
Effective GW++ Green’s Function g
eff( ω)
Damped qp Green’s function
Extrinsic + Intrinsic
-
2 x Interference geff(ω)=Spectral function: A(ω) = -(1/π) Im
g
eff( ω)
L. Campbell, L. Hedin, J. J. Rehr, and W. Bardyszewski, Phys. Rev. B 65, 064107 (2002)
+ - -