lectures on applications of tddft · 2019. 10. 2. · n r ] n 1 −[ exc[n] n r ] n o 1/n...
TRANSCRIPT
Angel Rubio
Benasque, September 2004
Dpto. de Física de Materiales, Universidad del País Vasco, Donostia International Physics Center (DIPC), and Centro Mixto CSIC-UPV/EHU, Donostia, Spain
http://dipc.ehu.es/arubio E-mail: [email protected]
Lectures on Applications of TDDFT
Motivation
I. Illustration of the physics for nano- and bio-structures
II. Extended systems: problems and new developments
Moore's Law
Evolution towards Nanoscience
Nanotechnology
Paradigm Shift
Current View Think Different
Classical continuum physics Quantum mechanics
Solid state properties Binding properties
Volume dominating Surfaces dominating
Homogeneous Materials Composite-inhomogeneous
Statistical ensemble Individual particles
Miniaturization Self-organization/Assembly
Optical properties of nanostructures
Nanoscience: different scales need to be bridged (multicomponent simulations)
THEORY and EXPERIMENTS on the same object
- spatial scale: 0D - 1D - 2D from molecule-nano-bio-meso-wires-leads
- time scale: femtosecond (electron excitations) picoseconds (phonon) nanosecond (charge-transfer) millisecond >>> bio-structural organization growth; self assembly
FIRST PRINCIPLES TOOLS: SPECTROSCOPIES
MotivationMotivation
Ab-initio approach for non-equilibrium electron dynamics
e-
e-
e-
excited states
Ground state
transport
1. Optical spectroscopy. Photoelectron and time-resolved spectroscopies2. Lifetimes: Electron-electron and electron-phonon coupling (adiabatic/nonadiabatic)4. Photo-induced dynamics (Ultrafast dynamics): e.g. Photoisomerisation in bio-photoreceptors
hv
Time-dependent approach: TDDFT?
Electronic coupling between a solid and an adsorbate governs chemical dynamics at surfaces
The dynamics of electronic excitations play an important role in molecule-surface interactions and reactivity, as in laser-driven surface reactions, and are also critical to technological applications of electronic materials
Photochemistry: Phonon- versus electron- mediated surface reactions:
HOMO
LUMO
Adsorbate
Vacuum level
probe photon
pump photon
EF
K. Yamanouchi, Science 295, 1659 (2002)
High-intense short-laser (fs) pulses
Quantum optimally controlled landscapes: time resolved dynamics
Motivation: non-linear phenomena and chemical reactivity
R. J. Levis et al, Science 292, 709 (2001)
Multiphoton <1014 Wcm2 Tunneling<1015 Wcm2
The spatial resolution depends on the wavelength, the spatial localization (STM, SNOM), the inelastic attenuation (PES, PD), the range of the interaction (EELS, MARPE)
1 10 100 10000,1
1
10
100
1000
10000
energía (eV)
LEED
EXAFS, NEXAFSPES, PD, Auger
SNOMphotonic materials
IMFP
electrons
fotones
energy (eV)
wav
elen
gth
(Å)
longitud de onda del electrón (A)
EELS
STEM
STM
NEXAFS
PES, PD,MARPE
o0,1 1 10
10
100
1000
10000
phot
on w
avel
engt
h (Å
)
electron wavelength (Å)
Characterisation tools of the “nanoworld“ Need of the knowledge of response functions!!!!!
See for a review: G. Onida, L. Reining and AR, Rev. Mod. Phys. 74, 601 (2002)
Electronic properties of a many body system: ab-initio approach
Ground state
density n(r)yes, LDA, GGA's
n(r,t) and/or j(r,t)
G1, total energy
G2
yes
Excited states
Corrections toKS eigenvalues?
Yes, non-linear phenomena and time-resolved spectroscopies
Quasiparticle excitationsIP and EAoptical absorption
yes, in some cases!
Scheme
DFT
TDDFT
MBPT
QM
Problems
XC-func.
(r,r') (w)dependence
Comput.
Time Dependent Density Functional Theory (Runge and Gross 1984)
Electron-dynamics first, then e-ion problem
HK-like theorem: v(r,t) <------> � (r,t)The time dependent density determines uniquely the time-dependent external potential and therefore all physical observables
Kohn-Sham formalism:The time dependent density of the interacting system can be calculated as the density of an auxiliary non-interacting system
Same time-dependent
density n(r,t)
i ℏd
dt=H i ℏ
d
dt
i=H
KS [ { j}]i, i=1,⋯N ?
HKS=ℏ2
2 mi∇−
e
c ℏAA
xc
2
Vexternal
Vhartree
Vexchange
Vcorrelation
M.A.L. Marques, A. Castro, G. Bertsch, AR Comp.Phys.Comm. (2002)C. Rozzi, M.A.L. Marques, A. Castro, E.K.U. Gross A. R. (to be published)
http://www.tddft.org/programs/octopus
The octopus project is aim to the first principle description of the excite state electron-ion dynamics of nanostructures and extended systems within TDDFTImplementation:✔ Numerical description of functions: real space discretization (1D-3D).✔ Auxiliary use of LCAO/ FFTs. QM/MM for biomolecular structures✔ Electon-ion coupling: pseudopotentials. Spin-Orbit , non-collinear magnetism, periodic systems .........
Linear Response Applications
Linear optical response: general aspects0r ,r ', w=∑ij
f j− f ii
c r jr ir ' jc r '
i− j−w
[1−v f xc 0]=0
Single-pole approximation
=0
0v f
xc
K ij=∫ dr ir d V xc r d r
jr
– LDA- thick solid line– EXX- dotted line– Bethe-Salpeter – Experiments
Transport: Current + E-field
Optical response: (Benzene)
7eV
� (HOMO)
M.A.L Marques, A. Castro, G.F. Bertsch and AR, Comp. Phys. Comm. (2002);K. Yabana and G. F. Bertsch, Int. J. Quantum Chem. 75, 55 (1999)
� * (LUMO)
Linear optical response: Geometry identification
discrimination of C20 isomers through optical spectroscopy
A. Castro, M. A. L. Marques, J. A. Alonso, G. F. Bertsch, K. Yabana and AR, J. Chem. Phys. 116, 1930 (2002).
• Real-space, real-time TDLDA yields reliable photoabsorption spectra of carbon-clusters.
• Photoabsorption spectra of C20 isomers are significantly different.
• Optical spectroscopy (absorption and emission) is proposed as an experimental tool to identify the cluster structure.
Optical spectroscopy of gold clusters: relativistic effects.
A "poor" condensed matter physicist view !!!!
Biological molecules: photoreceptors
QM/MM + TDDFT approach
M.A.L Marques, X. Lopez, D. Varsano, A. Castro, and A. R. Phys. Rev. Lett. 90, 158101 (2003)
F. Gai et al. Science 279, 1886 (1998)
QM/MM approach
Green Fluorescent Protein (GFP).
(Aequorea victoria: jellyfish)
Towards understanding biomolecular colors: the GFP case
T.M.H. Creemers et al, Proc. Natl. Acad. Sci.. USA (1999)
Guanine Cytosine
Absorption
Circulardichroism
ℜ j E∝∫0
∞dt ei Ei t L j t
RE=ℑℜE
rotational strength function
ℜE=ℜxℜ
yℜ
z;
Optical Rotatory Power
Azobenzene: spectroscopy along femtosecond-laser induced photoisomerization
S. Spörlein et al, Proc. Natl. Acad. Sci., 99, 7998 (2002)
HOMO
LUMO
T. Hugel et al, Science, 296, 1103 (2002) Y. Yu et al, Nature 425, 145 82003)
Azobenzene dyes are known to isomerize at the central N-N bond within fractions of picoseconds at high quantum yield [T. Nägele et al, Chem. Phys. Lett. 272, 489 (1997)].
One example is the APB optical trigger: Single-Molecule Optomechanical Cycle
Excited state electron-ion dynamics
CIS
Next step: QM/MM calculation of the chromophore+peptide system
TRANS
Azobenzene: spectroscopy along femtosecond-laser induced photoisomerization
Introduction: standard simulation of excited state dynamics?Introduction: standard simulation of excited state dynamics?
No
Observation of the nonradiative decay!lifetime, decay path
Yes Do MD.
Hellmann-Feynman theorem works
1. No need of level assignment for the hole and excited electron
2. Automatic monitoring of the nonradiative decay (lifetime, decay path).
|g>
Reaction coordinate
Pot
entia
l
t = 0: Promote the electronic occupations to mimic the excited states. Then perform a SCF calculation
t > 0: Solve � n(t+� t)=exp{- i � tH(t)} � n(t).
|e>
Is the matrix of HKohn-Sham
diagonal?
Our Method: TDDFT✔TDDFT: N-coupled one-particle equations.
i ℏ ∂∂t=H i ℏ ∂
∂t
i=H
KS [ { j}]i, i=1,⋯N ?
HKS=ℏ2
2 mi∇−
e
c ℏA
2
Vexternal
Vhartree
Vexchange
Vcorrelation
Gross et al (1985)-to date
✔Classical description of electromagnetic field. (Absence of radiative-decay channel!)
✔Classical description of nuclei (point particles): Ehrenfest path:
✔Solution of the TD-KS equations by unitary propagation schemes
Fat =−⟨t ∣∇ a
H∣t ⟩ t =Det {it }
it t =e
−iHKSt t t /2
e−iH
KSt t /2
it
Femtosecond dynamics: test photodissociation of a dimer (Na2
+).
80fs
A. Castro, M.L. Marques. J.A. Alonso, G.F. Bertsch and AR (2003)
ω=3.2eV
ω=2.5eV
Time resolved Vibrational Spectroscopy: Raman
Femtosecond dynamics (ongoing work!!!!)
Time resolved Vibrational Spectroscopy: Raman & IR.
Na-dimer
Photodissociation dynamics: case of He3
+
J. Chem. Phys. 103, 3450 (1995).
Photodissociation dynamics: case of He3
+
Linear Optical Response. Laser Pulse: I = 9 1011 W/cm2
=5eVI=1.310
11W cm
2
Femtosecond dynamics: Time-resolved photoelectron spectroscopy
D. Varsano, M.A.L. Marques, H. Appel, E.K.U Gross and AR (to be published)
-The simulation region is divided in two parts: A and B, separated by a smooth mask function- Electrons are not allowed to come back from B to A
We write the photoelectron spectrum as:
PES p=∑i ∣i
p , t∞∣2
The ultimate biomedical goal of The ultimate biomedical goal of nanotechnology:nanotechnology:
Nanotechnology: a higher form of evolution? (Humility is perhaps appropriate…)
II. Optical absorption and electron energy loss spectroscopy of extended systems
Benasque, September 2004
G. Onida, L. Reining and AR, Rev. Mod. Phys. 74, 601 (2002)
Introduction: how to handle the electron dynamics in extended systems under the influence of an external electromagnetic field?
TDDFT: - Problems with standard exchange-correlation functionals- A new fxc derived from Many-body perturbation theory
proper description of excitonic effects!!!- Applications to poliacetilene as one-dimensional system
Some Pathologies of the KS-DFT functional
Band-gap problemis discontinuous by a constant when an electron is added to the system(Sham, Schlüter and Perdew, Levy (1983)
Widely separated open-shell atomsAlmbladh and von Barth 1985
Exchange correlation electric field in insulators
Godby and Sham 1984, Gonze, Ghosez and Godby 1995
Some Pathologies of the KS-DFT functionalSome Pathologies of the KS-DFT functionalBan-gap problem(Sham, Schlüter and Perdew, Levy (1983)
Egap=N1KS −N
KSN1KS N1−N1
KS ≡EgapKSxc
Egap=I−A=LUMO−HOM O
xc=[Exc [n ]nr ]N1
−[Exc [n ]nr ]NO1 /N xc-discontinuity
Some Pathologies of the KS-DFT functional
Band-gap problemis discontinuous by a constant when an electron is added to the system(Sham, Schlüter and Perdew, Levy (1983)
Widely separated open-shell atomsAlmbladh and von Barth 1985
Exchange correlation electric field in insulators
Godby and Sham 1984, Gonze, Ghosez and Godby 1995
Some Pathologies of the KS-DFT functionalSome Pathologies of the KS-DFT functionalWidely separated open-shell atoms
Almbladh and von Barth 1985
The KS-effective potential takes a positive value IB-I
A around atom B
Some Pathologies of the KS-DFT functional
Band-gap problemis discontinuous by a constant when an electron is added to the system(Sham, Schlüter and Perdew, Levy (1983)
Widely separated open-shell atomsAlmbladh and von Barth 1985
Exchange correlation electric field in insulators
Godby and Sham 1984, Gonze, Ghosez and Godby 1995
Some Pathologies of the KS-DFT functionalSome Pathologies of the KS-DFT functional
Hedin Equation's (1965)
The GW ''soup''
G0W0 Band Structures of Insulators
From "Quasiparticle calculations in solids", W.G. Aulbur, L. Jönsson and J.W. Wilkins, Solid State Physics 54 1 (2000),
also available in preprint form at http://www.physics.ohio-state.edu/~wilkins/vita/publications.html#reviews
EnQ P≃n
KSn∣nKS−vxc∣n
Time-dependent approach for extended systems: a gauge formalismG.F. Bertsch, J.I. Iwata, AR, K. Yabana, PRB62, 7998 (2000)
E t =−1
c
d Ad t
The Hamiltonian of a periodic system in a volume V under a uniform field is
The equation of motion are:
H=∑i
i∣
1
2 mp
e
cA
2
Vion∣
iE
HartreeE
xc
V
8c2 d Adt
2
i ℏ∂∂t
i=[ 1
2 mp
e
cA
2
VionV
HV
xc ]i
d2 A
dt 2=−4e2 n
mA−4c
e
V∑
i
i∣pm∣
i
For the electric field is:d Edt=−4j j=
−e
V∑
i
i∣pm∣
i−
e2
cn Aand
Gonze,Ghosez,Godby; PRL74, 4035 (95)
Lithium Diamond
?
See for a review: G. Onida, L. Reining and AR, Rev. Mod. Phys. 74, 601 (2002)
Very encouraging results for Si already:
●TD-CDF: results of P. de Boeij (using Vignale,Kohn functional)
●TD-EXX: results of A. Goerling
●What about insulators, bound-excitons, etc...?
Non-local fxc for extended systems:
Why a non-local (static?) fxc for extended systems:
fxc=−/q2
=0
0v f
xc
- In the EELS spectra fxc is added to the full coulomb
that already contains a long range contribution
The lack of a long range term in fxc
LDA is much important
(crucial) in the absorption spectra than in the EELS!!!
EELS−1=1v
L. Reining, V. Olevano, AR, G. Onida, PRL88, 0664041 (2002); S. Botti et al, PRB (2004).
=0
G W0
G Wv fxc
M
RPA≡1 /[1−v0]G=G '=0
- In the absorption spectra fxc is added to the full coulomb
that does not contains a long range contribution (q=0)
M=1−v
Density Functional versus Manybody perturbation theory
Diagrammatic expansion
Density Functional Theory and Many-Body Perturbation Theory
Exchange-correlation Potential: Real, Local in space, Frequency independentDensity Functional Theory
Self-Energy: Complex, Non-local in space, Frequency dependent
Many-Body Perturbation Theory
F. Aryasetiawan, Rep. Prog. Phys. 61, 237-312 (1998)
R. O. Jones and O. Gunnarsson, Rev. Mod. Phys. 61, 689 (1989)
G. Onida, L. Reining and AR, Rev. Mod. Phys. 74, 601 (2002)
Many-Body approach to the Exchange-Correlation Kernel of TDDFT
A. Marini, R. Del Sole and AR, PRL (2003)
Hypothesis
It exists a ''many-body xc-kernel'' such that the TDDFT and Many-Body polarization
functions are identical
Consequently TDDFT equation can be used as an equation for the xc-kernel and as a formal solution can be found in terms of an iterative equation for
the nth order contribution
A diagrammatic approach
Many-Body approach to the Exchange-Correlation Kernel of TDDFT
TDDFT
MBPT
Bethe-Salpeter Equation
Iterative equation for fxc
A. Marini, R. Del Sole and AR, PRL (2003)
=
BSETDDFT Experiment
TDDFT, scalar fxc
Bound excitons in TDDFT
A. Marini, R. Del Sole and AR, PRL (2003)
=
-2=
BSE QP-RPA
C
How many terms ?
=
C
A many-body causal TDDFT kernel
Causal/T-ordered
Causal/T-ordered fxc
BSETDDFT 1st orderTDDFT 2nd order
Same agreement between BSE and TDDFT for the finite transferred momentum
absorption spectra...
...and for the off-diagonal elements of the microscopic dielectric function
Is TDDFT ''fast'' compared to the BSE ?
=
[# of frequencies] ×100 x10000
×W10000×10000
× '10000×100
When the only optical spectra is calculated TDDFT is as time consuming as BSE...
...but when the full dielectric matrix is needed TDDFT is more favorable than BSE
Low dimensional systems (1D): polyacetylane
M. van Faassen et al. PRL 88 186401 (2002) S.J.A. Van Gisbergen PRL 83 694 (1999)
Electric field dependence of the XC Potential in Molecular Chains
In LDA and GGA xc potential lack of a term counteracting the applied electric field
Low dimensional systems (1D): polyacetylane
fxc
BSE r , r' ,Isolated infinite Polyacetylene chain
G. Onida, L. Reining and AR, Rev. Mod. Phys. 74, 601 (2002)
What about the description of decaying quasiparticle processes
within TDDFT?
Lifetime of quasiparticles
interactions between quasiparticles limit how long the corresponding quantum states retain their identity, i.e., the lifetime
of the excitation. In combination with the velocity, this lifetime determines the mean free path, a measure of influence of the excitation
h
Importance of lifetime
- surface photochemistry- electron transfer across interfaces- electron dynamics and energy transfer
- screening in an electron gas- electron-phonon coupling- localization
from F.Reinert et al., PRB 63 (2001) 115415.
Experimental lifetimes change quickly with time!
Heimann et al. 1977
Kevan et al. 1987
Paniago et al. 1995
Nicolay et al. 2000
binding energy [meV]
Excitonic effects (via TDDFT) on the lifetimes of LiF
10'000x10'000 BS kernel
Up to 200 G-vectors dielectric
function, 256 Q-points in the
whole BZ
Small broadening stability
The fxc kernel remains stable and robust even with a 10 meV broadening and energies up to 20 eV
A BS-based calculation is, in this case,enormously less convenient than using TDDFT
Linear behavior Small penetration of
the RPA ''forbidden region''
www.tddft.org/programs/octopus
Summary
Ongoing work on applications to:
Time-resolved spectroscopies: pump-probe simulations.
High-harmonic generation from quantum dots.
QM/MM: for excited-state dynamics in biomolecules: photoisomerisation
Control chemical reactivity:Pulse optimization in laser induced reaction
Non-linear effects in One-dimensional systems
Time-dependent Molecular transport !!!!!!!!!
Quantum nuclei, etc......
TDDFT is a powerful tool to handle the combined dynamics of electron/ion in response to external electromagnetic fields of nanostructures, biological
molecules and extended systems.
Problem: we need better fxc functionals based on either DFT (or current-DFT) or MBPT approaches
Acknowledgments
A. Castro, A. Marini, X. López, L Wirtz, D. Varsano Department of Material Physics, Centro Mixto CSIC-UPV, University of the Basque
Country, and Donostia International Physics Center (DIPC), Donostia, Spain
L. Reining, V. Olevano
G.F. Bertsch
École Polytechnique, Palaiseau, France
Physics Department and Institute for Nuclear Theory University of Washinton, Seattle (USA)
R. Del Sole and G. OnidaINFM e Dipartimento di Fisica dell'Università di Roma ``Tor Vergata'', Roma, Italy
M. A. L. Marques, C.A. Rozzi, and E. K. U. GrossInstitut für Theoretische Physik, Freie Universität, Berlin, Germany
S.G. Louie and M.L. CohenDepartment of Physics, University of California at Berkeley, USA
Thank you
http://dipc.ehu.es/arubio E-mail: [email protected]
It is one of the first duties of a professor, in any
subject, to exaggerate a little both the importance of his
subject and his own importance in it.
G.H. Hardy (A Mathematician's Apology)
The 2-point vertex functionSuppose to have a good approximation fot the (TD)DFT potential...
PRL 91, 056402 (2003);PRL 91, 256402 (2003). PRL 88, 066404 (2002) etc etc
No difference with GoWo using ALDA or similar approaches PRL 62, 2718 (1989); PRB 49, 8024 (1994); PRB 56, 12832 (1997).
BUT IS ?
=
The 3-point vertex function
If
Closed expression for the ''on mass-shell'' electronic lifetime