GW and Bethe-Salpeter Equation Approach to Spectroscopic Properties Steven G. Louie Department of Physics, University of California at Berkeley and Materials Sciences Division, Lawrence Berkeley National Laboratory Supported by : National Science Foundation U.S. Department of Energy
70
Embed
GW and Bethe-Salpeter Equation Approach to Spectroscopic ...nano-bio.ehu.es/.../GW_and_Bethe-Salpeter...Louie.pdf · Steven G. Louie Department of Physics, University of California
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
GW and Bethe-Salpeter Equation Approach toSpectroscopic Properties
Steven G. Louie
Department of Physics, University of California at Berkeleyand
Materials Sciences Division, Lawrence Berkeley National Laboratory
Supported by: National Science FoundationU.S. Department of Energy
First-Principles Study of Material Properties
+
= iGW
Fermi sea
Fermi sea
(excitonic)
Content
• Quasiparticle excitations
- The GW approximation- Applications to solids, surfaces and nanostructures
• Excitons, optical response, and forces in the excited state
- The Bethe-Salpeter Equation- Applications to crystals, surfaces, nanotubes, self-
trapped excitons
• Some more-correlated systems
Quasiparticle Excitations
Kohn-Sham Eigenvalues QP Energies
One simple example: the Homogeneous Interacting Electron System
Standard K-S equation:
1
2
2+Vext +VH +
Exc(r)
(r)=
KS(r)
Vext +VH = constant
Vxc (r) =Exc(r)
constant Free electron dispersion (m* = me, infinite
lifetime, etc.)
WRONG!
Additional Theoretical Issues
• Kohn-Sham formulation is only one approach to DFT.- not unique- different formulation different eigenvalues
• How shall we interpret the K-S eigenvalues?- electron addition energies?- optical transition energies?
…
Diagrammatic Expansion of the Self Energy in Screened Coulomb Interaction
Hybertsen and Louie (1985)
H = Ho + (H - Ho)
Quasiparticle Band Gaps: GW results vs experimental values
Measured values: Bulk-state qp gap 1.2 eV Surface-state qp gap 0.7 eV Surface-state opt. gap 0.4 eV
Rohlfing & LouiePRL,1998.
Optical Absorption Spectrum of SiO2
Chang, Rohlfing& Louie.PRL, 2000.
M. Rohfling and S. G. Louie, PRL (1998)
Both terms important!
repulsive
attractive
Rohlfing & LouiePRL, 1998.
Optical Absorption Specturm of GaAs
Bound excitons
Optical Absorption Spectrum of SiO2
Chang, Rohlfing& Louie.PRL, 2000.
Nanostructures
• Size and restricted geometry => quantum confinement,enhanced many-electron interaction, reduceddimensionality, and symmetry effects
– Novel properties and phenomena– Useful in applications
Size
Bawendi Group: Colloidal CdSequantum dots dispersed in hexane.
• Small can be different!
Optical Excitations in Carbon Nanotubes
• Recent advances allowed the measurement of optical response of wellcharacterized, individual SWCNTs.
• Response is quite unusual and cannot be explained by conventionalpictures.
• Many-electron interaction (self-energy and excitonic) effects are veryimportant => interesting physics
(n,m) carbon nanotube
• Many-electron interaction effects
- Quasiparticles and the GW approximation
- Excitonic effects and the Bethe-Salpeterequation
• Single-walled carbon nanotubes
- Absorption spectra
- Exciton binding energies and wavefunctions
- Radiative lifetime, …
First-principles Study of Optical Properties
+
Quasiparticle Self-Energy Corrections
• Metallic tubes -- stretch of bands by ~15-25% (velocity renormalization)• Semiconductor tubes -- large opening (~ 1eV) of the gap
(8,0) semiconducting SWCNT(10,10) metallic SWCNT
Spataru, Ismail-Beigi, Benedict & Louie, PRL (2004)
GW Quasiparticle Band Dispersion of Metallic CNTs
Quasiparticle energy corrections:
• larger compared to graphite• increase with increasing diameter
Quasiparticle Fermi Velocities (106 m/s)
LDA QP GW shift
(3,3) 0.56 0.65 15%
(5,5) 0.72 0.85 19%
(10,10) 0.81 1.00 24%
Graphene 0.82 1.04 28%
Eq
p(e
V)
ELDA-EF
(10,10)
Absorption Spectrum of Semiconducting (8,0) Carbon Nanotube
• Long-range attractive electron-hole interaction• Spectrum dominated by bona fide and resonant excitons• Large binding energies ~ 1eV! [Experimental verification: Wang, Heinz et al, (2005); Ma, Fleming, et al.
(2005); Maultzsch, Molinari, et al, (2005), Avouris, et al …]
Spataru, Ismail-Beigi, Benedict &Louie, PRL 92, 077402 (2004)
(Not Frenkel-like)
| (re,rh)|2
1.41b1.441.67b1.741.19b1.21(11,0)
2.16a2.392.31a2.391.07a1.00(10,0)
1.17a1.161.88a1.801.60a1.55(8,0)
2.43a2.503.14a3.001.29a1.20(7,0)
Exp.TheoryExp.TheoryExp.Theory
E22/E112nd transition
(E22)1st transition
(E11)
Optical transition energies (in eV) of four semiconducting CNTs
aS. Bachilo, et al. (2002), bY-Z Ma, et al, (2005)
• Important Physical Effects: - band structure (~ eV shift each) - quasiparticle self-energy
- excitonic• Transport gap optical gap
Spataru, Ismail-Beigi, Benedict & Louie, PRL (2004)
(7,0) (8,0)
(10,0) (11,0)
Optical Spectrum of Semiconducting Carbon SWNTs
• Excitonic effects are equally dominant in BN nanotubes and Si nanowires!
Spataru, Ismail-Beigi, Benedict & Louie (2004)
Absorption Spectrum of (3,3) Metallic Carbon Nanotube
• Existence of bound excitons in metal tubes! (Eb = 86 meV)• Due to ineffective screening in 1D and symmetric gap• Similar results for the (10,10) and larger metallic tubes
• Phenomenon of electric field-induced half-metallicity
– Tunable spin carriers of one type (100%spin polarization)
– Could be useful for nanoscale spingeneration and injection
• Optical response is also dominated by excitons
Son, Cohen and Louie, Nature (2006)Son, Cohen and Louie, PRL (2006)Yang, Son, Cohen and Louie, (2007)
Graphene Electronic Structure
kx
ky
Ener
gy
kx' ky'
E
unoccupied
occupied
E =hvF
r k
EF
E2 = p2c2
2D massless Dirac fermion system
Graphene Nanoribbons with Homogenous Edges & Passivated -bonds
Armchair Graphene Nanoribbons(N-AGNRs)
Simple tight-binding:
Metal: Na = 3p+2 Semiconductor: Na = 3p or 3p+1
Zigzag Graphene Nanoribbons(N-ZGNRs)
Simple tight-binding: Always metal
Ab initio calculations predicted all GNRs have gaps!
Son, Cohen and Louie, PRL (2006)
Quasiparticle Band Structure and Optical Spectrum of 10-AGNR
Armchair-edgenanoribbon
• Width of w ~ 1.1nm• Large exciton binding energy of Eb ~1.3 eV• Similar strong exciton effects in other
nanoribbons
Yang, Park, Cohen and Louie (2007)
Forces in the Photo-Excited State:Self-trapped Exciton
Forces in Excited State
• For many systems, photo-induced structural changesare important– differences between absorption and luminescence– self-trapped excitons– molecular/defect conformation changes– photo-induced desorption
• Need excited-state forces– structural relaxation– luminescence study– molecular dynamics, etc.
• GW+BSE approach gives accurate forces in photo-excited state
Ismail-Beigi & Louie, Phys. Rev. Lett. 90, 076401 (2003)
Excited-state Forces
ES = E0 + S
RES = RE0 + R S
E0 & RE0 : DFT
S : GW+BSE
Ismail-Beigi & Louie, Phys. Rev. Lett. 90, 076401 (2003).
Verification on molecules
Ismail-Beigi & Louie, Phys. Rev. Lett. 90, 076401 (2003).
Excited-state force methodology
• Proof of principle: tests on molecules
- CO, NH3, …
• GW-BSE force method works well
• Forces allow us to efficiently find excited-stateenergy minima
[1] e.g. Itoh, Tanimura, &Itoh, J. Phys. C 21 (1988).
1. Start with 18 atom bulk cell
2. Randomly displace atoms by ±0.02 Å
3. Relax triplet exciton state
4. Repeat steps 2&3: same final config.
Ismail-Beigi & Louie, PRL (2005)
Structural Distortion from Self-Trapped Exciton in SiO2
Final configuration: Broken Si-O bond Hole on oxygen Electron on silicon Si in planar sp2 configuration
Ismail-Beigi & Louie, PRL (2005)
• Oxygen• Silicon
Atomic rearrangement for STE
No activation barrier!
Electron-Hole Wavefunction of Self-Trapped Exciton in SiO2
Hole probability distributionwith electron any where inthe crystal
Electron probabilitydistribution given thehole is in the colored box
Electron & Hole Distributions of Self-Trapped Exciton in SiO2
Final configuration: Broken Si-O bond Hole on oxygen (brown) Electron on silicon (green) Si in planar sp2 configuration
Ismail-Beigi & Louie, PRL (2005)
• Oxygen
• Silicon
Constrained DFT Calculations
Constrained LSDA: DFT with excited occupations
Problems:
• Relaxes back to ideal bulk from random initial displacements: excited-state energy surface incorrectly has a barrier.
• Large initial distortions needed for STE [1,2]
• Predicted Stokes shift and STE luminescence energy are very poor to correlate with experiments
[1] Song et al., Nucl. Instr. Meth. Phys. Res. B 166-167, 451 (2000).[2] Van Ginhoven and Jonsson, J. Chem. Phys. 118, 6582 (2003).
STE in SiO2: Comparison to Experiment
2.14
6.37
6.2-6.4
Stokes shift(eV)
----4.12CLSDA (forced)
0.48, 0.65,0.70
2.6, 2.74,2.75, 2.8
Expt. [1-6]
GW+BSE 2.85
Luminescencefreq.: T (eV)
LuminescencePol || z (*)
0.72
1. Tanimura et al., Phys. Rev. Lett. 51, 423 (1983).2. Tanimura et al., Phys. Rev. B 34, 2933 (1986).3. Itoh et al., J. Phys. C 21, 4693 (1988).4. Itoh et al., Phys. Rev. B 39, 11183 (1989).5. Joosen et al., Appl. Phys. Lett. 61, 2260 (1992).6. Kalceff & Phillips, Phys. Rev. B 52, 3122 (1996).
(*) Pol =Iz IxyIz + Ixy
Rohlfing & Louie,PRL, 1998.
Molecular energy levels at metal-organic interfaces