Excitation gaps of finite-sized systems from OptimallyTuned Range-Separated Hybrid Functionals: Leeor Kronik Department of Materials & Interfaces, Weizmann Institute of Science 5th Benasque TDDFT Workshop, January 2012 The Group Tami Zelovich Ido Azuri Eli Kraisler Andreas Baruch Karolewski Ariel Feldman (visiting) Biller Anna Ofer Sivan Hirsch Sinai Abramson Funding European Research Council Israel Science Foundation Germany-Israel Foundation US-Israel Binational Science Foundation Lise Meitner Center for Computational Chemistry Alternative Energy Research Initiative The people Roi Baer Tamar Stein (Hebrew U) Sivan RefaelyAbramson Natalia Kuritz (Weizmann Inst.) Kronik, Stein, Refaely-Abramson, Baer, J. Chem. Theo. Comp. (Perspectives Article), to be published Fundamental and optical gap – the quasi-particle picture Evac EA derivative discontinuity! IP Eg (a) Eopt (b) See, e.g., Onida, Reining, Rubio, RMP ‘02; Kümmel & Kronik, RMP ‘08 Mind the gap The Kohn-Sham gap underestimates the real gap Eg I A LUMO KS Perdew and Levy, PRL 1983; Sham and Schlüter, PRL 1983 HOMO KS xc derivative discontinuity! Kohn-Sham eigenvalues do not mimic the quasi-particle picture even in principle! H2TPP TD TD TD Energy [eV] -1.4 -1.7 -2.5 -2.9 1.8 -1.5 2.0 2.7 2.2 4.8 2.1 4.7 1.9 4.7 2.1 -4.7 -5.2 -6.2 -6.2 -6.4 -IP, -EA Eopt GGA B3LYP OT-BNL GW-BSE EXP Wiggle room: Generalized Kohn-Sham theory - Map to a partially interacting electron gas that is represented by a single Slater determinant. - Seek Slater determinant that minimizes an energy functional S[{φi}] while yielding the original density - Type of mapping determines the functional form Oˆ S [{ j }] V ion ( r ) v R ([ n ]; r ) i ( r ) i i ( r ) Seidl, Goerling, Vogl, Majevski, Levy, Phys. Rev. B 53, 3764 (1996). Kümmel & Kronik, Rev. Mod. Phys. 80, 3 (2008) Baer et al., Ann. Rev. Phys. Chem. 61, 85 (2010). - Derivative discontinuity problem possibly mitigated by non-local operator!! Hybrid functionals are a special case of Generalized Kohn-Sham theory! 1 2 sl sl ˆ V ( r ) V ([ n ]; r ) a V ( 1 a ) v ([ n ]; r ) v ([ n ]; r ) i ( r ) i i ( r ) ion H F x c 2 Does a conventional hybrid functional solve the gap problem? H2TPP TD TD TD Energy [eV] -1.4 -1.7 -2.5 -2.9 1.8 -1.5 2.0 2.7 2.2 4.8 2.1 4.7 1.9 4.7 2.1 -4.7 -5.2 -6.2 -6.2 -6.4 -IP, -EA Eopt GGA B3LYP OT-BNL GW-BSE EXP Need correct asymptotic potential! Can’t work without full exact exchange! But then, what about correlation? How to have your cake and eat it too? Range-separated hybrid functionals Coulomb operator decomposition: r 1 1 1 r erfc ( r ) r erf ( r ) Short Range Long Range Emphasize long-range exchange, short-range exchange correlation! 1 2 lr , sr , sl V ion ( r ) V H ([ n ]; r ) VˆF v x ([ n ]; r ) v c ([ n ]; r ) i ( r ) i i ( r ) 2 See, e.g.: Leininger et al., Chem. Phys. Lett. 275, 151 (1997) Iikura et al., J. Chem. Phys. 115, 3540 (2001) Yanai et al., Chem. Phys. Lett. 393, 51 (2004) Kümmel & Kronik, Rev. Mod. Phys. 80, 3 (2008). But how to balance ?? How to choose ? “Koopmans’ theorem” HOMO E gs ( N 1; ) E gs ( N ; ) Need both IP(D), EA(A) choose to best obey “Koopmans’ theorem” for both neutral donor and charged acceptor: Minimize J ( ) i0, ,i ( HOMO E gs ( N i 1; ) E gs ( N i ; )) i i 2 Tune, don’t fit, the range-separation parameter! Tuning the range-separation parameter J ( ) H ( N ) IP ( N ) H ( N 1) IP ( N 1) Neutral molecule (IP) J ( opt ) min J ( ) Anion (EA) H2TPP TD TD TD Energy [eV] -1.4 -1.7 -2.5 -2.9 1.8 -1.5 2.0 2.7 2.2 4.8 2.1 4.7 1.9 4.7 2.1 -4.7 -5.2 -6.2 -6.2 -6.4 -IP, -EA Eopt GGA B3LYP OT-BNL GW-BSE EXP Gaps of atoms Stein, Eisenberg, Kronik, Baer, Phys. Rev. Lett., 105, 266802 (2010). Fundamental gaps of acenes 10 Gap (eV) 9 Ref 0.31 BNL* orb. gap 0.28 8 7 0.25 6 0.22 5 0.20 0.19 4 0 1 2 3 4 5 6 7 n Stein, Eisenberg, Kronik, Baer, Phys. Rev. Lett., 105, 266802 (2010). Fundamental gaps of hydrogenated Si nanocrystals 14 0.41 12 0.33 Energy (eV) 10 0.24 8 6 -Lumo GW EA -HOMO Series4 Exp IP 4 2 0 0 0.13 0.14 0.12 5 10 15 Diameter (Å) GW: Tiago & Chelikowsky, Phys. Rev. B 73, 2006 DFT: Stein, Eisenberg, Kronik, Baer, PRL 105, 266802 (2010). Ionization Energy 11 10 - HOMO[eV] 9 8 7 GW data: Blasé et al., PRB 83, 115103 (2011) EXP 6 GW OT-BNL B3LYP 5 6 6.5 7 7.5 8 8.5 9 9.5 10 10.5 Experimental ionization energy [eV] S. Refaely-Abramson, R. Baer, and L. Kronik, Phys.Rev. B 84 ,075144 (2011) [Editor’s choice]. Optical gaps with Time-dependent DFT TDDFT: BNL results as accurate as those of B3LYP a – thiophene b – thiadiazole c – benzothiadiazole d – benzothiazole e – flourene f – PTCDA g – C60 h – H2P i – H2TPP j – H2Pc S. Refaely-Abramson, R. Baer, and L. Kronik, Phys.Rev. B 84 ,075144 (2011) The charge transfer excitation problem Time-dependent density functional theory (TDDFT), using either semi-local or standard hybrid functionals, can seriously underestimate charge transfer excitation energies! Biphenylene – tetracyanoethylene: B3LYP: 0.77 eV Experiment: 2 eV Liao et al., J. Comp. Chem. 24, 623 (2003). zincbacteriochlorin-phenylenebacteriochlorin: GGA (BLYP): 1.33 eV CIS: 3.75 eV Druew and Head-Gordon, J. Am. Chem. Soc. 126, 4007 (2004). The Mulliken limit In the limit of well-separated donor and acceptor: h CT IP ( D ) EA ( A ) 1 / R Coulomb attraction Neither the gap nor the ~1/r dependence obtained for standard functionals! Both obtained with the optimally-tuned range-separated hybrid! Results – gas phase Ar-TCNE Donor benzene toluene o-xylene Naphtha lene MAE TDPBE TDB3LYP TDBNL =0.5 TDBNL Best Exp G0W0BSE GWBSE (psc) 1.6 2.1 4.4 3.8 3.59 3.2 3.6 1.4 1.8 4.0 3.4 3.36 2.8 3.3 1.0 1.5 3.7 3.0 3.15 2.7 2.9 0.4 0.9 3.3 2.7 2.60 2.4 2.6 2.1 1.7 0. 8 0.1 --- 0.4 0.1 Thygesen Blase PRL ‘11 APL ‘11 Stein, Kronik, Baer, J. Am. Chem. Soc. (Comm.) 131, 2818 (2009). Wong, B. M.; Cordaro, J. G., J. Chem. Phys. 129, 214703 (2008). 0.25 0.20 Difference from CC2 (eV) 0.15 0.10 0.05 0.00 -0.05 -0.10 -0.15 -0.20 -0.25 C343 Fit to CC2 -0.08 NKX 2388 s-tran s 0.02 NKX 2388 s-cis NKX 2311 s-cis 0.05 NKX 2311 strans 0.02 NKX 2586 s-cis NKX 2677 Mean MAD 0.02 NKX 2586 s-tran s 0 0 -0.04 0.00 0.03 BNL* 1 0.01 0.09 0.11 0.03 0 -0.05 -0.05 -0.02 0.02 0.05 BNL* 2 or 3 -0.08 -0.06 -0.01 -0.1 -0.11 -0.17 -0.17 -0.23 -0.12 0.12 Stein, T.; Kronik, L.; Baer, R., J. Chem. Phys. 131, 244119 (2009). Optical excitations: Fixing the La, Lb problem of oligoacenes 4.60 B3LYP 5.10 BNL (Tuned) 4.60 BP86* 4.10 3.60 3.10 2.60 2.10 1.60 1.10 excitation energy [eV] La 5.10 excitation energy [eV] CC2 CC2 Lb B3LYP BNL (Tuned) BP86* 4.10 3.60 3.10 2.60 2.10 1.60 1.10 2 3 4 5 6 N – number of benzene rings 2 3 4 5 N-number of benzene rings Kuritz, Stein, Baer, Kronik, J. Chem. Theo. Comp. 7, 2408 (2011). 6 Where’s the charge transfer? LUMO +1 LUMO HOMO Energy LUMO HOMO HOMO-1 1L b excitation La excitation LUMO HOMO LUMO+HOMO LUMO-HOMO N. Kuritz, T. Stein, R. Baer, L. Kronik, JCTC 7, 2408 (2011). Conclusions Kohn-Sham quasi-particle Optical GW GW+BSE RSH TD-RSH Kronik, Stein, Refaely-Abramson, Baer, J. Chem. Theo. Comp. (Perspectives Article), to be published Two different paradigms for functional development and applications From To Tuning is NOT fitting! Tuning is NOT semi-empirical! Choose the right tool (=range parameter) for the right reason (=Koopmans’ theorem)