Non-equilibrium electron transport in dry DNA C D Pemmaraju, I Rungger and S Sanvito School of Physics and CRANN, Trinity College, Dublin http://www.spincomp.com QuantSim09, Warwick, UK Outline Motivation Electronic structure Spatial distribution of electronic wavefunctions Energy spectrum and level alignment Zero Bias results Dependence on geometry Effect of solvation layers Finite bias results Incoherent contributions Conclusions QuantSim09, Warwick, UK Intro to DNA A-DNA B-DNA Z-DNA 5' 3' So can DNA be used as a current carrying wire? QuantSim09, Warwick, UK Earlier experiments Insulator (12 eV) Nature 391, 775 (1998) Metal Nature 398, 407 (1999) PRL 85, 4992 (2000) QuantSim09, Warwick, UK Superconductor Science 291, 280 (2001) Earlier experiments Rev.Mod.Phys, 76, 195 (2004) QuantSim09, Warwick, UK Recent Experiments PNAS, 102, 11589 (2005) 26-bp long ds-DNA of a complex sequence Deposited on Au-111 and connected to Au nano particle Semi-conducting behaviour with a voltage gap of ~2 eV is observed Large currents of ~220 nA at 2.0 V QuantSim09, Warwick, UK Recent Experiments Appl. Phys. Lett. 87, 083902 (2005) SAM of 12 bp poly(GC)-poly(GC) DNA on Au(111) ●Voltage gap of ~2 V ●Current in the range of ~1 nA ● poly(G)-poly(C) DNA deposited flat on Au(111) Voltage gap of ~2.0 V is observed Current ~ 1nA above 2.0 V Nat.Mat, 7, 68 (2008) QuantSim09, Warwick, UK Why study dry poly(G)-poly(C) DNA? poly(G)-poly(C) is possibly the best case scenario for high conductance. Has the same base sequence (C or G) on each strand. Formation of extended states more likely. Has the smallest band gap Guanine has the lowest I.P Cytosine has the highest E.A Degree of solvation in STM set-ups is unclear Dry conditions are easier to handle ab-initio QuantSim09, Warwick, UK Computational approach ● ● The DFT+NEGF formalism is used Calculations carried out using the SMEAGOL code www.smeagol.tcd.ie – – two terminal device setup LDA+Approximate self interaction correction scheme (ASIC) Phys. Rev. B 75, 045101 (2007) ● poly(G)-poly(C) DNA with alkyl-thiol connectors on Au(111) leads ● System size: ~ 6000 orbitals ● Complex plane integration points: 128 ● ● Real energy points to integrate non-equilibrium contribution to DM: 1000-5000 Parallelized over 256-512 compute nodes QuantSim09, Warwick, UK SMEAGOL: NEGF + DFT Green's function: −1 G E =[ E i −H eff ] L R Coupling matrices: † L=i L− L L = F eV 2 H eff = H M L R † R =i R − R R =F − eV 2 SMEAGOL (NEGF): A. R. Rocha et al., PRB 73, 085414 (2006); Nature Materials 4, 335 (2005); http://www.smeagol.tcd.ie SIESTA (DFT): J. Soler et al., J. Phys.: Condens. Matter 14, 2745 (2002) QuantSim09, Warwick, UK Electronic structure QuantSim09, Warwick, UK periodic poly(G)-poly(C) A-DNA Back bone Cytosine Guanine derived HOMO band has a very small band-width ~60 meV QuantSim09, Warwick, UK Finite fragments of DNA? Guanine band splits into discrete levels. Levels spread over a much wider energy range ~0.6 eV QuantSim09, Warwick, UK Localization of electronic states HOMO localized at 5' end of the Guanine chain Energy is lowered moving from the 5' to 3' end QuantSim09, Warwick, UK Localization of electronic states QuantSim09, Warwick, UK Localization of electronic states Periodic 5' G G G G G G G G G G G 3' 3' C C C C C C C C C C C 5' Finite 5' G G G G G G G G G G G 3' 3' C C C C C C C C C C C 5' Surface dipole leads to localization of electronic wavefunctions QuantSim09, Warwick, UK Effect of a solvation layer Energy levels are rearranged as water layer screens the dipole QuantSim09, Warwick, UK Energy level alignment We need quasi-particle energies i.e I.P and E.A Semi-local DFT functionals perform poorly Self-interaction corrections + Correlation effects are needed Guanine Expt (eV) -EHOMO/ELUMO ΔSCF GGA ASIC GGA Ionization Potential (I.P) 8.2 5.2 8.8 7.9 Electron Affinity (E.A) 1.4 4.04 -1.2 3.8 4.76 >7.9 NA Band Gap Cytosine Expt (eV) -EHOMO/ELUMO ΔSCF GGA ASIC GGA Ionization Potential (I.P) 8.9 5.6 9.7 8.54 Electron Affinity (E.A) 0.2 2.1 4.8 -0.7 Band Gap 8.7 3.5 4.9 >8.54 ΔSCF Energies provide the best estimates of removal/addition energies I.P(ΔSCF)=E(N)-E(N-1), E.A(ΔSCF)=E(N+1)-E(N) QuantSim09, Warwick, UK Energy level alignment For short chains, I.P decreases with increasing chain length Hybrid-DFT gives higher I.P compared to ordinary GGA. I.P value converges at around 6 base pairs GGA HOMO eigenvalue much too high An approximate Self-Interaction correction is applied to “tune” the HOMO eigenvalue to match the ΔSCF I.P ΔSCF band-gap > 3.8 eV QuantSim09, Warwick, UK Energy level alignment GGA HOMO is pinned at EF leading to a sharp DOS The same feature is seen in the transmission coefficient The correct HOMO-EF distance is restored with ASIC ASIC LUMO states are roughly 1.5 eV too low as they do not correspond directly to E.A QuantSim09, Warwick, UK Zero Bias Results QuantSim09, Warwick, UK Adaptive mesh 1 meV Refinement: Start with uniform mesh and broadened GF Resolve peaks Add mesh points where needed and decrease broadening Iterate until true width of resonances is reached. QuantSim09, Warwick, UK 4.0 10-7 eV Zero bias properties HOMO QuantSim09, Warwick, UK Effect of alkyl-thiol connections QuantSim09, Warwick, UK Effect of a solvation layer HOMO HOMO QuantSim09, Warwick, UK Finite bias Results QuantSim09, Warwick, UK Voltage gap vs. transmission gap Vg LUMO ELUMO EFL EHOMO EFR HOMO 1)LUMO and HOMO do not move with bias: Vg=4*Min(EHOMO,ELUMO) ● 2)LUMO and HOMO move with one of the two leads (asymmetric coupling, STM): Vg=ELUMO+EHOMO ● 3)LUMO and HOMO avoid the bias window: Vg=4*Max(EHOMO,ELUMO) ● QuantSim09, Warwick, UK Finite bias setup: G3'C3' arrangement The 3' ends of a 6 base pair DNA are connected to Au via thiol groups. Coupling of localized states expected to be asymmetric QuantSim09, Warwick, UK G3'C3' geometry: PDOS and LDOS HOMO is localized at the G5' end LUMO is localized at the C5' end HOMO LUMO QuantSim09, Warwick, UK G3'C3' geometry: I-V Vgap ~ 2 eV QuantSim09, Warwick, UK G3'C3' setup: Bias dependent transmission HOMO peak conducts first for both positive and negative bias QuantSim09, Warwick, UK G3'G5' geometry Both the 5' and 3' ends of the Guanine chain are connected. Cytosine chain is left unconnected QuantSim09, Warwick, UK G3'G5' geometry: PDOS and LDOS HOMO is localized at G5' LUMO is localized at C5' Coupling of HOMO to leads expected to be even more asymmetric. LUMO HOMO QuantSim09, Warwick, UK G3'G5' arrangement: I-V Vgap ~ 2.2 eV QuantSim09, Warwick, UK G3'G5' geometry: Bias dependent transmission QuantSim09, Warwick, UK Incoherent contributions QuantSim09, Warwick, UK Incoherent contributions: floating probes To obtain a basic approximation for incoherent contributions to the current •We apply the idea of M. Büttiker [PRB 33, 3020 (1986)] and add a set of additional fictitious leads to the system, so that all states are coupled to at least one of the leads •The Fermi energy of each of the additional leads is set by the condition that no current can flow into these. −1 G E =[ E − H M− L− R −∑ n ] L R 1 Current conservation: I n =∫ dE T nm f n − f m =0 ⇒ † 3 occupied empty EF,n T nm=Tr [ n G m G ] EF ,L EF ,2 EF ,1 Total current: I =∫ dE T LR f L− f R ∑n ∫ dE T Ln f L− f n coherent 2 incoherent QuantSim09, Warwick, UK EF ,3 EF ,R Incoherent contributions:System setup 1 2 linker 4 6 8 10 G G G G linker 11 C C C C 3 5 7 9 12 The system is partitioned into 12 units each with its own floating lead and local Fermi level The coupling strength (broadening) is controlled by a parameter δ QuantSim09, Warwick, UK Incoherent contributions 1 2 linker 4 6 8 10 G G G G linker 11 C C C C 3 5 7 9 VB=0.2 V 12 The local Fermi level seems to follow the potential drop QuantSim09, Warwick, UK Incoherent contributions VB=0.2 eV Current without the incoherent contributions is rather small. At a given bias, increasing the strength of δ increases the current QuantSim09, Warwick, UK Incoherent contributions TnR exhibits a position dependence Electrons from probe “n” tunneling into the right lead “R”, only “see” electronic states localized to the right of “n”. 5' QuantSim09, Warwick, UK 1 2 3 4 G G G G C C C C 1 2 3 4 Conclusions Electronic states on DNA fragments are strongly localized. The Fermi level of Au is expected to lie within the HOMO-LUMO gap of DNA. Most of the states on DNA are expected to couple asymmetrically to the leads in SAM geometries A voltage gap of ~ 2 eV is expected for coherent tunneling conduction. The size of the voltage gap is most likely determined by Guanine HOMO levels as the empty Cytosine states are too high. Currents obtained from a purely coherent tunneling mechanism are several orders of magnitude smaller than those experimentally observed. Incoherent contributions may be treated in a qualitative way by introducing an additional broadening through local fictitious leads QuantSim09, Warwick, UK Acknowledgements Ivan Rungger Alexandre Rocha Stefano Sanvito Computational resources TCHPC Funding Thank You QuantSim09, Warwick, UK Weakly coupled states generalized bound state : Re n E W = E W L , R , and L , = L ∣Im n E W ∣ † R , = R † 1 Im = − L , R , 2 L , f L , R , f R , occupation = L , R , 2 e L , R , I = h L , R , •Mapping from DFT to single level tight-binding-like parameters •For a true bound state ɣLα=ɣRα = 0, so that Iα =0; the occupation however is undefined for states in the bias window (fLα ≄ fRα ). © CRANN www.crann.tcd.ie Some definitions γL εα γR ΣL ΣR γ α determ ines the broadening of trans m is s ion peaks rα is a m eas ure of coupling as y m m etry T α → 1 as rα → 0 (ev en coupling ) rα → + 1 if s tate is s trong ly coupled to the left lead rα → -1 if s tate is s trong ly coupled to the rig ht lead QuantSim09, Warwick, UK QuantSim09, Warwick, UK QuantSim09, Warwick, UK Localization in 2 base pair DNA G C C HOMO LUMO HOMO-1 LUMO-1 QuantSim09, Warwick, UK G