1 Supporting information for: Singlet fission in linear chains of molecules Francesco Ambrosio* and Alessandro Troisi* Department of Chemistry and Centre for Scientific Computing, University of Warwick, UK *E-mail: F.Ambrosio@warwick.ac.uk A.Troisi@warwick.ac.uk Supporting Information 1. Matrix elements of the SDCI Hamiltonian for a dimer model The following list of equations, adapted from ref.1, represents the full set of matrix elements of π»CI for the dimer model. For explanation of the different types of matrix elements, see main text. 1 1 2 2 β¨ |π»CI | β© = β¨ |π»CI | β© = πr − πa − π½r1 a1 + 2πΎr1 a1 1 1 2 2 (SI1) 2 2 1 1 β¨ |π»CI | β© = β¨ |π»CI | β© = πr − πa − π½r1 a2 1 1 2 2 (SI2) 11 11 22 22 |π»CI | β© = β¨ |π»CI | β© = 2(πr − πa ) − 4π½r1 a1 + π½r1 r1 + π½a1 a1 + 2πΎr1 a1 11 11 22 22 (SI3) 22 22 11 22 |π» | β© = β¨ |π»CI | β© = 2(πr − πa ) − 4π½r1 a2 + π½r1 r1 + π½a1 a1 11 CI 11 22 11 (SI4) β¨ β¨ β¨ 11 11 22 22 |π»CI | β© = β¨ |π»CI | β© == 2(πr − πa ) + πΎr1 a1 − 2π½r1 a1 + π½r1 r1 − 2π½r1 a2 + π½a1 a2 12 12 12 12 (SI5) β¨ 12 12 12 12 |π» | β© = β¨ |π»CI | β© == 2(πr − πa ) + πΎr1 a1 − 2π½r1 a1 + π½a1 a1 − 2π½r1 a2 + π½r1 r2 11 CI 11 22 22 (SI6) 12 12 (A)|π»CI | (A)β© = 2(πr − πa ) − 2π½r1 a1 12 12 (SI7) 12 12 (B)|π»CI | (B)β© = 2(πr − πa ) − 2π½r1 a1 + 4πΎr1 a1 12 12 (SI8) 2 1 β¨0|π»CI | β© = β¨0|π»CI | β© = πΉr1 a2 1 2 (SI9) 11 22 β¨0|π»CI | β© = β¨0|π»CI | β© = πΎr1 a1 11 22 (SI10) β¨ β¨ 2 12 β¨0|π»CI | (B)β© = −√3(r1 a1 |r 2 a2 ) 12 (SI11) 1 2 β¨ |π»CI | β© = 2(r1 a1 |r 2 a2 ) 1 2 (SI12) 1 2 2 1 β¨ |π»CI | β© = β¨ |π»CI | β© = πΉr1 r2 1 1 2 2 (SI13) 1 1 2 2 β¨ |π»CI | β© = β¨ |π»CI | β© = −πΉa1 a2 1 2 2 1 (SI14) 1 11 2 22 β¨ |π»CI | β© = β¨ |π»CI | β© = −πΉr1 a2 1 12 2 12 (SI15) 1 12 2 12 β¨ |π»CI | β© = β¨ |π»CI | β© = πΉr1 a2 1 11 2 22 (SI16) 2 22 1 11 β¨ |π»CI | β© = β¨ |π»CI | β© = √2πΉr1 a2 1 11 2 22 (SI17) 2 12 1 12 β¨ |π»CI | (A)β© = β¨ |π»CI | (A)β© = −√3/2πΉr1 a2 1 12 2 12 ( SI18) 11 11 22 22 |π» | β© = β¨ |π»CI | β© = √2πΉa1 a2 11 CI 12 22 12 (SI19) 11 12 22 12 β¨ |π»CI | β© = β¨ |π»CI | β© = √2πΉr1 r2 11 11 22 22 (SI20) 12 22 12 12 |π» | β© = β¨ |π»CI | β© = −(r1 a1 |r 2 a2 ) 11 CI 12 22 11 (SI21) 11 12 22 12 |π»CI | (A)β© = β¨ |π»CI | (A)β© = √2πΉa1 a2 12 12 12 12 (SI22) 12 12 12 12 |π» | (A)β© = β¨ |π»CI | (A)β© = √2πΉr1 r2 11 CI 12 22 12 (SI23) β¨ β¨ β¨ β¨ 2. Matrix elements of the SDCI Hamiltonian for a trimer model The following list of equations describes the additional types of relevant matrix elements that have to be considered in a trimer model, in addition to the types of matrix elements, considered in a ππ dimer model (here we only refer to A-type | 〉 configurations): ππ β¨ 13 12 |π»CI | β© = −(a2 a3 |r 2 r 3 ) 13 12 (SI24) 3 13 23 |π» | β© = −(a1 a2 |r1 r 2 ) 13 CI 23 (SI25) 23 12 |π»CI | β© = −(a1 a3 |r1 r 3 )~0 23 12 (SI26) β¨ β¨ where (a1 a3 |r1 r 3 )~0, since the matrix element is overlap dependent and, therefore, it will decay exponentially outside the overlap region, as in the case where are involved non-adjacent sites. ππ The indirect coupling between | 〉 configurations is mediated by ME-type CT configurations (we ππ are always considering A-type configurations): β¨ 23 23 13 13 |π» | β© = β¨ |π»CI | β© = −πΉa1 a2 13 CI 23 23 13 (SI27) β¨ 12 13 13 12 |π»CI | β© = β¨ |π»CI | β© = −πΉa2 a3 13 13 12 12 (SI28) 23 13 13 23 |π» | β© = β¨ |π»CI | β© = πΉr1 r2 13 CI 13 23 23 (SI29) 13 13 13 12 |π»CI | β© = β¨ |π»CI | β© = πΉr2 r3 13 12 12 12 (SI30) β¨ 23 12 23 12 |π» | β© = β¨ |π»CI | β© = −πΉa1 a3 ~0 23 12 CI 23 12 (SI31) β¨ 12 23 23 12 |π»CI | β© = β¨ |π»CI | β© = −πΉa1 a3 ~0 23 23 12 12 (SI32) β¨ β¨ Again, πΉa1 a3 ~0, because of the short-range nature of the overlap dependent matrix element. π Finally, | 〉 configurations couple with ME-CT configurations: π 1 13 β¨ |π»CI | β© = −πΉa2 r3 1 12 (SI33) 12 1 β¨ |π»CI | β© = −πΉa3 r2 13 1 (SI34) 3 13 β¨ |π»CI | β© = −πΉa2 r1 3 23 (SI35) 3 23 β¨ |π»CI | β© = −πΉa1 r2 3 13 (SI36) 4 21 2 β¨ |π»CI | β© = −πΉa3 r1 ~0 23 2 (SI37) 2 23 β¨ |π»CI | β© = −πΉa1 r3 ~0 2 21 (SI38) 3. On the effect of the relative sign of different π′ elements on the eingevalues The change in the eigenvalues, due to different relative signs of πΉr1 a2 , πΉr1 r2 ,πΉa1 a2 in a dimer P-like system (πΉr1 a2 and πΉa1 r2 are set to be equal), is evaluated. The lowest eigenvalues for different choices of the relative signs of the Fock elements are collected in Table SI1, showing that the differences are fairly modest and depend only on the relative sign of πΉr1 r2 and πΉa1 a2 . Table SI1: Lowest eigenvalues (eV) for a dimer P-like system for different relative signs of the π′ elements. πΉa1 a2 = +0.15 eV πΉa1 a2 = +0.15 eV πΉa1 a2 = +0.15 eV πΉa1 a2 = −0.15 eV πΉr1 r2 = +0.15 eV πΉr1 r2 = −0.15 eV πΉr1 r2 = +0.15 eV πΉr1 r2 = +0.15 eV πΉr1 a2 = +0.15 eV πΉr1 a2 = +0.15 eV πΉr1 a2 = −0.15 eV πΉt1 a2 = +0.15 eV 1 −0.34125 −0.34126 −0.34125 −0.34126 2 1.66507 1.66831 1.66507 1.66831 3 1.95438 1.99687 1.95438 1.99687 4 2.59596 2.53469 2.59596 2.53469 5 3.14609 3.14246 3.14609 3.14246 6 3.34761 3.36019 3.34761 3.36019 Eigenvalue Μ π −ππ and πππ 4. On the effect of the relative sign of different π′ elements on π½ π ππ ππ The change in πΜ π 1 −ME12 , due to different relative signs of πΉri aj , πΉri rj ,πΉai aj in a trimer P-like system along with the correspondent change in πME12 is evaluated. The results for different choices of the relative signs of the Fock elements are collected in Table SI2. When the signs of πΉr1 r2 and πΉa1 a2 . are opposite the computed rate decreases by one order of magnitude as a result of negative interference between different coupling paths. 5 Μ π¬ −ππ (eV) and ππ¨π (π€ ππ /π¬ −π ) for a trimer P-like system for different relative signs of the π′ elements. Table SI2: π π ππ ππ πΉai aj = +0.15 eV πΉai aj = +0.15 eV πΉai aj = +0.15 eV πΉai aj = −0.15 eV πΉri rj = +0.15 eV πΉri rj = −0.15 eV πΉri rj = +0.15 eV πΉri rj = +0.15 eV πΉri aj = +0.15 eV πΉri aj = +0.15 eV πΉri aj = −0.15 eV πΉri aj = +0.15 eV Μ π −ππ π½ π ππ −0.019467 0.005970 −0.019467 0.005970 ππ¨π (πππππ ) 13.151 12.123 13.151 12.123 5. Analysis of the accuracy of the matrix partitioning scheme The eigenvalues of the states of interest (S and ME states) for a linear tetramer (P-like parameters) obtained from a full SDCI Hamiltonian are compared with the ones obtained with the matrix partitioning approach, in order to test the accuracy of our approach. The results are collected in Table SI3. Table SI3: Eigenvalues (eV) of the states of interest for π = 4, πππ₯ = 0.10 eV, πΉ ′ = 0.15 eV, computed from a full SDCI Hamiltonian (column a) and with the matrix partitioning scheme (column b). State a b S1 1.9142 1.9171 S2 2.1086 2.1094 S3 2.3702 2.3805 S4 2.6322 2.6363 MEb1 1.6447 1.6450 MEb2 1.6452 1.6557 MEb3 1.6457 1.6559 MEu1 1.7168 1.7053 MEu2 1.7170 1.7121 MEu3 1.7421 1.7377 6. Reorganization energies and Huang-Rhys factors for the ME←S1 transition Reorganization energy for the T1←S0 transition has been computed at the B3LYP/6-31G* level of theory for a tetracene and pentacene molecule, as the difference between the energy of the optimized electronic structure of T1 (πΈ(T1 )T1 ) and the energy of T1 in the optimized geometry of S0 (πΈ(T1 )S0 ) following the procedure adopted in ref.2, 3 6 πT1 ←S0 = πΈ(T1 )T1 − πΈ(T1 )S0 (SI39) Therefore, πME←S1 for a transition taking place in a linear chain of molecules will be twice the value of πT1 ←S0 . We are considering that the geometry of S1 is identical to the one of S0, because of delocalization of the S1 state over several molecules, and that the ME state geometry is assumed to be the equilibrium geometry of two triplet states, because of localization of this type of state. Computed value of πT1 ←S0 is 0.217 eV for pentacene and 0.257 eV for tetracene, respectively, which gives values of πME←S1 equal to 0.434 eV for pentacene and 0.514 eV for tetracene, respectively. 7. Dependence of the rate from the value of π′ We compute the rate as a function of πΉ′ matrix element (we consider πΉri aj = πΉri rj = πΉai aj ) in a trimer model. In particular, we compute the rate of the transition from the S1 state to one of the two MEb states of the trimer (ME12), and to the MEu state (ME13) The rate is computed at πmin , for a fixed value of ΔπΈ (0.2 eV). Results are reported in Figure SI1: Figure SI1: Log(π) as a function of πΉ′, computed at πmin , for π = 3. Black dotted line shows an appropriately scaled (πΉ ′ )4 function. 7 The graph shows that the SF rate is roughly proportional to the fourth power of πΉ′ (πΜ π 1 −ME depends on the square of the πΉ′, and it is squared in the FGR equation). We also pinpoint that this contains only the effect of the πΉ′ element on the coupling, as πΉ′ matrix elements also modulate the difference ΔπΈ between MEb and MEu states, which decreases as πΉ′ decreases. References 1. A. Gamba, G.F. Tantardini and M. Simonetta, Spectrochim. Acta Mol. Spectros. 28A, 1877 (1972). 2. N. E. Gruhn, D. A. da Silva Filho, T. G. Bill, M. Malagoli, V. Coropceanu, A. Kahn and J.L. Brédas, J. Am. Chem. Soc. 124, 7918 (2002). 3. G. D. Scholes and K. P. Ghiggino, J. Chem. Phys. 101, 1251 (1994).