ebook img

Efficient Symmetrized DMRG Algorithm for Low-lying Excited States of Conjugated Carbon Systems: Application to Coronene, Ovalene and 1,12-Benzoperylene PDF

0.23 MB·
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Efficient Symmetrized DMRG Algorithm for Low-lying Excited States of Conjugated Carbon Systems: Application to Coronene, Ovalene and 1,12-Benzoperylene

Efficient Symmetrized DMRG Algorithm for Low-lying Excited States of Conjugated Carbon Systems: Application to Coronene, Ovalene and 1,12-Benzoperylene Suryoday Prodhan,1,a) Sumit Mazumdar,2,3,b) and S. Ramasesha1,c) 1)Solid State and Structural Chemistry Unit, Indian Institute of Science, Bangalore-560012, India. 2)Department of Physics, University of Arizona, Tucson, Arizona 85721, USA. 3)College of Optical Sciences, University of Arizona, Tucson, Arizona 85721, USA. (Dated: 5 January 2016) Symmetry adapted density matrix renormalization group (SDMRG) technique has been an efficient method for studying low-lying eigenstates in one- and quasi-one dimensional electronic systems. However, SDMRG 6 method had bottlenecks involving construction of linearly independent symmetry adapted basis states as 1 the symmetry matrices in the DMRG basis were not sparse. We have developed a modified algorithm to 0 overcome this bottleneck. The new method incorporates end-to-end interchange symmetry (C ), electron- 2 2 hole symmetry (J) and parity symmetry (P) in these calculations. The one-to-one correspondence between n direct-product basis states in the DMRG Hilbert space for these symmetry operations renders the symmetry a J matricesinthenewbasiswithmaximumsparseness,justonenon-zeromatrixelementperrow. Usingmethods similartothoseemployedinexactdiagonalizationtechnique forPariser-Parr-Pople(PPP)models,developed 1 in the eighties, it is possible to construct orthogonal SDMRG basis states while bypassing the slow step of ] Gram-Schmidtorthonormalizationprocedure. The methodtogether withthe PPPmodelwhichincorporates h long-range electronic correlations is employed to study the correlated excited state spectrum of coronene, p substituted coronene, ovalene and 1,12-benzoperylene; all these systems can be viewed as graphene dots. It - m is knownthatin linearpolyenes the correlationparametersofthe modelHamiltoniandetermine the ordering of the lowestone-photon, two-photonand triplet states, which are experimentally important. In the systems e h we have studied, the correlationparameters for all the orbitals of the model are same, however,the ordering c of the important states is very different in these systems. This underlines the importance of the topology of s. thehoppingterminthePPPmodelformolecules. Wealsodiscussindetailthe propertiesoftheseimportant c states in the molecules we have studied. The properties include bond orders, spin density distribution in the i s tripletstateandtransitiondipolemomentforthelow-lyingone-photonstatesfromthegroundstate. Besides, y the exciton binding energies and charge gaps are also reported for these systems. h p [ I. INTRODUCTION the Brillouin zone leading to formation of Dirac cones. 1 FractionalquantumHalleffectatroomtemperatureand v Successful exfoliation of a single atomic layer of tunneling throughclassicallyforbidden regionexhibiting 3 graphene from graphite crystal has led to an explosion Klein paradox are consequences of the dispersion1. 5 of research, both experimental and theoretical, in the Being a π-conjugated system, electron-electron corre- 0 area of graphene and related systems. A sheet of car- lation should play an important role in the physics of 0 0 bon atoms, arranged in honeycomb structure can be as- thistwo-dimensionalcarbonarray,althoughthestrength . sumed to be the parent structure from which carbon al- of correlation may get diminished from well-known π- 1 lotropes like carbon nanotubes and graphite can be con- conjugated systems like polyenes, linear polyacenes or 0 6 structed. Each of the carbon atoms in graphene being carbon nanotubes due to higher dimensionality. Tight- 1 sp2-hybridized are σ-bonded with three nearest neigh- binding calculations of graphene without explicitly con- : borswhilethe2p orbital,perpendiculartothemolecular sidering the electron-electron repulsion provides suffi- v plain provides thze conjugation network. The completely ciently accurate ground state description2, yet, ignoring i X filledσ-bandliesdeepdowninenergyandprimarilyren- explicitelectroncorrelationshasbeen criticizedrecently. r ders stability to the structure. On the other hand, the Onthecontrary,organicpolyeneandpolyacenesystems, a half filled π-band, due to the overlapof the 2p orbitals, which are quasi-one dimensional, have been well studied z ismostlyresponsibleforfascinatingpropertieslikelinear in the long-ranged Pariser-Parr-Pople(PPP) model and dispersion of electronic energy band near the vertices of the effect of electron-electron correlation is found to be qualitativelysignificant. Theorderingofexcitedstatesin polyenesdifferfromsingleCIcalculationsanditprovides anexplanationofthe opticalspectrainthesesystems3,4. a)Electronicmail: [email protected] Correlation also plays a significant role in dynamical b)Electronicmail: [email protected] properties of polyenes; observation of electrolumines- c)Electronicmail: [email protected] cenceefficiencywhichis farinexcessoftheupper bound 2 placedbythespinstatistics,getsresolvedonlyaftertak- and highly accurate DMRG algorithm for utilizing end- ing explicit electron correlations in account5. Crossover to-endinterchangesymmetry(C ),electron-holesymme- 2 of the dipole allowed or one-photon state (1B ) and the try (J) and spin-flip or parity symmetry (P). The ad- u dark two-photonstate (2A ) in polyacene series with in- vantage of this algorithm overthe previous symmetrized g creasing chain length, is observedin the PPP model cal- DMRG method is that the symmetry operators can be culations and agrees well with experiments6. Correlated expressedashighlysparsematriceswithonlyonenonzero electronic calculations even support the higher optical element per row. This allows representing the symme- gaps observed in carbon nanotubes with different diam- try operators by vectors of the same dimension as of the eters, contrary to tight binding calculations7. The ef- DMRG Hilbert space. Hence it is possible to employ fect of an additional dimension on the correlation effect, very large DMRG cut-offs in the new symmetry adap- therefore, is worthwhile investigating starting from sys- tation algorithm. We have used this new algorithm to tems suchaspolycyclic peri-condensedacenes(graphene calculate the ground state, lowest-lying two-photon and nano-discs), which resemble closely to two-dimensional one-photonstatesinM =0sectoralongwithafewlow- S graphene system. lying triplet states in M = 1 sector for coronene, ova- S lene, 1,12-benzoperylene and substituted coronene. No- While a number of optical studies have been car- riedoutonpolynucleararomatichydrocarbons8,mostof tably,coronenehasD6h symmetry,while the otherthree are of lowersymmetry groups;nonetheless, in our study, the theoretical studies are either in the non-interacting we have considered only one C symmetry axis in all limitorapproximatingthe electronrepulsionviaa mean 2 the molecules. We have also characterized these states field. Moffitt and Coulson calculated the bond order in by calculating bond orders along with the spin density coronene exploiting both Hu¨ckel’s tight binding method at different sites (in M = 1 sector). Charge gaps and and valence bond method without considering high en- S ergy ionic structures9 while Pritchard and Sumner cal- exciton binding energies for coronene, ovalene and 1,12- benzoperylene are also mentioned here. culated the ground state wavefunction and bond order The paper is organizedas follows. In the next section, for coronene and ovalene using self-consistent molecu- lar orbital theory10. Hummel and Ruedenberg also used we have given an account of the modified algorithm and compared it with exact calculations for 16-site polyene similar techniques to study other large polyaromatic hy- drocarbons. A number of electronic structure calcula- chain and tetracene with 18 pz-orbitals in conjugation. InsectionIII,wehavediscussedourresultsforcoronene, tions of these polycyclic compounds and their ions, em- ovalene, 1,12-benzoperylene and substituted coronene. ploying density functional theory (DFT) and its time- dependent variant (TDDFT) have been reported11. The In the last section we have summarized our study and concluded. accuracy of these calculations depend upon the reliabil- ity of the functional form of the exchange correlation potential which still is not beyond doubt for strongly correlated systems. Furthermore, implementation of the II. METHODOLOGY TDDFT method considers only single particle-hole exci- tations which lacks the desired accuracy. In recent past, A. Modified Algorithm For Symmetry Adaptation excited state ordering in some of these molecules were studiedusingmultiplereferenceconfigurationinteraction The DMRG procedureis a truncatedHilbertspace al- (MRCI) approach,which considers upto quadruple exci- gorithmwhichadoptsiterativeblock-buildingschemefor tations of each targeted state12. The theoretical predic- calculatinglow-lyingeigenstatesoflargeoneorquasi-one tions are in agreement with the optical and two-photon dimensional Hamiltonians14,16. The system block or left absorption data. Another study has speculated the ef- block(L)alongwiththeenvironmentblock,alsodenoted fect of different molecular geometries over the ordering asrightblock (R)arelinearlygrownthroughadditionof of excited states in polynuclear hydrocarbons13. While one new site to each block per iteration and these two theHartree-Focksemiempericalcalculationsareoversim- together form the superblock which corresponds to the plified, the accuracy of the other method used is still physical system to be studied. In the infinite DMRG al- dubious. This problem can be overcome by exact di- gorithm, superblock of size 2l consists of system and en- agonalizationtechnique in full Hilbert space, though the vironment blocks of same size (l). The reduced density sizeoftheHilbertspaceincreasesexponentiallywithsys- matrix of the system (environment) is generated from temsizeandmakesitcomputationallyexpensiveevenfor low-lying eigenstates of the superblock by tracing over moderate-sized systems. the environment (system) states. Matrix elements of The density matrix renormalization group method the reduced density matrix of the system (environment) (DMRG), put forward by White14, is a well-accepted are represented by ρm,m′ =Pnψm,nψm∗′,n, where m,m′ methodfor calculatinglow-lyingstatesofone andquasi- aresystem(environment)stateswhilenareenvironment one dimensional systems. Ramasesha et al. have suc- (system) states. Reduced density matrices for both the cessfullyimplemented spatialanddiscretesymmetriesin system and environment are diagonalized and Ml den- DMRG to target low-lying states in specific symmetry sity matrix eigenstates (DME) with highest eigenvalues subspaces6,15. In this study, we have employed a new ({µ }for system and {µ }for environment)arestoredas L R 3 columnvectorsofaMl−1d ×Ml matrix,whereMl−1 is The Hamiltonian of a half-filled bipartite system con- σ the number of truncated reduced density matrix eigen- serves electron-hole symmetry (J), where interchanging states retained at (l −1)-th iteration and d is the di- the creation and annihilation operators at each lattice σ mension of the Fock space of the newly added site in sites with an associated phase factor keeps the Hamilto- l-th iteration. The Hamiltonian and site operators of nianinvariant. Onapplicationofthissymmetryelement, the newly generated blocks are constructed in the basis Fock states of a single site transforms in the following µl−1σandthenrenormalizedusingtheMl−1d ×Ml ma- way: σ trix. Further addition of two new sites gives a system of 2l+2sites,whosebasis|µlLσiσi′µlRiisgeneratedbydirect Jˆi|0i=|↑↓i Jˆi|↑i=(−1)ηi|↑i productof |µlLi ofthe systemblock, |µlRi ofthe environ- Jˆi|↑↓i=−|0i Jˆi|↓i=(−1)ηi|↓i mentblockwith|σiiand|σi′iofthenewlyaddedsitesto the left and right blocks respectively. The Hamiltonian where η is 1 if ‘i’ belongs to A-sublattice and 0 if it i for the 2l+2 sites system is formed in this direct prod- belongs to B-sublattice. uctspaceanddiagonalizedtogetthe energyeigenstates, The result of parity or spin-flip symmetry (P) which which are utilized for the next iteration. In case of sys- flips the spin at each site is given by the following: temswithfinitesizeN,finiteDMRGalgorithmenhances the accuracy in eigenstate calculation; infinite algorithm Pˆ|0i=|0i Pˆ|↑i=|↓i i i is employed starting from a small system until the de- Pˆ|↑↓i=−|↑↓i Pˆ|↓i=|↑i sired superblock size is reached and then sweeping over i i thesuperblockisappliedbychangingthesystemanden- Non-relativistic Hamiltonians remain invariant under vironment sizes, keeping the superblock size fixed. The parity symmetry. This symmetry can only be employed basis at a step of the finite DMRG sweeping can be rep- in M = 0 space, as application of this symmetry maps resented by |µlLσiσi′µNR−l+2i, where l can vary between fromSone non-zero MS space to a −MS space. 1 and N −3. Once the size of the system (environment) BothJ andP operatorsaresiteoperatorsandhence, i i block reaches its maximum, the environment (system) the operators for the full system can be defined as: blockstartsto growattheexpense ofthe otherblock. A complete sweep consists of expansion of both blocks to Jˆ=YJˆi Pˆ =YPˆi their maximum and returning back to the block size of i i the final infinite DMRG step. The symmetryelementsconsideredinourmodifiedal- Thematrixelementofthesetwosymmetryoperatorsfor gorithm are same as in the work by Ramasesha et al.15. a superblock of size 2l+2 can be represented by: The C symmetry about an axis perpendicular to the 2 hµ′σ′σµ|Rˆ |νττ′ν′i plane of the molecule corresponds to the end-to-end in- 2l+2 (1) terchange of the superblock with a phase factor: =hµ|Rˆl|νihσ|Rˆi|τihσ′|Rˆi′|τ′ihµ′|Rˆl|ν′i Cˆ2|µlLσiσi′µlRi=(−1)γ|µlRσi′σiµlLi i.e. itisdirectproductofmatrixelementsofcorrespond- or in shorter notation (we will use onwards) ing symmetry operators in the left block DME space, rightblockDMEspaceandintheFockspacesofthetwo Cˆ2|µσσ′µ′i=(−1)γ|µ′σ′σµi newly added sites. γ =nµ(nσ+nσ′ +nµ′)+nσ(nσ′ +nµ′)+nσ′nµ′ The symmetry operators of the superblock follow the relations i.e. the system and environment blocks inter- change their density matrix eigenstates while the two (Cˆ )2 =1 Jˆ2 =(−1)(Ns+Ne) Pˆ2 =1 (2) 2 newly added sites also interchange their Fock states. nµ,nσ,nσ′ and nµ′ are the occupancies in the system where Ns and Ne are number of sites and number of block, environment block and in the two newly added electrons in a basis state, onwhich the symmetry opera- sites. Astheinterchangeshouldnotaffectthesuperblock tions are applied. In case of half-filled fermionic system Hilbert space, system and environment blocks should N = N ; subsequently, each of the operator becomes s e have same density matrix eigenspaces. In other words, their own inverse, the symmetry operators commute employing C symmetry approximately halves the effec- among themselves and form an abelian group. However, 2 tive size of the superblock Hilbert space beside reduc- the symmetry operators may not commute in individual ing the computational expense by avoiding environment block spaces and in Fock spaces of newly added sites. block calculations, at every infinite DMRG step and at Asmentionedearlier,the C symmetryoperatormaps 2 the last step of the finite DMRG iteration. In this sym- onedirectproductstateexclusivelytoanotherone. How- metry operation, each DMRG basis state is carried over ever, J and P operators do not produce a single state, to only one unique DMRG basis state, with a phase of due to the absence of one-to-one symmetry correspon- ±1. Thus the C symmetry operationcanbe storedas a dence between different density matrix eigenstates of a 2 column vector with entry ±j , resulting from C opera- particular block, although the mapping of Fock states k 2 tion on the state k. of a single site is exclusive. This absence of one-to-one 4 mappingofthe statesinthe sub-blocksandhence aone- where ηs are the corresponding phase factors. to-one mapping of the states of the superblock leads to (ii){µi(ns >ne,MS =0)}: ApplicationofJL symme- some serious problems in symmetry adaptation. Firstly, try operator on these density matrix eigenvectors gener- the matrix representation of the operators in the DME ate eigenvectors with particle number 2n −n ; yet, P s e L basis is not very sparse. Secondly, it is difficult to keep symmetryshouldmaptheeigenspacetoitself. Therefore, track of and thereby include all the symmetry partners to begin with, this group of eigenvectors is further sub- of the J and P symmetry while introducing the DMRG divided into smaller sub-groups {µ (n = n′,M = 0)}, i e e S cut-off. Thirdly, weeding out linearly dependent states where n′ is some particular value of n and n′ < n . e e e s after symmetry adaptation is a slow step as it involves Considering these smaller blocks, one at a time, we set Gram-Schmidt orthonormalization of the symmetrized up the matrix of Pˆ in this sub-space. This is diagonal- superblock states. Thus, a successful symmetry adap- ized to obtain the eigenvectors of Pˆ so that each of the tation scheme must have a one-to-one unique correspon- new DMEs map into themselves with a phase factor of dence between states in the DME basis space. ±1. These transformed µ (n =n′,M =0) provide a k e e S We invoke the one-to-one correspondence among dif- one-to-one mapping under the symmetry operations Jˆ, ferent DMEs which results in only one non-zero ele- Pˆ and PˆJˆ. ment in each row of the matrices of symmetry opera- (iii) {µi(ns =ne,MS <0)}: This set of eigenstates tors. This is achieved by considering truncated reduced are similar to the one discussed above, differing only in density matrix eigenstates restricted to a specific region the fact that J will map the space to itself instead of L of the Hilbert space, as by symmetry we can obtain the P . The space is sub-divided into smaller blocks on the L DMEs in the complementary spaces. For example, the basis of different negative M values and the matrices S DMEs with total number of electrons ne <ns, the num- of the operator Jˆ are setup in these smaller DME ba- berofsitesinthesystemblock,arerelatedtoDMEswith sis spaces. The eigenstates of this operator serve as the ne >ns by – new DMEs for obtaining one-to-one correspondence of the DMEs under the symmetry operations. Jˆ µ (n >n )=µ (n <n ) (3) L i s e j s e (iv) {µi(ns =ne,MS =0)}: This particular set of and similarly the DMEs with M > 0 are related to eigenvectors span a vector space which on application S DMEs with MS <0 by, by both JL and PL should map into itself. Hence, lin- ear combinations are found out which are eigenvectors PˆLµi(MS <0)=µj(MS >0). (4) of JL with eigenvalues (−1)ζJ, ζJ = 0 or 1. Using J- eigenvectorswithaparticulareigenvalue(−1)ζJ,another Hence,westartwiththefollowingsetofDMEofthesys- setoflinearcombinationsareformedwhichareeigenvec- tem block which satisfy the criteria {µ (n ≥ n ,M ≤ i s e S torsofP , witheigenvalues(−1)ζP, ζ =0or1. Tothis 0)}. L P end, we first construct the matrix representation of Jˆin We consider here only the left block, as the algorithm thebasisofDMEswithn =n andM =0. We obtain will be same for the right block as well. This particular s e S the eigenvectors of Jˆin this space by diagonalizing the set of eigenstates can be segregated into four groups – (i){µi(ns >ne,MS <0)}: ApplicationofJL, PL and mwhaitcrhixt.hWe emtahternixuosfePˆthiesesiegtenuvpe.ctTohresorfesJˆulatsintgheeibgeansivseicn- theirproductontheseeigenstatesgivethecorresponding tors of Pˆ replace the DMEs in this space and they will symmetry counterparts. As J and P may not com- L L be employed in symmetry adaptation. mute, we only considered the product P J ; the order L L While imposing the cut-off in the DMEs, we only con- of operationsof the two operatorswill become irrelevant sider the subspace of DMEs with n ≤ n and M ≤ 0. in the direct product basis space of a half-filled system, e s S Since the DMEs related to these states by the operation where J and P of the superblock commute. of Jˆ, Pˆ and PˆJˆwill also have the same eigenvalues, it is Jˆ µ (n > n ,M < 0) = µ (n < n ,M < 0) ensuredthatallDMEsaboveacut-offinthedensityma- L i s e S iJ s e S trixeigenvalueareincluded. Usuallythe cut-offimposed Pˆ µ (n > n ,M < 0) = µ (n > n ,M > 0) L i s e S iP s e S in the specified sector is about 0.4 times of the cut-off Pˆ Jˆ µ (n >n ,M <0)=µ (n <n ,M >0) intended for the full density matrix. Since we have a L L i s e S iPJ s e S one-to-one correspondence of the DMEs within a block The DMEs obtained from µi by symmetry operations for the operations Jˆ, Pˆ and PˆJˆ, it is a simple matter to are indexed and stored as eigenvectors {µi}. Symmetry extendtheone-to-onecorrespondencetothedirectprod- correspondence in every pair of these four eigenstates is uctbasisstatesformedbytheDMEsoftheleftandright alsodeducedusingtherelationPˆLJˆL =(−1)(ns+ne)JˆLPˆL blocksandtheFockspacestatesofthenewlyaddedsites. along with Eqn. 2 and stored separately. The matrix Symmetry adapted direct product bases in an irre- elements of the symmetry operators are given by, ducible representation Γ can be generated by employing projection operator, which is given by – (Jˆ ) =(−1)ηJδ (Pˆ ) =(−1)ηPδ L i,j i,iJ L i,j i,iP 1 (Pˆ Jˆ ) =(−1)ηPJδ P(Γ)= hXχ(Γ,Ri)Rˆi (5) L L i,j i,iPJ i 5 whereχ(Γ,R )isthecharacterofthesymmetryoperator action: i Rˆ intheirreduciblerepresentationΓandhbeingtheor- deirofthegroup. OperationbyP(Γ)oneachofthebasis H = X t0(cˆ†i,σcˆj,σ+ H.C.)+Xǫinˆi statesinthedirectproductspaceproducesthesymmetry <i,j>,σ i (7) adapted linear combinations. However, if P(Γ) acts on U +X nˆi(nˆi−1)+XVij(nˆi−zi)(nˆj −zj) eachofthestatesinthedirectproductspace,wewillend 2 i i>j up with a linearly dependent symmetry adapted basis. In the earlier symmetry adaptation procedure15, weed- t is the nearest-neighbor hopping integral between 0 ing out linear dependence by Gram-Schmidt orthonor- bonded sites ‘i’ and ‘j’, ǫ is the site energy of ‘i’-th car- i malization technique was an extremely expensive step, bonatomwhileU istheon-siteCoulombrepulsionterm. computationally. Withpresenttechnique,wherewehave Theintersiteinteractionenergies,V areinterpolatedby ij a one-to-one correspondence between symmetry related Ohno scheme18 assuming C-C bond length of 1.4 ˚A. cˆ† i,σ states,thisturnsouttobeextremelysimple andcompu- and cˆ are the creation and annihilation operators of i,σ tationally veryefficient. Firstly we note that, one-to-one an electron with spin σ in the 2p orbital of the carbon z correspondence between the DMRG basis states under atomatsite ‘i’andnˆ = cˆ† cˆ isthe corresponding the symmetry operationsimplies thata givenbasis state i Pσ i,σ i,σ number operator. For carbon atoms, the standard PPP can appear only once (or not at all) in a given symme- parametersaret =−2.4eVandU =11.26eVwhilethe 0 try space. Therefore, in the symmetry adaptation step, site energy(ǫ)for unsubstitutedcarbonatomistakento we sequentially go through the list of basis states, apply be zero. z is the local chemical potential, expressed by i the projection operator and construct the symmetrized the the number of electrons at site ‘i’ which leaves the combinationalongwitheliminationofthepartnerswhich site neutral; for carbon atoms in a π-conjugated system, haveappearedinthesymmetrycombinationfromthelist z = 1. Substitution effect by donor or acceptor groups i of unsymmetrized basis states. This ensures that a ba- canbe modeledbyconsideringthesiteenergyǫ positive i sis state occurs at most once in the symmetrized states. or negative respectively. Taken together with the orthonormality of the unsym- As DMRG calculations are done in constant M -basis S metrized basis states, this ensures orthogonality of the space,contrarytothespin-adaptedbasis(Diagrammatic basis states generated by the projection operator. Us- valencebondbasis)calculations,labellingstatesbytheir ing these symmetry combinations, we can generate the spin value is ambiguous. The parity symmetry partially transformationmatrix S, which is D×D matrix where Γ resolves this as the singlet state lies in the even-parity D is the dimensionality of the unsymmetrized space and subspace while the triplet state lies in the odd-parity D is the dimensionality of the symmetrized space. Γ subspace. Though, higher spin states also lie in the TheHamiltonianofthesuperblockcanbetransformed even (odd) parity sub-spaces, they remain high in en- into symmetrized basis space by operating with the ma- ergy compared to singlet and triplet states in moderate trix S as: sized carbon-based systems and so can be neglected6. H˜ =S†H S (6) 2l+2 2l+2 C. Comparison with exact calculation results ThetransformedHamiltonianisthendiagonalizedtoget lowest-lying eigenvectors in different irreducible repre- The accuracy of our modified algorithm can be estab- sentations labeled as eA+, eB+, eA−, eB−, oA+, oB+, lished by comparing the results obtained for small sys- oA− andoB−,wheretheleft-handsuperscriptrepresents temswithexactPPPresults,likein16-sitepolyenechain characterunderP-symmetry,the rightsuperscriptisthe andtetracenemoleculewhichhas18sites. Thedifference character under J-symmetry and the letter A/B is the betweengroundstateenergyof16-sitepolyenechain,ob- characterunderC symmetryintheirreduciblerepresen- 2 tained by symmetrized DMRG technique, and the exact tationΓ. Thesingletliesinthe‘e’spacewhilethetriplet calculation result is plotted for four different cutoff val- lies in the ‘o’ space. The space ‘+’ refers to ‘covalent’ ues M′l in the sector {µL(n ≥ n ,M ≤ 0)} of the space while ‘-’ to the ‘ionic’ space. A(B) represents the i s e S sub-block in Fig. 1. In all the calculations, the energy space even (odd) under C symmetry. The ground state 2 values are accurate at least up to fifth decimal places is usually in eA+ while the optically connected space is when compared with exact calculation results. We have eB−. The lowest triplet lies in the oB+ space. also plotted the fluctuations in the DMRG energy about the exact energy with number of finite DMRG sweeps for a fixed M′l value, which depicts that the fluctuation B. Model Hamiltonian is negligible with number of sweeps and therefore, 2−3 sweeps will suffice for acceptable convergence of energy. ThemodelHamiltonianemployedforthestudyofperi- In Table I, PPP energies of the ground state (11A+), condensed polyacenes is the well known Pariser-Parr- lowesttwo-photonstate(21A+), lowestone-photonstate Pople (PPP) Hamiltonian17, which considers long-range (11B−)andofthelowesttripletstate(13B+)ofboth16- Coulombic interactionalongwith on-site Hubbardinter- site polyene chain and tetracene molecule are presented. 6 1.5e-06 {+ε}1 17{−ε} 3 2 x 21 20 ∆E1.29ee--0067 ∆Esweep33..311.241eee---000777 541 2236311462157118051691134 112892 2201 x 46 54675 1832111321151110114 123212 181179 y 6e-07 1 2nswe3ep 4 {7−ε}8 9 24{+ε2}3 y 7 8 99 1014 15 16 3 1 23 3e-07 y 4 4 5 2 6 22 24 3 10 6 25 7 14915 1116 21 20 25 26 0 150 225 M’l 300 375 71 813 1217 1819 27 x 8 12 28 9 31 29 FIG. 1. (Color online) Deviation of ground state energy cal- 11 10 32 30 culated by symmetrized DMRGtechniquefrom exact energy valuefor differentcutoffM′l ofthedensitymatrix eigenstate FIG.2. (Coloronline)Schematicdiagramsofcoronene,1,12- space. The numberof finite DMRG sweeps (2) is fixed in all benzoperylene and ovalene with site (black and bold) and the cases. Inset: Fluctuation of DMRG ground state energy bond (red and italics) indices. The sites of substitution in with respecttoexactenergy valuewithincreasing numberof substitutedcoroneneareindicatedincurlybrackets;+ǫ(sites finite DMRG sweeps. The number of density matrix eigen- 1 and 24) represents a donor site while −ǫ (sites 8 and 17) states retained, M′l, is 225. represents an acceptor site. III. RESULTS AND DISCUSSION M′l is kept at 225 for 16-site polyene chain and 425 for tetracene molecule, thus total number of DMEs retained We have studied a few low-lying states of coronene, for the sub-block is ∼ 550 and ∼ 1000 respectively at ovalene and 1,12−benzoperylene in different symmetry each step of DMRG calculation. Full CI calculations, spaces considering ∼1000 DMEs in PPP model and the employingdiagrammaticvalencebondmethod(DVB)19, energy gaps are tabulated in Table. II and III. We have shows excellent agreementwith the results obtained and alsocalculatedenergiesintheabovesymmetryspacesfor the upper bound of errorin energies ofdifferent states is substituted coronene,with donorandacceptorgroupsof of the order of 10−4. equal strength, set to 1.0 eV (|ǫ| = 1.0 eV). We have employed only a few of the existing spatial symmetries of the molecules, namely C symmetry axis perpendic- 2 ular to the molecular plane (in plane C axis in case of 2 1,12−benzoperylene)forourcalculations,asourDMRG calculations cannot incorporate more than one symme- TABLEI.Comparison oftheenergiesofgroundstate,lowest two-photonstate,lowestopticalstateandlowest tripletstate try axis. Being treated as planar molecules, with Wan- of 16-site polyene chain and tetracene, obtained by modified nier orbitals, the reflection in the molecular plane is a symmetrized DMRG algorithm and exact calculation. The trivial symmetry. Yet, we can label the eigenstates ac- exactcalculationsarecarriedoutusingdiagrammaticvalence cordingtodifferentirreduciblerepresentationsofthe full bond method. group based on transition dipole moment direction for excitations from the ground state. In case of coronene, System State Exact calculation DMRGcalculation wehaveconsidereditssubgroupD ratherthanthe full 2h Polyeneof 11A+ −33.55575 −33.55575 group D for labelling the eigenstates. The lowest op- 6h 16 sites 21A+ −32.03385 −32.03385 tical states obtained may not be the states with high- est transition dipole moment, which will be prominent 11B− −30.43291 −30.43291 inUV-visiblespectroscopybutcorrespondstothelowest 13B+ −32.87993 −32.87993 energystateoftheappropriatesymmetrysubspace;how- Tetracene 11A+ −43.69083 −43.69076 ever,theresultsobtainedareingoodagreementwithgas phase molecular spectroscopic data available for above molecule 21A+ −40.55749 −40.55725 molecules8, suggesting that the states obtained are also 11B− −40.51439 −40.51419 the ones with considerable transition dipoles in the en- ergy range of the experiment. 13B+ −42.47046 −42.47038 The relative ordering of the energy levels are closely associated with the effective correlation strength in lin- 7 TABLE II. Lowest energy states in coronene and ovalene with the energy gaps in units of eV. Although coronene has D 6h symmetry, here the energy states are labelled according to Abelian D symmetry representations. The transition dipole 2h moment from ground state totheexcited states (µ ) along specific axes are also specified in thelast two columns in units tr,x/y of Debye. Coronene Natureof the state State level Energy Gap µ (D) µ (D) tr,x tr,y Two-photon (21A+/11B+)a 3.97 0.00 0.00 g 1g (31A+/21B+) 4.09 0.00 0.00 g 1g (41A+/31B+) 5.08 0.00 0.00 g 1g Optical 11B− 4.83 0.81b 8.43 2u 11B− 4.87 7.51 0.49 3u Triplet (13B+/13B+) 2.35 0.00 0.00 2u 3u (23B+/23B+) 3.00 0.00 0.00 2u 3u (33B+/33B+) 3.02 0.00 0.00 2u 3u (13A+/13B+) 3.35 0.00 0.00 g 1g Dark states (11B+/11B+) 2.82 0.00 0.00 2u 3u (x1B−/ x1B−)c 3.88 0.28 1.74 2u 3u (11A−/11B−) 4.73 0.00 0.00 g 1g Ovalene Two-photon (21A+/11B+) 3.10 0.00 0.00 g 1g (31A+/21B+) 3.75 0.00 0.00 g 1g (41A+/31B+) 5.48 0.00 0.00 g 1g Optical 11B− 3.32 0.29 6.58 2u 21B− 4.53 0.33 0.58 2u 11B− 4.64 7.28 0.22 3u Triplet (13B+/13B+) 1.49 0.00 0.00 2u 3u (13A+/13B+) 2.77 0.00 0.00 g 1g (23A+/23B+) 2.80 0.00 0.00 g 1g (23B+/23B+) 2.82 0.00 0.00 2u 3u Dark states (11B+/11B+) 2.65 0.00 0.00 2u 3u (11A−/11B−) 4.06 0.00 0.00 g 1g (21A−/21B−) 4.69 0.00 0.00 g 1g a Probablelabelsforstates whichcannot beidentifieduniquelyaregivenwithinparentheses. b Non-zerovalueoftransitiondipolemomentalongpolarizationdirectionforbiddenbysymmetryisanartifactofthecalculations as averagedensitymatrices,calculated fromeigenstates ofdifferentsymmetrysubspaces, areemployedtodeterminethetransitiondipole moment. However,theerrorsarenegligibleasintensitiesdependonthesquareofthetransitiondipolemoment. c ximpliesthatthelevelnumberingisunknown. ear π−conjugated systems. In extended systems effec- based on one-particle bandwidth which are similar in all tive correlation strength can be thought of as the ratio themoleculeswehavestudied. InPPPorextendedHub- of the on-site PPP correlation to the bandwidth of the bardmodel,theon-sitecorrelationstrengthtoafirstap- one-particle spectrum. However, in molecules with dis- proximation is expressed by U ∼ (U −V ), where PPP 12 crete excitation spectrum, it is difficult to use the width V is the nearest-neighbor PPP term. Our calculations 12 oftheenergyspectruminplaceofthebandwidth. Thus, yield the lowest two-photon gap in ovalene molecule to it is difficult to a prioripredictthe orderingofthe states be almost twice that of the lowest spin gap, which is a 8 TABLE III. Lowest energy states in 1,12-benzoperylene and substituted coronene with the energy gaps in units of eV. The transition dipole moment from ground state to theexcited states (µ ) along specific axesare also specified in thelast two tr,x/y columns in unitsof Debye. 1,12-Benzoperylene Natureof the state Statelevel Energy Gap µ (D) µ (D) tr,x tr,y Two-photon 21A+ 2.99 0.00 0.00 1 31A+ 4.26 0.00 0.00 1 41A+ 4.58 0.00 0.00 1 Optical 11B− 3.72 0.00 3.17 1 21B− 4.71 0.00 3.83 1 31B− 4.83 0.00 5.20 1 Triplet 13B+ 2.84 0.00 0.00 1 23B+ 3.65 0.00 0.00 1 33B+ 4.26 0.00 0.00 1 Substitutedcoronene Two-photon 21A 4.01 0.00 0.00 g 31A 4.81 0.00 0.00 g 41A 4.90 0.00 0.00 g Optical 11B 2.90 0.00 0.23 u 21B 3.87 0.12 1.46 u 31B 5.18 0.17 7.52 u 41B 5.97 4.10 0.59 u Triplet 13B 2.29 0.00 0.00 u 23B 2.92 0.00 0.00 u 33B 3.48 0.00 0.00 u characteristic feature of regular polyene systems where al.13. The two-photon gaps reported earlier are 3.96 eV two-photon state is predominantly two triplet excita- and 3.03 eV respectively for these systems, against 3.97 tions. Theothertwomoleculesandsubstitutedcoronene, eV and 3.10 eV, found in our calculations. The triplet on the other hand, do not show this signature. There- gapsearlierreportedare2.38eVand1.57eVforcoronene fore, it can be suggested that the effective electronic and ovalene respectively, which are close to 2.35 eV and correlation strength in ovalene molecule is comparable 1.49eV obtainedby us. However,the lowestone-photon to regular polyene systems while the π− electrons in gaps and relative positions of the lowest optical states other molecules are less correlated. The ordering of low- withrespecttothelowesttwo-photonstatesdonotagree lying singlet excited states in coronene and ovalene, al- wellwiththe previousstudy. Incoronene,the lowestop- though, are in accordance with those in strongly cor- tical gap reported earlier is 4.11 eV while we have found related systems; the lowest optical states lie above the two nearly degenerate excitations at higher energies of lowest two-photon states in both the systems. However, values, 4.83 and 4.87 eV. A low-lying state at an energy 1,12-benzoperyleneshowssomepeculiartrendasthelow- of3.88eV,whichhasasmalltransitiondipoleappearsin est two-photon state remains close to the lowest triplet thesamesubspaceasoftheopticalstates. Thisisproba- state and quite low in energy compared to 11B−, show- blybecausewehaveincorporatedonlyoneC symmetry 1 2 ing that in this system, we cannot view the 21A+ state axisinourcalculations. InDMRGcalculations,forcom- 1 as made up oftwo triplets generatedindifferent partsof puting transition dipole moments we have used average the molecule. density matrices, obtained from eigenstates of different The lowest two-photon gaps along with the lowest symmetry subspaces and hence the spatial symmetry is tripletgapscalculatedincoroneneandovaleneareinex- notretained. Thiscouldleadtoasmalltransitiondipole cellentagreementwiththeearlierreportbyAryanpouret moment as an artifact for this state (3.88 eV) although 9 0.24 TseActBoLrEanIVd.inLoMweSst=tri0plseetcsttoarteweitnherχgiPes=obt−a1inefodrincoMroSne=ne1, n n>αβii00..1281 Coronene ovalene and 1,12-benzoperylene in units of eV. < 0.15 0 1 2 3 4 5 6 7 8 9 10 11 12 M =1 M =0,χ =−1 0.24 S S P > β 0.21 Coronene −58.67378 −58.56182 nαi0.18 GTwroou-pnhdo sttoante state Ovalene −80.50885 −80.37843 <ni0.15 Ovalene OTrpitpicleatl ssttaattee 0.12 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 1,12-Benzoperylene −52.54237 −52.32442 0.24 > β 0.21 ni α 0.18 <ni0.15 1,12-Benzoperylene 0.12 0 1 2 3 4 5 6 7 8 9 10 11 Site index (i) this state correspond e−h allowed subspace and if spa- tialsymmetryisnotretainedstrictlywewillhaveaweak FIG.3. (Coloronline)Doubleoccupancyplotwithsiteindex transition dipole. On the other hand, the 2.82 eV exci- for coronene (top panel), ovalene (middle panel) and 1,12- tation in coronene is optically forbidden as it belongs to benzoperylene(bottom panel). Asall molecules aresymmet- the same e-h symmetry subspace as the ground state. ricaboutC axis,onlyhalfofthemolecularsitesareplotted, 2 This argument is supported by the fact that this state the other half being symmetry equivalent to it. The color acquires some intensity on breaking the e− h symme- schemefordoubleoccupancyindifferenteigenstatesaresame try by introducing substituents, as can be seen from Ta- inallthreepanelsandareasfollows: groundstate(blackfilled ble. III. Thus it appears that as in benzene, in coronene circles),lowesttwo-photonstate(redfilledsquare),lowestop- ticalstate(greenfilledtriangles)andlowesttripletstate(blue the lowest degenerate excitations are pushed up in en- open squares). Doubleoccupancy in theother partnerof the ergy while the higher energy non-degenerate states cross degenerate optical states in coronene is shown by green bro- these excitations on incorporating electron-electron in- kenlinewithopentriangles. Linesconnectingdatapointsare teractions. In case of ovalene, we have found the lowest given only as a guide to theeye. opticalgaptobegreaterthanthelowesttwo-photongap while reverse ordering of states has been reported ear- lier. The earlier reported energy gap between the lowest one-andtwo-photonstatesaswellasthatobtainedbyus Hilbert space becomes half of that in M =1 sector cal- S are relatively small. The earlier study used interaction culation. As DMRG is a variational technique, accuracy parameters optimized to yield experimental gaps in con- increases with the dimension of the Hilbert space and densed phase while our study employsparameterswhich hence,theM =1sectorcalculationsaremoreaccurate. S are applicable to isolated molecules. In condensed phase Spin gap is also a good measure of electronic correla- the one-photon state will shift more towards the longer tionstrength. ForaPPPchain,thespingapcanbeesti- wavelengthregimethantwo-photonstate,theformerbe- matedfromtheHeisenbergmodel,withanantiferromag- ing ionic in nature. Thus we expect that the one-photon netic J ∼ 4t2/U . This implies that with increasing PPP state will be below the two-photon state for ovalene in effectivecorrelationstrength,thegapbetweenthelowest solution. The polarization in ovalene in the lowest one- singletstateandthelowesttripletstatereduces. Accord- photon state is along the short axis while the higher en- ing to this line of argument, ovalene has the highest ef- ergyopticalexcitationis longaxispolarized. This is due fectivecorrelationstrengthwhile1,12-Benzoperylenehas to the fact that excitons with charge separation along thelowestamongthemoleculessitedinTable. IIandIII. the short axis are more stable than those with separa- Substitution by donor-acceptor groups does not seem tion along the long axis. Indeed, similar polarization for to have appreciable effect on correlation strength, al- dipole excitations is also found in polyacenes6. though, the lifting of electron-hole symmetry allows op- Thelowesttripletstateenergyineachofthemolecules tical transition to states which are labelled as dipole- can be calculated in two ways while employing both C forbidden in unsubstituted molecules. The extent of 2 and J symmetries – either by calculating the lowest en- mixing ofdifferent symmetry states ofthe unsubstituted ergy state in M =1 sector or by calculating the lowest system due to substitution depends on the strength of S energy state in the M = 0 space with χ = −1. The the donor-acceptor groups. In the present case, the S P energies obtained by both calculations are given in Ta- dipole-forbidden 11B+/11B+ state in unsubstituted 2u 3u ble. IV. It can be observed that the energies obtained coronenebecomes symmetry alloweddue to substitution in M = 1 sector calculations are lower. This can be andcorrespondstothelowestopticalstateinsubstituted S attributed to the larger Hilbert space of the M = 1 coronene. S state for a given DMRG cut-off compared to the Hilbert Itisevidentfromabovethattheeffectiveelectroncor- space dimension of the M =0 space with χ =−1. In relation strength is not controlled by the symmetry of S P fact,theparitysymmetryalmostbisectsthecompletedi- themolecule;whilecoroneneisthemoleculewithhighest rectproductspaceandtherefore,the effective size ofthe symmetry (D ), substitution by donor-acceptor groups 6h 10 0.1 0.1 of the substituents considered in this study. 0.08 Coronene 0.08 >i0.06 0.06 We have tabulated the bond order values of different <sz0.04 0.04 bonds,asindexedinFig. 2inlowestenergystatesofeach 0.02 0.02 symmetry subspaces, for all the four molecules studied 0 0 0 2 4 6 8 10 12 14 16 18 20 22 24 0.3 0.3 here. The bond order values are related to bond lengths >i0.2 Ovalene 0.2 asindicatedbyCoulson20;bondswithhigherbondorder sz0.1 0.1 valuehasmoredoublebondcharacterwhilelowvaluede- < 0 0 pictsthatthebondisprimarilyasinglebond. Thereper- 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 cussionofsimilarbondordervaluesfordifferentbondsin 0.2 0.2 >i0.01.51 1,12-Benzoperylene 00..115 coronene establishes D6h symmetry of the ground state. <sz0.05 0.05 This particular symmetry is not retained in the excited 0 0 statesindicatingdistortedequilibriumstructuresinthese -0.050 2 4 6 8 10 12 14 16 18 20 22 -0.05 states. In excited states the bonds are symmetric about Site index (i) twoin-planeC axes,whichbisectsthemolecule. Intwo- 2 photon state, a quinonoid structure emerges at the cen- FIG. 4. Spin density plot for coronene (top panel), ova- tralpartof the molecule, while in the doubly degenerate lene (middle) and 1,12-benzoperylene (bottom) in thelowest opticalstates,whichhavealmostsimilarbondordersand tripletstate. Thedatapointsareconnectedbyalineforsake in the lowest triplet state the central benzene ring gets of smooth vision. squeezed compared to ground state and its connectivity to the peripheralbenzene rings becomes weaker. Similar structures appear even in substituted coronene. Small variationinbondordersforbondsconnectedtodonoror lowersthesymmetrytoC ,yet,theeffectivecorrelation 2h acceptorsitesarequite expectedasthe effective hopping strengthinboththesystemsasindicatedbythespingaps amplitudes for bonds involving these sites will be differ- and relative positions of the energy levels, are compara- ent from the rest of the molecule. In ovalene, the edge ble. Ovalene is also a lower symmetry molecule (D ) 2h bonds switch their character from double bond to sin- than coronene, but appears to have large effective cor- gle bond and vice versa in two-photon state and triplet relation strength. On the contrary, 1,12-benzoperylene state, relative to those in the ground state. In optical (C )beingofsimilarsymmetrytosubstitutedcoronene, 2v state, the edge bonds become more or less of same char- seems to have rather low effective correlation strength acterasallthebondordersbecomenearlyequal. Yet,the compared with the latter. mostnotablebondvariationinovaleneisforthebonds11 Spindensityatdifferentsitesinthelowesttripletstate, and 12 and their symmetry partners, which also switch which is the expectation value of hSˆzi in this state is i their character like the edge bonds, but in all excited plotted for different molecules in Fig. 4. In the coronene states; other bonds in the inner part of the molecule re- moleculewenotedthatthespindensityishighestatsites mainalmostinvariant. However,the biggestvariationin 1, 8, 17 and 24 which are all related by symmetry of the bondorderisobservedinthe lowesttripletstate of1,12- molecule. Thefractionaldoubleoccupanciesattheabove benzoperylene when comparedto the groundstate bond mentioned sites are small (Fig. 3), which is consistent orders. The Pauliblockingatthe sites 2and21converts with the higher spin densities at these sites. However, the bond 1 from a single bond to a double bond. Weak- the bond order pattern along the periphery has period ening of neighboringbonds can also be attributed to the four andcorrespondto ( ==——== ) whichis un- n Pauli blocking. Bond orders of most of the peripheral usual as one would expect a quinonoidal structure from andringbondsbecomenearlyequalwhilebondsparallel thespindensitydata. Onthecontrary,thespindensities to bond 1 remain approximately same in all states. In are well-localized at sites 1 and 32 in ovalene and at 2 the two-photon state, overall expansion of the structure and 21 in 1,12-benzoperylene (Fig. 2 and 4). In ovalene is observed while in the optical state, most significant and 1,12-benzoperylene,there are many sites with nega- variation in bond order occurs in the lowermost ring as tive spin densities unlike in the case of coronene. Again, shown in Fig. 2. The bond order of the weakest bond of we noted that the fractional double occupancy is least the molecule increases both in the two-photon state and at sites with high positive spin densities (Fig. 3). The the opticalstate also,but rathersmall comparedto that peripheral bond order in ovalene has period four with in the triplet state. kinks at the high spin density sites at the middle of the The charge gap of a system is defined as the energy molecule. Bond 11 and its symmetry partners (Fig. 2) required to remove an electron from a neutral molecule intheinteriortwobenzeneringsareratherweakascom- andplacingitonanotherneutralmolecule. This process pared to that in the ground state and we can expect gives rise to two chargedspecies of different charges and puckering of the structure in the triplet state. In 1,12- can be represented by the reaction6: benzoperylene also we find bond order pattern with pe- riod four on the periphery if we exclude bonds involving 2M →P++P− the high spin density sites. In substituted coronene, the spin density distribution is not sensitive to the strength The charge gap for a system can be calculated by ob-

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.