05.02.2013 Views

Calculating Intermolecular Charge Transport Parameters in ...

Calculating Intermolecular Charge Transport Parameters in ...

Calculating Intermolecular Charge Transport Parameters in ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

A thesis entitled<br />

<strong>Calculat<strong>in</strong>g</strong> <strong>Intermolecular</strong> <strong>Charge</strong> <strong>Transport</strong><br />

<strong>Parameters</strong> <strong>in</strong> Conjugated Materials<br />

James Kirkpatrick<br />

Submitted for the degree of Doctor of Philosophy<br />

of the University of London<br />

Imperial College of Science, Technology and Medic<strong>in</strong>e<br />

February 2007


Contents<br />

1 Introduction 4<br />

1.1 Why is a Theoretical Study of <strong>Charge</strong> <strong>Transport</strong> Necessary to the Design<br />

of Better Devices? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4<br />

1.2 Conjugated Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6<br />

1.3 <strong>Charge</strong> <strong>Transport</strong> <strong>in</strong> Conjugated Materials . . . . . . . . . . . . . . . . . 8<br />

2 Background 17<br />

2.1 Marcus Theory <strong>in</strong> the Non-Adiabatic High Temperature Limit . . . . . . 18<br />

2.2 Reorganisation Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . 24<br />

2.3 Transfer Integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26<br />

2.3.1 Note on Systems with Degenerate Orbitals . . . . . . . . . . . . 28<br />

2.3.2 Splitt<strong>in</strong>g of the Molecular Orbitals . . . . . . . . . . . . . . . . . 29<br />

2.4 Site Energy Difference . . . . . . . . . . . . . . . . . . . . . . . . . . . 30<br />

2.4.1 Electrostatic Interactions <strong>in</strong> Molecules . . . . . . . . . . . . . . . 31<br />

2.4.2 Inductive Interactions <strong>in</strong> Molecules . . . . . . . . . . . . . . . . 35<br />

2.5 The Gaussian Disorder Model and Related Methods . . . . . . . . . . . . 36<br />

2.5.1 Computational Methods . . . . . . . . . . . . . . . . . . . . . . 37<br />

2.5.2 Distribution of <strong>Parameters</strong> of the Transfer Rates . . . . . . . . . . 40<br />

2.5.3 Predictions and Results . . . . . . . . . . . . . . . . . . . . . . . 43<br />

2.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46<br />

3 Development and Verification of Methods for <strong>Calculat<strong>in</strong>g</strong> <strong>Charge</strong> <strong>Transport</strong><br />

<strong>Parameters</strong> 53<br />

3.1 Projective Method for Calculation of Transfer Integrals . . . . . . . . . . 54<br />

i


3.2 Molecular Orbital Overlap Method for Calculation of Transfer Integrals . 57<br />

3.3 Calculation of Site Energy Difference from Electrostatic Interactions . . . 59<br />

3.3.1 Ethylene: Neutral Systems . . . . . . . . . . . . . . . . . . . . . 61<br />

3.3.2 Ethylene: <strong>Charge</strong>d Systems . . . . . . . . . . . . . . . . . . . . 66<br />

3.3.3 Pentacene . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70<br />

3.4 Comparison of Different Methods of <strong>Calculat<strong>in</strong>g</strong> Transfer Integrals . . . . 73<br />

3.4.1 Pentacene . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73<br />

3.4.2 Hexabenzocoronene . . . . . . . . . . . . . . . . . . . . . . . . 78<br />

3.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82<br />

4 Reorganisation Energy <strong>in</strong> PolyPhenylenev<strong>in</strong>ylene: Polaron Localisation 85<br />

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86<br />

4.2 Defect-free PPV . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89<br />

4.3 Torsional Defects <strong>in</strong> PPV . . . . . . . . . . . . . . . . . . . . . . . . . . 92<br />

4.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94<br />

5 Ambipolar <strong>Transport</strong> <strong>in</strong> PCBM 98<br />

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98<br />

5.2 Transfer Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99<br />

5.3 Reorganisation Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . 101<br />

5.4 Monte Carlo Simulations . . . . . . . . . . . . . . . . . . . . . . . . . . 103<br />

5.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108<br />

6 <strong>Charge</strong> <strong>Transport</strong> <strong>Parameters</strong> for Randomly Oriented Pairs of Dialkoxy Poly-<br />

paraphenylenev<strong>in</strong>ylene and Triarylam<strong>in</strong>e Derivatives 111<br />

6.1 Dialkoxy PPV Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112<br />

6.1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112<br />

6.1.2 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116<br />

6.1.3 Results and Comment . . . . . . . . . . . . . . . . . . . . . . . 118<br />

6.2 Spiro L<strong>in</strong>ked Triarylam<strong>in</strong>e Derivatives . . . . . . . . . . . . . . . . . . . 123<br />

6.2.1 Method and Results . . . . . . . . . . . . . . . . . . . . . . . . 124<br />

6.3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126<br />

ii


7 Simulation of <strong>Charge</strong> <strong>Transport</strong> <strong>in</strong> Assemblies of Hexabenzocoronenes and<br />

Fluorene Oligomers 130<br />

7.1 PFO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131<br />

7.2 HBC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141<br />

7.3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147<br />

8 Conclusion and Future Work 151<br />

iii


List of Figures<br />

1 Probability density distribution for receiv<strong>in</strong>g an email from Jenny. 41%<br />

of emails is received after five o’clock! . . . . . . . . . . . . . . . . . . . xiv<br />

1.1 Resonant Lewis structures for benzene with s<strong>in</strong>gle and double bonds (left)<br />

and representation of the aromatic structure (right). . . . . . . . . . . . . 7<br />

1.2 Lewis structure of a segment of PPV. . . . . . . . . . . . . . . . . . . . . 8<br />

1.3 Lewis structure of a charged segment of PPV before and after charge mi-<br />

gration. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8<br />

1.4 Eigenvalues of equation 1.1 as a function of the reaction coord<strong>in</strong>ate Q. . . 10<br />

1.5 Time evolution for three systems with the same reorganisation energy,<br />

the same angular frequency ω = 5 10 12 s −1 and three different transfer<br />

<strong>in</strong>tegrals J: 1 meV (top), 50 meV (middle) and 300 meV(bottom). . . . . 12<br />

1.6 Time evolution for three systems with the same reorganisation energy, the<br />

same transfer <strong>in</strong>tegral J = 10 meV and three different angular frequencies<br />

ω: 10 12 s −1 (top), 5 10 12 s −1 (middle) and 10 14 s −1 (bottom). . . . . . . . . 14<br />

2.1 Schematic view of a potential energy surface for asymmetric charge trans-<br />

port. λ denotes the reorganisation energy, ∆E the difference <strong>in</strong> energy<br />

between reactants and products <strong>in</strong> their relaxed geometries and ∆Q is the<br />

equilibrium reaction coord<strong>in</strong>ate for the products. . . . . . . . . . . . . . . 20<br />

2.2 Pictorial representation of the vectors discussed <strong>in</strong> deriv<strong>in</strong>g equation 2.26 31<br />

2.3 Pictorial representation of the vectors discussed <strong>in</strong> deriv<strong>in</strong>g equation 2.32 33<br />

iv


2.4 A schematic of one dimensional transport through an energetically and<br />

positionally disordered medium. This figure represents disorder <strong>in</strong> en-<br />

ergy (vertical placement of the transport levels) and <strong>in</strong> transfer <strong>in</strong>tegrals<br />

(thickness of arrows) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37<br />

2.5 σ versus the value of the dipole moment for five different small conju-<br />

gated molecules (right panel)[44], σd obta<strong>in</strong>ed from equation 2.41 (left<br />

panel) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42<br />

2.6 Pictorial representation of correlation <strong>in</strong> a 100x100x100 cubic lattice of<br />

randomly oriented dipoles. The diameter of the spheres is proportional to<br />

the magnitude of the dipole <strong>in</strong>teraction at that po<strong>in</strong>t, whereas the colour<br />

of the spheres represents the sign of the potential [45] . . . . . . . . . . . 43<br />

2.7 Energy distribution of charges as a function of time after generation [46]. 44<br />

2.8 Schematic of different routes from A to B <strong>in</strong> the direction of the field [46] 45<br />

3.1 A pair of ethylene molecules at a separation of 5 Å . Also shown is the<br />

axis around which one of the two molecules will be rotated to generate<br />

the geometries for which ∆E is calculated. . . . . . . . . . . . . . . . . 62<br />

3.2 Site energy difference for a pair of neutral ethylene molecules as a func-<br />

tion of rotation of one of the molecule about the the C=C bond calcu-<br />

lated by projection of PW91 MOs (circles) and calculated from the differ-<br />

ence <strong>in</strong> quadrupole-monopole <strong>in</strong>teractions calculated classically from the<br />

electrical moments calculated at the PW91 level on an isolated ethylene<br />

molecule (solid l<strong>in</strong>e). . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63<br />

3.3 Site energy difference for a pair of neutral ethylene molecules as a func-<br />

tion of rotation of one of the molecules about the the C=C bond calculated<br />

by projection of PW91 MOs (circles) and calculated from the difference<br />

<strong>in</strong> quadrupole-monopole <strong>in</strong>teractions corrected by the quadrupole <strong>in</strong>duced<br />

dipoles (solid l<strong>in</strong>es). Polarisabilities and electrical moments were calcu-<br />

lated at the PW91 level. . . . . . . . . . . . . . . . . . . . . . . . . . . . 65<br />

v


3.4 Distance dependence of the contribution to ∆E from the <strong>in</strong>duced dipole<br />

<strong>in</strong> a pair of ethylene molecules as a function of separation. The solid<br />

l<strong>in</strong>e corresponds to the contribution from the dipole moment <strong>in</strong>duced by a<br />

quadrupole (i.e. <strong>in</strong> a neutral system), whereas the dotted l<strong>in</strong>e corresponds<br />

to that from the dipole moment <strong>in</strong>duced by a monopole (i.e. <strong>in</strong> a charged<br />

system). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67<br />

3.5 Diagram show<strong>in</strong>g the dipoles <strong>in</strong>duced <strong>in</strong> two ethylene molecules oriented<br />

at right angles to each other. (a) Case of po<strong>in</strong>t quadrupoles (neutral cal-<br />

culation). (b) Case of po<strong>in</strong>t charges (charge calculation). For discussion,<br />

please see the text. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68<br />

3.6 Dipoles <strong>in</strong>duced on each molecule <strong>in</strong> a pair of neutral ethylene molecules<br />

(upper pannel) and <strong>in</strong> a positively charged pair (lower panel). Notice that,<br />

as shown <strong>in</strong> figure 3.5, <strong>in</strong> a charged system the <strong>in</strong>duced dipoles are larger<br />

than <strong>in</strong> the neutral case, but are <strong>in</strong> opposite directions. . . . . . . . . . . 70<br />

3.7 Two cofacial pentacene molecules and the axis around which one will be<br />

rotated and translated. The <strong>in</strong>set shows the chemical formula of pentacene. 71<br />

3.8 ∆E for pairs of pentacene molecules, calculated us<strong>in</strong>g projection of the<br />

PW91 results (red po<strong>in</strong>ts) and electrostatic results (diamonds). The elec-<br />

trostatic results are calculated us<strong>in</strong>g monopole monopole <strong>in</strong>teractions only<br />

(squares) and us<strong>in</strong>g all <strong>in</strong>teractions up to quadrupole-quadrupole (circles). 72<br />

3.9 Absolute value of the hole transfer <strong>in</strong>tegral as a function of displacement<br />

along the x axis of one of the two pentacene molecules calculated by<br />

projection of ZINDO orbitals (circles) and by the orbital splitt<strong>in</strong>g model<br />

(triangles). The length of the molecule is 14.1 Å . . . . . . . . . . . . . . 74<br />

3.10 HOMO for pentacene calculated at the ZINDO level. . . . . . . . . . . . 74<br />

3.11 Absolute value of the transfer <strong>in</strong>tegral for a pair of pentacene molecules as<br />

a function of rotation around the x axis. The transfer <strong>in</strong>tegral is calculated<br />

by projection of the ZINDO orbitals (circles) and by the orbital splitt<strong>in</strong>g<br />

method (triangles). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75<br />

vi


3.12 Absolute value of the transfer <strong>in</strong>tegral for a pair of ethylene molecules<br />

as a function of rotation angle. All curves were obta<strong>in</strong>ed by projective<br />

method applied to PW91 calculations; the different curves are obta<strong>in</strong>ed<br />

by us<strong>in</strong>g different basis sets. . . . . . . . . . . . . . . . . . . . . . . . . 76<br />

3.13 Absolute value of the transfer <strong>in</strong>tegral as a function of displacement along<br />

the x axis of one of the two pentacene molecules. The two curves show<br />

the results from the projective method us<strong>in</strong>g the PW91 functional and<br />

the ZINDO Hamiltonian. The PW91 calculation were performed with a<br />

TVZP basis set. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77<br />

3.14 Two cofacial HBC molecules and the cartesian axis used <strong>in</strong> the text. The<br />

<strong>in</strong>set shows the chemical formula of HBC. . . . . . . . . . . . . . . . . . 78<br />

3.15 Contour plots of the effective transfer <strong>in</strong>tegral as a function of rotation<br />

about the x and z axis. The molecules are first rotated about the z axis,<br />

then rotated about the x axis <strong>in</strong> such a way as to keep the m<strong>in</strong>imum dis-<br />

tance of approach between the molecules equal to 3.5 Å . . . . . . . . . . 79<br />

3.16 Effective transfer <strong>in</strong>tegral and its components as a function of rotation<br />

about the z axis calculated us<strong>in</strong>g the projective (PRO) and MOO (MOO)<br />

methods. The molecules are at a distance of 3.5 Å . . . . . . . . . . . . . 80<br />

3.17 Countour plots of the dependence on x-y displacement of two molecules<br />

of HBC. The two molecules are at a distance of 3.5 Å. The maximum<br />

radius of the molecule is 6.8 Å. . . . . . . . . . . . . . . . . . . . . . . . 82<br />

4.1 Two monomers of PPV and the scheme of the labels used <strong>in</strong> the rest of the<br />

chapter: ph1...phn will label the phenylene units and vi1...vi2 will label<br />

the v<strong>in</strong>ylene units. The first monomer also has the carbons with bonds to<br />

hydrogen labeled 1 to 6. . . . . . . . . . . . . . . . . . . . . . . . . . . 86<br />

4.2 Experimental and modelled ENDOR spectra of PPV with an applied mag-<br />

netic field parallel and perpendicular to the polymer cha<strong>in</strong> axis. The figure<br />

is from reference [2]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87<br />

vii


4.3 Sp<strong>in</strong> density of a polaron on a PPV cha<strong>in</strong> calculated with a PPP Hamil-<br />

tonian. Sites B and C’ correspond to sites 2 and 4 from figure 4.1, sites<br />

B’ and C correspond to sites 1 and 3 from figure 4.1 and sites E and F<br />

correspond to sites 5 and 6 from 4.1. The figure is from reference [2]. . . 88<br />

4.4 Mulliken sp<strong>in</strong> density for different length oligomers of PPV. The x axis<br />

labels position along the cha<strong>in</strong>: the labels ph1, ph2, etc. label the first,<br />

second, etc. phenylene unit of the PPV. Between the phenylene units phn<br />

and phn + 1 lies the v<strong>in</strong>ylene unit v<strong>in</strong>. . . . . . . . . . . . . . . . . . . . 90<br />

4.5 Polaron localisation energy for different length oligomers of PPV calcu-<br />

lated with BHandHLYP. . . . . . . . . . . . . . . . . . . . . . . . . . . 91<br />

4.6 Atomic Mulliken sp<strong>in</strong> density for the a cha<strong>in</strong> of 12 units calculated at the<br />

BHandHLYP (top panel) and AM1 (bottom panel) level. . . . . . . . . . 93<br />

4.7 Mulliken total sp<strong>in</strong> density <strong>in</strong>tegrated on each phenylene and v<strong>in</strong>ylene<br />

unit as a function of cha<strong>in</strong> torsion for an oligomer of length eight. . . . . 94<br />

5.1 The seven unique pairs with greatest hole transfer <strong>in</strong>tegral extracted from<br />

a cluster of twenty molecules of PCBM from its crystal structure grown<br />

<strong>in</strong> chlorobenzene. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100<br />

5.2 Distance dependence of the electron (black circles) and hole (red squares)<br />

transfer <strong>in</strong>tegrals as a function of displacement from the crystal structure<br />

separation. The l<strong>in</strong>es show exponential fits with natural lengths of respec-<br />

tively 0.78 Å and 0.56 Å. . . . . . . . . . . . . . . . . . . . . . . . . . . 102<br />

5.3 Scatter plot of the average sp<strong>in</strong> density for each bonded pair of atoms <strong>in</strong><br />

a cation (circles) or <strong>in</strong> an anion (crosses) versus the percentage change <strong>in</strong><br />

bond length upon charg<strong>in</strong>g. . . . . . . . . . . . . . . . . . . . . . . . . . 104<br />

5.4 Sp<strong>in</strong> density for cation and anion radicals of PCBM. The left panel shows<br />

a front view of the molecules, the right panel a side view. With<strong>in</strong> each<br />

panel the molecule to the left is the anion radical and the molecule to the<br />

right is the cation radical. . . . . . . . . . . . . . . . . . . . . . . . . . . 104<br />

5.5 Zero field mobility for electrons (squares) and holes (circles) <strong>in</strong> PCBM<br />

doped polystyrene as a function of PCBM concentration. . . . . . . . . . 105<br />

viii


5.6 Distance dependance for the transfer <strong>in</strong>tegral for neighbour and nearest<br />

neighbour hopp<strong>in</strong>gs, extracted from the cluster from the crystal structure<br />

of PCBM discussed <strong>in</strong> the text. The transfer <strong>in</strong>tegral is plotted both as a<br />

function of the distance between centres of mass of the PCBM molecules<br />

(above) and as a function of the distance between the centres of the C60<br />

cages (below). This is done because the C60 cages are the electronically<br />

active parts of the molecules. . . . . . . . . . . . . . . . . . . . . . . . 107<br />

5.7 Experimental (full symbols) and modelled (empty symbols) temperature<br />

dependence for prist<strong>in</strong>e MDMO-PPV (triangles) 1:1 MDMO-PPV:PCBM<br />

(circles) and 1:2 MDMO-PPV:PCBM (squares). . . . . . . . . . . . . . . 107<br />

6.1 Chemical formula of the three dialkoxy PPV derivatives studied <strong>in</strong> this<br />

chapter. a) is di-MethoxyPPV(dMeOPPV), b) is di-hexyloxyPPV (dHeOPPV)<br />

and f<strong>in</strong>ally c) is di-decyloxyPPV (dDeOPPV). . . . . . . . . . . . . . . . 113<br />

6.2 Electric field dependence of mobility for dMeOPPV (full squares, pre-<br />

annealed C1 ) for dHeOPPV (empty diamonds, C6) and for dDeOPPV<br />

(empty circles, C10). The figure also shows the electric field dependence<br />

of mobility for dMeOPPV annealed at low temperature (full triangles,<br />

non-annealed C1) and for MDMO-PPV (dotted l<strong>in</strong>e). . . . . . . . . . . . 115<br />

6.3 Representation of the three Euler angles def<strong>in</strong><strong>in</strong>g relative orientation: Φ<br />

and Θ represent the first two Euler rotations about the temporary z and x<br />

axis, Ψ represents the last rotation about the z axis. . . . . . . . . . . . . 117<br />

6.4 Probability distributions of the difference <strong>in</strong> site energies (left panel) and<br />

the logarithm of the transfer <strong>in</strong>tegral (right panel) of two hexamers of<br />

dMeOPPV with a m<strong>in</strong>imum separation of 3.5 Å . Probability distributions<br />

(vertical bars) are compared to a normal distribution (solid l<strong>in</strong>e). . . . . . 120<br />

6.5 Distance dependence of the standard deviation <strong>in</strong> values of ∆E. . . . . . 121<br />

6.6 Scatter plot of the absolute value of ∆E aga<strong>in</strong>st the logarithm of the trans-<br />

fer <strong>in</strong>tegral log(J) (a). Also shown is the scatter plot of log(J) versus the<br />

centre to centre separation d (b) and the scatter plot of ∆E versus d (c). . . 122<br />

ix


6.7 The three spiro-l<strong>in</strong>ked tryarilam<strong>in</strong>e derivatives considered <strong>in</strong> this study.<br />

From left to right they are 2,2’,7,7’-tetrakis-(N,N-diphenylhenylam<strong>in</strong>o)-<br />

9,9’-spirobifluorene (spiro-unsub), 2,2’,7,7’-tetrakis-(N,N-di-m-methylphenylam<strong>in</strong>o)-<br />

9,9’-spirobifluorene (spiro-Me) and f<strong>in</strong>ally 2,2’,7,7’-tetrakis-(N,N-di-4-<br />

methoxyphenylam<strong>in</strong>o)-9,9’-spirobifluorene (spiro-MeO) . . . . . . . . . . 123<br />

6.8 Histogram of ∆E for spiro-unsub (dashed l<strong>in</strong>e), spiro-Me (dotted l<strong>in</strong>e)<br />

and spiro-MeO (full l<strong>in</strong>e). The standard deviation for the three materials<br />

is 8 meV for spiro-unsub, 10 meV for spiro-Me and 15 meV spiro-MeO. . 125<br />

7.1 Schematic representation of the morphology model used by Dr. S. Athanosopou-<br />

los to describe PFO and def<strong>in</strong>itions of the parameters govern<strong>in</strong>g the rel-<br />

ative position and orientation of molecules. (a) The polymer cha<strong>in</strong>s are<br />

packed hexagonally, with the cha<strong>in</strong> axis perpendicular to the electric field.<br />

(b) Each trimer is allowed to rotate by a torsion φ and can be displaced<br />

by a distace ∆r <strong>in</strong> the direction perpendicular to the cha<strong>in</strong> axis and can<br />

slip by a distance dz parallell to the cha<strong>in</strong> axis. (c) Top view of the central<br />

planes of two trimers, show<strong>in</strong>g the torsion angles φ1 and φ2 (∆φ = φ2−φ1)<br />

and the polar angle α. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131<br />

7.2 HOMO (left) and HOMO-1 (right), of the restricted open shell orbitals of<br />

a fluorene trimer with one of the hydrogen end groups miss<strong>in</strong>g. . . . . . . 132<br />

7.3 Contour plot of log(|J| 2 ) as a function of relative torsion angle dφ and the<br />

polar angle α which describes position. Notice how the values of these<br />

<strong>in</strong>tercha<strong>in</strong> transfer <strong>in</strong>tegrals are always much smaller than the <strong>in</strong>tracha<strong>in</strong><br />

transfer <strong>in</strong>tegrals from figure 7.5. . . . . . . . . . . . . . . . . . . . . . 135<br />

7.4 Distance dependence of the transfer <strong>in</strong>tegral squared on distance for φ =<br />

150 ◦ and α = 0 ◦ (squares) and for φ = 0 ◦ and α = 0 ◦ (circles). . . . . . . 135<br />

7.5 Intracha<strong>in</strong> transfer <strong>in</strong>tegrals squared as a function of relative torsion φ for<br />

two trimers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136<br />

x


7.6 Difference between forward and backward rates for an ordered lattice<br />

as a function the torsion angle φ at which the trimers are set. √ F =<br />

200(V/cm) 1/2 (solid l<strong>in</strong>e), 400(V/cm) 1/2 (dotted l<strong>in</strong>e), 600(V/cm) 1/2 (dashed<br />

l<strong>in</strong>e), 800(V/cm) 1/2 (cha<strong>in</strong>ed l<strong>in</strong>e, one dot), 1000(V/cm) 1/2 (cha<strong>in</strong>ed l<strong>in</strong>e,<br />

two dots). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137<br />

7.7 Dependence of mobility on the square root of the applied electric field<br />

for: the ordered morphology with φ = 0 ◦ (open triangles), the ordered<br />

morphology with φ = 20 ◦ (diamonds), the morphology with disordered<br />

torsions (full squares), the optimally disordered morphology (full trian-<br />

gles) and the experimental data on alligned PFO from reference [3] (full<br />

circles).The optimally ordered morphology corresponds to when neigh-<br />

bours <strong>in</strong> the direction of the field have a relative torsion φ of 150 ◦ . The<br />

lattice constant for the hexagonal lattice is 6.5 Å . . . . . . . . . . . . . . 138<br />

7.8 Chemical structure of the various derivatives of HBC considered <strong>in</strong> this<br />

study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142<br />

7.9 ToF results for several selected snapshots, together with the averaged over<br />

the MD run displacement current (left). Block averages over 10, 20, ...<br />

190 snapshots. (right) . . . . . . . . . . . . . . . . . . . . . . . . . . . 144<br />

7.10 Displacement current for different numbers of molecules <strong>in</strong> a column.<br />

Average over 200 snapshots at an electric field of 10 5 Vcm −1 . . . . . . . . 145<br />

7.11 Radial distribution function along the z direction for all six derivatives<br />

considered <strong>in</strong> this study. . . . . . . . . . . . . . . . . . . . . . . . . . . . 145<br />

7.12 Distribution <strong>in</strong> logarithm of the effective transfer <strong>in</strong>tegrals for holes squared<br />

for 20 columns of 100 molecules each. . . . . . . . . . . . . . . . . . . . 148<br />

xi


List of Tables<br />

5.1 Table of the transfer <strong>in</strong>tegral for holes Jh and for electrons Je. n is the<br />

number of such pairs <strong>in</strong> a cluster of 20 molecules. dcoc is the separation<br />

between the centres of the fullerene cages for the pair of molecules. The<br />

labels a-g label the pairs accord<strong>in</strong>g to figure 5.1 . . . . . . . . . . . . . . 100<br />

5.2 Reorganisation energies for anion. . . . . . . . . . . . . . . . . . . . . . 102<br />

5.3 Reorganisation energies for cation. . . . . . . . . . . . . . . . . . . . . . 102<br />

6.1 GDM parameters for the various dialkoxy PPV derivatives. . . . . . . . . 115<br />

6.2 GDM energetic disorder σGDM and standard deviation <strong>in</strong> distribution of<br />

site energy difference calculated us<strong>in</strong>g distributed Multipole analysis (<br />

σDMA ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125<br />

7.1 Electron (e) and hole (h) mobilities (cm 2 V −1 s −1 ) of different compounds<br />

calculated us<strong>in</strong>g time-of-flight (ToF) and master equation (ME) methods,<br />

<strong>in</strong> comparison with experimentally measured PR-TRMC mobilities. Also<br />

shown are the nematic order parameter Q and the average vertical separa-<br />

tion between cores of HBC molecules h (Å). . . . . . . . . . . . . . . . . 147<br />

xii


Acknowledgments<br />

I would like to acknowledge all the help and support from my colleagues at Imperial<br />

College and my collaborators. None of my work would have been possible without my<br />

supervisor, Jenny Nelson, who provided guidance and advice whilst allow<strong>in</strong>g me absolute<br />

freedom. On no scientific topic was she unable of either provid<strong>in</strong>g expertise or po<strong>in</strong>t<strong>in</strong>g<br />

me <strong>in</strong> the correct direction. Her stakhanovist attitude to work is <strong>in</strong>spirational and is best<br />

exemplified by the probability distribution for receiv<strong>in</strong>g emails from her (shown <strong>in</strong> figure<br />

1).<br />

I received a lot of editorial help <strong>in</strong> the writ<strong>in</strong>g of the thesis: from my father, Rysia,<br />

most of the people I shared an office with <strong>in</strong> room H724 (especially Graeme) and ob-<br />

viously from Jenny! Rysia’s help <strong>in</strong> proof-read<strong>in</strong>g at the 11th hour was <strong>in</strong>dispensable.<br />

Kristian Sylvester-Hvid also provided useful <strong>in</strong>sight <strong>in</strong> the <strong>in</strong>troduction of the thesis.<br />

Thanks to Bhavna and Carolyn for their efficient runn<strong>in</strong>g of the group.<br />

I would like to thank, or rather apologise to, those who I have pestered with program-<br />

m<strong>in</strong>g issues, especially my housemates Pete and Joe.<br />

I would also like to thank those who helped straighten out a few of the bugs <strong>in</strong> my<br />

code: Joe Kwiatkowski, Stavros Athanasopoulos from Bath University and David Reha<br />

from Leeds University.<br />

A word of thanks should go to Ian Gould and Emilio Palomares for help<strong>in</strong>g me nav-<br />

igate my first steps <strong>in</strong> Quantum Chemistry. The staff at the EPSRC centre for computa-<br />

tional chemistry are also thanked for all the assistance <strong>in</strong> learn<strong>in</strong>g to use computational<br />

packages and computer time.<br />

Many of the <strong>in</strong>vestigations <strong>in</strong> this thesis were collaborative and I want to especially<br />

acknowledge the experimental efforts of Sachetan Tuladhar, whose results are discussed<br />

<strong>in</strong> chapter five and six . A special word of thank should go to Sachetan for be<strong>in</strong>g such a<br />

xiii


Probability Density<br />

0.1<br />

0.05<br />

0<br />

0 4 8 12 16 20<br />

Arrival Time (hr)<br />

Figure 1: Probability density distribution for receiv<strong>in</strong>g an email from Jenny. 41% of<br />

emails is received after five o’clock!<br />

pleasant room mate <strong>in</strong> the undercroft. The time of flight code I used has been developed by<br />

many: Jenny Nelson, Amanda Chatten, Rosemary Chandler and Joe Kwiatkowski. The<br />

work on hexabenzorononene is the fruit of a fruitful collaboration with Denis Andrienko,<br />

who was also a most courteous host <strong>in</strong> Ma<strong>in</strong>z. For the work on poly(dioctylfluorene) I<br />

would like to thank the group at Bath University, especially Pete Watk<strong>in</strong>s who did much<br />

of the work on develop<strong>in</strong>g their Monte Carlo code and morphology model.<br />

A word of thanks goes also to the research groups which I have been priviledged to<br />

visit and the staff who took care of me: Jérôme Cornil and Davide Beljonne from the Uni-<br />

versity of Mons-Ha<strong>in</strong>aut, Jean-Luc Brédas and Demetrio da Silvha Filho from Georgia<br />

Institute of technology and f<strong>in</strong>ally Kristian Sylvester-Hvid for host<strong>in</strong>g me <strong>in</strong> Copenhagen.<br />

F<strong>in</strong>ally I would like to thank all the rest of the people <strong>in</strong> the physics deparment who<br />

have provided lunch brakes and coffee break discussions: Graeme, the two Matts, Rob,<br />

Peter, Thil<strong>in</strong>i, Jessica, Joe, Jarvist, Dan, Rahul, Dave Johnson, Markus, Marianne, Ben,<br />

Just<strong>in</strong>, Tom, Mariano, Jony, Ben. Paul Stavr<strong>in</strong>ou also deserved mention for <strong>in</strong>terest<strong>in</strong>g<br />

discussion.<br />

xiv


Mike Bearpark and Martial Boggio-Pasqua have guided me through my PhD on CASSCF<br />

calculations. The results are not <strong>in</strong> this thesis, but I am greatly <strong>in</strong>debted to them for the<br />

<strong>in</strong>terest<strong>in</strong>g tutor<strong>in</strong>g they gave me.<br />

The last word of thanks is to the reader: I hope you do not f<strong>in</strong>d this too bor<strong>in</strong>g!<br />

xv


Abstract<br />

In this thesis we discuss the calculation of charge transport parameters <strong>in</strong> conjugated<br />

solids. We aim to elucidate the role of chemical structure and molecular pack<strong>in</strong>g <strong>in</strong> de-<br />

term<strong>in</strong><strong>in</strong>g charge transport characteristics. In chapter two we discuss the rate equation<br />

from Marcus theory used throughout the thesis and its parameters: electronic coupl<strong>in</strong>g,<br />

reorganisation energy and site energy difference. We also discuss the Gaussian Disor-<br />

der Model as an example of a lattice based simulation of charge transport. We show that<br />

lattice simulations are limited because of the difficulty <strong>in</strong> relat<strong>in</strong>g model parameters to mi-<br />

croscopic properties of a material. In chapter three we present fast methods of calculat<strong>in</strong>g<br />

transfer <strong>in</strong>tegrals and site energies for pairs of molecules. We will use these methods <strong>in</strong><br />

chapters six and seven, <strong>in</strong> which calculations on large numbers of pairs of molecules are<br />

carried out. These fast but approximate methods are shown to be <strong>in</strong> agreement with more<br />

ab-<strong>in</strong>itio methods. In chapter four we discuss localisation of charges along the backbone<br />

of conjugated polymers and argue that thermal fluctuations are of greater importance than<br />

polaron b<strong>in</strong>d<strong>in</strong>g energies <strong>in</strong> affect<strong>in</strong>g localisation. In chapter five we calculate transfer<br />

<strong>in</strong>tegrals and reorganisation energies for electrons and holes <strong>in</strong> a fullerene derivative <strong>in</strong><br />

order to justify the choice of parameters <strong>in</strong> a lattice based simulation. In chapter six we<br />

use the fast methods for calculat<strong>in</strong>g transfer <strong>in</strong>tegrals and site energies to rationalise the<br />

Gaussian Disorder Model parameters of two families of materials: dialkoxy polypara-<br />

phenylenev<strong>in</strong>ylenes and triarylam<strong>in</strong>e derivatives. The differences <strong>in</strong> energetic disoder<br />

between the materials is attributed to electrostatic effects. In chapter seven we calcu-<br />

late charge transport parameters for simulated morphologies of the crystall<strong>in</strong>e phase of<br />

poly(dioctylfluorene) and of the discotic liquid crystal mesophase of different hexabenzo-<br />

coronene derivatives. In the case of poly(dioctylfluorene) calculations of mobility help to<br />

shed light on molecular pack<strong>in</strong>g. In the case of hexabenzocoronene the availability of a<br />

good morphology model leads, with no free parameters, to values of mobility <strong>in</strong> excellent<br />

agreement with experimental data.


List of Symbols and Abbreviations<br />

J: Electronic overlap<br />

MOO: Molecular Orbital Overlap: a method to calculate J based on a simplification of<br />

ZINDO<br />

INDO: Intermediate Neglect of Differential Overlap method<br />

ZINDO: Zerner’s Intermediate Neglect of Differential Overlap method [1]<br />

∆E: Site energy difference<br />

λ: Reorganisation energy<br />

S CF: Self Consistent Field: a method to determ<strong>in</strong>e the electronic energy of a system<br />

where the Fock matrix is dependent on the wavefunction; the procedure is therefore re-<br />

peated until the wavefunction obta<strong>in</strong>ed by diagonalis<strong>in</strong>g the Fock matrix and the Fock<br />

matrix obta<strong>in</strong>ed from the wavefunction are consistent with each other<br />

HOMO: Highest Occupied Molecular Orbital<br />

LUMO: Lowest Unoccupied Molecular Orbital<br />

µ ν: Labels for molecular (or atomic) orbitals; µ is also the mobility.<br />

A B: Labels for molecules<br />

i j: Labels for electrons<br />

a b: Labels for atoms<br />

x y z α β: Labels for x,y and z coord<strong>in</strong>ates or for generic coord<strong>in</strong>ate labels (α β)<br />

Ψ: Electron wavefunction<br />

χ: Nuclear wavefunction<br />

Ψ: Slater determ<strong>in</strong>ant: a comb<strong>in</strong>ation of molecular orbitals which ensures the anti-symmetric<br />

nature of the wavefunction<br />

ψ: Molecular orbital<br />

φ: Atomic orbital<br />

S : Overlap matrix<br />

P: The density matrix P = � µ occupied φ t µφµ<br />

Q: Quadrupole moment<br />

P: Polarisability<br />

a0: Bohr radius<br />

1


e: Magnitude of the charge of an electron<br />

k: Boltzmann’s constant<br />

T: Temperature<br />

DMA: Distributed Multipole Analysis<br />

U: An electrostatic <strong>in</strong>teraction<br />

σ: Energetic disorder or a bond formed by sp hybridised orbitals or σ overlap (i.e. the<br />

overlap between p atomic orbitals symmetric about the vector jo<strong>in</strong><strong>in</strong><strong>in</strong>g the two atomic<br />

centres)<br />

π: Bonds formed by p orbitals or π overlap (i.e. the overlap between p atomic orbitals<br />

antisymmetric about the vector jo<strong>in</strong><strong>in</strong><strong>in</strong>g the two atomic centres)<br />

C: Matrix represent<strong>in</strong>g all the molecular orbitals <strong>in</strong> a basis set; each row corresponds to a<br />

molecular orbital<br />

Q: Nuclear coord<strong>in</strong>ates<br />

q: Electronic coord<strong>in</strong>ates<br />

ω: Angular frequency<br />

F: Electric field<br />

2


Bibliography<br />

[1] J. Ridley and M. Zerner, An Intermediate Neglect of Differential Overlap Technique<br />

for Spectroscopy: Pyrrole and the Az<strong>in</strong>es, Theoretica Chimica Acta, 32, 111 (1973)<br />

3


Chapter 1<br />

Introduction<br />

1.1 Why is a Theoretical Study of <strong>Charge</strong> <strong>Transport</strong> Nec-<br />

essary to the Design of Better Devices?<br />

Ever s<strong>in</strong>ce the discovery of conductivity <strong>in</strong> trans-polyacetylene [1], study of the electronic<br />

properties of organic molecules has been very <strong>in</strong>tense. The field was further stimulated<br />

by the discovery of light emitt<strong>in</strong>g polymers [2] and of photo<strong>in</strong>duced charge separation<br />

<strong>in</strong> polymer fullerene blends [3], which stimulated the development of, respectively, light<br />

emitt<strong>in</strong>g diodes and solar cells. The reason for these high levels of <strong>in</strong>terest is that if it were<br />

possible to design materials with the opto-electronic properties of <strong>in</strong>organic semiconduc-<br />

tors and the mechanical and process<strong>in</strong>g properties of plastics, the applications would be<br />

revolutionary. Solution processability would allow fabrication of devices <strong>in</strong> cont<strong>in</strong>uous<br />

<strong>in</strong>dustrial processes [4, 5], thus reduc<strong>in</strong>g costs. Plastics are also flexible and light-weight<br />

compared to rigid, brittle <strong>in</strong>organic semiconductors: this could open the door to <strong>in</strong>tegra-<br />

tion of electronic devices on different subtrates. It seems to us that the area which will<br />

show the greatest long term benefits is organic photovoltaics (PV): <strong>in</strong>organic PV is lim-<br />

ited <strong>in</strong> applicability by its high cost, both due to material costs (solar cells are large area<br />

devices) and process<strong>in</strong>g costs (fabricat<strong>in</strong>g a PV module is a complex process). For both<br />

these reasons, the potential low cost and ease of process<strong>in</strong>g of organic materials would<br />

have a great impact.<br />

Many different designs of solar cells based on organic materials have been proposed,<br />

4


for example: dye sensitised solar cells [6], bulk heterojunction cells [7, 8], composite<br />

cadmium selenide/polymer solar cells [9], cells based on self organised liquid crystals<br />

[11]. In all these systems three processes must occur <strong>in</strong> order for a photocurrent to be<br />

collected at the electrodes: photons must be absorbed, charges must be separated and fi-<br />

nally charges must be able to migrate to the electrodes. <strong>Charge</strong> mobility is important both<br />

because it sets the maximum space charge limited current obta<strong>in</strong>able by the device [12]<br />

and because, if transport is slower than charge absorption or charge separation, it causes<br />

bottlenecks and <strong>in</strong>creased recomb<strong>in</strong>ation [13]. In this thesis, we study the parameters con-<br />

troll<strong>in</strong>g charge dynamics <strong>in</strong> organic materials. In particular we will argue for the need of<br />

electronic structure calculations to elucidate the relationship between chemical structure<br />

and charge transport characteristics. We want to suggest that our understand<strong>in</strong>g of the<br />

fundamental mechanism beh<strong>in</strong>d charge transport and the speed of computers is now suf-<br />

ficient to allow an understand<strong>in</strong>g of charge mobility based on the fundamental electronic<br />

properties of molecules and that this is the only path which will allow disentanglement<br />

of the various chemical and physical causes for device performance, therefore allow<strong>in</strong>g<br />

better material and device design.<br />

The thesis is organised as follows: <strong>in</strong> chapter two we describe the non-adiabatic, Mar-<br />

cus high temperature charge transfer equation and methods from the literature to calculate<br />

its parameters; we also discuss some lattice based models of charge mobility. In chapter<br />

three we discuss fast methods of calculat<strong>in</strong>g the parameters of the Marcus charge transfer<br />

equation. In chapter four we discuss the nature of charged radicals on polymer cha<strong>in</strong>s.<br />

The rest of the thesis is dedicated to apply<strong>in</strong>g the methods developed to expla<strong>in</strong><strong>in</strong>g ex-<br />

perimental charge transport characteristics of conjugated materials. In chapter five we<br />

exam<strong>in</strong>e the dependence on fullerene concentration of electron and hole mobilities <strong>in</strong><br />

blends of fullerene and conjugated polymer. In chapter six we calculate charge transport<br />

parameters <strong>in</strong> pairs of molecules <strong>in</strong> random orientations and compare these results to pa-<br />

rameters from lattice models of mobility. In chapter seven we run full-blown simulations<br />

of molecular morphology and electronic properties for polymers and discotic liquid crys-<br />

tals <strong>in</strong> an effort to reproduce charge transport characteristics with a m<strong>in</strong>imal number of<br />

adjustable parameters. F<strong>in</strong>ally we will give a brief of conclusion of the thesis. The rest of<br />

5


this chapter <strong>in</strong>troduces some concepts about charge transfer <strong>in</strong> conjugated materials.<br />

1.2 Conjugated Materials<br />

The properties of conjugated materials are based on the nature of covalent bonds between<br />

carbon, nitrogen, sulfur and oxygen. Specifically, they are based on the properties of sp 2<br />

bonds <strong>in</strong> carbon atoms. In order to better understand the different types of bond<strong>in</strong>g that<br />

carbon is capable of let us consider three compounds: ethane (C2H6), ethene (C2H4) and<br />

ethyne (C2H2). In these three compounds, each carbon forms bonds with either four, three<br />

or two other atoms. These three different behaviours depend on how the 2s and 2p atomic<br />

orbitals on carbon hybridise to form bonds: <strong>in</strong> ethane four σ bonds are formed by mix<strong>in</strong>g<br />

a s<strong>in</strong>gle 2s orbital with three 2p orbitals. These σ orbitals are therefore known as sp 3<br />

orbitals. In ethene only three σ bonds are formed by mix<strong>in</strong>g a 2s orbital with two 2p<br />

orbitals. This is sp 2 hybridazation. In ethyne, a s<strong>in</strong>gle σ bond is formed by mix<strong>in</strong>g a 2s<br />

and a 2p orbital. This is sp 1 hybridazation. These three types of orbitals are responsible<br />

for the formation of the bonds which determ<strong>in</strong>e the fundamental geometrical features of<br />

the three molecules. Us<strong>in</strong>g the pr<strong>in</strong>ciple that orbitals will tend to repel each other we can<br />

predict the bond angles <strong>in</strong> the three materials: <strong>in</strong> ethane the bonds around each carbon<br />

are <strong>in</strong> a tetrahedral arrangement, <strong>in</strong> ethene they are triangular and <strong>in</strong> ethyne the bonds are<br />

l<strong>in</strong>ear. In the case of sp 2 and sp 1 hybridasation, one and two 2p orbitals respectively are<br />

left un-hybridised: these orbitals participate <strong>in</strong> bond<strong>in</strong>g by form<strong>in</strong>g a double and a triple<br />

bond respectively between the carbons. A conjugated material is one <strong>in</strong> which all bonds<br />

are either sp 2 or sp 1 hybridised, result<strong>in</strong>g <strong>in</strong> an alternation of s<strong>in</strong>gle and double bonds. As<br />

shown <strong>in</strong> figure 1.1, <strong>in</strong> compounds such as benzene different ways of alternat<strong>in</strong>g s<strong>in</strong>gle<br />

and double bonds leads to different structures with the same energy: these structures<br />

can mix together form<strong>in</strong>g lower energy structures where <strong>in</strong>stead of alternat<strong>in</strong>g s<strong>in</strong>gle and<br />

double bonds, there are only aromatic bonds. These resonances are responsible for the<br />

high stability of aromatic compounds.<br />

σ and π bonds are so called because they are either symmetric or antisymmetric with<br />

respect to the axis connect<strong>in</strong>g the two atoms participat<strong>in</strong>g <strong>in</strong> bond<strong>in</strong>g. This expla<strong>in</strong>s<br />

why σ bonds are strong and participate more <strong>in</strong> determ<strong>in</strong><strong>in</strong>g the geometrical structure<br />

6


Figure 1.1: Resonant Lewis structures for benzene with s<strong>in</strong>gle and double bonds (left) and<br />

representation of the aromatic structure (right).<br />

of molecules: σ orbitals are symmetric across the b<strong>in</strong>d<strong>in</strong>g direction and therefore have<br />

large overlap <strong>in</strong>tegrals, π orbitals, conversely, rely on overlap of the lobes of parallel 2p<br />

orbitals and are much weaker. This observation, that 2p orbitals must be parallel to form<br />

bonds, leads us to understand why ethene is planar: if the two triangles formed by each<br />

carbon and the two hydrogens bonded to them are not parallel, the 2p bonds on each atom<br />

are not capable of form<strong>in</strong>g a bond. So sp 2 hybridisation implies the formation of planar<br />

structures, with double bonds formed by the rema<strong>in</strong>g 2p orbitals.<br />

The <strong>in</strong>terest<strong>in</strong>g electronic properties of conjugated materials is delocalisation. Often<br />

the concept of delocalisation is expla<strong>in</strong>ed by argu<strong>in</strong>g that when the unhybridised p atomic<br />

orbitals on each atom <strong>in</strong>teract to form molecular π orbitals, these π orbitals are delocalised<br />

over the whole molecule. This is certa<strong>in</strong>ly true, however it is not a property unique to π<br />

orbitals, also the hybridised sp 2 atomic orbitals <strong>in</strong>teract to form delocalised σ orbitals. In<br />

my op<strong>in</strong>ion what is special about p conjugated systems is that charged or neutral excita-<br />

tions are delocalised, <strong>in</strong> the sense that they can migrate along the molecule with relatively<br />

small amounts of energy. As an example of this take polyparaphenylenev<strong>in</strong>ylene (PPV).<br />

This p conjugated polymer is shown <strong>in</strong> figure 1.2. If it becomes charged, its Lewis struc-<br />

ture is similar to figure 1.3; <strong>in</strong> this figure the dot represents an unpaired electron and the<br />

7


Figure 1.2: Lewis structure of a segment of PPV.<br />

Figure 1.3: Lewis structure of a charged segment of PPV before and after charge migration.<br />

”+” represents a charge. What we mean to represent <strong>in</strong> this figure is that, upon charg<strong>in</strong>g,<br />

the bond order<strong>in</strong>g changes mak<strong>in</strong>g the structure less aromatic and more qu<strong>in</strong>oid. This<br />

localised qu<strong>in</strong>oid deformation can easily travel along the cha<strong>in</strong> as it only <strong>in</strong>volves the<br />

break<strong>in</strong>g and reform<strong>in</strong>g of loosely bound bonds. This therefore seems to us to be the<br />

fundamental property of conjugated systems: the nature of π bonds is such that they can<br />

be easily broken and reformed without the significant rearrangement of bonds that would<br />

occur if σ bonds were broken.<br />

1.3 <strong>Charge</strong> <strong>Transport</strong> <strong>in</strong> Conjugated Materials<br />

In chapter two we will provide a formal description of non-adiabatic charge transfer, but <strong>in</strong><br />

this section we briefly describe three possible regimes for charge transfer <strong>in</strong> a conjugated<br />

8


material: delocalisation, adiabatic transfer and non-adiabatic transfer. S<strong>in</strong>ce most of our<br />

efforts are computational, rather than provid<strong>in</strong>g a review of analytical charge transfer<br />

models such as can be found <strong>in</strong> references [14, 15], we will implement a simple quantum<br />

dynamical simulation from a simple semi-classical Hamiltonian. In this section we mean<br />

to simply exemplify different possible regimes of charge transfer.<br />

We write a Hamiltonian for charge transfer with two non-<strong>in</strong>teract<strong>in</strong>g (diabatic) states<br />

whose energies depend quadratically on a s<strong>in</strong>gle reaction coord<strong>in</strong>ate Q. Electron-phonon<br />

coupl<strong>in</strong>g is <strong>in</strong>troduced by displac<strong>in</strong>g the potential energy parabula by a distance Q0. At<br />

the nuclear coord<strong>in</strong>ate where the energy of state |0 > is m<strong>in</strong>imised the potential energy for<br />

state |1 > is ω2 Q0 2<br />

2 . This quantity is called the reorganisation energy and is an expression of<br />

electron-phonon coupl<strong>in</strong>g. In the case of the polymers depicted <strong>in</strong> figure 1.3, the reaction<br />

coord<strong>in</strong>ate Q could represent go<strong>in</strong>g from an aromatic to a qu<strong>in</strong>oid structure. Electronic<br />

coupl<strong>in</strong>g between the two diabatic states is described by a transfer <strong>in</strong>tegral J. In state<br />

representation, us<strong>in</strong>g mass weighted coord<strong>in</strong>ates, the Hamiltonian for the system is:<br />

ˆHe = ω2 Q 2<br />

2 |0 >< 0| + ω2 (Q − Q0) 2<br />

2<br />

|1 >< 1| + (J|0 >< 1| + J ∗ |1 >< 0|) (1.1)<br />

Figure 1.4 shows the eigenvalues of this Hamiltonian represented as a function of the<br />

reaction coord<strong>in</strong>ate Q for typical values of ω and Q0.<br />

If the electronic coupl<strong>in</strong>g is of the same order of magnitude as ω2 Q0 2<br />

2 , the system is <strong>in</strong><br />

the delocalised regime. In order to dist<strong>in</strong>guish the other two regimes, we have to consider<br />

the relative time scales for nuclear and electronic transitions. Nuclear oscillations occur<br />

on a timescale of ω −1 ; electronic transitions between the diabatic states, conversely, occur<br />

on a time scale of �/J. If the timescale for electronic transitions is much faster than<br />

that for nuclear motion, i.e. if J is large compared to �ω, charge transfer will occur<br />

gradually as the system oscillates along the potential for the adiabatic states (that is for<br />

the eigenfunction of the Hamiltonian shown <strong>in</strong> equation 1.1 ). If J is small compared to<br />

�ω, the electronic states do not have the time to rearrange themselves as the nuclei vibrate<br />

and the system oscillates along one of the diabatic potential energy surfaces. Each time<br />

the potential energies for each diabatic state are resonant, i.e. when Q = Q0/2 is reached,<br />

the transfer <strong>in</strong>tegral J will set a f<strong>in</strong>ite probability for cross<strong>in</strong>g from one diabatic state to<br />

9


Energy / eV<br />

1<br />

0.5<br />

0<br />

-1 0 1 2<br />

Nuclear coord<strong>in</strong>ates Q / Q<br />

0<br />

Figure 1.4: Eigenvalues of equation 1.1 as a function of the reaction coord<strong>in</strong>ate Q.<br />

another.<br />

In order to perform a simulation of the dynamics of the evolution of the system, we<br />

will use the Ehrenfest approximation. In other words, we assume that the nuclear motion<br />

can be treated classically and that the force on the nuclei is simply the expectation value<br />

of the force operator ˆF = − ˆ He . The set of coupled differential equations we solve is:<br />

Q<br />

˙Pe = i<br />

� [ ˆHe, Pe] (1.2a)<br />

�<br />

F = Tr − d ˆHe<br />

dQ Pe<br />

�<br />

(1.2b)<br />

¨Q = F (1.2c)<br />

where Pe is the density matrix for the electronic system, the square brackets denote tak-<br />

<strong>in</strong>g the commutator and Tr{.} denotes tak<strong>in</strong>g the trace. We numerically <strong>in</strong>tegrate these<br />

equations of motion us<strong>in</strong>g the GNU octave program. In all simulations we scale the mass<br />

weighted length scale so that ω2 Q0 2<br />

2<br />

= 0.3eV. In all cases we start the simulation with a<br />

large potential energy correspond<strong>in</strong>g to a displacement of −Q0 to ensure that the system<br />

10


has sufficient energy to reach the cross<strong>in</strong>g po<strong>in</strong>t of the two diabatic surfaces. We also pre-<br />

pare the electronic system so that at time zero it is <strong>in</strong> the diabatic state |0 >, i.e. take the<br />

<strong>in</strong>itial condition Pe(t = 0) = |0 >< 0|. We then make two comparisons. First we keep the<br />

angular frequency fixed at ω = 5 10 12 s −1 and <strong>in</strong>crease the transfer <strong>in</strong>tegral from 1 meV to<br />

500 meV to illustrate the transition of charge transfer characteristics from non-adiabatic to<br />

adiabatic and f<strong>in</strong>ally to delocalised charge transfer. In the second comparison we keep the<br />

transfer <strong>in</strong>tegral fixed at 10 meV and vary ω from 10 12 s −1 to 10 14 s −1 to demonstrate that<br />

adiabaticity does not result so much from the relative values of J and the reorganisation<br />

energy, but more from the relative values of ω and J/�. In order to exemplify the time<br />

evolution of the system we look at two quantities: the nuclear coord<strong>in</strong>ate Q and Ple f t, the<br />

expectation value of f<strong>in</strong>d<strong>in</strong>g the system <strong>in</strong> diabatic state |0 >. Ple f t is calculated by tak<strong>in</strong>g<br />

the follow<strong>in</strong>g trace over the density matrix: Tr {|0 >< 0|Pe}. We would like to strongly<br />

stress that this simulation has only the purpose of illustrat<strong>in</strong>g graphically different transfer<br />

regimes and that it ignores issues crucial to determ<strong>in</strong><strong>in</strong>g charge transfer rates, such as the<br />

role of nuclear wavefunction overlap, the appropriateness of apply<strong>in</strong>g Ehrnfest dynamics<br />

to non-adiabatic problems and the role of decoherence.<br />

Figure 1.5 shows the time evolution for systems with the same angular frequency and<br />

different transfer <strong>in</strong>tegrals. The top panel shows an example of non-adiabatic transfer,<br />

with the timescale for transition between diabatic states beg<strong>in</strong> �/J = 6.6 10 −13 s and with<br />

the time scale for nuclear motion be<strong>in</strong>g 1/ω = 2.0 10 −13 s. The nuclear system is oscillat-<br />

<strong>in</strong>g along the diabatic surface and <strong>in</strong> fact the midpo<strong>in</strong>t for oscillation is 0, the position of<br />

the bottom of the left parabula. Each time the nuclear coord<strong>in</strong>ate passes through Q0<br />

2 , Ple f t<br />

jumps abruptly and there is a f<strong>in</strong>ite probability of f<strong>in</strong>d<strong>in</strong>g the system <strong>in</strong> the diabatic state<br />

|1 >.<br />

The middle panel of figure 1.5 shows time evolution for the same ω but for a dif-<br />

ferent value of J of 50 meV. In this case the timescale for diabatic transitions is �/J =<br />

1.3 10 −14 s: electronic transitions occur faster than the nuclear motion. In fact each time<br />

that the system goes through Q0<br />

2 , Ple f t switches from left to right, show<strong>in</strong>g that electronic<br />

transitions take place fully and not partially as <strong>in</strong> the non-adiabatic regime. Look<strong>in</strong>g at<br />

the nuclear displacement, we notice that the midpo<strong>in</strong>t of the oscillation occurs at Q0 , not 2<br />

11


Displacement / Q 0<br />

Displacement / Q 0<br />

Displacement / Q 0<br />

1.5<br />

1<br />

0.5<br />

0<br />

-0.5<br />

-1<br />

1.6<br />

1.4<br />

1.2<br />

1<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

0<br />

-0.2<br />

-0.4<br />

-0.6<br />

2<br />

1.5<br />

1<br />

0.5<br />

0<br />

-0.5<br />

-1<br />

0 1e-12 2e-12 3e-12 4e-12 5e-12 6e-12 7e-12 8e-12 9e-12<br />

0.88<br />

1e-11<br />

Time / s<br />

Displacement<br />

P left<br />

0 1e-12 2e-12 3e-12 4e-12 5e-12 6e-12 7e-12 8e-12 9e-12<br />

0<br />

1e-11<br />

Time / s<br />

Displacement<br />

P left<br />

0 1e-12 2e-12 3e-12 4e-12 5e-12 6e-12 7e-12 8e-12 9e-12 1e-11<br />

0<br />

Time / s<br />

Displacement<br />

P left<br />

Figure 1.5: Time evolution for three systems with the same reorganisation energy, the<br />

same angular frequency ω = 5 10 12 s −1 and three different transfer <strong>in</strong>tegrals J:<br />

1 meV (top), 50 meV (middle) and 300 meV(bottom).<br />

12<br />

1<br />

0.98<br />

0.96<br />

0.94<br />

0.92<br />

0.9<br />

1<br />

0.9<br />

0.8<br />

0.7<br />

0.6<br />

0.5<br />

0.4<br />

0.3<br />

0.2<br />

0.1<br />

1<br />

0.9<br />

0.8<br />

0.7<br />

0.6<br />

0.5<br />

0.4<br />

0.3<br />

0.2<br />

0.1<br />

P left<br />

P left<br />

P left


at 0, show<strong>in</strong>g that the system is oscillat<strong>in</strong>g along the adiabatic potential energy surface.<br />

Note however how the oscillation is far from harmonic and has a sharp k<strong>in</strong>k at Q0 . S<strong>in</strong>ce<br />

2<br />

<strong>in</strong> this regime the system never abandons the adiabatic surface we call it adiabatic.<br />

The bottom graph panel of figure 1.5 shows the time evolution for J = 300 meV. In<br />

this case the coupl<strong>in</strong>g between the diabatic states is stronger than the electron phonon<br />

coupl<strong>in</strong>g and the system is free to oscillate harmonically along the diabatic surface. Re-<br />

gardless of what value the nuclear coord<strong>in</strong>ate takes, Ple f t oscillates rapidly, <strong>in</strong> other words<br />

the system is <strong>in</strong> the delocalised regime.<br />

In order to underl<strong>in</strong>e the orig<strong>in</strong> of adiabatic or non-adiabatic transfer, we will look at<br />

three different time evolutions, all obta<strong>in</strong>ed us<strong>in</strong>g the same transfer <strong>in</strong>tegral J = 10 meV,<br />

but with three different angular frequencies: ω = 10 12 s −1 , 5 10 12 s −1 and 10 14 s −1 . These<br />

three time evolutions are show <strong>in</strong> figure 1.6. In the middle panel we show an <strong>in</strong>termediate<br />

regime between the adiabatic and non-adiabatic transfer. The top panel shows that by de-<br />

creas<strong>in</strong>g the frequency of nuclear oscillations, transfer becomes adiabatic. In the bottom<br />

panel we show that by mak<strong>in</strong>g nuclear oscillations very fast (notice the change <strong>in</strong> time<br />

scale), the electron transfer becomes non-adiabatic.<br />

In this thesis we will describe charge transport as a non-adiabatic processes. This<br />

can be justified by argu<strong>in</strong>g that neutral to charged reactions can have contributions from<br />

high frequency normal modes and that therefore the frequency of electronic relaxations<br />

is <strong>in</strong> general slower than the nuclear vibrations. In chapter two we give a more detailed<br />

explanation of the semiclassical formula that we use to describe transport and its essential<br />

characteristics.<br />

13


Displacement / Q 0<br />

Displacement / Q 0<br />

Displacement / Q 0<br />

2<br />

1.5<br />

1<br />

0.5<br />

0<br />

-0.5<br />

-1<br />

2<br />

1.5<br />

1<br />

0.5<br />

0<br />

-0.5<br />

-1<br />

1.5<br />

1<br />

0.5<br />

0<br />

-0.5<br />

-1<br />

0 1e-12 2e-12 3e-12 4e-12 5e-12 6e-12 7e-12 8e-12 9e-12<br />

0<br />

1e-11<br />

Time / s<br />

Displacement<br />

P left<br />

0 1e-12 2e-12 3e-12 4e-12 5e-12 6e-12 7e-12 8e-12 9e-12<br />

0.1<br />

1e-11<br />

Time / s<br />

Displacement<br />

P left<br />

0 1e-13 2e-13 3e-13 4e-13 5e-13 6e-13 7e-13 8e-13 9e-13 1e-12<br />

0.4<br />

Time / s<br />

Displacement<br />

P left<br />

Figure 1.6: Time evolution for three systems with the same reorganisation energy, the<br />

same transfer <strong>in</strong>tegral J = 10 meV and three different angular frequencies ω:<br />

10 12 s −1 (top), 5 10 12 s −1 (middle) and 10 14 s −1 (bottom).<br />

14<br />

1<br />

0.9<br />

0.8<br />

0.7<br />

0.6<br />

0.5<br />

0.4<br />

0.3<br />

0.2<br />

0.1<br />

1<br />

0.9<br />

0.8<br />

0.7<br />

0.6<br />

0.5<br />

0.4<br />

0.3<br />

0.2<br />

1<br />

0.9<br />

0.8<br />

0.7<br />

0.6<br />

0.5<br />

P left<br />

P left<br />

P left


Bibliography<br />

[1] C.K. Chiang, C.R. F<strong>in</strong>cher, Y.W. Park, A.J. Heeger, H. Shirakawa, E.J. Louis, S.C.<br />

Gau and A.G. MacDiarmid, Electrical Conductivity <strong>in</strong> Doped Polyacetylene, Physical<br />

Review Letters, 39, 1098 (1977)<br />

[2] J.H. Burroughes, D.D.C. Bradley, A.R. Brown, R.N. Marks, K. Mackay, R.H. Friend,<br />

P.L. Burns and A.B. Holmes, Light Emitt<strong>in</strong>g Diodes based on conjugated polymers.,<br />

Nature, 347, 539 (1990)<br />

[3] N.S. Sariciftci , L. Smilowitz, A.J. Heeger and F. Wudl, Photo<strong>in</strong>duced charge transfer<br />

from a conductive polymer to Buckm<strong>in</strong>sterfullerene, Science, 258, 1474 (1992)<br />

[4] J. Zamuseil, Z. Bao, Y.L. Loo, R. Cirelly and J.A. Rogers, Nanoscale organic tran-<br />

sistors that use source dra<strong>in</strong> electrodes supported by high resolution rubber stamps,<br />

Applied Physics Letters, 82, 793 (2003)<br />

[5] J.Z. Wang, Z.H. Zheng, H.W. Li, W.T.S. Huck and H. Sirr<strong>in</strong>ghaus , Dewett<strong>in</strong>g of<br />

conduct<strong>in</strong>g polymer <strong>in</strong>kjet droplets on patterned surfaces, Nature Materials , 3, 171<br />

(2004)<br />

[6] A. Hagfeldt and M. Grätzel, Molecular Photovoltaics, Account of chemical research,<br />

33, 269 (2000)<br />

[7] G. Yu and A.J. Heeger, <strong>Charge</strong> separation and photovoltaic conversion <strong>in</strong> polymer<br />

composites with <strong>in</strong>ternal donor/acceptor heterojunctions, Journal of Applied Physics,<br />

78, 4510 (1995)<br />

15


[8] S.E. Shaheen, C.J. Brabec, N.S. Sariciftci, F. Pad<strong>in</strong>ger, T. Fromherz and J.C. Hum-<br />

melen, 2.5 % efficient organic plastic solar cells, Applied Physics Letters, 78, 841<br />

(2001)<br />

[9] D.S. G<strong>in</strong>ger and N.C. Greenham, <strong>Charge</strong> Separation <strong>in</strong> Conjugated-Polymer<br />

Nanocrystal Blends, Synthetic Metals, 101, 425 (1999)<br />

[10] C.J. Brabec, F. Pad<strong>in</strong>ger, J.C. Hummelen, R.A.J. Janssen and N.S. Sariciftci, Reali-<br />

sation of Large Area Flexible Fullerene - Conjugated Polymer Photocells: A Route to<br />

Plastic Solar Cells, Synthetic Metals, 102,861 (1999)<br />

[11] L. Schmidt-Mende, A. Fechtenkotter, K. Mullen, R.H. Friend and J.D. MacKen-<br />

zie, Efficient organic photovoltaics from soluble discotic liquid crystall<strong>in</strong>e materials,<br />

Physica E, 14, 263 (2002)<br />

[12] A. Moliton and J.M. Nunzi, How to model the behaviour of organic photovoltaic<br />

cells, Polymer International, 55, 583 (2006)<br />

[13] J. Nelson, J. Kirkpatrick and P. Ravirajan, Factors limit<strong>in</strong>g the efficiency of molecu-<br />

lar photovoltaic devices, Physical Review B, 69, 035337 (2004)<br />

[14] M.D. Newton, Quantum Mechanical Probes of Electron-Transfer K<strong>in</strong>etics: the Na-<br />

ture of Donor-Acceptor Interactions, Chemical Reviews, 91, 767 (1991)<br />

[15] P.F. Barbara, T.J. Meyer and M.A. Ratner, Contemporary Issues <strong>in</strong> Electron Transfer<br />

Research, Journal of Chemical Physics, 100, 13148 (1996)<br />

16


Chapter 2<br />

Background<br />

<strong>Charge</strong> transport <strong>in</strong> disordered molecular solids proceeds by discrete charge transfer events<br />

between localised states. Because of orientational, positional and configurational disor-<br />

der the parameters govern<strong>in</strong>g charge transfer are also disordered and charge dynamics <strong>in</strong><br />

three dimensional networks are best solved with stochastic approaches, such as Monte<br />

Carlo methods, or by numerical solution of the underly<strong>in</strong>g Master equation.<br />

In this chapter we describe Marcus theory <strong>in</strong> the non-adiabatic high temperature limit.<br />

This leads to an expression for the charge transfer rate <strong>in</strong> terms of three parameters: the<br />

reorganisation energy λ, the transfer <strong>in</strong>tegral J and the site energy difference ∆E. We<br />

will then discuss commonly used methods to calculate J and λ and <strong>in</strong>troduce the theo-<br />

retical framework for calculat<strong>in</strong>g electrostatic and <strong>in</strong>ductive <strong>in</strong>teractions between charge<br />

distributions: these theoretical tools will be used <strong>in</strong> chapter three to calculate ∆E. The<br />

last part of the chapter is devoted to describ<strong>in</strong>g present empirical numerical models of<br />

charge transport <strong>in</strong> disordered materials, chiefly the Gaussian Disorder Model (GDM).<br />

The aim of this description is both to <strong>in</strong>troduce the numerical tools that will be used later<br />

<strong>in</strong> the thesis to describe charge dynamics, and to discuss the limitations of the GDM <strong>in</strong><br />

identify<strong>in</strong>g the physical and chemical basis for charge transport properties.<br />

17


2.1 Marcus Theory <strong>in</strong> the Non-Adiabatic High Tempera-<br />

ture Limit<br />

In this section we will provide a description for charge transfer based on time dependent<br />

perturbation theory and on the Marcus theory of oxidative-reductive reactions [1]. This<br />

description will lead to the expression [2] for the charge transfer rate used <strong>in</strong> the rema<strong>in</strong>-<br />

der of the text which we shall refer to as the Marcus charge transfer rate. We follow the<br />

derivation of this equation proposed by Jortner and co-workers [3, 4]. We will consider<br />

the high temperature (classical treatment of nuclear vibrations), non-adiabatic (small elec-<br />

tronic coupl<strong>in</strong>g) limit of the rate equation. The same equation can be obta<strong>in</strong>ed by treat<strong>in</strong>g<br />

the motion of the nuclei <strong>in</strong> a fully classical way, and <strong>in</strong>vok<strong>in</strong>g the Landau-Zener formula<br />

for the probability of non-adiabatic transitions [5]; however such a proof is less clear and<br />

satisfy<strong>in</strong>g than the follow<strong>in</strong>g which rests firmly on Fermi’s golden rule.<br />

The chemical reaction which we wish to describe is a charge transfer event from<br />

molecule DA to molecule DB:<br />

D + A + D 0<br />

B → D0A<br />

+ D+ B<br />

(2.1)<br />

where the superscripts represent the ionic state of the molecule, + for positive and 0<br />

for neutral. Note that although we will refer to the charged molecule with a + symbol<br />

thoughout this section, all arguments can be applied to transport of negative charge by<br />

chang<strong>in</strong>g the positive sign to a negative one.<br />

Firstly let us list the ma<strong>in</strong> assumptions used <strong>in</strong> this derivation:<br />

(a) The <strong>in</strong>itial and f<strong>in</strong>al localised electronic states are derived from a zeroeth order<br />

Hamiltonian which does not account for the off-diagonal elements of the Hamilto-<br />

nian describ<strong>in</strong>g the <strong>in</strong>teraction between the two molecules.<br />

(b) Initially, the molecule is assumed to be <strong>in</strong> an <strong>in</strong>coherent super-position of vibrionic<br />

states, due to its temperature.<br />

(c) The off-diagonal elements of the Hamiltonian describ<strong>in</strong>g the <strong>in</strong>teraction between<br />

molecules, J, is <strong>in</strong>troduced as a time dependent perturbation caus<strong>in</strong>g the <strong>in</strong>itial and<br />

18


f<strong>in</strong>al states to mix.<br />

(d) The matrix element J is assumed to be <strong>in</strong>dependent of the nuclear coord<strong>in</strong>ates Q.<br />

(e) The potential energy surfaces for the <strong>in</strong>itial and f<strong>in</strong>al states are assumed to be<br />

parabolae with identical curvature.<br />

Let us label the <strong>in</strong>itial and f<strong>in</strong>al zeroeth order electronic states Ψ+0 and Ψ0+ respec-<br />

tively. The vibrionic states will be labelled {χ+0,i} and {χ0+, f } respectively, where {} repre-<br />

sents the fact that these states form a set and the subscripts identify the quantum numbers<br />

for electronic and vibrionic states. Because of assumption (c) , the rate of transition from<br />

a particular <strong>in</strong>itial state Ψ+0χ+0,i to the set of f<strong>in</strong>al states {Ψ0+χ0+, f } can be described by a<br />

sum of rates of the form def<strong>in</strong>ed by Fermi’s Golden rule:<br />

�<br />

W+0,i;0+ =<br />

f<br />

2π<br />

� |J+0,i;0+, f | 2 δ(E+0,i − E0+, f ) (2.2)<br />

where W+0,i;0+ represents the rate of transition from an <strong>in</strong>itial state Ψ+0χ+0,i to a set of<br />

f<strong>in</strong>al states {Ψ0+χ0+, f } , J+0,i;0+, f represents the matrix element for the zeroeth order <strong>in</strong>itial<br />

and f<strong>in</strong>al states and E+0,i and E0+, f represent the total energy of the <strong>in</strong>itial and f<strong>in</strong>al states.<br />

Fermi’s golden rule is derived <strong>in</strong> Landau and Lifshitz [6].<br />

The system is assumed to be <strong>in</strong> a thermal average of vibrionic states <strong>in</strong>itially. There-<br />

fore to obta<strong>in</strong> the total rate of conversion of reactants <strong>in</strong>to products W+0,0+ one must sum<br />

over the <strong>in</strong>itial vibrionic states, weigh<strong>in</strong>g each term by the Boltzmann probability of be<strong>in</strong>g<br />

<strong>in</strong> that state:<br />

� �<br />

W+0,0+ =<br />

i<br />

f<br />

2π<br />

� |J+0,i;0+, f | 2 δ(E+0,i − E0+, f )exp(−E+0,i/kT)Z −1<br />

where Z = � i exp(−E+0,i/kT) is the partition function.<br />

(2.3)<br />

Let us take a moment to <strong>in</strong>terpret eq. 2.3. Figure 2.1 shows a schematic represen-<br />

tation of E+0,i and E0+, f as functions of Q. These two energy surfaces are referred to <strong>in</strong><br />

the literature [7] as diabatic surfaces, mean<strong>in</strong>g that the <strong>in</strong>teractions between the zeroeth<br />

order <strong>in</strong>itial and f<strong>in</strong>al states have been neglected. The δ function <strong>in</strong> equation 2.3 is stat<strong>in</strong>g<br />

that electronic tunnell<strong>in</strong>g has to occur at resonance to conserve energy. In other words<br />

19


λ<br />

0<br />

∆E<br />

∆Q<br />

Reactants<br />

Products<br />

Figure 2.1: Schematic view of a potential energy surface for asymmetric charge transport.<br />

λ denotes the reorganisation energy, ∆E the difference <strong>in</strong> energy between reactants<br />

and products <strong>in</strong> their relaxed geometries and ∆Q is the equilibrium<br />

reaction coord<strong>in</strong>ate for the products.<br />

- if molecular vibrations are treated classically - the reaction occurs at the <strong>in</strong>tersection<br />

between the two curves <strong>in</strong> figure 2.1. If molecular vibrations are treated quantum me-<br />

chanically, each quantum of vibrational energy (phonon) has energy �ω, where ω is the<br />

frequency of vibration, and the reaction can occur at a lower or higher energy than the<br />

<strong>in</strong>tersection by absorption or emission of phonons. The matrix element |J+0,i;0+, f | 2 repre-<br />

sents the probability of electron and nuclear tunnell<strong>in</strong>g, whereas the Boltzmann factor<br />

represents the probability of reach<strong>in</strong>g the transition po<strong>in</strong>t. The summation is over both<br />

the <strong>in</strong>itial and the f<strong>in</strong>al vibrational modes.<br />

At this po<strong>in</strong>t let us look at the matrix element J+0,i;0+, f . Us<strong>in</strong>g assumption (d), that is<br />

that the electronic part of the matrix element is <strong>in</strong>dependent of nuclear coord<strong>in</strong>ates, one<br />

can rewrite equation for J:<br />

to:<br />

� ∞<br />

J+0,i;0+, f =<br />

−∞<br />

�� ∞<br />

J+0,i;0+, f =<br />

−∞<br />

dQ<br />

� ∞<br />

−∞<br />

dq Ψ ∗<br />

0+ (q; Q)χ∗0+,<br />

f (Q) ˆH(q)Ψ+0(q; Q)χ+0,i(Q) (2.4a)<br />

dQ χ ∗<br />

0+, f (Q)χ+0,i(Q)<br />

� �� ∞<br />

dq Ψ<br />

−∞<br />

∗<br />

0+ (q; Q) �<br />

ˆH(q)Ψ+0(q; Q)<br />

20<br />

(2.4b)


where q represents electronic coord<strong>in</strong>ates. In Dirac notation:<br />

J+0,i;0+, f = 〈χ+0,i|χ0+, f 〉〈Ψ+0| ˆH|Ψ0+〉 = 〈χ+0,i|χ0+, f 〉J+0;0+<br />

(2.4c)<br />

If the results of equation 2.4 is substituted <strong>in</strong>to 2.3, the rate equation can be expressed<br />

<strong>in</strong> terms which separate the electronic and nuclear degrees of freedom. The result<strong>in</strong>g<br />

equation will then have the form:<br />

W+0;0+ = 2π<br />

�<br />

|J+0;0+| 2 �<br />

i f<br />

〈χ+0,i|χ0+, f 〉 2 δ(E+0,i − E0+, f )exp(−E+0,i/kT)Z −1<br />

(2.5)<br />

Equation 2.5 is still not at all trivial to simplify. Follow<strong>in</strong>g [4], a sketch of the method<br />

employed is as follows: the δ function is treated by us<strong>in</strong>g Fourier Transforms and an<br />

analytic form of the vibrational coupl<strong>in</strong>g is obta<strong>in</strong>ed by assum<strong>in</strong>g that all the vibrational<br />

modes i and f are quadratic and have the same angular frequency ω. F<strong>in</strong>ally, the equation<br />

is reduced <strong>in</strong>to a form similar to a standard rate equation by assum<strong>in</strong>g that we are <strong>in</strong> the<br />

semi-classical (high temperature) limit, i.e. by assum<strong>in</strong>g that the energy of a quantum of<br />

vibrational energy �ω is smaller than kT and that the electronic coupl<strong>in</strong>g J+0;0+ is small<br />

when compared to the reorganisation energy λ. The f<strong>in</strong>al form of the rate equation is<br />

therefore:<br />

W+0;0+ =<br />

|J+0;0+| 2<br />

�<br />

� π<br />

λkT<br />

+ λ)2<br />

exp(−(∆E ) (2.6)<br />

4λkT<br />

This form of the equation is pleas<strong>in</strong>g because of its ease of <strong>in</strong>terpretation. There are<br />

three ma<strong>in</strong> terms appear<strong>in</strong>g <strong>in</strong> it: the quantity |J+0;0+| represents the strength of the elec-<br />

tronic coupl<strong>in</strong>g, the exponential term exp(− (∆E+λ)2<br />

) represents the Boltzmann factor for a<br />

4λkT<br />

state with energy equal to the <strong>in</strong>tersection po<strong>in</strong>t between the diabatic energy surfaces de-<br />

picted <strong>in</strong> fig 2.1 and f<strong>in</strong>ally the normalisation term � π<br />

λkT<br />

normalises the Boltzmann factor<br />

to a density of states and weighs it by the vibrionic overlap. There are three parameters<br />

<strong>in</strong> Equation 2.6 which depend on material properties: the reorganisation energy λ, the<br />

electronic coupl<strong>in</strong>g J+0;0+ and f<strong>in</strong>ally the difference <strong>in</strong> site energies ∆E. In the rest of this<br />

21


thesis the electronic coupl<strong>in</strong>g will sometimes be referred to as the transfer <strong>in</strong>tegral and<br />

will be denoted by the symbol J, without suffixes.<br />

There are two important conditions which must be satisfied <strong>in</strong> order for equation 2.6<br />

to be valid: the reaction must be non-adiabatic and kT must be significantly greater than<br />

the vibrational modes. The first condition is truly fundamental and can be cast <strong>in</strong> many,<br />

equivalent ways: the reaction must occur between diabatic states or the reaction must be<br />

sudden. If this is not the case, a form for the reaction rate can be obta<strong>in</strong>ed by assum<strong>in</strong>g<br />

that the probability of cross<strong>in</strong>g diabatic surfaces is unity [2]. If the electronic coupl<strong>in</strong>g<br />

is actually larger than the barrier height, it seems to me that the concept of localised<br />

states poorly describes the physical situation and charge transfer will be a coherent, not<br />

temperature activated process. As we will discuss <strong>in</strong> chapter five, this may be the regime<br />

under which <strong>in</strong>tracha<strong>in</strong> transport <strong>in</strong> polymers occurs. The second assumption can be<br />

overcome by divid<strong>in</strong>g the nuclear modes <strong>in</strong>to low frequency and high frequency modes,<br />

one can then obta<strong>in</strong> a formula for the electron transfer rate, the Marcus-Levich-Jortner<br />

equation[8]:<br />

W =<br />

|J+0;0+| 2<br />

�<br />

� π<br />

λlkT exp(−S h)<br />

∞�<br />

n=0<br />

S n<br />

h<br />

n! exp(−(∆E + λl + n�ωh) 2<br />

4λlkT<br />

) (2.7)<br />

where λl represents the contribution of low frequency modes to the reorganisation energy,<br />

ωh represents the high frequency and S h represents the Huang-Rhys factor for that fre-<br />

quency which is equal to S h = λh/(�ωh). This formula has pretty much the same form<br />

as equation 2.6, with the energy <strong>in</strong> the exponential term somewhat lowered by allow<strong>in</strong>g<br />

for the absorption of n phonons. Equations 2.7 and 2.6 have the same form and the same<br />

parameters: the only difference is that the high frequency modes are treated quantum<br />

mechanically <strong>in</strong> equation 2.7, therefore it is necessary to dist<strong>in</strong>guish a high frequency<br />

mode that leads from the reactants energy m<strong>in</strong>imum to the seam between the diabatic en-<br />

ergy surfaces and its contribution to λh. This mode is often referred to as the promot<strong>in</strong>g<br />

mode. Accord<strong>in</strong>g to Brédas and co-workers [9], us<strong>in</strong>g the Marcus-Levich-Jornter formula<br />

<strong>in</strong>stead of the high temperature Marcus theory charge transfer rate leads only to a renor-<br />

malisation of the rate not to a change <strong>in</strong> its temperature dependence. For this reason we<br />

have <strong>in</strong> general used the high temperature form of the equation 2.6.<br />

22


When simulat<strong>in</strong>g charge transport, two other forms for the charge transfer rate are<br />

commonly used: the Miller-Abrahams and symmetric rate equations. The Miller-Abrahams<br />

rate expression is written as:<br />

⎧<br />

⎪⎨ νe kT , ∆E > 0<br />

W =<br />

⎪⎩ ν, ∆E < 0<br />

− ∆E<br />

(2.8)<br />

where ∆E represents the difference <strong>in</strong> energies, ν is the rate prefactor. The motivation for<br />

such an expression is an assumption on the system be<strong>in</strong>g weakly coupled to the phonon<br />

bath. In these cases it is easy for phonons to absorb downward steps <strong>in</strong> energy but very<br />

hard for the opposite to occur. What this means is that the charge carriers are limited<br />

to energy match<strong>in</strong>g conditions only for upwards jumps, whereas <strong>in</strong> downward jumps the<br />

phonon bath will absorb the excess energy without accelerat<strong>in</strong>g the rate. These equations<br />

were obta<strong>in</strong>ed orig<strong>in</strong>ally to describe hopp<strong>in</strong>g from shallow impurity traps at helium tem-<br />

peratures and are based on a perturbational treatment <strong>in</strong> <strong>in</strong>organic solids. More precise<br />

non-perturbative treatment of the electron-phonon coupl<strong>in</strong>g [10] shows that <strong>in</strong> amorphous<br />

solids this expression is good at low temperatures, but fails at higher temperature. The<br />

motivation for present<strong>in</strong>g this equation is that it is used <strong>in</strong> the Gaussian Disorder Model,<br />

an empirical model for describ<strong>in</strong>g charge transport that we discuss <strong>in</strong> section 2.5.<br />

The symmetric rate has form:<br />

∆E (−<br />

W = νe 2kT )<br />

(2.9)<br />

this equation, actually, has the same form as equation 2.6. In fact, for small enough site<br />

energy difference ∆E, equation 2.6 can be written as:<br />

W =<br />

|J+0;0+| 2<br />

�<br />

�<br />

π<br />

λkT exp<br />

�<br />

− λ<br />

� �<br />

exp −<br />

4λkT<br />

∆E<br />

�<br />

2kT<br />

(2.10)<br />

In this form, equation 2.6 has the same energy dependence as 2.9, the only difference be-<br />

<strong>in</strong>g that the reorganisation energy leads to a temperature dependence <strong>in</strong> the rate prefactor<br />

ν of equation 2.9. The symmetric form of the equation is used by many workers <strong>in</strong> the<br />

field [11, 12, 13] to simulate the field dependence. In the limit of large reorganisation en-<br />

ergy and small field, the equation gives the same field dependence as the Marcus charge<br />

23


transfer equation. In order to recover the observed field dependence, workers us<strong>in</strong>g the<br />

symmetric rate explicitally write the temperature dependence of the rate <strong>in</strong> the Marcus<br />

theory form [14].<br />

2.2 Reorganisation Energy<br />

As we have shown <strong>in</strong> figure 2.1, the reorganisation energy <strong>in</strong> a reaction from an <strong>in</strong>itial<br />

electronic state to a f<strong>in</strong>al one corresponds to the difference <strong>in</strong> energy of the f<strong>in</strong>al elec-<br />

tronic state <strong>in</strong> the f<strong>in</strong>al nuclear coord<strong>in</strong>ates and <strong>in</strong> the <strong>in</strong>itial electronic state’s m<strong>in</strong>imum<br />

energy nuclear coord<strong>in</strong>ates. In pr<strong>in</strong>ciple, these energies and nuclear coord<strong>in</strong>ates do not<br />

only belong to the molecules <strong>in</strong>volved <strong>in</strong> charge hopp<strong>in</strong>g (<strong>in</strong>ternal sphere reorganisation),<br />

but also the polarisation of surround<strong>in</strong>g molecules (outer sphere reorganisation). In the<br />

case of the earliest charge transfer systems studied, such as metal ions <strong>in</strong> acqueous solu-<br />

tion, the outer sphere reorganisation is vastly greater than the <strong>in</strong>ner sphere reorganisation,<br />

because of the high polarity of the solvents [15, 16]. In contrast, most organic materials<br />

have low polarity and therefore it can be argued that the ma<strong>in</strong> contribution to the reorgan-<br />

isation energy is from <strong>in</strong>ner sphere reorganisation. It seems that part of the motivation for<br />

ignor<strong>in</strong>g the outer reorganisation <strong>in</strong> the solid state is that the problem <strong>in</strong> the solid state is<br />

very complicated <strong>in</strong>deed: as Barbara [17] notes <strong>in</strong> a glassy film it is hard to dist<strong>in</strong>guish<br />

the <strong>in</strong>termolecular modes which are free to move and are able to modify their geometry<br />

dur<strong>in</strong>g charge transport from those which are ”frozen”. The ”free” modes would be able<br />

to participate <strong>in</strong> the polarisation of the surround<strong>in</strong>g, whereas the ”frozen” modes would<br />

only participate <strong>in</strong> the site energy difference. Because of the low polarisabilities of the<br />

materials considered <strong>in</strong> this thesis, outer sphere reorganisation will <strong>in</strong> general be ignored.<br />

In order to determ<strong>in</strong>e the <strong>in</strong>ner sphere reorganisation energy for charge transport be-<br />

tween identical molecules we follow the method of Sakanoue [18]. In this method, the<br />

<strong>in</strong>teraction between the two molecules <strong>in</strong>volved <strong>in</strong> charge transport is assumed to rema<strong>in</strong><br />

unaltered upon charge transfer. In this approximation, therefore, the reorganisation energy<br />

is unaffected by orientational effects: it depends uniquely on the properties of the isolated<br />

molecule. In order to estimate it one has to consider two contributions, the relaxation of<br />

the neutral molecule and that of the charged molecule. We can therefore write:<br />

24


λi = λi1 + λi2<br />

(2.11a)<br />

where i labels the <strong>in</strong>ternal reorganisation energy. The relaxation energy of the neutral<br />

molecule is then def<strong>in</strong>ed as:<br />

λi1 = E0(Q+) − E0(Q0) (2.11b)<br />

where Q+ and Q0 represent respectively the equilibrium coord<strong>in</strong>ates of the charged and<br />

neutral molecule. Similarly the relaxation energy of the charged molecule can be written<br />

as:<br />

λi2 = E+(Q0) − E+(Q+) (2.11c)<br />

From a computational po<strong>in</strong>t of view, it is necessary to conduct two geometry opti-<br />

misations and four energy calculations. Many different quantum chemical methods are<br />

available, but Sakanoue [18] f<strong>in</strong>ds that the hybrid Density Functional B3LYP and the<br />

semiempirical method AM1 yield consistent results. The reorganisation energy can also<br />

be deduced experimentally from the vibrionic structure of the photoelectron spectrum. It<br />

has been shown that results from B3LYP for pentacene are <strong>in</strong> good agreement with these<br />

experimental f<strong>in</strong>d<strong>in</strong>gs [19]. There are also other techniques used to estimate the <strong>in</strong>ternal<br />

sphere reorganisation energy directly from calculation of the electron-phonon coupl<strong>in</strong>g<br />

[20]. This approach allows a better understand<strong>in</strong>g of the contribution of each vibrational<br />

mode to the reorganisation energy - although the same result can be obta<strong>in</strong>ed by frequency<br />

analysis [9].<br />

Regard<strong>in</strong>g the outer sphere reorganisation, most exist<strong>in</strong>g theories are based on di-<br />

electric cont<strong>in</strong>uum models because the first systems to be studied were donor-acceptor<br />

charge transfer systems <strong>in</strong> solution, where the solvent can conveniently be substituted by<br />

a cont<strong>in</strong>uum characterised only by static (zero frequency) and optical (<strong>in</strong>f<strong>in</strong>ite frequency)<br />

dielectric constants. These models have also been expanded to treat ellipsoidal cavities,<br />

but rema<strong>in</strong> of similar form [16]. It is also possible to use Molecular Dynamics, <strong>in</strong>clud<strong>in</strong>g<br />

25


some of the solvent <strong>in</strong> a shell around the donor acceptor molecules explicitly and substi-<br />

tut<strong>in</strong>g the rest by a cont<strong>in</strong>uum dielectric [21]. This k<strong>in</strong>d of approach may be applicable to<br />

the solid state, allow<strong>in</strong>g one to dist<strong>in</strong>guish the free modes from the frozen ones.<br />

2.3 Transfer Integral<br />

The overall rate of charge transfer is dictated by the transfer <strong>in</strong>tegral J0+;+0 which we have<br />

def<strong>in</strong>ed <strong>in</strong> equation 2.4. In order to evaluate this matrix element, we must have some<br />

knowledge of the electronic Hamiltonian and of the wavefunctions Ψ+0 and Ψ0+. The<br />

electronic Hamiltonian can be written as:<br />

H e = T e + U = Σi(T e i<br />

ZA<br />

− ΣA<br />

riA<br />

1<br />

) + ΣiΣ j>i<br />

ri j<br />

(2.12)<br />

where T e i represents the k<strong>in</strong>etic energy of the ith electron, ZA the nuclear charge of the<br />

A th nucleus, riA the distance between the i th electron and the A th nucleus and ri j represents<br />

the distance between the i th and j th electrons. The first term of this Hamiltonian is a one<br />

electron property and will be represented from now on by O1 = Σi(T e i<br />

ZA − ΣA ); the second<br />

riA<br />

term is a two electron property and will be represented by the symbol O2 = ΣiΣ j>i 1<br />

ri j .<br />

It is also necessary to have forms for the wavefunctions represent<strong>in</strong>g the molecular<br />

system with charge localised on either the components. In order to correctly represent<br />

the anti-symmetric nature of the wavefunction, Slater determ<strong>in</strong>ants of the molecular sp<strong>in</strong>-<br />

orbitals are used [22]. Slater determ<strong>in</strong>ants are anti-symmetrised sums of all permutations<br />

of a particular set of one electron wavefunctions (molecular orbitals). Imag<strong>in</strong>e that these<br />

orbitals are all localised on one molecule or on the other and label the ones localised on<br />

molecule A ψ Ai , and the ones localised on molecule B ψ Bi , where i is a label enumerat<strong>in</strong>g<br />

the spatial part of the wavefunction. A bar will be used to label sp<strong>in</strong> state. A possible<br />

Slater determ<strong>in</strong>ant represent<strong>in</strong>g a state with the charge on molecule A would therefore be:<br />

Ψ+0 = |ψ A0 ¯ψ A0 ψ B0 ¯ψ B0 . . . ψ An−1 ¯ψ An−1 ψ Bn−1 ¯ψ Bn−1 ¯ψ An ψ Bn ¯ψ Bn > (2.13)<br />

where the bold Ψ represents a Slater determ<strong>in</strong>ant and ψ represents a molecular orbital. If<br />

the charge is localised on molecule B, the state might be represented as:<br />

26


Ψ0+ = |ψ A0 ¯ψ A0 ψ B0 ¯ψ B0 . . . ψ An−1 ¯ψ An−1 ψ Bn−1 ¯ψ Bn−1 ψ An ¯ψ An ¯ψ Bn > (2.14)<br />

It is easy to notice that these two Slater determ<strong>in</strong>ants differ <strong>in</strong> only one sp<strong>in</strong>-orbital:<br />

ψ An and ψ Bn , assum<strong>in</strong>g that all sp<strong>in</strong> orbitals are orthonormal. We are therefore easily able<br />

to use Slater matrix rules [22] and write the form of the matrix element:<br />

+<br />

�<br />

µ=A0,B0...An,Bn<br />

J0+;+0 =< Ψ0+|O1 + O2|Ψ+0 >= (2.15)<br />

< ψ An |T e − ΣA<br />

ZA<br />

rA<br />

|ψ Bn ><br />

(< ψ An ψ Bn | 1<br />

r |ψµ ψ µ > − < ψ An ψ µ | 1<br />

r |ψBnψ µ >)<br />

The right hand side of the equation is now a matrix element of the Fock operator [22]. In<br />

other words we have shown that <strong>in</strong> go<strong>in</strong>g from a multi-electron Hamiltonian to a s<strong>in</strong>gle<br />

electron picture of molecular orbitals, the transfer <strong>in</strong>tegral of the Hamiltonian becomes<br />

equal to the matrix element of the Fock operator:<br />

J0+;+0 =< ψ An | ˆF|ψ Bn > (2.16)<br />

So long as we can represent the charge localised states as Slater determ<strong>in</strong>ants with only<br />

one s<strong>in</strong>gly occupied orbital, the problem of f<strong>in</strong>d<strong>in</strong>g the transfer <strong>in</strong>tegral corresponds to<br />

f<strong>in</strong>d<strong>in</strong>g the s<strong>in</strong>gly occupied orbitals on each molecule <strong>in</strong> the pair then f<strong>in</strong>d<strong>in</strong>g the Fock<br />

matrix for the whole dimer and tak<strong>in</strong>g the matrix element. When we perform such analysis<br />

with density functional methods, we use the Kohn-Sham matrix <strong>in</strong>stead of the Fock matrix<br />

and the Kohn-Sham orbitals <strong>in</strong>stead of the molecular orbitals. Three methods of achiev<strong>in</strong>g<br />

this will be discussed: by perform<strong>in</strong>g SCF calculations on a pair of molecules and tak<strong>in</strong>g<br />

the splitt<strong>in</strong>g between the frontier orbitals, by project<strong>in</strong>g the orbitals of the pair onto the<br />

molecular orbitals of the s<strong>in</strong>gle molecules and f<strong>in</strong>ally by direct calculation of the Fock<br />

matrix. The latter two methods were designed and implemented as part of the work for<br />

this thesis and will be discussed <strong>in</strong> the next chapter.<br />

27


2.3.1 Note on Systems with Degenerate Orbitals<br />

It is <strong>in</strong>terest<strong>in</strong>g to discuss the validity of equation 2.16 for the case of molecules with<br />

degenerate frontier orbitals. This is relevant to chapter 7, <strong>in</strong> which we discuss charge<br />

transport <strong>in</strong> highly symmetric molecules such as hexabenzocoronene (HBC). Because of<br />

its S6 symmetry HBC possesses doubly degenerate highest occupied molecular orbitals<br />

(HOMOs). In this case we argue that represent<strong>in</strong>g the charged state with a s<strong>in</strong>gle Slater<br />

determ<strong>in</strong>ant of the type represented <strong>in</strong> equation 2.13 is not correct, as the choice of which<br />

degenerate orbital to leave s<strong>in</strong>gly occupied is arbitrary. We should note that, upon charg-<br />

<strong>in</strong>g, Jahn-Teller <strong>in</strong>stabilities would cause the molecule to deform along one of its three<br />

symmetry axes, lose its S6 symmetry and lift the degeneracy of the HOMOs. Presumably<br />

a pair of such molecules would then both be able to deform and charge hopp<strong>in</strong>g would<br />

occur along one of n<strong>in</strong>e possible seams on the diabatic energy surfaces describ<strong>in</strong>g charge<br />

transport. This description is overly complicated and not fully consistent with the assump-<br />

tions made <strong>in</strong> deriv<strong>in</strong>g the Marcus charge transfer rate, as J will be different along each<br />

of these n<strong>in</strong>e seams; we wish to describe charge transfer with a s<strong>in</strong>gle effective transfer<br />

<strong>in</strong>tegral Je f f . Follow<strong>in</strong>g Newton [7] we wish to show that it is appropriate to simply pick<br />

the root mean square value of the transfer <strong>in</strong>tegrals for each pair of degenerate orbitals.<br />

Instead of account<strong>in</strong>g for the lift<strong>in</strong>g of the degeneracy, we assume that the charged states<br />

can be expressed as a superposition of two Slater determ<strong>in</strong>ants:<br />

ΨA = cos(θ)Ψ HOMOA<br />

ΨB = cos(ω)Ψ HOMOB<br />

+ s<strong>in</strong>(θ)Ψ HOMO−1A<br />

+ s<strong>in</strong>(ω)Ψ HOMO−1B<br />

(2.17a)<br />

(2.17b)<br />

where ΨA represents the wavefunction with charge localised on A, ΨB represents the wave-<br />

function with charge localised on B, the bold Ψ represents Slater determ<strong>in</strong>ants, the labels<br />

HOMO A<br />

represent the orbital which is s<strong>in</strong>gly occupied <strong>in</strong> the Slater determ<strong>in</strong>ant and θ and<br />

ω represent two mix<strong>in</strong>g angles. Us<strong>in</strong>g the same arguments we used to write equation 2.16<br />

28


we can calculate a transfer <strong>in</strong>tegral for a particular pair of superpositions ΦA and ΦB:<br />

J(θ, ω) = cos(θ)cos(ω)Jhomo A ;homo B + cos(θ)s<strong>in</strong>(ω)Jhomo A ;homo−1 B<br />

+ s<strong>in</strong>(θ) cos(ω)Jhomo−1 A ;homo B + s<strong>in</strong>(θ)s<strong>in</strong>(ω)Jhomo−1 A ;homo−1 B<br />

where the expressions Jµ;ν represent the matrix element of the Fock operator for molecular<br />

orbitals µ and ν. In order to obta<strong>in</strong> such an expression for an effective <strong>in</strong>tegral this equation<br />

is squared and averaged over all possible angles θ and ω. One then obta<strong>in</strong>s the expression:<br />

J 2<br />

e f f<br />

= J2<br />

homo A ;homo B + J 2<br />

homo A ;homo−1 B + J 2<br />

homo−1 A ;homo B + J 2<br />

homo−1 A ;homo−1 B<br />

4<br />

this is the expression that we use when treat<strong>in</strong>g highly symmetric systems.<br />

2.3.2 Splitt<strong>in</strong>g of the Molecular Orbitals<br />

(2.18)<br />

A method for calculat<strong>in</strong>g transfer <strong>in</strong>tegrals often used <strong>in</strong> the literature [23, 24, 9] consists<br />

<strong>in</strong> equat<strong>in</strong>g half the splitt<strong>in</strong>g <strong>in</strong> the energies of the frontier orbitals of a pair of identical<br />

molecules to the coupl<strong>in</strong>g between the frontier orbitals of the constituent molecules. In<br />

the case of hole transport, for example, one assumes that the molecular orbitals which are<br />

s<strong>in</strong>gly occupied <strong>in</strong> equations 2.13 and 2.14 are the HOMOs of either molecule: this is<br />

equivalent to <strong>in</strong>vok<strong>in</strong>g the frozen orbitals approximation. It is then assumed that the only<br />

non-zero elements of the Fock matrix <strong>in</strong>volv<strong>in</strong>g the HOMOs of the isolated molecules are<br />

the diagonal elements and one transfer <strong>in</strong>tegral, which connects the two HOMOs on either<br />

molecule. This latter assumption is equivalent to assum<strong>in</strong>g that the HOMO and HOMO-1<br />

of the pair of molecules are simply formed by a unitary transformation of the HOMOs of<br />

the isolated molecules. In light of these assumptions the problem of determ<strong>in</strong><strong>in</strong>g the two<br />

highest molecular orbitals of the pair can be expressed as the follow<strong>in</strong>g matrix problem,<br />

<strong>in</strong> the basis set of the HOMOs of the isolated molecules:<br />

H = EA J<br />

J EB<br />

29<br />

(2.19)


where J is the shorthand for the transfer <strong>in</strong>tegral def<strong>in</strong>ed <strong>in</strong> equation 2.16 and EA and<br />

EB represent the expectation values of Fock operator of the pair of molecules for the<br />

HOMOs on molecules A and B respectively. Solv<strong>in</strong>g this eigenvalue problem, we f<strong>in</strong>d<br />

the expression for the two HOMOs of the pair:<br />

E± = EA + EB<br />

2<br />

± 1<br />

2 ( � (EA − EB) 2 + (2J) 2 ) (2.20)<br />

If EA and EB are the same, by tak<strong>in</strong>g the difference between the HOMOs of the pair<br />

one gets 2J. This method was widely applied to a variety of organic materials[23, 24,<br />

25, 9], both electron conductors and hole conductors; electron conductors are treated<br />

by look<strong>in</strong>g at the lowest unoccupied molecular orbital. The model chemistry used by<br />

Brédas and co-workers [25, 9] was <strong>in</strong>dependent neglect of differential overlap (INDO),<br />

with Zerner’s parametrisation for spectroscopy (ZINDO) [27]. The ZINDO parametri-<br />

sation was used both because it allows one to treat very large systems and because the<br />

spectroscopic parametrisation guarantees good treatment of both the occupied and unoc-<br />

cupied frontier orbitals.<br />

The ma<strong>in</strong> limitation of consider<strong>in</strong>g the transfer <strong>in</strong>tegral as half the difference <strong>in</strong> orbital<br />

energies lies <strong>in</strong> the fact that it cannot deal with cases <strong>in</strong> which the diagonal elements EA<br />

and EB are not the same. In this case other methods must be used such as the fragment<br />

orbitals methods [28]: <strong>in</strong> the next chapter we will discuss how to obta<strong>in</strong> similar results<br />

from projection of the orbitals onto localised basis set.<br />

2.4 Site Energy Difference<br />

One of the aims of this thesis is to f<strong>in</strong>d accurate and efficient methods to calculate the<br />

difference <strong>in</strong> site energy ∆E and to elucidate its orig<strong>in</strong>. There are many possible orig<strong>in</strong>s<br />

for such driv<strong>in</strong>g force: if the molecules have different ionisation potential or electron<br />

aff<strong>in</strong>ity or if an external electric field was present. We have shown how, us<strong>in</strong>g the frozen<br />

orbital approximation, we can def<strong>in</strong>e transfer <strong>in</strong>tegrals of a multi-electron wave-function<br />

<strong>in</strong> terms of matrix elements of the Fock matrix. Similarly the site energies can be obta<strong>in</strong>ed<br />

by look<strong>in</strong>g at the diagonal elements of the Fock matrix <strong>in</strong> a localised basis set, and we<br />

30


Figure 2.2: Pictorial representation of the vectors discussed <strong>in</strong> deriv<strong>in</strong>g equation 2.26<br />

can argue that the difference between these two energies is the site energy difference<br />

that appears <strong>in</strong> equation 2.6. In this thesis we are ma<strong>in</strong>ly concerned with charge transfer<br />

between identical molecules and <strong>in</strong> section 3.3 we show that, for these cases, it is sufficient<br />

to consider electrostatic and <strong>in</strong>ductive effects to expla<strong>in</strong> the observed variations <strong>in</strong> site<br />

energy difference. In this section we expla<strong>in</strong> how to calculate the electrostatic <strong>in</strong>teraction<br />

between two charge distributions. We also expla<strong>in</strong> how to calculate <strong>in</strong>ductive <strong>in</strong>teractions,<br />

that is how to calculate the dipole <strong>in</strong>duced by a particular electric field and its <strong>in</strong>fluence<br />

on the <strong>in</strong>teraction energy between two molecules.<br />

2.4.1 Electrostatic Interactions <strong>in</strong> Molecules<br />

SCF quantum chemical calculations will produce a form of the electron density <strong>in</strong> terms<br />

of atomic orbitals. Rather than calculat<strong>in</strong>g the electron density of two molecules at close<br />

po<strong>in</strong>ts <strong>in</strong> space and calculat<strong>in</strong>g a huge number of charge charge <strong>in</strong>teractions, we want to<br />

<strong>in</strong>tegrate the various moments of the charge density and deduce a simpler form for the<br />

<strong>in</strong>teraction energy. The resource I have found most useful <strong>in</strong> study<strong>in</strong>g this topic has been<br />

a set of notes by A.J. Stone, available from the Cambridge website [29], still I have found<br />

this topic so confus<strong>in</strong>g that I decided to repeat the derivation and obta<strong>in</strong>ed a somewhat<br />

different approach.<br />

Consider the potential V(r) of a charge distribution ρ(r) at a po<strong>in</strong>t r as illustrated <strong>in</strong><br />

figure 2.2. The potential will be described as a sum of potential from all po<strong>in</strong>ts at distances<br />

r + δr:<br />

�<br />

V(r) =<br />

1<br />

4πɛ0<br />

ρ(δr)<br />

|r + δr| d3 (δr) (2.21)<br />

where the vertical bars || represent tak<strong>in</strong>g the modulus, ρ represents charge density and ɛ0<br />

31


is the permittivity of free space. Let us <strong>in</strong>troduce a function v such that:<br />

v(r + δr) = 1<br />

4πɛ0<br />

1<br />

|r + δr|<br />

(2.22)<br />

So long as |δr| is smaller than |r| we can write the Taylor expansion of this function, and<br />

know that it will converge:<br />

�<br />

v(r + δr) = v(r) + δrα(∇αv)r + 1 �<br />

δrαδrβ(∇α∇βv)r + ... (2.23)<br />

2<br />

α<br />

where α and β represents the summation of the coord<strong>in</strong>ates x, y, z. Substitut<strong>in</strong>g this ex-<br />

pression back <strong>in</strong>to 2.21 we arrive at sums of <strong>in</strong>tegrals of the charge density multiplied by<br />

derivatives of the distance vector r:<br />

αβ<br />

�<br />

V(r) = v(r) d 3 (δr)ρ(δr) (2.24)<br />

+ � �<br />

α (∇αv)r δrαd 3 (δr)ρ(δr)<br />

+ 1<br />

�<br />

�<br />

αβ (∇α∇βv)r δrαδrβd<br />

2<br />

3 (δr)ρ(δr) + ...<br />

The <strong>in</strong>tegrals <strong>in</strong> equation 2.24 are readily manipulated to yield the first three electrical<br />

moments of the charge distribution depicted on the right of figure 2.2:<br />

c = �<br />

dα = �<br />

Qαβ = �<br />

d 3 (δr)ρ(δr) (2.25)<br />

δrαd 3 (δr)ρ(δr)<br />

( 3<br />

2 δrαδrβ − 1<br />

2 |r|2 δαβ)d 3 (δr)ρ(δr)<br />

The first two quantities are readily recognizable: they are the total charge c and the<br />

dipole moment d. The third quantity is the so-called traceless quadrupole moment. The<br />

reader will notice that, compared to the quantity shown <strong>in</strong> equation 2.24, we have multi-<br />

plied it by three and subtracted a quantity from the diagonal elements to ensure that the<br />

trace of the matrix is zero. We are justified <strong>in</strong> do<strong>in</strong>g this because a quadrupole moment<br />

which is proportional to the unit matrix does not contribute to the potential of a charge<br />

32


Figure 2.3: Pictorial representation of the vectors discussed <strong>in</strong> deriv<strong>in</strong>g equation 2.32<br />

distribution. To see why this is justified, we rewrite equation 2.24 <strong>in</strong> terms of the quanti-<br />

ties def<strong>in</strong>ed <strong>in</strong> equation 2.25, add<strong>in</strong>g a term proportional to the unit matrix to make up for<br />

the fact that we have def<strong>in</strong>ed a traceless form of the quadrupole moment:<br />

�<br />

V(r) = v(r)c + (∇αv)rdα + 1 �<br />

�<br />

(∇α∇βv)rQαβ + Γ (∇α∇αv)r + ... (2.26)<br />

3<br />

α<br />

αβ<br />

where Γ collects the proportionality factors and the trace of the quadrupole moment which<br />

we are go<strong>in</strong>g to ignore. The fourth term of this equation is equal to zero. To see why,<br />

evaluate the expression (∇α∇αv)r for α correspond<strong>in</strong>g to x. This term will have the value:<br />

(∇x∇xv)r = 1<br />

� 2 3x 1<br />

−<br />

4πɛ0 |r| 5 |r| 3<br />

�<br />

α<br />

(2.27)<br />

If we add the xx, yy, and zz contributions they will sum up to zero. It is therefore pre-<br />

ferrable to write the quadrupole moment <strong>in</strong> a traceless basis as shown <strong>in</strong> equation 2.25.<br />

Equation 2.26 therefore shows the first few terms of the potential due to a charge<br />

distribution at a particular po<strong>in</strong>t <strong>in</strong> space. However we want to be able to calculate the<br />

<strong>in</strong>teraction energy between two charge distributions. This is better described by figure 2.3<br />

than by figure 2.2. In order to dist<strong>in</strong>guish two charge distributions we use labels A and B<br />

to distiguish the electrical moments of one molecule from those of the other.<br />

Aga<strong>in</strong> let us write the <strong>in</strong>teraction energy UAB as an <strong>in</strong>tegral:<br />

�<br />

UAB =<br />

ρ(δrA)V B (r − δrA)d 3 (δrA) (2.28)<br />

Once aga<strong>in</strong> assume that |δrA| is smaller than r to ensure the convergence of a Taylor series:<br />

33


�<br />

UAB =<br />

d 3 ⎛<br />

(δrA)ρ(δrA) ⎜⎝ V B �<br />

(r) −<br />

α<br />

(∇αV B )rδrAα + 1<br />

2<br />

�<br />

αβ<br />

(∇α∇βV B )rδrAαδrAβ + ...<br />

⎞<br />

⎟⎠<br />

(2.29)<br />

Once aga<strong>in</strong> we can take the terms <strong>in</strong> equation 2.29 that depend on δrA and collect them<br />

as electrical moments of molecule A:<br />

UAB = c A V B �<br />

(r) − d A α(∇αV B � 1<br />

)r +<br />

3 QA αβ(∇α∇βV B )r + ... (2.30)<br />

α<br />

Now we substitute the expression for V B from equation 2.26 <strong>in</strong>to equation 2.30. Before<br />

show<strong>in</strong>g this equation, however, let us <strong>in</strong>troduce another handy piece of notation. In equa-<br />

tion 2.26 the only part which depends on δr are the derivatives of v(r), so for convenience<br />

let us def<strong>in</strong>e the operator Tαβγ... as:<br />

αβ<br />

Tαβγ... = (∇α∇β∇γ∇...v)(r)<br />

Us<strong>in</strong>g this convenient notation the <strong>in</strong>teraction energy UAB can be written as:<br />

UAB = c A Tc B + c<br />

−<br />

+<br />

A �<br />

Tαd B α + c<br />

A �<br />

1<br />

Tαβ<br />

3 QB αβ<br />

α<br />

αβ<br />

�<br />

d<br />

α<br />

A αTαc B �<br />

− d<br />

αβ<br />

A αTαβd B �<br />

β − d<br />

αβγ<br />

A 1<br />

αTαβγ<br />

3 QB βγ<br />

� 1<br />

3<br />

αβ<br />

QA αβTαβc B � 1<br />

+<br />

3<br />

αβγ<br />

QA αβTαβγd B � 1<br />

γ +<br />

3<br />

αβγδ<br />

QA αβTαβγδ<br />

1<br />

3 QB γδ + ...<br />

(2.31)<br />

(2.32)<br />

So f<strong>in</strong>ally we have an expression for the <strong>in</strong>teraction between two charge densities <strong>in</strong><br />

terms of <strong>in</strong>teractions between electrical moments.<br />

Note that if each electrical moment is to correspond to a whole molecule, this ex-<br />

pression is valid only for those cases <strong>in</strong> which the two molecules are further apart than the<br />

largest distance <strong>in</strong> each molecule. In this case the electrical moments can be obta<strong>in</strong>ed from<br />

a quantum chemical calculation on one isolated molecule; most quantum chemistry pack-<br />

ages will calculate these quantites. In the case <strong>in</strong> which two molecules are closer than the<br />

largest distance <strong>in</strong>side the molecule, we have to partition the space <strong>in</strong> each molecule and<br />

34


substitute each partitioned area with a collection of monopoles, dipoles, quadrupoles etc.:<br />

this scheme is known as distributed multipole analysis. Some computational chemistry<br />

packages, such as GAMESS-UK [30], have modules to do this. In the case of Gaussian 03<br />

[31] a helper program gdma by A. Stone [32] is available. Distributed multipole analysis<br />

is the technique we will use to determ<strong>in</strong>e the electrostatic <strong>in</strong>teraction of large molecules at<br />

short distances. The distribute multipoles are obta<strong>in</strong>ed from the gdma program and C++<br />

libraries were written to calculate the <strong>in</strong>teractions from equation 2.33.<br />

2.4.2 Inductive Interactions <strong>in</strong> Molecules<br />

In the treatment described so far, we have assumed that it is possible to take the elec-<br />

trical moments from a calculation on an isolated molecule and apply them to a pair of<br />

molecules: we have assumed that the electrical field due to one molecule does not <strong>in</strong>flu-<br />

ence the charge distribution of the other. In order to improve on this model it is possible<br />

to calculate the dipole-polarisability of a molecule: the tensor that determ<strong>in</strong>es the dipole<br />

<strong>in</strong>duced on one molecule by the electric field due to the other. This polarisability can<br />

be obta<strong>in</strong>ed from a quantum chemical calculation, provided that the first derivatives of<br />

the energy are calculated. The polarisability returned by Gaussian corresponds to a static<br />

polarisability, but excludes rotational and vibrational effects. The polarisability will <strong>in</strong><br />

general be a 3x3 matrix Pαβ. The <strong>in</strong>duced dipole at molecule A will therefore have form:<br />

d A<br />

<strong>in</strong>d α = PαβF B β<br />

(2.33)<br />

where FB is the electrical field from molecule B and dA <strong>in</strong>d is the <strong>in</strong>duced dipole moment.<br />

This <strong>in</strong>duced dipole will <strong>in</strong> turn change the value of the electric field due to molecule A<br />

at molecule B and <strong>in</strong>duce a different dipole on B, which will <strong>in</strong> turn change F B β<br />

and dA<br />

<strong>in</strong>d .<br />

In the implementation of the protocol that we use <strong>in</strong> section 3.3, a self consistent solution<br />

to this iterative procedure is implemented. In other words corrections are calculated until<br />

they are smaller than a token value. It is <strong>in</strong>tuitively obvious that this series will tend to<br />

converge rather rapidly: <strong>in</strong>duced dipoles tend to lower the electric field and therefore each<br />

term will be decreas<strong>in</strong>g <strong>in</strong> value.<br />

In the previous paragraph, we have described how to treat polarisabity by us<strong>in</strong>g a<br />

35


s<strong>in</strong>gle polarisability for each molecule. A caveat is that we have implicitly assumed the<br />

electric field to be uniform <strong>in</strong> the molecule; if the molecules are close relative to their<br />

overall size, this will not be a fair assumption. Schemes to treat this situation exist: for<br />

example G.A. Kam<strong>in</strong>ski et al [33] developed models of molecules where partitioned zones<br />

each have their own polarisability which is deduced by perform<strong>in</strong>g DFT calculations <strong>in</strong><br />

the presence of po<strong>in</strong>t charges, measur<strong>in</strong>g the change <strong>in</strong> the electrostatic potential of a<br />

molecule and fitt<strong>in</strong>g this potential to particular values of the polarisability. E.V. Tsiper<br />

and Z.G. Soos, <strong>in</strong> their description of the b<strong>in</strong>d<strong>in</strong>g energy of pentacene crystals [34], de-<br />

duce atom polarisabilities by perform<strong>in</strong>g semi-empirical calculation <strong>in</strong> the presence of an<br />

electric field, then separate the <strong>in</strong>duced dipole <strong>in</strong>to two contributions: one from charge<br />

reorganisation <strong>in</strong> the molecule, the other from the creation of atomic dipoles. From this<br />

latter contribution an atomic polarisability is deduced. Yet another technique is that of<br />

A. Stone [35] where distributed multipole analysis is performed for a variety of applied<br />

fields, lead<strong>in</strong>g to the deduction of a polarisability for each zone considered <strong>in</strong> the mul-<br />

tipole analysis. We have not implemented any of these approaches to calculate dipole-<br />

polarizabilities <strong>in</strong> non-uniform fields and correct our calculations by an <strong>in</strong>ductive term<br />

only <strong>in</strong> the case when molecules are far enough apart to consider the electric field across<br />

them uniform.<br />

2.5 The Gaussian Disorder Model and Related Methods<br />

All the methods discussed <strong>in</strong> this chapter treat charge transport events as hops between<br />

different localised states. In this section we give an overview of the methods to obta<strong>in</strong><br />

macroscopic charge mobilities from the microscopic rates that we discuss <strong>in</strong> the previous<br />

sections. <strong>Charge</strong> hops between molecules can be represented as hops between different<br />

sites on a lattice. Figure 2.4 shows how charge transport can be envisioned to occur <strong>in</strong><br />

a conjugated material. In this figure we are try<strong>in</strong>g to represent two possible forms of<br />

disorder: the position of the sites where the charges are localised and and the coupl<strong>in</strong>g<br />

between the sites (i.e. the rate of transfer of charge from site to site). Disorder <strong>in</strong> the<br />

energies of the charge transport<strong>in</strong>g site is represented <strong>in</strong> the figure by the distribution<br />

<strong>in</strong> heights of energy levels and disorder <strong>in</strong> the electronic coupl<strong>in</strong>g by a distribution <strong>in</strong><br />

36


Figure 2.4: A schematic of one dimensional transport through an energetically and positionally<br />

disordered medium. This figure represents disorder <strong>in</strong> energy (vertical<br />

placement of the transport levels) and <strong>in</strong> transfer <strong>in</strong>tegrals (thickness of<br />

arrows)<br />

orientations and separations which leads to different transfer <strong>in</strong>tegrals (represented by the<br />

thickness of the arrows). A typical model of charge transport <strong>in</strong> a material will therefore<br />

comprise three elements:<br />

(a) an appropriate computational or analytical method to solve charge dynamics<br />

(b) an appropriate microscopic equation for hopp<strong>in</strong>g rates<br />

(c) a distribution function for the energies of the sites and, <strong>in</strong> some cases, the electronic<br />

coupl<strong>in</strong>g between sites accord<strong>in</strong>g to a particular distribution.<br />

The second of these three aspects is discussed <strong>in</strong> section 2.1, the other two are discussed<br />

separately <strong>in</strong> the next sections. F<strong>in</strong>ally some results from the literature are reviewed<br />

briefly.<br />

2.5.1 Computational Methods<br />

Two numerical methods are commonly used to simulate microscopic charge transport: the<br />

Master equation and Monte Carlo approach. In order to model mobility, the experiment<br />

modelled is usually a time of flight (ToF) experiment. In a time of flight experiment charge<br />

pairs are generated near a semi-transparent, non-<strong>in</strong>jective electrode. Once separated by<br />

an electric fied F either electrons or holes (depend<strong>in</strong>g on the sign of the applied field)<br />

will drift towards the other electrode. This drift causes a displacement current which falls<br />

to zero as the carriers reach the counter electrode. Thus a time for carriers to reach the<br />

counter electrode electrode can be extracted and know<strong>in</strong>g the thickness of the film allows<br />

the measurement of the charge’s velocity and hence the charge mobility for a particular<br />

field and temperature. More <strong>in</strong>formation can be found <strong>in</strong> references [36] [37] .<br />

37


The Master equation is a differential equation that describes the time behaviour of<br />

stochastic systems [38], i.e. one whose state can only be described <strong>in</strong> terms of probabil-<br />

ities. If the states of a system are labelled 0, 1, 2, ..., m , for example, the state of the<br />

system at time t will be described by a set of probabilities P0(t), P1(t), Pm(t). The Master<br />

equation simply states that:<br />

dP(n, t)<br />

dt<br />

�<br />

= (AmnPm(t) − AnmPn(t)) (2.34)<br />

m�n<br />

where Amn is the transition rate from state m to state n. The problem can conveniently be<br />

cast <strong>in</strong> matrix form if one writes the diagonal elements of the matrix A as:<br />

�<br />

Ann = −<br />

m�n<br />

Anm<br />

(2.35)<br />

If the rates A are <strong>in</strong>dependent of the probabilities, the problem is l<strong>in</strong>ear and can be readily<br />

solved.<br />

The power of this method is that it is applicable to a vast number of systems and<br />

that it is possible to determ<strong>in</strong>e the steady state behavior by simply sett<strong>in</strong>g the right side<br />

of equation 2.34 to zero. For the particular case of charge transport, <strong>in</strong> the most formal<br />

approach Pm(t) would describe the probability of be<strong>in</strong>g <strong>in</strong> the m th many-electron state at<br />

time t. For example, for a system conta<strong>in</strong><strong>in</strong>g two states, each of which is allowed to carry<br />

either one electron or none, there would be four many-electron states: two where either<br />

of the states is occupied, one where neither is occupied and one where both are occupied.<br />

This approach becomes unwieldy very quickly, as the number of many-electron states<br />

<strong>in</strong>creases as 2 N where N is the number of sites. A common approach [39] [13] is to use<br />

Pm(t) as the probability of occupation of the m th site on the lattice at time t, the rates Amn<br />

are then assumed to be <strong>in</strong>dependent of P. For a one-dimensional model the steady state<br />

Master equation can be solved analytically, even <strong>in</strong> the presence of random disorder <strong>in</strong><br />

the hopp<strong>in</strong>g rates[39]. For higher dimensionalities, and with more physical values for<br />

disorder, the problem can be treated as a l<strong>in</strong>ear problem so long as carrier concentrations<br />

38


are low. Then, if the probabilities are written as a vector and the rates as a matrix:<br />

AP = 0 (2.36a)<br />

�<br />

Pi = 1 (2.36b)<br />

i<br />

The second equation states normalisation. In such an approach, equation 2.36a can be<br />

solved for the Pi and the velocity can be measured from the steady state values of P and<br />

A:<br />

v = d¯r<br />

dt =<br />

�<br />

¯rmnAmnPm<br />

m�n<br />

where rmn is the distance between sites m and n. Mobility is then obta<strong>in</strong>ed as µ = v<br />

F .<br />

(2.37)<br />

Monte Carlo methods are another way of calculat<strong>in</strong>g the properties of stochastic sys-<br />

tems. The term Monte Carlo refers to any method that uses random numbers. In this<br />

case the system is modelled as a large regular lattice or as a disordered assembly of lat-<br />

tice po<strong>in</strong>ts. Once energies are distributed accord<strong>in</strong>g to the particular distribution function<br />

used, carriers are generated and are assigned wait<strong>in</strong>g times and a dest<strong>in</strong>ation accord<strong>in</strong>g to<br />

the particular formula for the rates which is used [41] [42]. In the adaptive time step ver-<br />

sion of the method, time is advanced by the shortest wait<strong>in</strong>g time and the carrier with the<br />

shortest wait<strong>in</strong>g time is moved. If necessary, boundary conditions <strong>in</strong> the directions per-<br />

pendicular to the field are applied. When a carrier reaches the counter electrode its time<br />

of arrival is recorded. The simulation cont<strong>in</strong>ues until all or most carriers reach the counter<br />

electrode. Simulations are repeated over different <strong>in</strong>itial carrier distributions and realisa-<br />

tions of the lattice until a representative photocurrent transient is obta<strong>in</strong>ed. Then the mean<br />

arrival time is calculated and mobility is deduced. The great advantage of the adaptive<br />

time step Monte Carlo method is that it can deal with extremely non-l<strong>in</strong>ear, asymmetric<br />

hopp<strong>in</strong>g rates. For example, we have discussed how, <strong>in</strong> the Master equation approach, the<br />

rates Amn are rout<strong>in</strong>ely treated as <strong>in</strong>dependent of the occupation probabilities <strong>in</strong> order to<br />

make the problem l<strong>in</strong>ear. This makes it impossible to <strong>in</strong>clude effects where multiple occu-<br />

pation of a site is not allowed. In a Monte Carlo program, add<strong>in</strong>g this feature is extremely<br />

easy; one simply does not assign as a possible dest<strong>in</strong>ation a site which is occupied! Monte<br />

39


Carlo simulations are then run at a number of different fields and temperatures to model<br />

the results of ToF experiments.<br />

2.5.2 Distribution of <strong>Parameters</strong> of the Transfer Rates<br />

In this section, we will discuss some of the different models that have been used to assign<br />

distributions of energies. Two types of disorder can be considered: disorder <strong>in</strong> the energy<br />

of the hopp<strong>in</strong>g sites and disorder <strong>in</strong> the transfer <strong>in</strong>tegrals between molecules. These two<br />

types of disorder are sometimes called energetic and positional, they are also referred to<br />

as diagonal and off-diagonal disorder. In Equation 2.6, these two types of disorder would<br />

represent distributions <strong>in</strong> the values of ∆E and J respectively. The simplest assumption<br />

<strong>in</strong> the treatment of energetic disorder is that a material has no long range order and the<br />

only source of disorder is the random local energetic landscape. In this case the energetic<br />

disorder is simply treated as a Gaussian distribution of width σ. The off-diagonal disorder<br />

is modelled <strong>in</strong> the follow<strong>in</strong>g way by H. Bässler and co-workers [41] [42]. The transfer<br />

<strong>in</strong>tegral is assumed to vary exponentially with distance, with a wave function overlap<br />

parameter γi j for sites i and j. So:<br />

|J| 2 = |J0| 2 e −2γi jRi j (2.38)<br />

where Ri j is the distance between the sites. The wave function overlap γi j parameter<br />

represents the <strong>in</strong>verse of the spatial extent of the wave function. In the model γi j is allowed<br />

to vary <strong>in</strong> the follow<strong>in</strong>g way:<br />

2γi j = Γi + Γ j<br />

a<br />

= Γi j<br />

a<br />

(2.39)<br />

where Γi is a parameter specific to site i drawn from a gaussian distribution of width Σ<br />

and a is the lattice constant for the simulation. We can already see how the <strong>in</strong>terpretation<br />

of off-diagonal disorder <strong>in</strong> real materials will be difficult: whereas <strong>in</strong> the model it is<br />

assumed to simply represent the variation <strong>in</strong> <strong>in</strong>tersite separations, <strong>in</strong> a real material it will<br />

also depend on orientational effects. These forms for the distributions of site energy and<br />

transfer <strong>in</strong>tegral, together with the usage of a Miller-Abraham rate equation and a Monte<br />

Carlo rout<strong>in</strong>e, form the basis for the well known Gaussian Disorder Model (GDM) due to<br />

40


H. Bässler and coworkers [41]. The GDM was used to determ<strong>in</strong>e an empirical equation<br />

[42] that describes the temperature and field dependences of ToF mobility data <strong>in</strong> terms<br />

of the parameters σ and Σ. It was especially succesful <strong>in</strong> describ<strong>in</strong>g the Poole-Frenkel<br />

like field dependence of mobility under high fields:<br />

⎧<br />

⎪⎨ µ0e<br />

µ =<br />

⎪⎩ µ0e<br />

2 −( 3kT σ)2 σ C[( e kT )2−Σ2 ] √ F , Σ > 1.5<br />

2 −( 3kT σ)2 σ C[( e kT )2−2.25] √ F , Σ < 1.5<br />

(2.40)<br />

where µ0 represents the zero field mobility, high temperature mobility and C is a lattice<br />

related constant. Once data for a particular material are fitted with equation 2.40, that<br />

material is assigned particular values of the empirical parameters µ0 , Σ, σ and C.<br />

The value of σ has been <strong>in</strong>terpreted <strong>in</strong> terms of the dipole moment of the constituents<br />

of the system by several groups[44, 47, 11]. The assumption made by [44] is that the<br />

diagonal disorder has two contributions, from van der Waals and dipole <strong>in</strong>teractions.<br />

σ 2 = σ 2<br />

d + σ2vdw<br />

(2.41)<br />

The effect of randomly oriented dipoles on the energetic landscape was modelled and it<br />

was found that σd is proportional to the dipole moment [44]. When the dipole moment<br />

of small conjugated molecules was plotted aga<strong>in</strong>st the values of the diagonal disorder<br />

extracted by GDM analysis of measured mobility, the plot figure 2.5 was obta<strong>in</strong>ed [44].<br />

If the value of σvdW is extrapolated from the <strong>in</strong>tercept at p=0 and σd is extrapolated with<br />

equation 2.41 the graph on the left of figure 2.5 is obta<strong>in</strong>ed, <strong>in</strong>dicat<strong>in</strong>g that the dipole<br />

<strong>in</strong>duced disorder is proportional to the magnitude of the dipole moment, as expected. The<br />

observation that a possible source of disorder could be dipole-dipole <strong>in</strong>teractions led S.V.<br />

Novikov and coworkers to challenge the view that long range order is not present <strong>in</strong> these<br />

materials: dipole <strong>in</strong>teractions <strong>in</strong> fact are long range and would lead to a spatial correlation<br />

<strong>in</strong> site energies. Novikov calculated the value of the correlation [45] for a cubic lattice as:<br />

< E(r)E(0) >= 0.74σ 2 a<br />

d<br />

r<br />

(2.42)<br />

where the triangular brackets denote averag<strong>in</strong>g, r is the vector separat<strong>in</strong>g the sites and<br />

41


Figure 2.5: σ versus the value of the dipole moment for five different small conjugated<br />

molecules (right panel)[44], σd obta<strong>in</strong>ed from equation 2.41 (left panel)<br />

a is the lattice constant. Figure 2.6 [45] shows a pictorial representation of the dipole<br />

potential <strong>in</strong> a lattice of randomly oriented dipoles. It can be seen clearly from figure 2.6<br />

that there are rather large doma<strong>in</strong>s of similar magnitudes of energy, i.e. that the energies<br />

are correlated.<br />

Other models have been proposed to expla<strong>in</strong> how correlation <strong>in</strong> energy could arise<br />

<strong>in</strong> non-polar materials: <strong>in</strong>clud<strong>in</strong>g the quadrupolar glass model [48] and the Los Alamos<br />

model [13]. The quadrupolar glass model is similar to the correlated dipole disorder<br />

model, but substitutes dipole <strong>in</strong>teractions with higher electric moment <strong>in</strong>teractions (quadrupole)<br />

<strong>in</strong>teraction. Both these approaches are formally correct only <strong>in</strong> the case of small, well<br />

separated molecules. In the case of large densily packed molecules the electrostatic <strong>in</strong>-<br />

teractions should be treated us<strong>in</strong>g distributed multipole analysis. The Los Alamos [13]<br />

model assigns correlations <strong>in</strong> polymers such as polyphenylenev<strong>in</strong>ylene to fluctuations <strong>in</strong><br />

the torsional angle between neighbour<strong>in</strong>g benzene r<strong>in</strong>gs. The necessity for <strong>in</strong>troduc<strong>in</strong>g<br />

correlations <strong>in</strong> the models will be expla<strong>in</strong>ed <strong>in</strong> the next section, where we discuss the<br />

successful predictions and short-com<strong>in</strong>gs of these models.<br />

42


Figure 2.6: Pictorial representation of correlation <strong>in</strong> a 100x100x100 cubic lattice of randomly<br />

oriented dipoles. The diameter of the spheres is proportional to the<br />

magnitude of the dipole <strong>in</strong>teraction at that po<strong>in</strong>t, whereas the colour of the<br />

spheres represents the sign of the potential [45]<br />

2.5.3 Predictions and Results<br />

In this section we discuss the correct predictions and short-com<strong>in</strong>gs of the models de-<br />

scribed so far. The first model to be truly successful at expla<strong>in</strong><strong>in</strong>g the results from ToF<br />

experiments was the GDM. It successfully expla<strong>in</strong>s the process of energetic relaxation<br />

<strong>in</strong> non-dispersive materials, through which the energy of a packet of carriers relaxes at<br />

long times to a constant value. This process is shown schematically <strong>in</strong> figure 2.7. It<br />

also expla<strong>in</strong>s anomalous broaden<strong>in</strong>g, that is the fact that <strong>in</strong> materials with large values<br />

of σ, eD<br />

µkT<br />

>> 1 , contrary to what would be predicted by the E<strong>in</strong>ste<strong>in</strong> relation between<br />

diffusivity D and drift mobility µ.<br />

The GDM also expla<strong>in</strong>ed that dispersive transport, where mobility depends on the<br />

thickness of the material and energetic relaxation is not achieved, results from high values<br />

of σ. The model <strong>in</strong>vokes a critical temperature Tc, below which transport is dispersive.<br />

Importantly, the GDM leads to equations 2.40, express<strong>in</strong>g a non-Arrhenius temperature<br />

dependence and strong electric field dependence of mobility. When off-diagonal disorder<br />

is present, the GDM also expla<strong>in</strong>s the odd phenomenon of negative field dependence of<br />

mobility, that is the phenomenon by which mobility can decrease with <strong>in</strong>creas<strong>in</strong>g fields.<br />

43


Figure 2.7: Energy distribution of charges as a function of time after generation [46].<br />

This was expla<strong>in</strong>ed by argu<strong>in</strong>g that if the transfer <strong>in</strong>tegrals along the route parallel to the<br />

field are smaller than those along a route that goes aga<strong>in</strong>st the field, <strong>in</strong>creas<strong>in</strong>g the field<br />

will have the effect of slow<strong>in</strong>g down the charge transfer process by mak<strong>in</strong>g the route with<br />

steps aga<strong>in</strong>st the field slower.<br />

The correlated disorder models were <strong>in</strong>troduced ma<strong>in</strong>ly to expla<strong>in</strong> a shortcom<strong>in</strong>g of<br />

the GDM: the fact that it can only expla<strong>in</strong> Poole-Frenkel like behavior for electric fields<br />

greater than 10 6 V/cm, whereas such behaviour is observed experimentally for fields ten<br />

times smaller [47]. Y.N. Garste<strong>in</strong> and E.M. Conwell’s simulations of the correlated dipole<br />

disorder model, where energetic disorder conta<strong>in</strong>s a contribution from dipolar <strong>in</strong>teractions<br />

and no off-diagonal disorder is used, do <strong>in</strong> fact show Poole-Frenkel behaviour for far<br />

lower fields. They also show that field dependence is not as dependent on the particular<br />

type of hopp<strong>in</strong>g equation used. The quadrupolar glass model and the Los Alamos model,<br />

are simply explanations of how correlations can occur for non polar materials. The Los<br />

Alamos model predicts a different dependence on temperature: ln(µ) ∝ 1<br />

kT<br />

as opposed to<br />

the behaviour predicted by the GDM: ln(µ) ∝ 1 2<br />

however it is very difficult to recog-<br />

kT<br />

nize which of the two expressions is most <strong>in</strong> agreement with experiment. An important<br />

achievement of these models is <strong>in</strong> expla<strong>in</strong><strong>in</strong>g how Poole-Frenkel behavior is observed <strong>in</strong><br />

such a large class of materials, and <strong>in</strong> achiev<strong>in</strong>g this by us<strong>in</strong>g a relatively small set of<br />

tunable parameters.<br />

44


Figure 2.8: Schematic of different routes from A to B <strong>in</strong> the direction of the field [46]<br />

The ma<strong>in</strong> limitation of the GDM is the difficulty to relate the parameters σ and Σ to<br />

microscopic physical and chemical properties of a material. This difficulty is due to two<br />

factors: the use of a Miller-Abrahams rate and of a cubic lattice. Us<strong>in</strong>g a Miller-Abrahams<br />

rate makes it difficult to calculate the paramters for charge transport because the correct<br />

models for non-adiabatic charge transport <strong>in</strong> conjugated materials must be consisted with<br />

Marcus theory, as described <strong>in</strong> section 2.1. This problem has been corrected by P.E. Parris<br />

and co-workers [14] by us<strong>in</strong>g a rate equation equivalent to equation 2.6. P.E. Parris’s<br />

approach, which we shall call polaronic GDM, was compared to the GDM by T. Kreouzis<br />

and co-workers [49]. They found that charge transport <strong>in</strong> polyfluorene is better described<br />

by the polaronic version than by the GDM, confirm<strong>in</strong>g that Marcus theory is <strong>in</strong>deed the<br />

correct form for charge transfer rate <strong>in</strong> conjugated materials. In chapter six of this thesis<br />

we will compare the GDM parameters of different materials to microscopic calculations<br />

of site energy differences and transfer <strong>in</strong>tegrals <strong>in</strong> order to show that the trends <strong>in</strong> GDM<br />

parameters are nevertheless consistent with the microscopic parameters.<br />

The use of a cubic lattice <strong>in</strong> the GDM is also a shortcom<strong>in</strong>g, as it makes it impossible<br />

to dist<strong>in</strong>guish the effects of electronic structure from those deriv<strong>in</strong>g from morphology.<br />

We try and approach this problem <strong>in</strong> chapter seven where we describe charge transport<br />

<strong>in</strong> polyfluorene and <strong>in</strong> hexabenzocoronene by us<strong>in</strong>g models of the morphology of the<br />

material to determ<strong>in</strong>e the irregular lattice that charges will hop <strong>in</strong>, to calculate the mi-<br />

croscopic parameters of equation 2.6 and f<strong>in</strong>ally to run Monte Carlo and Master equation<br />

45


simulations of the charge dynamics. This approach allows us to p<strong>in</strong>po<strong>in</strong>t exactly which<br />

microscopic parameters determ<strong>in</strong>e charge transport properties.<br />

The GDM and the other models we have briefly discussed have the great power of<br />

be<strong>in</strong>g able to describe charge transport <strong>in</strong> many different materials <strong>in</strong> a unified theoret-<br />

ical framework. We believe that this advantage is also a shortcom<strong>in</strong>g as a microscopic<br />

understand<strong>in</strong>g of the chemical and physical properties which lead to charge transport<br />

characteristics <strong>in</strong> a particular material must be based on the properties of that particular<br />

material.<br />

2.6 Conclusions<br />

In this chapter we have <strong>in</strong>troduced the theoretical framework for treat<strong>in</strong>g microscopic <strong>in</strong>-<br />

termolecular charge hopp<strong>in</strong>g events: the non-adiabatic high temperature Marcus theory<br />

charge transfer rate equation. We have described the three parameters of this equation:<br />

the transfer <strong>in</strong>tegral J, the reorganisation energy λ and the site energy differene ∆E. We<br />

have also presented some background <strong>in</strong>formation of how to calculate <strong>in</strong>teractions be-<br />

tween charge distributions, which we will use to calculate site energy differences <strong>in</strong> the<br />

next chapter. In the second part of the chapter we have discussed empirical models of<br />

charge transport, especially the Gaussian Disorder Model. We have argued that its ma<strong>in</strong><br />

shortcom<strong>in</strong>g is the difficulty of relat<strong>in</strong>g its parameters with chemical and physical char-<br />

acterstics of molecules. In the next chapters we will further discuss methods to calculate<br />

the Marcus charge transfer parameters, with an emphasis on f<strong>in</strong>d<strong>in</strong>g efficient methods for<br />

calculat<strong>in</strong>g J and ∆E and on def<strong>in</strong><strong>in</strong>g the charge transport<strong>in</strong>g unit <strong>in</strong> a polymer. We will<br />

also consider three different types of uses of these methods: to justify choice of param-<br />

eters <strong>in</strong> Monte Carlo simulation of ambi-polar mobility <strong>in</strong> fullerene doped polystyrene,<br />

to compare the GDM parameters of materials to the distribution <strong>in</strong> Marcus charge trans-<br />

fer parameters of randomly oriented pairs of molecules and f<strong>in</strong>ally to parametrise Monte<br />

Carlo simulations on non-cubic lattices derived from morphology simulations of polyflu-<br />

orene and hexabenzocoronene.<br />

46


Bibliography<br />

[1] R.A. Marcus, On the Theory of Oxidation Reduction Reaction Involv<strong>in</strong>g Electron<br />

Transfer, Journal of Chemical Physics, 24, 966 (1955)<br />

[2] R.A. Marcus and N. Sut<strong>in</strong>, Electron transfer <strong>in</strong> chemistry and biology, Biochimica et<br />

Biophysica acta 811, 265 (1985)<br />

[3] N.R. Kestner, J. Logan and J. Jortner, Thermal Electron Transfer Reactions <strong>in</strong> Polar<br />

Solvents, Journal of Chemical Physics, 21 , 2148 (1974)<br />

[4] K.F. Freed and J. Jortner, Multiphonon Processes <strong>in</strong> the Nonradiative Decay of Large<br />

Molecules, Journal of Chemical Physics, 52 , 6272 (1970)<br />

[5] B.S. Brunshwig, J. Logan, M.D. Newton and N. Sut<strong>in</strong>, A Semiclassical Treatment of<br />

Electron Exhange Reactions. Application to the Hexaaquoiron(II)-Hexaaquoiron(III)<br />

System, Journal of the American Chemical Society, 102, 5798 (1980)<br />

[6] L.D Landau, E.M Lifshitz, Quantum Mechanics non-relativistic theory, Pergamon<br />

Press (1965), 42<br />

[7] M.D. Newton, Quantum Mechanical Probes of Electron-Transfer K<strong>in</strong>etics: the Na-<br />

ture of Donor-Acceptor Interactions, Chemical Reviews, 91, 767 (1991)<br />

[8] J. Jortner, M. Bixon, T. Lagenbacher and M.E. Michele-Beyerle, <strong>Charge</strong> Transfer and<br />

transport <strong>in</strong> DNA, Proceed<strong>in</strong>gs of the National Academy of Science 95, 12759 (1997)<br />

[9] J.L. Brédas, D. Beljonne, V. Coropceanu and J. Cornil, <strong>Charge</strong> Transfer and Energy<br />

Transfer processes <strong>in</strong> π-conjugated Oligomers and Polymers: A Molecular picture,<br />

Chemical Reviews, (104), 4971, (2004)<br />

47


[10] D. Em<strong>in</strong>, Phonon-Assisted Jump Rate <strong>in</strong> Non-Crystall<strong>in</strong>e Solids, Physics Review<br />

Letters 32, 303 (1973)<br />

[11] S.V. Novikov, D.H. Dunlap, V.M. Kenkre, P.E. Parris and A.V. Vannikov, Essential<br />

Role of Correlations <strong>in</strong> Govern<strong>in</strong>g <strong>Charge</strong> <strong>Transport</strong> <strong>in</strong> Disordered Organic Materials,<br />

Physical Review Letters 81, 4472 (1998)<br />

[12] Z.G. Yu, D.L. Smith, A. Saxena, R.L. Mart<strong>in</strong> and A.R. Bishop, Molecular Geometry<br />

Fluctuation Model for the Mobility of Conjugated Polymers, Physical Review Letters<br />

84, 721 (2000)<br />

[13] Z.G. Yu, D.L. Smith, A. Saxena, R.L. Mart<strong>in</strong>, and A.R. Bishop, Molecular Geometry<br />

and field-dependent mobility <strong>in</strong> conjugated polymers, Physical Review B 63, 085202<br />

(2001)<br />

[14] P.E. Parris, V.M. Kenkre and D.H. Dunlap, Nature of <strong>Charge</strong> Carriers <strong>in</strong> Disordered<br />

Molecular Solids: Are Polarons Compatible with Observations?, Physical Review Let-<br />

ters 87, 126601 (2001)<br />

[15] R.A. Marcus and N. Sut<strong>in</strong>, Electron Transfer <strong>in</strong> chemistry and biology, Biochimica<br />

et biophysica Acta, (811), 265 (1985)<br />

[16] B.S. Bruceschwig, S. Ehrenson and N. Sut<strong>in</strong>, Solvent reorganisation <strong>in</strong> Optical and<br />

Thermal Electron-Transfer Processes, Journal of Physical Chemistry, 90, 3657 (1986)<br />

[17] P.F. Barbara, T.J. Mayer and M.A. Ratner, Contemporary Issues <strong>in</strong> Electron Trans-<br />

fer, Journal of Physical Chemistry, 100, 13148 (1996)<br />

[18] K. Sakanoue, M. Motoda, M. Sugimoto and S. Sakaki, A Molecular Orbital Study<br />

on the Hole <strong>Transport</strong> Property of Organic Am<strong>in</strong>e Compounds, Journal of Chemical<br />

Physics A, 103, 551 (1999)<br />

[19] N.E. Gruhn, D.A. da Silvha Filho, T.G. Bill, M. Malagoli, V. Coropceanu, A. Kahn<br />

and J.L Brédas, The Vibrational Reorganisation Energy <strong>in</strong> Pentacene: Molecular In-<br />

fluence on <strong>Charge</strong> <strong>Transport</strong>, Journal of the American Chemical Society, 124, 7918<br />

(2001)<br />

48


[20] T. Kato and T. Yamabe, Electron-phonon Coupl<strong>in</strong>g <strong>in</strong>teractions <strong>in</strong> charged cubic<br />

fluorocarbon cluster (CF)8, Journal of Chemical Physics, 120, 1006 (2004)<br />

[21] I.V. Leontyev, M.V. Basilevski and M.D. Newton, Theory and computation of elec-<br />

tron transfer reorganisation energies with cont<strong>in</strong>uum and molecular solvent models.,<br />

Theretical Chemistry Accounts, 111, 110 (2004)<br />

[22] A. Szabo and N.S. Ostlund, Modern Quantum Chemistry, Dover Publications, 1989<br />

[23] M.N. Paddon-Row and S.W. Wong, A Correlation between the Rates of Pho-<br />

to<strong>in</strong>duced Long-Range Intramolecular Electron Transfer <strong>in</strong> Rigidly L<strong>in</strong>ked Donor-<br />

Acceptor Systems and Computed π and π∗ Splitt<strong>in</strong>g Energies <strong>in</strong> Structurally Related<br />

Dienes, Chemical Physics Letters 167, 432 (1990)<br />

[24] C. Lian and M.D. Newton, Ab-<strong>in</strong>itio Studies of Electron Transfer: Pathway Analysis<br />

of Effective Transfer Integral, Journal of Physical Chemistry 96, 2855 (1992)<br />

[25] J.L. Brédas, J.P. Calbert, D.A. da Silva Filho and J. Cornil, Organic Semiconductors:<br />

A Theoretical Characterisation of the Basic <strong>Parameters</strong> Govern<strong>in</strong>g <strong>Charge</strong> <strong>Transport</strong>,<br />

Proceed<strong>in</strong>s of the National Accademy of Science 99, 5804 (2002)<br />

[26] P.G. da Costa, R.G. Dandrea and E.M. Conwell, First-pr<strong>in</strong>ciples calculation of the<br />

three-dimensional band structure of poly(phenylene v<strong>in</strong>ylene), Physycal Review B, 47,<br />

1800, (1993)<br />

[27] J. Ridley and M. Zerner, An Intermediate Neglect of Differential Overlap Technique<br />

for Spectroscopy: Pyrrole and the Az<strong>in</strong>es, Theoretica Chimica Acta, 32, 111 (1973)<br />

[28] K. Senthilkumar, F.C. Groezema, F.M. Bickelhaupt and L.D.A. Siebbeles, <strong>Charge</strong><br />

transport <strong>in</strong> columnar stacked triphenylenes: Effects of conformational fluctuations<br />

on charge transfer <strong>in</strong>tegrals and site energies, Journal of Chemical Physics, 32, 9809<br />

(2003)<br />

[29] A.J. Stone, http://www-teach.ch.cam.ac.uk/teach/M11/handout.pdf, <strong>Intermolecular</strong><br />

Forces, University of Cambridge (2004)<br />

49


[30] M.F. Guest, I.J. Bush, H.J.J. van Dam, P. Sherwood, J.M.H. Thomas, J.H. van<br />

Lenthe, R.W.A Havenith and J. Kendrick, The GAMESS-UK electronic structure pack-<br />

age: algorithms, developments and applications, Molecular Physics, 103,719 (2005)<br />

[31] Gaussian 03, Revision C.02, M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E. Scuse-<br />

ria, M. A. Robb, J. R. Cheeseman, J. A. Montgomery, Jr., T. Vreven, K. N. Kud<strong>in</strong>, J. C.<br />

Burant, J. M. Millam, S. S. Iyengar, J. Tomasi, V. Barone, B. Mennucci, M. Cossi, G.<br />

Scalmani, N. Rega, G. A. Petersson, H. Nakatsuji, M. Hada, M. Ehara, K. Toyota, R.<br />

Fukuda, J. Hasegawa, M. Ishida, T. Nakajima, Y. Honda, O. Kitao, H. Nakai, M. Klene,<br />

X. Li, J. E. Knox, H. P. Hratchian, J. B. Cross, V. Bakken, C. Adamo, J. Jaramillo, R.<br />

Gomperts, R. E. Stratmann, O. Yazyev, A. J. Aust<strong>in</strong>, R. Cammi, C. Pomelli, J. W.<br />

Ochterski, P. Y. Ayala, K. Morokuma, G. A. Voth, P. Salvador, J. J. Dannenberg, V.<br />

G. Zakrzewski, S. Dapprich, A. D. Daniels, M. C. Stra<strong>in</strong>, O. Farkas, D. K. Malick,<br />

A. D. Rabuck, K. Raghavachari, J. B. Foresman, J. V. Ortiz, Q. Cui, A. G. Baboul, S.<br />

Clifford, J. Cioslowski, B. B. Stefanov, G. Liu, A. Liashenko, P. Piskorz, I. Komaromi,<br />

R. L. Mart<strong>in</strong>, D. J. Fox, T. Keith, M. A. Al-Laham, C. Y. Peng, A. Nanayakkara, M.<br />

Challacombe, P. M. W. Gill, B. Johnson, W. Chen, M. W. Wong, C. Gonzalez, and J.<br />

A. Pople, Gaussian, Inc., Wall<strong>in</strong>gford CT, 2004.<br />

[32] A.J. Stone, Distributed multipole analysis: Stability for large basis sets, Journal of<br />

chemical theory and computation, 1, 1128 (2005)<br />

[33] G.A. Kam<strong>in</strong>ski, H.A. Stern, B.J. Berne, R.A. Friesner, Y.X. Cao, R.B. Murphy, R.<br />

Zhou and T.A. Halgren, Development of a polarisable Force Field For Prote<strong>in</strong>s via Ab<br />

Initio Quantum Chemistry: First Generation Model and Gas Phase Tests, Journal of<br />

Computational Chemistry, 23, 1515 (2002)<br />

[34] E.V. Tsiper and Z.G. Soos, <strong>Charge</strong> redistribution and polarisation energy of organic<br />

molecular crystalst, Physics Review B, 64, 195124 (2001)<br />

[35] A.J. Stone, Distributed polarisabilities, Molecular Physics, 56, 1065 (1985)<br />

[36] I.H. Campbell and D.L. Smith, Physics of organic electronic devices, Chapter 4<br />

(2001)<br />

50


[37] H. Scher and E. Montroll, Anomalous transit-time dispersion <strong>in</strong> amorphous solids ,<br />

Physical Review B, 12, 2455 (1975)<br />

[38] I. Oppenheim, K.E. Shuler and G.H. Weiss, Stochastic processes <strong>in</strong> chemical reac-<br />

tions, MIT press (1977)<br />

[39] B. Derrida, Velocity and diffusion constant of a periodic one-dimensional hopp<strong>in</strong>g<br />

model , Journal of Statistical Physics, 31, 433 (1982);<br />

[40] J. Nelson, J. Kirkpatrick and P. Ravirajan, Factors limit<strong>in</strong>g the efficiency of molecu-<br />

lar photovoltaic devices, Physical Review B, 69, 035337 (2004)<br />

[41] L. Pautmeier , R. Richert and H. Bässler , Poole-Frenkel behaviour of <strong>Charge</strong> Trans-<br />

port <strong>in</strong> organic solids with off-diagonal disorder studied by Monte Carlo simulation,<br />

Synthetic Metals, 37, 271 (1990)<br />

[42] P.M. Borsenberger, L. Pautmeier and H. Bässler, <strong>Charge</strong> <strong>Transport</strong> <strong>in</strong> disordered<br />

solids, Journal of Chemical Physics, 94, 5447 (1991)<br />

[43] R.A. Marcus , Electron Transfer Reactions <strong>in</strong> chemistry. Theory and experiment,<br />

Reviews of Modern Physics 65, 599 (1993)<br />

[44] A. Dieckamnn, H. Bässler and P.M. Borsenberger, An Assessment of the role of<br />

dipoles on the density of states of function of disordered molecular solics, Journal of<br />

Chemical Physics, 99, 8136 (1993)<br />

[45] S.V. Novikov and A.V. Vannikov, Cluster Structure and Distribution of the Elec-<br />

trostatic <strong>in</strong> a Lattice of Randomly oriented dipoles, Journal of Physical Chemistry 99,<br />

14573 (1995)<br />

[46] P.M. Borsenberger, D. Weiss, Organic Photoreceptors for Xerography, Marcel<br />

Dekker <strong>in</strong>c., 1998<br />

[47] Y.N. Garste<strong>in</strong>, E.M. Conwell, High Field Hopp<strong>in</strong>g mobility <strong>in</strong> molecular systems<br />

with spatially correlated energetic disorder, Chemical Physics Letters 245, 351 (1995)<br />

[48] S.V. Novikov, <strong>Charge</strong>-Carrier <strong>Transport</strong> <strong>in</strong> Disordered Polymers, Journal of Poly-<br />

mer Science: Part B: Polymer Physics 41, 2584 (2003)<br />

51


[49] T. Kreouzis, D. Poplavskyy, S.M. Tuladhar, M. Campoy-Quiles, J. Nelson, A.J.<br />

Campbell and D.D.C. Bradley, Temperature and field dependence of hole mobility <strong>in</strong><br />

poly(9,9-dioctylfluorene), Physical Review B 73, 235201 (2006)<br />

52


Chapter 3<br />

Development and Verification of<br />

Methods for <strong>Calculat<strong>in</strong>g</strong> <strong>Charge</strong><br />

<strong>Transport</strong> <strong>Parameters</strong><br />

In chapter 2 we have shown how hopp<strong>in</strong>g transport <strong>in</strong> disordered molecular assemblies<br />

can be described <strong>in</strong> terms of a number of fundamental electronic parameters of molecules.<br />

We <strong>in</strong>troduced the parameters that determ<strong>in</strong>e the charge transfer rate between two molecules:<br />

the transfer <strong>in</strong>tegral J, the site energy difference ∆E and the reorganisation energy λ. Of<br />

these three parameters, J and ∆E will <strong>in</strong> general have different values for each pair of<br />

molecules <strong>in</strong> a slab of material, depend<strong>in</strong>g on the relative position and conformation of<br />

molecules. In order to calculate charge transfer rates for all such pairs of molecules, J<br />

and ∆E have to be calculated many times, and therefore fast and efficient methods are<br />

desirable. This chapter presents the methods that we developed <strong>in</strong> this thesis for this<br />

purpose.<br />

In section 2.3 we described a method for calculat<strong>in</strong>g the transfer <strong>in</strong>tegral J from the<br />

splitt<strong>in</strong>g <strong>in</strong> orbital energies that has been widely used <strong>in</strong> the literature. We also <strong>in</strong>tro-<br />

duced the theoretical framework that can be used to describe the electrostatic and <strong>in</strong>-<br />

ductive <strong>in</strong>teractions between two charge distributions. In this chapter we expla<strong>in</strong> how<br />

to express the Fock matrix for a pair of molecules <strong>in</strong> a basis set of orbitals localised on<br />

each molecule. This allows the transfer <strong>in</strong>tegrals and energies of the localised orbitals to<br />

be obta<strong>in</strong>ed directly. We then present two methodologies for determ<strong>in</strong><strong>in</strong>g J and ∆E for<br />

53


pairs of molecules <strong>in</strong> which SCF calculations have to be carried out only on the isolated<br />

molecules. The approximate method to calculate J is based on a simplified version of<br />

the ZINDO [1] Hamiltonian and will <strong>in</strong>volve calculation of the Molecular Orbital Over-<br />

lap, this method will be referred to as MOO. The method to calculate ∆E is based on the<br />

classical tools for calculat<strong>in</strong>g <strong>in</strong>teraction energies between charge distributions that were<br />

<strong>in</strong>troduced <strong>in</strong> section 2.4. The results of these classical calculations will be compared to<br />

the difference <strong>in</strong> site energies obta<strong>in</strong>ed by the projective method.<br />

The last section of the chapter is devoted to compar<strong>in</strong>g the results of the three methods<br />

available for calculat<strong>in</strong>g J: the projective and orbital splitt<strong>in</strong>g methods are shown to be<br />

<strong>in</strong> agreement <strong>in</strong> the cases where site energies are the same. The projective method is<br />

used to compare results from the DFT Kohn-Sham and the ZINDO Fock matrix, to show<br />

the ZINDO method gives very similar values for the transfer <strong>in</strong>tegral to the pure DFT<br />

calculations. The projective method with the ZINDO Fock matrix is then used as the<br />

reference to evaluate the accuracy of the approximate MOO method. The systems used for<br />

this comparison are ethylene, pentacene and hexabenzocoronene. Ethylene and pentacene<br />

are chosen as typical conjugated molecules; hexabenzocoronene is selected because of<br />

its degenerate orbitals and because it is used as a model system for the study of charge<br />

transport <strong>in</strong> chapter seven.<br />

The notation and description of methods used <strong>in</strong> this chapter applies to charge transfer<br />

between identical molecules, as they will be used on identical molecules <strong>in</strong> the rest of the<br />

thesis. It would be straightforward to generalise these methods to be applicable to different<br />

molecules.<br />

3.1 Projective Method for Calculation of Transfer Inte-<br />

grals<br />

The orbital splitt<strong>in</strong>g method, where the transfer <strong>in</strong>tegral is approximated by half the split-<br />

t<strong>in</strong>g of the HOMO energies of a pair of molecules, is only valid if EA and EB, the site<br />

energies of equation 2.19, are the same. In order to f<strong>in</strong>d the general solution of equa-<br />

tion 2.16 K. Senthilkumar and co-workers [2] have exploited a feature of the Amsterdam<br />

54


Density Functional [3] package that allows the orbitals of molecular fragments of a larger<br />

system to be used as the basis set <strong>in</strong> which to solve the Kohn Sham equations. In order to<br />

be able to perform fragment type calculations us<strong>in</strong>g the ZINDO method as well as DFT<br />

functionals, we developed a procedure whereby the orbitals of a pair of molecules are pro-<br />

jected onto a basis set def<strong>in</strong>ed by the orbitals of each <strong>in</strong>dividual molecule. The molecular<br />

orbital energies of the pair of molecules are then used to rewrite the Fock matrix <strong>in</strong> the new<br />

localised basis set. We call this method the projective method. This method is based on<br />

application of the spectral theorem [4]. In the case of model chemistries such as Hartree<br />

Fock or Density Functional Theory, where the atomic orbitals do not form an orthogonal<br />

set, symmetric Lowd<strong>in</strong> renormalisation [5] is used to re-orthogonalise the orbitals. Com-<br />

pared to the orbital splitt<strong>in</strong>g method, the projective method has several advantages. First<br />

it is not necessary to assume that the two energies of the HOMOs are equal and therefore<br />

the projective method allows us to look at pairs of different molecules, or pairs of identical<br />

molecules <strong>in</strong> such arrangments as to lead to differences <strong>in</strong> EA and EB due to polarisation.<br />

Another advantage is that we are no longer limited <strong>in</strong> the type of orbitals we use to def<strong>in</strong>e<br />

the isolated molecule’s Slater determ<strong>in</strong>ant: we could, for example, use the orbitals from<br />

a restricted open shell calculation, or the natural orbitals from an unrestricted open shell<br />

calculation, or from a multi-reference method. The third advantage is that it is possible,<br />

by look<strong>in</strong>g at the diagonal elements of the Fock matrix <strong>in</strong> the localised basis set, to deter-<br />

m<strong>in</strong>e the expectation values of orbitals of the isolated molecules; if the orbitals chosen are<br />

the HOMO (or the LUMO) of each molecule this allows us to determ<strong>in</strong>e the site energies<br />

for the positive (or negative) charge localised states.<br />

Let us show how this projection is carried out. The orbitals of the pair of molecules<br />

will be written as a matrix Cpair, where each column corresponds to a molecular orbital<br />

and each row to an atomic orbital. The vector space that we wish to project these orbitals<br />

onto will be written <strong>in</strong> terms of the molecular orbitals of the isolated molecules. We write<br />

these orbitals as Cloc where the first N/2 columns correspond to the molecular orbitals<br />

of the first molecule and the last N/2 columns correspond to the molecular orbitals on<br />

the second molecule, similarly the first N/2 rows correspond to the atomic orbitals on the<br />

first molecule and the second N/2 to those on the second molecule. The localised basis<br />

55


set Cloc can be written as:<br />

Cloc =<br />

φ 1<br />

1<br />

φ 1<br />

2<br />

φ 2<br />

1 ... φ N 2<br />

1 0 0 0 0<br />

φ 2<br />

2 ... φ N 2<br />

2 0 0 0 0<br />

... ... ... ... 0 0 0 0<br />

φ 1 N 2<br />

φ2 N<br />

2<br />

... φ N 2<br />

N<br />

2<br />

0 0 0 0<br />

0 0 0 0 φ N 2 +1<br />

N<br />

2 +1 φ N 2 +2<br />

N<br />

2 +1<br />

... φN N<br />

2 +1<br />

0 0 0 0 φ N 2 +1<br />

N 2 +2 φ N 2 +2<br />

N 2 +2<br />

... φ N N 2 +2<br />

0 0 0 0 ... ... ... ...<br />

0 0 0 0 φ N 2 +1<br />

N<br />

φ N 2 +2<br />

N ... φN N<br />

where φ µ ν labels the contribution of molecular orbitals µ from atomic orbital (AO) ν.<br />

(3.1)<br />

If the AOs are orthogonal, as they are <strong>in</strong> the ZINDO method, the orbitals of the pair of<br />

molecules are written <strong>in</strong> the new basis set simply by calculat<strong>in</strong>g the dot products between<br />

the localised orbitals and the orbitals of the pair of molecules. In matrix form this is<br />

written as:<br />

where C loc<br />

pair<br />

C loc<br />

pair = CTlocCpair<br />

(3.2)<br />

represents the orbitals of the pair of molecules <strong>in</strong> the localised basis set of<br />

the isolated molecule orbitals, and the superscript T represents transposition. Once an<br />

orthogonal set of orbitals for the pair of molecules is found, we can rewrite the Fock<br />

matrix for the pair of molecules, Floc pair , as:<br />

F loc<br />

pair<br />

T loc<br />

= Cloc pair ɛpairCpair (3.3)<br />

where the molecular orbital energies of the pair of molecules have been written as a diag-<br />

onal matrix ɛpair.<br />

The off-diagonal element µ, ν of this Fock matrix will represent the transfer <strong>in</strong>tegral<br />

of the µ th and ν th isolated molecule molecular orbitals. The µ, µ diagonal elements will be<br />

equal to the expectation value of the µth molecular orbital of the isolated molecule with<br />

the Fock operator of the pair of molecules: if µ labels the HOMO of molecule A, this will<br />

correspond to the quantity EA from equation 2.19.<br />

56


If the molecular orbitals are expressed <strong>in</strong> a non-orthogonal basis set they can be or-<br />

thogonalised by us<strong>in</strong>g Lowd<strong>in</strong> orthogonalisation and def<strong>in</strong><strong>in</strong>g the matrix D :<br />

S = D T D (3.4)<br />

where S is the overlap matrix, the µ th , ν th element S µ,ν of the overlap matrix corresponds<br />

to the <strong>in</strong>tegral over all space of the product of the µ th and ν th atomic orbitals. The non-<br />

orthogonal orbitals C ′ loc are then orthogonalised by writ<strong>in</strong>g:<br />

Cloc = DC ′ loc<br />

(3.5)<br />

To see why this is the case, consider the def<strong>in</strong>ition of the orthonormality condition for<br />

the canonical molecular orbitals C ′ loc <strong>in</strong> a non-orthogonal basis set with overlap matrix S:<br />

C ′ T<br />

loc SC′ loc = 1 (3.6a)<br />

C ′ T<br />

loc DT DC ′ loc = 1 (3.6b)<br />

C T<br />

loc Cloc = 1 (3.6c)<br />

show<strong>in</strong>g that our new orbitals are orthonormal. In the case of DFT calculation the or-<br />

thogonalisation procedure is carried out on all isolated molecule and system orbitals. We<br />

can then apply equations 3.2 and 3.3 to the orthogonalised orbitals and obta<strong>in</strong> transfer<br />

<strong>in</strong>tegrals and site energies.<br />

3.2 Molecular Orbital Overlap Method for Calculation<br />

of Transfer Integrals<br />

The fast method we use to calculate transfer <strong>in</strong>tegrals is based on calculat<strong>in</strong>g the ZINDO<br />

Fock matrix of a pair of molecules directly. We use an SCF ZINDO calculation to ob-<br />

ta<strong>in</strong> the isolated molecule’s molecular orbitals and wrote custom code to calculate atomic<br />

overlaps and derive an approximate value for the transfer <strong>in</strong>tegral. Because this method is<br />

based on calculat<strong>in</strong>g Fock matrix elements from atomic overlap, we dub this method and<br />

57


the program we devised to carry it out molecular orbital overlap (MOO). The approxima-<br />

tions made <strong>in</strong> MOO are discussed <strong>in</strong> this section.<br />

We use the same nam<strong>in</strong>g convention for atomic orbitals which we have used <strong>in</strong> writ<strong>in</strong>g<br />

equation 3.1. In this convention molecular orbitals localised on molecule A will have<br />

zeroes as the first N/2 elements, whilst orbitals localised on molecule B will have zeroes<br />

as their last N/2 elements. We are go<strong>in</strong>g to calculate transfer <strong>in</strong>tegrals between orbitals<br />

localised on either molecule, therefore the only elements of the Fock matrix we need to<br />

calculate are the off-diagonal blocks with one <strong>in</strong>dex runn<strong>in</strong>g from 0 to N/2 and the other<br />

runn<strong>in</strong>g from N/2 to N. This is because the diagonal blocks of the Fock matrix will be<br />

cancelled out by the rows of zeros present on either localised molecular orbitals. The<br />

elements of the Fock matrix we need to calculate are those for atomic orbitals on different<br />

atoms, therefore the form of these elements <strong>in</strong> the INDO approximation is [1]:<br />

(βa<br />

Fµν = ¯S<br />

+ βb) γab<br />

µν + Pµν<br />

2 2<br />

(3.7)<br />

where ¯S represents the overlap matrix of atomic orbitals (with σ and π overlap between<br />

p orbitals weighted differently, see below), a and b label the two atomic centres that the<br />

µ and ν atomic orbitals are centred on, βa is the ionisation potential of atom a, Pµν labels<br />

the density matrix and γ is the Mataga-Nashimoto potential. The approximation we make<br />

<strong>in</strong> this calculation is to assume that Pµν is block diagonal, <strong>in</strong> the sense that all elements<br />

enter<strong>in</strong>g equation 3.7 are equal to zero; the density matrix and Mataga-Nashimoto poten-<br />

tial therefore do not contribute to the elements of the Fock matrix we are <strong>in</strong>terested <strong>in</strong><br />

calculat<strong>in</strong>g. This assumption for the orbitals will hold either if the orbitals of the pair are<br />

all localised or if each pair of orbitals can be written as a constructive/destructive combi-<br />

nation of a correspond<strong>in</strong>g pair of orbitals <strong>in</strong> the isolated molecule. To see why the latter is<br />

the case, consider two particular doubly occupied orbitals ψ µ and ψ µ+1 which are formed<br />

from bond<strong>in</strong>g and anti-bond<strong>in</strong>g comb<strong>in</strong>ations of two of the isolated molecule’s occupied<br />

orbitals ψ Aν and ψ Bν . The contribution δP of these two orbitals to the density matrix will<br />

be of the form:<br />

δP = 2(ψ µ T ψ µ + ψ µ+1 T ψ µ+1 ) (3.8a)<br />

58


= 2((ψ Aν + ψ Bν ) T (ψ Aν + ψ Bν ) + (ψ Aν − ψ Bν ) T (ψ Aν − ψ Bν )) (3.8b)<br />

= 2(2ψ Aν T ψ Aν + 2ψ Bν T ψ Bν ) (3.8c)<br />

Because all isolated molecule orbitals ψ Aν and ψ Bν are localised on one molecule only, this<br />

contribution will be block diagonal. Furthermore, s<strong>in</strong>ce all contributions to the density<br />

matrix will be of this form, the density matrix will be overall block diagonal. Assum<strong>in</strong>g<br />

that the density matrix is block diagonal can be thought of as assum<strong>in</strong>g that both the<br />

bond<strong>in</strong>g and antibond<strong>in</strong>g comb<strong>in</strong>ation of molecular orbitals are doubly occupied, that is<br />

assum<strong>in</strong>g that the two molecules are not covalently bonded to each other.<br />

The task of determ<strong>in</strong><strong>in</strong>g values for the Fock matrix has therefore been reduced to<br />

the comparatively simple task of determ<strong>in</strong><strong>in</strong>g ¯S , the weighted atomic orbital overlap.<br />

Atomic orbital overlaps can be determ<strong>in</strong>ed analytically us<strong>in</strong>g the expressions derived by<br />

Mulliken [6] and the π and σ components of the p-orbital overlaps must be weighed by<br />

the appropriate proportionality factors, <strong>in</strong> accordance with the scheme devised by Zerner<br />

and coworkers [1].<br />

The great advantage of this approach is that it requires only one SCF calculation to<br />

be carried out to obta<strong>in</strong> the orbitals of the isolated molecule. All other quantities can be<br />

derived simply from the relative geometry of the two molecules. Additionally, s<strong>in</strong>ce the<br />

heaviest computational task is comput<strong>in</strong>g the atomic overlaps, the s, σ and π overlaps can<br />

be tabulated as a function of distance for the most common atom types encountered <strong>in</strong><br />

organic molecules, carbon and hydrogen, allow<strong>in</strong>g a significant <strong>in</strong>crease <strong>in</strong> computational<br />

speed for a modest <strong>in</strong>crease <strong>in</strong> memory usage.<br />

3.3 Calculation of Site Energy Difference from Electro-<br />

static Interactions<br />

In this section we show how to determ<strong>in</strong>e the contribution to ∆E from classical electric<br />

<strong>in</strong>teractions only and compare the results with those from ab-<strong>in</strong>itio calculation. The elec-<br />

trostatic contribution will be determ<strong>in</strong>ed from the electrical moments of two molecules<br />

<strong>in</strong> their charged and neutral states. Assum<strong>in</strong>g that the two molecules are identical, the<br />

59


procedure to determ<strong>in</strong>e the electrostatic contribution to the site energy difference will be<br />

as follows:<br />

(a) distributed multipole analysis (DMA) is carried out for a charged and a neutral<br />

molecule<br />

(b) the two sets of distributed multipoles are rotated to match the orientations of molecule<br />

A and molecule B<br />

(c) equation 2.32 is used to calculate the <strong>in</strong>teraction energies UA0B+ between the mul-<br />

tipoles calculated for the neutral molecule <strong>in</strong> the orientation of molecule A and the<br />

charged radical <strong>in</strong> the orientation of molecule B, and UA+B0 between the multipoles<br />

calculated for the charged radical <strong>in</strong> the orientation A and the neutral molecule <strong>in</strong><br />

the orientation of B<br />

(d) these two <strong>in</strong>teraction energies are subtracted to obta<strong>in</strong> ∆E<br />

If the calculation of the charged radical is performed us<strong>in</strong>g the orbitals of the neutral<br />

molecule as a guess without optimis<strong>in</strong>g the orbitals, this corresponds to the frozen orbital<br />

approximation and can be compared to the results from the projective method.<br />

The only aspect we have not already discussed <strong>in</strong> this scheme is the rotation of elec-<br />

trical multipoles. <strong>Charge</strong>s are clearly <strong>in</strong>dependent of coord<strong>in</strong>ate transformations, dipoles<br />

transform as vectors and the elements of the quadrupole moment Qαβ will transform as<br />

rαrβ, that is Q behaves as a rank two tensor:<br />

Q ′ αβ =<br />

� �<br />

γ<br />

δ<br />

RγαRδβQγδ<br />

(3.9)<br />

The multipole expansion is truncated so that the highest moments <strong>in</strong>cluded are quadrupoles<br />

because <strong>in</strong> conjugated materials the π system creates two layers of negative charge sand-<br />

wich<strong>in</strong>g the positive nuclei. This leads to a large quadrupole moment which we expect<br />

to be important <strong>in</strong> <strong>in</strong>teractions between π systems; for example, it is known that simple<br />

models of π stack<strong>in</strong>g can be made which treat molecules as a collection of quadrupoles<br />

[7].<br />

60


In the follow<strong>in</strong>g sections we carry out this procedure for two molecules: ethylene<br />

and pentacene. Ethylene is chosen because it is small enough to permit substitut<strong>in</strong>g each<br />

molecule with a s<strong>in</strong>gle set of multipoles even at relatively small <strong>in</strong>termolecular distances.<br />

The projective method will be used on systems with small <strong>in</strong>termolecular separations,<br />

because under these conditions differences <strong>in</strong> site energy are larger than the error <strong>in</strong> the<br />

SCF calculations. In the first subsection, we perform DFT calculations on neutral pairs<br />

of ethylene molecules and show that electrostatics can be used to calculate the difference<br />

<strong>in</strong> site energies between the localised HOMOs on the two molecules. We also show<br />

that the small discrepancy between the density functional and electrostatic results can<br />

be accounted for by <strong>in</strong>clud<strong>in</strong>g the <strong>in</strong>duction energy. In the second part we address the<br />

question of how the presence of charge would modify the <strong>in</strong>duced dipoles and ∆E <strong>in</strong><br />

pairs of ethylene molecules. The nature of the site energy difference and the role of<br />

asymmetry <strong>in</strong> the properties that determ<strong>in</strong>e it will also be discussed. In the third part<br />

we will present a comparison of ∆E calculated from projection of MOs calculated from<br />

pure DFT results with ∆E calculated from the electrostatic <strong>in</strong>teraction between pairs of<br />

pentacene molecules. In this case the size of the molecules is large compared to their<br />

separation and therefore we must use distributed multipole analysis (DMA) to calculate<br />

the electrostatic <strong>in</strong>teraction, as discussed <strong>in</strong> section 2.4.1.<br />

3.3.1 Ethylene: Neutral Systems<br />

In this section we calculate site energy difference for the HOMO of neutral pairs of ethy-<br />

lene molecules. The aim is to show that both projection of DFT MOs and classical calcu-<br />

lations of electric <strong>in</strong>teractions lead to similar values of the site energy difference. Ethylene<br />

is chosen as an example of a molecule small enough that it can be substituted entirely by<br />

a s<strong>in</strong>gle set of electrical moments. This has the advantage that the multipole moments can<br />

be obta<strong>in</strong>ed exactly from a SCF calculation on an isolated molecule, without the need for<br />

the somewhat arbitrary partition<strong>in</strong>g schemes necessary to perform DMA. A represention<br />

of the pair of ethylene molecules which we consider is shown <strong>in</strong> figure 3.1; this figure<br />

also shows the x axis, around which one of the molecules will be rotated to generate the<br />

geometries <strong>in</strong> which we will calculate ∆E. The two molecules are at centre of mass sep-<br />

61


Figure 3.1: A pair of ethylene molecules at a separation of 5 Å . Also shown is the axis<br />

around which one of the two molecules will be rotated to generate the geometries<br />

for which ∆E is calculated.<br />

aration of 5 Å . This large separation justifies the use of a s<strong>in</strong>gle electrical moment to<br />

describe each molecule: the neutral molecule will be described by a po<strong>in</strong>t quadrupole, the<br />

charged one by a po<strong>in</strong>t charge. In the rema<strong>in</strong>der of this chapter we refer to po<strong>in</strong>t charges<br />

as monopoles to keep the term<strong>in</strong>ology consistent.<br />

The density functional used to determ<strong>in</strong>e the value of the quadrupole moment of ethy-<br />

lene of was PW91, the basis set employed was a triple zeta basis set with extra polari-<br />

sation basis sets (TVZP) and all quantum chemical calculations were performed <strong>in</strong> the<br />

GAMESS-UK [8] suite. In atomic units (units of ea 2<br />

0 , where a0 is the Bohr radius) the<br />

non-zero values of the quadrupole moment are:<br />

Qxx = 1.51 (3.10)<br />

Qyy = 1.34<br />

Qzz = −2.85<br />

�<br />

|Q| = Q2 xx + Q2 yy + Q2 zz = 3.49<br />

where the x axis is def<strong>in</strong>ed along the C=C double bond and the z axis is perpendicu-<br />

lar to the plane of the molecule. Consideration of these three numbers shows that the<br />

quadrupole of the molecule is essentially a l<strong>in</strong>ear quadrupole aligned <strong>in</strong> the z direction.<br />

62


site energy difference / eV<br />

0.2<br />

0.15<br />

0.1<br />

0.05<br />

PW91 ∆E<br />

Quadrupole-Monopole ∆E<br />

0<br />

0 50 100<br />

rotation angle / degrees<br />

150 200<br />

Figure 3.2: Site energy difference for a pair of neutral ethylene molecules as a function of<br />

rotation of one of the molecule about the the C=C bond calculated by projection<br />

of PW91 MOs (circles) and calculated from the difference <strong>in</strong> quadrupolemonopole<br />

<strong>in</strong>teractions calculated classically from the electrical moments calculated<br />

at the PW91 level on an isolated ethylene molecule (solid l<strong>in</strong>e).<br />

A l<strong>in</strong>ear quadrupole corresponds to two dipole moments emanat<strong>in</strong>g from the same po<strong>in</strong>t,<br />

but po<strong>in</strong>t<strong>in</strong>g <strong>in</strong> opposite directions. We can therefore th<strong>in</strong>k of ethylene as hav<strong>in</strong>g a partial<br />

positive charge <strong>in</strong> the plane of the atoms, sandwiched between two negative charge clouds<br />

caused by the π electrons.<br />

In figure 3.2 we compare the site energy difference computed us<strong>in</strong>g the projective<br />

method applied to MOs from a PW91/TVZP DFT calculation with that obta<strong>in</strong>ed from a<br />

classical calculation based on treat<strong>in</strong>g the neutral molecule as po<strong>in</strong>t quadrupole and the<br />

charged molecule as a po<strong>in</strong>t charge. It is immediately clear that the two distributions<br />

follow the same form and are <strong>in</strong>deed rather close <strong>in</strong> value: the greatest disagreement is<br />

roughly 10%. Also we notice that, hav<strong>in</strong>g excluded the <strong>in</strong>ductive term, we have overes-<br />

timated the value of the site energy difference. Let us compare this last po<strong>in</strong>t to a recent<br />

study by E.M. Valeev and coworkers [9]. In that study, calculations of ∆E are carried out<br />

for pairs of ethylene molecules and for an isolated ethylene molecule <strong>in</strong> the presence of<br />

63


po<strong>in</strong>t charges on each atom chosen to fit the electrostatic potential of the molecule as best<br />

as possible. The aim was to establish whether ∆E is electrostatic <strong>in</strong> orig<strong>in</strong> and can there-<br />

fore be reproduced by us<strong>in</strong>g po<strong>in</strong>t charges. The fact that their calculation underestimates<br />

∆E, can be expla<strong>in</strong>ed by the neglect of the most important electrical moment of the charge<br />

distribution of ethylene: the zz element of the quadrupole moment. In fact, it is impos-<br />

sible for 6 po<strong>in</strong>t charges <strong>in</strong> the xy plane to replicate the zz component of the quadrupole<br />

moment, which is due to sheets of negative charge be<strong>in</strong>g present above and below of the<br />

xy plane.<br />

In order to establish whether <strong>in</strong>ductive <strong>in</strong>teractions are responsible for the discrep-<br />

ancy <strong>in</strong> the two curves <strong>in</strong> figure 3.2, we now compute the polarisability of an ethylene<br />

molecule and use it to compute the <strong>in</strong>duced dipole on each molecule. We recalculate ∆E<br />

tak<strong>in</strong>g these <strong>in</strong>duced dipoles <strong>in</strong>to account. PW91 calculations are run on neutral systems,<br />

therefore we must calculate the dipoles <strong>in</strong>duced by the electrostatic moments of a neutral<br />

molecule. For this reason, we calculate the quadrupole <strong>in</strong>duced dipoles, then repeat the<br />

procedure iteratively until the <strong>in</strong>duced dipoles converge. The convergence criterion we<br />

used was when the root mean square difference of the n th and (n + 1) th iteration of the<br />

calculated <strong>in</strong>duced dipole should be smaller than 10 − 10ea0. As mentioned <strong>in</strong> subsection<br />

2.4.2, convergence should be very rapid. In fact, less than 10 cycles were always suffi-<br />

cient. In atomic units (units of a3 0 ) the non-zero elements of the polarisability calculated<br />

at the PW91 level are:<br />

Pxx = 34.80 (3.11)<br />

Pyy = 21.82<br />

Pzz = 13.93<br />

Us<strong>in</strong>g these values for the polarisability, we were able to f<strong>in</strong>d the correction to the<br />

quadrupole monopole <strong>in</strong>teraction energy. Compar<strong>in</strong>g the results thus obta<strong>in</strong>ed with those<br />

obta<strong>in</strong>ed by the projective method yields figure 3.3: the two curves are now much closer.<br />

The ma<strong>in</strong> purpose of this exercise was to show that us<strong>in</strong>g simple classical methods,<br />

we can obta<strong>in</strong> results which are very close to those obta<strong>in</strong>ed from a SCF calculation on<br />

64


site energy difference / eV<br />

0.15<br />

0.1<br />

0.05<br />

Quadrupole+Induced Dipole<br />

PW91 ∆E<br />

0<br />

0 50 100<br />

rotation angle / degrees<br />

150<br />

Figure 3.3: Site energy difference for a pair of neutral ethylene molecules as a function of<br />

rotation of one of the molecules about the the C=C bond calculated by projection<br />

of PW91 MOs (circles) and calculated from the difference <strong>in</strong> quadrupolemonopole<br />

<strong>in</strong>teractions corrected by the quadrupole <strong>in</strong>duced dipoles (solid<br />

l<strong>in</strong>es). Polarisabilities and electrical moments were calculated at the PW91<br />

level.<br />

65


a pair of molecules, whilst perform<strong>in</strong>g a s<strong>in</strong>gle SCF calculation on an isolated molecule<br />

only. The next section will be devoted to mak<strong>in</strong>g the calculations somewhat more relevant<br />

to the case of charge transport by consider<strong>in</strong>g a charged pair of molecules.<br />

3.3.2 Ethylene: <strong>Charge</strong>d Systems<br />

In this section we <strong>in</strong>tend to calculate ∆E for charged pairs of molecules. It would be<br />

difficult and beyond the scope of our <strong>in</strong>vestigation to run DFT calculations on charged<br />

systems while ensur<strong>in</strong>g the localisation of a charge on one molecule only, therefore we<br />

will only calculate the site energy difference from electrostatic and <strong>in</strong>ductive <strong>in</strong>teractions.<br />

In particular, we ignore the change <strong>in</strong> quadrupole moment upon charg<strong>in</strong>g and only look at<br />

the change <strong>in</strong> <strong>in</strong>duced dipole <strong>in</strong> go<strong>in</strong>g from a neutral pair of molecules (where the dipole<br />

is <strong>in</strong>duced by a quadrupole) to a charged pair of molecule (where the dipole is <strong>in</strong>duced by<br />

a quadrupole and a monopole).<br />

We imag<strong>in</strong>e that two effects might come <strong>in</strong>to play upon charg<strong>in</strong>g a molecule <strong>in</strong> a pair:<br />

• charg<strong>in</strong>g a molecule will modify its electrical moments<br />

• a po<strong>in</strong>t charge will <strong>in</strong>duce a much larger dipole than a po<strong>in</strong>t quadrupole<br />

We ignore the first po<strong>in</strong>t, as this is expected to be a small effect. For ethylene the lead-<br />

<strong>in</strong>g term <strong>in</strong> the electrostatic <strong>in</strong>teraction is unchanged as we do not expect any <strong>in</strong>tramolec-<br />

ular charge transfer to occur upon charg<strong>in</strong>g; the second largest electrical moment after<br />

the monopole will still be the quadrupole. We also ignore the small change <strong>in</strong> quadrupole<br />

moment.<br />

Regard<strong>in</strong>g the second po<strong>in</strong>t, putt<strong>in</strong>g a monopole on one of the two molecules is ex-<br />

pected to have a strong <strong>in</strong>fluence on the value of the <strong>in</strong>duced dipole, and thereby on the<br />

contribution of the <strong>in</strong>duced dipole to ∆E.<br />

In order to test this hypothesis, we consider two ethylene molecules with their π sys-<br />

tems perpendicular and consider the separation dependance of the various <strong>in</strong>duced-dipole<br />

monopole <strong>in</strong>teractions. We focus on the po<strong>in</strong>t <strong>in</strong> figure 3.3 where ∆E is the greatest <strong>in</strong><br />

neutral molecules (θ = 90 ◦ ). We compare the <strong>in</strong>duced dipole monopole <strong>in</strong>teraction <strong>in</strong><br />

neutral and charged systems, calculated us<strong>in</strong>g classical arguments. In the neutral system,<br />

66


Energy / eV<br />

0.03<br />

0.02<br />

0.01<br />

0<br />

-0.01<br />

-0.02<br />

-0.03<br />

positive pair<br />

neutral pair<br />

4 6 8 10 12<br />

Separation / A<br />

Figure 3.4: Distance dependence of the contribution to ∆E from the <strong>in</strong>duced dipole <strong>in</strong> a<br />

pair of ethylene molecules as a function of separation. The solid l<strong>in</strong>e corresponds<br />

to the contribution from the dipole moment <strong>in</strong>duced by a quadrupole<br />

(i.e. <strong>in</strong> a neutral system), whereas the dotted l<strong>in</strong>e corresponds to that from the<br />

dipole moment <strong>in</strong>duced by a monopole (i.e. <strong>in</strong> a charged system).<br />

67


Figure 3.5: Diagram show<strong>in</strong>g the dipoles <strong>in</strong>duced <strong>in</strong> two ethylene molecules oriented at<br />

right angles to each other. (a) Case of po<strong>in</strong>t quadrupoles (neutral calculation).<br />

(b) Case of po<strong>in</strong>t charges (charge calculation). For discussion, please see the<br />

text.<br />

the <strong>in</strong>duced dipole is <strong>in</strong>duced by a quadrupole. In the charged system the <strong>in</strong>duced dipole<br />

is <strong>in</strong>duced both by a monopole and a quadrupole. Figure 3.4 shows that, contrary to<br />

our expectations, the contribution to ∆E of dipoles <strong>in</strong>duced <strong>in</strong> a neutral and <strong>in</strong> a charged<br />

system are not very different <strong>in</strong> magnitude. This is because it is not the magnitude of<br />

the molecules’ electric fields and their polarisabilities which determ<strong>in</strong>es the magnitude of<br />

∆E, but their asymmetry. The reason for this is rather subtle and is best expla<strong>in</strong>ed with<br />

the aid of a diagram, which is shown <strong>in</strong> figure 3.5. The sign and magnitude of the <strong>in</strong>duced<br />

dipole for perpendicular ethylene molecules is represented by the direction and width of<br />

arrows. Two cases are considered: a) shows the dipoles <strong>in</strong>duced by po<strong>in</strong>t quadrupoles<br />

(correspond<strong>in</strong>g to a neutral system) and b) shows the dipole <strong>in</strong>duced by po<strong>in</strong>t charges<br />

(correspond<strong>in</strong>g to a charged system).<br />

Let us consider case a). The quadrupole of ethylene is depicted as a l<strong>in</strong>ear quadrupole,<br />

which is represented by a circle with a ”+” surrounded by two circles with ”-”s. The shape<br />

to the left represents an ethylene molecule with the molecular plane ly<strong>in</strong>g on the xy plane<br />

and therefore its π electrons are <strong>in</strong> two sheets above and below the plane of the page; the<br />

68


ethylene molecule to the right is rotated by 90 ◦ and has its π electrons ly<strong>in</strong>g to the left and<br />

to the right of the figure. The molecule to the left has the larger dipole, represented by a<br />

wider arrow, for two reasons:<br />

• the field caused by the molecule to the right is larger<br />

• the polarisability Pyy is larger than Pzz<br />

The most important aspect about this part of the diagram is that both dipoles are ly<strong>in</strong>g<br />

<strong>in</strong> the same direction: therefore their contributions to the components UA0B+ and UA+B0<br />

have opposite signs and, when subtracted to obta<strong>in</strong> ∆E, would <strong>in</strong>crease the site energy<br />

difference by the sum of their magnitudes.<br />

Let us compare this case to that represented <strong>in</strong> b). In this case the two dipoles act <strong>in</strong><br />

opposite directions. Therefore ∆E would be <strong>in</strong>creased only by the difference <strong>in</strong> magni-<br />

tudes of the contributions from the dipole <strong>in</strong>teractions to UA0B+ and UA+B0. There will still<br />

be a net contribution to ∆E, because the two components of the polarisability are different<br />

for the two different orientations and hence the molecule to the left will have the larger<br />

<strong>in</strong>duced dipole.<br />

In order to demonstrate this effect, <strong>in</strong> figure 3.6 we plot the dipoles <strong>in</strong>duced on each<br />

molecule <strong>in</strong> a neutral and <strong>in</strong> a positive ethylene system as function of separation. This<br />

figure shows that the dipoles <strong>in</strong>duced by monopoles are <strong>in</strong>deed larger but act <strong>in</strong> opposite<br />

direction, whereas the dipoles <strong>in</strong>duced by quadrupoles are smaller but act <strong>in</strong> the same<br />

direction. In our calculation for the charged system, we <strong>in</strong>clude the <strong>in</strong>ductive effects of<br />

both the po<strong>in</strong>t charge and the po<strong>in</strong>t quadrupole and therefore we observe a sum of these<br />

two effects. Our explanation why the <strong>in</strong>duced dipole has a slightly smaller effect on ∆E<br />

<strong>in</strong> the charged system than <strong>in</strong> the neutral system is as follows:<br />

• the monopole <strong>in</strong>duced dipoles contribute to ∆E only because of the asymmetry <strong>in</strong><br />

the polarisability<br />

• the quadrupole <strong>in</strong>duced dipole is asymmetric because of the asymmetry <strong>in</strong> the field<br />

of the differently oriented molecules<br />

69


Figure 3.6: Dipoles <strong>in</strong>duced on each molecule <strong>in</strong> a pair of neutral ethylene molecules<br />

(upper pannel) and <strong>in</strong> a positively charged pair (lower panel). Notice that, as<br />

shown <strong>in</strong> figure 3.5, <strong>in</strong> a charged system the <strong>in</strong>duced dipoles are larger than <strong>in</strong><br />

the neutral case, but are <strong>in</strong> opposite directions.<br />

• for positively charged systems, the quadrupole <strong>in</strong>duced dipole actually dim<strong>in</strong>ishes<br />

the asymmetry <strong>in</strong> the monopole <strong>in</strong>duced dipoles and reduces the overall contribu-<br />

tion to ∆E, mak<strong>in</strong>g it similar <strong>in</strong> magnitude to the quadrupole <strong>in</strong>duced dipole.<br />

The lesson from this <strong>in</strong>vestigation is that one has to take great care <strong>in</strong> predict<strong>in</strong>g the<br />

<strong>in</strong>ductive contributions to ∆E, because the site energy difference is determ<strong>in</strong>ed not only by<br />

the magnitude of the <strong>in</strong>teractions <strong>in</strong>volved but also, most importantly, by their asymmetry.<br />

3.3.3 Pentacene<br />

In this section we will compare the site energy difference calculated us<strong>in</strong>g distributed mul-<br />

tipole analysis for pairs of pentacene molecules with that by the projective method applied<br />

to PW91/TVZP calculations. Pentacene is chosen as a typical conjugated molecule that<br />

has been studied <strong>in</strong> the literature [11]. The projection of the DFT MOs is carried out<br />

exactly as <strong>in</strong> the case of ethylene: namely calculations of the overlap matrix, the orbitals<br />

of the pair of molecules and of each molecule separately are carried out <strong>in</strong> GAMESS-<br />

UK; the orbitals are then orthogonalized and projected; f<strong>in</strong>ally the Kohn-Sham matrix is<br />

recreated us<strong>in</strong>g the mathematical package Octave with some custom written scripts. The<br />

70


Figure 3.7: Two cofacial pentacene molecules and the axis around which one will be rotated<br />

and translated. The <strong>in</strong>set shows the chemical formula of pentacene.<br />

calculation of the isolated molecule’s electrical moments is carried out us<strong>in</strong>g Gaussian ’03<br />

with the same functional and basis set specified <strong>in</strong> GAMESS-UK; the DMA is then car-<br />

ried out by the helper program gdma [10]. The calculations on a cation were carried out<br />

us<strong>in</strong>g the guess wavefunction from the calculation on the neutral molecule and without<br />

carry<strong>in</strong>g out the SCF procedure, <strong>in</strong> an effort to replicate the frozen orbital approximation<br />

assumed <strong>in</strong> the projective method. It was necessary to carry out the DMA <strong>in</strong> gdma, rather<br />

than us<strong>in</strong>g the values given by GAMESS-UK because the latter gave multipoles which<br />

did not reflect the symmetry of the molecule.<br />

The geometry of the pair of pentacenes considered is shown <strong>in</strong> figure 3.7: the two<br />

pentacene molecules are kept at a distance of m<strong>in</strong>imum approach of 3.5 Å, and one of<br />

them is rotated along its long axis. The results for ∆E as a function of rotation angle are<br />

shown <strong>in</strong> figure 3.8. This figure shows how the classical DMA results are <strong>in</strong> relatively<br />

good agreement with the DFT results, and how <strong>in</strong>clud<strong>in</strong>g only the monopole <strong>in</strong>teractions<br />

accounts for only half of the electrostatic <strong>in</strong>teraction energy calculated us<strong>in</strong>g DMA. It<br />

also shows that the DMA results somewhat underestimate the PW91 results; this is a<br />

rather surpris<strong>in</strong>g result as we would expect that <strong>in</strong>clud<strong>in</strong>g <strong>in</strong>ductive <strong>in</strong>teractions would<br />

further lower the electrostatically calculated ∆E. Three possible reasons could expla<strong>in</strong> this<br />

discrepancy: us<strong>in</strong>g Gaussian <strong>in</strong>stead of GAMESS-UK might mean we are us<strong>in</strong>g slightly<br />

71


Figure 3.8: ∆E for pairs of pentacene molecules, calculated us<strong>in</strong>g projection of the PW91<br />

results (red po<strong>in</strong>ts) and electrostatic results (diamonds). The electrostatic results<br />

are calculated us<strong>in</strong>g monopole monopole <strong>in</strong>teractions only (squares) and<br />

us<strong>in</strong>g all <strong>in</strong>teractions up to quadrupole-quadrupole (circles).<br />

different implementations of the functional; <strong>in</strong> fact, Gaussian calculates a total energy<br />

for a molecule of pentacene which is 57 meV smaller than GAMESS-UK, demonstrat<strong>in</strong>g<br />

the two programs are not exactly identical. It is also possible that the multipole expansion<br />

does not actually reproduce the field of the molecule predicted by PW91, either because of<br />

the method gdma uses to partition the molecule <strong>in</strong>to atoms that the moments are <strong>in</strong>tegrated<br />

over or because the expansion is truncated at quadrupoles. It should be noted, however,<br />

that the agreement between the DFT results and the classical results is still rather good:<br />

the relative difference between the two method is approximately 10%, for an angle of 90 ◦ .<br />

The conclusion of the past three sections is that most of the site energy difference can<br />

be expla<strong>in</strong>ed <strong>in</strong> terms of the electrostatic <strong>in</strong>teraction between two molecules, which is<br />

best calculated us<strong>in</strong>g distributed multipole analysis. The great advantage of electrostatic<br />

<strong>in</strong>teraction is that they are pairwise additive. Therefore it is easy to go from describ<strong>in</strong>g<br />

site energies <strong>in</strong> pairs of molecules to describ<strong>in</strong>g them <strong>in</strong> large assemblies of molecules by<br />

add<strong>in</strong>g the contributions to the energy from all the pairs <strong>in</strong> the assembly.<br />

The role of <strong>in</strong>ductive <strong>in</strong>teractions is still difficult to identify exactly. So far we have<br />

been able to perform these calculations only for molecules at large separation, as we have<br />

not implemented any method to calculate polarisabilities <strong>in</strong> non-uniform electric fields.<br />

72


For ethylene pairs we found that the contributions to ∆E from <strong>in</strong>ductive <strong>in</strong>teractions is<br />

much smaller than from the electrostatic <strong>in</strong>teractions. Also, it is not strongly dependent<br />

on whether we consider the dipole <strong>in</strong>duced <strong>in</strong> neutral pairs or <strong>in</strong> charged ones, although<br />

this observation might be specific to the system considered. Another complication is that<br />

<strong>in</strong>ductive <strong>in</strong>teractions are not at all additive, therefore the self-consistent procedure used<br />

to determ<strong>in</strong>e <strong>in</strong>teraction for pairs of molecules would have to be carried out for the whole<br />

assembly of molecules.<br />

3.4 Comparison of Different Methods of <strong>Calculat<strong>in</strong>g</strong> Trans-<br />

fer Integrals<br />

In this section we will compare the three methods described to calculate the transfer <strong>in</strong>-<br />

tegral J: splitt<strong>in</strong>g of frontier orbital energies, projective method and MOO. The aim is to<br />

show that the use of MOO is well justified when compared to both SCF and more ab-<strong>in</strong>itio<br />

methods. We will use three different molecules to carry out these comparisons: pentacene,<br />

hexabenzocoronene and ethylene. Ethylene is <strong>in</strong>cluded because the previous calculations<br />

of ∆E furnished us with the <strong>in</strong>formation necessary to study the basis set dependence of<br />

transfer <strong>in</strong>tegrals.<br />

3.4.1 Pentacene<br />

In this section, pentacene will be used to compare the orbital splitt<strong>in</strong>g method with the<br />

projective method. We will also show that the projective method can be used with either<br />

a ZINDO Hamiltonian or a PW91 functional to yield very similar results. The geome-<br />

tries we will <strong>in</strong>vestigate are represented <strong>in</strong> figure 3.7. We consider the cases where one<br />

molecule is translated along the long axis of the pentacene molecule, which we label the<br />

x-axis, and where one molecule is rotated around that axis.<br />

Because of the D2h symmetry of pentacene, a pair of two parallel molecules with<br />

one molecule translated along the x axis possesses a po<strong>in</strong>t of <strong>in</strong>version: that is, there is<br />

a unitary transformation which br<strong>in</strong>gs one molecule onto the other. This means that the<br />

site energies EA and EB <strong>in</strong> equation 2.19 must be the same: we should therefore be <strong>in</strong><br />

73


Figure 3.9: Absolute value of the hole transfer <strong>in</strong>tegral as a function of displacement along<br />

the x axis of one of the two pentacene molecules calculated by projection of<br />

ZINDO orbitals (circles) and by the orbital splitt<strong>in</strong>g model (triangles). The<br />

length of the molecule is 14.1 Å .<br />

Figure 3.10: HOMO for pentacene calculated at the ZINDO level.<br />

74


Figure 3.11: Absolute value of the transfer <strong>in</strong>tegral for a pair of pentacene molecules as a<br />

function of rotation around the x axis. The transfer <strong>in</strong>tegral is calculated by<br />

projection of the ZINDO orbitals (circles) and by the orbital splitt<strong>in</strong>g method<br />

(triangles).<br />

the regime when the projective and orbital splitt<strong>in</strong>g methods give the same results. The<br />

hole transfer <strong>in</strong>tegral as calculated by each method is shown <strong>in</strong> figure 3.9. The orig<strong>in</strong><br />

for the oscillations <strong>in</strong> |J| with displacement <strong>in</strong> the x direction can be understood <strong>in</strong> terms<br />

of orbital overlap. As the two molecules are translated over each other, the nodes and<br />

ant<strong>in</strong>odes of the HOMOs on each molecule pass over each other; if the nodes of the<br />

HOMOs of one molecule are over the nodes, the contributions add and one obta<strong>in</strong>s a<br />

large transfer <strong>in</strong>tegral. If the nodes are over the ant<strong>in</strong>odes the contributions cancel out and<br />

the transfer <strong>in</strong>tegral vanishes [12]. This explanation is consistent with the period of the<br />

oscillation, which is approximately the same as the width of a benzene r<strong>in</strong>g <strong>in</strong> pentacene,<br />

and with the shape of the HOMO of pentacene - shown <strong>in</strong> figure 3.10. The reason why<br />

the peak amplitude is decreas<strong>in</strong>g with displacement is because of the dim<strong>in</strong>ished overlap<br />

between the two molecules. It is evident from figure 3.9 that, for this symmetric geometry,<br />

the projective and splitt<strong>in</strong>g methods are <strong>in</strong> very close agreement.<br />

Now let us look at the two pentacene molecules separated by 4.5 Å where one molecule<br />

is rotated relative to the other, as shown <strong>in</strong> figure 3.7. For any angle different from zero<br />

there will be no unitary transformation that moves the atoms of one molecule onto the<br />

other; therefore symmetry will not ensure that the site energies will be the same. Further-<br />

75


Transfer <strong>in</strong>tegrals / eV<br />

0.06<br />

0.05<br />

0.04<br />

0.03<br />

0.02<br />

0.01<br />

sto-3g<br />

DVZP<br />

TVZP<br />

CC-PVQZ<br />

0<br />

0 50 100<br />

Rotation angle / degrees<br />

150 200<br />

Figure 3.12: Absolute value of the transfer <strong>in</strong>tegral for a pair of ethylene molecules as a<br />

function of rotation angle. All curves were obta<strong>in</strong>ed by projective method<br />

applied to PW91 calculations; the different curves are obta<strong>in</strong>ed by us<strong>in</strong>g<br />

different basis sets.<br />

more the electric field of pentacene is asymmetric and as we discussed <strong>in</strong> the previous<br />

section this will lead to a difference <strong>in</strong> site energies. In this regime we do not expect the<br />

two methods to agree and <strong>in</strong> fact they do not, as shown <strong>in</strong> figure 3.11. The results from<br />

the projective method are sensible: the transfer <strong>in</strong>tegral falls to zero as the π systems on<br />

the two molecules become orthogonal. The orbital splitt<strong>in</strong>g method, conversely, suggests<br />

<strong>in</strong>correctly that the transfer <strong>in</strong>tegral would <strong>in</strong>crease at large angles. The discrepancy is<br />

expla<strong>in</strong>ed by the fact that at larger angles the site energy difference <strong>in</strong>creases.<br />

Hav<strong>in</strong>g confirmed the accuracy of calculat<strong>in</strong>g |J| us<strong>in</strong>g the projective method, we now<br />

address the accuracy of results obta<strong>in</strong>ed by the projective method applied to ZINDO re-<br />

sults and by the projective method applied to PW91 calculations. We made a prelim<strong>in</strong>ary<br />

study on the pairs of ethylene molecules as a function of rotation angle us<strong>in</strong>g the pure DFT<br />

functional PW91 and study<strong>in</strong>g the dependence of the transfer <strong>in</strong>tegral on the basis set. A<br />

strong dependence is expected because the transfer <strong>in</strong>tegral is <strong>in</strong>timately connected to or-<br />

bital overlap, and therefore the ability of a basis set to describe charge density far from<br />

76


Figure 3.13: Absolute value of the transfer <strong>in</strong>tegral as a function of displacement along<br />

the x axis of one of the two pentacene molecules. The two curves show<br />

the results from the projective method us<strong>in</strong>g the PW91 functional and the<br />

ZINDO Hamiltonian. The PW91 calculation were performed with a TVZP<br />

basis set.<br />

a molecule is important. We compare four different basis sets: m<strong>in</strong>imal Slater type or-<br />

bital (sto-3g), a double and a triple zeta basis set with extra polarisation functions (DVZP<br />

and TVZP) and f<strong>in</strong>ally a huge correlation-corrected qu<strong>in</strong>tuple zeta set with extra polari-<br />

sation basis sets (cc-QVZP). The results of this study show that a TVZP set is sufficient<br />

to describe transfer <strong>in</strong>tegrals appropriately: as shown <strong>in</strong> figure 3.12, both the TVZP and<br />

the cc-QVZP sets give similar results. Hav<strong>in</strong>g established this, we compare the results<br />

for the case of pentacene molecules with relative displacement us<strong>in</strong>g the ZINDO and<br />

PW91/TVZP model chemistries. The results from this comparison are shown <strong>in</strong> figure<br />

3.13. We can see that the two curves are very similar, justify<strong>in</strong>g the use of ZINDO. It is<br />

pretty astonish<strong>in</strong>g that the ZINDO Hamiltonian, with its m<strong>in</strong>imal basis set, produces such<br />

good agreement with the results from a much more time consum<strong>in</strong>g PW91 calculation! It<br />

should also been noted that these results are at odds with the <strong>in</strong>vestigation of Huang and<br />

coworkers [13] who f<strong>in</strong>d significant discrepancies <strong>in</strong> the predicted value of J us<strong>in</strong>g PW91<br />

and ZINDO methods for pairs of ethylene molecules. It is possible that these differences<br />

are less pronounced for larger molecules than for smaller oleculas such as ethylene.<br />

In this section we have validated the use of the projective method by comparison to<br />

77


Figure 3.14: Two cofacial HBC molecules and the cartesian axis used <strong>in</strong> the text. The<br />

<strong>in</strong>set shows the chemical formula of HBC.<br />

the orbital splitt<strong>in</strong>g method when the site energies of two molecules are the same. When<br />

site energies are different, the projective method gives sensible results. F<strong>in</strong>ally, we show<br />

that the projective method can be applied both to ZINDO and PW91 results and, <strong>in</strong> both<br />

cases, the transfer <strong>in</strong>tegral obta<strong>in</strong>ed is similar: for the rest of the thesis we will use the<br />

ZINDO method because of its greater computational efficiency.<br />

3.4.2 Hexabenzocoronene<br />

In this section we compare transfer <strong>in</strong>tegrals calculated us<strong>in</strong>g the projective and MOO<br />

methods, us<strong>in</strong>g hexabenzocoronene (HBC) as an example molecule. The structure of two<br />

cofacial molecules and the def<strong>in</strong>ition of the axes are shown <strong>in</strong> figure 3.14.<br />

Before show<strong>in</strong>g the comparisons of the projective and MOO methods, we briefly re-<br />

view some of the literature on HBC. HBC has been studied <strong>in</strong> the literature by Lemaur<br />

and co-workers [14]: <strong>in</strong> that work, HBC and other discotic liquid crystals were compared<br />

by calculat<strong>in</strong>g at the relative values of the reorganisation energy and the dependence of<br />

the transfer <strong>in</strong>tegral on rotation about the z axis. The rotational potential energy for ro-<br />

tation about the z axis was obta<strong>in</strong>ed from molecular mechanics calculations on pairs of<br />

molecules and an expression for the relative mobility <strong>in</strong> the various compounds was ob-<br />

ta<strong>in</strong>ed us<strong>in</strong>g a mean field treatment. We would like to expand on this work by consider<strong>in</strong>g<br />

78


Figure 3.15: Contour plots of the effective transfer <strong>in</strong>tegral as a function of rotation about<br />

the x and z axis. The molecules are first rotated about the z axis, then rotated<br />

about the x axis <strong>in</strong> such a way as to keep the m<strong>in</strong>imum distance of approach<br />

between the molecules equal to 3.5 Å .<br />

the dependence of the transfer <strong>in</strong>tegral on the degrees of freedom neglected <strong>in</strong> this study,<br />

rotation about the x axis and slip <strong>in</strong> the x-y plane. In section 7.2 we comb<strong>in</strong>e calculation of<br />

transfer <strong>in</strong>tegrals with accurate atomistic molecular dynamics modell<strong>in</strong>g of entire stacks<br />

of HBC. It is <strong>in</strong> order to carry out this modell<strong>in</strong>g that we aim to show the projective and<br />

MOO methods to be equivalent. S<strong>in</strong>ce the HOMO of HBC is doubly degenerate, <strong>in</strong> this<br />

section and <strong>in</strong> 7.2 we use an effective transfer <strong>in</strong>tegral Je f f , as def<strong>in</strong>ed <strong>in</strong> equation 2.18,<br />

to determ<strong>in</strong>e the charge transfer rates. In reference [14], <strong>in</strong>stead of the effective transfer<br />

<strong>in</strong>tegral the splitt<strong>in</strong>g between the degenerate HOMOs of a pair of HBC molecules is taken<br />

as the electronic coupl<strong>in</strong>g. In this section we expla<strong>in</strong> why, for certa<strong>in</strong> relative orientations,<br />

this approach is equivalent to ours except for a difference <strong>in</strong> a constant scal<strong>in</strong>g factor.<br />

We wish to carry out two comparisons of the effective transfer <strong>in</strong>tegral: as a function<br />

of x-y displacement and as a function of rotation about the z axis and about the x axis.<br />

First we show the results for rotation about the x and z axes. The relative geometry of the<br />

molecules is obta<strong>in</strong>ed by start<strong>in</strong>g with two identical molecules A and B: first molecule<br />

B is rotated about the z axis, then it is rotated about the x axis, f<strong>in</strong>ally molecule B is<br />

translated along the z axis by as much as is necessary to make the m<strong>in</strong>imum distance of<br />

approach equal to 3.5 Å. The results for the effective transfer <strong>in</strong>tegral Je f f are shown <strong>in</strong><br />

figure 3.15. It can immediately be seen that both methods predict very similar results. We<br />

have chosen to limit the z rotation to between 0 ◦ and 180 ◦ <strong>in</strong> order to show that the transfer<br />

<strong>in</strong>tegral varies with a period of 120 ◦ , as suggested by the S6 symmetry of the molecule. If<br />

we had constra<strong>in</strong>ed the HBC molecule to a planar geometry, the overall symmetry would<br />

have been D6h and the period of Je f f would have been 60 ◦ . Rotations about the x axis are<br />

79


Transfer <strong>in</strong>tegral / eV<br />

0.4<br />

0.2<br />

0<br />

-0.2<br />

MOO J (HOMO HOMO)<br />

MOO J (HOMO HOMO-1)<br />

MOO J (HOMO-1 HOMO)<br />

MOO J (HOMO-1 HOMO-1)<br />

PRO Jeff MOO J<br />

eff<br />

-0.4<br />

0 50 100<br />

Rotation angle / degrees<br />

150 200<br />

Figure 3.16: Effective transfer <strong>in</strong>tegral and its components as a function of rotation about<br />

the z axis calculated us<strong>in</strong>g the projective (PRO) and MOO (MOO) methods.<br />

The molecules are at a distance of 3.5 Å .<br />

limited to ±30 ◦ because atomistic modell<strong>in</strong>g of various HBC derivatives has suggested<br />

that this is the average tilt <strong>in</strong> the HBC derivatives with most disorder[15]. Figure 3.16<br />

also shows that, for a particular rotational angle about the x axis, rotation about the z axis<br />

can cause a variation <strong>in</strong> the transfer <strong>in</strong>tegral of roughly a factor of 6. In contrast figure<br />

3.15 shows that a rotation of 30 ◦ about the x axis produces a variation <strong>in</strong> transfer <strong>in</strong>tegral<br />

of a factor of 20, because of the greater variation <strong>in</strong> π system overlap. In a discotic liquid<br />

crystal such as HBC we expect that rotations about the x axis should be small, therefore<br />

we expect that the reduction <strong>in</strong> order caused by <strong>in</strong>creas<strong>in</strong>g the temperature or by hav<strong>in</strong>g<br />

shorter side cha<strong>in</strong> lengths will have drastic effects on the charge transport characteristics.<br />

It is <strong>in</strong>terest<strong>in</strong>g at this po<strong>in</strong>t to compare these results with the results calculated by<br />

Lemaur and co-workers [14]. In that work, the transfer <strong>in</strong>tegral was calculated from the<br />

splitt<strong>in</strong>g between the top two orbitals of the pair of molecules and the next two. Sur-<br />

pris<strong>in</strong>gly, this treatment produced similar results for the dependence of J on z rotation as<br />

those from equation 2.18 shown <strong>in</strong> figure 3.16. This can be expla<strong>in</strong>ed by consider<strong>in</strong>g the<br />

values of the transfer <strong>in</strong>tegral as a function of rotation about the z axis. These results are<br />

shown <strong>in</strong> figure 3.16 for the MOO method, with the results for Je f f calculated from the<br />

projective method shown for comparison. This figure clearly shows how - for this geome-<br />

try - JHOMO,HOMO is equal to JHOMO−1,HOMO−1 and JHOMO,HOMO−1 is equal to JHOMO−1,HOMO,<br />

as expected for a system with C3 symmetry such as these systems [2]. Therefore, us<strong>in</strong>g<br />

80


similar arguments to those used for systems with non-degenerate orbitals, we can write<br />

the matrix govern<strong>in</strong>g the orbitals of the pair of molecules as:<br />

F =<br />

E 0 JHOMO,HOMO JHOMO,HOMO−1<br />

0 E −JHOMO,HOMO−1 JHOMO,HOMO<br />

JHOMO,HOMO −JHOMO,HOMO−1 E 0<br />

JHOMO,HOMO−1 JHOMO,HOMO 0 E<br />

(3.12)<br />

Diagonalis<strong>in</strong>g the Fock matrix F, we obta<strong>in</strong> two sets of doubly degenerate eigenvalues<br />

for the pair of molecules, with energies correspond<strong>in</strong>g to:<br />

E± = E ±<br />

�<br />

J2 homo,homo + J2<br />

homo,homo−1<br />

(3.13)<br />

This expla<strong>in</strong>s why the difference between the splitt<strong>in</strong>g from a pair of molecules re-<br />

ported by Lemaur <strong>in</strong> [14] and the values shown <strong>in</strong> figure 3.16 is a factor of 2 √ (2). In the<br />

case when JHOMO,HOMO is different from JHOMO−1,HOMO−1 and JHOMO−1,HOMO is different<br />

from −JHOMO,HOMO−1, for example if x rotations are <strong>in</strong>cluded, equation 3.13 does not hold<br />

anymore and it is not straightforward to deduce the transfer <strong>in</strong>tegral from the splitt<strong>in</strong>g of<br />

the frontier orbitals, especially if polarisation makes the diagonal terms of the 4x4 matrix<br />

from equation 3.12 different. Therefore consider<strong>in</strong>g the splitt<strong>in</strong>g of the orbitals seems<br />

valid only <strong>in</strong> the specific cases where the pair of molecules has a particular symmetry.<br />

It is also <strong>in</strong>terest<strong>in</strong>g to look at figure 3.16 and note how neither the JHOMO,HOMO nor<br />

the JHOMO−1,HOMO possess the S6 symmetry of the molecule, this gives a graphic explana-<br />

tion of why - if Jahn-Teller distortions are ignored - it is necessary to look at the effective<br />

transfer <strong>in</strong>tegral <strong>in</strong> order to respect the symmetry of the system analysed. The degener-<br />

acy of the frontier orbitals is not accidental, but reflects the symmetry properties of the<br />

molecules <strong>in</strong>volved.<br />

F<strong>in</strong>ally let us look briefly at displacements <strong>in</strong> the x-y direction. A comparison of the<br />

results for the projective and MOO methods is shown <strong>in</strong> figure 3.17. Aga<strong>in</strong> it can be seen<br />

that the two contour plots are very similar. We do not expect HBC molecules <strong>in</strong> a discotic<br />

liquid phase to be able to slip freely <strong>in</strong> the x-y direction, as later movements should<br />

81


Figure 3.17: Countour plots of the dependence on x-y displacement of two molecules of<br />

HBC. The two molecules are at a distance of 3.5 Å. The maximum radius of<br />

the molecule is 6.8 Å.<br />

be restricted by the <strong>in</strong>teractions with the surround<strong>in</strong>g columns. However a variation <strong>in</strong><br />

transfer <strong>in</strong>tegral of a factor of two is obta<strong>in</strong>ed by displacements of less than 2 Å.<br />

To conclude this section we have shown how the MOO method adequately reproduces<br />

results from the projective method applied to the ZINDO Hamiltonian. This is at a consid-<br />

erable computational advantage, s<strong>in</strong>ce all calculations on pairs of HBC molecules require<br />

only a s<strong>in</strong>gle ZINDO calculation to be performed on an isolated HBC molecules, whereas<br />

the projective method requires a SCF calculation for each pair considered. We have also<br />

shown that tilt<strong>in</strong>g of the disks will has a tremendous effect on transfer <strong>in</strong>tegral and have<br />

given a graphical demonstration of the significance of molecular orbital degeneracy <strong>in</strong> the<br />

calculation of transfer <strong>in</strong>tegrals.<br />

3.5 Conclusion<br />

In this chapter we have presented fast methods which allow the calculation of J and ∆E<br />

for very large numebers of relative orientations and positions of molecules. The method<br />

to calculate ∆E is based on the classical treatment of electrostatic <strong>in</strong>teraction discussed<br />

<strong>in</strong> chapter two, whereas the method to calculate J is based on an approximation of the<br />

ZINDO Hamiltonian. The site energy differences thus calculated were found to be <strong>in</strong><br />

agreement with site energy differences deduced from projection of DFT calculations,<br />

whereas the MOO method for calculat<strong>in</strong>g J was found to be <strong>in</strong> agreement with the pro-<br />

jective method applied to ZINDO claculations.<br />

82


Bibliography<br />

[1] J. Ridley and M. Zerner, An Intermediate Neglect of Differential Overlap Technique<br />

for Spectroscopy: Pyrrole and the Az<strong>in</strong>es, Theoretica Chimica Acta, 32, 111 (1973)<br />

[2] K. Senthilkumar, F.C. Groezema, F.M. Bickelhaupt and L.D.A. Siebbeles, <strong>Charge</strong><br />

transport <strong>in</strong> columnar stacked triphenylenes: Effects of conformational fluctuations<br />

on charge transfer <strong>in</strong>tegrals and site energies, Journal of Chemical Physics, 32, 9809<br />

(2003)<br />

[3] G. te Velde, F. M. Bickelhaupt, E. J. Baerends, C. Fonseca Guerra, S.J.A. Van Gis-<br />

bergen, J.G. Snijders, and T. Ziegler, Chemistry with ADF, Journal Computational<br />

Chemistry, 22, 931 (2001).<br />

[4] L.D. Landau and E.M. Lifshitz, Quantum Mechanics non-relativistic theory, Perga-<br />

mon Press (1965)<br />

[5] P.O. Lowd<strong>in</strong>, On the Non-Orthogonality Problem Connected with the Use of Atomic<br />

Wave Functions <strong>in</strong> the Theory of Molecules and Crystals, Journal of Chemical Physics,<br />

18, 365 (1950)<br />

[6] R.S. Mulliken, C.A. Reike, D. Orloff and H. Orloff, Formulas and numerical table<br />

for overlap <strong>in</strong>tegrals, Journal of Chemical Physics, 17, 1248 (1949)<br />

[7] C.A. Hunter and J.K.M. Sanders, The nature of π − π <strong>in</strong>teractions, Journal of the<br />

American Chemical Society, 112, 5525 (1990)<br />

[8] M.F. Guest, I.J. Bush, H.J.J. van Dam, P. Sherwood, J.M.H. Thomas, J.H. van Lenthe,<br />

R.W.A. Havenith and J. Kendrick, The GAMESS-UK electronic structure package:<br />

algorithms, developments and applications, Molecular Physics, 103, 719 (2005)<br />

83


[9] E.F. Valeev, V. Coropceanu, D.A. da Silva Filho, S. Salman, and J.L. Brédas, Ef-<br />

fect of Electronic Polarisation on <strong>Charge</strong>-<strong>Transport</strong> <strong>Parameters</strong> <strong>in</strong> Molecular Organic<br />

Semiconductors, Journal of the American Chemical Society, 128, 9882 (2006)<br />

[10] A.J. Stone, Journal of chemical theory and computation, Distributed multipole anal-<br />

ysis: Stability for large basis sets, 1, 1128 (2005)<br />

[11] O. Kwon, V. Coropceanu, N.E. Gruhn, J.C. Durivage, J.G. Laqu<strong>in</strong>danum, H.E. Katz,<br />

J. Cornil and J.L. Brédas, Characterisation of the molecular parameters determ<strong>in</strong><strong>in</strong>g<br />

charge transport <strong>in</strong> anthradithiophene, Journal of Chemical Physics 120, 8186 (2004)<br />

[12] J.L. Brédas, J.P. Calbert, D. A. da Silva Filho and J. Cornil, Organic semiconductors:<br />

A theoretical characterisation of the basic parameters govern<strong>in</strong>g charge transport,<br />

Proceed<strong>in</strong>gs of the National Academy of Science 59, 5804(2002)<br />

[13] J. Huang, M. Kertesz, Validation of <strong>in</strong>termolecular transfer <strong>in</strong>tegral and band-<br />

width calculations for organic molecular materials, Journal of Chemical Physics, 122,<br />

234707 (2005)<br />

[14] V. Lemaur, D.A. da Silva Filho, V. Coropcenau, M. Lehmann, Y. Geerts, J. Piris,<br />

M.G. Debije, A.M. van der Craats, K. Senthilkumar, L.D.A. Siebbeles, J.M. Warman,<br />

J.L. Brédas and J. Cornil, Journal of the American Chemical Society, <strong>Charge</strong> <strong>Transport</strong><br />

Properties <strong>in</strong> Discotic Liquid Crystals: A Quantum-Chemical Insight <strong>in</strong>to Structure-<br />

Property Relationships, 126126, 3271 (2004)<br />

[15] D. Andrienko, V. Marcon, K. Kremer, Atomistic simulation of structure and dy-<br />

namics of columnar phases of hexabenzocoronene derivatives, Journal of Chemical<br />

Physics, 125, 124902 (2006)<br />

84


Chapter 4<br />

Reorganisation Energy <strong>in</strong><br />

PolyPhenylenev<strong>in</strong>ylene: Polaron<br />

Localisation<br />

In the previous chapters, we have described ways of calculat<strong>in</strong>g the parameters of the<br />

Marcus electron transfer equation <strong>in</strong> terms of s<strong>in</strong>gle molecule properties. We discussed<br />

how the reorganisation energy can be calculated from the relaxation energy of radical<br />

cations (or anions) for s<strong>in</strong>gle molecules and how the site energy difference and transfer<br />

<strong>in</strong>tegral can be computed from the diagonal and off-diagonal elements of the Fock matrix<br />

written <strong>in</strong> the localised basis set of the molecular orbitals of each molecule. A problem<br />

arises when we want to extend this treatment to polymers: what segment length should<br />

be considered as the charge transport<strong>in</strong>g unit? The oligomer segment length considered<br />

is important as one would expect that <strong>in</strong>creas<strong>in</strong>g its length would decrease the ionisa-<br />

tion potential and <strong>in</strong>crease the electron aff<strong>in</strong>ity. In analogy to a potential well, if the<br />

well is made wider the energy levels become shallower. In this chapter we will consider<br />

electron-phonon coupl<strong>in</strong>g as a process which can limit the localisation of charge <strong>in</strong> a poly-<br />

mer, thereby creat<strong>in</strong>g an effective cha<strong>in</strong> length beyond which charged radical excitations<br />

do not spread. The polymer we take as an example is polyparaphenylenev<strong>in</strong>ylene (PPV),<br />

shown <strong>in</strong> figure 4.1. The figure also shows the labell<strong>in</strong>g convention of the phenylene units<br />

and v<strong>in</strong>ylene units and of atoms we use <strong>in</strong> the rest of the chapter. In particular, we will<br />

look at two possible models for polarons: those predicted by semi-empirical calculations<br />

85


Figure 4.1: Two monomers of PPV and the scheme of the labels used <strong>in</strong> the rest of the<br />

chapter: ph1...phn will label the phenylene units and vi1...vi2 will label the<br />

v<strong>in</strong>ylene units. The first monomer also has the carbons with bonds to hydrogen<br />

labeled 1 to 6.<br />

and those predicted by hybrid density functional methods. We will show that even though<br />

the sp<strong>in</strong> density distributions obta<strong>in</strong>ed from DFT calculations are more delocalised than<br />

those obta<strong>in</strong>ed from semi-empirical methods, neither method is <strong>in</strong> disagreement with the<br />

sp<strong>in</strong> distribution observed by electron nuclear double resonance (ENDOR) experimental<br />

spectra on PPV. We will then discuss the <strong>in</strong>fluence of torsional fluctuations on polaron lo-<br />

calisation to show how the charge transport<strong>in</strong>g units are liable to be affected by fluctuation<br />

<strong>in</strong> torsional angles small enough to be easily accessible at room temperature.<br />

4.1 Introduction<br />

A process that can limit the spatial extent of a charge on a polymer cha<strong>in</strong> is polaron<br />

self-localisation, that is an <strong>in</strong>teraction of a charge with the polymer backbone such that<br />

the charged excitation and the cha<strong>in</strong> deformation localise. The term polaron was origi-<br />

nally used <strong>in</strong> the context of describ<strong>in</strong>g quasi-particles created by the <strong>in</strong>teraction between<br />

lattice vibrations (phonons) and electrons <strong>in</strong> <strong>in</strong>organic polar crystals. In the case of or-<br />

ganic solids, there is evidence for the formation of polarons <strong>in</strong> stretched films of poly<br />

para phenylenev<strong>in</strong>ylene (PPV) from electron nuclear double resonance spectroscopy (EN-<br />

DOR) measurements on PPV crystals [1, 2, 3]. This technique measures the hyperf<strong>in</strong>e<br />

<strong>in</strong>teraction between carbon and hydrogen atoms and is therefore very sensitive to the ab-<br />

solute value of the sp<strong>in</strong> density on the carbon atoms bonded to hydrogen atoms. The<br />

ENDOR spectrum consists of a series of resonance peaks measured at different magnetic<br />

86


Figure 4.2: Experimental and modelled ENDOR spectra of PPV with an applied magnetic<br />

field parallel and perpendicular to the polymer cha<strong>in</strong> axis. The figure is from<br />

reference [2].<br />

fields, as shown <strong>in</strong> the lowest panel of figure 4.2. The position of each peak can be deter-<br />

m<strong>in</strong>ed by the equation:<br />

ν± =<br />

� �<br />

i=x,y,z<br />

(ν0 ± Ai) 2 p 2 i<br />

(4.1)<br />

where ν0 is the free proton frequency, ± labels the shift on either branch of the ENDOR<br />

spectra, pi is the i th component of the magnetic field and f<strong>in</strong>ally Ai is the i th component<br />

of the hyperf<strong>in</strong>e coupl<strong>in</strong>g tensor, which is proportional to the absolute value of the sp<strong>in</strong><br />

density on the carbon atom multiplied by a geometry dependent proportionality factor.<br />

This expression is derived and discussed <strong>in</strong> reference [4]. Figure 4.2 shows three dis-<br />

t<strong>in</strong>ct ENDOR peaks observed experimentally, therefore because of equation 4.1 the sp<strong>in</strong><br />

densities of all carbon atoms bonded to hydrogens must take three dist<strong>in</strong>ct values. In ref-<br />

erence [2], Shimoi and co-workers f<strong>in</strong>d that such sp<strong>in</strong> density values can be obta<strong>in</strong>ed by<br />

us<strong>in</strong>g a Pariser-Parr-Pople (PPP) Hamiltonian, which takes electron-phonon coupl<strong>in</strong>g as<br />

a parameter. The authors f<strong>in</strong>d that the parameters which allow the experimental data to<br />

be fitted also lead to localisation of the charged excitation on a segment of roughly four<br />

units of PPV, as shown <strong>in</strong> figure 4.3. An important po<strong>in</strong>t to make is that these ENDOR<br />

measurements are made <strong>in</strong> the dark, <strong>in</strong> the absence of dop<strong>in</strong>g ions and at low temperature:<br />

<strong>in</strong> other words, the polaron seems to be an <strong>in</strong>tr<strong>in</strong>sic property of the polymer.<br />

The PPP Hamiltonian is an ad-hoc Hamiltonian, which has to be parametrised to<br />

match the experimental ENDOR data. Attempts have been made by Gesk<strong>in</strong> and co-<br />

87


Figure 4.3: Sp<strong>in</strong> density of a polaron on a PPV cha<strong>in</strong> calculated with a PPP Hamiltonian.<br />

Sites B and C’ correspond to sites 2 and 4 from figure 4.1, sites B’ and C<br />

correspond to sites 1 and 3 from figure 4.1 and sites E and F correspond to<br />

sites 5 and 6 from 4.1. The figure is from reference [2].<br />

workers [5, 6] to repeat such calculations us<strong>in</strong>g methods which are more ab-<strong>in</strong>itio and<br />

do not require such parametrisation. Three classes of methods were used by Gesk<strong>in</strong><br />

and co-workers for the geometry optimisation of radical cations on different lengths of<br />

polymer: pure DFT methods, hybrid DFT methods and semi-empirical methods (Aust<strong>in</strong><br />

Model 1). It was found that pure DFT methods show no localisation, semi-empirical<br />

methods show strong localisation and hybrid DFT methods an <strong>in</strong>termediate level of lo-<br />

calisation. In the first section of this chapter we repeat the calculation from reference [6]<br />

us<strong>in</strong>g hybrid DFT, with particular emphasis on the variation <strong>in</strong> polaron localisation energy<br />

with <strong>in</strong>creas<strong>in</strong>g oligomer length. We also argue that hybrid DFT, even though it does not<br />

show significant polaron localisation, still predicts variations <strong>in</strong> site sp<strong>in</strong> denisities which<br />

are not <strong>in</strong>consistent with the ENDOR spectra of PPV. In the second section of the chapter<br />

we <strong>in</strong>troduce torsional defects <strong>in</strong> the middle of an oligomer of PPV and show how they<br />

would cause polaron localisation by effectively break<strong>in</strong>g conjugation. We also show that<br />

the thermal energy required for such breaks of conjugation is relatively low. The conclu-<br />

sion we draw from this chaper is that polaron localisation is likely to be weak factor <strong>in</strong><br />

defect-free cha<strong>in</strong>s of PPV. A suitable model for charge transport <strong>in</strong> three dimensional PPV<br />

networks is one where torsional defects create conjugated segments with<strong>in</strong> which charge<br />

88


is delocalised and between which charge transport occurs <strong>in</strong> the temperature activated,<br />

<strong>in</strong>coherent Marcus regime. The boundaries of these conjugated segments would have to<br />

be def<strong>in</strong>ed dynamically, as the breaks <strong>in</strong> conjugation would be thermally activated.<br />

4.2 Defect-free PPV<br />

In this section we study polaron localisation <strong>in</strong> different length oligomers of PPV by<br />

calculat<strong>in</strong>g the sp<strong>in</strong> density distribution and polaron b<strong>in</strong>d<strong>in</strong>g energy. The calculations<br />

were performed with the BHandHLYP functional and the 6-31g* basis set, <strong>in</strong> accordance<br />

with the work of [5, 6]. Us<strong>in</strong>g a hybrid density functional with a large proportion of<br />

Hartree Fock exchange leads to qualitative localisation of the sp<strong>in</strong> density for cha<strong>in</strong>s of<br />

length smaller than ten, as shown <strong>in</strong> figure 4.4. In figure 4.4 we show the total Mulliken<br />

sp<strong>in</strong> density per phenylene and v<strong>in</strong>ylene unit on the cha<strong>in</strong>. To obta<strong>in</strong> these numbers, we<br />

perform Mulliken analysis [7] on the density matrix for each sp<strong>in</strong>, then subtract the up and<br />

down sp<strong>in</strong> populations and f<strong>in</strong>ally sum the sp<strong>in</strong> density over all the atoms belong<strong>in</strong>g to the<br />

same phenylene or v<strong>in</strong>ylene unit. Mulliken analysis tends to overlocalise charge densities<br />

on each atom, so that the difference <strong>in</strong> sp<strong>in</strong> density between neighbour<strong>in</strong>g phenylene and<br />

v<strong>in</strong>ylene units might be somewhat exaggerated. Nevertheless the qualitative observation<br />

is clear: sp<strong>in</strong> density is peaked around the middle of the oligomer cha<strong>in</strong>. The spatial<br />

extent of the sp<strong>in</strong> density is approximately four phenylene units, <strong>in</strong> accordance with [2].<br />

On the other hand it is also evident that for the longest oligomers considered the<br />

sp<strong>in</strong> density is more evenly distributed along the cha<strong>in</strong>. In order to assess whether this<br />

functional is predict<strong>in</strong>g polaron localisation we calculate the relaxation energy of a radical<br />

cation ( λi2 from equation 2.11). Figure 4.5 shows the polaron b<strong>in</strong>d<strong>in</strong>g energy for different<br />

length oligomers: as the oligomer becomes longer this quantity decreases and seems to<br />

tend to values smaller than 75 meV for <strong>in</strong>f<strong>in</strong>ite cha<strong>in</strong>s. This suggests that, although figure<br />

4.4 suggests some degree of localisation of the sp<strong>in</strong> density <strong>in</strong> short cha<strong>in</strong>s, the polaron is<br />

actually only very weakly localised <strong>in</strong> a longer, defect-free cha<strong>in</strong>.<br />

In order to show that the hybrid DFT results are not <strong>in</strong>consistent with ENDOR data,<br />

we compare the total sp<strong>in</strong> density on each carbon atom bonded to a hydrogen for a PPV<br />

oligomer of length 12 obta<strong>in</strong>ed with the BHandHLYP functional and the AM1 (Aust<strong>in</strong><br />

89


total sp<strong>in</strong> density<br />

0.4<br />

0.3<br />

0.2<br />

0.1<br />

0<br />

12PV<br />

10PV<br />

8PV<br />

6PV<br />

4PV<br />

2PV<br />

ph1 ph2 ph3 ph4 ph5 ph6 ph7 ph8 ph9 ph10 ph11 ph12<br />

r<strong>in</strong>g number<br />

Figure 4.4: Mulliken sp<strong>in</strong> density for different length oligomers of PPV. The x axis labels<br />

position along the cha<strong>in</strong>: the labels ph1, ph2, etc. label the first, second, etc.<br />

phenylene unit of the PPV. Between the phenylene units phn and phn + 1 lies<br />

the v<strong>in</strong>ylene unit v<strong>in</strong>.<br />

90


polaron localization energy / eV<br />

0.25<br />

0.2<br />

0.15<br />

0.1<br />

polaron localization energy<br />

0 2 4 6 8 10 12<br />

length of oligomer<br />

Figure 4.5: Polaron localisation energy for different length oligomers of PPV calculated<br />

with BHandHLYP.<br />

Model 1) Hamiltonian. In accordance with reference [8] we use a restricted open shell<br />

wavefunction when us<strong>in</strong>g AM1. It should also be noted that on a polymer of such a<br />

great length, the convergence of the geometry optimisation us<strong>in</strong>g AM1 is very difficult,<br />

as the potential energy surface of such a large polymer is very flat. Therefore we were<br />

obliged to relax the convergence criteria of Gaussian 1 . The comparison of the AM1 and<br />

BHandHLYP results is shown <strong>in</strong> figure 4.6. AM1 predicts a very similar sp<strong>in</strong> density<br />

distribution to that from figure 4.3, obta<strong>in</strong>ed by Shimoi and co-workers [2] us<strong>in</strong>g the PPP<br />

Hamiltonian. Even though the extent of the polaron obta<strong>in</strong>ed by PPP and AM1 are similar,<br />

the values of the sp<strong>in</strong> density are much greater with AM1 than with PPP. It should be noted<br />

that because we are us<strong>in</strong>g a restricted open shell wavefunction for AM1, we are unable to<br />

reproduce the small negative sp<strong>in</strong> densities seen with the PPP Hamiltonian. BHandHLYP,<br />

conversely, shows a very large alternation of positive and negative Mulliken sp<strong>in</strong> densities,<br />

but with absolute values which are smaller than those necessary to reproduce the ENDOR<br />

1 The geometry we shows had a maximum force of 4.15 10 −4 a.u., a root-mean-square force of<br />

4.2 10 −5 a.u., a maximum displacement of 2.345 10 −3 a.u. and a root-mean-square displacement of<br />

2.55 10 −4 a.u.<br />

91


peaks seen by Shimoi [2]. Additionally, the BHandHLYP functional shows negative sp<strong>in</strong><br />

densities which are not picked up <strong>in</strong> ENDOR triple resonance (TRIPLE) experiments [3]:<br />

ENDOR is not sensitive to the sign of the sp<strong>in</strong> density on the carbon atoms, whereas<br />

TRIPLE can determ<strong>in</strong>e whether different shifts are caused by sp<strong>in</strong> densities of different<br />

signs. Still, one could argue that s<strong>in</strong>ce a lot of positive and negative sp<strong>in</strong> densities have<br />

similar values, TRIPLE might simply not be able to pick out the fact that the different<br />

ENDOR shifts are caused by sp<strong>in</strong> densities of both signs of different magnitudes.<br />

In conclusion, we f<strong>in</strong>d that AM1 reproduces the PPP sp<strong>in</strong> density distribution but not<br />

its magnitude, while BHandHLYP reproduces the magnitude of the sp<strong>in</strong> densities slightly<br />

better than AM1 but has a very different sp<strong>in</strong> distribution. Neither method is entirely<br />

consistent with the ENDOR data, but even so neither set of results can be excluded on<br />

the basis of the ENDOR data. Evidence of a strongly localised polaron was found <strong>in</strong><br />

reference [2] us<strong>in</strong>g PPP simulations, which are not ab-<strong>in</strong>itio and are designed to create<br />

localised excitation. In my op<strong>in</strong>ion, the ENDOR data does not unambigously demonstrate<br />

the existence of strongly localised charged excitations: all ENDOR suggests is that dif-<br />

ferent sp<strong>in</strong> densities exist <strong>in</strong> the charged radical and all that TRIPLE suggests is that the<br />

two largest sp<strong>in</strong> densities are either both of the same sign or they are both positive and<br />

negative.<br />

4.3 Torsional Defects <strong>in</strong> PPV<br />

In order to model the effect of defects <strong>in</strong> PPV on polaron localisation we consider an<br />

oligomer of PPV of length eight and rotate the cha<strong>in</strong> around the bond connect<strong>in</strong>g the<br />

central v<strong>in</strong>ylene unit to a phenylene unit by a certa<strong>in</strong> torsion angle θ. We then perform<br />

constra<strong>in</strong>ed geometry optimisations for different values of θ and study the variation of<br />

the sp<strong>in</strong> density distribution. This evolution is shown <strong>in</strong> figure 4.7: it can be clearly seen<br />

that <strong>in</strong>creas<strong>in</strong>g the torsion <strong>in</strong> the cha<strong>in</strong> above an angle of 55 ◦ critically affects the sp<strong>in</strong><br />

distrribution. An angle of ≈ 45 ◦ is sufficient to heavily affect this localisation, this is<br />

consistent with studies [9] that show such an angle is sufficient to effectively break the<br />

conjugation <strong>in</strong> a π system. If a charge was travell<strong>in</strong>g down such a polymer, it would have<br />

to hop over such a defect.<br />

92


Mulliken Sp<strong>in</strong> Density<br />

Mulliken Sp<strong>in</strong> Density<br />

0.1<br />

0.05<br />

0<br />

0.2<br />

0.15<br />

0.1<br />

0.05<br />

BHandHLYP<br />

0 10 20 30 40 50 60 70<br />

AM1<br />

0<br />

0 10 20 30 40 50 60 70<br />

Carbon Atom Number<br />

Figure 4.6: Atomic Mulliken sp<strong>in</strong> density for the a cha<strong>in</strong> of 12 units calculated at the<br />

BHandHLYP (top panel) and AM1 (bottom panel) level.<br />

In order to estimate how much thermal energy would be necessary to achieve such<br />

angles we consider second order perturbation theory MP2 calculation of the oligomer of<br />

length two of PPV (stillbene) [10]. The barrier of rotation calculated <strong>in</strong> reference [10] is<br />

just 0.17eV and angles of 45 ◦ can be achieved with just 40 meV; similar values were ob-<br />

ta<strong>in</strong>ed by us<strong>in</strong>g us<strong>in</strong>g DFT. Therefore thermal fluctuations should be able to partition the<br />

polymer cha<strong>in</strong> <strong>in</strong>to conjugated segments with<strong>in</strong> which charge is effectively delocalised.<br />

This underl<strong>in</strong>es the need to be able to dynamically describe the charge transport<strong>in</strong>g unit <strong>in</strong><br />

conjugated polymers. Quasi-coplanar segments would have charge delocalised on them<br />

and transport <strong>in</strong>side these segments would be treated as a coherent quantum mechanical<br />

process [10], whereas transport between such segments would be treated <strong>in</strong> the temper-<br />

ature activated, non-adiabatic, <strong>in</strong>coherent limit described by the Marcus charge transfer<br />

equation.<br />

Another observation is that AM1, which predicts a very sharp localisation <strong>in</strong> space,<br />

also predicts a twisted structure for neutral state of PPV, because it predicts shorter bond<br />

lengths between v<strong>in</strong>ylene and phenylene units which lead to steric <strong>in</strong>teraction between<br />

93


total sp<strong>in</strong> density<br />

0.25<br />

0.2<br />

0.15<br />

0.1<br />

0.05<br />

0<br />

θ=0<br />

θ=18<br />

θ=36<br />

θ=47<br />

θ=55<br />

θ=90<br />

ph1 ph2 ph3 ph4 ph5 ph6 ph7 ph8<br />

r<strong>in</strong>g number<br />

Figure 4.7: Mulliken total sp<strong>in</strong> density <strong>in</strong>tegrated on each phenylene and v<strong>in</strong>ylene unit as<br />

a function of cha<strong>in</strong> torsion for an oligomer of length eight.<br />

the hydrogen groups on these units and torsion of ≈ 22 ◦ between the units. BHandHLYP<br />

predicts angles of ≈ 8 ◦ degrees between phenylene and v<strong>in</strong>ylene units, whereas B3LYP<br />

predicts planar structures. MP2 predicts planar structures for neutral stillbene. Maybe the<br />

fact that DFT results are <strong>in</strong> better agreement with MP2 results suggests that they represent<br />

the electronic structure of PPV better than AM1 does.<br />

4.4 Conclusion<br />

We have shown how even the hybrid density funtional which was previously shown to lead<br />

to most polaron localisation, BHandHLYP, shows polaron localisation energies which<br />

tend to dim<strong>in</strong>ish<strong>in</strong>g values for long cha<strong>in</strong>s. We have argued that the sp<strong>in</strong> density distribu-<br />

tions predicted by this functional are consistent with ENDOR data. We have also shown<br />

how <strong>in</strong>troduc<strong>in</strong>g torsions <strong>in</strong> an oligomer cha<strong>in</strong> tends to localise the charged excitation to<br />

one side of the defect. Whether or not charged radicals are localised and the magnitude<br />

of the polaron localisation energy rema<strong>in</strong>s an open question which depends on the model<br />

94


chemistry chosen. The fact the DFT methods reproduces the geometry of stillbene calcu-<br />

lated at the MP2 level more closely than AM1 leads us to tentatively suggest that polarons<br />

may not be <strong>in</strong>tr<strong>in</strong>sically localised on perfect cha<strong>in</strong>s of PPV, or at least that their localisa-<br />

tion length is very long. This view is supported by calculations of the electronic structure<br />

of crystall<strong>in</strong>e PPV, which show that polarons should not exist <strong>in</strong> perfect crystals of PPV<br />

[11] and therefore the actual existence of polarons <strong>in</strong> crystall<strong>in</strong>e PPV must be expla<strong>in</strong>ed<br />

by the presence of chemical or physical defects <strong>in</strong> the conjugated backbone. Similarly,<br />

the difference between the high <strong>in</strong>tracha<strong>in</strong> charge mobilities measured for s<strong>in</strong>gle cha<strong>in</strong>s<br />

of PPV <strong>in</strong> solution and the lower values obta<strong>in</strong>ed for solid films is attributed to defects<br />

due to chemical or structural irregularities [12], a conclusion supported by tight-b<strong>in</strong>d<strong>in</strong>g<br />

calculations of the effects of geometric defects on <strong>in</strong>tracha<strong>in</strong> mobilities <strong>in</strong> s<strong>in</strong>gle polymer<br />

cha<strong>in</strong>s [10]. Recent measurements of metallic conductivity <strong>in</strong> highly ordered polyanil<strong>in</strong>e<br />

[13] re<strong>in</strong>forces the hypothesis that <strong>in</strong> perfect polymers, polaron localisation is not a strong<br />

effect.<br />

The relevance of these observations to charge transport <strong>in</strong> polymers is gleaned from<br />

the realisation that the k<strong>in</strong>ds of torsional defects sufficient to break conjugation are easily<br />

accessible at room temperature. Even if polarons were not <strong>in</strong>tr<strong>in</strong>sic properties of perfect<br />

PPV cha<strong>in</strong>s, the picture of a polymer as a series of segments that charge is delocalised<br />

would still apply. <strong>Transport</strong> events between such segments should be treated <strong>in</strong> the tem-<br />

perature activated regime, whereas with<strong>in</strong> the coplanar segments values of the transfer<br />

<strong>in</strong>tegral would be much larger than the reorganisation energy and coherent conduction<br />

could occur.<br />

95


Bibliography<br />

[1] S. Kuroda, T. Noguchi and T. Ohnishi, Electron Nuclear Double Resonance Obser-<br />

vation of π -Electron defect States <strong>in</strong> Undoped poly(paraphenylenev<strong>in</strong>ylene), Physics<br />

Review Letters 72, 286 (1994)<br />

[2] Y. Shimoi, S. Abe, S. Kuroda and K. Murata, Polarons and their ENDOR spectra <strong>in</strong><br />

poly p-phenylene v<strong>in</strong>ylene, Solid State Communications 95, 137 (1995)<br />

[3] S. Kuroda, Y. Shimoi, S. Abe, T. Noguchi and T. Ohnishi, Electron Nuclear Double<br />

Resonance Spectra of Polarons <strong>in</strong> poly (Phenylene V<strong>in</strong>ylene) , Journal of the Physical<br />

Society of Japan 67, 3936 (1998)<br />

[4] H. Muto and M. Iwasaki, ENDOR studies of the superhyperf<strong>in</strong>e coupl<strong>in</strong>gs of<br />

hydrogen-bonded protons. I. Carboxyl radical anions <strong>in</strong> irradiated L-anal<strong>in</strong>e s<strong>in</strong>gle<br />

crystals, Journal of Chemical Physics, 59, 4821 (1973)<br />

[5] V.M. Gesk<strong>in</strong>, A. Dkhissi and J.L. Brédas, Oligothioohene radical cations: polaron<br />

structure <strong>in</strong> hybrid DFT and MP2calculations , Internation Journal of Quantum Chem-<br />

istry 91, 350 (2003)<br />

[6] V.M. Gesk<strong>in</strong>, F.C Grozema, L.D.A. Siebbeles, D. Beljonne, J.L. Brédas and J.<br />

Cornil, Impact of computational method on the geometric and electronic properties<br />

of oligo(phenylene) v<strong>in</strong>ylene radical cations, Journal of Physical Chemistry B 109,<br />

20237 (2005)<br />

[7] R.S. Mulliken, Electronic Population Analysis on LCAO-MO Molecular Wave Func-<br />

tions, Journal of Chemical Physics 23, 1833 (1955)<br />

96


[8] V.M. Gesk<strong>in</strong>, J. Cornil and J.L. Brédas, Comment on ’Polaron formation and symme-<br />

try break<strong>in</strong>g’ by L. Zuppiroli at al[Chem Phys Lett. 374 (2003) 7], Chemical Physics<br />

Letters 403, 228 (2005)<br />

[9] J.L. Brédas, G.B. Street, B. Themans and J.M. Andre, Organic Polymers Based<br />

on aromatic r<strong>in</strong>gs (polyparaphenylen, polypyrrole, polythiophene) - evolution of the<br />

elctronic-properties as a function of the torsion angle between adjacent r<strong>in</strong>gs., Journal<br />

of Chemical Physics 83, 1323, 1985<br />

[10] F.C. Grozema, P.T. van Duijnen, Y.A. Berl<strong>in</strong>, M.A. Ratner and L.D.A. Siebbeles,<br />

Intramolecular charge transport along isolated cha<strong>in</strong>s of conjugated polymers: effect<br />

of torsional disorder and polymerisation defects, Journal of Physical Chemistry B 106,<br />

7791 (2002)<br />

[11] P.G. da Costa, R.G. Dandrea and E.M. Conwell, First-pr<strong>in</strong>ciples calculation of the<br />

three-dimensional band structure of poly(phenylene v<strong>in</strong>ylene), Physical Review B 47,<br />

1800, (1993)<br />

[12] R.J.O. Hoofman, M.P. de Haas, L.D.A. Siebbeles and J.M. Warman, Highly mobile<br />

electrons and holes on isolated cha<strong>in</strong>s of the semiconduct<strong>in</strong>g polymer poly(phenylene<br />

v<strong>in</strong>ylene), Nature 392, 54, (1998)<br />

[13] K. Lee, S. Cho, S.H. Park, A.J. Heeger, C.W. Lee and S.H. Lee, Metallic transport<br />

<strong>in</strong> polyanil<strong>in</strong>e, Nature 441, 65 (2006).<br />

97


Chapter 5<br />

Ambipolar <strong>Transport</strong> <strong>in</strong> PCBM<br />

In this chapter we present calculations of transfer <strong>in</strong>tegrals and reorganisation energies for<br />

electrons and holes <strong>in</strong> a methanofullerene. The aim is to support experimental observation<br />

of ambipolar transport <strong>in</strong> this material and to provide order of magnitude estimates for the<br />

parameters which should be used <strong>in</strong> lattice simulations of charge transport <strong>in</strong> polymer<br />

blends doped with methanofullerene.<br />

5.1 Introduction<br />

[6,6]-phenyle-C61 butyric acid methyl ester (PCBM) is a popular choice as electron ac-<br />

ceptor for organic photo-voltaic devices, when used <strong>in</strong> comb<strong>in</strong>ation with a donor polymer<br />

such as the PPV derivative poly[2-methoxy-5-(3,7-dimethyloctyloxy)-1,4-phenylenev<strong>in</strong>ylene]<br />

(MDMO-PPV). A curious property of such cells is that they achieve maximum efficiency<br />

at a relatively high concentration of PCBM of 80% by weight. This is especially confus<strong>in</strong>g<br />

as one would th<strong>in</strong>k that a higher concentration of MDMO-PPV would be more desirable<br />

as MDMO-PPV is the light absorb<strong>in</strong>g material. The high concentration of PCBM needed<br />

for optimal devices has been expla<strong>in</strong>ed by the observation that as the PCBM concentra-<br />

tion is <strong>in</strong>creased, both electron and hole transport become faster[1]. This expla<strong>in</strong>s the <strong>in</strong>-<br />

creased efficiency of photovoltaic cells but leaves a major question open: how can PCBM,<br />

traditionally viewed as an electron conductor, enhance hole mobility <strong>in</strong> MDMO-PPV?<br />

Recent time of flight measurements of PCBM doped polystyrene have shown that<br />

PCBM has similar electron and hole mobilities [2] and that PCBM acts as an ambipolar<br />

98


transporter. Our colleagues Prof. Nelson and Dr. Chatten wanted to run simulations of<br />

hole transport on a cubic lattice partially occupied by PCBM and MDMO-PPV, with the<br />

aim of expla<strong>in</strong><strong>in</strong>g the experimentally observed PCBM concentration dependence of mo-<br />

bilty by allow<strong>in</strong>g PCBM occupied sites to participate <strong>in</strong> hole transport. In order to help<br />

parametrise such simulations, <strong>in</strong> this chaper we calculate the Marcus charge trasfer equa-<br />

tion parameters for electrons and holes. We calculate the reorganisation energy for radical<br />

cations and anions and compare them. Pairs of PCBM molecules are extracted from the<br />

crystal structure of PCBM [3] and their transfer <strong>in</strong>tegral for both electrons and holes are<br />

compared. The crystal structure used is that of crystals grown from chlorobenzene solu-<br />

tions, this structure is more tightly packed than the one grown from ortho-dichlorobenzene<br />

solutions and is therefore thought to be responsible for the higher efficiency of devices<br />

fabricated from that solvent [3]. In the Monte Carlo simulations of charge transport, <strong>in</strong>-<br />

clud<strong>in</strong>g hopp<strong>in</strong>g to next-nearest neighbours proved necessary. We therefore calculate the<br />

distance dependence of the transfer <strong>in</strong>tegral for PCBM, to provide an order of magnitude<br />

estimate for the natural length of decay of the transfer <strong>in</strong>tegral with distance.<br />

PCBM has a similar electronic structure to fullerene (C60). C60 belongs to the icosa-<br />

hedral po<strong>in</strong>t group and because of this its HOMO is fivefold degenerate and its LUMO<br />

threefold degenerate [4]. It has also been demonstrated via cyclic voltammetry that C60<br />

has multiple reduction potentials [5]. PCBM, because of its sidecha<strong>in</strong>, does not pos-<br />

sess icosahedral symmetry and therefore its frontier orbitals are quasi degenerate. At the<br />

ZINDO level the separation between the top two HOMOs is 64 meV and the separation<br />

between the bottom two LUMOs is 132 meV. Even though these energy differences are<br />

not very large we feel that, s<strong>in</strong>ce we are only <strong>in</strong>terested <strong>in</strong> order of magnitude estimates<br />

of the magnitudes of hole and electron transfer rates, it is sufficient to look at the reor-<br />

ganisation energies for s<strong>in</strong>gle ionisation and to the transfer <strong>in</strong>tegrals for the HOMO and<br />

LUMO only of the isolated molecule only.<br />

5.2 Transfer Integrals<br />

A cluster of 20 PCBM molecules is generated from the crystal structure obta<strong>in</strong>ed <strong>in</strong> ref-<br />

erence [3]. We generate all possible pairs with<strong>in</strong> this cluster and calculate their transfer<br />

99


Figure 5.1: The seven unique pairs with greatest hole transfer <strong>in</strong>tegral extracted from<br />

a cluster of twenty molecules of PCBM from its crystal structure grown <strong>in</strong><br />

chlorobenzene.<br />

pair n Jh / meV Je /meV dcoc / Å<br />

a 4 26 5 6.78<br />

b 4 23 32 6.76<br />

c 4 20 5 6.93<br />

d 4 16 9 6.86<br />

e 2 14 26 6.87<br />

f 8 5 11 6.71<br />

g 1 4 3 6.71<br />

Table 5.1: Table of the transfer <strong>in</strong>tegral for holes Jh and for electrons Je. n is the number<br />

of such pairs <strong>in</strong> a cluster of 20 molecules. dcoc is the separation between the<br />

centres of the fullerene cages for the pair of molecules. The labels a-g label the<br />

pairs accord<strong>in</strong>g to figure 5.1<br />

<strong>in</strong>tegrals us<strong>in</strong>g the projective method described <strong>in</strong> section 3.1 applied to ZINDO Hamil-<br />

tonian. All calculations are performed with Gaussian 03. We consider the seven unique<br />

pairs with the greatest magnitude hole transfer <strong>in</strong>tegrals; these seven unique pairs also<br />

have the highest electron transfer <strong>in</strong>tegrals. These pairs are shown <strong>in</strong> figure 5.1 and their<br />

transfer <strong>in</strong>tegral values are shown <strong>in</strong> table 5.1. The average value of the transfer <strong>in</strong>tegrals<br />

for these pairs of molecules was found to be 15meV for holes and 14meV for electrons.<br />

These two values are very similar therefore electron and hole transfer rates should be<br />

similar if the reorganisation energies for electrons and holes are similar.<br />

In order to allow Monte Carlo simulations to <strong>in</strong>clude both nearest and next nearest<br />

100


neighbour hops, it is necessary to know the distance dependence of the transfer <strong>in</strong>tegral.<br />

In order to calculate this we consider pairs of PCBM <strong>in</strong> the relative orientation b) from<br />

figure 5.1, displaced them by a certa<strong>in</strong> distance from their crystal centre of mass dis-<br />

tance (keep<strong>in</strong>g the angles describ<strong>in</strong>g their position and orientation constant) and calculate<br />

transfer <strong>in</strong>tegrals us<strong>in</strong>g the projective method applied to the ZINDO Hamiltonian. This<br />

particular pair was chosen because it has similar values of the hole and electron transfer<br />

<strong>in</strong>tegral. The results of these calculation are shown <strong>in</strong> figure 5.2. For displacements from<br />

the crystal separation greater than 2 Å, both transfer <strong>in</strong>tegrals fall exponentially with a<br />

natural length scale of 0.5 Å and 0.53 Å for electrons and hole respectively. At shorter<br />

displacements the electron transfer <strong>in</strong>tegral falls far more slowly, with a decay length of<br />

0.70 Å . The difference <strong>in</strong> decay lengths between electron and hole transfer <strong>in</strong>tegrals for<br />

short displacement can be expla<strong>in</strong>ed by notic<strong>in</strong>g that PCBM is not a planar molecule. To<br />

illustrate how this might affect the value of the natural decay length of transfer <strong>in</strong>tegrals<br />

let us assume that two particular faces of PCBM are closest together. If the two molecules<br />

are close, as the separation between molecules is <strong>in</strong>creased the distance and overlap be-<br />

tween atomic orbitals on those faces becomes smaller; but maybe some of the other faces<br />

will f<strong>in</strong>d themselves <strong>in</strong> such positions as to have better overlap. The net result of this effect<br />

is that the decay would be slightly less steep at short distances, depend<strong>in</strong>g on which faces<br />

of PCBM are closest and whether the molecular orbitals are <strong>in</strong>volved. We have therefore<br />

established a lower bound for the natural length of decay of the transfer <strong>in</strong>tegrals of ap-<br />

proximataly 0.5 Å; an exact value for short distances will be more difficult to ascerta<strong>in</strong><br />

and would depend heavily on the relative orientation of the molecules <strong>in</strong>volved.<br />

5.3 Reorganisation Energy<br />

We have already stated that <strong>in</strong> order to calculate the reorganisation energy it is necessary<br />

to compute the geometries of both the charged and neutral molecules. We will follow<br />

Sakanoue [6] and use both semiempirical and DFT methods to compute these geometries.<br />

The semiempirical method we pick is AM1 and the DFT functional is the hybrid function<br />

B3LYP. The values of λ1, λ2 and λ <strong>in</strong> the methods are tabulated <strong>in</strong> tables 5.2 and 5.3.<br />

It can be clearly seen that both methods predict similar reorganisation energies for<br />

101


|J| / eV<br />

0.01<br />

0.001<br />

0.0001<br />

exp(-r/0.7)<br />

exp(-r/0.53)<br />

exp(-r/0.5)<br />

|J| electrons<br />

|J| holes<br />

1e-05<br />

0 1 2 3<br />

Displacement from crystal position r / A<br />

Figure 5.2: Distance dependence of the electron (black circles) and hole (red squares)<br />

transfer <strong>in</strong>tegrals as a function of displacement from the crystal structure separation.<br />

The l<strong>in</strong>es show exponential fits with natural lengths of respectively<br />

0.78 Å and 0.56 Å.<br />

method λ1 λ2 λ<br />

B3LYP 0.062 0.059 0.121<br />

AM1 0.113 0.93 0.206<br />

Table 5.2: Reorganisation energies for anion.<br />

method λ1 λ2 λ<br />

B3LYP 0.064 0.060 0.124<br />

AM1 0.086 0.082 0.168<br />

Table 5.3: Reorganisation energies for cation.<br />

102


positive or negative charge transport, but that AM1 predicts larger values of the reorgan-<br />

isation energy compared to B3LYP. A good illustration of the relation between charge<br />

localisation and lattice distortion can be obta<strong>in</strong>ed by compar<strong>in</strong>g the change <strong>in</strong> geome-<br />

try of the molecule that occurs upon charg<strong>in</strong>g to the sp<strong>in</strong> density distribution. In both the<br />

cases of positive and negative charg<strong>in</strong>g, some bond lengths change up to 1-2%. In order to<br />

show the correlation between sp<strong>in</strong> density and bond length deformation, <strong>in</strong> figure 5.3 we<br />

show a scatter plot of the percentage change <strong>in</strong> absolute value of a bond length versus the<br />

average Mulliken sp<strong>in</strong> density on the two atoms <strong>in</strong>volved <strong>in</strong> the bond. This figure clearly<br />

shows that there is a very strong correlation <strong>in</strong>deed between sp<strong>in</strong> density and bond length<br />

deformation; for an anion the Pearson’s product moment correlation 95 % <strong>in</strong>terval is 0.76<br />

to 0.87 and for cation it is 0.63 to 0.80. In order to show where the bond deformation<br />

occurs and the sp<strong>in</strong> density is localised, <strong>in</strong> figure 5.4 we plot the total sp<strong>in</strong> isosurface for<br />

an anion radical (left) and a cation (right) radical. It can be seen that <strong>in</strong> the anion radical<br />

the sp<strong>in</strong> density is ma<strong>in</strong>ly concentrated around the equator of the molecule, whereas <strong>in</strong><br />

the cation radical the sp<strong>in</strong> density is ma<strong>in</strong>ly located at the top pole, near the sp 3 carbon<br />

which l<strong>in</strong>ks the fullerene cage to the side cha<strong>in</strong>. When the bonds with the highest sp<strong>in</strong><br />

density rearrange <strong>in</strong> a radical anion, the bonds which cross the equator of the molecule<br />

are elongated and the molecule becomes more ”rugby ball” shaped. In a radical cation,<br />

the bonds which rearrange the most are near the sp 3 carbon, result<strong>in</strong>g <strong>in</strong> an open<strong>in</strong>g of<br />

the fullerene cage.<br />

The conclusions from these calculations are that <strong>in</strong> PCBM the reorganisation energies<br />

for both cations and anions are very similar and that therefore the magnitude of charge<br />

transfer rates should be similar for both positive and negative charges. This expla<strong>in</strong>s the<br />

similarity of zero field electron and hole mobilities for PCBM doped polystyrene, shown<br />

<strong>in</strong> figure 5.5.<br />

5.4 Monte Carlo Simulations<br />

In this section we describe the results of the Monte Carlo simulations run by our col-<br />

leagues Dr. Chatten and Prof. Nelson on charge transport <strong>in</strong> blends of MDMO-PPV:PCBM.<br />

The aim of this exercise is to fit the experimental temperature dependence of hole trans-<br />

103


Absolute value of the bond length distorsion / %<br />

2<br />

1.5<br />

1<br />

0.5<br />

anion (Pearson’s correlation=0.76-0.87)<br />

cation (Pearson’s correlation=0.63-0.80)<br />

0<br />

-0.02 0 0.02 0.04 0.06 0.08<br />

Average Mulliken sp<strong>in</strong> density<br />

Figure 5.3: Scatter plot of the average sp<strong>in</strong> density for each bonded pair of atoms <strong>in</strong> a<br />

cation (circles) or <strong>in</strong> an anion (crosses) versus the percentage change <strong>in</strong> bond<br />

length upon charg<strong>in</strong>g.<br />

Figure 5.4: Sp<strong>in</strong> density for cation and anion radicals of PCBM. The left panel shows a<br />

front view of the molecules, the right panel a side view. With<strong>in</strong> each panel<br />

the molecule to the left is the anion radical and the molecule to the right is the<br />

cation radical.<br />

104


zero field mobility / cm 2 V -1 s -1<br />

0.0001<br />

1e-05<br />

1e-06<br />

hole mobility<br />

electron mobility<br />

1e-07<br />

20 25 30 35<br />

PCBM concentration / %<br />

40 45 50<br />

Figure 5.5: Zero field mobility for electrons (squares) and holes (circles) <strong>in</strong> PCBM doped<br />

polystyrene as a function of PCBM concentration.<br />

port <strong>in</strong> prist<strong>in</strong>e MDMO-PPV and <strong>in</strong> blends of MDMO-PPV:PCBM us<strong>in</strong>g the same pa-<br />

rameters. The model used is very similar to that used to describe hole transport at low<br />

PCBM concentration [7]. Its ma<strong>in</strong> features are: use of the Marcus form for sett<strong>in</strong>g charge<br />

hopp<strong>in</strong>g rates, use of a distribution of energies based on a Gaussian with an exponential<br />

distribution of traps and use of a cubic lattice. Holes are allowed to be transported <strong>in</strong> the<br />

polymer donor material, <strong>in</strong> the fullerene and between the fullerene and the polymer; hole<br />

hops from the donor to the acceptor are decelerated by <strong>in</strong>clusion of an additional energy<br />

difference ∆ between the two materials. The lattice is allowed to be partially occupied<br />

by MDMO-PPV sites and PCBM sites. The density of states for hole transport sites are<br />

picked from a Gaussian of width 75 meV for PCBM [8] and 25 meV for MDMO-PPV;<br />

for MDMO-PPV 2% of site energies are drawn from an exponential trap distribution with<br />

characteristic energy 65 meV [7]. The reorganisation energy for PCBM was 0.5 eV, or<br />

roughly twice the <strong>in</strong>ternal sphere reorganisation energy we have calculated. This value<br />

can be justified by argu<strong>in</strong>g the importance of the outer sphere reorganisation energy. The<br />

value of the reorganisation energy used to model hole transport <strong>in</strong> MDMO-PPV was also<br />

105


0.5 eV, follow<strong>in</strong>g reference [9], although a value of 0.25 eV could also reproduce the ex-<br />

perimental data. The transfer <strong>in</strong>tegral for nearest neighbours <strong>in</strong> PCBM which fitted the<br />

experimental mobility is 28 meV, which correlates rather well with the values we have<br />

calculated.<br />

The lattice used is cubic with a lattice constant of 1nm, whilst hops with<strong>in</strong> a 3nm<br />

radius are allowed. In order to calculate the transfer <strong>in</strong>tegrals to next nearest neighbours,<br />

we assume that the transfer <strong>in</strong>tegral falls exponentially with a natural length of 2 Å; this<br />

length is somewhat longer than the one we calculated. The localisation length is very<br />

important <strong>in</strong> simulations on partially occupied lattices, as it will have a very large ef-<br />

fect on the percolation threshold and on the formation of efficient pathways for charge<br />

transport. The fact that such a large natural length is required to reproduce experimental<br />

data might suggest that cubic lattices do not represent the pack<strong>in</strong>g of PCBM very well:<br />

next nearest neighbours are actually closer <strong>in</strong> a real solid than a cubic lattice is capable<br />

of represent<strong>in</strong>g. Also, at the relatively short distance variations between nearest and next<br />

nearest neighbours the exponential dependence of transfer <strong>in</strong>tegrals on distance might not<br />

hold, as orientational effects become stronger. To show this we plot the transfer <strong>in</strong>tegral<br />

for pairs of molecules from the cluster we ran the calculations <strong>in</strong> table 5.1 from, and plot<br />

them aga<strong>in</strong>st centre of mass distance between each pair. We also plot them aga<strong>in</strong>st the<br />

distance between the centre of the C60 cages, as these are the parts of the molecules that<br />

conta<strong>in</strong> most of the HOMOs and LUMOs and therefore it is the proximity of these areas<br />

which is more relevant to transfer <strong>in</strong>tegrals. This plot is shown <strong>in</strong> figure 5.6. An expo-<br />

nential fit is not terribly good for either the dependence on centre of mass distance or for<br />

distance between the centres of the C60 cages, but it is somewhat better <strong>in</strong> the latter case.<br />

This is easily understood by look<strong>in</strong>g at <strong>in</strong>set c) <strong>in</strong> figure 5.1. In orientations of that k<strong>in</strong>d,<br />

with the side cha<strong>in</strong>s opposite the closest po<strong>in</strong>t of the two cages, look<strong>in</strong>g at the distance be-<br />

tween the centres of mass overestimates separation because of the effect of the side cha<strong>in</strong><br />

<strong>in</strong> determ<strong>in</strong><strong>in</strong>g the centre of mass. However, the rate of decrease with distance of transfer<br />

<strong>in</strong>tegral is certa<strong>in</strong>ly not as fast as with a natural length of 0.5 Å, the transfer <strong>in</strong>tegral falls<br />

two orders of magnitude over approximately 3 Å , which would correspond to a natural<br />

length of approximately 1.5 Å.<br />

106


J /eV<br />

J / eV<br />

0.01<br />

0.001<br />

0.0001<br />

0.01<br />

0.001<br />

0.0001<br />

distance between centres of mass<br />

8 9 10 11 12 13 14<br />

6 7 8 9 10<br />

distance between c60 cage centres<br />

Figure 5.6: Distance dependance for the transfer <strong>in</strong>tegral for neighbour and nearest neighbour<br />

hopp<strong>in</strong>gs, extracted from the cluster from the crystal structure of PCBM<br />

discussed <strong>in</strong> the text. The transfer <strong>in</strong>tegral is plotted both as a function of<br />

the distance between centres of mass of the PCBM molecules (above) and<br />

as a function of the distance between the centres of the C60 cages (below).<br />

This is done because the C60 cages are the electronically active parts of the<br />

molecules.<br />

£��<br />

� ¡¢<br />

�<br />

�<br />

�<br />

£© � ¡¢<br />

�<br />

�<br />

£¨�<br />

¡¢<br />

�<br />

�<br />

�<br />

£§ � ¡¢<br />

�<br />

£¦ � � ¡¢<br />

�<br />

�<br />

£¥ � ¡¢<br />

�<br />

�<br />

�<br />

�<br />

�<br />

�<br />

�<br />

�<br />

¡¢ £ ¡¢ £¤ ¡¢ £� ¡¢ £ �� £¡¢ £��� ¨¡¢ £�©� £¡¢ £�©�<br />

¡¢<br />

Figure 5.7: Experimental (full symbols) and modelled (empty symbols) temperature dependence<br />

for prist<strong>in</strong>e MDMO-PPV (triangles) 1:1 MDMO-PPV:PCBM (circles)<br />

and 1:2 MDMO-PPV:PCBM (squares).<br />

107<br />

¨¡¢<br />

£�¨�<br />

£¡¢<br />

£�¨�<br />

¨¡¢<br />

£�§�<br />

����<br />

���<br />

����<br />

��<br />

��<br />

����<br />

�����<br />

���<br />

����<br />

�����<br />

����<br />

������<br />

�<br />

����<br />

�����<br />

����<br />

������<br />

�<br />

��<br />

��<br />

��<br />

��<br />

����<br />

�����<br />

���<br />

��<br />

��<br />

�����<br />

����<br />

������<br />

�<br />

��<br />

��<br />

�����<br />

����<br />

������<br />

�<br />

£¡¢<br />

£�


The most crucial parameter <strong>in</strong> determ<strong>in</strong><strong>in</strong>g whether the simulation is capable of re-<br />

produc<strong>in</strong>g experimental observation is the offset <strong>in</strong> site energies between PCBM and<br />

MDMO-PPV ∆. Nom<strong>in</strong>ally, the difference between the HOMO of PCBM and the HOMO<br />

of MDMO-PPV is 0.9 eV. If simulations are run with such a high offset, charges cannot<br />

hop from the polymer to the fullerene and hole mobilities are decreased as the PCBM<br />

concentration is <strong>in</strong>creased. If, however, a far smaller value for ∆ is picked of between 0.1<br />

and 0.2 eV, the experimental temperature dependence of the zero field mobility can be<br />

reproduced, as figure 5.7 shows. Values of ∆ as large as 0.25 eV could also be made con-<br />

sistent with the experimental data. The temperature dependence is shown because hops<br />

between MDMO-PPV and PCBM are temperature activated: if our model of transport<br />

was correct and if the barrier to hopp<strong>in</strong>g between the materials ∆ was large, we would<br />

predict that there would be a very strong temperature dependence of the mobility, but this<br />

is not observed experimentally. The conclusion from this <strong>in</strong>vestigation is that <strong>in</strong> order for<br />

ambipolar transport <strong>in</strong> PCBM to expla<strong>in</strong> the <strong>in</strong>creased hole mobility <strong>in</strong> PCBM:MDMO-<br />

PPV blends, the offset between the hole transport levels must be much smaller than one<br />

would imag<strong>in</strong>e from look<strong>in</strong>g at the PCBM HOMO. Such small offset could be justified<br />

by <strong>in</strong>vok<strong>in</strong>g the formation of an <strong>in</strong>terface state, with a much lower ionisation potential<br />

compared to PCBM.<br />

5.5 Conclusion<br />

In this chapter we have shown results of calculations of transfer <strong>in</strong>tegrals and reorganisa-<br />

tion energies for electron and hole transport. These calculations imply that hole transport<br />

is plausible <strong>in</strong> PCBM. Simulation of hole transport <strong>in</strong> MDMO-PPV:PCBM blends show<br />

that experimental mobility data can be consistent with ambipolar transport <strong>in</strong> PCBM only<br />

if the energy offset between the hole transport<strong>in</strong>g states on MDMO-PPV and on PCBM<br />

is between 0.1 and 0.25 eV.<br />

108


Bibliography<br />

[1] V.D. Mihailetchi, B. de Boer, C. Melzer, L.J.A. Koster and P.W.M. Blom, Electron<br />

and hole transport <strong>in</strong> poly(para-phenylenev<strong>in</strong>ylene):methanofullerene bulk heterojunc-<br />

tion solar cells, proceed<strong>in</strong>g of the SPIE conference (2004)<br />

[2] S.M. Tuladhar, D. Poplavskyy, S.A. Choulis, J.R. Durrant, D.D.C. Bradley<br />

and J. Nelson, Ambipolar charge transport <strong>in</strong> films of methanofullerene and<br />

poly(phenylenev<strong>in</strong>ylene) blends, Advanced functional Materials, 15, 1171 (2005)<br />

[3] M.T. Rispens, A. Meetsma, R. Rittberger, C.J. Brabec, N.S. Sarificiftci and J.C. Hum-<br />

melen, Influence of the solvent on the crystal structure of PCBM and the efficiency of<br />

MDMO-PPV:PCBM ’plastic’ solar cells Chemical Communications, 2006 (2003)<br />

[4] P.J. Benn<strong>in</strong>g, J.L. Mart<strong>in</strong>s, J.H. Weaver, L.P.F. Chibante and R.E. Smalley, Electronic<br />

structure of C60: Insulat<strong>in</strong>g Metallic and Superconduct<strong>in</strong>g character, Science, 252,<br />

1417 (1991)<br />

[5] P.M. Allemand, A. Koch, F. Wudl, Y. Rub<strong>in</strong>, F. Diederich, A.M. Alvarez, S.J. Anz<br />

and R.L. Whetten, Two different fullerenes have the same cyclic voltametry, Journal of<br />

the American Chemical Society, 113, 1050 (1991)<br />

[6] K. Sakanoue, M. Motoda, M. Sugimoto and S. Sakaki, A Molecular Orbital Study<br />

on the Hole <strong>Transport</strong> Property of Organic Am<strong>in</strong>e Compounds, Journal of Chemical<br />

Physics A, 103, 551 (1999)<br />

[7] A.J. Chatten, S.M. Tuladhar, S.A. Choulis, D.D.C. Bradley and J. Nelson, Monte<br />

Carlo modell<strong>in</strong>g of hole transport <strong>in</strong> MDMO-PPV:PCBM blends, Journal of Material<br />

Science, 40, 1393 (2005)<br />

109


[8] V.D. Mihailetchi, J.K.J. van Duren, P.W.M. Blom, J.C. Hummelen, R.A.J. Janssen,<br />

J.M. Kroon, M.T. Rispens, W.J.H. Verhees and M.M. Wienk, Electron <strong>Transport</strong> <strong>in</strong> a<br />

Methanofullerene, Advanced Functional Materials, 13, 43<br />

[9] A.J. Chatten, S.M. Tuladhar, D.D.C. Bradley and J. Nelson, Modell<strong>in</strong>g charge trans-<br />

port <strong>in</strong> gated polymers for solar cell applications, proceed<strong>in</strong>gs of the Photovoltaic<br />

Solar Energy Conference and Exhibition (2004)<br />

110


Chapter 6<br />

<strong>Charge</strong> <strong>Transport</strong> <strong>Parameters</strong> for<br />

Randomly Oriented Pairs of Dialkoxy<br />

Poly-paraphenylenev<strong>in</strong>ylene and<br />

Triarylam<strong>in</strong>e Derivatives<br />

This chapter reports calculations of J and ∆E on huge samples of pairs of molecules <strong>in</strong><br />

random relative orientations and positions. We use the fast methods of calculat<strong>in</strong>g these<br />

parameters described <strong>in</strong> chapter three, namely molecular orbital overlap based calcula-<br />

tions for J and distributed multipole electrostatic calculations for ∆E. We will compare<br />

the distributions of J and ∆E from these calculations to the trends <strong>in</strong> GDM parameters for<br />

two different series of materials: dialkoxy PPV with different length side cha<strong>in</strong>s and spiro<br />

l<strong>in</strong>ked triarylam<strong>in</strong>e derivatives with different side groups. Hole transport <strong>in</strong> these two<br />

groups of materials has been studied experimentally at Imperial College with the explicit<br />

<strong>in</strong>tention of study<strong>in</strong>g the relationship between charge mobility and chemical structure.<br />

Our aim is to provide an explanation <strong>in</strong> terms of microscopic properties of the physical<br />

orig<strong>in</strong> of the GDM parameters. As mentioned <strong>in</strong> chapter two, the GDM is a model which<br />

provides an expression for the temperature and field dependence of mobility to equation<br />

2.40, namely:<br />

111


⎧<br />

⎪⎨ µ0e<br />

µ =<br />

⎪⎩ µ0e<br />

2 −( 3kT σ)2 σ C[( e kT )2−Σ2 ] √ F , Σ > 1.5<br />

2 −( 3kT σ)2 σ C[( e kT )2−2.25] √ F , Σ < 1.5<br />

(6.1)<br />

this equation takes four parameters: µ0, σ, Σ and C. The GDM is developed from an<br />

idealised model of charge transport on a cubic lattice, where the form for the transfer rates<br />

is the Miller-Abrahams rate equation (equation 2.8). With<strong>in</strong> this model the four parame-<br />

ters σ, Σ, µ0 and C have very specific mean<strong>in</strong>gs: the energetic disorder σ represents the<br />

width of the distribution of site energies, the positional disorder Σ represents the width of<br />

the distribution of the logarithm of the transfer <strong>in</strong>tegral, µ0 scales the absolute value of the<br />

mobility and C is related to the lattice constant. When applied to experimental data, how-<br />

ever, it is not always possible to relate these parameters to the microscopic properties they<br />

represent <strong>in</strong> the model. Strictly speak<strong>in</strong>g, the GDM parameters are really fitt<strong>in</strong>g parame-<br />

ters for the Poole-Frenkel like dependence of mobility on electric field and temperature:<br />

a more positive field dependence and temperature dependence of mobility corresponds to<br />

greater energetic disorder and a more negative field dependence but unchanged tempera-<br />

ture dependence of mobility corresponds to a greater positional disorder. In this chapter<br />

we calculate distributions of site energy differences and of transfer <strong>in</strong>tegrals and compare<br />

the results to GDM parameters <strong>in</strong> order to make a comparison of the GDM with chemical<br />

structure grounded on physical computation.<br />

6.1 Dialkoxy PPV Series<br />

6.1.1 Introduction<br />

This part of the study is aimed at expla<strong>in</strong><strong>in</strong>g the trends <strong>in</strong> GDM hole transport parameters<br />

for a series of dialkoxy polyparaphenylenev<strong>in</strong>ylene (PPV) derivatives with different length<br />

side cha<strong>in</strong>s. Figure 6.1 shows the three derivatives considered <strong>in</strong> this study: they all have<br />

the same conjugated backbone and all are symmetrically substituted with alkoxy side<br />

cha<strong>in</strong>s which vary only by their lengths.<br />

In polymers with a conjugated backbone such as PPV, the <strong>in</strong>fluence of side cha<strong>in</strong>s on<br />

electronic properties is often thought of as aris<strong>in</strong>g from their role <strong>in</strong> determ<strong>in</strong><strong>in</strong>g cha<strong>in</strong><br />

112


Figure 6.1: Chemical formula of the three dialkoxy PPV derivatives studied <strong>in</strong> this chapter.<br />

a) is di-MethoxyPPV(dMeOPPV), b) is di-hexyloxyPPV (dHeOPPV) and<br />

f<strong>in</strong>ally c) is di-decyloxyPPV (dDeOPPV).<br />

pack<strong>in</strong>g, which <strong>in</strong> turn affects the degree of coplanarity of the backbone. This effect is<br />

certa<strong>in</strong>ly observed <strong>in</strong> regio-regular polythiophenes with alkyl side groups [1], where pro-<br />

cess<strong>in</strong>g methods which lead to high charge mobility also lead to the formation of ordered<br />

micro-crystallites. X-ray diffraction studies [2] proved that these crystallites are formed<br />

thanks to <strong>in</strong>terdigitation of the alkyl side cha<strong>in</strong>s, which aid the formation of long, coplanar<br />

segments of the backbone and to supramolecular arrangements of the backbones <strong>in</strong> lamel-<br />

lae. Martens et al. [3] measure hole mobilities for symmetrically and asymmetrically sub-<br />

stituted PPV polymers and extract the correspond<strong>in</strong>g parameters for the correlated GDM:<br />

symmetrically substituted PPV are shown to have lower energetic disorder and greater<br />

mobility, which were both attributed to the hypothesis that symmetrical side cha<strong>in</strong>s aid<br />

order<strong>in</strong>g of the backbones. Kemer<strong>in</strong>k and co-workers [4] also observe that asymmetric<br />

substitution leads to stronger <strong>in</strong>tracha<strong>in</strong> <strong>in</strong>teractions and <strong>in</strong> some cases <strong>in</strong>tracha<strong>in</strong> aggre-<br />

gation <strong>in</strong> PPV polymers, whereas symmetric subsitution leads to <strong>in</strong>tercha<strong>in</strong> aggregation<br />

and greater mobilities.<br />

In this study we are not compar<strong>in</strong>g the effect of symmetric versus asymmetric substi-<br />

tution on backbone conformation, but rather the <strong>in</strong>fluence of side cha<strong>in</strong> length on charge<br />

transport parameters. We use the simplest possible <strong>in</strong>terpretation of the role of side cha<strong>in</strong>s:<br />

longer side cha<strong>in</strong>s keep the backbones of different polymers at larger separations, but oth-<br />

erwise they do not affect the relative orientation of molecules or their electronic structure.<br />

Our aim <strong>in</strong> this <strong>in</strong>vestigation is to evaluate whether or not this naïve assumption can be<br />

consistent with the observed trends <strong>in</strong> charge transport characteristics of the polymers<br />

with different length side cha<strong>in</strong>s. The hypothesis we want to test <strong>in</strong> this study is that the<br />

113


qualititative dependence of charge mobility on side cha<strong>in</strong> length for symmetrically sub-<br />

stituted PPV derivatives can be expla<strong>in</strong>ed uniquely <strong>in</strong> terms of changes <strong>in</strong> the m<strong>in</strong>imum<br />

distance of separation between the π conjugated backbones, without <strong>in</strong>vok<strong>in</strong>g changes <strong>in</strong><br />

backbone conformation.<br />

Of the three materials considered and def<strong>in</strong>ed <strong>in</strong> figure 6.1, dHeOPPV and dDeOPPV<br />

can be processed from solution, whereas dMeOPPV cannot. In order to make films of<br />

this materials for time-of-flight charge mobility measurements the technique described<br />

<strong>in</strong> [5] was used, namely <strong>in</strong>-situ polymerisation of dMeOPPV by thermal conversion of a<br />

soluble precursor. A study by Dr Marc Sims [6] and coworkers at Imperial College also<br />

demonstrates by Raman spectroscopy that convert<strong>in</strong>g dMeOPPV at a high temperature<br />

(185 ◦ C) produces films with micron-sized doma<strong>in</strong>s show<strong>in</strong>g high Raman signal contrast,<br />

compared to lower contrast for films converted at lower temperatures. The authors relate<br />

these high contrast Raman regions to the presence of an ordered, tightly packed mor-<br />

phology caused by the <strong>in</strong>terdigitation of methoxy group on neighbour<strong>in</strong>g polymer cha<strong>in</strong>s.<br />

Such <strong>in</strong>terdigitation has also been postulated to expla<strong>in</strong> the high degree of translational<br />

order<strong>in</strong>g observed by electron diffraction <strong>in</strong> crystals of dMeOPPV [7]. Ellipsometry mea-<br />

surements also reveal that the films converted at high temperatures have a density greater<br />

by 6 ± 2% than the films converted at a lower temperature.<br />

The time of flight mobility of these three materials was measured by our colleague Mr<br />

Sachetan Tuladhar. He found that as the side cha<strong>in</strong>s get longer:<br />

• the overall mobility becomes smaller<br />

• the field dependence becomes less positive (see figure 6.2 )<br />

These observations are reflected <strong>in</strong> the GDM parameters of the three materials. These<br />

parameters are shown <strong>in</strong> table 6.1. Note that the stronger field dependence of mobility<br />

of dMeOPPV is reflected both by its higher energetic disorder and its lower positional<br />

disorder, and its higher mobility is reflected by higher mobility prefactor µ0.<br />

The trend <strong>in</strong> the mobility prefactor µ0 can be <strong>in</strong>tuitively expla<strong>in</strong>ed by argu<strong>in</strong>g that<br />

longer side cha<strong>in</strong>s lead to greater <strong>in</strong>tercha<strong>in</strong> distances and hence smaller <strong>in</strong>tercha<strong>in</strong> trans-<br />

fer rates. The trend <strong>in</strong> positional disorder Σ seems also to be <strong>in</strong> agreement with the emer-<br />

gence of more crystall<strong>in</strong>e regions <strong>in</strong> dMeOPPV: <strong>in</strong> a crystal the positions and orientations<br />

114


Figure 6.2: Electric field dependence of mobility for dMeOPPV (full squares, preannealed<br />

C1 ) for dHeOPPV (empty diamonds, C6) and for dDeOPPV (empty<br />

circles, C10). The figure also shows the electric field dependence of mobility<br />

for dMeOPPV annealed at low temperature (full triangles, non-annealed C1)<br />

and for MDMO-PPV (dotted l<strong>in</strong>e).<br />

material µ0/cm 2 V −1 s −1 σ /eV Σ C/cmV −1<br />

dMeOPPV 2.5 10 −1 0.110 2.0 5.10 10 −4<br />

dHeOPPV 4.25 10 −3 0.091 3.05 3.20 10 −4<br />

dDeOPPV 2.73 10 −4 0.085 3.92 3.26 10 −4<br />

Table 6.1: GDM parameters for the various dialkoxy PPV derivatives.<br />

115


of molecules are more ordered and therefore the logarithms of the transfer <strong>in</strong>tegral will<br />

also be more narrowly distributed. The trend <strong>in</strong> energetic disorder is somewhat harder<br />

to expla<strong>in</strong>. When compar<strong>in</strong>g the energetic disorder of symmetrically and asymmetrically<br />

subsituted PPV derivatives earlier <strong>in</strong> the <strong>in</strong>troduction of this chapter, we quoted Martens’<br />

and coworkers’ observation that this could be due to differences <strong>in</strong> backbone conforma-<br />

tion. Apply<strong>in</strong>g the same reason<strong>in</strong>g to the comparison of the more crystall<strong>in</strong>e dMeOPPV<br />

with the longer side cha<strong>in</strong> derivatives, we would conclude that stronger <strong>in</strong>tercha<strong>in</strong> <strong>in</strong>ter-<br />

actions <strong>in</strong> dMeOPPV and the postulated <strong>in</strong>terdigitation of side cha<strong>in</strong>s would lead to a<br />

stiffer polymer backbone with less energetic disorder than the more amorphous polymers<br />

with longer side cha<strong>in</strong>s. However, the GDM parameters <strong>in</strong>dicate that energetic disorder <strong>in</strong><br />

dMeOPPV is greater than <strong>in</strong> both dHeOPPV and dDeOPPV. In this study we argue that<br />

electrostatic <strong>in</strong>teractions can expla<strong>in</strong> the observed trend <strong>in</strong> energetic disorder.<br />

In this <strong>in</strong>vestigation we assume that all three polymers can be modelled as dMeOPPV<br />

oligomers and that longer side cha<strong>in</strong>s only lead to greater <strong>in</strong>tercha<strong>in</strong> separations without<br />

affect<strong>in</strong>g any other aspect of the relative geometry of pairs of oligomers. Calculations<br />

of the transfer <strong>in</strong>tegral and site energy difference are performed on pairs of oligomers<br />

at fixed m<strong>in</strong>imum distances of approach and the results are compared to the mobility<br />

prefactor and the energetic disorder respectively. The conclusion of the study is that the<br />

variations <strong>in</strong> energetic disorder are consistent with electrostatic calculations of ∆E, which<br />

predict that more tightly packed molecules will have stronger electrostatic <strong>in</strong>teractions<br />

and hence greater site energy differences. We also show an unexpected result concern<strong>in</strong>g<br />

the correlation between ∆E and J.<br />

6.1.2 Method<br />

We consider pairs of hexamers of dMeOPPV. Calculations on shorter oligomers were car-<br />

ried out and yielded similar results to those described here. The geometry of the isolated<br />

oligomer is obta<strong>in</strong>ed at the B3LYP/6-31g* level us<strong>in</strong>g Gaussian 03 [8]. The site energy<br />

difference ∆E is calculated as the difference <strong>in</strong> electrostatic <strong>in</strong>teraction energy estimated<br />

us<strong>in</strong>g Distributed Multipole Analysis, as described <strong>in</strong> section 3.3. Briefly, two <strong>in</strong>teraction<br />

energies are calculated correspond<strong>in</strong>g respectively to the charge be<strong>in</strong>g localised on either<br />

116


Figure 6.3: Representation of the three Euler angles def<strong>in</strong><strong>in</strong>g relative orientation: Φ and<br />

Θ represent the first two Euler rotations about the temporary z and x axis, Ψ<br />

represents the last rotation about the z axis.<br />

oligomer. The difference <strong>in</strong> these <strong>in</strong>teraction energies is then the site energy difference<br />

∆E. Calculations of the charge density for the radical cation and neutral state are per-<br />

formed at the B3LYP/6-31g* level us<strong>in</strong>g Gaussian 03[8]. The distributed multipoles for<br />

either molecule are obta<strong>in</strong>ed us<strong>in</strong>g Stone’s gdma program [9]. Inductive <strong>in</strong>teractions are<br />

ignored <strong>in</strong> the calculation of ∆E. The transfer <strong>in</strong>tegral J is calculated us<strong>in</strong>g the MOO<br />

method, as described <strong>in</strong> section 3.2. ZINDO/S orbitals for the <strong>in</strong>dividual oligomers were<br />

calculated us<strong>in</strong>g the ZINDO/S Hamiltonian, as implemented <strong>in</strong> Gaussian 03[8]. In order<br />

to model the <strong>in</strong>crease <strong>in</strong> side cha<strong>in</strong> length we <strong>in</strong>crease the separation dm<strong>in</strong> between the pair<br />

of oligomers, def<strong>in</strong>ed as the m<strong>in</strong>imum distance of approach between any two atoms on<br />

either oligomer.<br />

In order to <strong>in</strong>troduce disorder we orient one of the two cha<strong>in</strong>s at random with respect<br />

to the other. Five quantities def<strong>in</strong>e the relative orientation of the two cha<strong>in</strong>s: an azimuthal<br />

and a polar angle def<strong>in</strong>e the unit dispalcement vector of the centre of the second molecule<br />

from the first molecule; three angles def<strong>in</strong>e the rotation matrix that, if the two molecules<br />

had the same centre, would br<strong>in</strong>g the coord<strong>in</strong>ates of one onto the other. The rotation<br />

matrix is def<strong>in</strong>ed <strong>in</strong> terms of Euler angles, that is three successive rotations about the z,x<br />

and z axes respectively. In terms of two sets of <strong>in</strong>ternal Cartesian coord<strong>in</strong>ates on either<br />

molecule, the first two Euler angles def<strong>in</strong>e the position of the z axis of the <strong>in</strong>ternal coordi-<br />

117


nate system of one molecule with respect to the other, while the third Euler angle places<br />

the x axis of the <strong>in</strong>ternal coord<strong>in</strong>ate system of a molecule <strong>in</strong> a particular position; the three<br />

Euler angles Φ, Θ and Ψ are represented pictorially <strong>in</strong> figure 6.3. In order to achieve a<br />

uniform distribution of the position of the centre of the molecule, the azimuthal angle and<br />

the cos<strong>in</strong>e of the polar angle are varied uniformly. To achieve a uniform distribution of<br />

orientations, we guarantee that the <strong>in</strong>ternal z axis of the second molecule is distributed<br />

uniformly on the surface of a sphere by vary<strong>in</strong>g the Euler angle Φ and the s<strong>in</strong>e of the<br />

Euler angle Θ uniformly; the Euler angle Ψ is also varied uniformly. In order to obta<strong>in</strong> a<br />

particular m<strong>in</strong>imum separation between the two molecules, we vary the distance between<br />

the centres of mass of the two molecules until the m<strong>in</strong>imum separation has the required<br />

value. A b<strong>in</strong>omial search algorithm is used to search the separation d between the centres<br />

of the two molecules <strong>in</strong> order to achieve greater computational efficiency. A separation d<br />

is picked and the distance of closest approach dm<strong>in</strong> is calculated; if this distance is smaller<br />

than the m<strong>in</strong>imum separation desired then the separation d is doubled, if it is bigger then<br />

d is halved. The procedure is iterated until the desired value of dm<strong>in</strong> is achieved.<br />

Another aspect which should be mentioned is the computational efficiency of the<br />

methods used: for each separation dm<strong>in</strong> the 10 5 calculations took roughly 90 m<strong>in</strong>utes<br />

on a s<strong>in</strong>gle processor of an Opteron 24x workstation. For comparison, on the same com-<br />

puter a s<strong>in</strong>gle SCF calculation with the ZINDO/S Hamiltonian on a pair of oligomers of<br />

PPV takes roughly 2 m<strong>in</strong>utes. Perform<strong>in</strong>g 10 5 calculations us<strong>in</strong>g the SCF method would<br />

require roughly one month: therefore repeat<strong>in</strong>g such calculations for many separations<br />

would require a huge amount of computational resources.<br />

6.1.3 Results and Comment<br />

For each m<strong>in</strong>imum distance of separation, we generate 10 5 configurations of two oligomers<br />

with random orientations; for each of these configurations we calculate the site energy dif-<br />

ference ∆E and the transfer <strong>in</strong>tegral J. We then calculate the root-mean-square average of<br />

the tranfer <strong>in</strong>tegral Jrms and the standard deviation of the distribution of energy differences<br />

σDMA. The distributions of log(J) and ∆E obta<strong>in</strong>ed for a m<strong>in</strong>imum separation dm<strong>in</strong> of 3.5<br />

Å are shown <strong>in</strong> figure 6.4. A separation of 3.5 Å was chosen to match the sum of two van<br />

118


der Waals radii of carbon (the van der Waals radius of carbon is approximately 1.7 Å ).<br />

The distribution of site energy differences and the distribution of logarithms of the trans-<br />

fer <strong>in</strong>tegral can both be approximated by normal distributions. One should bear <strong>in</strong> m<strong>in</strong>d<br />

that we are consider<strong>in</strong>g only the electrostatic <strong>in</strong>teraction of pairs of molecules, not the<br />

<strong>in</strong>teractions <strong>in</strong> the whole solid, therefore we are unable to observe the spatial correlations<br />

<strong>in</strong> site energies discussed <strong>in</strong> section 2.5.2 which would be expected to arise because of the<br />

long range nature of electrostatic <strong>in</strong>teractions. When compar<strong>in</strong>g to values of the energetic<br />

disorder σ one must bear <strong>in</strong> m<strong>in</strong>d two factors: <strong>in</strong> the first place the GDM predicts values<br />

for the distribution <strong>in</strong> site energy, not <strong>in</strong> site energy difference. Our values of σDMA should<br />

therefore be divided by √ 2 1 . In the second place, the site energy difference we calculate<br />

is only for a pair of molecules and therefore does not take <strong>in</strong>to account <strong>in</strong>teractions with<br />

the rest of the molecules <strong>in</strong> the solid. In a solid several pairs of molecules are present<br />

and thus several contributions to ∆E would be summed and we expect that the site energy<br />

would be distributed more widely than for a s<strong>in</strong>gle pair.<br />

Next, we plot the dependence of the standard deviation of ∆E and of the root mean<br />

square value of J on dm<strong>in</strong>, shown <strong>in</strong> figure 6.5. Both Jrms and σDMA decrease with <strong>in</strong>-<br />

creas<strong>in</strong>g separation, J falls exponentially with a natural length scale of 0.42 Å , while ∆E<br />

falls far more slowly and can be fitted to a polynomial. The dependence on separation of<br />

Jrms can be expla<strong>in</strong>ed by the decrease <strong>in</strong> orbital overlap, whereas the separation depen-<br />

dence of ∆E can be expla<strong>in</strong>ed by argu<strong>in</strong>g that the difference <strong>in</strong> electrostatic <strong>in</strong>teractions<br />

is correlated to the absolute values of the two electrostatic <strong>in</strong>teractions and therefore falls<br />

with distance. With regard to site energy difference this study <strong>in</strong>dicates that, <strong>in</strong> the pres-<br />

ence of orientational disorder, tighter pack<strong>in</strong>g leads to greater energetic disorder because<br />

of stronger electrostatic <strong>in</strong>teractions. This observation would be unchanged by <strong>in</strong>clud<strong>in</strong>g<br />

<strong>in</strong>ductive <strong>in</strong>teractions.<br />

Once aga<strong>in</strong>, we want to stress that the site energy difference we calculated reflects the<br />

distribution of site energies of pairs of molecules, not of a solid. In a solid, <strong>in</strong> fact, several<br />

1To see why this is the case, consider that f (E) is the site energy distribution and is distributed normally:<br />

f (E) = 1<br />

√ exp(−<br />

2πσ E2<br />

2∗σ2 ) . To estimate the distribution of the difference of two site energies E1 and E2 we<br />

must therefore estimate the <strong>in</strong>tegral: f (u) = � � f (E1) f (E2)δ((E1 − E2) − u)dE1dE2. If this expression is<br />

evaluated, the result<strong>in</strong>g distribution will be: f (u) = 1<br />

2 √ u2 exp(−<br />

πσ 4σ2 ), or <strong>in</strong> other words a normal distribution<br />

with a standard deviation σ ′ = √ (2)σ<br />

119


f(∆E)<br />

15<br />

10<br />

5<br />

-0.15 -0.1 -0.05 0 0.05 0.1 0.15<br />

∆E / eV<br />

f(log 10 J)<br />

0.5<br />

0.4<br />

0.3<br />

0.2<br />

0.1<br />

0<br />

-8 -6 -4<br />

log J / eV 10<br />

-2 0<br />

Figure 6.4: Probability distributions of the difference <strong>in</strong> site energies (left panel) and the<br />

logarithm of the transfer <strong>in</strong>tegral (right panel) of two hexamers of dMeOPPV<br />

with a m<strong>in</strong>imum separation of 3.5 Å . Probability distributions (vertical bars)<br />

are compared to a normal distribution (solid l<strong>in</strong>e).<br />

pairs of molecules would be present and the site energies would depend on the sum of<br />

all pairwise <strong>in</strong>teraction energies. We have already calculated the separation dependent<br />

distributions of electrostatic <strong>in</strong>teractions between pairs of molecules (which we used to<br />

generate figure 6.4). If the radial distribution function of hexamers <strong>in</strong> the solid film was<br />

known, we could have used that <strong>in</strong>formation to reconstruct the distribution of site energies<br />

for the whole solid. This approach would be questionable for two reasons: first there are<br />

many different ways of subdivid<strong>in</strong>g a polymer <strong>in</strong>to such conjugated segments and hence a<br />

radial distribution function of hexamers <strong>in</strong> the solid is somewhat arbitrary. Second, simply<br />

pick<strong>in</strong>g <strong>in</strong>teraction energies at random from a separation dependent distribution ignores<br />

the long range nature of electrostatic <strong>in</strong>teractions and the correlations <strong>in</strong> <strong>in</strong>teraction ener-<br />

gies to which this would lead. For both these reasons it seems to us that the best approach<br />

would be to generate realistic morphologies for slabs of polymer, use the relative orienta-<br />

tion and position of polymer segments to calculate all pairwise electrostatic <strong>in</strong>teractions<br />

and thereby obta<strong>in</strong> the distribution of site energies for the slab. This approach depends on<br />

the ability to generate accurate morphologies for polymers, a formidable computational<br />

task which we have not attempted. Nevertheless, the pairwise <strong>in</strong>teractions we have cal-<br />

culated would be the fundamental build<strong>in</strong>g block for such an approach. It can be safely<br />

120


σ DMA / eV<br />

0.05<br />

0.04<br />

0.03<br />

0.02<br />

0.01<br />

0<br />

e -d/λ λ=0.42A<br />

10 -10<br />

10 -12<br />

10 -14<br />

10 -16<br />

5 10<br />

distance of closest approach d / A m<strong>in</strong><br />

15<br />

10<br />

20<br />

-18<br />

Figure 6.5: Distance dependence of the standard deviation <strong>in</strong> values of ∆E.<br />

assumed that if the distribution <strong>in</strong> these pairwise <strong>in</strong>teractions is wider, so will be the dis-<br />

tribution <strong>in</strong> site energies for the whole solid. We expect that, all else be<strong>in</strong>g equal, tighter<br />

pack<strong>in</strong>g leads to wider distributions of site energies.<br />

Another aspect we <strong>in</strong>vestigate is whether, for a particular separation, a correlation ex-<br />

ists between J and ∆E. We know from section 2.5 that spatial correlations <strong>in</strong> site energies<br />

have a significant effect on charge transport. A correlation between J and ∆E would have<br />

an effect on transport because <strong>in</strong> the Marcus equation the transfer <strong>in</strong>tegral sets the overall<br />

rate, whereas the site energy difference sets the ratio of forward and backward rates for<br />

hops. We consider the absolute value of the site energy difference and the logarithm of<br />

J. The decision to pick the absolute value of ∆E is that its sign is arbitrary: if we swap<br />

the labels of each molecule, ∆E changes sign. The decision to consider the logarithm of<br />

J is that its distribution seemed similar ot that of ∆E. If we calculate the Pearson coef-<br />

ficient of these two quantities for a separation between dMeOPPV molecules of 3.5 Å ,<br />

we f<strong>in</strong>d a value of 0.11, which suggests a weak, but not zero, correlation. Also, this does<br />

not exclude a correlation between the two quantities which is not l<strong>in</strong>ear. A scatter plot<br />

of the absolute value of ∆E versus log(J) is shown <strong>in</strong> figure 6.6 (a). This figure shows<br />

how non-l<strong>in</strong>ear the correlation between the two quantities is. Let us also make a scatter<br />

plot of log(J) and ∆E as a function of the centre to centre distance, shown <strong>in</strong> figures 6.6<br />

(b) and (c). The assumption beh<strong>in</strong>d these plots is that, for a given m<strong>in</strong>imum distance of<br />

121<br />

10 -2<br />

10 -4<br />

10 -6<br />

10 -8<br />

J rms / eV


Figure 6.6: Scatter plot of the absolute value of ∆E aga<strong>in</strong>st the logarithm of the transfer<br />

<strong>in</strong>tegral log(J) (a). Also shown is the scatter plot of log(J) versus the centre<br />

to centre separation d (b) and the scatter plot of ∆E versus d (c).<br />

separation, shorter centre to centre distances correspond to closer packed, more coplanar<br />

morphologies. The scatter plot of transfer <strong>in</strong>tegrals suggests an isotropic distribution for<br />

distances over 10 Å and a positive correlation for shorter distances. The scatter plot for<br />

∆E shows that the site energy difference <strong>in</strong>creases with decreas<strong>in</strong>g centre to centre dis-<br />

tances, but that for centre to centre distances <strong>in</strong>ferior to 8 Å or so ∆E becomes smaller<br />

with shorter centre of mass distances. This suggests that as pack<strong>in</strong>g becomes denser the<br />

electrostatic <strong>in</strong>teractions become stronger. Below certa<strong>in</strong> centre to centre distances, how-<br />

ever, the site energy differences decrease because more coplanar pack<strong>in</strong>g <strong>in</strong>creases the<br />

symmetry <strong>in</strong> electrostatic <strong>in</strong>teractions. It seems therefore that the fact that both ∆E and<br />

J have a dependence - albeit a complex one - on the centre of mass separation, this leads<br />

to some degree of correlation between these two values. This relationship is also highly<br />

non-l<strong>in</strong>ear: large transfer <strong>in</strong>tegrals would correspond to either very small or very large<br />

∆E values, but not to <strong>in</strong>termediate ones; for this reason the correlation between the two<br />

quantities is probably badly represented by Pearson’s coeffiecient.<br />

Symmetrically substituted derivatives of PPV were studied because of their high mo-<br />

bilities compared to asymmetrically substituted derivatives. It was expected that shorter<br />

side cha<strong>in</strong>s would lead to greater mobilities. However, an unexpected result was that the<br />

122


Figure 6.7: The three spiro-l<strong>in</strong>ked tryarilam<strong>in</strong>e derivatives considered <strong>in</strong> this study.<br />

From left to right they are 2,2’,7,7’-tetrakis-(N,N-diphenylhenylam<strong>in</strong>o)-<br />

9,9’-spirobifluorene (spiro-unsub), 2,2’,7,7’-tetrakis-(N,N-di-mmethylphenylam<strong>in</strong>o)-9,9’-spirobifluorene<br />

(spiro-Me) and f<strong>in</strong>ally 2,2’,7,7’tetrakis-(N,N-di-4-methoxyphenylam<strong>in</strong>o)-9,9’-spirobifluorene<br />

(spiro-MeO)<br />

.<br />

field dependence of mobility also became more positive as side cha<strong>in</strong>s became shorter.<br />

Analysis us<strong>in</strong>g the GDM attributes such a change <strong>in</strong> field dependence to greater energetic<br />

disorder and lower positional disorder. However, the physical orig<strong>in</strong> of these changes <strong>in</strong><br />

disorder is unclear. Without perform<strong>in</strong>g electronic structure calculations one would have<br />

to speculate about the orig<strong>in</strong> of this difference <strong>in</strong> energetic disorder. For example it could<br />

have been tempt<strong>in</strong>g to attribute the differences <strong>in</strong> charge transport characteristics to those<br />

differences between the films that were observed with the Raman technique, namely the<br />

fact that films made from the dMeOPPV cha<strong>in</strong>s showed doma<strong>in</strong>s of high density and were<br />

therefore less homogeneous than films made from the longer sidecha<strong>in</strong> derivatives. Elec-<br />

tronic structure calculations revealed that it is entirely reasonable to attribute the changes<br />

<strong>in</strong> energetic disorder to the denser pack<strong>in</strong>g <strong>in</strong> the shorter side cha<strong>in</strong> derivatives and to<br />

the fact that electrostatic <strong>in</strong>teractions are stronger. They also confirm that the <strong>in</strong>crease <strong>in</strong><br />

overall mobility with denser pack<strong>in</strong>g can be attributed to greater root-mean-square trans-<br />

fer <strong>in</strong>tegrals.<br />

6.2 Spiro L<strong>in</strong>ked Triarylam<strong>in</strong>e Derivatives<br />

In this section, we analyse the dependence on side group substitution of energetic dis-<br />

order for the three differently substituted spiro-l<strong>in</strong>ked tryarilam<strong>in</strong>e derivatives shown <strong>in</strong><br />

figure 6.7. The three compounds considered all share the same core: two spiro-l<strong>in</strong>ked<br />

bi-phenyl units, where all four phenyl units are bonded to nitrogen atoms, each of which<br />

<strong>in</strong> turn is bonded to two more phenyl r<strong>in</strong>gs. The eight outer phenyl r<strong>in</strong>gs are either unsub-<br />

123


stitued (spiro-unsub), or substituted with methyl groups <strong>in</strong> the para position (spiro-Me),<br />

or substituted with methoxy groups <strong>in</strong> the meta position (spiro-MeO). These compounds<br />

benefit both from high hole mobilities and from high glass transition temperatures and<br />

have found application <strong>in</strong> LEDs [13] and dye sensitised solar cells [14]. Other than the<br />

technological <strong>in</strong>terest <strong>in</strong> these materials, the motivation for compar<strong>in</strong>g their charge trans-<br />

port characteristics is similar to the motivation for study<strong>in</strong>g different dialkoxy PPVs: to<br />

understand the relationship between chemical structure and charge transport characteris-<br />

tics. The time-of-flight hole mobility of spiro-MeO and spiro-unsub was measured by U.<br />

Bach et al [11], whereas the time of flight hole mobility of spiro-MeO was measured by<br />

D. Poplavskyy and J. Nelson [12]. D. Poplavskyy observes that the energetic disorder<br />

calculated via the GDM for the three materials is larger for spiro-MeO (0.101eV) than for<br />

either spiro-Me (0.080eV) or spiro-unsub (0.080eV). They attribute this observation to<br />

two possible <strong>in</strong>terpretations: either the methoxy group <strong>in</strong>creases the permanent dipole of<br />

the spiro-MeO, or the different techniques used to produce the samples for time of flight<br />

measurements (evaporation for spiro-Me and spiro-unsub and solution cast<strong>in</strong>g for spiro-<br />

MeO) lead to different amounts of disorder. Because all molecules belong to the S 4 po<strong>in</strong>t<br />

group, the hypothesis of a permanent dipole moment can be excluded. Even though the<br />

molecules <strong>in</strong> their ground state have no permanent dipole, they can still have electrostatic<br />

<strong>in</strong>teractions. In the case of well separated small polar molecules, the lead<strong>in</strong>g component<br />

of the electrostatic <strong>in</strong>teraction is the permanent dipole-monopole <strong>in</strong>teraction; however <strong>in</strong><br />

the case of densely packed large non-polar molecules, electrostatic <strong>in</strong>teractions should<br />

be calculated from distributed multipole analysis. The aim of this study was to assess<br />

whether the differences <strong>in</strong> energetic disorder extracted via the GDM from mobility data<br />

for the different spiro l<strong>in</strong>ked compounds can be attributed to electrostatic <strong>in</strong>teractions.<br />

6.2.1 Method and Results<br />

Geometries for the spiro-MeO, spiro-Me, and spiro-unsub were obta<strong>in</strong>ed at the AM1 level<br />

us<strong>in</strong>g Gaussian 03 and assum<strong>in</strong>g S 4 symmetry. Calculation of the charge density was<br />

done us<strong>in</strong>g the B3LYP functional with the 6-31g* basis set and the distributed multipole<br />

analysis was performed us<strong>in</strong>g Prof. Stone’s gdma program. Geometries were generated<br />

124


number of calculations<br />

7000<br />

6000<br />

5000<br />

4000<br />

3000<br />

2000<br />

1000<br />

spiro-unsub<br />

spiro-Me<br />

spiro-MeO<br />

0<br />

-0.06 -0.04 -0.02 0<br />

∆E / eV<br />

0.02 0.04<br />

Figure 6.8: Histogram of ∆E for spiro-unsub (dashed l<strong>in</strong>e), spiro-Me (dotted l<strong>in</strong>e) and<br />

spiro-MeO (full l<strong>in</strong>e). The standard deviation for the three materials is 8 meV<br />

for spiro-unsub, 10 meV for spiro-Me and 15 meV spiro-MeO.<br />

material σGDM/eV σDMA/eV<br />

spiro-unsub 0.080 0.08<br />

spiro-Me 0.080 0.010<br />

sprio-MeO 0.101 0.015<br />

Table 6.2: GDM energetic disorder σGDM and standard deviation <strong>in</strong> distribution of site<br />

energy difference calculated us<strong>in</strong>g distributed Multipole analysis ( σDMA )<br />

<strong>in</strong> the same way as <strong>in</strong> the previous section, us<strong>in</strong>g a m<strong>in</strong>imum distance of approach of<br />

3.5 Å and ensur<strong>in</strong>g that pairs of molecules are <strong>in</strong> relative orientations which are evenly<br />

distributed <strong>in</strong> orientational phase space. A m<strong>in</strong>imum separation of 3.5 Å was chosen<br />

tak<strong>in</strong>g <strong>in</strong>to account the van der Waals radius of carbon. The distributions of values of ∆E<br />

are shown <strong>in</strong> figure 6.8, whereas the values of standard deviations are shown <strong>in</strong> table 6.2.<br />

The standard deviation for spiro-MeO is considerably larger than for spiro-Me or spiro-<br />

unsub, while those for spiro-Me and spiro-unsub are similar, reflect<strong>in</strong>g the trend <strong>in</strong> the<br />

value of the energetic disorder parameters with chemical structure. We have also carried<br />

out these calculations for a constant centre to centre separation of 18.5 Å while vary<strong>in</strong>g<br />

125


the relative orientation at random to ensure that the m<strong>in</strong>imum distance of approach is 3.4<br />

Å to 3.6 Å. These calculations show the same trend <strong>in</strong> variations of the standard deviation<br />

of ∆E for the three materials.<br />

The values of σDMA are rather different <strong>in</strong> magnitude from the energetic disorder pa-<br />

rameter σ. Several possible reasons could be responsible for this: first, consider<strong>in</strong>g <strong>in</strong>ter-<br />

actions with more neighbours might <strong>in</strong>crease ∆E. Second, local m<strong>in</strong>imum energy ground<br />

state structures are likely to exist <strong>in</strong> such a large molecule, and these would contribute to<br />

the distribution of ∆E. Third, without a good morphology model it is impossible to deter-<br />

m<strong>in</strong>e which particular pair geometries exist <strong>in</strong> the solid and would therefore contribute to<br />

the site energies. What is clear, however, is that add<strong>in</strong>g polar side groups to a molecule<br />

will <strong>in</strong>crease its short range electric field and therefore <strong>in</strong>crease the magnitude of ∆E even<br />

if it does not modify the molecule’s permanent dipole.<br />

6.3 Conclusion<br />

In this chapter we have shown how the fast methods developed for calculat<strong>in</strong>g ∆E and<br />

J can be used to improve our understand<strong>in</strong>g of charge transport <strong>in</strong> disordered solids by<br />

<strong>in</strong>terpret<strong>in</strong>g the orig<strong>in</strong> of GDM parameters <strong>in</strong> terms of molecular properties. In particu-<br />

lar, the use of DMA for calculat<strong>in</strong>g ∆E allows us to make qualitative statements on the<br />

strength and asymmetry of electrostatic <strong>in</strong>teraction and allows us to understand the factors<br />

contribut<strong>in</strong>g to energetic disorder.<br />

126


Bibliography<br />

[1] H. Sirr<strong>in</strong>ghaus, P.J. Brown, R.H. Friend, M.M. Nielsen, K. Bechgaard, B.M.W.<br />

Langeveld-Voss, A.J.H. Spier<strong>in</strong>g, R.A.J. Janssen, E.W. Meijer, P. Herwig and D.M.<br />

de Leeuw, Two-dimensional charge transport <strong>in</strong> self-organized, high-mobility conju-<br />

gated polymers, Nature, 401, 605 (1999)<br />

[2] M.J. W<strong>in</strong>okur and W. Chunwachirasiri, Nanoscale StructureProperty Relationships<br />

<strong>in</strong> Conjugated Polymers: Implications for Present and Future Device Applications,<br />

Journal of Polymer Science: Part B: Polymer Physics, 41, 2630, (2003)<br />

[3] H.C.F. Martens, P.W.M. Blom and H.F.M. Schoo, Comparative study of hole trans-<br />

port <strong>in</strong> poly(p-para-phenylne v<strong>in</strong>ylene) derivatives, Physical Review B, 61, 7489<br />

(2000)<br />

[4] M. Kemer<strong>in</strong>k, J.K.J. van Duren, A.J.J.M. van Breemen, J. Wildeman, M.M. Wienk,<br />

P.W.M. Blom, H.F.M. Schoo and R.A.J. Janssen, Substitution and Preparation Effects<br />

on the Molecular-Scale Morphology of PPV Films, Macromolecules 38, 7784 (2005)<br />

[5] P.L. Burn, D.D.C. Bradley, R.H. Friend, D.A. Halliday, A.B. Holmes, R.W. Jack-<br />

son and A. Kraft, Precursor route chemistry and electronic properties of poly(p-<br />

phenylnev<strong>in</strong>ylene), poly[(2,5-dimethyl-p-phenylne)v<strong>in</strong>ylene] and poly[(2,5-dimethoxy-<br />

p-phenylne)v<strong>in</strong>ylene], Journal of the Chemical Society, Perk<strong>in</strong> Transaction 1 23, 3231<br />

(1992)<br />

[6] M. Sims, S.M. Tuladhar, R.C. Maher, M. Campoy-Quiles, S.A. Choulis, J. Nelson,<br />

D.D.C. Bradley, P.G. Etchego<strong>in</strong>, C. Jones, A. Connor, K. Suhl<strong>in</strong>g, D.R. Richards,<br />

P. Massiot, C. Nielsen and J.H.G Ste<strong>in</strong>ke, Structure-charge transport correlations <strong>in</strong><br />

poly(2,5-dimethoxy-1,4-phenylnev<strong>in</strong>ylene) (PDMeOPV). , manuscript <strong>in</strong> preparation<br />

127


[7] J.H.F. Martens, E.A. Marseglia, D.D.C. Bradley, R.H. Friend, P.L. Burn and A.B.<br />

Holmes The Effect of Side Groups on the Structure and Order<strong>in</strong>g of poly(p-Phenylene<br />

V<strong>in</strong>ylene) Derivatives, Synthetic Metals, 55, 449 (1993)<br />

[8] Gaussian 03, Revision C.02, M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E. Scuse-<br />

ria, M. A. Robb, J. R. Cheeseman, J. A. Montgomery, Jr., T. Vreven, K. N. Kud<strong>in</strong>, J. C.<br />

Burant, J. M. Millam, S. S. Iyengar, J. Tomasi, V. Barone, B. Mennucci, M. Cossi, G.<br />

Scalmani, N. Rega, G. A. Petersson, H. Nakatsuji, M. Hada, M. Ehara, K. Toyota, R.<br />

Fukuda, J. Hasegawa, M. Ishida, T. Nakajima, Y. Honda, O. Kitao, H. Nakai, M. Klene,<br />

X. Li, J. E. Knox, H. P. Hratchian, J. B. Cross, V. Bakken, C. Adamo, J. Jaramillo, R.<br />

Gomperts, R. E. Stratmann, O. Yazyev, A. J. Aust<strong>in</strong>, R. Cammi, C. Pomelli, J. W.<br />

Ochterski, P. Y. Ayala, K. Morokuma, G. A. Voth, P. Salvador, J. J. Dannenberg, V.<br />

G. Zakrzewski, S. Dapprich, A. D. Daniels, M. C. Stra<strong>in</strong>, O. Farkas, D. K. Malick,<br />

A. D. Rabuck, K. Raghavachari, J. B. Foresman, J. V. Ortiz, Q. Cui, A. G. Baboul, S.<br />

Clifford, J. Cioslowski, B. B. Stefanov, G. Liu, A. Liashenko, P. Piskorz, I. Komaromi,<br />

R. L. Mart<strong>in</strong>, D. J. Fox, T. Keith, M. A. Al-Laham, C. Y. Peng, A. Nanayakkara, M.<br />

Challacombe, P. M. W. Gill, B. Johnson, W. Chen, M. W. Wong, C. Gonzalez, and J.<br />

A. Pople, Gaussian, Inc., Wall<strong>in</strong>gford CT, 2004.<br />

[9] A.J. Stone, Distributed multipole analysis: Stability for large basis sets, Journal of<br />

chemical theory and computation, 1, 1128 (2005)<br />

[10] J. Caillet and P. Claverie, Theoretical Evaluation of the <strong>Intermolecular</strong> Interaction<br />

Energy of a Crystal: Application to the Analysis of Crystal Geometry, Acta Crystallo-<br />

graphica, A31, 448 (1975)<br />

[11] U. Bach, K. de Cloedt, H. Spreitzer and M. Grätzel Characterisation of Hole Trans-<br />

port <strong>in</strong> a New Class of Spiro-L<strong>in</strong>ked Oligotriphenylam<strong>in</strong>e Compounds, Advanced Ma-<br />

terials, 12, 1060 (2000)<br />

[12] D. Poplavskyy and J. Nelson, Nondispersive hole transport <strong>in</strong> amorphous films of<br />

methoxy-spirofluorene-arylam<strong>in</strong>e organic compounds, Journal of Applied Physics, 93,<br />

341 (2003)<br />

128


[13] F. Steuber, J. Staudigel, M. Stossel, J. Simmerer, A. W<strong>in</strong>nacker, H. Spreitzer, F.<br />

Weissortel and J. Salbeck, White Light Emission from Organic LEDs Utiliz<strong>in</strong>g Spiro<br />

Compounds with High-Temperature Stability, Advanced Materials 12, 130, (2000)<br />

[14] J. Kruger, R. Plass, L. Cevey, M. Piccirelli, M. Grätzel and U. Bach, High efficiency<br />

solid-state photovoltaic device due to <strong>in</strong>hibition of <strong>in</strong>terface charge recomb<strong>in</strong>ation,<br />

Applied Physics Letters 79, 2085, (2001).<br />

129


Chapter 7<br />

Simulation of <strong>Charge</strong> <strong>Transport</strong> <strong>in</strong><br />

Assemblies of Hexabenzocoronenes and<br />

Fluorene Oligomers<br />

In the previous chapters, calculation of charge transfer parameters are carried out to give<br />

qualitative understand<strong>in</strong>g, either by help<strong>in</strong>g parametrize lattice Monte Carlo simulations<br />

(chapter five) or to expla<strong>in</strong> GDM parameters (chapter six). In this chapter we present two<br />

studies <strong>in</strong> which calculations of transfer <strong>in</strong>tegrals are used to model charge mobility <strong>in</strong> as-<br />

semblies of molecules, as models of solid films of two particular materials. The transfer<br />

<strong>in</strong>tegrals are used to calculate charge transfer rates us<strong>in</strong>g the Marcus charge transfer equa-<br />

tion and these rates are <strong>in</strong> turn used to perform Monte Carlo simulations of charge dynam-<br />

ics. S<strong>in</strong>ce the relative position and orientation of neighbour<strong>in</strong>g molecules must be known<br />

to calculate J, it is necessary to have a model of the molecular pack<strong>in</strong>g of a material.<br />

The two material systems we consider were the crystall<strong>in</strong>e phase of poly(dioctylfluorene)<br />

(PFO) and the hexagonal mesophase of the discotic liquid crystal hexabenzocoronene<br />

(HBC). In both cases the morphology model is provided by collaborators: <strong>in</strong> the case<br />

of PFO Dr. S. Athanosopoulos and Dr. A. Walker from the University of Bath designed<br />

morphologies us<strong>in</strong>g an empirical morphology model. In the case of HBC classical molec-<br />

ular dynamics methods were used by Dr. V. Marcon and Dr. D. Andrienko from the Max<br />

Planck Institute for Polymer Research <strong>in</strong> Ma<strong>in</strong>z to obta<strong>in</strong> an atomistic simulation of the<br />

morphology of columns of HBC. For the PFO calculations, Monte Carlo simulations<br />

130


(a)<br />

z<br />

Electric field<br />

x<br />

(b)<br />

φ 1<br />

∆∆∆∆r<br />

Figure 7.1: Schematic representation of the morphology model used by Dr. S.<br />

Athanosopoulos to describe PFO and def<strong>in</strong>itions of the parameters govern<strong>in</strong>g<br />

the relative position and orientation of molecules. (a) The polymer cha<strong>in</strong>s<br />

are packed hexagonally, with the cha<strong>in</strong> axis perpendicular to the electric field.<br />

(b) Each trimer is allowed to rotate by a torsion φ and can be displaced by a<br />

distace ∆r <strong>in</strong> the direction perpendicular to the cha<strong>in</strong> axis and can slip by a<br />

distance dz parallell to the cha<strong>in</strong> axis. (c) Top view of the central planes of<br />

two trimers, show<strong>in</strong>g the torsion angles φ1 and φ2 (∆φ = φ2 −φ1) and the polar<br />

angle α.<br />

were performed by our colleagues <strong>in</strong> Bath, while for HBC we used some <strong>in</strong>-house code,<br />

adapted from Dr. A. Chatten’s and Prof. J. Nelson’s [1] code to be able to accomodate<br />

lattices where the transport sites are arbitrarily distributed. By apply<strong>in</strong>g calculations of<br />

transfer <strong>in</strong>tegrals to simulations of hopp<strong>in</strong>g transport the work described <strong>in</strong> this chapter<br />

goes beyond previous studies <strong>in</strong> relat<strong>in</strong>g charge transport parameters to charge mobilities.<br />

7.1 PFO<br />

<strong>Charge</strong> transport <strong>in</strong> polyfluorene derivatives is heavily dependent on morphology and<br />

orientation. For example, <strong>in</strong> poly(9,9 dioctylfluorene) (PFO) studies have shown [2] [3]<br />

that changes <strong>in</strong> thermal treatment and alignment can lead to changes <strong>in</strong> mobility of over<br />

two orders of magnitude. The aim of this part of the study is to exam<strong>in</strong>e the effect of<br />

variations <strong>in</strong> the morphology of a film of crystall<strong>in</strong>e polyfluorene on simulated charge<br />

transport behaviour. The morphology simulation was developed by our colleagues <strong>in</strong><br />

Bath, Dr. A. Walker, Dr. P. Watson and Dr. S. Anthopoulos. This morphology simulation<br />

is <strong>in</strong>spired by studies of PFO crystal structure by S. Kawana and co-workers [4]. In<br />

that work, the structure of crystall<strong>in</strong>e PFO that best fitted the observed X-ray spectra of<br />

aligned films of PFO was a hexagonal arrangement of parallel cha<strong>in</strong>s. In our morphology<br />

131<br />

φ 2<br />

y<br />

(c)<br />

φ 1<br />

α<br />

φ 2<br />

x


Figure 7.2: HOMO (left) and HOMO-1 (right), of the restricted open shell orbitals of a<br />

fluorene trimer with one of the hydrogen end groups miss<strong>in</strong>g.<br />

model, polymer cha<strong>in</strong>s are packed <strong>in</strong> a hexagonal lattice with a particular lattice constant<br />

a. The cha<strong>in</strong>s are segmented <strong>in</strong>to charge transport<strong>in</strong>g units, which we choose to represent<br />

as trimers. This length oligomer is to 23 Å and is consistent with the proposed length of<br />

polarons <strong>in</strong> PPV (25 Å [5]) and <strong>in</strong> polythiophene (20 Å [6]). The excluded volume of octyl<br />

side cha<strong>in</strong>s was calculated and corresponds to a m<strong>in</strong>imum distance of approach between<br />

neighbour<strong>in</strong>g polymer cha<strong>in</strong>s of 6.3 Å. It must be noted that the morphology model we<br />

use is not truly ab <strong>in</strong>itio. In particular, the hexagonal lattice constant was allowed to vary.<br />

In fact the fundamental difference between this approach and models of charge transport<br />

on lattices such as the GDM is that whereas <strong>in</strong> the lattice models the parameters of the<br />

transfer rate equation are picked from arbitrary distributions, <strong>in</strong> our model the morphology<br />

is chosen <strong>in</strong> a parametrized fashion but the transfer rates are then derived rigorously.<br />

Figure 7.1 shows the model of morphology we use. The position and orientation of<br />

a trimer is def<strong>in</strong>ed by its torsion angle φ (def<strong>in</strong>ed as the angle between the plane of the<br />

central monomer to with respect to the central axis of the hexagonal axis), and its x,y,z<br />

coord<strong>in</strong>ates. In a torsionally disordered system φ is chosen at random for each monomer<br />

and <strong>in</strong> a laterally disordered system the displacement <strong>in</strong> x and y are chosen at random,<br />

132


ensur<strong>in</strong>g that neighbour<strong>in</strong>g trimers never come closer than 6.3 Å. Then, the relative po-<br />

sition of two trimers is therefore uniquely def<strong>in</strong>ed by a displacement ∆r perpendicular to<br />

the cha<strong>in</strong> axis, a displacement dz parallel to the cha<strong>in</strong> axis, the difference dφ between the<br />

torsion angles and the polar angle α def<strong>in</strong><strong>in</strong>g the position of the second trimer with respect<br />

to the plane of the central monomer of the first trimer.<br />

For each pair of nearest neighbour<strong>in</strong>g trimers we use the Marcus charge transfer equa-<br />

tion (equation 2.6) to f<strong>in</strong>d the transfer rates, hav<strong>in</strong>g calculated the transfer <strong>in</strong>tegrals us<strong>in</strong>g<br />

MOO and the reorganisation energy us<strong>in</strong>g a scheme such as that described <strong>in</strong> section 2.2<br />

and us<strong>in</strong>g B3LYP/6-31g* for the geometry optimisations. The site energy difference ∆E<br />

is calculated from the external electric field F and the relative position of the centres dr<br />

as: ∆E = F.dr. We exclude electrostatic effects from our calculation of ∆E. Note that<br />

s<strong>in</strong>ce we treat all transport sites as trimers, no site energy disorder due to a distribution <strong>in</strong><br />

conjugation lengths is allowed. By elim<strong>in</strong>at<strong>in</strong>g such sources of site energy disorder this<br />

approach should help to clarify the <strong>in</strong>fluence of morphology on charge transport charac-<br />

teristics.<br />

Four different morphologies are compared <strong>in</strong> this study:<br />

• an ordered morphology where the centre of each trimer lies on the lattice po<strong>in</strong>ts and<br />

each trimer has the same torsion angle φ<br />

• a torsionally disordered morphology where the torsion angles are varied at random<br />

• a regular morphology with a unit cell composed of several trimers with the relative<br />

torsion angles chosen to maximise transfer rates. We refer to this morphology as<br />

optimally ordered<br />

• a laterally disordered morphology where the trimers are displaced by random amounts<br />

x, y.<br />

Once the morphology has been generated, it is necessary to compute two different<br />

types of transfer <strong>in</strong>tegral to describe charge transport <strong>in</strong> such assemblies of molecules:<br />

<strong>in</strong>tramolecular transfer <strong>in</strong>tegrals between neighbour<strong>in</strong>g trimers <strong>in</strong> the same cha<strong>in</strong> and <strong>in</strong>-<br />

termolecular <strong>in</strong>tegrals between neighbour<strong>in</strong>g trimers on different cha<strong>in</strong>s. Both were cal-<br />

culated us<strong>in</strong>g the MOO method described <strong>in</strong> section 3.2. In this paragraph we discuss<br />

133


how the end groups of the oligomers were treated. When calculat<strong>in</strong>g the transfer <strong>in</strong>te-<br />

grals for <strong>in</strong>termolecular transfer, we added two hydrogen end groups to each trimer to<br />

obta<strong>in</strong> the molecular orbitals. For the <strong>in</strong>tramolecular transfer <strong>in</strong>tegrals we could only add<br />

a s<strong>in</strong>gle hydrogen end group to each trimer and had to therefore calculate the trimer’s<br />

molecular orbitals <strong>in</strong> a doublet sp<strong>in</strong> state. We performed this calculation of the orbitals<br />

us<strong>in</strong>g a restricted open shell wavefunction: this way we obta<strong>in</strong> a s<strong>in</strong>gly occupied HOMO<br />

which corresponds to the unpaired electron which, <strong>in</strong> the presence of another hydrogen<br />

end group, would form a σ bond. We also obta<strong>in</strong> a doubly occupied HOMO-1 which is<br />

virtually <strong>in</strong>dist<strong>in</strong>guishable from the HOMO of the closed shell s<strong>in</strong>glet with two hydrogen<br />

end group. These orbitals are shown <strong>in</strong> figure 7.2. S<strong>in</strong>ce the end group hydrogens hardly<br />

contribute to the HOMO of the closed shell trimer and the HOMO-1 of the open shell<br />

trimer, the choice of whether or not to <strong>in</strong>clude these end groups is somewhat arbitrary, re-<br />

member<strong>in</strong>g that <strong>in</strong> the MOO scheme described <strong>in</strong> section 3.2 the density matrix does not<br />

contribute to the calculation of the Fock matrix and hence the transfer <strong>in</strong>tegral depends<br />

only on the orbitals of <strong>in</strong>terest.<br />

In this paragraph we show the results for the calculations of transfer <strong>in</strong>tegrals per-<br />

formed on pairs of trimers on the same cha<strong>in</strong> and on neighbour<strong>in</strong>g cha<strong>in</strong>s. Figure 7.3<br />

shows the dependence of the transfer <strong>in</strong>tegral on the polar and torsional angles for a dis-<br />

tance of 6.3 Å, while figure 7.4 shows its dependence on separation for particular values<br />

of α and ∆φ. When calculat<strong>in</strong>g the <strong>in</strong>tramolecular <strong>in</strong>tegrals, the trimers are assumed to be<br />

col<strong>in</strong>ear and hence the transfer <strong>in</strong>tegral is assumed to depend only on the relative torsion<br />

angle. Figure 7.5 shows the dependence of the modulus square of the transfer <strong>in</strong>tegral<br />

on ∆φ. It should be noted that our treatment of <strong>in</strong>tramolecular transport is still somewhat<br />

unsatisfactory, because the largest <strong>in</strong>tramolecular transfer <strong>in</strong>tegrals obta<strong>in</strong>ed are compa-<br />

rable <strong>in</strong> size to the reorganisation energy λ. Therefore us<strong>in</strong>g the non-adiabatic Marcus<br />

equation is not appropriate <strong>in</strong> this case. However the hopp<strong>in</strong>g rate predicted by the Mar-<br />

cus transport equation along the polymer cha<strong>in</strong> are still significantly faster than that for<br />

<strong>in</strong>tercha<strong>in</strong> hopp<strong>in</strong>g and s<strong>in</strong>ce we are measur<strong>in</strong>g transport <strong>in</strong> the direction perpendicular to<br />

the cha<strong>in</strong> axis, <strong>in</strong>tercha<strong>in</strong> hops limit the mobility. Also, no model exists, to the best of our<br />

knowledge that allows the treatment of both coherent and <strong>in</strong>coherent transport.<br />

134


Polar angle α<br />

350<br />

300<br />

250<br />

200<br />

150<br />

100<br />

50<br />

0<br />

0 100 200 300<br />

Torsion angle dφ<br />

Figure 7.3: Contour plot of log(|J| 2 ) as a function of relative torsion angle dφ and the<br />

polar angle α which describes position. Notice how the values of these <strong>in</strong>tercha<strong>in</strong><br />

transfer <strong>in</strong>tegrals are always much smaller than the <strong>in</strong>tracha<strong>in</strong> transfer<br />

<strong>in</strong>tegrals from figure 7.5.<br />

|J| 2 / eV 2<br />

0.0001<br />

1e-06<br />

1e-08<br />

1e-10<br />

1e-12<br />

6 7 8<br />

centre to centre separation / A<br />

9 10<br />

Figure 7.4: Distance dependence of the transfer <strong>in</strong>tegral squared on distance for φ = 150 ◦<br />

and α = 0 ◦ (squares) and for φ = 0 ◦ and α = 0 ◦ (circles).<br />

135<br />

−10<br />

−9<br />

−8<br />

−7<br />

−6<br />

−5<br />

−4<br />

−3<br />

log(|J| 2 ) /eV 2


transfer <strong>in</strong>tegral |J| 2 / eV 2<br />

0.04<br />

0.03<br />

0.02<br />

0.01<br />

0<br />

0 100 200<br />

torsion angle ∆φ / degrees<br />

300 400<br />

Figure 7.5: Intracha<strong>in</strong> transfer <strong>in</strong>tegrals squared as a function of relative torsion φ for two<br />

trimers.<br />

The reorganisation energy for the trimer was calculated us<strong>in</strong>g the protocol described<br />

<strong>in</strong> section 2.2 with the hybrid density functional B3LYP and Pople’s double split basis set<br />

with extra polarisation basis functions 6-31g*. λ was found to be equal to 0.21eV.<br />

Once the parameters for the Marcus charge transfer equation have been calculated<br />

for all <strong>in</strong>tercha<strong>in</strong> and <strong>in</strong>tracha<strong>in</strong> neighbours, charge transport can be simulated us<strong>in</strong>g the<br />

Monte Carlo method. These transport simulations were carried out by Dr. S. Athanosopou-<br />

los us<strong>in</strong>g a cont<strong>in</strong>uous time random walk algorithm with cont<strong>in</strong>uous boundary conditions<br />

as described <strong>in</strong> section 2.5.1. First we simulate hole transport <strong>in</strong> an ordered lattice where<br />

all the trimers have the same torsion φ. To choose the optimal torsion, we study the net<br />

transfer <strong>in</strong> the field direction as a function of φ for an ordered lattice. The overall mobility<br />

will be higher when the rates along the field (forward rates) are faster compared to rates<br />

aga<strong>in</strong>st the field (backward rates). The difference between forward and backward rates as<br />

a function of φ is shown <strong>in</strong> figure 7.6 for several different fields. It can be seen that the<br />

fastest net forward rate is obta<strong>in</strong>ed for φ = 20 ◦ . Thus, for the ordered morphology we<br />

pick a torsion of 20 ◦ . The results for the hole mobility µh as a function of electric field<br />

136


net rate (s -1 )<br />

2.0×10 10<br />

1.5×10 10<br />

1.0×10 10<br />

5.0×10 9<br />

0.0<br />

0 20 40 60 80<br />

torsion angle φ (degrees)<br />

Figure 7.6: Difference between forward and backward rates for an ordered lattice as<br />

√<br />

a function the torsion angle φ at which the trimers are set. F =<br />

200(V/cm) 1/2 (solid l<strong>in</strong>e), 400(V/cm) 1/2 (dotted l<strong>in</strong>e), 600(V/cm) 1/2 (dashed<br />

l<strong>in</strong>e), 800(V/cm) 1/2 (cha<strong>in</strong>ed l<strong>in</strong>e, one dot), 1000(V/cm) 1/2 (cha<strong>in</strong>ed l<strong>in</strong>e, two<br />

dots).<br />

are shown <strong>in</strong> figure 7.7 (diamonds) for a lattice constant of 6.5 Å (the choice of lattice<br />

constant is justified below). Also shown are the results for φ = 0 ◦ (open triangles) and<br />

a torsionally disordered system (squares). Results for lateral disorder are not shown, as<br />

it has little effect on muh at this lattice constant because the trimers have little freedom<br />

to move outside the excluded volume of PFO, which has a diamater of 6.3 Å. The fourth<br />

plot (closed triangles) is for optimally ordered systems. The optimally ordered system<br />

was designed ensur<strong>in</strong>g that each trimer has a neighbour <strong>in</strong> the direction of the field with<br />

∆φ = 150 ◦ (this is the difference <strong>in</strong> torsion angles that ensures greatest possible J 2 for<br />

α = 0 ◦ as can be seen <strong>in</strong> figure 7.3). Also shown <strong>in</strong> figure 7.7 is the experimental mobility<br />

for aligned films of PFO from reference [3]. As can be seen, the disordered morphology<br />

with a lattice constant of 6.5 Å produces the best fit with experimental data.<br />

The experimental hole mobility is weakly dependent on electric field. Equation 2.10<br />

shows that for small fields the Marcus transfer rates are proportional to exp(−∆E/2kT);<br />

137


Hole mobility µ h (cm 2 /Vs)<br />

10 -1<br />

10 -2<br />

10 -3<br />

0 200 400 600 800 1000<br />

F 1/2 (V/cm) 1/2<br />

10 -4<br />

Figure 7.7: Dependence of mobility on the square root of the applied electric field for:<br />

the ordered morphology with φ = 0 ◦ (open triangles), the ordered morphology<br />

with φ = 20 ◦ (diamonds), the morphology with disordered torsions (full<br />

squares), the optimally disordered morphology (full triangles) and the experimental<br />

data on alligned PFO from reference [3] (full circles).The optimally<br />

ordered morphology corresponds to when neighbours <strong>in</strong> the direction of the<br />

field have a relative torsion φ of 150 ◦ . The lattice constant for the hexagonal<br />

lattice is 6.5 Å .<br />

138


therefore if the only contribution to the site energy is the electric field via ∆E = qF.r and if<br />

the fields is small, we expect the velocity of charges to be proportional to the electric field<br />

and the mobility to be field <strong>in</strong>dependent. S<strong>in</strong>ce our simulations were made <strong>in</strong> the absence<br />

of disorder, the weak electric field dependence can be expla<strong>in</strong>ed this way. To fit the<br />

magnitude of the hole mobility we treat the lattice constant as a fitt<strong>in</strong>g parameter. There<br />

are two <strong>in</strong>terest<strong>in</strong>g observations emerg<strong>in</strong>g from this study: first we notice that disorder<br />

<strong>in</strong> the transfer <strong>in</strong>tegrals can <strong>in</strong>crease mobility and second that the lattice constant (6.5 Å)<br />

which needed to fit the experimental data is much shorter than the lattice constant (9.5 Å)<br />

expected from a density of 1 gcm −3 .<br />

The fact that disorder <strong>in</strong> the <strong>in</strong>termolecular transfer <strong>in</strong>tegral distribution can <strong>in</strong>crease<br />

mobility can be readily understood if one bears <strong>in</strong> m<strong>in</strong>d that <strong>in</strong>tramolecular charge transfer<br />

is almost always faster than <strong>in</strong>termolecular charge transfer. S<strong>in</strong>ce charges are therefore<br />

free to travel quickly along the cha<strong>in</strong>s, they will always be able to f<strong>in</strong>d locations with<br />

highest transfer <strong>in</strong>tegrals for hopp<strong>in</strong>g to a neighbour<strong>in</strong>g cha<strong>in</strong> <strong>in</strong> the field direction. It ap-<br />

pears therefore that <strong>in</strong> this system mobility is controlled by hot spots where the <strong>in</strong>tercha<strong>in</strong><br />

transfer <strong>in</strong>tegral takes the largest values. The large variability of transfer <strong>in</strong>tegrals on α<br />

and ∆φ means that <strong>in</strong>troduc<strong>in</strong>g disorder leads to wider distributions <strong>in</strong> |J| 2 . S<strong>in</strong>ce mobility<br />

is set by the greatest transfer <strong>in</strong>tegrals <strong>in</strong> the distribution and not by the average value of<br />

|J| 2 , a wide distribution <strong>in</strong> J 2 can lead to higher mobilities than a narrower one with a<br />

greater average.<br />

In order to fit the experimental data a lattice constant used of 6.5 Å was used. Such a<br />

lattice constant is much higher than that which is necessary for PFO to have a density of<br />

1 gcm −3 (approximately 9 Å). There are different possible <strong>in</strong>terepretations of this obser-<br />

vation. The mobility could have been underestimated because of an underestimation of<br />

the transfer <strong>in</strong>tegrals or an overestimation of the reorganisation energy. Alternatively, one<br />

could accept that <strong>in</strong> the polymer pathways with such high densities and short <strong>in</strong>tercha<strong>in</strong><br />

distances do <strong>in</strong>deed form. Another possible <strong>in</strong>terpretation would be that the morphology<br />

model we employed was fundamentally <strong>in</strong>correct. We will discuss these three possibilities<br />

<strong>in</strong> order.<br />

We argue that the discrepancy cannot be expla<strong>in</strong>ed by underestimation of the transfer<br />

139


ates. The transfer <strong>in</strong>tegral would have to be greater by a factor of 1000 (based on the nat-<br />

ural length of the decay of the transfer <strong>in</strong>tegral) to obta<strong>in</strong> similar mobilities to those we ob-<br />

ta<strong>in</strong>ed with a lattice constant of 6.5 Å with a lattice constant of 9.0 Å. Even if the reorgani-<br />

sation energy had a value of 0.1 eV <strong>in</strong>stead of the 0.21 eV we calculated, this would lead to<br />

an <strong>in</strong>crease <strong>in</strong> transfer rates for mobilities of a factor of √ (0.21/0.1)exp((0.21−0.1)/4kT)<br />

which would be equal to roughly a factor of five, not the factor of 10 6 needed. Smaller<br />

values of the reorganisation energy would easily lead to the electric field caus<strong>in</strong>g energy<br />

differences greater than the reorganisation energy, under which the Marcus equation pre-<br />

dicts the so-called <strong>in</strong>verted regime and rates become smaller with greater driv<strong>in</strong>g forces.<br />

This would lead to negative field dependences which are not experimentally observed.<br />

Also, we have ignored outer sphere reorganisation and as such it is highly unlikely that<br />

we underestimated the reorganisation energy.<br />

The hypothesis that mobility is dom<strong>in</strong>ated by high density regions is more compell<strong>in</strong>g.<br />

Studies show that the same thermal treatment which leads to higher µh also leads to an <strong>in</strong>-<br />

creased average density [2]. Unfortunately this experimental technique cannot dist<strong>in</strong>guish<br />

between an even or an uneven <strong>in</strong>crease <strong>in</strong> pack<strong>in</strong>g density. In order to test this hypothesis,<br />

we design morphologies with lattice constants <strong>in</strong> the range of 6.5 Å to 10 Å, with maximal<br />

lateral disorder. The <strong>in</strong>crease <strong>in</strong> average <strong>in</strong>tercha<strong>in</strong> distance from 6.5 Å to 10 Å should<br />

lead to a decrease <strong>in</strong> mobility of five orders of magnitude <strong>in</strong> an ordered system. However,<br />

<strong>in</strong> the presence of lateral disorder, we f<strong>in</strong>d that the mobility is reduced by less than three<br />

orders of magnitude; the reduction would be even smaller if longer polymer cha<strong>in</strong>s were<br />

used to run the Monte Carlo simulations.<br />

The third hypothesis, that the morphology of PFO is not hexagonal, rests on transmis-<br />

sion electron microscopy by S.H. Chen and coworkers [7], which came to our attention<br />

after our model had been designed. They f<strong>in</strong>d that the unit cell of PFO is rather large (or-<br />

thorhombic a=2.56 nm, b=2.34 nm, c=3.32 nm) compared to the unit cell we assumed.<br />

When they perform molecular mechanics simulation of 8 cha<strong>in</strong>s of PFO <strong>in</strong> such a unit<br />

cell, they obta<strong>in</strong> a morphology with <strong>in</strong>tercha<strong>in</strong> distances as short as approximately 6 Å.<br />

Our mobility calculation would be entirely consistent with such <strong>in</strong>tercha<strong>in</strong> distances. The<br />

last two hypotheses are not <strong>in</strong> disagreement with each other: <strong>in</strong> both cases we are argu-<br />

140


<strong>in</strong>g that there is experimental evidence refut<strong>in</strong>g a perfect hexagonal lattice structure <strong>in</strong><br />

PFO. At the moment it is still unclear whether the crystal structure from S.H. Chen and<br />

coworkers would fit experimental data, or whether lateral disorder on larger assemblies<br />

of molecules would be sufficient. In fact an <strong>in</strong>terest<strong>in</strong>g possibility from this study is that<br />

calculations of mobility can be used to test the validity of a morphology model.<br />

In conclusion, our <strong>in</strong>vestigation strongly suggests that charge transport <strong>in</strong> PFO is de-<br />

term<strong>in</strong>ed by <strong>in</strong>tercha<strong>in</strong> hops where the conjugated backbones are separated by a distance<br />

of approximately 6.5 Å. In order for charge transport to be determ<strong>in</strong>ed by the fastest <strong>in</strong>-<br />

tercha<strong>in</strong>, only a m<strong>in</strong>ority of <strong>in</strong>tercha<strong>in</strong> distances between charge transport<strong>in</strong>g units need<br />

to be so short.<br />

7.2 HBC<br />

Discotic thermotropic liquid crystals are formed by flat molecules with a central aromatic<br />

core and several aliphatic cha<strong>in</strong>s attached at its edges [8, 9, 10, 11, 12]. Discotics can<br />

form columnar phases, where the molecules stack on top of each other and the result<strong>in</strong>g<br />

columns are arranged <strong>in</strong> a regular lattice[9]. The self-organisation <strong>in</strong>to stacks with aro-<br />

matic cores surrounded by saturated hydrocarbons results <strong>in</strong> the onedimensional transport<br />

of charge with<strong>in</strong> the core, along columns [13, 14]. By modify<strong>in</strong>g the periphery or shape<br />

of the core it is possible to obta<strong>in</strong> materials with different electronic characteristics and<br />

phase behaviour. <strong>Charge</strong> transport occurs preferentially along the columns and the mo-<br />

bility <strong>in</strong> directions perpendicular to the column axis is much smaller, s<strong>in</strong>ce <strong>in</strong> this case<br />

the charges have to tunnel through the <strong>in</strong>sulat<strong>in</strong>g side cha<strong>in</strong>s. Therefore the discotics <strong>in</strong><br />

a columnar phase can be seen as nano-wires and an alternative to conjugated polymers,<br />

where <strong>in</strong>tracha<strong>in</strong> transport is fast but where charge mobility <strong>in</strong> devices is often limited by<br />

<strong>in</strong>efficient <strong>in</strong>tercha<strong>in</strong> hops.<br />

In this part of the study we simulate charge transport <strong>in</strong> different derivatives (shown<br />

<strong>in</strong> figure 7.8) of the discotic liquid crystal hexabenzocoronene (HBC) <strong>in</strong> a column and<br />

compare the results to pulse radiolysis time resolved microwave conductivity (PR-TRMC)<br />

studies of the mobility <strong>in</strong> these materials. The structure of the simulation we employ is<br />

very similar to that of the previous simulation on PFO: a morphology for stacks of HBC<br />

141


R<br />

R<br />

R<br />

R<br />

R<br />

R<br />

C12<br />

PhC12<br />

C10-6<br />

Figure 7.8: Chemical structure of the various derivatives of HBC considered <strong>in</strong> this study<br />

is generated, the transfer <strong>in</strong>tegrals are calculated for the nearest neighbours <strong>in</strong> the stack,<br />

the transfer <strong>in</strong>tegrals are used to calculate charge transfer rates us<strong>in</strong>g the Marcus charge<br />

transfer rate equation and f<strong>in</strong>ally the charge transfer rates are used to simulate charge<br />

transport us<strong>in</strong>g Monte Carlo and Master Equation methods. The ma<strong>in</strong> difference between<br />

this study and the study on polyfluorene is that the morphology model we use to determ<strong>in</strong>e<br />

the relative position and orientation of HBC molecules is not an empirical model, but is the<br />

result of an atomistic molecular dynamics simulation of HBC molecules. The result<strong>in</strong>g<br />

simulation is able to replicate the trends and magnitudes of mobilities measured <strong>in</strong> the<br />

different derivatives considered <strong>in</strong> this study.<br />

The use of the Marcus equation is justified by experiments by T. Kreouzis and co-<br />

workers [15] who found that Holste<strong>in</strong>’s small polaron model of charge transport is best<br />

capable of describ<strong>in</strong>g charge transport <strong>in</strong> the discotic liquid crystal triphenylene. The<br />

parameters required for the Marcus electron transfer equation, i.e. transfer <strong>in</strong>tegrals for<br />

nearest neighbours and reorganisation energies, are calculated <strong>in</strong> a manner very similar to<br />

the previous study on polyfluorene. The reorganisation energies for electrons and holes<br />

are calculated us<strong>in</strong>g the scheme described <strong>in</strong> section 2.2 us<strong>in</strong>g the functional B3LYP with<br />

the basis set 6-31g**. We f<strong>in</strong>d that the reorganisation energies are of 0.13 eV and 0.11<br />

eV for negative and positive charges respectively. A technical difference from the case<br />

of polyfluorene is that the HOMO and LUMO of HBC are doubly degenerate, thus we<br />

use an effective transfer <strong>in</strong>tegral as def<strong>in</strong>ed <strong>in</strong> section 2.3.1; that is we consider all four<br />

transfer <strong>in</strong>tegral between the HOMOs (LUMOs) and HOMO-1s (LUMO+1s) on either<br />

molecule then take the root mean square of the set.<br />

The molecular dynamics simulation of the morphology of HBC derivatives was car-<br />

142


ied out by Dr. V. Marcon and Dr. D. Andrienko at the Max Planck Institute for Polymer<br />

Research <strong>in</strong> Ma<strong>in</strong>z. Molecular Dynamics (MD) simulation is based on us<strong>in</strong>g parametrised<br />

potentials (force fields) to describe the bonded <strong>in</strong>teractions (harmonic bond, angle and tor-<br />

sions) and non-bonded <strong>in</strong>teraction (van der Waals and electrostatic <strong>in</strong>teractions) between<br />

the atoms of several different molecules of HBC. The potentials are then used to <strong>in</strong>tegrate<br />

Newton’s equations. In our case some atoms were substituted by pseudoatoms, namely<br />

the hydrogens <strong>in</strong> the aliphatic side cha<strong>in</strong>s were <strong>in</strong>corporated <strong>in</strong>to the carbons they were<br />

bonded to. The exact force field used is described <strong>in</strong> reference [16]. The molecules were<br />

arranged <strong>in</strong> 16 columns with 10, 20, 60 or 100 molecules <strong>in</strong> each column. The time step<br />

for <strong>in</strong>tegration of Newton’s equations of motion was 2 f s and each run lasted 100 ns. The<br />

temperature and pressure of the simulation were kept fixed at 300 K and 0.1MPa us<strong>in</strong>g<br />

the Berendsen method. A snapshot of the system was taken every 20 ps; such large time<br />

<strong>in</strong>tervals are necessary <strong>in</strong> order to ensure that the configurations are not correlated with<br />

one another. For each of these snapshots we recorded the relative position and orienta-<br />

tion between nearest neighbours <strong>in</strong> a column. The relative position and orientation are<br />

then used to calculate the effective transfer <strong>in</strong>tegral between two conjugated cores with<br />

a fixed <strong>in</strong>ternal geometry: <strong>in</strong> other words, before the MOO method is used to calculate<br />

transfer <strong>in</strong>tegrals the cores from the MD simulation are substituted with rigid copies of<br />

ones obta<strong>in</strong>ed from geometry relaxation of the neutral geometry. We consider only near-<br />

est neighbours on the same column because the distance between conjugated cores on<br />

separate columns is much larger than the distance between cores on the same column<br />

and therefore the transfer <strong>in</strong>tegral between columns will be negligibly small compared to<br />

transfer along the columns.<br />

Hav<strong>in</strong>g calculated the transfer <strong>in</strong>tegrals for a particular snapshot, we simulate time<br />

of flight (ToF) experiments us<strong>in</strong>g a cont<strong>in</strong>uous time random walk algorithm. The details<br />

of the simulation are as follows. A column of HBC molecules of a height of 1µm (cor-<br />

respond<strong>in</strong>g to approxiately 2800 molecules) is obta<strong>in</strong>ed by stack<strong>in</strong>g copies of the same<br />

stack from the MD simulation on top of each other. At the beg<strong>in</strong>n<strong>in</strong>g of the simulation<br />

a charge is positioned at random near the beg<strong>in</strong>n<strong>in</strong>g of the column. The motion of the<br />

charge and the result<strong>in</strong>g displacement current is calculated us<strong>in</strong>g the cont<strong>in</strong>uous time ran-<br />

143


Displacement current<br />

1.8<br />

1.6<br />

1.4<br />

1.2<br />

1<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

TOF, s<strong>in</strong>gle configurations<br />

C12, 300K, 10V<br />

1<br />

10<br />

20<br />

30<br />

40<br />

50<br />

60<br />

70<br />

80<br />

90<br />

100<br />

average (190)<br />

0<br />

0 5e-10 1e-09 1.5e-09<br />

time (sec)<br />

2e-09 2.5e-09 3e-09<br />

Displacement current<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

TOF block averages<br />

C12, 300K, 1000 Nsims<br />

0<br />

0 1e-09 2e-09<br />

time (sec)<br />

Figure 7.9: ToF results for several selected snapshots, together with the averaged over<br />

the MD run displacement current (left). Block averages over 10, 20, ... 190<br />

snapshots. (right)<br />

dom walk Monte Carlo (MC) algorithm described <strong>in</strong> section 2.5.1. Two thousand such<br />

displacement currents are then averaged to ensure that the displacement current reflects<br />

the properties of that column.<br />

Several displacement currents are shown <strong>in</strong> the left panel of figure 7.9. They were<br />

obta<strong>in</strong>ed for different columns from different snapshots for the C12 derivative with stacks<br />

10 molecules high. It can be noticed that the transit time is rather different between<br />

different columns, hence it will be necessary to average the displacement currents for<br />

different columns. The averaged current transients fall off less rapidly than the transients<br />

for <strong>in</strong>dividual columns; this dispersion is due to the distribution <strong>in</strong> arrival times for the<br />

different columns. The right panel of figure 7.9 shows that 200 snapshots are sufficient<br />

to obta<strong>in</strong> a converged displacement current. The observation of variation <strong>in</strong> mobility for<br />

different columns is the result of the one dimensional nature of transport along columns<br />

of HBC: mobility is heavily determ<strong>in</strong>ed by the smallest transfer <strong>in</strong>tegral, whose value<br />

fluctuates between different columns.<br />

Figure 7.10 shows the variation of displacement currents for columns composed of<br />

different length stacks of the C12 derivative. Note that the size of column that the MC<br />

simulation is run over is unchanged, what changes is the number of molecules <strong>in</strong> each<br />

column over which MD is carried out. As the number of molecules <strong>in</strong> a stack becomes<br />

larger, the transit time <strong>in</strong>creases slightly. This aga<strong>in</strong> can be understood by remember<strong>in</strong>g<br />

that the smallest transfer <strong>in</strong>tegral sets the mobility: as stack length <strong>in</strong>creases the chances<br />

144<br />

10<br />

30<br />

50<br />

70<br />

90<br />

110<br />

130<br />

150<br />

170<br />

190


Displacement current<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

0<br />

TOF, column length dependence<br />

C12, 400K, 200 configurations average, 10V<br />

10 molecules<br />

20 molecules<br />

100 molecules<br />

0 1e-09 2e-09 3e-09<br />

time (sec)<br />

Figure 7.10: Displacement current for different numbers of molecules <strong>in</strong> a column. Average<br />

over 200 snapshots at an electric field of 10 5 Vcm −1 .<br />

of smaller transfer <strong>in</strong>tegrals occur<strong>in</strong>g <strong>in</strong>creases. A stack length of 100 molecules seems<br />

sufficient to converge a value of the transit time and will be used throughout <strong>in</strong> the rest of<br />

the study. In the next part of the study we calculate mobilities for holes and electrons for<br />

all the derivatives shown <strong>in</strong> figure 7.8.<br />

Figure 7.11: Radial distribution function along the z direction for all six derivatives considered<br />

<strong>in</strong> this study.<br />

Hav<strong>in</strong>g established the size of the system to use for the Monte Carlo simulation, let us<br />

compare the morphology of the different materials we look at <strong>in</strong> this study. In particular<br />

we consider two measures, the centre of mass separation between neighbour<strong>in</strong>g molecules<br />

<strong>in</strong> a column, h, and the nematic order parameter, Q. Figure 7.11 shows the radial distri-<br />

bution function for all six derivatives, that is the probability of f<strong>in</strong>d<strong>in</strong>g a molecule at a<br />

particular distance along the z axis, def<strong>in</strong>ed as the symmetry axis of the stack. The four<br />

145


derivatives with unbranched aliphatic cha<strong>in</strong>s C10, C12, C14 and C16 all show a high de-<br />

gree of order and have a similar mode and average <strong>in</strong>termolecular separation of 3.6 Å.<br />

The derivative with the branched side cha<strong>in</strong> C10-6 is considerably more disordered. Al-<br />

though the mode distance to the nearest neighbour (the position of the first peak <strong>in</strong> figure<br />

7.11) is identical to that for unbranched side cha<strong>in</strong>s, the probability does not fall to zero<br />

for positions away from the crystal positions. As we shall see <strong>in</strong> the next paragraph this<br />

is a consequence of misalignment. The derivative with phenyl r<strong>in</strong>gs Ph-C12 shows larger<br />

nearest neighbour distances than the unbranched side cha<strong>in</strong>s, but is more crystall<strong>in</strong>e than<br />

the branched side cha<strong>in</strong>.<br />

The nematic order parameter Q is obta<strong>in</strong>ed by calculat<strong>in</strong>g the vector normal to the<br />

molecular plane of the i th HBC molecule u (i) . A matrix Qαβ is then calculated by tak<strong>in</strong>g:<br />

Qαβ =<br />

� 1<br />

N<br />

N�<br />

i=1<br />

�<br />

3<br />

2 u(i) α u (i) 1<br />

β −<br />

2 δαβ<br />

��<br />

(7.1)<br />

This matrix is diagonalized and the greatest eigenvalue is called the nematic order<br />

paramater Q. The maximum value that the order parameter can take is 1, denot<strong>in</strong>g perfect<br />

order<strong>in</strong>g. The m<strong>in</strong>imimum is 0, denot<strong>in</strong>g an isotropic arrangements. Table 7.1 shows the<br />

order parameter Q for the various HBC derivatives considered. Q takes very large val-<br />

ues for all the unbranched aliphatic derivatives, and slightly smaller ones for the C10-6<br />

and Ph-C12 derivatives respectively. This confirms the <strong>in</strong>formation from the radial distri-<br />

bution function, that C10-6 and Ph-C12 form less well ordered stacks than the aliphatic<br />

unbranched side cha<strong>in</strong> derivatives. The fact that columns of C10-6 have a slightly higher<br />

order parameter than Ph-C12, suggests that C10-6 stacks are composed of short segments<br />

of well ordered molecules, which do not stack well with each other. The small mode<br />

distance <strong>in</strong> C10-6 stacks is expla<strong>in</strong>ed by the fact that most molecules are <strong>in</strong> well ordered<br />

segments, but the badly def<strong>in</strong>ed peaks at greater distances correspond to the fact that these<br />

segments are misaligned.<br />

Figure 7.12 shows histograms of the values of the calculated hole transfer <strong>in</strong>tegrals<br />

for the six derivatives. As can be seen, the distribution of transfer <strong>in</strong>tegrals for each of the<br />

unbranched side cha<strong>in</strong> derivatives is very similar. Ph-C12 has a wider peak and a smaller<br />

mean than the aliphatic side cha<strong>in</strong>s reflect<strong>in</strong>g the greater degree of disorder <strong>in</strong> columns<br />

146


compound µPR−TRMC µ e<br />

ToF µ h<br />

ToF µ e<br />

ME µ h<br />

C10 0.5 [17] 0.22 0.75 0.14<br />

ME<br />

0.49<br />

Q<br />

0.98<br />

h<br />

3.6<br />

C12 0.9 [17] 0.23 0.76 0.14 0.49 0.98 3.6<br />

C14 1 [17] 0.27 0.80 0.17 0.59 0.98 3.6<br />

C16 - 0.29 0.91 0.17 0.58 0.98 3.6<br />

C10−6 0.08 [18] - 0.01 6e-4 0.003 0.96 3.7<br />

PhC12 0.2 [17] 0.036 0.13 0.012 0.045 0.95 4.1<br />

Table 7.1: Electron (e) and hole (h) mobilities (cm 2 V −1 s −1 ) of different compounds calculated<br />

us<strong>in</strong>g time-of-flight (ToF) and master equation (ME) methods, <strong>in</strong> comparison<br />

with experimentally measured PR-TRMC mobilities. Also shown are the<br />

nematic order parameter Q and the average vertical separation between cores<br />

of HBC molecules h (Å).<br />

and the greater separation between molecules. C10-6 has a peak <strong>in</strong> values of log(|J 2 |)<br />

with values which are actually slightly larger than those of the aliphatic unbranched side<br />

cha<strong>in</strong>s, reflect<strong>in</strong>g the similar value of the mode of the separation between nearest neigh-<br />

bours. However, C10-6 has also a tail of low values of log(|J 2 |), reflect<strong>in</strong>g the high degree<br />

of disorder <strong>in</strong> the columns. These histograms expla<strong>in</strong> the trend <strong>in</strong> calculated mobilities<br />

shown <strong>in</strong> table 7.1. Because of the one dimensional nature of transport <strong>in</strong> HBC columns,<br />

mobility is determ<strong>in</strong>ed by the lowest transfer <strong>in</strong>tegrals <strong>in</strong> the column. For this reason<br />

C10-6 has smaller mobilities than any other derivative, even though it has the highest<br />

mode transfer <strong>in</strong>tegrals. Table 7.1 shows mobility as calculated by both a Monte Carlo<br />

and a Master Equation formalism and compares the mobilities to results from PR-TRMC,<br />

which probes the sum of electron and hole mobility. Our calculations match both the<br />

trends and magnitudes of the charge mobility, with no adjustable parameters.<br />

7.3 Conclusion<br />

In this chapter we have shown calculations of electric field dependence of mobility for two<br />

materials. We argue that the fact that we are able to reproduce experimental mobility for<br />

HBC with no free parameters proves the validity of the methods used to calculate transfer<br />

<strong>in</strong>tegrals and reorganisation energies. Agreement with experiment could also be obta<strong>in</strong>ed<br />

<strong>in</strong> the case of PFO, however the morphology model required empirical parameters.<br />

147


counts<br />

200<br />

150<br />

100<br />

50<br />

200<br />

150<br />

100<br />

C10<br />

0<br />

-8 -6 -4 -2 0<br />

50<br />

200<br />

150<br />

100<br />

0<br />

-8 -6 -4 -2 0<br />

50<br />

C12<br />

C14<br />

0<br />

-8 -6 -4 -2 0<br />

200<br />

150<br />

100<br />

50<br />

C10_6<br />

0<br />

-8 -6 -4 -2 0<br />

70<br />

60<br />

50<br />

40<br />

30<br />

20<br />

10<br />

0<br />

-8 -6<br />

C12_Ph<br />

-4 -2 0<br />

200<br />

150<br />

100<br />

50<br />

2<br />

log (J ) 10 eff<br />

C16<br />

0<br />

-8 -6 -4 -2 0<br />

Figure 7.12: Distribution <strong>in</strong> logarithm of the effective transfer <strong>in</strong>tegrals for holes squared<br />

for 20 columns of 100 molecules each.<br />

148


Bibliography<br />

[1] A.J. Chatten, S.M. Tuladhar, S.A. Choulis, D.D.C. Bradley, and J. Nelson, Monte<br />

Carlo modell<strong>in</strong>g of hole transport <strong>in</strong> MDMO-PPV:PCBM blends, Journal of Materials<br />

Science, 40, 1393 (2005)<br />

[2] T. Kreouzis, D. Poplavskyy, S.M. Tuladhar, M. Campoy-Quiles, J. Nelson, A.J.<br />

Campbell, and D.D.C. Bradley, Temperature and field dependence of hole mobility<br />

<strong>in</strong> poly(9,9-dioctylfluorene), Physical Review B 73, 235201 (2006)<br />

[3] M. Redecker, D.D.C. Bradley, M. Inbasekaran and E.P. Woo, Mobility enhancement<br />

through homogeneous nematic alignment of a liquid-crystall<strong>in</strong>e polyfluorene, Applied<br />

Physics Letters 74, 1400 (1999)<br />

[4] S. Kawana, M. Durrell, J. Lu, J.E. MacDonald, M. Grell, D.D.C. Bradley, P.C. Jukes,<br />

R.A.L. Jones and S.L. Bennett, X-Ray diffraction study of the structure of th<strong>in</strong> polyflu-<br />

orene films, Polymer, 43, 1907 (2002)<br />

[5] Y. Shimoi, S. Abe, S. Kuroda and K. Murata, Polarons and their ENDOR spectra <strong>in</strong><br />

poly p-phenylene v<strong>in</strong>ylene, Solid State Communications 95, 137 (1995)<br />

[6] M. Knupfer, J. F<strong>in</strong>k and D. Fichou, Strongly conf<strong>in</strong>ed polaron excitations <strong>in</strong> charged<br />

organic semiconductors, Physical Review B, 63, 165203 (2001)<br />

[7] S. H. Chen, H.L. Chou, A.C. Su and S.A. Chen, Molecular Pack<strong>in</strong>g <strong>in</strong> Crystall<strong>in</strong>e<br />

Poly(9,9-di-n-octyl-2,7-fluorene), Macromolecules, 37, 6833 (2004)<br />

[8] M. Müller, C. Kubel and K. Müllen, Giant polycyclic aromatic hydrocarbons, Chem-<br />

istry -A European Journal 4, 2099 (1998)<br />

149


[9] S. Chandrasekhar and G.S. Ranganath, Discotic liquid-crystals, Reports on Progress<br />

<strong>in</strong> Physics, 53, 57 (1990)<br />

[10] J.D. Brand, C. Kubel, S. Ito and K. Müllen, Functionalized hexa-peri-<br />

hexabenzocoronenes: Stable supramolecular order by polymerisation <strong>in</strong> the discotic<br />

mesophase, Chemistry Of Materials , 12, 1638 (2000)<br />

[11] R.J. Bushby and O.R. Lozman, Discotic liquid crystals 25 years on, Current Op<strong>in</strong>ion<br />

In Colloid and Interface Science, 7, 343 (2002)<br />

[12] S. Kumar, Recent developments <strong>in</strong> the chemistry of triphenylene-based discotic liq-<br />

uid crystals, Liquid Crystals, 31, 1037 (2004)<br />

[13] A.M. van de Craats, J.M. Warman, A. Fechtenkotter, J.D. Brand, M.A. Harbison<br />

and K. Müllen, Record charge carrier mobility <strong>in</strong> a room-temperature discotic liquid-<br />

crystall<strong>in</strong>e derivative of hexabenzocoronene, Advanced Materials, 11, 1469 (1999)<br />

[14] L. Schmidt-Mende, A. Fechtenkotter, K. Müllen, E. Moons, R. H. Friend and J. D.<br />

MacKenzie, Self-organized discotic liquid crystals for high-efficiency organic photo-<br />

voltaics, Science, 293, 1119 (2001)<br />

[15] T. Kreouzis, K.J. Donovan, N. Boden, R.J. Bushby and O.R. Lozman, Temperature-<br />

<strong>in</strong>dependent hole mobility <strong>in</strong> discotic liquid crystals, Journal of Chemical Physics, 114,<br />

1797 (2001)<br />

[16] D. Andrienko, V. Marcon and K. Kremer, Atomistic simulation of structure and<br />

dynamics of columnar phases of hexabenzocoronene, Journal of Physical Chemistry,<br />

125, 124902 (2006)<br />

[17] A.M. van de Craats, L.D.A. Siebbeles, I. Bleyl, D. Haarer, Y.A. Berl<strong>in</strong>, A.A.<br />

Zharikov and J.M. Warman, Mechanism of <strong>Transport</strong> along Columnar Stacks of a<br />

Triphenylene Dimer, Journal of Physical Chemistry B, 102, 9625 (1998)<br />

[18] W. Pisula, M. Kastler, D. Wasserfallen, M. Mondeshki, J. Piris, I. Schnell and<br />

K. Müllen, Relation between Supramolecular Order and <strong>Charge</strong> Carrier Mobility of<br />

Branched Alkyl Hexa-peri-hexabenzocoronenes, Chemical Materials, 18, 3634 (2006)<br />

150


Chapter 8<br />

Conclusion and Future Work<br />

We have discussed the role of electronic structure theory calculations <strong>in</strong> help<strong>in</strong>g to under-<br />

stand charge transport characteristics of conjugated materials. The start<strong>in</strong>g po<strong>in</strong>t of this<br />

approach is the Marcus charge transfer theory rate equation, which takes as parameters<br />

three quantities: the transfer <strong>in</strong>tegral, the reorganisation energy and the site energy dif-<br />

ference. Ideally we would calculate the charge transfer rates for all nearest neighbours<br />

<strong>in</strong> a material and would then simulate charge dynamics <strong>in</strong> a realistic simulation. S<strong>in</strong>ce<br />

charge dynamics simulation volumes can be very large, it is of paramount importance<br />

to be able to calculate the parameters of the Marcus charge transfer equation us<strong>in</strong>g fast<br />

methods. For this reason we developed methods to calculate transfer <strong>in</strong>tegrals and site<br />

energy differences for pairs of molecules without perform<strong>in</strong>g SCF calculations.<br />

In chapter five we applied these methods to justify the choice of parameters to be used<br />

<strong>in</strong> a lattice simulation of charge transport <strong>in</strong> blends of MDMO-PPV:PCBM. In chapter<br />

six, calculations were performed on pairs of molecules <strong>in</strong> random relative orientation,<br />

with the aim of compar<strong>in</strong>g the distributions <strong>in</strong> transfer <strong>in</strong>tegrals and site energy difference<br />

to the distributions obta<strong>in</strong>ed us<strong>in</strong>g the Gaussian Disorder Model. In both these applica-<br />

tions, the aim of our calculations was qualitative. Chapter seven conta<strong>in</strong>s two full-blown<br />

simulations of charge transport <strong>in</strong> a conjugated polymer (polyfluorene) and <strong>in</strong> a liquid<br />

crystal (hexabenzocoronene). The study on hexabenzocoronene demonstrates that if a<br />

good morphology model is available, it is possible to quantitatively predict the mobility<br />

of a material.<br />

Future work will be directed towards two ma<strong>in</strong> areas: extend<strong>in</strong>g the work on hex-<br />

151


abenzocoronene to polymers and simulat<strong>in</strong>g temperature as well as field dependence of<br />

mobility. Two challenges will have to be resolved for a treatment of charge transport <strong>in</strong><br />

polymers: first, calculat<strong>in</strong>g morphologies for polymers will be substantially harder than<br />

for hexabenzocoronene and second the issue of charge delocalisation along the polymer<br />

cha<strong>in</strong> discussed <strong>in</strong> chapter four will have to be resolved. In order to simulate the tem-<br />

perature dependence of mobility it will be necessary to calculate site energies accurately,<br />

calculat<strong>in</strong>g electrostatic and <strong>in</strong>ductive <strong>in</strong>teractions. This will require extend<strong>in</strong>g the meth-<br />

ods from chapter two to treat slabs of material rather than pairs of molecules only.<br />

152

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!