1: \documentclass[prl,twocolumn,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}

2: \usepackage{graphicx}

3: 

4: \begin{document}

5: 

6: \title{Structural relaxations in electronically excited

7: poly(\textit{para}-phenylene)}

8: 

9: 

10: \author{Emilio~Artacho$^1$, M.~Rohlfing$^2$, M.~C\^ot\'e$^3$, 

11: P.~D.~Haynes$^4$, R.~J.~Needs$^4$, and C.~Molteni$^5$}

12: \affiliation{

13: $^1$Department of Earth Sciences, University of Cambridge,

14:       Downing Street, Cambridge, CB2 3EQ, UK\\

15: $^2$School of Engineering and Science,

16:       International University Bremen, P.O.Box 750561,

17:       28725 Bremen, Germany\\

18: $^3$D\'epartement de Physique, Universit\'e de Montr\'eal, C.P.\ 6128, 

19:       succ.\ Centre-Ville, Montreal, Quebec, H3C 3J7, Canada\\

20: $^4$TCM Group, Cavendish Laboratory, University of Cambridge,

21:       Madingley Road, Cambridge, CB3 0HE, UK\\

22: $^5$Physics Department, King's College London, Strand, London WC2R 2LS, UK

23: }

24: 

25: \date{6 Februrary 2004}

26: 

27: \begin{abstract}

28: Structural relaxations in electronically excited

29: poly(\textit{para}-phenylene) are studied using many-body perturbation

30: theory and density-functional-theory methods.  A sophisticated

31: description of the electron-hole interaction is required to describe

32: the energies of the excitonic states, but we show that the structural

33: relaxations associated with exciton formation can be obtained quite

34: accurately within a constrained density-functional-theory approach.

35: We find that the structural relaxations in the low-energy excitonic

36: states extend over about 8 monomers, leading to an energy reduction of

37: 0.22 eV and a Stokes shift of 0.40 eV.

38: \end{abstract}

39: 

40: \pacs{71.15Mb, 71.15.Qe, 71.20.Rv, 71.35.Aa}

41: 

42: \maketitle

43: 

44: The sustained growth of interest in conjugated polymers arises from

45: their potential as the active material in low-cost field-effect

46: transistors, photovoltaic devices and light emitting diodes

47: (LEDs).\cite{greenham_1995,heeger_2001} Investigating electronically

48: excited states is an important part of this effort, which has proved a

49: great challenge for theoretical techniques.  An accurate description

50: of the excited electronic states of conjugated polymers requires the

51: inclusion of the strong electron-hole interaction and the structural

52: relaxations which occur in the excited state.

53: 

54: In its simplest form, a polymer LED consists of a layer of conjugated

55: polymer sandwiched between two electrodes.  Electrons and holes are

56: injected into the polymer where they are attracted to one another and

57: form bound excitons. Light is then emitted by exciton recombination.

58: In an excited state the polymer may lower its energy by structural

59: relaxations so that the peak optical emission frequency is lower than

60: the peak absorption frequency.  Although this ``Stokes shift'' can, at

61: least in principle, be measured, the actual relaxations have not been

62: determined experimentally.  These relaxations localize excitons and

63: are relevant for the technologically important processes of exciton

64: migration and recombination.

65: 

66: We have investigated structural relaxations in the low-energy excited

67: states of poly(\textit{para}-phenylene) (PPP).  A PPP LED has been

68: demonstrated which emits blue light in a band around

69: 2.7~eV.\cite{grem_1992,grem_1993} PPP is, however, insoluble and

70: therefore difficult to process and, instead, soluble derivatives of

71: PPP with various side groups are preferred for manufacturing LEDs.

72: The low-energy optical properties of PPP and its derivatives are

73: similar, although the structural relaxations in the excited states

74: depend on the nature of the side groups.  We have chosen to study PPP

75: because it has a simpler structure than its derivatives.

76: 

77: \begin{figure}[ht]

78: \centering\includegraphics[width=7cm]{ppp_structure.eps}

79: \caption{The structure of a PPP chain with the C atoms represented by

80: filled circles and the H atoms by open circles.  (a) Side view showing

81: the A and B bonds.  (b) End view showing the torsional angle

82: $\theta$.}

83: \label{fig:structure}

84: \end{figure}

85: 

86: As we show below, systems containing more than 100 atoms are required

87: for studying the excited-state relaxations of PPP, which pose a

88: significant problem to theory.  Such systems can easily be treated by

89: density-functional theory (DFT) \cite{footnote_lda}, but DFT alone

90: often has difficulties in describing excited states.  Excited states

91: can be described by many-body perturbation theory (MBPT) \cite{bse},

92: but this is numerically demanding and cannot currently be applied to

93: such large systems. However, we find that the forces in the excited

94: states of PPP obtained within DFT are very similar to those within

95: MBPT, which indicates that the relaxed structures and Stokes shift

96: obtained within DFT are reliable.

97: 

98: A PPP chain (Fig.~\ref{fig:structure}) consists of phenyl rings joined

99: by single C-C bonds, with adjacent rings being rotated with respect to

100: one another by a torsional angle $\theta$.  Our DFT structures for the

101: ground states of the single chain and crystal are in excellent

102: agreement with a previous DFT calculation and in accord with

103: experiment.\cite{ambrosch-draxl_1995} Our DFT band gap for the single

104: chain is 2.44 eV and that for the crystal is 1.70 eV, which is in good

105: agreement with a previous DFT calculation.\cite{ambrosch-draxl_1995}

106: 

107: We began by using a simple model of how an optical excitation affects

108: the bonding in PPP: we performed DFT calculations of the changes in

109: the electronic charge distribution due to a single excitation from the

110: highest occupied molecular orbital (HOMO) to the lowest unoccupied

111: molecular orbital (LUMO). In both the isolated chain and crystalline

112: forms we found an increase in the electronic charge in the A bonds and

113: to a lesser extent in the B bonds (see Fig.~\ref{fig:structure}).

114: These changes indicate the development of double-bonding character

115: which is expected to favor both a more planar arrangement of rings

116: (reduction in $\theta$) and compression of the A and B bonds.

117: 

118: In a perfect chain or crystal an exciton would, in principle, be

119: delocalized because localization within some region of space costs

120: kinetic energy.  However, a larger energy may be gained by structural

121: relaxations so that an exciton may localize itself.  The low-energy

122: excitonic states of PPP may therefore be self-localized by structural

123: relaxations, i.e., local reductions in $\theta$ and the lengths of the

124: A and B bonds.

125: 

126: The structural relaxations in the lowest excited states of a single

127: PPP chain can be investigated using a constrained DFT method.  To this

128: end we model the excited state by promoting an electron from the HOMO

129: to the LUMO at wave vector $k=0$ and performing full structural

130: relaxations in unit cells of various sizes.  A unit cell containing 28

131: monomers was found to be sufficient to converge the excited state

132: structure.  The important parameters of the minimum energy structure

133: are shown in Fig.~\ref{fig:bond+torsion}.  The relaxations are largely

134: confined to a region of approximately 8 monomers over which the

135: torsional angle and the lengths of the A bonds are significantly

136: reduced, and the lengths of the B bonds are somewhat reduced.  The

137: gain in energy from the relaxation is 0.22 eV, and the Stokes shift,

138: calculated as the difference between the band gaps of the ground and

139: excited state structures, is 0.40 eV.\cite{footnote_gaps}

140: 

141: We have analyzed the localization of the exciton in terms of the

142: compression of the A and B bonds and the reduction in the torsional

143: angle, $\theta$.  We performed two constrained DFT calculations for

144: the excited state in which (1) only relaxations along the axis of the

145: chain and (2) only relaxations in the plane perpendicular to the chain

146: were allowed.  The relaxation energies of 0.14 eV and 0.17 eV,

147: respectively, add up to significantly more than the full relaxation

148: energy of 0.22 eV, indicating that the modes are strongly coupled.

149: The relaxations along the chain led to a Stokes shift of only 0.06 eV

150: and to very weak localization of the exciton, whereas the relaxations

151: perpendicular to the chain gave a Stokes shift of 0.16 eV and much

152: stronger localization.  The torsional relaxation is therefore very

153: important in determining the Stokes shift, although the coupling to

154: the bond compression is strong and both types of relaxation are

155: important in determining the full relaxation energy and Stokes shift.

156: 

157: \begin{figure}[ht]

158: \centering\includegraphics[width=8cm]{bond+torsion.eps}

159: \caption{The torsional angle, $\theta$, and the lengths of the A and B

160: bonds (see Fig.~\ref{fig:structure}) as a function of monomer number

161: for the excited state of a single PPP chain.}

162: \label{fig:bond+torsion}

163: \end{figure}

164: 

165: Stokes shifts in PPP derivatives have been measured in dilute

166: solutions and solid films.  A common approach is to use ``laddered''

167: PPP derivatives in which some or all of the phenyl rings are joined to

168: their neighbors by chemical ``bridges''.  These bridges hinder

169: rotation about the A bonds which, according to our picture, is

170: expected to reduce the Stokes shift.  Hertel \textit{et

171: al.}~\cite{hertel_2001} report absorption and emission measurements on

172: a series of laddered and non-laddered polymers, which clearly show

173: that laddering reduces the Stokes shift. As shown in Fig.~3b of Hertel

174: \textit{et al.}~\cite{hertel_2001}, in dilute solution the

175: non-laddered polymer dodecyloxy-poly(\textit{para}-phenylene)

176: (DDO-PPP) exhibits a Stokes shift of 0.6$\pm 0.2$ eV.  Considering the

177: presence of the solvent and different side chains in the experimental

178: system the agreement with our single chain PPP value of 0.40 eV is

179: reasonable.  The measured Stokes shifts for the laddered polymers are

180: smaller, in agreement with our picture that changes in $\theta$ couple

181: strongly to the band gap.  Our picture also suggests that structural

182: relaxations for excitons in laddered PPP derivatives should only lead

183: to a weak localization effect.

184: 

185: We now investigate the reliability of the constrained-DFT results for

186: the relaxation and Stokes shift of the exciton.  The formation of a

187: completely delocalized exciton leads to a vanishingly small change in

188: the charge density and therefore vanishingly small forces on the

189: atoms.  If, however, a lattice distortion occurs which reduces the

190: band gap in some region of space then the exciton will be attracted to

191: this region.  If the lowering of the energy of the exciton is greater

192: than the energy required to form the distortion then the exciton will

193: stabilize the distortion, leading to a self-localized exciton.  The

194: reduction in the band gap due to a lattice distortion is an effect

195: which is described approximately within our DFT calculations.  The

196: excitonic energies themselves, however, are strongly modified by

197: many-body effects arising from the electron-hole interaction.  We now

198: investigate the dependence of these many-body effects on the lattice

199: distortion.  If this dependence is sufficiently weak the constrained

200: DFT approach will be reliable.

201: 

202: A rigorous approach to electronic many-body effects for excited

203: electronic states is given by many-body perturbation theory

204: (MBPT).\cite{bse} We have used the $GW$ approximation to describe the

205: addition or removal of an electron and the Bethe-Salpeter equation

206: (BSE) for the excitation of an electron including the electron-hole

207: interaction.  These techniques have recently been used to describe the

208: excitonic states of solids~\cite{rohlfing_2000},

209: clusters~\cite{rohlfing_1998}, and polymers.\cite{rohlfing_1999,horst}

210: MBPT is numerically very demanding and cannot currently be applied

211: directly to a polymer with a self-localized exciton.

212: 

213: To investigate the effects of electronic correlation and validate the

214: constrained-DFT approach we proceeded as follows.  We performed DFT

215: and MBPT calculations for a series of structures involving torsional

216: angles and compressions of the A and B bonds which are similar to

217: those found in the self-localized exciton of

218: Fig.~\ref{fig:bond+torsion}.  We considered the relaxation of the two

219: central monomers of Fig.~\ref{fig:bond+torsion} as indicating the

220: maximum relaxations which occur in the exciton.  We then periodically

221: repeated this structure, obtaining a reference system which can be

222: investigated within both constrained DFT and MBPT.  We defined a

223: structural parameter $x$ which takes the value $x$=0 for the ground

224: state structure ($\theta$=33.7$^{\circ}$, A=1.456 \AA, and B=1.380

225: \AA), and $x$=1 for the geometry corresponding to the central monomers

226: of the self-localized exciton ($\theta$=9.1$^{\circ}$, A=1.427 \AA,

227: and B=1.366 \AA).  Intermediate values of $x$ correspond to linearly

228: interpolating $\theta$, A and B between the extremal values.

229: Fig.~\ref{fig:energy_x} shows the excitation energies of this periodic

230: structure as a function of $x$.  The four curves denote the DFT energy

231: gap, the quasiparticle (QP) energy gap, and the transition energies of

232: the lowest spin-singlet and spin-triplet excitons obtained from

233: solving the BSE.\cite{note_bse_technical} The QP gaps are 2.3-3.0 eV

234: larger than the DFT gaps due to the significant QP corrections typical

235: of semiconducting systems.  The exciton energies, on the other hand,

236: are 1.5-2.9 eV smaller than the QP gaps due to the attractive

237: electron-hole interaction.  The singlet (triplet) excitation energy

238: ranges from 3.41 eV (2.62 eV) at $x$=0 to 2.18 eV (1.61 eV) at $x$=1.

239: 

240: The change in the excitation energy when going from $x$=0 to $x$=1

241: constitutes the excited-state contribution to the force on the

242: structure projected onto the relaxations studied.  The most important

243: feature of Fig.~\ref{fig:energy_x} is that this change (--1.22 eV or

244: -1.01 eV, respectively, for the singlet and triplet state) is very

245: similar to the change in the DFT gap energy (--0.95 eV).  Changes in

246: the DFT gap energy, in turn, correspond to the excited-state forces

247: given by constrained DFT.  Our results therefore imply that the forces

248: are indeed given reliably by the constrained DFT theory, even though

249: the exciton binding energies are large and the absolute values of the

250: DFT band gaps are in poor agreement with the excitonic

251: energies.\cite{note:singlet-triplet_relaxations}

252: 

253: The singlet excitation energy for the ground state structure of the

254: single chain (3.41 eV) is in good agreement with experimental

255: absorption spectra which show maximum absorption around 3.5

256: eV.\cite{grem_1992,grem_1993} We find that the singlet-triplet

257: splitting depends rather weakly on the relaxations, changing from 0.79

258: eV at $x$=0 to 0.57 eV at $x$=1. We therefore expect the splitting for

259: the relaxed excited state structure to be in the region of 0.5 eV,

260: which is consistent with the splitting of 0.7-0.8 eV obtained from

261: emission spectra of PPP derivatives.\cite{hertel_2001}

262: 

263: \begin{figure}[ht]

264: \centering\includegraphics[width=8cm]{energy_x.eps}

265: \caption{Excitation energies calculated within the DFT and $GW$

266: approaches and the $GW$-BSE singlet and triplet energies as a function

267: of the structural parameter $x$ for a single PPP chain.}

268: \label{fig:energy_x}

269: \end{figure}

270: 

271: We have analyzed the reasons for the success of the constrained DFT

272: approach for calculating excited state geometries.  The details of

273: this analysis will be published elsewhere, but we conclude that the

274: various approximations should work well for the chain or solid when

275: the excitons are delocalized over many atoms and when the relaxations

276: are relatively small and do not include reconstructive changes such as

277: bond breaking.  The constrained DFT approach would be inaccurate if

278: the excitonic wave function contained contributions from more than one

279: electron or hole band, but this mixing could be calculated from BSE

280: calculations on high symmetry structures and then used in DFT studies

281: of large structures.  Our study suggests that the constrained DFT

282: approach is likely to work for other conjugated polymers.

283: 

284: Our main calculations are for single chains, in which the

285: electron-hole interaction is larger than in solids.  Therefore our

286: calculations are an even stronger test of the idea of neglecting the

287: electron-hole interaction when calculating the excited state

288: relaxations than would be encountered in the more technologically

289: relevant solid state.  In the solid, many chains are packed together

290: at van der Waals distances, which gives rise to three main additional

291: features: (i) electronic overlap between chains, (ii) van der Waals

292: interactions, and, in particular, (iii) the effect of inter-chain

293: dielectric screening on the electron-hole interaction.  The first of

294: these is well described within DFT while MBPT includes the second and

295: third.

296: 

297: To study the effects of inter-chain dielectric screening we considered

298: a 3D crystalline array of PPP chains in which the electronic overlap

299: between chains is small and consequently the DFT results are only

300: weakly perturbed.  The calculated QP correction to the DFT gap and the

301: electron-hole interaction are, however, both reduced by more than 1~eV

302: as a consequence of the more effective dielectric screening in a 3D

303: solid.  The QP state, which describes an additional electron or hole

304: on one chain, shows significant interaction with the polarizable

305: neighboring chains, and the QP gap closes.  In addition, the QP

306: corrections also depend on $x$, indicating that the intra-chain

307: screening changes when the geometry relaxes.  The exciton energies

308: are, however, only weakly affected by the changes in the inter- or

309: intra-chain screening.  In the solid the singlet (triplet) energy

310: ranges from 3.66 eV (3.14 eV) at $x$=0 to 2.55 eV (2.23 eV) at $x$=1,

311: which is only a little higher than the single-chain results.  This

312: insensitivity derives from the charge-neutral character of the

313: exciton, which is much less influenced by electrostatic screening than

314: charged single-particle excitations.  Similar effects have also been

315: observed for other polymers by van der Horst \textit{et

316: al}.\cite{horst} The changes in the excitation energies as $x$ varies

317: from 0 to 1 (--1.11 eV for the singlet, --0.91 eV for the triplet) are

318: very similar to the change of the DFT gap (--0.95 eV), confirming the

319: reliability of the constrained-DFT forces.  The weak interactions

320: between chains have to be handled with care, however, and the

321: self-trapping of the exciton in the solid state will be the subject of

322: a further study.

323: 

324: In summary, we have shown that while the electron-hole interaction is

325: very important in determining the excitonic energies in PPP it is less

326: important for the excited state geometries.  We have demonstrated that

327: a simple constrained DFT approach is sufficient for calculating

328: relaxations in the low-energy excited states of PPP.  In this case we

329: find that the structural relaxations extend over about 8 monomers,

330: leading to an energy reduction of 0.22 eV and a Stokes shift of 0.40

331: eV.  The DFT method is tractable for studying the influence of

332: structural disorder on the optical properties of polymers.

333: 

334: We thank Neil Greenham for discussions.  We acknowledge financial

335: support from the Engineering and Physical Sciences Research Council of

336: the United Kingdom and the Deutsche Forschungsgemeinschaft.  PDH

337: acknowledges the support of Sidney Sussex College, Cambridge.

338: 

339: 

340: \begin{thebibliography}{}

341: 

342: \bibitem{greenham_1995} N.~C. Greenham and R.~H. Friend, in

343: \textit{Solid State Physics, Advances in Research and Application},

344: edited by H. Ehrenreich and F. Spaepen (Academic, New York, 1995),

345: Vol. 49, p.1.

346: 

347: \bibitem{heeger_2001} A.~J. Heeger, Rev. Mod. Phys. \textbf{73}, 681

348:   (2001).

349: 

350: \bibitem{grem_1992} G. Grem, G. Leditzky, B. Ullrich, and G. Leising,

351: Advanced Materials \textbf{4}, 36 (1992).

352: 

353: \bibitem{grem_1993} G. Grem and G. Leising, Synthetic Metals

354: \textbf{55-57}, 4105 (1993).

355: 

356: \bibitem{footnote_lda} We used the local density approximation (LDA)

357: for the exchange-correlation energy.  The calculations were performed

358: using SIESTA with a DZP basis, see J.~M. Soler, E. Artacho, J.~D. Gale, 

359: A. Garcia, J. Junquera, P. Ordejon, and D. Sanchez-Portal, J. Phys.:

360: Condens. Matter \textbf{14}, 2745 (2002) and P. Ordejon, E. Artacho,

361: and J.~M. Soler, Phys. Rev. B \textbf{53}, 10441 (1996), and PARATEC

362: (PARAllel Total Energy Code) by B. Pfrommer, D. Raczkowski,

363: A. Canning, S.~G. Louie, Lawrence Berkeley National Laboratory (with

364: contributions from F. Mauri, M. C\^ot\'e, Y. Yoon, C. Pickard and

365: P. Haynes). For more information see www.nersc.gov/projects/paratec.

366: 

367: \bibitem{bse} G. Onida, L. Reining, and A. Rubio, Rev. Mod. Phys. {\bf

368:   74}, 601 (2002).

369: 

370: \bibitem{ambrosch-draxl_1995} C. Ambrosch-Draxl, J.~A. Majewski,

371: P. Vogl, and G. Leising, Phys. Rev. B \textbf{51}, 9668 (1995), and

372: references therein.

373: 

374: \bibitem{footnote_gaps} The Stokes shift can also be calculated as the

375: difference between the constrained DFT total energy for the excited

376: state and the DFT energy for the ground state, but this gives an

377: almost identical result.

378: 

379: \bibitem{hertel_2001} D. Hertel, S. Setayesh, H.-G. Nothofer,

380: U. Scherf, K. M\"ullen, and H. B\"assler, Advanced Materials

381: \textbf{13}, 65 (2001).

382: 

383: \bibitem{rohlfing_2000} M. Rohlfing and S.~G. Louie, Phys. Rev. B

384: \textbf{62}, 4927 (2000).

385: 

386: \bibitem{rohlfing_1998} M. Rohlfing and S.~G. Louie,

387: Phys. Rev. Lett. \textbf{80}, 3320 (1998).

388: 

389: \bibitem{rohlfing_1999} M. Rohlfing and S.~G. Louie,

390: Phys. Rev. Lett. \textbf{82}, 1959 (1999).

391: 

392: \bibitem{horst} J.-W. van der Horst \textit{et al.},

393: Phys. Rev. Lett. \textbf{83}, 4413 (1999);

394: Phys. Rev. B \textbf{61}, 15817 (2000);

395: Phys. Rev. B \textbf{66}, 035206 (2002).

396: 

397: \bibitem{note_bse_technical} The technical details of our $GW$-BSE

398: calculations are similar to those of Refs.~\cite{rohlfing_1999}

399: and~\cite{rohlfing_2000} except for the screening of the

400: electron-electron interaction.  Refs.~\cite{rohlfing_1999}

401: and~\cite{rohlfing_2000} considered a single chain but included

402: additional screening to model the effect of the other chains in a

403: solid, while in the present single-chain calculations no additional

404: screening was added. 

405: 

406: \bibitem{note:singlet-triplet_relaxations} These results also imply

407: that the structural relaxations and Stokes shifts in the lowest

408: singlet and triplet states of PPP should be very similar.

409: 

410: \end{thebibliography}

411: 

412: \end{document}

413: 

414: 

415: 

416: