Chain Conformations and Photoluminescence in Poly(di-n-octylfluorene)

0.97 MB·English

by  W. Chunwaschirasiri

Chain Conformations and Photoluminescence in Poly(di-n-octylfluorene)
W. Chunwaschirasiri, B. Tanto, D.L. Huber, and M.J. Winokur
Department of Physics, University of Wisconsin, Madison, WI 53706
(Dated: February 2, 2008)
Thediversesteady-statespectroscopicpropertiesofpoly(di-n-octylfluorene)areaddressedfroma
molecular-level perspective. Modeling of representative oligomers support the experimental obser-
5
vationofatleastthreedistinguishableclassesofconformationalisomerswithdifferingchaintorsion
0
angles. Oneclass appearstobepopulated byasingle compact structuralisomerandthisconforms
0
2 to the so called β phase. A rigorous Franck-Condon analysis of the photoluminescence in conjunc-
tion with Frenkel-type exciton band structure calculations is performed. These results accurately
n
reproduceall major spectral features of thephotoabsorption and those of singlet exciton emission.
a
J

4 Recentyearshavewitnessedwidespreadactivityinthe culated optical spectra, including a full Franck-Condon

field of polymer electronics. Integral to this progress (FC) vibronic progression,manifest many key attributes
]
ci is development of realistic molecular-level models that seen in the data. As a general feature PF8 β-type (Cβ)
can address fundamental issues of photophysics and emissionincorporatesa power law lineshape and not the
s
- chargetransport. Π-conjugatedorganics,includingsmall compositeGaussian/Lorentziantypicallyassumed[7,15].
l
r molecules,oligomersandpolymers,thattendtopackinto This approach also provides a clear microscopic starting
t
m locally well-ordered crystalline structures have received point for more quantitative analysis.
broad attention because of notable advances[1, 2, 3] in All of the PF8 polymers in this study were obtained
.
at boththeoreticalchemistryandphysics. Thesegainshave fromAmericanDyeSourceandusedasreceived. Various
m not been matched by parallel progress in the burgeon- solutionswereprepared,typically1%w/winsolvent,and
ing area of the functionalized polymers because of sam- the specific formulations (and some optical spectra) are
-
d ple or processing derived heterogeneities and, more im- already published[12, 14]. Thin-film samples were cast
n portantly, limited knowledge of the main chain and side onto sapphire substrates. All PA and PL spectra were
o
chain conformations. Thus many recent reports rely on recorded from the same physical location under steady-
c
[ qualitative underpinnings with few explicit references to state conditions using a dual PA/PL spectrometer[12].
the molecular level construction[4, 5, 6, 7]. ThusallPLspectraarecorrectedforself-absorptionand,
4
Polyfluorenes(PFs)[8] exemplify the extreme sensitiv- additionally, normalized for detector response.
v
0 ity of conjugated polymer photophyics to minor changes The three sets of PL/PA spectra in Fig. 1 demon-
1 in molecular architecture and film-forming conditions. strate the wide range of PF8’s optical properties. Each
2 Poly(di-n-octylfluorene),PF8assketchedintheFig.1in- PL spectrum has its emission dominated by one of the
9 set, is best known for polymorphic phase behavior[9, 10] three claimed conformational isomer families. The most
0 and the striking presence of a low energy emitting “β-
4
type” chromophore. This often cited “phase” always
0
d-mat/ atphaipbhiogplonhyteloa(ysPlhrusAaimrr)rapiesnbgaeuuatstlcarmedrendoincimuneociren(tidPhtayatLtete)cs,motthnbphesyeetsrieptaoutopfueeltynaricmettasue[lir7nree,sms1tthb1iase,ensc1idpoo2anhm]r.[od9etCs,o,ra1aeid0bms]bs.oayarrlIspketost-- unt, arb. units) PCBPLkAag l cdd..aattaa 0−0 ~8.4A C4 Cφ3Cτ23τC21Cτ1bC1
rXiv:con agdafolipe.orpT[pn9eeeh]sxnaipohsdrflaelaaevnesitnetntwhieppnarerrgnlooalc[tppd1ehoev2dersa,tecnd1idheci3vase,teishna1raasn4pet]dn.lasteotnhaetairahsrdmeietrymyo-.elsiemtsUcasuitnitoelutanisornu-poglaterliivccgtehaielnrlmofaprmptareeomosrppaeiehnwtruotorrrireeeekss-- ce Intensity (photon co 0−4 0−130x0−2 0−1 Cβ CCγαR R λλEExx1==1 1 330899 0 0KK nnmm 00..06Photoabsorption
a omtrfiofsduinerclgialnetgtioeonxfcorieftpotnrhesesiennlotPwaFtie8vn.eeAorgliysgiomcmhpaelerinsemysiteprliudrcisctaualrsefuosgrtcgoee-sfigteiivvldee uminescen " ienmteirscshioanin" Cα λE1x=1 040 K0 nm00..60 (absolute)
three nominally distinct classes or families of conforma- ol 0.0
tionalisomers(referredtoasC ,C andC )with10-20◦ hot 2.0 2.5 3.0 3.5 4.0
α β γ P Photon energy (eV)
stepsintheaveragefluorene-fluorenetorsionangles(|φ|).
The C type “family” is comprised of just a single iso-
β FIG. 1: Representative PF8 PA and PL spectra from the
mer. The inferred nature of these families is then used three claimed conformational isomer families in combination
toframeFrenkel-typetight-binding(t.b.) bandstructure with fitsto thedata using Eq. 1. Inset: Sketchof PF8.
calculations of the single chain PA and PL spectra. Cal-
2
∗
intense “0-0” peak is a superposition of the π-π transi-
TABLE I: Eq.1 parameters from fitsto PL data in Fig. 1.
tion and, on the low-energy tailing edge, vibronic over-
tones from low energy torsional modes[16]. Additional Data set ⇒ Cβ type Cγ type Cα type
peaks arise from other FC vibronic subbands. The top ¯hω0/Drel (eV) 2.806/0.12 2.839/0.11 2.922/0.21
spectraarerepresentativeofPF8films[14]spincastfrom ¯hωi (eV) HR factors (i=1to 9)
p-xylene and contain a coexistence of C and C iso-
α β 1: 0.0075 6: 0.159 0.75 0.17 0.98 0.13 2.17 0.21
mers. Because of large scale exciton energy migration
2: ∼0.07 7: 0.167 0.02 0.04 0.03 0.08 0.03 0.17
the lowest energy chromophores dominate the PL (here
3: ∼0.09 8: 0.176 0.02 0.05 0.03 0.05 0.03 0.05
C isomers). The middle spectra are from a C isomer
β γ
∗ 4: ∼0.11 9: 0.199 0.02 0.30 0.03 0.19 0.03 0.46
dominated sample and, in this case, the π-π transition
energiesareintermediatetothoseoftheproposedC and 5: 0.140 0.11 0.06 0.14
α
C families. This occurs in samples cycled above ∼430
β
K[14]. The bottommost spectra, obtained from casting R
e
PF8 dissolved in a toluene/tetrahydrofuranmixture[12], ~16.8 A 4lativ
isdominatedbyemissionfromCα typeisomers. Thisne- e e
n
matic glass sample also includes a broad low-energy PL e
2rg
backgrounddue to interchainexcitations[17]. Formation y (k
of the Cβ conformer has been suppressed by quenching J/m
to -30 ◦C from a thermotropic liquid crystal state. All 0ol p
P(aLgasipnedctureataoreennerogtyabmlyigsrhaatiropne)rbtuhtanthtehdeisrteisnpcetcntiavreroPwA- Cα Cγ Cβ er mo
ing of the Cβ emission is quite evident. 80 100 120 140 160 −2nom
Also included in Fig. 1 are non-linear least-square fits Mean fluorene−fluorene torsion angle (deg.) er)
to the PL data using a slightly modified FC formula:
∞ ∞ p FIG.2: LowenergyPF8conformers(referencedtotheC iso-
PL(h¯ω)∝(n(ω)h¯ω)3 ··· (1+c) mer). Eachsymbolcorrespondstoasingle,uniquetestedβcon-
nX1=0 nXp=0kY=1 former. Cα,Cβ andCγ correspondtotheproposedconformer
e−SkSnk p families. Inset: Model of Cβ isomer highlighting side chains.
k Γ (δ[h¯ω−(h¯ω − n ¯hω )]) (1)
o k k
(cid:20) n ! (cid:21)
k kX=1
For structural modeling we used a combinatorial
where n(ω) is the index of refraction (with data taken method approach in which a large number of possi-
from Ref. 7), h¯ω0 is the π-π∗ transition energy, h¯ωk ble single chain conformers were examined using em-
are vibrational mode energies for each mode k with pirically determined force-field parameters (MM3*[19])
nk = 0,1,2,... vibronic overtones. Sk is the conven- at zero-temperature in the gas-phase. Recent work in
tional Huang-Rhys (HR) coefficient but c = 0.2(p−1) polysilanes[20]andarelatedPF[21]suggestthatthisap-
(for p ≥ 1) is an ad-hoc term that increases the rela- proach, though limited in scope, reveals some basic at-
tive contribution by higher order modes. Because the tributesofthelocalintrachainstructure. Inthiscaseover
18K Cβ type PL is exceptionally well-definednine sepa- 700distinctstartingF8 decamerswereconstructed. The
ratebaseωk’swerenecessary(andextendinguptoseven fluorene backbone consisted of a repeating dyad (vari-
overtones). Whenpossiblethesefrequencieswerederived ants of a 2/1 helix as suggested in Ref. 10) in which the
fromRamanscatteringdata[18]. Thelineshapeoperator, starting torsion angles were alternately set to ±160◦ re-
Γ, is actually a composite function with, for example in spectively. Thethreecorealkylsidechaincarbon-carbon
the 0-0 case, (¯hω −¯hω0)3/2exp[(h¯ω0−¯hω)]/kbT when dihedralangles(seeFig.1)werethenassignedtoallpos-
(ω > ω0) and a small Urbach type tail (for ω < ω0) to sible combinations of the three conventional alkyl low
account for low energy states extending into the band energyconformers(i.e., anti, gauche andgauche′) within
gap. These first of these two Γ factors is essential for each alkyl chain. Every fluorene monomer received the
quantitative agreementalongthe high energyside of the identicalsidechainconformer. Additionaldetailsappear
0-0 PL band in Cβ-type PL spectra (both at 18 K and in Ref. 21. More extensive modeling (to identify viable
higher temperatures). Γ also includes convolution with dyad and larger conformational repeats) and molecular
a narrow Gaussian representing instrumental resolution simulations are needed to fully validate this approach.
and residual disorder. The only essentials needed in the Figure 2 displays the lowest zero-temperature energies
following text are the HR parameters, the ωk’s (for use of the trial starting conformers in terms of the aver-
afterthet.b.calculations)andthedisplacementparame- age backbone torsion angle |φ|. At torsion angles above
9 ◦
ter,Drel = i=1¯hωiSi. Drel measurestherelativediffer- 120 , there is some indication of three distinct classes
◦ ◦
ence betweePn the ground and excited state coordinates. of low energy chain structures near |φ| =135 , 150 and
◦
160 (or C , C and C respectively). Only a single
α γ β
3
unique isomer represents the C type conformer. The
Fig.2 insetdepicts the resultingβcompactmolecularcon- nits) CDaaltca. 110 K
struction. The alkyl chains adopt a trasoid (i.e., nearly b. u Cα Cα Pho
anti),gauche,gauchearrangementofthethreeinnermost ar to
cinasrebto)n.-VcaarnbodnertoWrsaioanlsainntgelreasc(tτi1o,nτs2baentwdeτe3ninnetihghebFoirgi.n1g nsity ( Cγ 110 K absorp
aliiflzktehytelhsaiisldkseytclrhuchactianuinsreal.ennTdghtthihseicsflotunwofoorreCmneHerb2aaucpnkpibtesoanrssmetwaololberekrautonndssttsaaibmblie-- escence Inte CγCβ Cβ 210 K tion (arb. un
iohlnyalrpyoatnmhaeolsnyisoziemstoohdfeaplopldryei(ssdterin-ibc2e-uetotihofyntlsh-nroe-fheehlxeoylwilcfleeusn.oerreWgnyee)c[t2lha1es]rsgeefsiovreoesf Photolumin Cβ Cβ Cα waornm18in gK its)
chain structures with differing average chain conforma-
2.5 3.0 3.0 3.5 4.0
tions and that Cβ segments are well ordered. Energy (eV) Energy (eV)
This suggestive finding is further tested by employing
a conventional t.b. Hamiltonian[22, 23, 24]
FIG. 3: PA and PL data in comparison with those obtained
N from the Eq.2 based tight-bindingcalculations.
H = E |nihn|+ J cos(φ )|mihn| (2)
n mn mn
nX=1 mX=n+1
where E and J are on-site excitation energy and merism we specified a different average fluorene-fluorene
n mn
nearest-neighbor intersite transfer energies respectively. torsionangle,|φ|,foreachfamilyandtothiswetypically
In this analysis we assume no on-site disorder and each addeda Gaussiandistributionwith astandarddeviation
of the two coefficients (E and J ) are given a single σ. In mostcasesonly twoof these three claimedfamilies
n mn
unique value. The second term has been modified by a werechosenbyspecifyingxand1−xrelativephasefrac-
cosine function to mimic the nature of π-conjugation[25] tions assuming the minority phase segment length was
asthebackbonetorsionangleφ varies. Afterdiagonal- dictated by Poisson statistics. In this way features of
mn
ization of the matrix the density of states and oscillator both segmental and worm-like disorder[23] are included.
strengths were calculated in standard fashion to give an Over 4000 single chain trials were averagedfor adequate
absorptionlineshape A(ω)=h µ2δ(ω−ω )i where µ2 statistics. A(ω) was sensitive to even small parameter
j j j j
changesandthusmanysuccessiveiterationswereneeded.
is the transition dipole momenPt of the jth eigenstate and
averaged over the entire ensemble. This function was Figure 3 displays direct comparisons of both PA and
further convoluted with the appropriate FC expression PLinfourrepresentativedatasets(threearefromFig.1)
for absorption[16] using HR parameters identified with with those calculated. Table II lists the associated pa-
the conformerdominatingthe respectivePL(asinTable rameters. These parameters best approximate the lead-
I) and a narrow Gaussian representing the spectrometer ing edge of the PA; the region that is most critical for
resolution plus residual disorder. simultaneously assessing both PA and PL. In particular
Calculations of the PL employed A(ω) directly. To the anomalous lineshape of the Cβ 0-0 is accurately re-
modelemissionfromexcitonsinthermodynamic equilib- flectedinthiscalculation. TheleadingedgeofthePAhas
rium at the bottom of the band A(ω) was multiplied by an extended region over which a near power law rise oc-
a Boltzmann factor, exp(h¯ω0−¯hω)/kbT (ω ≥ω0), next cursanditisthisfeaturethatgivestheatypicallineshape
convolutedwith the composite resolution/residualdisor- on the high energy side of the 0-0 peak and, in regards
derfunctionandthenexpressedusingaconventionalFC to peak position and width, the systematic thermal pro-
formula. Matching the exact energies required a small gression in Ref. 12. Replacing the Poisson distribution
Stokes shift defined as ∆. Clearly this simple approach ofthe minorityisomersegmentallengthwith aGaussian
cannot address the myriad of processes present in real worsensthesefits. ThisattributeiswashedoutintheCα
PFs. There is no accounting for either time-dependence andCγ dominatedPLspectrabecauseσandtheresidual
orenergymigration. Inadditionthereisnothingthatre- disorder (see text on Γ) are many times larger.
flectsanexcitonbindingenergyorinterchainexcitations. An important factor for estimating the relative phase
Only singlet exciton emission is modeled. fraction is the energy dependence of the oscillator
In all these calculations each phenylene ring (i.e., 1 of strength. Spectra similar to the top PA in Fig. 1 are
2
a fluorene unit) was deemed the basic structural unit so assigned up to a 25% C concentration based on life-
β
thateveryotheroff-sitetermwassetto−J becauseof time measurements[7]. If the actual t.b. results are di-
mn
thechemicallyimposedplanarizationofthefluoreneunit. rectly employed then the actual fraction drops to under
A single chain was truncated to 100 monomer units (or 10%. This“phase”,assuggestedbyakineticanalysis[12],
200 sites) and this qualitatively reflects the experimen- appears to be isolated one-dimensional threads of en-
tal length of the PF8. To represent conformational iso- hanced structural order embedded in sea of more dis-
4
try approaches planarity[29] we naively expect emission
TABLE II: Tight-bindingmodel parameters.
from the most planar ground state conformations (i.e.,
T Minority conformer Cα-typeregion ∆ shift Cβ-type) to yield the smallest displacement parameter.
(K) type conc. hni |φ| σ hnαi |φα| σα (meV) We actually find that the smallest Drel occurs when the
18 Cβ 9.5% 5.0 165◦ 3◦ 32 136.5◦ 13◦ 10 PF8 PL is dominated by the claimed Cγ-type family.
Quantum chemical calculations and other model studies
210 C 9.3 5.1 165 6 33 135 15 24
β
are still needed to establish the veracity of this result.
110 C 0.7 3.0 172 3 55 129 31† 9
β
110 C‡ 9.1 6.2 155 15 29 127 31† 11 As a final comment we note that this discrete three
γ
family model is overly abrupt and that one may ex-
†Assumes a uniform distribution of torsion angles.
pect more gradualcrossoversin terms of local structure.
‡ Additionally incorporates 0.5% of C conformers.
β The modeling procedure itself has merit for identifying
prospectivechainconformationsincludingthesidechains
and, with additional effort, conformational disorder as
orderedconformers. Moreoverthese t.b. calculationsob-
well. Successfully decipheringthe singlechainproperties
tain relatively short segmental lengths for the emitting
is one very important step for developing a full micro-
chromophore unit and, even in the locally well-ordered
scopic understanding of structure and photophysics.
C isomer, averages only five monomers.
β
All calculated |φ| and σ parameters adopt physically NSF DMR-0350383 grant support is gratefully ac-
realistic values and, in regards to |φ|, match the nomi- knowledged.
naloligomermodelingvalues. Matchingthetemperature
dependenceoftheC dominatedsamplerequiredfewpa-
β
rameter changes except for σ (which doubles at 210 K).
Increasingσ,becauseoftheoscillatorstrengthweighting
effects, enhances the leading edge of the PA. The Boltz- [1] J. Cornil, et al., Adv.Mat. 13, 1053 (2001).
[2] S. Tretiak, et al., Phys. Rev.Lett. 89, 097402 (2002).
mann weighting factor provides for an immediate fit of
[3] A. Ruini,et al., Phys. Rev.Lett. 88, 206403 (2002).
the PL data although ∆, a Stokes shift, goes from 10 to
[4] S.C.J.Meskers,etal.,J.Phys.Chem.105,9139(2001).
24meV.Thismayreflectthenatureofexcitonenergymi-
[5] O.J.KorovyankoandZ.V.Vardeny,Chem.Phys.Lett.
gration. Thet.b.resultsyieldEn andJm,m valuesof5.45 356, 361 (2002).
∗
and 1.35 eV. These give a π-π threshold energy of 2.75 [6] M.Wohlgenannt,X.M.Jiang, andZ.V.Vardeny,Phys.
eV and a band width of 2.70 eV. This threshold approx- Rev. B 69, 241204 (2004).
imates those of the planarized ladder type polymers[26]. [7] A. L. T. Khan, et al., Phys. Rev.B 69, 085201 (2004).
There are additional factors that need to be empha- [8] D. Neher, Macro. Rap.Comm. 22, 1366 (2001).
[9] A. J. Cadby,et al., Phys. Rev.B 62, 15604 (2000).
sized. Mimicking the slowrise onthe leadingedge ofthe
[10] M. Grell, et al., Macromolecules 32, 5810 (1999).
PA in samples dominated by C or C type emission re-
α γ [11] M. Ariu,et al., Phys.Rev. B 67, 195333 (2003).
quired very small but finite fractions of C . Modeling of
β [12] M. J. Winokur, J. Slinker, and D. L. Huber, Phys. Rev.
the broad PA at higher energy in these two cases is im- B 67, 184106 (2003).
perfect. The quenched (Cα type) sample includes a uni- [13] M. Ariu,etal., J.Phys.Condens.Mat. 14,9975 (2002).
formdistributionoftorsionanglesbetween98◦ and160◦. [14] H. Cheun, et al., Appl.Phys. Let. 84, 22 (2004).
More complex distributions are necessary to address the [15] T. W. Hagler, et al., Phys. Rev.B 44, 8652 (1991).
failure athigh energies. Other shortcomingsarepresent. [16] J. Gierschner, et al., J. Chem. Phys. 116, 8596 (2002).
All FC progressions for the PAs assume fixed HR coef- [17] J.M.Lupton,M.R.Craig,andE.W.Meijer,App.Phys.
Lett. 80, 4489 (2002).
ficients regardless of ¯hω. The implicit heterogeneity of
[18] S. Guha, privatecommunication.
these PF8 models should produce systematic variations
[19] StandardMM3 parameters withminormodifications for
in these coefficients. Moreover the PA HR coefficients
π-conjugation(see http://www.schrodinger.com).
are unlikely to be identical to those for PL. [20] M. J. Winokur and R. West, Macromolecules 36, 7338
We also resolve an order dependent systematic failure (2003).
of the FC model. As the number of vibronic overtones [21] B. Tanto, et al., Macromolecules 37, 9438 (2004).
rises the PL fit becomes progressively worse. This even [22] H. Port, H. Nissler, and R. Silbey, J. Chem. Phys. 87,
occursonthelowenergysideofthe0-0peak. Thiseffect 1994 (1987).
[23] A. Tilgner, et al., J. Chem. Phys. 96, 781 (1992).
may, in part, be due to the limited number of the ω ’s
k
[24] M. Bednarz, V. A. Malyshev, and J. Knoester, Phys.
used. However the 0-4 FC vibronic sub-band structure
Rev. Lett.91, 217401 (2003).
(near 2.1 eV) can still be resolved experimentally (see [25] S.N.YalirakiandR.J.Silbey,J.Chem.Phys.104,1245
Fig. 1[27]) and this is many times that given by Eq. 1.
(1996).
Given the simplistic assumptions usedto specify the HR [26] J. G. Mu¨ller, et al., Phys. Rev.Lett. 26, 267403 (2003).
parameters this failing is not surprising[28]. [27] Lower resolution dataset with better signal/noise.
One unexpected result relates to the Drel values listed [28] H. Kikuchi,et al., J. Chem. Phys. 119, 729 (2003).
in Table I. Because the equilibrium excited state geome- [29] J. F. Wang, et al., Macromolecules 37, 3451 (2004).