Controlling molecular conformation for highly efficient and stable deep-blue copolymer light-emitting diodes

.


■ INTRODUCTION
Since the discovery of electroluminescence (EL) from conjugated polymers in 1989, 1 there has been significant interest in solution-processable polymer light-emitting diodes (PLEDs) as potential candidates for low cost, energy efficient display and lighting applications.Tremendous efforts have been made to develop deep-blue (usually defined by the EL emission having Commission International de L'Eclairage (CIE) (x, y) coordinates both ≤0.15) 2 light-emitting polymers (LEPs) for use in high-luminance and high-efficiency PLED displays, 3−5 this requirement being essential to achieve the color gamut needed for high-quality display applications.An additional commercialization challenge is the limited stability of blue fluorescent LEPs, 6 for which the operational device lifetime is relatively short compared to red and green phosphorescencebased emitters; the latter have encapsulated lifetimes of over 100 000 h. 7 In this paper, we report a novel approach for the achievement of deep-blue, high-efficiency, and long-lived PLEDs via the introduction of a conformation change in the conjugated backbone of a fluorene−arylamine copolymer.
Conjugated polymers based on fluorene backbones have been extensively studied as blue organic light-emitting diode emission materials on account of both (i) their wide optical gaps, for example, ∼3.0 eV for poly(9,9-dioctylfluorene) (PFO), 8 that are favorable for deep-blue emission and (ii) their high photoluminescence (PL) quantum yields, for example, up to 50−60% for glassy phase PFO (see Figure 1a). 9,10−26 The β-phase has an increased backbone planarity within a fraction of chain segments, with the corresponding torsion angle between adjacent fluorene units ≈180°, resulting in the octyl substituent groups for neighboring monomers lying on opposite sides of the chain.−28 The β-phase has attracted much attention because of the action of its extended chain segments as a "self-dopant" 20 within an otherwise glassy matrix.These segments constitute the most ordered, lowest energy states and trigger an efficient energy transfer from the surrounding high-energy state glassy segments. 14,20,22,27,28−32 As a consequence, β-phase PFO PLEDs have been reported with a nearly 2-fold increase in the luminous efficiency from 1.0 to 1.9 cd/A compared to their glassy counterparts. 30,33Such an approach has not been applied to copolymers yet.Finally, there are numerous processing methods by which to induce β-phase chain segments, with reported methods including (i) thermal cycling, 20,21,34 (ii) Langmuir−Bodgett film formation and transfer, 35 (iii) deposition of films from mixed solvent, 23 solvent/additive 33 and high boiling point solvent 28 solutions, (iv) postdeposition film exposure to solvent vapors, 21 and (v) dipping in/flooding with a solvent. 23,30pproaches to high efficiency within simple PLED device architectures require that individual layers be optimized to perform more than one function.In contrast to blending materials with different functionalities, 36 covalently linked copolymers combine functionalities in a way that is resistant to phase separation.−39 In this study, we focus on F8:BSP copolymers (see Figure 1b) and, in particular, 95% F8:5% BSP or for short, the 95F8:5BSP copolymer.During synthesis, the insertion of BSP units into each copolymer chain is subject to the constraint that as BSP−BSP couplings are not possible, each BSP unit must have F8 neighbors.The chain formation process is otherwise statistical in nature, dependent on the BSP monomer concentration.The low fraction of BSP monomer units in the reaction mixture then ensures that the copolymer will contain long sequences of F8 units, interrupted only by sparsely distributed BSP units.Further details of the F8:BSP copolymer synthesis are given in the experimental section below.We directly compare copolymer properties with the 100F8 homopolymer PFO and a blend of 90% PFO and 10% poly(9,9-dioctylfluorene-alt-bis-N,N′-(4-butylphenyl)-bis-N,N′phenyl-1,4-phenylenediamine) (PFB) (with a corresponding volume fraction of 95% F8 and 5% BSP units), which we label 90PFO/10PFB.Further results are presented in the Supporting Information for the alternating 50F8:50BSP (PFB) and statistical 97F8:3BSP, 90F8:10BSP, and 80F8:20BSP copolymers.
−42 Varying both the arylamine moiety and its fractional incorporation within an F8-based copolymer thus provides an approach to tuning the hole injection and carrier balance properties of blue PLEDs. 6,38,40,43,44The associated optical gap remains large 40,45 but because the emission acquires a more charge transfer-like character with a broadened spectrum, it is no longer suitable to address the display requirement for deep-blue luminescence despite the improvement in efficiency.
Our study looks to combine the electrical benefits of arylamine incorporation with a conformational approach to spectral control.We demonstrate deep-blue, high-efficiency, and stable PLEDs by inducing the β-phase conformation within long, uninterrupted F8 chain segments of the 95F8:5BSP copolymer.This allows a significant improvement in CIE (x, y) from (0.149, 0.175) to (0.145, 0.123) and yields a peak luminous efficiency, η = 3.60 cd/A (at 146.5 cd/m 2 ), and a luminous power efficiency, η w = 2.44 lm/W (at 10.8 cd/m 2 ).The latter efficiency values represent the state-of-the-art performance for simple copolymer deep-blue PLEDs 3,4,46 being more than 5-fold better than for otherwise-equivalent, glassy PFO emission layer (EML) PLEDs (0.70 cd/A and 0.38 lm/W) and 14 and 60 times higher than for our first-reported PFO EML PLEDs (0.25 cd/A and 0.04 lm/W). 10Even at 1000 cd/m 2 , the efficiencies remain high, with η = 3.50 cd/A and η w = 1.50 lm/W.

■ RESULTS AND DISCUSSION
Optical Absorption and PL Spectroscopy.Optical absorption spectra for PFO, 95F8:5BSP copolymer, and 90F8/10PFB blend films on Spectrosil substrates are shown in Figure 2a−c, respectively, before and after solvent treatment to generate F8 β-phase chain segments.The glassy phase spectra of the 95F8:5BSP copolymer (Figure 2b) and 90F8/ 10PFB blend (Figure 2c) films closely resemble that of PFO with the main π−π* absorption peak located at ∼384 nm.After solvent vapor annealing (SVA) with toluene, the characteristic β-phase absorption peak at 433 nm was observed for all three samples, including the copolymer.This confirms that the BSP moieties within the copolymer backbone do not prevent βphase chain segment formation.The small BSP fraction (5 wt %) ensures that there are sufficiently long F8 segments within which the β-phase can form; oligofluorenes with as few as five F8 repeat units are reported to show β-phase spectral features. 47he proportion of F8 β-phase segments can be estimated by subtraction of a suitably normalized glassy phase absorption spectrum and comparison of the integrated residual (β-phase) and subtracted areas (Figure S1).Table 1 summarizes the estimated β-phase fractions for each film type.
A higher fraction of β-phase chain segments is formed in films of the PFO homopolymer (10%) and 90PFO/10PFB blend (12%) than that of the 95F8:5BSP copolymer (5%).The β-phase fraction in the homopolymer is broadly consistent with previous results in the literature. 22,48The 50F8:50BSP alternating copolymer (PFB) does not have any extended sequences of F8 units in which the β-phase can form, and hence in the blend, only the PFO chains support β-phase segments.We also know that bulky BSP moieties disrupt close chain packing, leading to a more disordered glassy microstructure for PFB, with no observed crystallization on thermal annealing. 40,45 is perhaps not, therefore, surprising that the 95F8:5BSP copolymer has a smaller fraction of β-phase chain segments formed during SVA than PFO.Our study additionally shows that β-phase segment formation still occurs (to a lesser degree) for 90F8:10BSP and (marginally) 80F8:20BSP copolymer films (see Figure S2).In the case of the blend films, the 10% fraction of PFB chains (with 50% BSP content) will not support βphase segment formation, so one might expect a proportionate reduction in the overall β-phase fraction relative to PFO.However, this is not seen, most likely as a result of an increase in the free volume that compensates by facilitating the conformation change in the PFO chains.
The PL emission spectra for each of the three film types are shown in Figure 2a−c before and after solvent treatment to generate F8 β-phase chain segments.All spectra were excited at λ ex = 385 nm.The glassy homopolymer PFO film spectrum in Figure 2a is consistent with the literature reports, with S 1 −S 0 0−0 and 0−1 vibronic peaks at 421 and 447 nm, respectively. 14,24,25,27,28The glassy 95F8:5BSP copolymer film shows a red-shifted, broad, asymmetric (with long wavelength tail), and largely featureless PL spectrum, which is very similar (but slightly red-shifted (455 nm peak) and broadened) to that of the PFB component (450 nm peak) in the blend film and indeed those of the 97F8:3BSP, 90F8:10BSP, and 80F8:20BSP copolymers (see Figure S2).The 95F8:5BSP copolymer film spectrum does not reveal an obviously PFO-like (F8-localised) component, indicating efficient energy transfer from locally excited F8 excitons to BSP-centered excitons. 49he excited states responsible for the glassy copolymer PL have been shown to have a significant charge-transfer (CT) character, 49 as observed for PFB and poly(9,9-dioctylfluoreneco-N-(4-butylphenyl)diphenylamine) (TFB). 50The dilute solution PL spectra of the PFO homopolymer, PFB, and 95F8:5BSP for solvents of different polarities (toluene, tetrahydrofuran, and dichlorobenzene) are shown in Figure S3.The PFO PL shows little change with the solvent polarity, whereas for PFB, there is a large red shift and broadening with increasing polarity, strongly indicative of CT character.For 95F8:5BSP, the emission comprises both vibronically structured F8 and broadband BSP-related contributions, with the former experiencing no solvatochromic shift, whereas the latter red shifts and broadens, confirming the coexistence of both bound neutral exciton and CT emission states.The CT character originates from a differential spatial partitioning of the highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO) wave functions across the BSP and F8 units, leading to a displacement in the associated hole and electron charge densities. 49This is supported by cyclic voltammetry (CV) measurements (see Supporting Information Figure S4).
The PL spectra are significantly altered by the generation of β-phase chain segments.All three film types then display a wellresolved vibronic structure with a close match of the blend and copolymer PL spectral features to those of PFO; the characteristic β-phase vibronic peaks appear at 437, 465, and 498 nm.However, differences do exist, with the vibronic peaks best resolved for PFO, less so for the blend, and least for the copolymer.In addition, the apparent strength of the S 1 −S 0 0−1 and 0−2 vibronic peaks relative to 0−0 is greater for the 90PFO/10PFB blend than for PFO and greatest for the 95F8:5BSP copolymer, consistent with the blend and copolymer spectra comprising a superposition of β-phase PFO-like structured excitonic emission and residual PFB-like  broadband CT emission.Separation of the 95F8:5BSP copolymer and 90PFO/10PFB blend spectra in this way (Figure S5) reveals that 62% of the copolymer emission is βphase structured emission, whereas 38% is residual PFB-like emission.The 90PFO/10PFB blend shows 82% β-phase emission and 18% residual PFB emission.Efficient energy transfer from high-energy glassy to lowenergy β-phase segments in PFO has been extensively studied, with only a few % of β-phase segments needed for dominant βphase emission. 14,25,28Intriguingly, we observe similar behavior here with the presence of β-phase segments in the copolymer and blend films leading to a strong promotion of structured vibronic emission.This is despite the fact that, as a consequence, there is a net increase in the mean PL emission energy relative to the glassy film spectra (with dominant CTlike emission); this can be explained by the very small Stokes shift (λ ≈ 5 nm) for β-phase segments.β-phase PL spectral components are also evident in the solvent vapor annealed 97F8:3BSP, 90F8:10BSP, and 80F8:20BSP copolymer films but not for PFB (50F8:50BSP alternating copolymer) (Figure S2).As the fraction of BSP units increases, the fraction of β-phase segments formed after SVA decreases (Table S1), resulting in larger residual fractions of PFB-like emission.Radiative decay times for the competing emissive species are likely to be important in this context.
Despite both having the same volume fraction of F8 and BSP units, the glassy phase 90PFO/10PFB blend film PL spectrum contains emission components from both PFO (evidenced by the shoulder at 425 nm) and PFB (main peak at 450 nm) polymer chains, whereas the 95F8:5BSP copolymer seemingly does not.This indicates that the energy transfer from majority F8-to minority BSP-centered sites is less efficient in the blend (PFO to PFB interchain transfer) than that found in the copolymer (for combined inter-and intrachain transfer); the blend microstructure will clearly also play a role in this.
The copolymer, in addition, shows less intense emission around 530 nm (see Figure S6d), where "green-band" (fluorenone defect enabled) excimer emission occurs for PFO and related materials. 12,13A reduction in the green-band emission is an advantage of incorporating bulky BSP moieties within the copolymer backbone, reducing fluorenone-tofluorenone cofacial π-stacking and hence excimer formation and emission.
Time-Dependent PL Spectroscopy.To further characterize the emissive species contributing to the PL spectra, timecorrelated single-photon counting (TCSPC) measurements were used to record PL decay transients under (unless otherwise indicated) 404 nm excitation and 460 nm detection.Figure 3 shows the results, from top to bottom, for the 95F8:5BSP copolymer, 90PFO/10PFB blend, and PFO homopolymer thin films with both glassy (left) and β-phase (right panel) microstructures.Table 2 collects together the fitted decay times, fractional amplitudes, average decay times, and PL quantum efficiencies (PLQEs).Additional decay transients collected at a range of different wavelengths between 420 and 540 nm are shown in Figure S7, and equivalent data for PFB are shown in Figure S8.PL spectra and decay transients were also measured for dilute toluene solution samples of PFO, PFB, and 95F8:5PFB and are shown in Figure S9.
−55 Consistent with this assignment, the PFO decay transients for dilute solutions are monoexponential at all detection wavelengths, with ∼356 ps decay time (Figure S9b).For PFO β-phase film samples, the 460 nm decay times reduce marginally to ∼270 ps and ∼1.15 ns, which (given no PLQE decrease (Table 2)) point to an increase in the transition dipole moment, consistent with the known increase in the conjugation length. 19,26he PL transients of the 95F8:5BSP copolymer in dilute solution (Figure S9c) show a biexponential decay at shorter wavelengths (420 and 440 nm), attributed to combined PFOlike excitonic emission (with τ 1 ≈ 140 ps) and PFB-like CT emission (with τ 2 ≈ 1.4 ns).At longer wavelengths, beyond 460 nm, a monoexponential decay with τ ≈ 1.4 ns is observed, identical to the dilute solution decay for PFB (Figure S9d).In the case of glassy 95F8:5BSP copolymer films, the overall decay at 460 nm can be described by three components with τ 1 ≈ 580 ps (43%), τ 2 ≈ 2.18 ns (45%), and τ 3 ≈ 11.9 ns (12%), each attributable to CT emission.The average lifetime is then τ av ≈ 2.66 ns.These time constants all differ from the solution CTstate decay, consistent with the influence of heterogeneity in solid-state packing and inter-chain/-segment interactions, as also reported in the previous studies of fluorene−amine copolymers. 49,56,57We further note that the decay times vary  substantially with the emission wavelength (Figure S7c), again suggesting a distribution of CT lifetimes within the copolymer.Interestingly, the longest time constant (τ 3 ≈ 11.9 ns) decay is not observed in PFB (50F8:50BSP alternating copolymer) (Figure S8), and its observation here then points to a potentially more substantial spatial separation of electron and hole wave functions in glassy 95F8:5BSP.One possibility would be interchain excitations formed between high electron affinity F8 units in one chain and low ionization potential BSP units in a neighboring chain.However, as no distinct exciplex peak is seen in either PL or EL, this remains speculative.
The striking spectral changes that occur when β-phase chain segments are induced in 95F8:5BSP copolymer films (Figure 2b) are accompanied by a strong change in the 460 nm PL decay dynamics (Figure 3 and Table 2).Each of the fitted decay times reduces to τ 1 ≈ 319 ps (56%), τ 2 ≈ 1.32 ns (35%), and τ 3 ≈ 7.43 ns (9%), with a large shift in the fractional weighting toward the τ 1 component.The average lifetime correspondingly reduces to τ av ≈ 1.31 ns.The majority sub-320 ps decay component is consistent with the spectral dominance of the vibronic F8-based emission (Figure 2b), whereas the 1.32 and 7.43 ns time constants signal the presence of residual CT emission.As for the glassy case, the decays are slower for longer collection wavelengths (Figure S7d).However, unlike the situation for PFO and blend films, a modest decrease in the PLQE from 45 to 40% was observed when the β-phase was induced in copolymer films (Table 2), albeit that the PLQE itself remains relatively high (cf.35% for PFO and 25% for the blend).Among possible explanations, fluorenone-centered quenching is plausible but remains unproven.
The 460 nm PL decay transient for glassy 90PFO/10PFB blend films can also be fit to three exponentials, although their relative fractions and decay times, not surprisingly, differ from those of the 95F8:5BSP copolymer (Table 2).Excitonic emission from the PFO chains is dominant, with CT-like emission accounting for much smaller fractions than in the copolymer; corresponding decay times are τ 1 ≈ 295 ps (64%), τ 2 ≈ 1.48 ns (27%), and τ 3 ≈ 8.2 ns (9%).As for the copolymer, when emission is collected at longer wavelengths, we see that longer lifetime emissive species are increasingly important (Figure S7e).In the blend case, the longest time constant excited states (τ 3 ≈ 8.2 ns) have previously been shown to be exciplexes generated between PFO and PFB, with thermally assisted energy transfer to a PFB CT-like exciton. 58pon β-phase induction, the decay becomes almost monoexponential (Figure 3), with excitonic emission from PFO βphase chain segments totally dominant; decay times are τ 1 ≈ 247 ps (97%) and τ 2 ≈ 1.93 ns (3%).The lack of a longer lived time constant indicates that the majority excitons generated on PFO chains no longer form exciplexes with PFB and instead efficiently transfer to β-phase segments before undergoing radiative decay.As the detection wavelength is increased (Figure S7f), the short-lived decay time remains relatively constant (τ ≈ 271−325 ps) until 500 nm, where CT emission increases.The PLQE values for glassy and β-phase 90PFO/ 10PFB blends were 20 and 25% respectively, smaller than for both PFO and 95F8:5BSP copolymer films, indicating more substantial nonradiative decay.
Display-Related Device Characteristics.To test the effect of β-phase segment formation on device performance, PLEDs were fabricated (see Experimental Methods) with a conventional bottom-emitting device architecture, comprising glass substrate/indium tin oxide (ITO) anode/poly(3,4-ethylenedioxythiophene) (PEDOT:PSS) hole injection layer/TFB electron-blocking interlayer/EML/LiF/Ca/Al cathode.Schematic device and energy level diagrams are shown in Figure 4.The energy levels of PFO, 95F8:5BSP, and PFB were deduced from CV measurements (Figure S4 and Table S2).
Figure 5a,b compares the current density and luminance versus voltage (J−V−L) characteristics for, respectively, glassy and β-phase PFO homopolymer, 95F8:5BSP copolymer, and 90PFO/10PFB blend EML devices.Figure 5c,d shows the corresponding luminance-dependent glassy and β-phase PLED efficiencies η (cd/A) and η w (lm/W), and Figure 5e,f shows the associated external quantum efficiencies η eqe (EQEs).Other parameters (turn-on voltage, peak η , η w , and η eqe ) for these devices are collated in Table 3 and their EL spectra are shown in Figure 6a.
In terms of PLED efficiency, PFO glassy EML devices show a peak η = 0.43 cd/A at 6.8 V, a peak η w = 0.22 lm/W at 5.6 V, and a peak η eqe = 0.31% at 6.6 V, with 1000 cd/m 2 luminance at 7 V.The PFO β-phase EML devices show a significant improvement, with peak efficiencies of η = 0.70 cd/A at 6.2 V, η w = 0.38 lm/W at 5.4 V, η eqe = 0.43% at 6.0 V, and 1000 cd/m 2 luminance now at 5.9 V.This improvement is consistent with previous reports for β-phase PFO devices. 30,31,33,59,60  shows, however, that the efficiency gains are at the expense of a detrimental change in the EL emission color; a consequence not previously emphasized.The shift from glassy (peak λ = 425 nm) to β-phase (peak λ = 440 nm) alters the CIE (x, y) coordinates from (0.155, 0.098) to (0.157, 0.117), with a corresponding shift in the dominant wavelength from 470 to 475 nm and a decrease in the color saturation from 86 to 82%.This behavior limits the achievable display color gamut.It also helps to explain why the η eqe enhancement is more modest than the gains in η and η w ; a shift to the green leads to a better overlap with the photopic eye sensitivity function that peaks at 555 nm.The 90PFO/10PFB blend EML PLEDs show significantly better efficiency characteristics than the corresponding PFO devices (Figure 5), with the PFB fraction strongly assisting hole injection (Figure 4b). 40,45The glassy blend EML gives peak efficiencies η = 1.52 cd/A at 5.6 V, η w = 0.97 lm/W at 4.6 V, and η eqe = 0.89% at 5.6 V, with a luminance of 1000 cd/m 2 at 6.3 V, whereas β-phase blend EML devices show peak efficiencies η = 1.31 cd/A at 5.6 V, η w = 0.79 lm/W at 4.4 V, and η eqe = 0.80% at 5.6 V and reach 1000 cd/m 2 at 6.2 V. βphase devices are, therefore, somewhat less efficient than glassy devices, but in terms of CIE (x, y) color coordinates, β-phase segment formation leads to a shift from (0.162, 0.174) to (0.151, 0.118), resulting in a highly desirable, deeper-blue emission (Figure 6b).The corresponding dominant wavelength shifts from 478 to 474 nm, and color saturation increases from 73 to 83%.This color shift also helps to explain, at least in the proportionately larger decrease in η and η w values than in η eqe , as the emission then has reduced overlap with the photopic eye sensitivity function.
Glassy and β-phase 95F8:5BSP copolymer EML devices show yet further enhanced PLED efficiency (Table 3).Glassy devices give peak efficiencies η = 4.05 cd/A at 6.2 V, η w = 2.62 lm/W at 4.0 V, and η eqe = 2.40% at 6.0 V, with a luminance of 1000 cd/m 2 at 6.6 V, whereas for β-phase devices, η = 3.60 cd/ A at 5.4 V, η w = 2.44 lm/W at 4.2 V, and η eqe = 2.40% at 5.6 V, with 1000 cd/m 2 at 7.2 V. Interestingly, here η eqe is unaltered by β-phase induction, whereas (see Figure 6b) the CIE (x, y) coordinates still shift positively from (0.149, 0.175) to (0.145, 0.123), resulting in a dominant wavelength decrease from 479 to 474 nm and a color saturation increase from 77 to 85%. Figure S10 shows the deconvolution of the EL spectra of βphase copolymer and blend devices, showing that β-phase emission accounts for ∼67% of the total EL emission in the copolymer device and ∼80% emission in the blend device.
The copolymer EML efficiency improvement is largely attributed to the BSP units both assisting hole injection and, because of their sparse distribution, acting as deep hole traps.Glassy PFO EML devices display the highest turn-on voltages (defined as the applied bias at which L = 1 cd/m 2 ), namely 4.2 V, reducing to 4.0 V on induction of β-phase chain segments, consistent with previous reports. 30,31,59The 90PFO/10PFB blend devices conversely show the same turn-on voltage (3.8 V) irrespective of the glassy or β-phase microstructure, as also do the 95F8:5BSP copolymer devices (3.2 V); injection is clearly controlled by BSP rather than β-phase units.The better turn-on behavior for the copolymer devices is likely because of the more uniform distribution of BSP units; they are present within every 95F8:5BSP polymer chain.Atomic force microscopy topography images show evidence of phase segregation between PFO and PFB chains in 90PFO/10PFB blend samples (Figure S11) along with an increase in roughness (Table S3), whereas the 95F8:5BSP microstructure appears featureless.Evidence of phase segregation between PFO and PFB chains has also been previously observed in the literature. 58ther studies have additionally shown that such copolymers have 3 to 4 orders of magnitude lower time-of-flight photocurrent hole mobility (μ h ToF ) than equivalent films of PFO and alternating fluorene−arylamine copolymers. 40,43,45he measured μ h ToF ≈ 10 −7 cm 2 /Vs for glassy films indicates substantially deeper hole trapping than for β-phase segment formation in PFO.Electrochemical data (Figure S4 and Table S2) for 95F8:5BSP films suggest that the trap depth is ∼0.3 eV above the ∼5.80 eV HOMO level of PFO.This trapping effect is also evident in the reduction in the current density between the PFO homopolymer and 95F8:5BSP copolymer EML devices.We find J = 231.4mA/cm 2 at 7 V for glassy PFO and only 41.4 mA/cm 2 at the same voltage for glassy 95F8:5BSP despite the turn-on voltage for the copolymer (3.2 V) being 1 V lower than for PFO (4.2 V).
A well-known drawback to using PFO as the EML in PLEDs is its low spectral stability under operation because of the appearance of a low-energy "green-band" emission when driven at higher voltages, the origin of which is interchain emission from fluorenone-based defects. 13,54,59,61Figure 7 shows the EL spectra as a function of the applied voltage for glassy and βphase PFO and 95F8:5BSP devices.As the voltage is driven beyond 7 V for the glassy PFO device and beyond 9 V for the β-phase PFO device, the emergence of a broad, low-energy component ("green-band") is clearly observed.However, for both glassy and β-phase 95F8:5BSP devices, no green band is observed when driving devices between 5 and 13 V, indicating that 95F8:5BSP devices show much improved color stability compared with PFO PLED devices.This is likely to be due to the bulky BSP units causing increased disruption in the packing structure between the polymer chains, thereby preventing green-band emission; similar reductions in the green band have been observed when small amounts of carbazole units or large amine group end caps have been incorporated into polyfluorene chains. 62,63he BSP-centered hole trapping increases the likelihood of exciton formation and improves the charge-carrier balance. 6,64opolymer devices based on 97F8:3BSP and 90F8:10BSP EMLs were also fabricated (Figures S12 and S13), revealing that the 95F8:5BSP copolymer gives the optimal EML efficiency performance in this composition sequence (Table S4); one needs enough but not too much BSP incorporation, and the way in which the BSP units are incorporated also matters.
In summary, the device efficiency and display color parameter data (Table 4) show that in the absence of BSP units (PFO homopolymer), the induction of β-phase chain segments is advantageous to device efficiency but only at the expense of a shift away from the desirable deep-blue emission.The introduction of BSP units into the EML (blend and copolymer) yields a significant overall enhancement in the device efficiency and color stability, with the copolymer performing substantially better than the blend.Again, the efficiency enhancement is achieved at the expense of color response, with the BSP-related CT-like emission yielding less deep-blue color coordinates (green shift in the dominant wavelength and reduction in saturation).Induction of β-phase chain segments is then strongly beneficial to the color response, and for copolymer EML devices, this occurs without any appreciable decrease in the quantum efficiency; the β-phase copolymer EML PLEDs thus provide the best overall combination of efficiency and color performance.
As a final device test, motivated by literature reports that βphase formation in PFO increases the device lifetime, 54 encapsulated pre-and post-solvent vapor annealed 95F8:5BSP copolymer PLEDs were subjected to accelerated lifetime testing under nitrogen using a constant current source set to deliver 4 mA (i.e., J ≈ 90 mA/cm 2 for the 4.5 mm 2 pixels under test).The luminance was measured at 60 s intervals starting from 2821 cd/m 2 for glassy and 2300 cd/m 2 for βphase devices.Initially (cf. Figure S14), the glassy and β-phase 95F8:5BSP PLED luminance values decayed at a similar rate, with half decay times T 50% (glassy) = 176 min and T 50% (βphase) = 180 min.The subsequent decay was much more rapid in the glassy copolymer EML devices, especially beyond 400  a Note the close agreement for β-phase EML color saturation and dominant wavelength for all three EML types and likewise the close agreement for glassy EML copolymer and blend devices.
min.The luminance took 490 min to drop to 30% of its starting value for the glassy 95F8:5BSP PLED, but took 630 min to reach the same fractional output for the β-phase device, and while the latter was still emitting some 250 cd/m 2 at 1200 min, the glassy device luminance had fallen below 4 cd/m 2 by that time.The operational stability of β-phase 95F8:5BSP copolymer devices is, therefore, significantly greater than that of otherwise-equivalent glassy devices.Several factors are expected to contribute to this, including a desirable distribution of the recombination sites vertically through the EML, enhanced charge-carrier balance, and a faster radiative decay time for the exciton.We now consider further the effects of β-phase formation on the energy transfer of emissive species to explain the desirable PLED emission for 95F8:5BSP copolymer EML devices.We explore, in particular, the origin of efficient energy transfer to βphase F8-centred excitons, with a concomitant net increase in the mean photon energy and deeper-blue emission.
The HOMO and LUMO of glassy phase PFO homopolymer films were deduced from CV measurements to be 5.80 and 2.10 eV, respectively (Figure S4 and Table S2), similar to previously reported values. 42Upon β-phase chain segment formation, a smaller optical gap component is introduced into the ensemble of absorbing chromophores, with the resolved S 0 −S 1 0−0 peak at 433 nm. 24,27Additionally, as a result of rapid energy migration to β-phase segments, there is a ∼16 nm red shift in S 1 −S 0 0−0 PL emission.CV measurement-based HOMO and LUMO values were also determined for PFB and 95F8:5BSP, yielding 5.05 and 2.00 and 5.49 and 2.10 eV, respectively (Table S2).
The schematic energy level diagram for glassy 95F8:5BSP copolymer chains (Figure 8a) shows how the energy levels vary spatially along the copolymer chain, with the electron-rich BSP unit being raised in energy relative to the F8 units.Although electrical excitation should (at least initially) predominantly produce CT excitons because of the strong hole-trapping nature of the BSP units, 6,38,43 under optical excitation, exciton states will also form on longer F8 segments (labelled I in Figure 8).In this case, efficient inter-(not shown) and intrachain energy transfers of F8 excitons to CT states (labelled II in Figure 8) are observed. 49pon β-phase formation, the corresponding F8 HOMO level will move up and the LUMO will move down, forming a type I quantum well-like structure along the polymer chain (Figure 8b). 65Under optical excitation, excitons formed locally on glassy F8 segments undergo efficient energy transfer to either a β-phase F8 segment 15,27,52 (labelled III in Figure 8) or BSPbased CT states 49 before decaying (Figure 8b).As such, the 95F8:5BSP β-phase PL spectrum is a superposition of both PFO-like β-phase emission (∼62% of total emission) and residual CT emission (∼38%).The reason β-phase emission dominates the spectrum despite the film having roughly the same proportion of β-phase segments and BSP units (∼5%) is likely to be the result of the faster decay for β-phase F8 excitons.Longer-lived CT states will be more prone to thermal energy transfer before decay.Nevertheless, as expected, the fraction of CT-type emission increases (and β-phase emission correspondingly decreases) as more BSP units are incorporated into the copolymer chain (see Figure S2 for 90F8:10BSP and 80F8:20BSP ultraviolet−visible (UV−vis) absorption and PL emission data).
To further probe the energy-transfer mechanism in the 95F8:5BSP copolymer, additional low-temperature PL measurements (λ ex = 400 nm) were taken for β-phase PFO and both glassy and β-phase 95F8:5BSP copolymer samples (Figure S15) from 290 K down to 10 K. Low-temperature β-phase 95F8:5BSP PL measurements (Figure S15c,d) reveal that the spectrum remains a superposition of both vibronic β-phase emission (Figure S15a) and CT-like emission from BSPcentered states (Figure S15b) over the whole 10 to 290 K range.Spectral deconvolution reveals that the component spectra are identical to the PFO β-phase and glassy 95F8:5BSP spectra at each temperature, as for example shown at 10 K (Figure S15e).By integrating the component spectra at each temperature, we find that (Figure S15f) the fraction of β-phase emission decreases as the temperature decreases, from ∼64% at 290 K to ∼32% at 10 K.
These results support the energy-transfer model outlined above because the energy transfer in β-phase PFO has been proposed to be a two-step process of thermally assisted exciton diffusion followed by Forster resonance energy transfer from high-energy glassy phase F8 to low-energy β-phase segments. 24he decrease in β-phase emission at low temperature could then be explained as a result of energy transfer to β-phase units being more dependent on thermally assisted diffusion than is the case for BSP sites.Alternatively, enhanced polaron formation has been previously observed for β-phase PFO, and polarons are known to act as emission-quenching sites, especially at low temperatures where their lifetimes are long, leading to an increasing β-phase exciton quenching at low temperatures. 24Similar effects are seen in other polymers when segmentation by conjugation breaks leads to polaron trapping. 66nder electrical excitation of copolymer EML devices, it is likely that a much greater fraction of initially formed excitations are CT states because of the hole-trapping nature of the BSP unit (trap depth ∼0.3 eV).Despite this, F8 β-phase emission provides the dominant (∼67%) contribution to the EL spectrum, with residual CT-like emission delivering a more modest (∼33%) contribution.This suggests that either endothermic energy transfer from BSP-centered CT states to β-phase F8 excitons or "trap-filling" of BSP sites occurs.In the latter case, as the BSP "traps" fill, β-phase F8 charge localization is expected to play an increasing role.Evidence for this is clearly seen in Figure S16, where the relative fraction of β-phase exciton EL emission increases with voltage.The shorter emission decay time for F8 β-phase segment than BSPlocalized states also feeds into the higher relative fraction of βphase EL.
A key difference for 90PFO/10PFB blend samples is that intermolecular energy transfer between neighboring PFB and PFO chains becomes important.Under optical excitation of the glassy phase blend, excited states are formed directly on both PFO and PFB chains.However, owing to the raised HOMO and LUMO energies of PFB relative to PFO, a type II heterojunction occurs at the PFO/PFB interfaces, resulting in exciplex formation. 58At room temperature, the exciplex undergoes endothermic energy transfer to the emissive PFB CT state, yielding its characteristic spectrum (Figure S17). 58owever, not all PFO excitons generated will form an exciplex, and we consequently also observe PFO exciton emission as a shoulder at 425 nm (Figure 2c).Under electrical excitation, as holes will tend to localize on PFB chains and electrons on PFO (see Figure 4 energy levels), it is likely that exciplexes are directly formed by electron−hole Coulomb capture 67 before undergoing endothermic energy transfer to PFB CT states (Figure S18).Consistent with this, Figure 6a indeed shows a much weaker 425 nm PFO emission shoulder for glassy phase blend EL than for PL.
For β-phase blends, the majority of optically excited glassy segment F8 excitons will tend to transfer their energy to βphase sites without forming exciplexes (Figure S17).Additionally, under electrical excitation, exciplex states that form between PFB and β-phase PFO chains (from direct electron−hole capture) will likely undergo endothermic energy transfer to PFO β-phase excitons (Figure S18).As a consequence, both the PL (Figure 2c) and EL (Figure 6a) βphase spectra are dominated by structured vibronic emission.Some residual CT emission remains, respectively, 18 and 20% for PL and EL (Figures S5 and S10), but these values are significantly lower than for the 95F8:5BSP copolymer case despite the same volume fraction of BSP units being present.This is most likely because of the greater fraction of β-phase segments generated and the consequences of phase separation.

■ CONCLUSIONS
In summary, we have demonstrated a novel route to highefficiency, deep-blue emitting PLEDs by introducing a simple molecular level conformation change in the F8 sequences of 95F8:5BSP copolymers from the disordered glassy phase to the rigid β-phase microstructure via solvent vapor annealing in toluene.UV−vis absorption and PL spectroscopy measurements of solvent-annealed F8:BSP copolymers at 3, 5, 10, and 20% BSP fraction showed β-phase formation in the F8 segments, with the appearance of characteristic red-shifted absorption peaks and the promotion of well-structured vibronic emission.
Incorporating 5% BSP units into the conjugated backbone of an otherwise F8 polymer (thus yielding 95F8:5BSP) produces a 5-fold performance enhancement in the PLED luminous and luminous power efficiencies relative to PFO (i.e., 100F8).The BSP units (with significantly lower ionization potential) enhance hole injection and act as hole-trapping sites that assist efficient exciton formation.The BSP units have an additional benefit in increasing the color stability of the PLEDs by suppressing the green-band emission when driven at higher voltages.The BSP unit itself causes an undesirable shift of the EL to a sky-blue emission (CIE (x, y) = (0.149, 0.175)) due to its CT character formed in the 95F8:5BSP copolymer.Subsequent introduction of β-phase chain segments within the copolymer restores a highly desirable deep-blue, vibronically structured EL emission with CIE (x, y) = (0.145, 0.123) while retaining the high efficiency and increasing operational stability.The PLEDs using β-phase 95F8:5BSP EMLs show η = 3.60 cd/A at 5.4 V and η w = 2.44 lm/W at 4.2 V, with 1000 cd/ m 2 luminance at 7.2 V.
The spatial distribution of BSP units in the active layer is also found to play an important role in device function, with the 95F8:5BSP copolymer (homogeneous) and 90PFO/10PFB blend (heterogeneous) EMLs showing distinctly different optoelectronic properties despite containing the same volume fraction of BSP units.Both the glassy and β-phase 90PFO/ 10PFB blend devices performed less efficiently than equivalent copolymer devices, with EQE lower by a factor of 2.5−3.0,but still significantly better than for PFO-only devices.
Our study represents the first demonstration of the use of a molecular level chain conformation change as a simple but effective method to control the optoelectronic properties of a fluorene-based copolymer.It will be interesting to see how broadly such a conformation control approach can be applied to other material systems for various device applications, including solar energy conversion, electronics, and sensing.
UV−Vis Absorption, PL Spectroscopy, and Transient PL Decay Measurements.Optical spectroscopy was undertaken on ∼60 nm thickness PFO homopolymer, F8:BSP copolymer, and polymer/polymer blend films spin-coated from toluene solution (10 mg/mL) onto quartz substrates.The substrates had been precleaned by sequential 15 min sonications in acetone, isopropanol, and detergent (Hellmanex III, 2% by volume in deionized (DI) water) prior to a 3 min plasma ash in air at 80 W.
Steady-state transmittance was measured using a Shimadzu UV-2550 UV−vis spectrophotometer.Absorbance was calculated directly from transmittance (no scattering or reflection correction) based on the natural logarithm, and the substrate contribution was simply subtracted.PL spectra were recorded in reflection geometry using a Jobin Yvon Horiba Fluoromax-3 spectrofluorometer (excitation wavelength λ ex = 385 nm).PLQE was measured using a Jobin Yvon Horiba Fluoromax-3 spectrofluorometer equipped with a diffusely reflecting integrating sphere.
Time-resolved PL decay measurements used an IBH fluorescence lifetime spectrometer operating in the TCSPC mode.The excitation source was a 404 nm-pulsed LED operating at 1 MHz rep-rate with a pulse temporal width of 200 ps.IBH Datamax software was used to deconvolve the instrument response function from the data and to fit multiexponential decay functions.
Low-Temperature PL Measurements.Samples were held inside a helium-filled, closed-cycle cryostat, with spectra recorded at 20 K intervals.The cryostat temperature was held constant for 5 min prior to each measurement, enabling the sample to reach thermal equilibrium.Excitation (0.13 mW) was with 400 nm light from a monochromated supercontinuum laser source (Fianium).PL was collected at right angles to excitation using a 100 μm diameter optical fiber and fed into a spectrometer (Andor SR-163) equipped with a charge-coupled device (CCD) detector (Andor i-Dus).To enhance the signal-to-noise ratio, one hundred 0.1 s duration measurements were averaged.A dark background was subtracted from all spectra before correcting with a calibration spectra measured for a standard light source that accounts for the detector spectral response.The experimental configuration was unaltered between sequential temperature measurements.
PLED Fabrication and Characterization.The PLED device architecture consists of a multilayer stack comprising ITO/ PEDOT:PSS/TFB/EML/LiF/Ca/Al.ITO anode structures on glass substrates (size 12 mm × 8 mm), which were cleaned for 15 min each in a sequence of ultrasonic baths using acetone, isopropanol, and detergent (Hellmanex III, 2% by volume in DI water).This was followed by oxygen plasma ashing in an Emitech K1050X.Next, a 35 nm thickness film of PEDOT:PSS (Clevios P VP) was deposited as a hole-injecting layer by spin-coating at 3000 rpm and annealing in air for 15 min at 135 °C.This was followed by spin-coating (at 1000 rpm) a 15 nm thickness electron-blocking TFB interlayer from 2 mg/mL toluene solution and then baking in nitrogen at 180 °C for 1 h.The 95F8:5BSP EML (60 nm thickness) was deposited on top of the TFB interlayer, again by spin-coating at 2500 rpm from a 10 mg/mL toluene solution.For the 90PFO/10PFB blend EML, PFO and PFB toluene solutions were separately prepared and mixed to give the desired weight ratio before spin-coating (2500 rpm) to a thickness of 60 nm.Finally, for PFO EML samples, toluene solutions were spincoated (2500 rpm) to a thickness of 60 nm.To induce β-phase chain segments in the EMLs, each sample was solvent vapor annealed in a toluene atmosphere at 50 °C for 2 h.An MBraun thermal evaporator was used to deposit the top cathode comprising a triple layer of LiF (2 nm), calcium (30 nm), and aluminum (100 nm).
PLEDs were characterized at room temperature in a sealed sample chamber under nitrogen, using a computer-controlled Keithley source measure unit to apply a bias voltage to the chosen pixel (each substrate accommodated 6 PLED pixels) and to measure the resultant current.A Minolta LS100 spot luminance meter measured the corresponding pixel luminance, and EL spectra were recorded using an Ocean Optics USB 2000 CCD spectrometer equipped with a fiber light collection bundle.Accelerated lifetime testing was performed using the same experimental apparatus.

Figure 2 .
Figure 2. Peak normalized optical absorption (solid) and PL emission (dashed) spectra for (a) PFO, (b) 95F8:5BSP copolymer, and (c) 90PFO/10PFB blend films spin-coated on Spectrosil substrates.Glassy phase data are shown with black lines, whereas β-phase data are shown with red lines.PL emission spectra were excited at λ ex = 385 nm.

Figure 3 .
Figure 3. PL decay transients for (a) glassy and (b) β-phase films of, from top to bottom, the 95F8:5BSP copolymer (red), 90PFO/10PFB blend (blue), and PFO homopolymer (black).The measured instrument response function is also shown (dashed gray line).Samples were excited at 404 nm, and the resulting PL emission intensity was monitored at 460 nm.

Figure 4 .
Figure 4. (a) Device structure of blue PLEDs and (b) the corresponding schematic energy level diagram.The polymer energy levels were deduced from CV measurements.The shaded gray area indicates the smaller energy gap for β-phase segments in PFO and 95F8:5BSP, whereas the pale green area indicates the energy gap for BSP units within the 95F8:5BSP copolymer.

Figure 5 .
Figure 5. PLED characteristics for glassy (filled symbols, (a), (c) and (e)) and β-phase (open symbols, (b), (d) and (f)) EML film microstructure devices.J−V−L data are plotted in (a) and (b), luminous (cd/A) and luminous power (lm/W) efficiency data as a function of luminance are plotted in (c) and (d), and associated EQE data are plotted in (e) and (f).95F8:5BSP copolymer EML data are plotted as red circles, 90PFO/10PFB blend data as blue triangles, and PFO homopolymer data as black squares.

Figure 8 .
Figure 8. Schematic energy level diagrams (left column) for intrachain energy transfer processes in 95F8:5BSP (a) glassy and (b) β-phase chains following optical excitation.Jablonski diagrams for each process are shown to the right of each schematic.See text for the explanation of numberings I, II, and III.

Table 3 .
Summary of Best PLED Performance for the Current Study Showing Turn-on Voltages (Defined as the Voltage at Which Luminance Reaches 1 cd/m 2 ), Peak Luminous Efficiency (cd/A), Luminous Power Efficiency (lm/W), and EQE (%) and Their Values at 100 and 1000 cd/m 2 and CIE (x, y) Color Coordinates for PFO, 95F8:5BSP Copolymer, and 90PFO/10PFB Blend EML Films with Both Glassy and β-Phase Microstructures

■
ASSOCIATED CONTENT * S Supporting Information The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsami.8b00243.UV−vis absorption and PL measurements for 97F8:3BSP, 90F8:10BSP, 80F8:20BSP, and PFB; additional TCSPC measurements in solution and in thin film; E-mail: donal.bradley@mpls.ox.ac.uk (D.D.C.B.).*E-mail: ji-seon.kim@imperial.ac.uk (J.-S.K.).ACKNOWLEDGMENTS I.H., N.C., N.J.C., and M.D. acknowledge the provision of UK Engineering and Physical Sciences Research Council (EPSRC) Plastic Electronics Doctoral Training Centre (EP/G037515/1) and Cambridge Display Technology Ltd (CDT) EPSRC CASE studentships.The authors would also like to acknowledge additional research funding from CDT, the University of Oxford, and EPSRC (EP/G037515/1) and thank CDT for supplying PFO, 95F8:5BSP, and PFB polymers.In addition, M.D. gratefully acknowledges support from the Marie Sklodowska-Curie Actions Innovative Training Networks "H2020-MSCA-ITN-2014 INFORM -675867".The authors also thank Dr Toshihiro Ohnishi of the Sumitomo Chemical Company Tsukuba Research Laboratory for supplying 97F8:3BSP, 95F8:5BSP, 90F8:10BSP, and 80F8:20BSP copolymers.The authors further thank Alexander Giovannitti for help with gel permeation chromatography and CV measurements.
*NotesThe authors declare no competing financial interest.■