Solar Energy Materials & Solar Cells 116 (2013) 262–269

Contents lists available at SciVerse ScienceDirect

Solar Energy Materials & Solar Cells

0927-02http://d

n CorrE-m1 Cu

Qassim

journal homepage: www.elsevier.com/locate/solmat

The influence of the nematic phase on the phase separation of blendedorganic semiconductors for photovoltaics

Steven A. Myers a,n, Manea S. Al Kalifah a,1, Chunghong Lei a, Mary O'Neill a,Stuart P. Kitney b, Stephen M. Kelly b

a Department of Physics and Mathematics, University of Hull, Hull HU6 7RX, UKb Department of Chemistry, University of Hull, Hull HU6 7RX, UK

a r t i c l e i n f o

Article history:Received 28 November 2012Received in revised form25 April 2013Accepted 7 May 2013Available online 6 June 2013

Keywords:PhotovoltaicsLiquid crystalsNematicSelf-assemblyNano-morphologyPhase transitions

48/$ - see front matter & 2013 Elsevier B.V. Ax.doi.org/10.1016/j.solmat.2013.05.006

esponding author. Tel.: +44 1482465396.ail address: S.A.Myers@hull.ac.uk (S.A. Myers)rrent Address: Department of Physics, FacultyUniversity, Kingdom of Saudi Arabia.

a b s t r a c t

Differential scanning calorimetry in combination with atomic force microscopy is used to examine thephase separation of a blended nematic liquid crystalline electron-donor and crystalline perylene electron-acceptor mixture. Separate domains of donor and acceptor material are mostly retained in the blend,although a small proportion of the acceptor, increasing with increasing donor concentration, is mixed inwith the donor domains. Annealing in the nematic phase allows the donor and acceptor molecules to moveand generate phase-separated domains of the required size, thus enhancing the performance of bulkheterojunction photovoltaic devices based on these blends. We show that the optimum annealingtemperature can be controlled by manipulation of the temperature range of the nematic phase of thedonor.

& 2013 Elsevier B.V. All rights reserved.

1. Introduction

Columnar, smectic and, more recently, nematic liquid crystalshave been shown to be very promising charge-transportingorganic semiconductors due to their ability to spontaneouslyself-assemble into highly ordered domains in uniform, thin filmsover large areas [1–3]. This combination of properties allied withbroad absorption spectra render them particularly suitable asactive materials for organic photovoltaics, such as bulk hetero-junction devices based on blends of liquid crystalline electrondonors and crystalline electron acceptors. Photogenerated excitonsare dissociated at the interface between the phase-separateddomains of the donor and acceptor components. The separatedelectrons and holes then travel along different and distinct con-tinuous pathways in opposite directions to the electrodes. Photo-voltaic cells comprising a coronene compound as a donor withcolumnar liquid crystalline phases blended with an electronacceptor showed power conversion efficiencies up to 1.5% understandard measurement conditions [4–6]. The performance of theseorganic photovoltaic (OPV) cells was optimized by effective use ofthermal annealing to control the degree of molecular ordering

ll rights reserved.

.of Science and Art at Alrass,

within the individual domains of electron donors and electronacceptors. Thioenothiophene polymers, which have high-temperature liquid crystalline phases, were blended with solublefullerene derivatives [7,8], with optimized devices having a powerconversion efficiency of 2.5%. It is noted that annealing in theliquid crystal phase improved the performance of photovoltaicdevices incorporating donors with either columnar or lamellarpolymeric phases [6,9,10]. A nematic liquid crystal composite witha porous surface and sub-micron scaled grooves has also beenused to provide a distributed interface to vertically separateelectron-donating and electron-accepting films in a bilayer PVdevice [11,12]. However, quite surprisingly, the use in OPVs of low-molar-mass nematic liquid crystals (small molecules) has not beenstudied to any significant degree [13]. The nematic phase of suchcompounds possesses a much lower viscosity than that of eitherhighly ordered columnar liquid crystals or high molecular weightliquid crystalline polymers, which are exceptionally viscous mate-rials. Therefore, nematic liquid crystals should offer significantadvantages in controlling the morphology of the donor–acceptorcomposites for organic photovoltaics, especially when using thermalannealing to control domain size and morphology.

This work addresses the important question of how the self-assembly properties of liquid crystals can be exploited to controlthe morphology of the donor–acceptor composites for organicphotovoltaics. Our model system consists of a nematic donorblended with a crystalline perylene based acceptor. We propose

S.A. Myers et al. / Solar Energy Materials & Solar Cells 116 (2013) 262–269 263

a combination of macroscopic and nanoscale techniques, differ-ential scanning calorimetry and atomic force microscopy, as aneffective method to investigate how the liquid crystalline phase ismodified by blending and whether the donors and acceptors areintimately mixed or occupy distinct domains. We examine whyannealing in the nematic phase improves morphology and opti-mises PV performance. We choose two distinct donor–acceptorblends, where both LC donors are electrically and optically verysimilar but have significant different LC temperature ranges, toconfirm the relevance of the LC phase.

Table 2The glass transition temperature (tg), melting point (Cr–N and Cr–I), clearing point(N–I) and recrystallisation temperature (I–Cr) for each of the four blends studied.The fifth column 5 records the enthalpy of melting (ΔE, per gram) of the acceptormolecule 3.

Material Blend molarratio (D:A)

tg(1C)

Cr–N(1C)

Cr–I(1C)

ΔE(J/g of 3)

N–I(1C)

I–Cr(1C)

1 80 2182 94 180 3033 – – 274 19.1 – 2562 and 3 1:3 93 180 256 16.4 – 209

1:2 93 180 257 11.5 – 2061:1 94 181 250 10.4 – 1814:3 94 180 250 8.7 – 180

1 and 3 1:3 80 – 265 10.1 206 225

2. Experimental

Table 1 shows the chemical structure of the materials used inthis investigation. The synthesis of the electron-donor materials1 and 2 is described in the Supplementary information. Thesynthesis of perylene acceptor 3 has been reported previously[12]. Cyclic voltammetry and absorption spectroscopy are used toobtain the ionization potential and electron affinity of the materi-als, as discussed in the supplementary information.

A Perkin-Elmer differential scanning calorimeter (DSC 7) wasused to measure the phase transition temperatures of the purematerials and blends, which are summarized in Table 2. Blendswere prepared by dissolving either donor 1 or donor 2 along withthe appropriate amount of acceptor 3 in dichloromethane as asolvent and then drop-casting the resultant solution into astandard aluminium differential scanning calorimetry (DSC) pan(Perkin-Elmer) using a pipette. Around 4–5 mg of material wasused for each DSC sample. The pan was prepared in this way inorder to simulate the blend mixture produced during thin filmprocessing. Donor 1 and acceptor 3 were mixed in a 1:3 mol ratio,whilst donor 2 and acceptor 3 were mixed in mole ratios of 4:3,1:1, 1:2 and 1:3. A standard indium sample was used to calibratethe DSC prior to all measurements. The scan rate was 10 1C min−1

and two heating–cooling cycles were performed for each blend.The morphology of the blended thin films was investigated withan Atomic Force Microscope (AFM) developed by MolecularImaging (Agilent) in the tapping mode. MAC Levers Type IIcantilevers (Agilent) were used for all measurements. PicoscanV.5.2 was used to produce the images, RMS roughness and heightdistribution data. Fast Fourier Transforms (FFT) and the powerspectral density data (PSD) were obtained using Gwyddion free-ware software. We investigated areas of 5�5 mm2 and 1�1 mm2,each area being composed of 1024�1024 data points.

Soda lime glass substrates were provided by UQG optics with a100 nm thick indium tin oxide (ITO) layer, giving a sheet resistanceof ≤20 Ω/sq. Acid etching was used to remove strips of ITO fromthe edges to allow contact to the cathode. The substrateswere cleaned by sonification and treated with oxygen plasma.

Table 1The chemical structure of the liquid crystalline electron-donor materials 1 and 2 and th

1

2

3

Poly(3,4-ethylenedioxythiophene) (PEDOT) doped with poly(styr-enesulfonate) (PSS) supplied by Baytron (P AI4083) was sonicatedfor 5 min, spin cast onto the ITO substrate at 4000 rpm for 30 s andthen annealed at 120 1C for 30 min followed by 5 min at 220 1C.The active layer was spincast from 1.5% by weight solution of thedonor and acceptor compounds mixed in various mole ratios inchlorobenzene at 2000 rpm for 30 s. This was followed by anneal-ing on a hot-plate under various conditions as described later. Thecathode consisted of a thin layer (∼0.6 nm) of lithium fluoride (LiF)and a layer (100 nm) of aluminium deposited by thermal evapora-tion under vacuum. The completed PV device was mounted in avacuum sealed test chamber for characterisation outside a glovebox. The devices were illuminated with a Xenon lamp (Bausch &Lomb), dispersed through a monochromator and attenuated withneutral light filters over an area of 0.25 cm2. The current–voltagecharacteristics of the photovoltaic devices were measured in aninert atmosphere using a Visual-Basic controlled picoammeter.

3. Results and discussion

3.1. Phases of binary donor–acceptor mixtures

The ionization potential and electron affinity of liquid crystal-line compounds 1 and 2 are the same (5.52 eV and 3.15 eV,respectively) within experimental error. Both of them act aselectron donors, when blended with the crystalline electronacceptor 3, see Table 1, which has an electron affinity of 4.19 eV.Fig. 1 shows the DSC traces for the donors 1 and 2 and the acceptor3, the blend comprising 2 and 3 as well as the blend 1 and 3 in a1:3 mol ratio. The transition temperatures of these compoundsand their blends, as well as those of other blends of 2 and 3 withdifferent mole ratios, are summarised in Table 2.

e crystalline perylene acceptor 3.

I N

Hea

t Flo

w (m

W)

Temperature (°C)

TgN I

18

19

20

21

22N ICr N

Tg

Hea

t Flo

w (m

W)

1st Cycle 2nd Cycle

Sample Temperature (°C)

16

18

20

22

24

26

Sample Temperature (°C)

Hea

t Flo

w (m

W)

Cr I

I Cr16

18

20

22

24

Hea

t Flo

w (m

W)

Sample Temperature (°C)

1st Cycle 2nd Cycle

Tgof 2 Cr N of 2Cr Iof 3

I C of 3

0 50 100 150 200 250 300 0 50 100 150 200 250 300 350

0 50 100 150 200 250 300 -50 0 50 100 150 200 250 300 350

0 50 100 150 200 250 300

10

15

20

25

30

35

40

18

20

22

24

26

I Cr of 3

Hea

t Flo

w (m

W)

Sample Temperature (°C)

1st Cycle2nd Cycle

Tg of 1 N I of 1

Crystallisation of 3

Cr I of 3

I N of 1

Fig. 1. DSC scans for compounds (a) 1, (b) 2, (c) 3, blends (d) 2 and 3 and (e) 1 and 3 in a 1:3 mol ratio.

S.A. Myers et al. / Solar Energy Materials & Solar Cells 116 (2013) 262–269264

The DSC data shown in Fig. 1a indicate that the donor 1 is anematic glass at room temperature. It has a glass transition at80 1C on heating and a nematic–isotropic phase transition at218 1C. Donor 2 is crystalline at room temperature and melts toform a nematic phase at 180 1C as illustrated in Fig. 1b. Theclearing point (N–I) is substantially higher (+85 1C) than that of1, see Table 1. The nematic phase is formed again on cooling2 from the isotropic liquid followed by the formation of a glassyphase (vitrification) on further cooling significantly below themelting point. As a consequence a nematic glass transition isobserved at 94 1C in the second heating cycle of 2. Acceptor3 exhibits no observable mesophase as shown in Fig. 1c. Com-pounds 2 and 3 were mixed together in four different mole ratios(4:3, 1:1, 1:2 and 1:3). Fig. 1d shows that the 1:3 mole ratio blendof donor 2 and acceptor 3 has a crystal–nematic melting transitionat 180 1C in the first heating cycle, most probably attributable tomelting of compound 2 present in the blend. This meltingtransition is replaced by a weak glass transition at temperature93 1C in the second heating cycle, implying that the nematic phasehas vitrified on cooling significantly below the melting point ofboth 2 and 3. Table 2 shows that the glass transition temperature(tg) and melting point (Cr–N) during heating, are fairly constant forall mole ratios of the blends comprising components 2 and 3 andthat they correlate closely with those of 2. A double-peakedcrystal–isotropic transition (Cr–I) is observed on heating in bothcycles at about 256 1C, see Fig. 1d. As Table 2 shows, this

temperature decreases only slightly for blends with higher con-centrations of donor material 2 and most probably corresponds tothe melting of the acceptor component 3 containing traces of thedonor 2. On cooling of the 1:3 blend, both components of themixture remain isotropic until 209 1C, when the isotropic–crystal(I–Cr) phase transition, i.e., recrystallisation, is observed. Thedegree of super-cooling observed depends on the mole ratio ofthe mixture. These results suggest that most of donor 2 is presentas phase-separated nematic domains in the blend alongsidecrystalline domains of the acceptor 3. However, the phase separa-tion of the two components is not complete as shown by examin-ing the enthalpy change per gram of acceptor 3 integrated over themelting points for the different blends. As the concentration ofdonor increases in the samples, proportionally less acceptor isincorporated into the distinct phase-separated domains. Thisimplies that small, but varying, fractions of donor and acceptorare intimately mixed.

The first heating and cooling cycle of the DSC scan of the blendcontaining donor 1 and acceptor 3 in the mole ratio 1:3, see Fig. 1e,shows distinct transitions for each material. The glass transitiontemperature and clearing point of the blend are very similar tothose of the pure donor 1 material. However, the melting pointand the recrystallisation transition of the blend are similar, but alittle lower, than those of the acceptor 3. However, we note thatthe enthalpy of the recrystallisation transition is only 68% of thatat the melting point, suggesting that a significant amount of

Table 4Data from bulk heterojunction PV devices obtained on excitation with 22 mW cm−2

at 466 nm. The blended film contained donor 1 and acceptor 3 in a 1:3 ratio bymole. Different annealing temperatures were used.

Devicenumber

Max. anneal. temp.(1C)

η

(%)EQE(%)

Jsc(A cm−2)

Voc

(V)FF

S.A. Myers et al. / Solar Energy Materials & Solar Cells 116 (2013) 262–269 265

3 remains mixed with 1 on cooling. This is confirmed in the secondcycle where an exothermic recrystallisation transition is observedon heating at 239 1C, which is probably attributable to the acceptor3 partially re-crystallising out of the blend. Thus, the DSC mea-surements suggest that mostly separate phases of donor andacceptor are retained in the blend.

D9 – 0.4 5.2 0.436 0.95 0.21D10 50 0.32 4.6 0.381 0.85 0.22D11 100 0.65 7.1 0.592 0.95 0.26D12 120 1.1 8.3 0.708 1.0 0.36D13 150 0.65 6.6 0.54 0.95 0.28D14 200 0.19 2.6 0.21 0.85 0.24

0.5 1.0 1.5 2.0 2.5 3.00.0

0.2

0.4

0.6

0.8

1.0

Blend Ratio Acceptor/Donor (Mol)

120C200C

� (%

)

3.2. Photovoltaic devices

Fig. 2 shows the current–voltage characteristics for a bulkheterojunction device containing a blend of 1 and 3 in the moleratio 1:3. The devices were irradiated with light of wavelength466 nm and of different irradiances as shown in the figure.

Photovoltaic devices were fabricated using the liquid crystallinecompound 2 as the donor and the crystalline perylene derivative3 as the acceptor, mixed in different ratios as collated inTables 3 and 4, which also tabulate the power conversion effi-ciency (η), external quantum efficiency (EQE), short circuit currentdensity (Jsc), open circuit voltage (Voc) and fill factor (FF) for thedevices on irradiation with 24 mW cm−2 at a wavelength of462 nm. The maximum value of Voc approaches the differenceequal to 1.34 eV between the LUMO energy of the acceptor and theHOMO energy of the donor. The fill factor is poor for all thedevices, but improves on annealing at 200 1C. The results showthat the performance of devices is better for blends with a greaterpercentage of the acceptor 3 in the blend. Table S1 in theSupporting information shows that the performance deterioratesfor blend ratios of 2 and 3 greater than 3:1. Similar trends werefound for the blend of 1 and 3, see Table S2. It is also improved by

-0.5 0.0 0.5 1.0 1.5 2.0

-2.0x10-4

0.0

2.0x10-4

4.0x10-4

6.0x10-422 mW cm-2

14.7 mW cm-2

9.28 mW cm-2

2.55 mW cm-2

Dark

Cur

rent

(A)

Voltage (V)

Fig. 2. The current versus voltage of the photovoltaic device incorporating a thinfilm of a 1:3 blend of 1 and 3. The inset labels the irradiance in mW cm−2 of theinput light source of wavelength 466 nm.

Table 3Data from bulk heterojunction PV devices obtained on excitation with 24 mW cm−2

at 462 nm. Different mole ratios of donor 2 and acceptor 3 and different annealingtemperatures are used.

Devicenumber

D:A moleratio

Max. anneal.temp. (1C)

η

(%)EQE(%)

Jsc(A cm−2)

Voc

(V)FF

D1 4:3 120 0.04 0.75 6.7�10−5 0.50 0.31D3 1:1 120 0.11 1.07 9.6�10−5 0.87 0.30D5 1:2 120 0.30 2.72 2.4�10−4 0.91 0.32D7 1:3 120 0.39 3.73 3.4�10−4 0.94 0.30D2 4:3 200 0.32 2.77 2.5�10−4 0.88 0.35D4 1:1 200 0.49 3.40 3.1�10−4 1.06 0.36D6 1:2 200 0.86 5.34 4.8�10−4 1.20 0.36D8 1:3 200 0.98 6.46 5.8�10−4 1.17 0.35

Fig. 3. The η of bulk heterojunction photovoltaic devices as a function of the moleratio of acceptor 3 to the donor 2 in the blended film annealed at either 120 1C or200 1C.

annealing at 200 1C as summarized in Fig. 3, which shows η as afunction of blend composition and annealing temperature. AsFig. 1d shows, the donor component of the blend is in thecrystalline state at 120 1C and in the nematic phase at 200 1C,suggesting that annealing in the (fluid) nematic phase is beneficialfor the device performance.

Based on the simulation of Peumans et al.[14], who studied theeffect of annealing at increasing temperatures on the morphologyof a donor–acceptor blend, it is assumed that the surface topo-graphy can give an insight into the bulk morphology and that thelateral features of the surface roughness correlate with the donor–acceptor domains. Fig. 4 shows 2D atomic force microscopy imagesof all eight films. The main images are 5�5 mm2 with the1�1 mm2 image inset. It is clear from the images for each blendratio that the roughness increases significantly when the sample isannealed at 200 1C, see Fig. 5. For the samples annealed at 120 1C,the 1:2 ratio blend exhibits the maximum rms roughness ampli-tude (equal to 1.6 nm). Annealing at 200 1C increases the rough-ness for all the mixtures up to a maximum value of 2.6 nm for theblend with the largest acceptor concentration. Fig. 4 also shows aclear difference in the texture of the samples annealed at differenttemperatures. The films annealed at 120 1C show phase separationon a significantly finer spatial scale compared with those annealedat 200 1C, where many of the grains are elongated and apparentlycrystalline, with well defined edges, particularly when the accep-tor concentration is high. The narrow diameter of the grain is ofthe order 50 nm, which is relatively large compared with theexpected exciton diffusion length of about 10 nm.

It is clear from the analysis of the blend of 2 and 3 that annealingin the nematic phase promotes phase separation of mixtures ofdonor and acceptor materials formed as a result of deposition fromsolution by spin casting and subsequent evaporation of the solvent.

Fig. 4. AFM topography image for thin films of donor 2 and acceptor 3 blended in the mole ratio (a) 4:3 mol, (b) 1:1 mol, (c) 1:2 mol, and (d) 1:3 mol. The samples on the leftand right were annealed at 120 1C and at 200 1C respectively. The main image size is 5 mm�5 mm with 1 mm�1 mm inset.

S.A. Myers et al. / Solar Energy Materials & Solar Cells 116 (2013) 262–269266

We now investigate the significance of annealing in the nematicphase for the blended materials 1 and 3, which have very differenttransition temperatures, correlating morphology and photovoltaicperformance.

Fig. 6 shows that the power conversion efficiency of devicesD9–D14 based on blends of 1 and 3mixed in a 1:3 mol ratio is verysensitive to the annealing temperature. Annealing at 50 1C appears

to have little effect whilst an annealing temperature of 120 1Cappears to be optimal for PV device performance, which issubstantially inferior at higher annealing temperatures. Table 2shows the same trends for EQE, Jsc, Voc and the fill factor. Fig. 1e(DSC) suggests that this particular blend is a nematic glass below108 1C, and then forms a nematic phase up to the clearing point of206 1C. Some of the dissolved acceptor crystallizes out at higher

1 2 3

1

2

31μm 120°C1μm 200°C

RM

S R

ough

ness

(nm

)

Blend Mole Ratio (Acceptor/Donor)

Fig. 5. RMS roughness measured from the 1 mm�1 mm image of each sampleshown in Fig. 4.

50 100 150 200

0.2

0.4

0.6

0.8

1.0

1.2

� (%

)

T (°C)

Fig. 6. η of bulk heterojunction photovoltaic devices as a function of annealingtemperature. Blends of donor 1 and acceptor 3 in the mole ratio 1:3 were used inthe devices.

S.A. Myers et al. / Solar Energy Materials & Solar Cells 116 (2013) 262–269 267

temperatures. Fig. 7 shows the surface morphology of the films onannealing at 50 1C, 120 1C and 200 1C. The rms roughness increaseswith temperature and is equal to 0.5 nm, 1.0 nm and 1.7 nm for thethree films studied. The AFM image shown in Fig. 7c clearly showsthat the acceptor has crystallized out at 200 1C, indicating that thenematic clearing point is lower than that determined for a bulksample using the DSC. This is not surprising as phase transitiontemperatures tend to be lower in thin films than in the bulk [15].Fig. 8 shows the normalized power spectral density for the threefilms. The plot is split into four regions to aid analysis of thedomain size. The lowest frequency (i) relates to domain sizeslarger than 100 nm, the second section (ii) relates to domain sizesfrom 20 nm to 100 nm, the third (iii) relates to domains from 5 nmto 20 nm and the highest frequency range (iv) indicates domainssmaller than 5 nm. Ideally, a morphology with domain sizes inregion (iii) will allow the majority of excitons to reach a donor–acceptor interface within the assumed 10 nm diffusion length.Domain sizes smaller than 5 nm, whilst good for charge separa-tion, may not allow efficient charge transport. These results showthat annealing in the nematic phase at 120 1C allows optimalphase separation by creating more domains of the required size.Larger domains are obtained by annealing at 200 1C, whilstannealing at low temperatures, i.e., in the nematic glassy state,gives a larger fraction of smaller domains. AFM images of sampleswhich were annealed in the nematic phase at 150 1C show asimilar morphology to that obtained at 120 1C. However, the rmsroughness of the former sample is significantly higher, equal to1.7 nm, which may account for the inferior performance of thecorresponding photovoltaic device.

3.3. Discussion

Despite the very different temperature ranges of the nematicphases, the best photovoltaic performance is found for the blendswhen the bulk heterojunction donor–acceptor films are annealedin the nematic phase. This is particularly surprising for the blendof 2 and 3, since annealing in the nematic phase also results inhigh surface roughness, which is normally associated with poorphotovoltaic performance [16,17]. The DSC for both blends sug-gests that the donor and acceptor components retain theirindividual phase behaviour to a large extent in the blended film.Annealing in the fluid nematic phase then allows the material tomove to phase-separated regions of improved size. Other authorsfind improved photovoltaic performance following annealing inliquid crystalline phases, for example, Sun et al. [18] reportedincreases of 200–400% in the Jsc using PCBM with liquid crystallineporphyrins. The improvement in efficiency was attributed tothermally induced alignment of the porphyrins in the columnardiscotic phase. This alignment provides efficient charge transportalong the columnar axis and optimized light harvesting. Recentlyan increase in efficiency from 0.22% to 0.65% was reported for asolar cell containing a blend of poly(3-hexylthiophene) mixedwith a liquid crystalline fullerene derivative, N-methyl-2-{4-[6-(4′-cyanobiphenyl-4-oxy)hexyloxy]phenyl}-3,4-fulleropyrrolidine,when annealing above the liquid crystalline transition tempera-ture. [19].

The intensity dependence of the photocurrent density can beexpressed by Jsc¼AIin

α, where Iin is the incident optical irradianceand A and α are constants; α¼1 implies monomolecular recombi-nation whereas the recombination of nongeminate electrons andholes is bimolecular with α¼0.5. Fig. 9 shows α as a function of theannealing temperature for both blend systems with the donor:acceptor mole ratio 1:3. It shows that annealing at the optimumtemperature produces photovoltaic devices, which have the high-est values of α, showing mono-molecular recombination. Attemperatures below the nematic phase the molecules are notsufficiently mobile in the vitrified glassy state to move from theirnon-ideal positions reached on evaporation of the solvent from thespin-coated film. The domain size is small, and possibly not wellphase-separated, increasing the probability of recombination ofthe free electrons and holes. This result agrees with a recent studythat found the optimum domain size for poly(3-hexylthiophene):PCBM blends to be around 6 nm, with respect to photovoltaicperformance [20]. Domains smaller than this were found to resultin a poorly percolated domain structure and hence increase theprobability of charge recombination. For the blends studied,bimolecular recombination also becomes more probable on theformation of large acceptor domains following high temperatureannealing of the blended 1 and 3.

Another key result is that the device performance of the blendof donor 2 and acceptor 3 is best for the highest concentration ofthe acceptor in the blend, which is well in excess of an equal moleratio with the donor. We have found this to be a general resultfound for other bulk heterojunction devices involving a nematicliquid crystalline donor combined with the acceptor 3. It may beassociated with more complete phase separation of the donor andacceptor, which also occurs with greater acceptor concentration asindicated by DSC in donor rich samples; proportionally lessacceptor is incorporated into the distinct phase separateddomains, resulting in poorer photovoltaic devices.

4. Conclusion

We show that differential scanning calorimetry combined withatomic force microscopy is an extremely effective method to

Fig. 7. AFM topography image for thin films of donor 1 and acceptor 3 blended in the mole ratio 1:3 on annealing at (a) 50 1C, (b) 120 1C and (c) 200 1C. The main image sizeis 5 mm�5 mm with a 1 mm�1 mm inset.

0.01 0.1 11E-6

1E-5

1E-4

1E-3

0.01

0.1

1

<5nm

(iv)<20nm>5nm

<100nm>20nm

Nor

mal

ized

PS

D

Spatial frequency (nm)-1

50C120C200C

(i) (ii) (iii)

>100nm

Fig. 8. Radially integrated power spectral density obtained using Fourier analysis ofa 2-dimensional topography image of the surface of films of the blended materials1 and 3 annealed for the temperatures stated in the inset.

0 50 100 150 200

0.6

0.7

0.8

0.9

1.0

Slo

pe α

T (°C)

Blend: 1 and 3Blend: 2 and 3

Fig. 9. Variation of α, defined in text, with annealing temperature for devices basedon blended thin films of donor and acceptors, mixed in the mole ratio 1:3; α¼1implies monomolecular recombination whereas the recombination of nongeminateelectrons and holes is bimolecular with α¼0.5.

S.A. Myers et al. / Solar Energy Materials & Solar Cells 116 (2013) 262–269268

S.A. Myers et al. / Solar Energy Materials & Solar Cells 116 (2013) 262–269 269

examine the phase separation of blended donors and acceptors forbulk heterojunction photovoltaic devices. Such studies would beusefully extended to more commonly investigated blends invol-ving crystalline polymeric donors and fullerene acceptors. Ourresults also highlight the usefulness of the nematic phase incontrolling phase separation. As demonstrated here, the temperaturerange of the nematic phase can be varied independently of theelectronic properties of the molecule. Hence, effective low tempera-ture annealing of blends could be achieved by the incorporation ofmolecules with low temperature nematic phases. These moleculescould also be polymerizable so that the optimized morphology issubsequently locked in by crosslinking.

Appendix A. Supporting material

Supplementary data associated with this article can be found inthe online version at http://dx.doi.org/10.1016/j.solmat.2013.05.006.

References

[1] M. Funahashi, J.I. Hanna, High carrier mobility up to 0.1 cm2 V−1 s−1 atambient temperatures in thiophene-based smectic liquid crystals, AdvancedMaterials 17 (2005) 594–598.

[2] A.M. Van De Craats, J.M. Warman, A. Fechtenkötter, J.D. Brand, M.A. Harbison,K. Müllen, Record charge carrier mobility in a room-temperature discoticliquid-crystalline derivative of hexabenzocoronene, Advanced Materials11 (1999) 1469–1472.

[3] K.L. Woon, M.P. Aldred, P. Vlachos, G.H. Mehl, T. Stirner, S.M. Kelly, M. O'Neill,Electronic charge transport in extended nematic liquid crystals, Chemistry ofMaterials 18 (2006) 2311–2317.

[4] L. Schmidt-Mende, A. Fechtenkötter, K. Müllen, E. Moons, R.H. Friend,J.D. MacKenzie, Self-organized discotic liquid crystals for high-efficiencyorganic photovoltaics, Science 293 (2001) 1119–1122.

[5] J. Li, M. Kastler, W. Pisula, J.W.F. Robertson, D. Wasserfallen, A.C. Grimsdale,J. Wu, K. Müllen, Organic bulk-heterojunction photovoltaics based on alkylsubstituted discotics, Advanced Functional Materials 17 (2007) 2528–2533.

[6] W.W.H. Wong, T. Birendra Singh, D. Vak, W. Pisula, C. Yan, X. Feng,E.L. Williams, K.L. Chan, Q. Mao, D.J. Jones, M. Chang-Qi, K. Müllen,P. BÃuerle, A.B. Holmes, Solution processable fluorenyl hexa-peri-

hexabenzocoronenes in organic field-effect transistors and solar cells,Advanced Functional Materials 20 (2010) 927–928.

[7] I.W. Hwang, J.Y. Kim, S. Cho, J. Yuen, N. Coates, K. Lee, M. Heeney,I. McCulloch, D. Moses, A.J. Heeger, Bulk heterojunction materials composedof poly(2,5-bis(3- tetradecylthiophen-2-yl)thieno[3,2-b;]thiophene): ultra-fast electron transfer and carrier recombination, Journal of Physical Chem-istry C 112 (2008) 7853–7857.

[8] N.C. Cates, R. Gysel, Z. Beiley, C.E. Miller, M.F. Toney, M. Heeney, L. McCulloch,M.D. McGehee, Tuning the properties of polymer bulk heterojunction solarcells by adjusting fullerene size to control intercalation, Nano Letters 9 (2009)4153–4157.

[9] J.P. Schmidtke, R.H. Friend, M. Kastler, K. Müllen, Control of morphology inefficient photovoltaic diodes from discotic liquid crystals, Journal of ChemicalPhysics 124 (2006).

[10] K. Yao, Y. Chen, L. Chen, F. Li, X. Li, X. Ren, H. Wang, T. Liu, Mesogens mediatedself-assembly in applications of bulk heterojunction solar cells based on aconjugated polymer with narrow band gap, Macromolecules 44 (2011)2698–2706.

[11] M. Carrasco-Orozco, W.C. Tsoi, M. O'Neill, M.P. Aldred, P. Vlachos, S.M. Kelly,New photovoltaic concept: liquid-crystal solar cells using a nematic geltemplate, Advanced Materials, 18 (2006) 1754-1758.-+.

[12] W.C. Tsoi, M. O'Neill, M.P. Aldred, S.P. Kitney, P. Vlachos, S.M. Kelly, Dis-tributed bilayer photovoltaics based on nematic liquid crystal polymernetworks, Chemistry of Materials 19 (2007) 5475–5484.

[13] C. Lei, M.S. Al Khalifah, M. O'Neill, M.P. Aldred, S.P. Kitney, P. Vlachos,S.M. Kelly, Calamitic liquid crystal blends for organic photovoltaics, in:Proceedings of SPIE—The International Society for Optical Engineering, 2008.

[14] P. Peumans, S. Uchida, S.R. Forrest, Efficient bulk heterojunction photovoltaiccells using small-molecular-weight organic thin films, Nature 425 (2003)158–162.

[15] R.A.L. Jones, The dynamics of thin polymer films, Current Opinion in Colloidand Interface Science 4 (1999) 153–158.

[16] S.H. Lee, D.H. Kim, J.H. Kim, G.S. Lee, J.G. Park, Effect of metal-reflection andsurface-roughness properties on power-conversion efficiency for polymerphotovoltaic cells, Journal of Physical Chemistry C 113 (2009) 21915–21920.

[17] H.C. Hesse, J. Weickert, M. Al-Hussein, L. Dassel, X. Feng, K. Müllen,L. Schmidt-Mende, Discotic materials for organic solar cells: effects ofchemical structure on assembly and performance, Solar Energy Materialsand Solar Cells 94 (2010) 560–567.

[18] Q. Sun, L. Dai, X. Zhou, L. Li, Q. Li, Bilayer and bulk heterojunction solar cellsusing liquid crystalline porphyrins as donors by solution processing, AppliedPhysics Letters 91 (2007) 253505.

[19] P. Wang, K. Yao, L. Chen, Y. Chen, F. Li, H. Wang, S. Yu, Self-assembledmesogens modified fullerene for efficiently stable bulk heterojunction solarcells, Solar Energy Materials and Solar Cells 97 (2012) 34–42.

[20] Y.M. Nam, H. Huh, W.H. Jo, Optimization of thickness and morphology ofactive layer for high performance of bulk-heterojunction organic solar cells,Solar Energy Materials and Solar Cells 94 (2010) 1118–1124.