Electromodulation and magnetomodulation of exciton dissociation in electron donor (starburst amine) : electron acceptor (bathocuproine) system

Organic Electronics 25 (2015) 362–376

Contents lists available at ScienceDirect

Organic Electronics journal homepage: www.elsevier.com/locate/orgel

Electromodulation and magnetomodulation of exciton dissociation in electron donor (starburst amine) : electron acceptor (bathocuproine) system Daniel Pelczarski a, Piotr Grygiel a,⇑, Karol Falkowski a, Maciej Klein a,b, Waldemar Stampor a a b

´ sk University of Technology, 80-233 Gdan ´ sk, Poland Department of Physics of Electronic Phenomena, Gdan ´ sk, Poland Center for Plasma and Laser Engineering, The Szewalski Institute of Fluid-Flow Machinery (PASci), 80-231 Gdan

a r t i c l e

i n f o

Article history: Received 1 June 2015 Received in revised form 29 June 2015 Accepted 30 June 2015

Keywords: Photoconduction Charge photogeneration Magnetic effects Starburst amines

a b s t r a c t Electric field dependencies of electromodulated photoluminescence and photocurrents as well as the magnetic field effects on photocurrents, photovoltaic characteristics, electromodulated photoluminescence and photoluminescence have been investigated in vacuum evaporated films of m-MTDATA:BCP (4,40 ,400 -tris(N-(3-methylphenyl)-N-phenylamino)triphenylamine and bathocuproine) system. The electromodulation processes do remain in accordance with Onsager as well as with the Sano–Tachiya– Noolandi–Hong formalisms of electron–hole pair separation. While the electromodulated photocurrents are due to operation of both long-radius and short-radius e–h pairs, the electromodulated photoluminescence quenching is related to merely short-radius fraction of the e–h pairs involved in the exciplex creation process. The photocurrents, photovoltaic characteristics, photoluminescence and electromodulated photoluminescence are influenced by external magnetic field of the hyperfine coupling (HFC) scale which modulates the singlet–triplet intersystem crossing of long-radius e–h pairs. Ó 2015 Elsevier B.V. All rights reserved.

1. Introduction The rapid increase in application of organic materials for manufacturing photovoltaic (PV) cells and electroluminescent (EL) diodes is nowadays accompanied by an intensive research in electrical and optical properties of organic thin films [1–3]. Among others, the charge recombination and exciton dissociation are the matter of exceptional interest since they are generally recognized as the basic electronic processes limiting the efficiency of organic EL and PV devices. These mechanisms in low-mobility organic materials are usually assumed to proceed via the intermediate stage of an electron–hole (e–h) pair of finite lifetime, but the knowledge on the subject is still far from complete. In particular, it is not known whether the same e–h pairs are involved in dissociation/recombination processes in the PV and EL devices which in turn addresses the issue of the intercarrier distance within these pairs [3]. An effective tool to resolve this problem is to trace the e–h distance by observation of influence of applied external electric and magnetic fields on the e–h pair dissociation and recombination products. In particular, under strong electric field the sample photocurrent and the electric-field assisted ⇑ Corresponding author. E-mail address: [email protected] (P. Grygiel). http://dx.doi.org/10.1016/j.orgel.2015.06.050 1566-1199/Ó 2015 Elsevier B.V. All rights reserved.

photoluminescence quenching are investigated as a result of the Coulomb-bound e–h pair (exciton) dissociation. To develop the models of relevant mechanisms the Onsager approach to geminate recombination as well as the Langevin approach to the bimolecular recombination are typically used. The usage of magnetic field in turn makes it possible to probe effectively the different intercarrier distances in spin-correlated e–h pairs by investigation of magnetic field effects on photocurrent, photoluminescence and electroluminescence [3,4]. Essentially, if the spin coherence time of the e–h pairs is long enough and the electrostatic electron exchange interaction is sufficiently weak for efficient spin evolution to occur, the magnetic field-dependent intersystem crossing (ISC) between the singlet, 1 ðe—hÞ, and triplet, 3 ðe—hÞ, pair spin states can take place [5]. This process does modulate the effectiveness of various e–h pair decay channels, hence the production of emissive states in the EL processes and the charge carrier population in the dissociation events in the PV devices [6–9]. It should be noted here that determination of the intercarrier distance and binding energy of the e–h pairs is essential for construction of exciton-dissociation models to optimize the efficiency of EL and PV devices. Indeed, sufficiently low electrostatic exchange energy of the e–h pairs allowing effective singlet–triplet spin mixing is required if the improvement of the electrofluorescent diodes over the pure spin

D. Pelczarski et al. / Organic Electronics 25 (2015) 362–376

statistic 25%-limit is considered using the reverse intersystem crossing (RISC) [10]. The origin of the magnetic field effects remains, however, under current debate (for recent reviews refer to Refs. [4,11]. Although the effects in weak magnetic fields are commonly interpreted in terms of the electron–hole conversion (EHP) model (see e.g. Refs. [7–9,12]), the trion model [4,13,14] as well as the bipolaron (BP) mechanism (see Refs. [15,16]), some experimental results obtained under strong magnetic fields cannot be rationalized in such a way. In particular, in Ref. [17] a simultaneous increase in fluorescence and phosphorescence in electrically excited layers of PhLPPP (phenyl-substituted derivative of the prototypical conjugated polymer ladder-type poly(p-phenylene) with trace concentrations of palladium atoms) under magnetic field intensity up to 8 T was reported which rejects any mechanisms involving the magnetic-field-driven conversion between the singlet and triplet spin pairs leading to the fluorescence intensity increase accompanied by the decrease in phosphorescence output or vice versa. It is generally recognized that formation of intermolecular excited-state complexes (exciplex states) can take place as a result of charge transfer processes at the interface between the separate layers or mixed molecules of an organic hole transporter (electron donor, D, with lower ionization potential) and an organic electron transporter (electron acceptor, A, with larger electron affinity), one of them being excited [3,18]. The photoinduced electron transfer between D and A molecules promotes the intermolecular charge separation which is profitable in PV applications whereas the broadband emission from electrically induced exciplex states is favoured in development of new-generation sources of light. Therefore, the exciplex-based optoelectronic elements are nowadays the subject of intense studies with substantial effort aimed at fabrication of effective bifunctional, EL and PV, devices. For instance, the usage of m-MTDATA (4,40 ,400 -tris(N-(3-methylpheny l)-N-phenylamino)triphenylamine) and BCP (bathocuproine), respectively, as the electron donor and the electron acceptor to construct a PV/EL diode has been demonstrated with expectations of good performance applications [19], therefore, we decided to examine the exciton dissociation and charge recombination processes limiting the efficiency of such devices. In this paper we investigate the vacuum-evaporated solid films of m-MTDATA:BCP system sandwiched between Al electrodes. The electric field dependencies of electromodulated photoluminescence and photocurrents as well as the magnetic field effects on photocurrents, photovoltaic characteristics, electromodulated photoluminescence and photoluminescence are analyzed to address the problem of the intercarrier distance of the e–h pairs involved in the relevant dissociation/recombination processes. An attempt has been undertaken to unveil the mechanisms responsible for spin-dependent processes underlying the observed magnetic field effects.

2. Experimental details The quartz/Al/m-MTDATA:BCP[1:1]/Al sandwich structures were prepared as follows. Films of organic material mixture (for molecular structures of both compounds see Fig. 1a) were obtained by simultaneous thermal vacuum (5  103 Pa) deposition of the components from separate quartz crucibles onto quartz slides with previously evaporated semitransparent substrate electrodes. With temperatures of the crucibles adjusted to 270 °C for m-MTDATA and 175 °C for BCP the common average evaporation rate of ca. 15 nm/s was achieved to obtain the mixture films of controlled, ca. 1:1, mass ratio (the densities of both components are similar). The layer thicknesses were measured by means of a Tencor Alpha Step 500 Profiler. As results from the X-ray diffraction

363

measurements the obtained co-evaporated layers exhibit the quasi-amorphous structure [20]. Sandwich structures were formed by vacuum deposition of semitransparent counter-electrodes. Both metal films were evaporated at pressure of about 9  103 Pa giving optical transmittance of 5–20% in visible and next dried in air for ca. 10 min. The energy level diagram with the Fermi level of Al electrodes and frontier molecular orbital HOMO (highest occupied MO) and LUMO (lowest unoccupied MO) levels of m-MTDATA and BCP solid films are depicted in a flat band fashion in Fig. 1b. Note that the dark carrier injection ability of the electrodes into the solid mixture is rather poor due to high energy barriers both for the electron and hole injection. As usual for organic materials, the samples were stored in darkness and in ambient dried air atmosphere for several days before measurements. When necessary the electrical contacts between the sample electrodes and the relevant circuits were accomplished by drops of carbon-containing emulsion. The absorption (ABS) spectra of system single components as well as component mixtures, both in the diluted and the solid phase, were recorded with the use of a Perkin Elmer, UV/VIS Lambda 10 spectrometer at fixed spectral line width of 2 nm. To obtain solutions the tetrahydrofuran (THF) was utilized as a solvent since its absorption spectrum does not coincide with the spectra of m-MTDATA and BCP. The photoluminescence (PL) spectra of all solutions and evaporated layers were in turn recorded with a Perkin Elmer, LS55B spectrofluorometer, also at fixed spectral width of excitation and emission lines being as high as 7.5 and 2.5 nm, respectively. Additionally, a Schott WG2 optical filter was used for the FL measurements of the BCP and m-MTDATA films, whereas for m-MTDATA:BCP ones a Corning 3-72 filter was applied. The spectral measurements were carried out with the spectral interval equal to 0.5 nm. In the photoluminescence (PL) and photocurrent experiments a mercury lamp (Narva, HBO 200 W) or a xenon lamp (Osram, XBO 150W) followed by a Zeiss, SPM-2 monochromator was used as a source of sample illumination light. The photoluminescence of the m-MTDATA:BCP mixture layers was excited through the active electrode area of 0.2 cm2 (mercury lamp, kexc = 313 nm) and the electric field-modulated photoluminescence (EML) was induced by a sinusoidally time-dependent field, FðtÞ ¼ F 0 sinðxtÞ (at typical value of x/2p = 175 Hz), applied to the electrodes of the sample. The sample PL light was collected with a quartz lightguide followed by an Optometrics, SDMC1 monochromator together with a set of appropriate Schott and Corning cut-off glass filters and then its intensity was measured by an EMI, 9863QB photomultiplier tube driving a EG&G Princeton Applied Research, model 181 current preamplifier. To avoid effectively the undesirable detection of exciting light, the lightguide rod was placed behind the sample in a position out of the excitation light path, at the angle of 27° towards the layer plane (this angle was chosen arbitrary and has no special meaning). The voltage signal from the preamplifier consisted of a steady-state and alternating components corresponding to the field-modulated photoluminescence of the organic layer, I(t), according to the Fourier series, IðtÞ ¼ RInx ðtÞ, (n = 0, 1, 2,. . .). The zeroth-order component, I0x, was measured with a dc-voltmeter whereas the second harmonic, I2x(t), was extracted and measured using of a lock-in amplifier, EG&G Princeton Applied Research, model 5210 referenced by the PL-modulating signal. The point of interest in this EML experiment is the value of the ratio recorded as a function of the rms electric field strength, FRMS, and excitation wavelength, here defined as

ð2xÞEML ¼

I 2x I0x

ð1Þ

where I0x denotes the steady-state Fourier component and I2x – the rms value of the second-order Fourier component of the

364

D. Pelczarski et al. / Organic Electronics 25 (2015) 362–376

(a)

CH3

N

(b)

m-MTDATA CH3

2.0 eV

N

3.0 eV N

N

m-MTDATA

Al 4.2 eV

H3C

BCP

Al 4.2 eV

5.0 eV 6.5 eV

BCP N

H3C

N

CH3

Fig. 1. Molecular structure (a) and energy levels (b) of a sandwich cell Al/m-MTDATA:BCP/Al.

electromodulated sample photoluminescence intensity (for detailed description of the EML method see the Ref. [21]). Note that the Formula (1) represents the relative change of PL intensity caused by external electric field. A positive (2x)EML signal means the appearance of PL quenching, i.e. the PL intensity being diminished by increasing electric field. The photocurrent (PC) measurements were performed utilizing a conventional measuring circuit with a Keithley, 485 electrometer and an adjustable stabilized power supply The sample holder having an aperture with the diameter of 2 mm determining the active area of the sandwiches was mounted between the pole pieces of an electromagnet in the way that the magnetic field was parallel to the plane of the samples. The samples were excited by the illumination setup with a mercury or xenon lamp through a 1 meter-long linear quartz waveguide. The sample photon flux was controlled by measuring of photocurrent (Keithley, 485 picoammeter) generated by a Si-photodiode (Hitachi, S1336-18BQ) illuminated via light beam splitter and a reference metal layer which was evaporated onto additional quartz slides placed in the evaporation chamber together with sample substrates. For further analysis, the photocurrent density action spectra and the photocurrent densities versus exciting photon flux have been recorded. To investigate the electron–hole pair dissociation leading to the photoconductivity of the samples, from the recorded photocurrents the charge photogeneration efficiencies have been derived according to the formula

g¼

j=e /0 =hm½1  expðadÞ

ð2Þ

with j being the photocurrent density, e – the elementary charge, /0 – the incident photon flux per cm2, a – the linear absorption coefficient of the organic material defined by the Bouguer– Lambert formula, / ¼ /0 expðadÞ (d stands for a layer thickness). For the measurements of magnetic field effects on the photocurrents the sample electromagnet was driven by an adjustable stabilized dc-power supply. A Hall-effect probe of a Wuntronics, Koshava 5 magnetometer was mounted close to the sample to determine the magnetic field intensity. The magnetic field effect (hence the MPC signal) is here defined as a relative change of the photocurrent density being the function of the field of strength B applied to the sample. Thus, the magnitude of the MPC signal is given by the ratio

MPC ¼

jðBÞ  jð0Þ jð0Þ

ð3Þ

where j(B) and j(0) are the photocurrent densities with and without magnetic field, respectively. The magnetic field effect on photovoltaic (PV) characteristics was investigated on the heterojunction-type devices of the structure Al/LiF/m-MTDATA:BCP/MoO3/ITO prepared by thermal vacuum deposition of subsequent layers on ITO-covered quartz substrate, as already described in this Section, with the LiF and MoO3 layers having the thickness of 2–5 nm. The samples positioned in the electromagnet-magnetometer-quartz waveguide system were excited by the illumination setup with a mercury lamp. To record the PV characteristics the current intensities were measured at constant illumination as functions of external (bias) voltage applied to the structures from an adjustable source. As a measuring device a Gamry Instruments, potentiostat/galvanostat Interface 1000 was used. The magnitudes of the magnetic field effect were calculated from the Formula (3) taking relevant intensities of the device photocurrents. In the experimental setup for determining the magnetic field effect on the photoluminescence (MPL) the mercury lamp was used as a source of sample illumination. To cut-off the parasitic light, an additional 313 nm-interference filter (Edmund Optics or Zeiss Jena) was mounted between the monochromator output and a 1 m long quartz lightguide rod positioned in the way that the exciting beam was perpendicular to the sample surface. The illuminated active area of the sandwiches was determined by the sample holder aperture with the diameter of 2 mm. Close to the sample a Hall-effect detector was placed to control the magnetic field strength. The FL light was collected by an another 1 m-long quartz lightguide positioned behind the sample at the angle of ca. 27° towards the layer plane, out of the excitation beam range. A set of cut-off Schott filters, GG7 positioned between the lightguide and the photomultiplier tube (EMI, 9863QB) for detection of the FL organic layer response ensured the satisfactory elimination of the exciting light. The current from the photomultiplier was then converted into a voltage signal by a Stanford Research Systems, model SR570 current preamplifier. Since the useful signal was usually very weak and noisy, the phase-sensitive technique had to be utilized for effective measurements. In this technique the sample is influenced by a magnetic field (parallel to the layer surface) which consists of a constant component modulated by an alternating one varying sinusoidally with time. While the constant field is delivered by a permanent magnet, the alternating component is generated by a system of coils driven by a set of generator and an operational amplifier. The voltage signal from the preamplifier is measured by a lock-in amplifier (EG&G Princeton

D. Pelczarski et al. / Organic Electronics 25 (2015) 362–376

Applied Research, model 5210) referenced by the signal driving the coils. The lock-in readouts, corresponding to the change in sample photoluminescence, dI, was divided by the field modulation amplitude to get the dI(B)/dB (see Ref. [22]) from which, by integration, the MPL curves were obtained. All measurements were carried out at room temperature in ambient atmosphere conditions. During the construction of the experimental setups special care was taken to reduce the parasitic magnetic and electric fields nearby their sensitive parts to a negligible level which was carefully controlled prior to the measurements. 3. Results and discussion 3.1. Absorption and fluorescence spectra The absorption (ABS) and photoluminescence (PL) spectra of the single components and two-component system in tetrahydrofuran (THF) 105 M solution and in vacuum-evaporated films are displayed in Fig. 2a and b, respectively. The star-burst amine m-MTDATA is a first-generation dendrimer built on the basis of a triphenylamine (TPA). The m-MTDATA molecule contains three methyl-substituted TPA derivative branching units centered in a nitrogen core. The TPA units are in strong conjugation with each other via the lone pair at the nitrogen atom (cf. Fig. 1a). The ABS spectrum of m-MTDATA diluted in THF (Fig. 2a) represents a structure of two peaks at 313.5 and 341 nm being more complex and distinctively shifted to the red as compared to corresponding spectrum of TPA monomer which exhibits a single asymmetric curve with maximum at 299 nm (see Ref. [23]). These features, also reported for other TPA-based dendrimers have been interpreted in the

(a) 8

ABS

ABS 10 M

0.8

-5

FL 10 M

BCP Mixture m-MTDATA : BCP

4

4

-3 -3 FL 10 M :10 M ( λexc = 310 nm )

-4 -3 FL 10 M : 10 M

2

(b)

(λexc = 280 nm)

300

400 500 λ [nm] BCP MTDATA

ABS

0.4

600

0.0

BCP

1.2

0.8

0.4

FL [a.u.]

-5

0

ABS, FL, EML [a.u.]

1.2

Single components

6

-1

-1

ε [10 M cm ]

MTDATA

in THF

FL BCP MTDATA

MTDATA : BCP

FL

I 2ω

EML

MTDATA

λexc = 313nm

EML I2ω/ I0ω

0.0

λem = 525nm

300

400

λ [nm]

500

600

Fig. 2. Absorption (ABS) and fluorescence (FL) spectra of single components (mMTDATA, BCP) and mixtures (m-MTDATA:BCP) in THF (a) and in solid films (b). The wavelengths of excitation light were kexc ¼ 280 nm for BCP and kexc ¼ 310 nm for m-MTDATA and m-MTDATA:BCP system. In part (a) the FL spectra for concentrated solutions of 103 M m-MTDATA:103 M BCP (dashed line) and 104 M mMTDATA:103 M BCP (dotted line) were compared with diluted solutions (105 M) of single components (thick solid lines). In part (b) the excitation spectra (squares) and emission spectra (circles) of EML signal for vacuum evaporated (1:1) solid films of m-MTDATA:BCP are also shown.

365

framework of the Frenkel exciton theory [24]. A high delocalization degree of primary excitons over the branches of the entire m-MTDATA molecule after its photoexcitation has also been proved by means of the electric-field absorption (EA) spectroscopy technique [23] since the EA curves could be qualitatively rationalized in terms of exciton interaction between the TPA segments of the dendrimer. Moreover, for consistent interpretation of the EA spectra the presence of degenerate m-MTDATA states had to be assumed. It should be noted here that the ABS spectrum of m-MTDATA film (Fig. 2b) with maxima detected at 321 and 346 nm resembles the solution curve from Fig. 2a being matrix-shifted towards the red only by several nanometers which implies the molecular origin of the solid phase absorption. The bathocuproine (BCP) molecule consists of a phenanthroline core having an aromatic p-conjugated frame that contains two nitrogen atoms placed in the 1 and 10 positions together with two phenyl rings substituted in the positions 4 and 7 as well as with two methyl groups substituted in the positions 2 and 9 (cf. Fig. 1b). The BCP molecule is non-planar and the dihedral angle between the core plane and the phenyl groups is as high as 47.9° [25]. As seen from Fig. 2, the ABS spectrum of BCP film resembles that of BCP diluted in THF except for the long-wavelength tail clearly seen beyond the solution absorption limit of approx. 370 nm which may be related to the dimers and defect sites as well as to light scattering on crystallites present in the evaporated layer [26]. The rather small shift to the red of the film spectrum with maximum at 281 nm with respect to that of diluted BCP with maximum at 278 nm suggests the insignificance of the matrix shift induced by non-resonance interaction between excited molecules and their ground state solid environment. Therefore, the electronic states in a film should be spatially well-localized within the BCP molecules and can be regarded as Frenkel excitons. The recognition of low-energy singlet excited states (in a solid film and a solution) as Frenkel excitons carrying intramolecular charge transfer (CT) states was also necessary to rationalize the results of the EA-spectra measurements in vacuum evaporated BCP layers [27]. The photoluminescence (fluorescence, FL) spectrum of THF-diluted m-MTDATA exhibits a maximum at 423 nm whereas that of evaporated film is similar in shape, but apart from a maximum at 428 nm its vibronic replica at 460 nm is clearly seen (cf. thick solid lines in Fig. 2a and b, respectively). Furthermore, the FL spectrum of solid BCP reproduces the spectral curve for THF solution of BCP (with maxima for 384 and 388 nm) except for pronounced long-wavelength tail which should be rather assigned to associative emission of dimers with decay time significantly longer than that of singlet excitons of monomers [26]. As regards both compounds, the matrix (solvent) shift of the fluorescence is similar to that seen for the ABS curves. This similarity implies that short-wavelength FL of solid films, as in the case of solutions, should be of molecular origin. The origin of the Stokes shift of FL spectra for m-MTDATA and BCP layers may in turn be assigned to a direct capture of migrating extended excitons by structural defects or self-trapping caused by strong exciton–phonon coupling [28]. Consider now the FL spectra recorded for THF-diluted m-MTDATA:BCP mixtures from Fig. 2a. Here, the dashed line represents the fluorescence of a 103 M: 103 M system for excitation wavelength kexc = 310 nm from the range of very high absorption of m-MTDATA exceeding substantially that of BCP. As seen, the fluorescence curve of the mixture reproduces approximately the FL spectrum of pure m-MTDATA. Interesting results have been obtained for 104 M: 103 M solution of m-MTDATA and BCP excited with kexc = 280 nm for which the light absorption of BCP prevails the absorption of m-MTDATA. Indeed, despite the tenfold higher concentration of BCP molecules the location and shape of the spectrum plotted in Fig. 2a by dotted line remain again very

366

D. Pelczarski et al. / Organic Electronics 25 (2015) 362–376

close to that recorded for m-MTDATA diluted solution (solid line) implying that the fluorescence of the latter compound contributes predominantly to the FL spectrum of the mixture. Provided similar values of the quantum fluorescence yields of both materials, this feature suggests the non-radiative transfer of electronic excitation energy between the BCP and m-MTDATA molecules in accordance with the Förster mechanism [29] which seems to be justified since the FL spectrum of BCP overlaps the ABS spectrum of m-MTDATA (cf. corresponding curves in Fig. 2a). Assuming that the unknown fluorescence quantum yield of the energy donor (BCP) is similar to that of 1,10-phenanthroline in a polar solvent [30], uD = 4  103, with THF refractive index, n = 1.4, one can then calculate the value of the Förster radius, R0 = 1.43 nm. Analogically, for quantum yield of a typical fluorescent dye as high as uD = 4  101 one obtains R0 = 3.07 nm which remain in accordance with the literature data [31]. From the plots of Fig. 2b it is seen that in the case of evaporated layers an additional structureless FL broad band with maximum at 522 nm appears shifted towards the red as compared to the FL spectra of each of the system monomer components. This spectrum ascribed to the emission of exciplexes is similar to the spectral curve recorded for m-MTDATA:BCP 1:1 mixed films from Ref. [19] as well as to that for m-MTDATA:BPT 1:1 blended films (BPT – the bathophenanthroline i.e. BCP modified by removing the 2,9-positioned methyl groups) [32]. The exciplex emission, however, has not been detected in our investigations of the THF solutions, also for high concentrations of the mixture components. This indicates that torsional degrees of freedom of molecules do prevent from special orientations characteristic for exciplex formation. It is generally recognized that a singlet exciplex, 1 ðDAÞ , is created via electron transfer in an encounter of a donor (D) and an acceptor (A), one of them being excited, D⁄ + A0 or D0 + A⁄. The wavefunction of an exciplex can then be written in the form of [31]

wexciplex ¼ awD A0 þ bwD0 A þ cwDþ A þ dwD Aþ ;

ð4Þ

where the functions awD A0 and bwD0 A are related to the electronic excitation of donor and acceptor whereas the components cwDþ A and dwD Aþ – to the charge transfer process (the symbols D and A+ are for the donor and acceptor ions, respectively). The relation between corresponding amplitudes of the components of Function (4) depends on the ionization potential of the donor, ID, as well as on the acceptor electron affinity, EA. These two quantities determine also the maximum energy of the exciplex emission band [3],

hmexciplex ¼ ID  EA  EC

ð5Þ

where EC > 1 eV stands for the coulombic attraction energy between the D and Aþ ions. It is known from the literature that upon the optical excitation of donors, D0 , to the singlet states, 1 D , the formation of singlet exciplexes, 1 ðDAÞ , followed by exciplexes emission can proceed in a direct manner proposed by Weller [33,34] and widely accepted [35,36] or indirectly (see Ref. [37,38]), the latter process involving the geminate electron–hole (e–h) pairs, 1 ðDþ ::A Þ, which corresponds to the radical ion pairs from the dynamic spin chemistry in solutions. The pathways of both mechanisms are depicted in Fig. 3. Accordingly, the geminate e–h pairs are created from the excited encounter complexes, ð1 D ::A0 Þ, at the expense of their electron affinity energy difference due to electron transfer between neighboring D and A molecules over a certain intrapair distance with probability g0. The 1 ðDþ ::A Þ pairs can then dissociate with the rate constant k1 escaping the Coulombic attraction of the parent countercharges to create the Dþ and A ions (which leads to the photoconduction) or geminately recombinate with the rate constant k1 forming fluorescent exciplex states, 1 ðDAÞ : Alternatively, the process of direct, i.e. without involvement of

+

D +A 1

D

*

+A 0

k -1 1

η0

*

( D ..A 0)

1

1 -η 0 1

+hν

-

(D A )

*

+

-

(D ..A ) k1

exciplex FLU O R E S C E N C E

D0

D0 + A0

Fig. 3. The diagram of energy levels and photophysical processes in the model of charge photogeneration in electron donor–electron acceptor m-MTDATA:BCP system.

the intermediate e–h pairs, creation of 1 ðDAÞ exciplexes from the encounter complex, ð1 D ::A0 Þ, is possible with the probability 1  g0 . Thus, with the 1 ðDþ ::A Þ pairs involved the indirect process of exciplex formation should be sensitive to an external electric field, that is a significant electromodulation of exciplex photoluminescence should be observed in an experiment. Since the analysis of relevant FL and ABS spectra from Fig. 2b imply the efficient non-radiative transfer of electronic excitation energy from the molecules of BCP to those of m-MTDATA, in the investigated blend independent of excitation energy range m-MTDATA should be regarded as the electron donor whereas BCP – the electron acceptor. 3.2. Electromodulation of photoluminescence First, we shall consider the spectrally integrated (global) EML signals depicted in Fig. 2b, i.e. the (2x)EML defined by Formula (1) measured at emission wavelength kem = 525 nm corresponding to the maximum of exciplex florescence (squares) and the rms value of the second-order Fourier component of the sample response to modulating field, I2x, recorded at excitation wavelength kexc = 313 nm (circles). These data have been obtained for a 120-nm thick film of m-MTDATA:BCP system at electric field strength FRMS = 5  105 V/cm. As seen from the figure, the (2x)EML signal is positive for all values of kexc < 450 nm which means that the FL is reduced by an external electric field. In particular, the FL quenching is approximately constant for kexc < 360 nm that is for energy of photons within the first absorption band of m-MTDATA (electron donor) together with maxima at 321 and 346 nm ascribed to the presence of intramolecular Frenkel excitons. For longer wavelengths, kexc > 360 nm, the FL quenching monotonically diminishes up to kexc ffi 420 nm. As for the I2x signal measured at excitation wavelength kexc = 313 nm and (circles in Fig. 2b), the FL intensity is reduced by electric field at all emission wavelengths and the EML spectrum resembles roughly the curve of the exciplex fluorescence. Consider now the electric field characteristic of the global EML signal as measured for a 120 nm-thick m-MTDATA:BCP film with excitation wavelengths kexc = 365 nm (circles in Fig. 4a) and kexc = 313 nm (squares in Fig. 5a), corresponding to the photon energy from spectral range of main low-energy absorption band of the electron donor m-MTDATA. For the sake of comparison in these figures the electroabsorption signals, (2x)EA versus FRMS, recorded at the wavelength of absorbed light equal to the kexc in the EML measurements representing changes in the number of absorbed photons due to the Stark effect (see Ref. [23]) is shown. As seen, the (2x)EA signals, marked in the Figs. 4a and 5a by

367

D. Pelczarski et al. / Organic Electronics 25 (2015) 362–376

λexc = 365 nm

-1

10

Onsager model

-2

10

EML

r0/rc = 0.117

EA

-3

10

Poole-Frenkel model

-4

10

slope 2.0

-5

10

4

10

5

10

6

10

FRMS [ V/cm] 0

10

STNH model

(2ω)EML

-1

10

a/rc = 0.05

-2

10

r0/rc = 0.117

-3

κ rc/D = 0.1

10

κ rc/D = 1

-4

10

κ rc/D = 3

-5

10

4

10

5

10

6

10

FRMS [ V/cm] (c)

0

10

STNH model

(a)

(2ω)EML

-1

10

a/rc = 0.05

-2

10

r0/rc = 0.117

-3

κ rc/D = 0.001

10

κ rc/D = 0.01

-4

10

κ rc/D = 0.1

-5

10

4

10

5

10

6

10

FRMS [ V/cm] Fig. 4. The double logarithmic plots displaying the dependence of (2x)EML signal on electric field at the excitation wavelength kexc ¼ 313 nm for a 120 nm-thick mMTDATA:BCP film. The abscissa of the plots stands for the rms value of the external electric field strength. In part (a) of the figure the squares stand for experimental data. The lines represent the best fit according to the Onsager (solid line, er ¼ 3:0, r0 =rc ¼ 0:125) and Poole–Frenkel (dashed line, APF ¼ 5  104 , e ¼ 3:0) model. The electroabsorption (EA) change in the transmitting light intensity with kexc ¼ 313 nm caused by the Stark effect is shown for comparison (diamonds). The thin dotted line in the figure have the slope of 2.0 which corresponds to the Stark effect. In part (b) of the figure the EML experimental data (squares taken from the part (a) of the figure) are compared with theoretical curves based on the STNH model assuming a=r c ¼ 0:05 and r 0 =r c ¼ 0:125 for three different values of capture velocity, j, as indicated in the figure.

(2ω)EML, (2ω)EA

(b)

by electron transfer between the electron donor and acceptor as indicated in the scheme of Fig. 3. Importantly, it is recognized in the literature that due to the high dipole moment of a radical ion pair (here an e–h pair) the electron transfer rate (so called Marcus electron transfer rate constant) is a function of the free Gibbs energy difference, DG, of the charge separation process which in turn represents a second-order function of the external electric field strength due to the Stark effect. Therefore, the applied field F may influence the value of DG and hence the probability (g0) of process in which the e–h pairs are generated giving a certain contribution to the electric field effect on FL quenching (compare scheme in Fig. 3). To summarize, the strong electromodulation of FL observed in the m-MTDATA:BCP system should be assigned to the Stark effect on the DG as well as to the electric-field-assisted dissociation of geminate e–h pairs. Extensive theoretical studies of the DG(F) contribution to the FL quenching have been published in the Ref. [39] for the case of electron donor and acceptor randomly distributed within a rigid sample and in the Ref. [40] for D and A separated at a fixed distance. As indicated in the Ref. [39], for the DG = 1 eV which is the case of the m-MTDATA:BCP system (cf. the LUMO levels of m-MTDATA and BCP from Fig. 1) at highest F  106 V/cm used in our measurements one should expect the maximal FL quenching of approx. 2%. This value seems to be not significant as compared to the maximum quenching of fluorescence of about 20% at F  106 V/cm recorded in our experiment and, therefore, in further considerations we shall omit the influence of electric field strength on the DG.

0

10

λexc = 313 nm

Onsager model r0/rc = 0.125

-1

10

-2

10

EML EA

-3

10

Poole-Frenkel model

-4

10

slope 2.0

-5

10

4

10

5

10

6

10

FRMS [ V/cm]

(b)

0

10

STNH model

-1

(2ω)EML

(2ω)EML, (2ω)EA

(a) 100

10

a/rc = 0.05

-2

r0/rc = 0.125

10

κ rc/D = 0.01

-3

10

κ rc/D = 0.1

-4

κ rc/D = 1

10

-5

10 diamonds and dotted lines, are the second-order functions of applied electric field with the magnitude at least two-orders lower than that of EML signals which means that the EA contribution to the electromodulation of fluorescence can be safely neglected in further considerations. In both cases the plots of (2x)EML reveal significantly increasing dependencies of the FL quenching on FRMS which tend to saturate and reach the value of about 80% at maximal field of 2  106 V/cm used in experiment. Next, the (2x)EML plots do rather depart form a Stark-type second-order functions. The strong electromodulation of FL enables us to conclude that the exciplexes are formed rather indirectly in the investigated m-MTDATA:BCP system, that is via 1 ðDþ ::A Þ e–h pairs produced

4

10

5

10

6

10

FRMS [ V/cm] Fig. 5. The double logarithmic plots displaying the dependence of (2x)EML signal on electric field at the excitation wavelength kexc ¼ 365 nm for a 120 nm-thick mMTDATA:BCP film. In part (a) of the figure the circles stand for experimental data. The lines represent the best fit according to the Onsager (the solid line, er ¼ 3:0, r0 =r c ¼ 0:117) and Poole–Frenkel (dashed line, APF ¼ 105 , e ¼ 4:0) model. The electroabsorption (EA) change in the transmitting light intensity with kexc ¼ 365 nm caused by the Stark effect is shown for comparison (diamonds). In part (b) and (c) of the figure the EML experimental data (circles taken from the part (a) of the figure) are compared with theoretical curves based on the STNH model with final recombination parameter, r 0 =r c ¼ 0:117, and for various values of capture velocity, j, as indicated in the figure.

368

D. Pelczarski et al. / Organic Electronics 25 (2015) 362–376

Now we shall recall the scheme of Fig. 3 and assume that the electron transfer between neighbouring molecules to create the geminate e–h pairs, 1 ðDþ ::A Þ, from the excited encounter complexes, ð1 D ::A0 Þ, proceeds over a certain intrapair distance, r0, and that the relevant probability, g0, is independent of the electric field strength. We also assume that the 1 ðDþ ::A Þ pairs do start subsequently a Brownian walk under the influence of combined Coulomb and external electric field until they dissociate escaping the Coulombic attraction of the parent countercharges (to create the free Dþ and A charge carriers) with escape probability increasing with the applied field, XðFÞ and that the probability of their geminate recombination (to form the fluorescent exciplex states) is as high as 1  XðFÞ. We shall note that the involvement of intermediate 1 ðDþ ::A Þ pairs in the creation of exciplexes and the electric field-induced FL quenching has been the subject of former considerations as a modification of the scheme originally proposed in Refs. [33,34]. In particular, in the Ref. [41] the reduction of the exciplex fluorescence of methylene-linked D–A systems (carbazole and terephthalic acid methyl ester doped in poly(methyl methacrylate) polymer films) were rationalized in terms of the mechanism in which the contact complexes, created by relaxation of photoexcited electron donors and the electron acceptors, via the electron transfer do create the radical ion pairs. These pairs, in turn, do produce the fluorescent exciplex states. The population of exciplexes may, however, be reduced by the electric field-assisted dissociation of the radical ion pairs. We shall also remark that the presented approach corresponds to the commonly accepted two-step dissociation process of molecular exciton states Refs. [1,42,43]. Moreover, the involvement of the 1 ðDþ ::A Þ pairs in the creation of exciplexes seems to be supported by results of time-resolved measurements of FL quenching for solid films of similar D:A (m-MTDATA:BPT) blends reported in Ref. [32] which will be discussed further in this Section. Since, according to the scheme from Fig. 3, the FL intensity, I, results here from competitive processes of dissociation of 1 ðDþ ::A Þ pairs and their geminate recombination into emissive exciplex states, the relevant dependence on electric field strength should take the form of

IðFÞ ¼

kf ½1  g0 XðFÞI kf þ kn

ð6Þ

where kf and kn stand for the rate constants of radiative and non-radiative decay pathways of exciplexes, I* – for the production rate of encounter complexes, ð1 D ::A0 Þ. The electric field-assisted dissociation process of e–h pairs may in general be rationalized in terms of various theoretical models, with commonly used Poole–Frenkel (PF) [44], three-dimensional (3-D) Onsager [45] as well as Sano–Tachiya–Noolandi–Hong (STNH) [46,47] ones. In what follows we shall examine the applicability of these approaches to results of our experiments by fitting the calculated values of (2x)EML signals according to the Formula (1) with the experimental data. As for the PF model, the dissociation of geminate e–h pairs occurs via a single thermally-activated jump of carriers over the Coulombic barrier being lowered by the external electric field in the down-field direction. At high electric fields the 1-D formalism is usually applied and the dissociation process is characterized by the first-order kinetic rate constant, keh ¼ k0 exp½bPF F 1=2 , where k0 is the zero-field 1

1=2

rate constant and bPF ¼ ð2kB TÞ ½e3 =ðpe0 er Þ – the PF coefficient with the (kB stands for the Boltzmann constant, T – the absolute temperature, e – the elementary charge, e0 – the vacuum permittivity, er – the dielectric constant of an organic solid). The results of fitting procedure using two values of the constant APF which determines the branching ratio of recombination and generation

channels at F = 0 are plotted by a dashed line in Fig. 4a for APF ¼ 1  105 with er = 4.0 and in Fig. 5a for APF ¼ 5  104 with er = 3.0 (for detailed description of the PF model parameters see Ref. [28]). As seen, both curves reveal that the PF approach is rather unsuitable for description of the e–h separation process in the whole range of applied electric fields. Accordingly, the carrier one-step jump (i.e. without the diffusive motion) over the potential barrier is rather impossible in narrow-band organic solids due to the magnitude of the Onsager radius (Coulombic capture radius), rC ¼ e2 =4pe0 er kB T, which is equal to many intermolecular distances (rC = 187 Å for er = 3.0 and T = 298 K), as well as due to strong spatial localization of molecular electronic wavefunctions and such a failure has been frequently reported in the literature (see e.g. Refs. [48,49]). As far as the Onsager and STNH models are concerned, the escape probability, X(F), was calculated by solving the Smoluchowski equation regarding the electric field-enhanced 3D diffusive motion of an ion pair in a continuous medium. Importantly, in the STNH approach the recombination of charge carriers occurs on a sphere of finite radius, a, with a finite velocity, j, whereas in the Onsager formalism the a = 0 is assumed which makes the required mathematical manipulations much less tedious and, therefore, the latter model is often chosen to rationalize the dissociation of e–h pairs in organic solids [42,43]. Hence, the averaged escape probability is calculated from the formula

XðFÞ ¼

Z

Xðr; hÞgðr; hÞds

ð7Þ

with Xðr; hÞ given respectively by Onsager in Ref. [45] and by Sano and Tachiya in Ref. [46] as well as, independently, by Noolandi and Hong in Ref. [47] (r stands for the initial intrapair separation, h – the polar angle, ds – the volume element). The function gðr; hÞ represents the initial distribution of pair separation assumed here to be isotropic and described by a Dirac delta,

gðr; hÞ ¼

1 dðr  r 0 Þ; 4pr 2

ð8Þ

where r0 denotes the mean intrapair distance. The available relevant analytical formulas for the X(F) together with numerical procedures used in the calculations are collected in our previous paper [50]. In the fitting procedures, the ratio r 0 =rC and the g0 (here the primary quantum yield of e–h pairs) were adjusted assuming the dielectric constant of the solid, er = 3. For the Onsager approach (plotted with solid lines) the curve from Fig. 4a was obtained with r 0 =r C ¼ 0:117 (r0 = 21.9 Å) whereas that from Fig. 5a with r0 =r C ¼ 0:125 (r0 = 23.4 Å) with approximately the same value of the primary quantum yield, g0 = 0.9. As seen, in both cases the Onsager-based functions follow well the measured (2x)EML signal but the reproductions distinctly fail at electric fields, FRMS, exceeding 106 V/cm. In Figs. 4b, c and 5b, together with corresponding experimental data taken from Figs. 4a and 5a (points), we present the results of fitting calculations performed with the use of the STNH model (lines in Figs. 4b, c and 5b). The calculations were performed for various values of the sphere radius, a=r C ; and capture velocity parameter, jr C =D (with D = De + Dh, the relative carrier diffusion coefficient), describing the final recombination step and assuming the initial e–h radii (r0 =r C ) as equal to those obtained according to the Onsager model. In general, for each value of the a=r C ratio, in the low-electric field regions the SNTH-based functions do fit well the (2x)EML experimental data independently of the value of jr C =D and come close to the Onsager-based curves. In the intermediate and high-field regions, however, in most cases the upward-deviation of the curves from the experimental data can be observed, somehow diminishing with increasing values of the jr C =D ratio. Next, the lower radii of the final recombination sphere (a), the lower values of the final recombination speed parameter (jr C =D) are required to obtain the satisfactory

369

D. Pelczarski et al. / Organic Electronics 25 (2015) 362–376

hν

quartz/Al1/m-MTDATA:BCP(1:1)/Al2

1.5

Al1(+)

1.6 1.2

2

ð9Þ

strength, F ¼ 4  104 V/cm, with photon flux / ¼ 1013 cm2 s1. For comparison, as solid lines we plot the spectra of the linear absorption coefficient, a(k) (k is the wavelength of illuminating light) defined by the Bouguer–Lambert formula, / ¼ /0 expðadÞ, with d being layer thickness, /0 – the incident photon flux. As seen, the behavior of the photocurrent does depend on the bias of the illuminated substrate electrode. In particular, for Al1(+) the photocurrent j+ (squares) decreases monotonically with the k whereas with the Al1 negatively polarized the corresponding j density curve (circles) of magnitude significantly lower than that of j+ forms a pronounced maximum. To rationalize qualitatively these features of the

-1

Iðt; FÞ ¼ Að0; FÞ exp½t=sðFÞ;

At first one should note that the single components of the investigated system are generally considered as conductors of one-sign carriers being the holes in the case of m-MTDATA and electrons in the case of BCP. In a layer of these two compounds mixed the photocurrent will then arise from the motion of both types of carriers towards the oppositely charged electrodes of a sandwich structure. The photoconductivity measurements of the system have been performed with the layer illuminated through the substrate electrode, that is in the experimental configuration hm ? quartz/ Al1/m-MTDATA:BCP/Al2. For the purpose of this paper we shall denote the registered photocurrent densities as the j+ and j for positive, Al1(+), and negative, Al1(), bias of the illuminated contact. In Fig. 6 we present exemplary action spectra of photocurrent densities j+ and j for a 250 nm-thick layer obtained for electric field

5

where t denotes time, and s – the lifetime of emissive exciplexes. As seen from the Formula (9), the electric field may generally reduce both the initial amplitude, A, of the FL intensity as well as the lifetime, s. Importantly, the time-resolved FL measurements should be carried out at low intensities of excitation light which ensures the lifetime of excited states not to be significantly reduced by the charge of field-modulated concentration in the sample. Nevertheless, according to the experimental outcomes given in [32], the amplitude quenching significantly exceeds the lifetime quenching in the m-MTDATA:BPT systems which suggests the dissociation of the emissive exciplex precursors (the e–h pairs) to dominate the field-assisted FL reduction. The dissociation process was in turn rationalized in terms of the hopping separation (HS)

1.0

Al1(-) ×0.56

0.5 ABS of BCP

0.0

300

ABS of m-MTDATA

350

0.8

-9

BCP the electron mobility le ¼ 6  106 cm2/Vs at 7  105 V/cm is given and we presume that the lh as well as the le in the m-MTDATA:BCP blend are somehow lowered due to disordered structure of evaporated films but do still fulfill the inequality lh > le , as in the case of single components of the system. Thus, taking into account the dominating mobility of holes in m-MTDATA one obtains the value of diffusion coefficient D  7  107 cm2/s which in turn using the STNH fitting parameter, jrC =D ¼ 1 (from Figs. 4b and 5b), enables us to calculate the value of the final 1 ðDþ ::A Þ pair recombination velocity, j  0.4 cm/s, consistent with j  0.4 as determined in the same manner in Ref. [50] for vacuum-evaporated layers of Alq3. It is interesting then to calculate the value of pair recombination time, s = a/j  225 ns (with the final recombination sphere radius, a = 9 Å) which remains close to the photoluminescence decay time in m-MTDATA:BPT blends from Ref. [32] being as high as 200 ns. Finally we note that the recombination sphere radii, a, resulting from our calculations remain close to the average intermolecular distance, rav = 8.5 Å, determined for the m-MTDATA:BCP solid. Summing up, one should assume that in the m-MTDATA:BCP system the final 1 ðDþ ::A Þ pair recombination process proceeds with a finite speed at the recombination sphere of radius as high as one intermolecular distance. It is recognized in the literature (see e.g. Refs. [53,54]) that the electric field-assisted dissociation of excited states in electron D–A organic solid films may, in general, occur from the emissive states in themselves or from their precursor ones. When the influence of the external electric field on time-resolved FL of these systems is examined as reported in Ref. [32] for the case of m-MTDATA:BPT blends, the resulting luminescence decay curves can be described by the exponential function,

3.3. Photocurrents

j [10 A/cm ]

m-MTDATA the hole mobility lh ¼ 2:7  105 cm2/Vs at an electric field of 105 V/cm is reported whereas, according to the Ref. [52], for

model in which the e–h pairs weakly bounded by the Coulombic forces dissociate due to the hopping of the more mobile holes between the TPA units of m-MTDATA with the rate constant, kdiss  exp(eqF/kT) where q stands for the carrier hopping distance. According to HS model used in Ref. [32], the Coulombic interaction between components of a geminate pair is neglected which is difficult to justify for e–h pairs with r0 being of several times smaller than rC. Since the measurements of FL quenching in our experiment were carried out at rather low intensities of excitation light (with photon flux of approx. 1013 cm2 s1), the modulation of excited states lifetime by the charge in the sample could be neglected (note that the carrier transporting properties of both m-MTDATA and BCP are rather good). Therefore, the observed FL quenching could safely be related to the population of weakly bonded fraction of the 1 ðDþ ::A Þ pairs predominantly involved in producing exciplexes, being reduced due to their field-assisted dissociation, as mentioned earlier in this Section.

α [10 cm ]

reproduction of experimental points. As seen from the Fig. 4b (solid line), the fitting procedure made it possible to determine the value of the jr C =D being as high as 3 in the case of the STNH approach at a=r C ¼ 0:05 (a = 9.35 Å) at kexc = 365 nm, or jr C =D ¼ 1 at kexc = 313 nm (the solid line in Fig. 5b). On the contrary, when the radius of recombination sphere is reduced, i.e. for a=r C ¼ 0:035 (a = 6.5 Å) which actually brings the STNH approximation closer to the Onsager one the best fitting of experimental data has been obtained for jr C =D ¼ 0:1 and 0.01 that is for the lower carrier recombination velocities, j (cf. dashed and solid lines in Fig. 4c). The values of j could be, in principle, determined from these results of fitting procedure, however, one has to know the value of carrier mobilities, l, to calculate the diffusion coefficient assuming the Einstein relation, D = lkBT/e. Tough, the carrier mobilities in solid organic mixtures do depend, among others, on morphological features as well as, importantly, on post-treatment conditions of a film and their reliable values are not known for our m-MTDATA:BCP system. Nevertheless, in the Ref. [51] for amorphous glass of

0.4

400

0.0

450

λ [nm] Fig. 6. The action spectra of photocurrent density (j) for positively (squares) and negatively (circles) biased illuminated electrode (d ¼ 250 nm, F ¼ 4  104 V=cm , / ¼ 1013 ph/(cm2 s)). The absorption spectrum of a 120 nm-thick BCP and 150 nmthick m-MTDATA films are added for comparison.

370

D. Pelczarski et al. / Organic Electronics 25 (2015) 362–376

photocurrent action spectra we shall recall that the photocurrents are assumed to be induced by simultaneous movement of both holes (of mobility lh ) and electrons (of mobility le ) in the m-MTDATA:BCP system with m-MTDATA and BCP being respectively the hole- and electron-transporting component and that lh > le (cf. Section 3.2). Consider now the range of wavelengths, k, between ca. 310 and 380 nm (including the maximum of the j photocurrent action spectrum) in which the value of the linear absorption coefficient of BCP exceeds considerably that of m-MTDATA. This implies that the energy of exciting photons is absorbed predominantly by the molecules of BCP and subsequently transferred to the molecules of m-MTDATA in the non-radiative Förster process (cf. the Section 3.1). The charge carriers are then produced in a dissociation process of the electron–hole pairs (cf. scheme in Fig. 3) at the interfaces between the molecules of both system components as enhanced by high difference (ca. 1 eV) between the electron affinity of BCP and that of m-MTDATA. It should be noted here that in considered range of excitation wavelengths the 250 nm-thick layer is only partially penetrated by incident light; for instance, at k = 350 nm the penetration depth of light,  1  113 nm (a  denotes the mean value of the linear absorpLA ¼ a tion coefficients of the mixture components). Thus, the positively biased illuminated substrate electrode, Al1(+), does attract and neutralize the free electrons produced in the illuminated sublayer of the sample nearby whereas the holes moving towards the counter-electrode do contribute to the j+ photocurrent. However, the negatively biased illuminated substrate electrode, Al1(), does on the contrary attract and neutralize the holes and, therefore, the j photocurrent is induced by the less-mobile electrons (lh > le ). Hence, the j+ > j relation is observed in the given spectral range. Consequently, for uniformly illuminated layer, i.e. at low values of the m-MTDATA and BCP linear absorption coefficients the magnitudes of both photocurrents should be the same whether the sample is illuminated through the positively or negatively biased substrate electrode which is the case in Fig. 6 for wavelengths exceeding ca. 380 nm. Accordingly, one may conclude that the photocurrents are primarily generated in the bulk of the layer. That the photocurrents are bulk-generated also implies from the dependencies of j+ (squares) and j (circles) on the photon flux, /, depicted in Fig. 7 which were recorded for 250 nm-thick sample excited with light strongly absorbed by both materials (k = 313 nm). In particular, at strong electric field strength, F = 4  105 V/cm, the relevant characteristics are practically linear in full range of used light intensities (1012–1015 ph/cm2 s) which is commonly related to the geminate (monomolecular) recombination of electrons and holes. With decreasing electric field (F = 4  104 V/cm in Fig. 7) the j+(/) and j(/) functions become

hν

quartz/Al1/m-MTDATA:BCP(1:1)/Al2

Al1(+)

-6

10

-7

2

j [A/cm ]

slop

Al1(-)

10

slo

-8

10

pe

=0

e slop

-9

10

.9

e=

0.9

-1.0

e=0 slop

= 0.

7

12

10

13

.6

(2) 5

F [10 V/cm] (1) 4 (2) 0.4

-10

10

(1)

14

10 10 2 φ [1/(cm s)]

15

10

Fig. 7. The log–log plots of the photocurrent density versus excitation light intensity for positively (squares) and negatively (circles) biased illuminated substrate electrodes at the excitation wavelength kexc ¼ 313 nm for a 250 nm-thick m-MTDATA:BCP film. The solid lines represent linear fits.

sublinear, of the slope = 0.7 or 0.6 depending on the photon flux, which in turn reveals the bimolecular recombination of holes with less-mobile electrons present within the sample. Accordingly, the decreasing slope of the photocurrent versus photon flux dependencies for rising sample electric field strength have been reported in Ref. [55] for m-MTDATA:BPT blends. Now, let us analyze the process of 1 ðDþ ::A Þ e–h pair dissociation leading to the carrier photogeneration in the bulk of the m-MTDATA:BCP films as influenced by the strength of electric field, F, applied to the sandwich structure. We shall recall here that, according to the scheme from Fig. 3, the dissociation of 1 ðDþ ::A Þ pairs is alternative to the geminate recombination forming fluorescent exciplex states, 1 ðDAÞ ; the latter process being investigated in our electromodulation-of-photoluminescence experiment from the previous Section. As for the analysis of the bulk 1 ðDþ ::A Þ pair dissociation, we shall remain in terms of the 3-D Onsager as well as the STNH approaches by applying the corresponding analytical formulas to fit the experimental data of the charge photogeneration efficiencies gþ ðFÞ and g ðFÞ derived with the use of Formula (2) from the photocurrent densities j+(F) and j(F) recorded at the experimental configuration hm ? quartz/Al1/m-MTDATA:BCP/Al2 at positively and negatively biased illuminated Al1 electrode, respectively. In the fitting procedures analogous to those performed for our electromodulation-of-photoluminescence results from Section 3.2 (cf. the Formulae (7), (8) and following remarks), the mean intrapair distance/Onsager radius ratio, r 0 =r C , and the primary quantum yield of 1 ðDþ ::A Þ pairs, g0, were adjusted at the dielectric constant of the solid, er = 3. Another fitting parameters in the case of the STNH approach were the values of the recombination sphere radius, a=rC ; and the capture velocity parameter, jrC =D. Results of the fitting calculations are depicted in Fig. 8a for g+(F) (squares) and g(F) (circles) experimental dependencies determined for d = 250 nm-thick sample uniformly illuminated by exciting light of kexc = 390 nm (with photon flux, / ¼ 1013 cm2 s1) deeply penetrating the organic layer to accomplish the desired carrier bulk-generation regime. Firstly, independent of the theoretical approach used, for a low electric-field range the upward-deviation of the curves from the experimental data are observed in Fig. 8a which is the common issue in such a type of plots since in this range the bimolecular recombination is involved and, in addition, the actual value of F in the sample, as modified by a spatial charge trapped within the layer, does not match that determined by the V/d ratio (with V the sample voltage). Consider now the STNH-based curves obtained for the sphere radius a=rC ¼ 0:045, that is for a = 8.4 Å. The relevant fits were performed using g0 = 0.8, r0 =rC ¼ 0:135 (r0 = 25.3 Å) but for two values of the capture velocity parameter: jrC =D ¼ 10 (dotted line) and jr C =D ¼ 1 marked in Fig. 8a by a dashed line. As seen, for sufficiently high electric field strengths exceeding approx. 105 V/cm both curves do reproduce well the experimental g+(F) and g(F) outcomes. Nevertheless, in the range of higher F, exceeding 7  8  105 V/cm, the approximation by the jr C =D ¼ 1 curve tends to upward-deviate slightly from the experimental results. Now, it is worth to compare the STNH- curves with the result of fitting procedure using the 3-D Onsager model as performed for the same values of r 0 =r C ¼ 0:135 and g0 = 0.8 which is shown by a solid line in Fig. 8a. As seen, the Onsager curve remains rather close to the measured g(F) dependencies in the whole range of considerable field strength and does not deviate significantly from both STNH fits as well. This seems to confirm the validity of the values of the r 0 =r C – and g0 – parameters used in the STNH approximations. It should be noted here that the Onsager model, although often used to describe the dissociation process of e–h pairs in organic solids due to less effort-consuming calculations required, does not enable to conclude about the value of the e–h pair

D. Pelczarski et al. / Organic Electronics 25 (2015) 362–376

(a)

hν

quartz/Al1/m-MTDATA:BCP(1:1)/Al2

0

10

STNH (a/rc = 0.045) -1

Al1(-)

r0/rC = 0.135 ; κrC/D = 10

10

Al1(+)

η

r0/rC = 0.135 ; κrC/D = 1 -2

10

Onsager

-3

10

r0/rC = 0.135

-4

10

4

5

10

10

6

10

F [V/cm]

(b)

0

10

= -1

η

10

-2

Onsager sh o rt + lo n g

(r0/rC)short = 0.117 (r0/rC)long = 0.160

10

Onsager

-3

10

(r0/rC) = 0.135

-4

10

4

10

5

10

6

10

F [V/cm] Fig. 8. The electric field dependence of the charge photogeneration efficiency (g). The photocurrent was measured with kexc ¼ 390 nm light incident on the substrate electrode penetrating deeply the bulk of a 250 nm-thick m-MTDATA:BCP layer. The experimental data are displayed by squares and circles for a positively and negatively biased illuminated electrode, respectively. The solid lines are calculated according to the 3D-Onsager model with g0 ¼ 0:8 and for a single value of r 0 =rc as indicated in the figures. In part (a) the broken lines represent the best fit according to the STNH model with parameter values as indicated in the figure. In part (b) of the figure the dotted line represents the fit according to the 3D-Onsager model assuming a combination of short and long radius-pairs involved in photogeneration process (see text for details).

recombination sphere radius, a, since the a = 0 is assumed in this formalism (note that the a = 0 does not require the recombination velocity j = 1 as indicated in our previous paper [50]). Importantly, the application of the STNH model to the recorded g-efficiencies makes it possible to compare the parameters of the e–h pair dissociation accompanying the charge photogeneration process with those of the e–h pairs involved in the pair geminate recombination leading to luminescent exciplexes which are known from our EML experiment (see Section 3.2). Here, we shall recall the best results of the STNH-fittings from Fig. 4b and c obtained for the recombination sphere radius a=r C ¼ 0:05 (a = 9.4 Å) and a=rC ¼ 0:035 (a = 6.6 Å), somehow different but still favourably close to the average intermolecular distance, rav = 8.5 Å of the m-MTDATA:BCP solid mixture. From Fig. 4b and c it is also seen that the lower limit of the final recombination speed parameter remains in the range of 0:1 < jr C =D < 3 whereas the relevant value of the r0 =rC ratio, being as high as 0.117, implies that the r0 = 22 Å is the mean intrapair distance of 1 ðDþ ::A Þ pairs which are supposed to undergo the geminate recombination process do form the fluorescent exciplex states, 1 ðDAÞ (cf. Fig. 3). Analogically, from Fig. 8 one may observe that at the sphere radius a = 8.4 Å there is 1 < jr C =D < 10 and r 0 =r C ¼ 0:135, the latter value yielding r0 = 25.3 Å of the 1 ðDþ ::A Þ pairs which are supposed to dissociate producing the photo-generated charge carriers. Therefore, the presence of pairs of different values of the initial e–h radius has been detected by two experimental techniques in our m-MTDATA:BCP system. To conclude about the contribution of different r0-radii e–h pairs to the photocurrent in a

371

m-MTDATA:BCP system we have performed another fitting procedure in which the experimental results from Fig. 8a have been reproduced by an Onsager-based curve representing a two-component charge photogeneration efficiencies, g ¼ gshort þ glong , with gshort and glong induced by pairs of short and long initial e–h radii, respectively. The results of such calculations are presented by a dotted line in Fig. 8b for the case of the r0 =rC ¼ 0:117 (r0 = 22 Å) as taken for reproduction of the (2x)EML signals from Fig. 4a, together with the r0 =rC ¼ 0:160 (r0 = 30 Å). For the sake of comparison, the single-g Onsager curve for r0 =rC ¼ 0:135 form Fig. 8a has been shown by a solid line. As seen, both the double-g and single-g Onsager functions do reproduce satisfactory the experimental results in the range of middle and high electric fields.Interestingly, the two-component curve tends to perform better for initial low values of F as compared to the single-component one. One may then presume that the photocurrents measured in our system are induced by presence of the long-radius (r0 = 30 Å) and short-radius (r0 = 22 Å) 1 ðDþ ::A Þ pairs with low and high exchange energy, respectively (we suppose that these pairs are created from corresponding sorts of excited encounter-complexes, ð1 D ::A0 Þ, from Fig. 3). Thus, it is also expected that the contribution of the long- and short-radius 1 ðDþ ::A Þ pairs to the magnetic field effects on photocurrents and photoluminescence should be different which will be discussed in the following Sections. Finally, it is worth noting that the electric-field dependent e–h pair dissociation as inducing the considerable photoconduction and exciplex photoluminescence quenching in a similar electron donor–acceptor m-MTDATA:BPT system has been investigated in Ref. [55]. However, in the hopping separation (HS) model developed in that paper the initial electron– hole separation (r0) is presumed to be unrealistically large to neglect practically the Coulombic attractive forces. 3.4. Magnetic field effects on photocurrents The plots of the curves representing the magnetic field effect (the MPC signal) on photocurrent recorded as a function of the field of strength B applied to the sample at the experimental configuration hm ? quartz/Al1/m-MTDATA:BCP/Al2 have been depicted in Fig. 9a. The results are obtained for d = 250 nm thick sample illuminated by exciting light of kexc = 313 nm (with photon flux, / ¼ 2  1013 cm2 s1). Note that for used value of kexc the sample  1  74 nm. This means that the light penetration depth, LA ¼ a organic layer is only partially penetrated by incident light which ensures the carrier bulk-generation regime during the measurements. In Fig. 9a the MPC signal data points calculated from the Formula (3) as related to the j+ (measured at positively biased Al1 electrode) and j (measured at negatively biased Al1 electrode) photocurrents are presented by squares and circles, respectively. One may observe in the figure that the effect on both photocurrent densities is positive and the initial rapid rise of the curves detected at field strengths of several mT with increasing B becomes more and more moderate whereas for strong magnetic fields (i.e. for B > 100 mT) the MPC signals distinctly tend to level-off within the range of experimental error, at the magnitude of approximately 6%. To conclude about the mechanism involved in the origin of MPC signals the data points have been fitted with a double-Lorentzian and a single non-Lorentzian function commonly used in such experiments [56], having the form of

MPC ¼ ALFE B2 =ðB2 þ B2LFE Þ þ AHFE B2 =ðB2 þ B2HFE Þ

ð10Þ

and

MPC ¼ ALFE B2 =ðB þ BLFE Þ2 ;

ð11Þ

372

D. Pelczarski et al. / Organic Electronics 25 (2015) 362–376

(a) 8

MPC [%]

j+

6

HFC

4

FSM

2

100

300

(b)

-

D + A (free carriers)

1

+

-

(D ..A ) long

3

0

2

+

4 2 0

k -3 k ST

2

4

5

6

8

10

6

500

k’-1

4

(d)

B [mT] +

2

ph/(cm s) B = 50 mT

F [10 V/cm]

2 x Lorentz 1 x non Lorentz

0

13

φ = 2×10

6

0

MPC [%]

MPC [%]

(c)

j-

B = 50 mT 5

F = 1.6×10 V/cm 12

10

2

φ [1/(cm s)]

10

13

(e)

-

(D ..A ) long

6

k

nr 3

MPC [%]

nr

k1

D 0 + A 0 (ground state)

4 2 0 300

B = 50 mT 4

F = 1.6×10 V/cm 350

λexc [nm]

400

Fig. 9. The MPC effect for j+ (squares) and j (circles): (a) versus the magnetic field strength (/ ¼ 2  1013 ph/(cm2 s), F ¼ 4  104 V/cm), (c) versus the electric field strength, (d) versus the excitation light intensity and (e) versus the excitation light wavelength. The sample thickness d = 250 nm, the wavelength of excitation light kexc ¼ 313 nm. In part (a) of the figure the lines represent the best fit according to the double-Lorentzian (dashed line) and single non-Lorentzian (solid line) function. The two Lorentzian MPC components, HFC and FSM are depicted by dotted lines. The schematic diagram for the EHP mechanism of MPC effect is displayed in part (b) of the figure.

respectively, with the relevant components representing the low-field (LFE) and high-field (HFE) effects. In particular, the ALFE and AHFE denote the MPC signal magnitudes for B ? 1, the BLFE and BHFE in the Lorentzian Formula (10) are the fitting parameters which determine the half width at half MPC signal maximum whereas the BLFE in the non-Lorentzian Formula (11) is the fitting parameter denoting the half width at quarter MPC signal maximum. The results of fitting procedures are presented in Fig. 9a using a dashed (the double-Lorentzian) and a solid (the single non-Lorentzian) line which, as seen, do reproduce well both the j+ and j data points in the initial B-range of several mT, with the double-Lorentzian function performing somehow better at higher values of B, up to B = 600 mT applied in the experiment. For the double-Lorentzian and the non-Lorentzian function we have obtained the same value of low-field fitting parameter, BLFE = 4 mT, whereas the high-field parameter of the Lorentzians is as high as BHFE = 50 mT. Note that these values are consistent with literature data since, as published in Ref. [9], one has BLFE = 5.5 mT for poly(phenylene-ethynylene) (PPE) and BLFE = 5.4 mT for polyfluorene (PFO) and aluminum quinolate complex (Alq3). Next, for comparison in our previous paper [57] for neat m-MTDATA layers we received BLFE = 5 mT and BHFE = 40 mT for double-Lorentzian function. In order to rationalize the effect on the low magnetic-field-scale of several mT in the case of m-MTDATA:BCP mixture we shall propose a scheme depicted in Fig. 9b which uses the mechanism of singlet–triplet electron–hole pair intersystem crossing (or intersystem conversion, ISC) embedded in the electron–hole pair conversion process. This model has primary been applied to explain the magnetic effects on photoconductivity in the layers of

poly-p-phenylenevinylene (PPV) [6], Alq3 [7], on electroluminescence in Alq3 films [8] and, later on, discussed widely in the literature concerning the origin of magnetic effects in organic solids [4,11,56]. Next, the ISC mechanism turned out to be appropriate to explain the origin of the MPC signals observed in neat films of m-MTDATA in our previous work [57]. Prior to the further considerations it has to be noted that, as generally recognized, the remarkable efficiency of the intersystem crossing requires the quasi-degeneracy of the singlet and triplet energy levels which implies a moderate value of electrostatic exchange energy of sufficiently long e–h pairs (importantly, the exchange energy decays exponentially with the e–h distance, [5]). For these pairs, in turn, in the ISC process the asymmetric hyperfine coupling (HFC) of the electron and hole spins with magnetic moments of surrounding nuclear environment is involved. Therefore, among the long-radius and short-radius geminate 1 ðDþ ::A Þ pairs originating from the relevant encounter complexes ð1 D ::A0 Þ (see Fig. 3) prevailingly the longer ones, 1 ðDþ ::A Þlong , are supposed to undergo the kST-rated process of intersystem conversion to produce the triplet pairs, 3 ðDþ ::A Þlong as depicted in Fig. 9b. Both singlet and tri0

plet long-radius pairs may dissociate respectively with k1 and k3 rate constants to the Dþ as well as the A ions hence contributing 0 to the photocurrent. Note that usually k1 k3 due to stronger coupling of the singlet pairs with ionic reaction products of electrons and holes as compared to the coupling of the triplet ones (see e.g. Refs. [4,58,59]) Alternatively, the 1 ðDþ ::A Þlong and 3

ðDþ ::A Þlong pairs may undergo the non-radiative relaxation pronr

nr

cesses rated with corresponding constants k1 and k3 to achieve the donor and acceptor ground states, D0 and A0 . Essential for

373

D. Pelczarski et al. / Organic Electronics 25 (2015) 362–376

the rationalization of the magnetic field effect on the MPC signal in the several-mT range from Fig. 9a is that the photocurrents can arise predominantly from the dissociation of singlet pairs 0 (k1 k3 in Fig. 9b) and that the triplet energy levels become Zeeman-split under external magnetic field. Does the energy of Zeeman splitting overcome that of the HFC, the value of the kST rate constant is reduced with increasing magnetic field intensity which makes the intersystem crossing process less and less probable but nr nr leaves the activity of both the k1 and k3 channels of the pair non-radiative relaxation unaffected. As a result, the magnetic field of several-mT-hyperfine scale increases the population of 1 ðDþ ::A Þlong pairs and decreases that of 3 ðDþ ::A Þlong pairs which, in turn, enhances the population of the Dþ and A ions produced 0 by the k1 -dissociation process, hence contributing to the photocurrent. Next, the saturation of the MPC curves from Fig. 9a does occur when the B-induced energy difference between the 3 ðDþ ::A Þlong -pairs split components becomes much greater than the energy of the HFC. To interpret the MPC effects from Fig. 9a corresponding to the high magnetic field fitting constant, BHFE = 50 mT, we shall recall our further considerations from Ref. [57] to explain the origin of analogous effect in single m-MTDATA layers. In that paper the model of the fine structure (electron–electron) modulation (FSM) of triplet-free carrier interactions in the presence of an external magnetic field has been employed and in the corresponding mechanism, based on relevant reactions presented originally in Ref. [13], the concept of trions, developed in Ref. [14] to rationalize the effect of magnetoresistance, is essential for explanation of the MPC signals. Accordingly, the doublet and quartet trions are formed by interaction between spin-1/2 free charge carriers and spin-1 triplet exciplexes (presumably trapped in defect sites of an organic solid) populated in the m-MTDATA:BCP system. Note that the energy level of triplet exciplexes in this mixture is expected to be lower than that of singlet exciplexes, ES = 2.4 eV as well as of triplet states in single system components, ET = 2.5 eV (BCP, [60]), ET = 2.6 eV (m-MTDATA, [10]). Essential for the functionality of this model is that the trions operate actually as free-carrier-capturing centers hindering the carrier mobility. As a consequence, the overall process may be interpreted as a scattering of free carriers on triplet states. Since the recombination process of doublet trions is spin-allowed with the recombination of the quartet trions remaining spin-forbidden, the lifetime of the doublet trions is much shorter than that of quartet ones. Therefore, the quartet trions are more efficient in capturing free carriers and, hence, in reducing the photocurrent. Importantly, in the FSM-scale magnetic field of the several-tens-of-mT, the population of the quartet trions is reduced whereas the population of the doublet ones increases. This leads to the lower contribution of the quartet trions and, simultaneously, higher contribution of the doublet trions to the overall scattering process. One then should observe the increase in photoconductivity of an organic solid due to rising mobility of charge carriers which is the case from Fig. 9a. It should be noted here that the validity of the trion model is, however, a matter of controversy since the observations of B-rising mobility of charge carriers on the FSM (or comparable) magnetic field scale are still ambiguous. Indeed, while no influence of the B  200 mT on the hole and electron mobility in Alq3 films measured by the transient electroluminescence method was reported in Ref. [61], an increasing mobility of holes in TPD (N,N-diphenyl-N,N-bis(3-methylphe nyl)-[1,1-biphenyl]-4,4-diamine) layers after application of B  500 mT has been observed using the dark injection transient technique in Ref. [62]. With suggested mechanisms that are responsible for the MPC signals of Fig. 9a on the HFC and FSM scales of magnetic field, now it is worth determining their contribution to the total

magnitude of the effect. We shall define this contribution as the AHFE =ALFE ratio where the AHFE and ALFE denote respectively the double-Lorentzian function low-field and high-field component magnitudes for B ? 1 (cf. Formula (10)). Thus, taking the corresponding values obtained from the fitting procedure of the j+ and j data points we get 2:5%=4:1%  0:6 which means that the contribution of the FSM-component to the total magnitude of the MPC signals is rather significant. One may observe this in Fig. 9a where the plots of both components of the Function (10) are depicted by dotted lines. Consider now the Fig. 9c in which the magnetic field effects on the j+ (squares) and j (circles) photocurrents are presented as functions of electric field strength applied to the sample obtained at fixed magnetic field intensity, B = 50 mT. Although the magnitudes of the MPC signals are somehow different in predominant range of the electric field strengths (the differences are diminishing with increasing F), the shape of both MPC curves, as seen, are similar. In particular, from the initial MPC magnitudes both curves do reach their maxima of ca. 6% at F  105 V/cm which are followed by a monotonic decrease to reach the low common value less than 1% for the highest applied electric field of F  106 V/cm. To interpret this we shall recall the Ref. [7] where the analogous behavior of the magnetic field effect on photoconductivity in Alq3 is presented as induced by an electric-field-assisted increase in the mean e–h intrapair distance, r0. Accordingly (see Fig. 9c), in the initial range of F the increase in r0 reduces the electrostatic exchange energy of the 1 ðDþ ::A Þlong e–h pairs, hence the energy levels of the 1 ðDþ ::A Þlong and the 3 ðDþ ::A Þlong states do become successively quasi-degenerate. As a result, at certain kST(0) and fixed value of the rate constant kST(B), the singlet–triplet conversion process becomes more and more efficient which makes the MPC magnitude curves from Fig. 9c reaching their maxima. Note that in 0 corresponding electric field strength range the rate constant, k1 , is expected to be negligibly small as compared to those of other mechanisms from Fig. 10 which implies the relative inefficiency of the 1 ðDþ ::A Þlong pair dissociation process. For higher F, however, 0

the electric-field-induced dissociation with increasing k1 becomes more efficient and hence the lifetime of the 1 ðDþ ::A Þlong pairs becomes too short to accomplish spin-precession in a meaningful way. Interestingly, a number of similar-in-shape MC (magnetoconductance) curves as functions of the applied voltage are presented in Ref. [14] for the case of Alq3-, PFO- and PPV-based organic light emitting diode (OLED) structures. In those structures, however, the current is induced by electrical excitation of electrons and holes and the origin of the MPC-voltage characteristics is explained in terms of the trionic model where the occupation of traps is driven by external voltage. Finally, we shall note that no mutual exciton–exciton interactions occur in the layers of m-MTDATA:BCP blend at light intensities used in measurements as the MPC signals (measured on the set of several samples) do not depend on sample photon flux (see Fig. 9d). Next, the short and long initial radii of the e–h pairs seem to be independent of the excitation wavelength, k, since the MPC signals (Fig. 9e) as well as the EML signals (Fig. 2b) are roughly independent of k within the lowest-energy exciton spectral range. 3.5. Magnetic field effects on photovoltaic characteristics In order to investigate the photovoltaic (PV) characteristics a bulk-heterojunction device of the experimental configuration Al/LiF/m-MTDATA:BCP[1:1]/MoO3/ITO hm has been used (see the inset of Fig. 10a), with the thickness of the m-MTDATA:BCP active layer blend as high as approx. 100 nm. Note that the LiF and MoO3 layers are commonly used in such structures to improve

374

D. Pelczarski et al. / Organic Electronics 25 (2015) 362–376

(a)

of volume (bimolecular) recombination of carriers, especially, in the range of low resultant electric field. To conclude about the magnetic field effect on the device photocurrent, we plotted the measured effect magnitude, DISC/ISC versus the field strength, B, (points in Fig. 10a) as fitted by a double-Lorentzian Function (10) (solid line). Here, the obtained curves do resemble those of the MPC signals from Fig. 10a but reveal somehow lower level-off magnitude of ca. 4%, with the values of the low- and high-field fitting parameter of the Lorentzian, BLFE = 4 mT and BHFE = 50 mT, respectively. Therefore, the magnetic field effect on photocurrent in the m-MTDATA:BCP-based PV devices can be rationalized in terms of the singlet–triplet intersystem crossing of long-radius electron–hole pairs at the HFC field scale and in terms of the trionic model of the carrier mobility modulation at the FSM scale, as we have concluded in the case of the MPC signals from Section 3.4.

Al LiF m-MTDATA:BCP MoO3 hν

3

ITO

2 300

500

B [mT] 0

100 mT 0 mT ONSAGER

I [nA]

50

Uoc

100 150 200

3.6. Magnetic field effects on electromodulated photoluminescence and photoluminescence

Isc = 4% Isc

To examine the magnetic field effect on electromodulated photoluminescence we have recorded the magnitudes of the (2x)EML signals defined by the Formula (1) as a function of the magnetic field

= 3.8 % Isc 0.0

0.5

1.0

1.5

2.0

U [V] Fig. 10. (a) The relative change of short-circuit photocurrent intensity (DISC =ISC ) versus magnetic field strength. The solid line stands for the best fit according to the double-Lorentzian function with the same parameter values as in Fig. 9a. Inset: the pictorial view of a photovoltaic device. (b) The voltage dependence of the photocurrent measured under magnetic field (circles) and with no magnetic field (squares). The solid line is calculated according to the 3D-Onsager model of geminate recombination with parameter values as described in the text.

the charge carrier extraction efficiency from the active film (see e.g. Refs. [63,19]). In the measurements the ITO device electrode was positively polarized whereas the Al electrode was negatively polarized by the external source of voltage. The value of the cell current I varied between the short-circuit current, ISC = 200 nA (recorded at U = 0 V bias voltage) and the I = 0 corresponding to the cell open-circuit voltage, UOC = 1.68 V. Under cell photon flux of approx. 5  1014 cm2 s1 the achieved photoconversion efficiency was as high as 0.16%. Importantly, in the heterojunction-type PV devices an internal, built-in, electric field is considered to enhance the dissociation process of geminate e–h pairs (hence the generation of photocurrent), this field resulting mainly from relevant energy levels alignments of organic layer and used electrodes. That the intensity of the internal electric field affects the carrier photogeneration (hence the photocurrent intensity) in heterojunction polymer/perylene PV cells has been proved in Ref. [64] by means of the electroabsorption (EA) technique. Therefore, we attempted to rationalize the PV characteristics of our m-MTDATA:BCP-based structure in terms of the Onsager model of e–h pair dissociation, as in the case of photoconductivity of the m-MTDATA:BCP-layers from Section 3.3. However, in the photovoltaic device, the dissociation driving electric field, F = FINT  FBIAS, is the sum of the internal electric field intensity and the opposite external field induced by the forward biased voltage (FBIAS = U/d). Here, we analyse the current–voltage I–U characteristics (Fig. 10b) as successfully reproduced experimental data (points) applying the Onsager formalism for solely geminate recombination losses and neglecting any non-geminate mechanisms (solid line). An excellent reconstruction of the I–U characteristics is obtained which confirms that geminate recombination of e–h pairs can be suppressed by the applied electric field. Nevertheless, we are aware of possible considerable contribution

(a)

6

0.15

(2ω) EML

(b)

1.8×10 V/cm

0.10

5

8.4×10 V/cm 0.05

5

4.7×10 V/cm 0.00

0

20

40

60

B [mT]

(b)

0.3

MPL [%]

100

0.2

MPL = A 0.1

B2 B + B 02 2

B0 = 2.8 mT 0.0 0

20

40

60

B [mT]

(c) d = 100 nm

0.20

MPL [%]

ΔIsc/Isc [%]

4

0.15

B = 6 mT

0.10 0.05

φ = 1×10 0.0

14

2

ph/(cm s) 1.0

2.0

6

F [10 V/cm] Fig. 11. (a) The (2x)EML signal displayed as a function of magnetic field (circles) at various electric fields strength. The MPL effect: (b) versus the magnetic field strength, (c) versus the electric field strength. The single Lorentzian function is depicted by solid line in part (b) of the figure. The solid line in part (c) is a guide to the eye. The photoluminescence was excited by light of kexc ¼ 313 nm.

D. Pelczarski et al. / Organic Electronics 25 (2015) 362–376

strength, B, at different values of electric field strength, F, applied to the m-MTDATA:BCP layers -Fig. 11a. From the figure it is seen that, despite the electric field F varying up to 2  106 V/cm, the (2x)EML signals within the measuring technique sensitivity of ca. 0.2% are apparently independent of the applied magnetic field strengths. Here we shall recall our former considerations that the long-radius, 1 ðDþ ::A Þlong , and short-radius, 1 ðDþ :A Þshort , pairs are generated in the m-MTDATA:BCP layers, independently, from the relevant sorts of excited encounter-complexes, ð1 D ::A0 Þ (cf. Fig. 3) and that prevailingly the 1 ðDþ ::A Þlong pairs do dissociate to produce free charge carriers or via the intersystem crossing generate the triplet, 3 ðDþ ::A Þlong , ones, the latter giving practically no contribution to the photoconductivity of the organic solid (see Fig. 9b). To rationalize the apparent magnetic field independence of the (2x)EML signals from Fig. 11a it is to be assumed that the 1 ðDþ ::A Þlong pairs can mainly recombine non-radiatively in some sample regions in contrast to the fairly rare regions with special (probably sandwich) orientation of paired molecules contributing prevailingly to the radiative recombination. It should be noted here that the ISC process is rather hindered between the 1 ðDþ :A Þshort and 3 ðDþ :A Þshort pairs as well as the 1 ðDAÞ and 3 ðDAÞ exciplexes due to the rather high values of relevant electrostatic exchange energy induced by the short distances in these excited entities. Therefore, only a minor increase in the sample photoluminescence should occur when the magnetic field-suppressed ISC process, 1 ðDþ ::A Þlong ! 3 ðDþ ::A Þlong , gives rise to the population of ðDþ ::A Þlong pairs. In fact, the low-level (down to 0.02%) magnetic field effect on photoluminescence (MPL) signals could be detected using higher-sensitivity magnetomodulation technique as seen from Fig. 11b where the experimental data (points) are fitted well by a single Lorentzian function (cf. Formula (11)) plotted by a solid line. With magnetic field intensity exceeding the HFC scale the ISC is gradually switched-off and the MPL(B) characteristic becomes flat. The MPL Lorentzian width B0 = 2.8 mT is consistent with the MPC low-field fitting parameter BLFE = 4 mT (Fig. 9a) which indicates that both effects are originated from the similar (long-radius) e–h pairs. As in the case of the MPC signal from Fig. 9c, the MPL signal magnitude at fixed B of the HFC scale should be a monotonically decreasing function of electric field strength, F, due to F-induced shortening of the e–h pair lifetime. In fact, as seen in Fig. 11c the MPL magnitude recorded at fixed B = 6 mT decreases monotonically with F exceeding ca. 8  105 V/cm, in a manner somehow analogous to the curve of Fig. 9c. To sum up, in low-dielectric-constant organic systems with luminescent exciplex states one should not expect any significant effects of magnetic field on photoluminescence which results from rather high electrostatic exchange energy (the singlet–triplet splitting, DST) within an exciplex. Indeed, in another electron donor– electron acceptor system, m-MTDATA:PBD, where PBD stands for oxadiazole derivative, the DST  50 meV is reported as obtained by the thermally activated delayed fluorescence (TADF) technique [10], later on this value being reduced to the DST  5 meV as also received from the TADF measurements [65] which is at least three orders of magnitude larger than a typical interaction energy of an electron magnetic moment with hyperfine-scale magnetic field (glBBHFC = 1 lV with lB denoting the Bohr magneton, the giromagnetic Lande factor g = 2 and BHFC = 5 mT). The weak MPL effects in such materials are then to be ascribed to the operation of this minority of long-radius e–h pairs (with DST on the HFC scale) which recombine geminately to the ground state via fluorescent exciplex states. Interestingly, unlike under optical excitation, in electrically excited m-MTDATA:BCP system (a heterojunction-based OLED device) a significant magneto-electroluminescence (MEL) signal of ca 10%-magnitude has been recorded [66]. We consequently 1

375

strongly suggest that a dominant part of this effect originates from the long-radius e–h pairs which as intermediate states are involved in the volume (bimolecular) recombination process in an OLED structure. 4. Summary and conclusions Electric field dependencies of electromodulated photoluminescence and photocurrents along with the magnetic field effects on photocurrents, photovoltaic characteristics, electromodulated photoluminescence and photoluminescence have been investigated in quartz/Al/m-MTDATA:BCP[1:1]/Al sandwich structures to trace the different electron–hole distances in pairs involved in corresponding generation processes. The electromodulation processes are rationalized in terms of the bulk, 3-D Onsager as well as with the Sano–Tachiya–Noolandi–Hong mechanisms of e–h pair separation. Assuming that differently distanced e–h pairs do originate from the relevant encounter complexes we found the electromodulated photocurrents to be determined by operation of both long-radius and short-radius e–h pairs (respectively with low and high exchange energy) whereas the electromodulated photoluminescence quenching can be related to predominant involvement of merely short-radius fraction of the e–h pairs in the exciplex creation process. The photocurrents, photovoltaic characteristics, electromodulated photoluminescence and photoluminescence are influenced by external magnetic field of the hyperfine coupling (HFC) scale which modulates the singlet–triplet intersystem crossing of long-radius e–h pairs. Acknowledgements This work was financially supported by National Sience Centre in Poland under grant DEC-2011/03/B/ST7/01888. One of the authors (M.K.) was also financially supported by the Ministry of Science and Higher Education in Poland under ‘‘Diamond Grant’’ 0228/DIA/2013/42. The authors wish to thank Paulina Morawska for assistance during the EML measurements and Maciej Mis´nik for sample preparation. References [1] A. Köhler, H. Bässler, Electronic Processes in Organic Semiconductors: An Introduction, Willey-VCH, Weinheim, 2015. [2] W. Brütting, Ch. Adachi (Eds.), Physics of Organic Semiconductors, Wiley-VCH, Weinheim, 2012. [3] J. Kalinowski, Organic Light Emitting Diodes: Principles, Characteristics, and Processes, Marcel Dekker, New York, 2005. [4] B. Hu, L. Yan, M. Shao, Adv. Mater. 21 (2009) 1500. [5] H. Hayashi, Introduction to Dynamic Spin Chemistry, World Scientific, Singapore, 2005. [6] E. Frankevich, A. Zakhidov, K. Yoshino, Y. Maruyama, K. Yakushi, Phys. Rev. B 53 (1996) 4498. [7] J. Kalinowski, J. Szmytkowski, W. Stampor, Chem. Phys. Lett. 378 (2003) 380. [8] J. Kalinowski, M. Cocchi, D. Virgili, P. Di Marco, V. Fattori, Chem. Phys. Lett. 380 (2003) 710. [9] Y. Sheng, T.D. Nguyen, G. Veeraraghavan, Ö. Mermer, M. Wohlgenannt, S. Qiu, U. Scherf, Phys. Rev. B 74 (2006) 045213. [10] K. Goushi, K. Yoshida, K. Sato, C. Adachi, Nat. Photon. 6 (2012) 253. [11] W. Wagemans, B. Koopmans, Phys. Stat. Sol. (b) 248 (2011) 1029. [12] V.N. Prigodin, J.D. Bergeson, D.M. Lincoln, A.J. Epstein, Synth. Met. 156 (2006) 757. [13] A.J. Schellekens, W. Wagemans, S.P. Kersten, P.A. Bobbert, B. Koopmans, Phys. Rev. B 84 (2011) 075204. [14] M. Cox, P. Janssen, F. Zhu, B. Koopmans, Phys. Rev. B 88 (2013) 035202. [15] P.A. Bobbert, T.D. Nguyen, F.W.A. van Oost, B. Koopmans, M. Wohlgenannt, Phys. Rev. Lett. 99 (2007) 216801. [16] W. Wagemans, F.L. Bloom, P.A. Bobbert, M. Wohlgenannt, B. Koopmans, J. Appl. Phys. 103 (2008) 07F303. [17] M. Reufer, M.J. Walter, P.G. Lagoudakis, A.B. Hummel, J.S. Kolb, H.G. Roskos, U. Scherf, J.M. Lupton, Nature 4 (2005) 340. [18] A.C. Morteani, P. Sreearunothai, L.M. Herz, R.H. Friend, C. Silva, Phys. Rev. Lett. 92 (2004) 247402.

376 [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42]

D. Pelczarski et al. / Organic Electronics 25 (2015) 362–376 L.L. Chen, W.L. Li, M.T. Li, B. Chu, J. Lum. 122–123 (2007) 667. M. Gazda, Private communication. J. Kalinowski, W. Stampor, P.G. Di Marco, J. Chem. Phys. 96 (1992) 4136. P. Janssen, W. Wagemans, W. Verhoeven, E.H.M. van der Heijden, M. Kemerink, B. Koopmans, Synth. Met. 161 (2011) 617. W. Stampor, W. Mróz, Chem. Phys. 331 (2007) 261. M.I. Ranasinghe, O.P. Varnavski, J. Pawlas, S.I. Hauck, J. Louie, J.F. Hartwig, T. Goodson III, J. Am. Chem. Soc. 124 (2002) 6520. H.Z. Gao, C.S. Qin, H.Y. Zhang, S.X. Wu, Z.M. Su, Y. Wang, J. Phys. Chem. A 112 (2008) 9097. M. Mis´nik, K. Falkowski, W. Mróz, W. Stampor, Chem. Phys. 410 (2013) 45. W. Stampor, A. Tykocki-Piłat, Photonics Lett. Pol. 3 (2011) 64. W. Stampor, Chem. Phys. 315 (2005) 259. T. Förster, Discuss. Faraday Soc. 27 (1959) 7. B.N. Bandyopadhyay, A. Harriman, J. Chem. Soc.: Faraday Trans. 1 73 (1977) 663. J.A. Barltrop, J.D. Coyle, Principles of Photochemistry, Wiley, Chichester, New York, Brisbane and Toronto, 1979. J. Kalinowski, M. Cocchi, D. Virgili, V. Fattori, J.A.G. Williams, Chem. Phys. Lett. 432 (2006) 110. A. Weller, Z. Physik, Chemie NF 130 (1982) 129. A. Weller, H. Staerk, R. Treichel, Faraday Discuss. Chem. Soc. 78 (1984) 271. J.J. McCullough, Chem. Rev. 87 (1987) 811. S.A. Jenekhe, J.A. Osaheni, Science 265 (1994) 765. N. Ohta, S. Umeuchi, Y. Nishimura, I. Yamazaki, J. Phys. Chem. 102 (1998) 3784. A.I. Burshtein, J. Chem. Phys. 117 (2002) 7640. M. Hilczer, S. Traytak, M. Tachiya, J. Chem. Phys. 115 (2001) 11249. M. Hilczer, M. Tachiya, J. Chem. Phys. 117 (2002) 1759. N. Ohta, M. Koizumi, Y. Nishimura, I. Yamazaki, Y. Tanimoto, Y. Hatano, M. Yamamoto, H. Kono, J. Phys. Chem. 100 (1996) 19295. M. Pope, E.C. Swenberg, Electronic Processes in Organic Crystals and Polymers, Oxford Univ. Press, New York, 1999.

[43] [44] [45] [46] [47] [48] [49] [50] [51] [52] [53] [54] [55] [56]

[57] [58] [59] [60] [61] [62] [63] [64] [65] [66]

M. Schwoerer, H.C. Wolf, Organic Molecular Solids, Wiley, Weinheim, 2007. J. Frenkel, Phys. Rev. 54 (1938) 647. L. Onsager, Phys. Rev. 54 (1938) 554. H. Sano, M. Tachiya, J. Chem. Phys. 71 (1979) 1276. J. Noolandi, K.M. Hong, J. Chem. Phys. 70 (1979) 3230. W. Stampor, Chem. Phys. 256 (2000) 351. B. Ries, G. Schönherr, H. Bässler, M. Silver, Philos. Mag. B 48 (1983) 87. K. Falkowski, W. Stampor, P. Grygiel, W. Tomaszewicz, Chem. Phys. 392 (2012) 122. Y. Shirota, K. Okumoto, Proc. SPIE 4105 (2001) 158. Z. Xie, T. Wong, L.S. Hung, S. Lee, Appl. Phys. Lett. 80 (2002) 1477. M.I. Khan, G.C. Bazan, Z.D. Popovic, Chem. Phys. Lett. 298 (1998) 309. Z.D. Popovic, M.I. Khan, A.M. Hor, J.L. Goodman, J.F. Graham, J. Phys. Chem. B 106 (2002) 8625. J. Kalinowski, M. Cocchi, D. Virgili, Chem. Phys. Lett. 441 (2007) 286. Ö. Mermer, G. Veeraraghavan, T.L. Francis, Y. Sheng, D.T. Nguyen, M. Wohlgenannt, A. Köhler, M.K. Al-Suti, M.S. Khan, Phys. Rev. B 72 (2005) 205202. D. Pelczarski, P. Grygiel, M. Mis´nik, W. Stampor, Org. Electron. 21 (2015) 54. M. Wohlgenannt, Z.V. Vardeny, J. Phys.: Condens. Matter 15 (2003) R83. D.R. McCamey, H.A. Seipel, S.-Y. Paik, M.J. Walter, N.J. Borys, J.M. Lupton, C. Boehme, Nat. Mater. 7 (2008) 723. M.A. Baldo, S.R. Forrest, Phys. Rev. B 62 (10) (2000) 958. F. Li, L. Xin, S. Liu, B. Hu, Appl. Phys. Lett. 97 (2010) 073301. J.Y. Song, N. Stingelin, A.J. Drew, T. Kreouzis, W.P. Gillin, Phys. Rev. B 82 (2010) 085205. Y.W. Su, S.C. Lan, K.H. Wei, Mater. Today 15 (2012) 554. I. Hiromitsu, S.I. Mada, A. Inoue, Y. Yoshida, S. Tanaka, Jpn. J. Appl. Phys. 46 (2007) 7241. D. Graves, V. Jankus, F.B. Dias, A. Monkman, Adv. Funct. Mater. 24 (2014) 2343. P. Chen, Q. Peng, L. Yao, N. Gao, F. Li, Appl. Phys. Lett. 102 (2013) 063301.