www.MaterialsViews.com

FULL P

APER

www.advenergymat.de

Investigation of Driving Forces for Charge Extraction in Organic Solar Cells: Transient Photocurrent Measurements on Solar Cells Showing S-Shaped Current–Voltage Characteristics

Wolfgang Tress , * Steef Corvers , Karl Leo , and Moritz Riede

The role of drift and diffusion as driving forces for charge carrier extraction in fl at heterojunction organic solar cells is examined at the example of devices showing intentional S-shaped current–voltage ( J-V ) characteristics. Since these kinks are related to energy barriers causing a redistribution of the elec-tric fi eld and charge carrier density gradients, they are suitable for studying the limits of charge extraction. The dynamics of this redistribution process are experimentally monitored via transient photocurrents, where the current response on square pulses of light is measured in the μ s to ms regime. In combination with drift-diffusion simulation data, we demonstrate a pile-up of charge carriers at extraction barriers and a high contribution of diffusion to photocurrent in the case of injection barriers. Both types of barrier lead to S-kinks in the J-V curve and can be distinguished from each other and from other reasons for S-kinks (e.g. imbalanced mobilities) by applying the presented approach. Furthermore, it is also helpful to investigate the driving forces for charge extraction in devices without S-shaped J-V curve close to open circuit to evaluate whether their electrodes are optimized.

1. Introduction

Organic electronics is progressing rapidly due to large efforts in scientifi c research and recently in industrial development. Apart from promising technological aspects like large-scale production with low amounts of material and at low cost, organic semiconductor devices are interesting for fundamental research in physics. Organic photovoltaics, in particular, is a major fi eld where research has focused on in recent years. Better materials and understanding of the devices have led to steadily increasing power-conversion effi ciencies recently exceeding 10%. [ 1 , 2 ]

© 2013 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

DOI: 10.1002/aenm.201200931

Dr. W. Tress, [†] S, Corvers, Prof. K, Leo, Dr. M. RiedeInstitut für Angewandte Photophysik Technische Universität Dresden George-Bähr-Str, 1, 01069 Dresden, Germany E-mail: woltr@ifm.liu.se [†] Present Address: Biomolecular and Organic Electronics IFM Linköping University SE-581 83 Linköping, Sweden

Adv. Energy Mater. 2013,DOI: 10.1002/aenm.201200931

A basic characterization of working organic solar cells usually relies on steady-state measurements like current–voltage characteristics (J-V curve) or external quantum effi ciency data. Impedance spec-troscopy measurements are performed on organic solar cells to get a closer insight into charge distribution, recombination and trapping dynamics. [ 3 , 4 ] Transient tech-niques like photoinduced charge extrac-tion under linearly increasing voltage (photo-CELIV) are employed to extract mobilities and to investigate the dynamics of recombination. [ 5 , 6 ] Recombination is also investigated using the transient photo-voltage method combined with charge carrier extraction measurements. [ 7 ]

A simple approach to characterize dynamics is a transient photocurrent measurement, where the current response on square pulses of light is monitored. It has been used to determine mobili-

ties [ 8 ] and to describe trapping of charge carriers. [ 9 ] Here, we apply and extend this method to small molecule solar cells to get a closer insight into the working principle of multilayer devices. As the driving force for charge carrier extraction is a matter of debate, [ 10 ] we identify the role of drift and diffusion separately. In the transient measurement, we vary the light intensity and the applied DC voltage bias to probe the whole voltage range of the J-V curve. This is in particular interesting for devices showing infl ection points in the J-V curve (so called S-kinks), [ 11–21 ] as these points indicate a change in the working regime of the solar cell. In Ref. [19] we have shown in experi-ment and simulation that charge transport layers energetically misaligned to the energy levels of the active materials in planar heterojunction solar cells are one reason leading to S-kinks. According to simulation, these kinks are caused either by pho-togenerated charge carriers piling up at an extraction barrier, or by charges diffusing against an electrical fi eld within the active layers in the case of an injection barrier.

In this article we confi rm these predictions by visualizing a predominantly diffusion driven current and the pile-up of charge carriers indicated by characteristic overshoots in the transient current response. A time integration of the current decay gives us an estimate of the piled-up charge carrier density

1wileyonlinelibrary.com

www.MaterialsViews.com

FULL

PAPER

www.advenergymat.de

at an energy barrier. To correctly interpret the experimental data, we extend our drift-diffusion model [ 22 ] to simulate the photocurrent-transient experiment. We focus on the qualitative difference in the photocurrent response, dependent on the lim-iting processes in the solar cell. Starting from a well working solar cell, we show data of solar cells with extraction barriers for holes, followed by devices with an injection barrier, and fi nally compare these data to data of solar cells with an S-kink due to a very low donor mobility. [ 17 ]

2. Investigated Devices

Solar cells are prepared in p-i-n or p-i-metal geometry. [ 23 , 24 ] Employing this approach, the photoactive stack, which consists of a donor-acceptor bilayer, is sandwiched between a p-doped (transparent) hole transport layer (HTL) and an n-doped elec-tron transport layer or a BPhen/Al contact. This structure allows ohmic contacts to the electrodes independent of the used donor material. Figure 1 (a) shows the stacks of the investigated devices in detail. These devices are selected as representatives of a spe-cifi c group, where each group is assigned to a specifi c type A, B, or C. The type refers to the dominating property independent of the specifi c materials employed. Devices of type A and B are planar heterojunctions in a stack ITO/p-HTL (20 nm)/HTL/donor/C 60 (40 nm)/BPhen(6 nm)/Al. The HTL in devices of type

2 wileyonlinelibrary.com © 2013 WILEY-VCH Verlag G

Figure 1 . (a) Schematic solar-cell stacks with Al cathode and ITO anode cohole transport layer (HTL). The donor-acceptor fl at-heterojunction consistmaterials combined with the acceptor C 60 . Devices of type A with HTL = docells without energy barriers for charge extraction. Devices B with HTL ≠ barriers at the HTL/donor interface. Device C contains a donor with a ver(b) Schematic energy diagram at short circuit for devices B1 and B2 with extrbarrier, respectively, formed by a HOMO level offset at the HTL/donor inte

A consists of the same material as the donor. This means that an energy barrier for hole transport is not formed at the HTL/donor interface. C 60 /BPhen/Al allows an electron transport without any signifi cant energy barrier. In device A1, a layer of 16 nm of MeO-TPD (N4,N4,N4 ′ ,N4 ′ -tetrakis(4-methoxyphenyl)-[1,1 ′ -biphenyl]-4,4 ′ -diamine) and in device A2 BPAPF (9,9-bis4-[di-(p-biphenyl)aminophenyl]uorene) are employed as donor and HTL materials. The ionization potential (IP, measured by photoelectron spectroscopy) of MeO-TPD is reported to be 5.1 ± 0.1 eV and of BPAPF 5.6 ± 0.1 eV 19 .

Devices of type B contain different materials for donor and HTL. In device B1, MeO-TPD (20 nm) is employed as donor and BPAPF (10 nm) as HTL. As the IP is a measure for the hole transport level, an extraction barrier for holes at the donor/HTL interface is expected, as depicted in the sketch of the energy diagram of Figure 1 (b). Interchanging MeO-TPD with BPAPF and vice versa leads to device B2, exhibiting an injec-tion barrier for holes at the HTL(8 nm)/donor(8 nm) interface. The stack of device C consists of ITO/p-doped BPAPF(25 nm, 10 weight%)/Ph4-Ph4-DIP(20 nm)/C 60 (30 nm)/n-C 60 (10 nm, 3 wt%)/Al. The key difference is the donor material Ph4-Ph4-DIP (1,2,3,4,9,10,11,12-octaphenyl-diindeno[1,2,3-cd:1 ′ ,2 ′ ,3 ′ -lm]perylene), which shows a very low hole mobility. [ 17 ] A signifi cant barrier to the HTL is not expected, because the IP of Ph4-Ph4-DIP is close to the IP of BPAPF.

mbH & Co. KGaA, Weinh

vered by a p-doped s of different donor nor represent solar donor give rise for y low hole mobility. action and injection rface.

3. Modeling

The model is based on a one-dimensional transient drift-diffusion approach. As the main physical models and parameters are described in Refs. [19,22] and the basic algo-rithm (fi nite differences, here with Schar-fetter-Gummel method) is published in Staudigel et al., [ 25 ] we do not report details and refer to the given sources. A compre-hensive description of the drift-diffusion approach for modeling transients of a bulk heterojunction single layer can be found in Ref. [8]. In this Reference, also the dis-sociation of bound electron hole pairs (CT states) at a donor/acceptor (D/A) interface is discussed in detail, assuming the applica-bility of the highly debated Onsager-Braun theory. [ 26–28 ] As we do not have any indica-tion that this process plays a signifi cant role for the investigated devices, we assume an instantaneous exciton separation at the D/A interface in the simulations.

To model photocurrent transients, the transient algorithm starts with a previ-ously obtained steady-state solution in dark or under constant illumination and calcu-lates the current response iteratively after switching on/off the illumination intensity. Steady state is reached as soon as the elec-trical particle current is spatially constant over the whole device. To extract a current

eim Adv. Energy Mater. 2013, DOI: 10.1002/aenm.201200931

FULL P

APER

www.advenergymat.de

www.MaterialsViews.com

transient J ( t ) from the simulation data, the displacement cur-rent JD(t) = ε0εr

∂ E (t)∂t has to be taken into account. Here, E

denotes the electric fi eld, ε 0 ε r the permittivity of the material, and t the time. According to Ampère’s circuital law (4th Max-well equation, � c HdS = ∫ ∫ S ( J e / h + J D ) dA with the magnetic fi eld H ), the sum of displacement and real (electron and hole) cur-rent ( J e/h ) at a fi xed time is independent of the position in the device, because one can select an arbitrary area A cutting the device for the same border curve S . This law is fulfi lled in our simulation. All displayed fi gures will show J e / h ( t ) + J D ( t ), which corresponds to the current J ( t ) measured at the (low) input resistance of the oscilloscope.

The authors emphasize that a complicated organic solar cell stack with several interfaces and organic materials cannot yet be treated predictively or seriously quantitatively with a one-dimensional drift-diffusion approach. However, this model is very valuable, because it helps to identify and explain domi-nating physical mechanisms. Furthermore, this kind of simula-tions describes conditions regarding a characteristic behavior of devices. Thus, the strategy of this paper is not to fi nd the best fi t to selected experimental data, because it most probably will not fi t an extended data range, but to reproduce the experimental trends identifying the important parameters while using a rea-sonable and consistent data set.

4. Results and Discussion

Figure 2 (a) shows the J-V curves of the samples discussed in the following. These steady-state J-V data are discussed in detail for devices of types A and B in Ref. [19] and for device C in Ref. [17]. To understand these curves, we briefl y sum-marize the conclusions from these steady-state data: The dif-ference in open-circuit voltage ( V oc ) of all devices is attributed to the different ionization potentials (IP ≈ –HOMO) values of the donor materials and roughly scales with the effective gap

© 2013 WILEY-VCH Verlag Gm

Figure 2 . (a) J-V curves of the samples investigated. Devices of type A do nis defi ned by the donor material independent of the barrier. The device stacferent types. Devices B1 + B2: The ionization potential of donor and hole trais 0.5 eV, the extraction barrier 0.3 eV considering a dipole at the interface BPare set to 10 − 5 cm 2 /Vs and electron mobilities to 5 × 10 − 4 cm 2 /Vs. Recombconstant of 7.23 × 10 − 11 cm − 3 s − 1 . The simulation parameters of the low-mob

−1 −0.5 0 0.5 1 1.5

−2

0

2

4

6

voltage / V

curr

ent d

ensi

ty /

mA

cm

−2

A1 MeO−TPDA2 BPAPFB1 Extra. Barr.B2 Injec. Barr.C low μ

(a)

Adv. Energy Mater. 2013,DOI: 10.1002/aenm.201200931

(IP donor – EA acceptor ). Here, EA denotes the electron affi nity ( ≈ –LUMO). Apart from little changes due to varying recom-bination probabilities, [ 7 ] this correlation is accepted and well-known. [ 29 ] The devices of type A show reasonable fi ll factors (43 to 48%), whereas the other devices show distorted J-V curves with an S-kink and a FF of 9 to 28%. The combination BPAPF/MeO-TPD (B1) gives rise to an extraction barrier at the interface donor/HTL, which reduces hole extraction and causes an S-kink. [ 18 , 19 ] In the case of an injection barrier (B2), the slight S-kink close to V oc describes the region where the extracted photocurrent is only diffusion driven with a drift component in the “wrong” direction. [ 12 , 19 ] For the Ph4-Ph4-DIP device (C) no signiffi cant barriers are expected. The S-kink in the J-V curve is attributed to a high mismatch in charge carrier mobilities in donor and acceptor. [ 17 , 20 ]

These explanations for the reasons of the S-kinks are based on the results of drift-diffusion simulations, which reproduce the different characteristics of the S-kinks well (Figure 2 (b)). The most important input parameters of the simulation are given in the fi gure caption. In Refs. [ [ 17 – 19 ] ] calculated fi eld pro-fi les and charge carrier density distributions within the devices are used to visualize the origins of the S-kinks. An experimental verifi cation of these predictions is presented in the following by an analysis of transient photocurrent data.

4.1. Well Working Solar Cells

Figure 3 (a) shows transient current curves for device A1 rep-resenting a solar cell without any signifi cant energy bar-riers for charge carrier extraction or injection. The current signal follows the illumination pulse and reaches steady state within few μ s. This is a commonly observed feature. [ 9 ] The delay time is attributed to limited charge transport proper-ties of the intrinsic organic layers. Decay dynamics with time constants in the range of 1 μ s and lower can be attributed to

3wileyonlinelibrary.combH & Co. KGaA, Weinheim

ot show any S-kink, whereas the other devices do. The open-circuit voltage ks are shown in Figure 1 . (b) Simulation data of exemplary devices of dif-nsport layer are varied to realize barriers. The value of the injection barrier

APF/MeO-TPD measured by photoelectron spectroscopy. Hole mobilities ination at the D/A interface is described by a bimolecular recombination ility ( μ ) device can be found in Ref. [ 17 ].

−1 −0.5 0 0.5 1 1.5

−2

0

2

4

6

voltage / V

curr

ent d

ensi

ty /

mA

cm

−2

IP 5.1 eVIP 5.6 eVExtra. Barr. Injec. Barr.low μ

(b)

www.MaterialsViews.com

FULL

PAPER

www.advenergymat.de

Figure 3 . Transient photocurrent responses of device A1 on an illumination pulse of 50 μ s as a function of (a) illumination intensity and (b) applied voltage. The shape of the transients hardly changes. Part (b) indicates the selectivity of a fl at heterojunction device, as the photocurrent in forward direction does not change sign.

0 10 20 30 40 50 600

0.5

1

1.5

2

2.5

3

3.5

4

time / µs

−J

/ mA

cm

−2

(a)1.7 suns

1.2 suns

0.84 suns

0.50 suns

0.17 suns

0 10 20 30 40 50 60−53

−16

−2

−1

0

1

2

3

time / µs

−J

/ mA

cm

−2

(b)

Voc

= 0.56 V

−1.0 V−0.5 V0.0 V0.2 V0.4 V

0.6 V

0.8 V

1.0 V

the macroscopic series resistance caused by the lateral cur-rent through ITO (30 … 40 Ω ) and the measurement resistor (50 Ω input impedance of oscilloscope), rather than to the device itself. Assuming that the intrinsic layers behave like a simple planar capacitor with a capacity of ε r ε 0 × area/thickness ≈ 3 × ε 0 × 6.44 mm 2 / 56 nm ≈ 3 nF leads to a time constant τ = RC ≈ 0.3 μ s. As we do not discuss the decay time in detail, the RC limitation does not signifi cantly infl uence the conclusion drawn in this work. The transient currents measured at dif-ferent light intensities visualize that the photocurrent increases with light intensity without changing the shape of the transient. This indicates that in MeO-TPD/C 60 devices, a possible depend-ence of mobilities on intensity cannot be observed in this meas-urement and intensity range.

Figure 3 (b) shows measurements repeated at several applied DC voltages. The data indicate that the shape of the current response is mainly independent of the applied voltage. Even for voltages larger than V oc (0.56 V), photocurrent is still extracted in the same direction as under reverse bias and approaches 0 for voltages larger than 0.8 V. This indicates the selectivity

4 wileyonlinelibrary.com © 2013 WILEY-VCH Verlag Gm

Figure 4 . Transient photocurrent response on a 300 μ s light pulse for the extment and (b) simulation ( V bi = 1.16 V). The transients show overshoots whe

0 100 200 300 400

−0.5

0

0.5

1

1.5

2

2.5

time / µs

−J

/ mA

cm

−2

−1−0−0−0−0 0 0 0 0

(a)

−1 V

(

Voc

= 0.0.6 V

of the device given by the realization as a fl at heterojunction. Device A2 shows qualitatively the same behavior (not shown).

4.2. Extraction Barrier Device

In Figure 4 photocurrent transients at different applied voltages are shown for device B1 comprising an extraction barrier. The steady-state current under illumination at each voltage point matches the current expected from the J-V curve displayed in Figure 2 (a). In contrast to the response of devices of type A, overshoots are observed when switching on the illumination for applied voltages larger than − 0.5 V. According to Figure 2 (a) this is the voltage range where the S-shape can be seen. Furthermore, negative overshoots when switching off the illu-mination occur for voltages larger than 0 V. Overshoots when switching on the light have already been observed and attrib-uted to traps in the active layer. [ 9 ] However, in the case of traps governing the transient current curve, an overshoot is not expected when switching off the light. The current decay is

bH & Co. KGaA, Weinheim

raction barrier device B1 as a function of applied DC voltage in (a) experi-n switching on/off the illumination.

0 100 200 300 400

−0.5

0

0.5

1

1.5

2

2.5

time / µs

−J

/ mA

cm

−2

.0 V

.8 V

.6 V

.4 V

.2 V.0 V.2 V.4 V.6 V

b)

54 V

−1 V

0.6 V

Adv. Energy Mater. 2013, DOI: 10.1002/aenm.201200931

www.MaterialsViews.com

FULL P

APER

www.advenergymat.de

then qualitatively comparable to the data of device A1, however showing slower dynamics. This is due to the contribution from trapped charges, which are extracted after a certain detrapping time commonly in the range of μ s to ms.

Device B1, however, shows current in reverse direction upon switching off the light when applying voltages close to V oc . This behavior is not expected and means that instead of the remaining photogenerated charge carriers leaving the device slowly, charge carriers are entering the device. We have observed this shape for devices with other material combina-tions and modifi ed extraction barriers as well. Furthermore, bulk heterojunction devices with extraction barriers show this behavior, too (see supplementary information). We conclude that this kind of double overshoot curve is a fi ngerprint of an extraction barrier. This conclusion is verifi ed by the simu-lated photocurrent transients shown in Figure 4 (b). These data reproduce the experimental data qualitatively well and are characteristic when simulating extraction-barrier devices. One can deduce from the voltage dependence that the overshoots are signifi cant in the voltage region showing the S-shape [cf. Figure 2 (b)], where the effects of the extraction barrier become visible. Under forward bias the overshoots vanish and the pho-tocurrent signal completely disappears verifying the selectivity of the device. (The full voltage sweep is shown in the supple-mentary information.)

This overshoot behavior is explained as follows: Upon switching on the light and hence electron/hole pair generation at the D/A interface, electrons leave the device quickly. This leads to a high displacement current over the intrinsic HTL due to the fact that there are no free charge carriers. Holes drift and diffuse away from the D/A interface to the donor/HTL inter-face, where they pile up at the extraction barrier. If the fi eld is not suffi cient to completely extract the photocurrent, over-shoots can be seen. The current causing these overshoots stops as soon as the device is in steady state. Thus, this steady-state current is lower. The piled-up space charge reduces the fi eld in donor and acceptor and leads to a fi eld mainly located in the intrinsic HTL as shown in Figure 3 of Ref. [18].

When switching off the light, the electron/hole source at the D/A interface vanishes, which leads to a strong diffusion gra-dient for the holes in the donor towards the D/A interface, as the holes cannot overcome the barrier and cannot penetrate the HTL. The holes stored at the extraction barrier then fl ow back to the D/A interface where they recombine with electrons pro-vided by the cathode. This implies a reverse current compared to the photocurrent. Such a current as response to switching off the light exists if the holes are stored at some distance from the position of generation. This proves the predictions of the simulation that holes are stored close to the extraction barrier. Furthermore, it demonstrates the role of diffusion. The height of the overshoots in simulation data strongly correlates with electron and hole mobilities in acceptor and donor, respectively, and can also be signifi cantly suppressed by a high peripheral series resistance. Furthermore, the overshoots are reduced for lower light intensities as the amount of charges to be displaced decreases (see Figure 1 in supplementary information). When leaving the S-kink region (V < − 0.5 V) the positive and nega-tive overshoots vanish, because the blocking effect of the bar-rier is superposed by a fi eld driven current over the barrier. The

© 2013 WILEY-VCH Verlag GAdv. Energy Mater. 2013,DOI: 10.1002/aenm.201200931

steady-state photocurrent approaches the one without barrier as obvious from the J-V curves in Figure 2 . Simulation data show this transition as well.

Another simple picture for explaining these overshoots is an analysis of the device in an equivalent-circuit model. The intrinsic HTL is then represented by a huge capacitance and a resistor in parallel which depends on the applied voltage. In the S-kink region, the capacitance dominates the current response because extraction is improbable and the resistance large. The current response when switching on the light is then a current which charges this capacity consisting of a planar capacitor formed by the doped HTL and the interface HTL/donor as elec-trodes. The dielectric layer is formed by the 10 nm intrinsic HTL and a depletion region in the p-HTL close to the interface p-HTL/intrinsic HTL. Charging this capacity requires some time which is represented by the decay and is limited by hole transport in the donor MeO-TPD and the external resistance as well. When switching off the light, the voltage source biasing the capacitor vanishes as it was provided by the charge carriers generated at the D/A interface. Therefore, the capacitor is dis-charged which results in an overshoot in negative direction.

Integrating the transient current in the overshoot regions over time gives an estimate of the stored charge at the extrac-tion barrier. From Figure 4 it can already be seen that the charge stored when switching on the illumination is in the same range as the charge released when switching off. However, the dis-placed charge upon switching off the light is slightly smaller as there is some charge leakage due to some holes crossing the barrier as already evident from the steady-state current. When the applied voltage approaches V oc and the steady-state pho-tocurrent zero, the displaced charges become roughly equal. Assuming that the charge is located and equally distributed in a region of, e.g. 3 nm, in front of the extraction barrier means a charge carrier density of 2 × 10 17 cm − 3 at 0.6 V. Access to these values of charge carrier densities might open the opportunity for quantitatively studying the effect of charge carrier densi-ties on the probability for charge carrier extraction over energy barriers.

Here, we qualitatively discuss the displaced charge (injected electrons) upon switching off the light as a function of illumi-nation intensity. The data are shown for different voltages in Figure 5 . Independent of voltage, the absolute charge Q increases logarithmically with light intensity. This is expected, as V oc and thus the voltage V over the HTL-capacitance C and in turn the charge Q = C × V scale logarithmically with light intensity. Thus, a change of the permeability of the extraction barrier as a func-tion of illumination intensity in the investigated range cannot be observed. As already seen in the transients (Figure 4 ), the extracted, i.e. piled-up charge increases from 0.1 to 0.5 V as the fi eld assisting charge carrier extraction decreases with applied voltage. However, when further increasing the voltage to the range of V oc and above, a signifi cant force for a redistribution of charges is not present any more as the fi eld in the device is very low in this voltage range. This is obvious as a device without barrier which shows the same V oc (cf. Figure 2 ) does not deliver any current at this point, i.e. charge that could be piled-up at an extraction barrier.

Comparing the voltage of the device with and without extrac-tion barrier at the same current gives an estimate on how

5wileyonlinelibrary.commbH & Co. KGaA, Weinheim

www.MaterialsViews.com

FULL

PAPER

www.advenergymat.de

Figure 5 . Injected charge when switching off the illumination measured at the extraction barrier device B1 as a function of illumination intensity and for different applied bias voltages. The logarithmic scaling follows the dependence of V oc on light intensity. The effect of the extraction barrier is most signifi cant for voltages close, however lower than V oc (0.56 V).

10−1

100

−15

−10

−5

0x 10

−9

intensity / "suns"

char

ge /

C c

m−

2

0.2 V

0.7 V

0.1 V

0.5 V

much voltage is “lost” due to the pile-up of charge carriers at the extraction barrier. This can be done because the D/A interface at a certain current should deliver a defi ned voltage. Changes in the J-V curve are then due to voltage-drops in other regions of the device. Doing this analysis with the aid of the J-V data of devices A1 and B1, e.g. at a current of − 0.7 mA cm − 2 , delivers a voltage difference of ≈ 0.5 V due to the HTL-capac-itor. Combining this voltage with the displaced charge at this point gives a capacitance per area C/A = Q /( V × A ) ≈ 4 × l0 − 5 mC cm − 2 /0.5 V = 80 nF cm − 2 and in turn a distance of 30 nm assuming a plate capacitor. This value is in a realistic range, combining 10 nm HTL and a 20 nm depletion region. This calculation is only a rough estimation for an intuitive visuali-zation. More accurate is the simulation data. It is once more

6 wileyonlinelibrary.com © 2013 WILEY-VCH Verlag G

Figure 6 . Transient photocurrent response of the injection barrier device B2by diffusion: (a) experiment; (b) simulation ( V bi = 0.86 V).

0 10 20 30 40

0

0.5

1

1.5

2

2.5

time / µs

−J

/ mA

cm

−2

(a)

0 V

Voc

=

1 V

emphasized that voltage is only “lost” as long as current is extracted - not at open circuit.

4.3. Injection Barrier Device

Figure 2 visualizes that device B2 with an injection barrier shows an S-kink as well, which results from a diffusion cur-rent against the electric fi eld, as demonstrated by simula-tion data in Figure 5 of Ref. [19]. Such a situation is caused, because the injection barrier decreases the built-in fi eld, whereas a device with selective contacts can deliver a V oc inde-pendent of the injection barrier height. This means that V oc can exceed the built-in potential. This leads to a region in the J-V curve, where the photocurrent is driven by diffusion only towards extraction. Such an effect is expected to have conse-quences on the photocurrent transients as well. The transient photocurrent responses of device B2 are plotted in Figure 6 for several applied voltages and show mainly a shape compa-rable to the one of devices of type A, because charge extraction works well (see the supplementary information for the whole sweep including forward bias). However, there is a difference for voltages, where the S-kink is present. There, a small posi-tive and negative overshoot after switching the light on (and off, respectively), can be seen. Simulations of injection barrier devices reproduce this effect, where it is indeed attributed to the dominance of diffusion. Upon switching on the light the diffusion gradient is highest. This leads to the overshoot with a reduced steady-state current as soon as the charge-carrier den-sity gradient is established giving rise to high charge carrier densities and therefore relatively high recombination at the D/A interface. When switching off the light, the diffusion force is reduced, however the inverse fi eld in the device accelerates charges towards the D/A interface. As there is a considerable amount of charge carriers present, which constituted the diffu-sion gradient, these holes and electrons fl ow back to the inter-face and recombine. However, this effect is relatively weak and was not observed for all injection barrier devices as it strongly depends on the charge transport properties of the donor.

mbH & Co. KGaA, Weinheim

as a function of applied voltage. Slight overshoots indicate currents driven

0 10 20 30

0

0.5

1

1.5

2

2.5

time / µs

−J

/ mA

cm

−2

0.0 V0.4 V0.5 V0.6 V0.7 V0.8 V0.9 V1.0 V

(b)

1 V

1 V

0 V

Adv. Energy Mater. 2013, DOI: 10.1002/aenm.201200931

www.MaterialsViews.com

FULL P

APER

www.advenergymat.de

Figure 7 . Transient photocurrent response of low-mobility device C showing response times in the μ s to ms range.

0 1 2 3 4 5

−11

−10

−9

0

1

2

3

time / ms

−J

/ mA

cm

−2

Voc

= 1.15 V2.0 V

1.5 V

1.0 V

0.5 V

−1.0 V−0.5 V0 V

4.4. Unbalanced Mobilities

The J-V curves of device C in Figure 2 and Ref. [17] show that unbalanced mobilities in donor and acceptor can also cause S-kinks. As we claimed the discovery of the finger-prints of energy barriers in the transient photocurrent data in the previous sections, this method is supposed to be capable of discriminating the reasons for other sources of S-kinks as well. Transient current data of device C for the voltage range covering the S-kink (cf. Figure 2 ) are shown in Figure 7 .

Despite a similar shape of the J-V curve, the transients look very different compared to the barrier devices, in par-ticular the timescale is orders of magnitude larger. Switching on the light leads to a slight overshoot, which is strongly dependent on light intensity. This overshoot might be attrib-uted to trapping of holes in the donor material, as described in Ref. [9].

In contrast to the barrier samples, the negative overshoot when switching off the light is missing. This means, that there is no charge stored at a barrier, which can move easily. At V oc the transient photocurrent vanishes completely, because hole transport in the donor away from the D/A interface is the lim-iting process. When applying forward bias, the current under illumination changes direction. This is due to photoconduc-tivity of the donor reducing the series resistance of the device, which directly results in a crossing point of the steady-state J-V curves in dark and under illumination. This photoconductivity effect in the donor can also be seen at single-carrier devices [ 22 ] showing signifi cantly higher currents under illumination. The photocurrent delivered by a fl at heterojunction is not supposed to change its direction. Another effect giving higher forward currents under illumination is predicted by drift-diffusion simulation [ 22 , 30 ] showing that a photoenhanced increase in dark injection current can be obtained under special circumstances. The data here demonstrate that photoconductivity is apart from this effect a very simple reason for an apparent positive pho-tocurrent in case of a selective device architecture (fl at hetero-junction) or selective contacts.

© 2013 WILEY-VCH Verlag GAdv. Energy Mater. 2013,DOI: 10.1002/aenm.201200931

5. Conclusions

We have used transient photocurrent measurements to inves-tigate the driving forces and charge-carrier distribution in organic solar cells. Currents driven by diffusion and by drift could be identifi ed separately in solar cells with high energy barriers at the contacts. We proved that predominantly fi eld-driven charge is indeed piled-up at an extraction barrier as predicted by simulation. Furthermore, we showed that the S-kink caused by an injection barrier is due to the fact that photocurrent extraction is only diffusion-driven. The investiga-tion of a device, where the S-kink is caused by another reason, i.e. imbalanced mobilities, shows that this method is capable of discriminating reasons for S-kinks in a simple way. This method of transient photocurrents, therefore, is very inter-esting for further quantitative and more detailed investigations of multilayer solar cells with a complicated stack, as well as for the identifi cation of the reasons for S-shaped J-V curves also in polymer solar cells in general. Furthermore, this method will allow to determine contact properties and driving forces for photocurrent close to open circuit in devices with non-S-shaped J-V curves.

6. Experimental Section The multilayer solar cells are prepared by subsequent evaporation in a vacuum system (K.J. Lesker company, UK) at a base pressure of 10 − 8 mbar and deposition rates of 0.3–0.5 Å/s. NDP2 (1 wt% in MeO-TPD), NDP9 (5 … 10 wt% in BPAPF) (p-doping), and NDN1 (3 wt% n-doping) (Novaled AG, Dresden, Germany) are used as molecular dopants. All other organic materials have been purifi ed at least twice by vacuum gradient sublimation, before being deposited. Indium-tin oxide coated glass (TFD, USA) with a sheet resistance of 30 Ohm/sq, pre-treated with acetone and oxygen plasma is used as substrate.

Current–voltage characteristics are recorded under simulated sunlight (16S-002 Solar Light Company Inc., Glennside USA), with an intensity of 120 mW/cm 2 . Transient photocurrent measurements are performed using a pulse generator (Hewlett Packard 8114A) driving a white high intensity light-emitting diode (Lumitronix Luxeon, peak emission wavelength of 450 nm, broad emission from 500 to 700 nm, rise and fall time < 100 ns). The light intensity is adjusted by the driving current of the LEDs which was calibrated using a silicon photodiode. The photocurrent of the devices is measured using an HP Agilent infi nium 54815A (500 MHz/1 Gs) oscilloscope with input impedance of 50 Ω . As low-noise voltage source, a battery with a resistive voltage divider (potentiometer) is used to apply the bias voltage.

Acknowledgements The research leading to these results has received funding from the BMBF (OPEG, grant no. 13N9720) and the European Research Council under the European Union's Seventh Framework Programme (FP7(2007–2013)/ERG grant agreement no. 267995). WT kindly thanks the Reiner Lemoine foundation for funding. Max Tietze is gratefully acknowledged for UPS measurements and Christopher McNeill and Zhe Li for the introduction into transient photocurrent measurements.

Received: November 12, 2012 Published online:

7wileyonlinelibrary.commbH & Co. KGaA, Weinheim

www.MaterialsViews.com

FULL

PAPER

www.advenergymat.de

[ 1 ] M. A. Green , K. Emery , Y. Hishikawa , W. Warta , E. D. Dunlop , Prog. Photovoltaics 2012 , 20 , 12 .

[ 2 ] Heliatek, press release :http://www.heliatek.com/newscenter/latest_news/neuer-weltrekord-fur-organische-solarzellen-heliatek-behauptet-sich-mit-12-zelleffi zienz-als-technologiefuhrer/?lang = en (accessed March 2013) .

[ 3 ] J. Bisquert , Phys. Rev. B 2008 , 77 , 1 . [ 4 ] L. Burtone , D. Ray , K. Leo , M. Riede , J. Appl. Phys. 2012 , 111 ,

064503 . [ 5 ] A. J. Mozer , N. S. Sariciftci , L. Lutsen , D. Vanderzande , R. Osterbacka ,

M. Westerling , G. Juska , Appl. Phys. Lett. 2005 , 86 , 112104 . [ 6 ] A. Pivrikas , N. S. Sariciftci , G. Juška , R. Österbacka , Prog. Photovol-

taics 2007 , 15 , 677 . [ 7 ] A. Maurano , R. Hamilton , C. G. Shuttle , A. M. Ballantyne , J. Nelson ,

B. O’Regan , W. Zhang , I. McCulloch , H. Azimi , M. Morana , C. J. Brabec , J. R. Durrant , Adv. Mater. 2010 , 22 , 4987 .

[ 8 ] N. S. Christ , S. W. Kettlitz , S. Valouch , S. Züfl e , C. Gärtner , M. Punke , U. Lemmer , J. Appl. Phys. 2009 , 105 , 104513 .

[ 9 ] C. R. McNeill , I. Hwang , N. C. Greenham , J. Appl. Phys. 2009 , 106 , 024507 .

[ 10 ] T. Ripolles-Sanchis , A. Guerrero , J. Bisquert , G. Garcia-Belmonte , J. Phys. Chem. C 2012 , 116 , 120723094950001 .

[ 11 ] M. Glatthaar , M. Riede , N. Keegan , K. Sylvester-Hvid , B. Zimmermann , M. Niggemann , A. Hinsch , A. Gombert , K. Sylvesterhvid , Solar Energy Mater. Solar Cells 2007 , 91 , 390 .

[ 12 ] C. Uhrich , D. Wynands , S. Olthof , M. Riede , K. Leo , S. Sonntag , B. Maennig , M. Pfeiffer , J. Appl. Phys. 2008 , 104 , 43107 .

[ 13 ] A. Kumar , S. Sista , Y. Yang , J. Appl. Phys. 2009 , 105 , 094512 . [ 14 ] M. R. Lilliedal , A. J. Medford , M. V. Madsen , K. Norrman ,

F. C. Krebs , Solar Energy Mater. Solar Cells 2010 , 94 , 2018 .

8 wileyonlinelibrary.com © 2013 WILEY-VCH Verlag G

[ 15 ] A. Wagenpfahl , D. Rauh , M. Binder , C. Deibel , V. Dyakonov , Phys. Rev. B 2010 , 82 , 115306 .

[ 16 ] J. Wang , X. Ren , S. Shi , C. Leung , P. K. Chan , Org. Electron. 2011 , 12 , 880 .

[ 17 ] W. Tress , A. Petrich , M. Hummert , M. Hein , K. Leo , M. Riede , Appl. Phys. Lett. 2011 , 98 , 063301 .

[ 18 ] W. Tress , S. Pfuetzner , K. Leo , M. Riede , J. Photon. Energy 2011 , 1 , 011114 .

[ 19 ] W. Tress , K. Leo , M. Riede , Adv. Funct. Mater. 2011 , 21 , 2140 . [ 20 ] M. Zhang , H. Wang , C. W. Tang , Appl. Phys. Lett. 2011 , 99 ,

213506 . [ 21 ] J. Wagner , M. Gruber , A. Wilke , Y. Tanaka , K. Topczak , A. Steindamm ,

U. Hermann , A. Opitz , Y. Nakayama , H. Ishii , J. Pfl aum , N. Koch , and W. Brutting , J. Appl. Phys. 2012 , 111 , 054509 .

[ 22 ] W. Tress , Dissertation, TU Dresden 2011 , 99–129 , 167 , 216 . [ 23 ] B. Maennig , J. Drechsel , D. Gebeyehu , P. Simon , F. Kozlowski ,

A. Werner , F. Li , S. Grundmann , S. Sonntag , M. Koch , K. Leo , M. Pfeiffer , H. Hoppe , D. Meissner , N. S. Sariciftci , I. Riedel , V. Dyakonov , J. Parisi , Appl. Phys. A 2004 , 79 , 1 .

[ 24 ] M. Riede , T. Mueller , W. Tress , R. Schueppel , K. Leo , Nanotechnology 2008 , 19 , 424001 .

[ 25 ] J. Staudigel , M. Stössel , F. Steuber , J. Simmerer , J. Appl. Phys. 1999 , 86 , 3895 .

[ 26 ] L. Onsager , Phys. Rev. 1938 , 54 , 554 . [ 27 ] C. L. Braun , J. Chem. Phys. 1984 , 80 , 4157 . [ 28 ] M. Wojcik , M. Tachiya , J. Chem. Phys. 2009 , 130 , 104107 . [ 29 ] M. C. Scharber , D. Mühlbacher , M. Koppe , P. Denk , C. Waldauf ,

A. J. Heeger , C. J. Brabec , Adv. Mater. 2006 , 18 , 789 . [ 30 ] D. Wehenkel , L. Koster , M. Wienk , R. Janssen , Phys. Rev. B 2012 ,

85 , 1 .

mbH & Co. KGaA, Weinheim Adv. Energy Mater. 2013, DOI: 10.1002/aenm.201200931