The Influence of Nanocrystal Aggregates on Photovoltaic Performance in Nanocrystal–Polymer Bulk Heterojunction Solar Cells

CdSe nanocrystals (NCs) can be used as an electron acceptor in solar cells, employing organic ligands to passivate their surface and make them processable from solution. The nature and abundance of impurities present after NC ligand exchange from oleic acid to n‐butylamine are identified. A further purification step using hexane as a selective solvent is described, which excludes NC aggregates from solution. The influence of NC aggregates on photovoltaic device performance is studied in a CdSe:poly[2‐methoxy‐5‐(3′,7′‐dimethyloctyloxy)‐1,4‐phenylene vinylene] (MDMO‐PPV) bulk heterojunction solar cell. The exclusion of NC aggregates leads to a four‐fold increase in device power conversion efficiency (PCE) in optimized devices. A superior blend morphology leading to improved charge generation and a better NC percolation network is identified as the main causes of this increased solar cell performance.


Introduction
Nanocrystal-polymer hybrid bulk heterojunction solar cells offer a promising way to overcome limitations of all-organic heterojunctions. In particular, the inorganic component offers a high dielectric constant, which translates into effective charge carrier screening. [ 1 ] Furthermore, the possibility of tuning the band gap by adjusting the nanocrystal (NC) dimensions offers a way to align energy levels at the bulk heterojunction. [ 2 ] In a hybrid bulk heterojunction with a polymer, the advantageous properties of NCs are complemented by the high absorption cross-section of the organic material, which allows effi cient light harvesting. [ 3 ] Because both components are solution FULL PAPER 1400139 (2 of 8) wileyonlinelibrary.com mented in optoelectronic devices, some of the identifi ed impurities may act as electron [ 15,16 ] and hole [ 14,15,17 ] trap sites. However, a detailed picture of the direct infl uence of these impurities on PV performance has not yet been provided.
Here we study the infl uence of aggregated and poorly passivated NCs on photovoltaic device performance in the model system of a bulk heterojunction solar cell consisting of colloidal CdSe NCs and poly[2-methoxy-5-(3′,7′-dimethyloctyloxy)-1,4phenylene vinylene] (MDMO-PPV). The application of an additional purifi cation steps after a post-synthetic ligand exchange treatment using hexane as a selective solvent [ 13 ] allows us to separate well-passivated NCs from aggregated and poorly ligated NCs in solution. Here, we identify and quantify the impurities and the relative ligand coverage before and after purifi cation, and show a marked improvement in PV performance after purifi cation. We investigate the differences in blend morphology, charge generation, recombination and extraction that are responsible for the improved PV performance.

Post-Synthetic Ligand Exchange and Purifi cation
The ligand exchange reaction from oleic acid (OA) to n -butylamine following standard procedures [ 7 ] is expected to produce aggregated NCs, residual native and exchanged ligands, and traces of solvents [ 11,13 ] as impurities. In order to minimize the amount of residue, we apply a post-ligand-exchange purifi cation using hexane as a selective solvent for well-passivated NCs, as proposed by Morris-Cohen et al. [ 18 ] In the remainder of this work we will refer to this procedure as hexane treatment. In contrast to the strongly binding X-type alkylphosphonates used in ref. [ 13 ] we employ NCs with weaker binding native X-type OA ligands in order to obtain a more complete removal of native ligands. Furthermore, we use n -butylamine as exchange ligand species, which is highly miscible in the precipitation solvents used in this work, compared to the relatively insoluble hexadecylamine (HDA) used in ref. [ 13 ] . We therefore do not expect any free n -butylamine to be present after precipitation. As a by-product of the ligand exchange reaction, free OA and n -butylamine may dimerise to an amide or ammonium salt species. Most of the ammonium salt should be removed from the dispersion upon fl occulation with polar solvents like methanol. Furthermore, we do not expect the for-mation of any amides, as it has been shown that the present carboxylic acid and amine species do not form amides without the aid of a catalyst at room temperature. [19][20][21] Considering the choice of ligands and purifi cation method, we anticipate the fi nal NC dispersion to only contain well-passivated, n-butylamine-ligated NCs and negligible amounts of ammonium salt. A schematic depiction of the expected products after the ligand exchange reaction and subsequent hexane treatment is shown in Figure 1 .
We fi rst use proton nuclear magnetic resonance spectroscopy ( 1 H NMR) to verify if OA is still attached to the NC or has been removed from the NCs upon ligand exchange. In Figure 2 a, the large resonance at 2.4 ppm can be accounted for by residual toluene that could not be removed from the dispersion. The signal at ≈5.35 ppm in both OA-containing samples can be identifi ed as the resonance for the olefi n double bond of OA. As expected for the sample containing OA bound onto NCs, this signal is broader and shifted compared to the sample with unbound OA. In fact, the rotational mobility of the bound ligand is reduced compared to the unattached ligand, and this leads to an observable broader resonance compared to the clear multiplet fi ne structure for free ligands. [ 22 ] The peak at 2.7 ppm in both n -butylamine samples represents the two protons in α-position to the amine functionality. The extreme broadness of the signal in the n -butylamine capped NCs sample again supports the binding of n -butylamine onto NCs. Pure n-butylamine does not have an olefi n bond in its backbone and therefore lacks an olefi nic resonance at 5.35 ppm (see Figure 2 a). Since we do not observe any signal at 5.35 ppm for the sample containing n -butylamine capped NCs we can conclude that no OA, in the guise of free OA or as an attached species, is present in the sample, i.e., the applied ligand exchange from OA to n -butylamine successfully removed the OA molecules from the NC surface.
Treating amine-ligated CdSe NC dispersions with hexane has been shown to partially remove impurities from the original NC synthesis [ 8 ] or the post-synthetic ligand exchange. [ 13 ] We note that the analyses in ref. [ 8,13 ] are based upon strong, native X-type ligands that can still cover up to 80% of the NC surface after purifi cation, thus maintaining dispersability of the NC. [ 18 ] The ligand exchange presented here, however, produces NCs without any original, X-type ligand coverage (see Figure 2 a). As a consequence, the fraction of temporarily unpassivated NC surface is greater, leading to a higher risk of NC agglomeration, due to a more likely encounter of unpassivated surface sites of NCs in close proximity. We expect, however, that the presented  post-ligand exchange hexane treatment only selects NCs that are well-covered with n-butylamine, and leaves out aggregated and poorly passivated NCs. In order to confi rm this hypothesis, we employ X-ray photoelectron spectroscopy (XPS) to analyze the nitrogen content as an indicator of the relative n -butylamine coverage. The direct comparison of nitrogen signal in the two n-butylamine containing samples is complicated by the overlapping Se LM Auger peak and N1s signal (see Figure 2 b). We therefore reconstruct the N1s signal, by subtracting the Se LM Auger peak obtained from the nitrogen-free OA sample from the superimposed signal of both n-butylamine samples. Having reconstructed the N1 signal, we compare the nitrogen content in the two n -butylamine samples by considering the nitrogen-to-cadmium ratio. As mentioned above, free ligands and dimer species are either excluded from solution during fl occulation or have only been formed in negligible amount, i.e., any nitrogen signal is indicative of attached n-butylamine. The measured greater abundance of nitrogen in the hexanetreated sample compared to its non-hexane-treated analogue therefore implies a higher relative n -butylamine coverage by ≈30%, after subtraction of the Se signal. We note that this is a lower limit, since any attenuation of the Se signal by OA compared to the shorter n -butylamine would increase the relative coverage estimate.
At this point we can conclude that the ligand exchange presented here to replace native OA ligands with n -butylamine produces NCs that are completely stripped of OA molecules. Furthermore, we identify hexane treatment as a method to enrich the NC ligand shell with n -butylamine by at least 30%. Next, we investigate the infl uence of this higher n-butylamine surface coverage on solar cell performance in the model system of a NC:polymer bulk heterojunction solar cell. Figure 3 shows the device performance of bulk-heterojunction solar cells of MDMO-PPV blended with both hexane-treated and non-hexane-treated CdSe NCs. We observe a dramatic difference in the current-voltage characteristics under AM1.5G illumination conditions, namely an increase in short-circuit current ( J SC ) by a factor of 2.9, whereas the open-circuit voltage ( V OC ) is reduced by about 50 mV. The external quantum efficiency (EQE) shows increased contributions from both polymer (400-600 nm) and NCs (600-660 nm) in solar cells incorporating hexane-treated NCs. The power conversion effi ciency (PCE) increases from 0.4% to 1.0%. Although we have achieved up to 1.7% PCE for hexane-treated NCs by optimizing the active layer thickness (see Supporting Information S3), we restrict our analysis to the two devices shown in Figure 3 , as they contain NCs from the same synthesis, and the layer thickness and absorption are very similar (see next section and Supporting Information S4), allowing a direct comparison. In the remainder of this work we study the mechanism by which the hexane treatment leads to superior device performance.

Morphology
The results of AFM and TEM analysis are shown in Figure 4 . Panels a and b show the fi lm surface of non-hexane and hexanetreated blend fi lms, as measured by AFM. A cross-sectional height profi le of the two fi lms is shown in panels c and d. While the hexane-treated fi lm appears smooth, the non-hexanetreated fi lm shows features several µm wide and >100 nm high, embedded in a continuous blend fi lm. Averaging over  the continuous fi lm thickness excluding aggregates in the AFM image (see Supporting Information S4), we fi nd that the non-hexane fi lm has an average height of 62.9 nm ± 3.9 nm, compared to 59.5 nm ± 3.9 nm for the hexane-treated fi lm. Likewise, the absorbance of both fi lms as measured using a UV-vis spectrometer is very similar (see Supporting Information S5), ruling out differences in absorption as a cause for the observed performance difference. We note, however, that blends containing hexane-treated NCs show a higher contribution from NCs (600-660 nm), but a lower contribution from the polymer (400-600 nm), which may indicate a difference in fi lm composition. Although both hexane-treated NCs and non-hexane-treated NCs containing fi lms were spun using solutions of identical NC to polymer ratio (9:1 by mass), it is possible that some aggregated NCs fell out of solution during processing, leading to a slightly higher NC loading in hexanetreated fi lms.
As mentioned before, non-hexane-treated NCs display a less passivated surface compared to the hexane-treated analogs and hence are more prone to aggregation. [ 23 ] It is therefore likely that the observed particulates in the non-hexane-treated fi lm are NC aggregates. TEM images of hexane and non-hexanetreated blend fi lms are shown in Figure 4 e-h. While individual NCs can be recognized in the hexane-treated fi lm, the nonhexane-treated fi lm shows signifi cantly larger NC aggregates. The size and number of NC aggregates on the surface of nonhexane-treated fi lms implies a signifi cantly lower NC loading in the underlying blend. As a consequence, the NC percolation network within the blend will be sparser. [ 7 ] Indeed, Figure 4 e,f reveal only short-range connectivity of the NC network, in contrast to well-connected NC domains for hexane-treated samples. Furthermore, larger polymer domain sizes are recognizable in samples containing non-hexane-treated NCs whereas a fi ner and more intermixed morphology is observable in their hexane-treated analogs. The better connected NC network and the smaller polymer domains in samples containing hexanetreated NCs is likely to have a benefi cial impact on charge carrier pathways and may be the origin for the observed higher photocurrent. In order to investigate charge extraction and recombination we next study the transient photocurrent and transient photovoltage decay in both devices.

Charge Evolution on the Microsecond Time Scale
Charge carrier extraction and recombination dynamics on the microsecond timescale were studied by investigating the transient photocurrent (TPC), where the current is monitored after the light is turned off, and transient photovoltage (TPV), where the voltage decay is measured following a small perturbation light pulse under variable white light bias.
The decay of the photocurrent yields a measure of the charge extraction timescales in photovoltaic devices. In Figure 5 a the photocurrent decay of hexane and non-hexane-treated devices after a 2 ms light pulse (λ = 500 nm) is shown, in absence of white light bias. We note that a measurement artefact appears at the onset of the TPC transient, the origin of which is explained in the Supporting Information S6. The hexane-treated device clearly shows faster current decay, especially at short times (<1 µs). We attribute this to an improved charge extraction, as recombination in these devices occurs on longer timescales, as probed by TPV and discussed below. The observed improvement in connectivity of the NC network is likely to improve charge extraction. [ 7,24 ] Since there is no current fl owing under open-circuit conditions, the decay of the photovoltage perturbation Δ V at V OC provides a direct measure of the recombination of photogenerated charge carriers. [ 25 ] Devices containing hexane-treated NCs show faster photovoltage decay than their non-hexane-treated analogs over a wide range of white light biases. The evolution of τ , defi ned as the time it takes for the additional photovoltage Δ V to decay to 1/e of its initial value, with increasing white light bias is shown in Figure 5 b. We note that recombination is much slower than typical TPC decay timescales, indicating that the TPC decay is indeed dominated by extraction rather than recombination.
Nelson, [ 26 ] Heinemann et al. [ 27 ] and Rauh et al. [ 28 ] have proposed the presence of recombination that is limited by the  de-trapping of one carrier. This could be the case for a trapped charge that is spatially removed from the heterojunction interface, for instance in a cluster of NCs. [ 26 ] In hexane-treated samples, the fi ner morphology could prevent trapping in NC clusters, which explains the faster TPV decay observed. An increase in recombination rate is likely to lead to a reduction in V OC . [ 29 ] However, only a modest reduction of 50 mV is observed under AM1.5G illumination (see Figure 3 ). This supports our earlier observation that charge extraction can successfully outcompete recombination under operating conditions. We note that the recombination rate is likely to be density dependent, [ 25 ] such that an improvement in charge generation will also reduce the value of τ . [ 35 ] Next we employ transient absorption (TA) spectroscopy in order to investigate charge carrier generation in the nanosecond region.   Figure 6 shows the near-IR TA spectra of hexane and nonhexane-treated devices, excited at 532 nm at a pump fl uence of 1.6 µJ and measured 3 ns after the pump pulse. At 3 ns, neither non-geminate recombination nor charge extraction is expected to play a signifi cant role, especially at the relatively low excitation fl uence used. Hexane-treated devices were found to show a much larger photoinduced absorption than non-hexane-treated devices at the same excitation density. MDMO-PPV polarons are well known to have an absorption peak between 860 nm and 920 nm. [ 9,30,31 ] We thus attribute the negative TA signal to the photoinduced absorption of the hole polaron residing on the polymer. Contributions from NC trapped or free charges have been shown to be more than an order of magnitude weaker than the polaronic contribution and thus can be neglected. [ 9 ] We fi nd that hexane-treated devices show a superior charge carrier generation by a factor of 2.2 compared to their nonhexane-treated counterparts (see Figure 6 ). This marked difference in charge generation indicates that the higher J SC of hexane-treated devices is mainly due to improved charge generation. Time-resolved photoluminescence measurements (see Supporting Information S7) indicate similar exciton quenching rates for devices consisting of hexane and non-hexane-treated NCs. This was further confi rmed by similar photoluminescence quantum effi ciencies for both blends. We show elsewhere that excitons generated in the polymer domain of CdSe:MDMO-PPV blends are quenched effi ciently by either Förster resonance energy transfer or electron transfer to the nanocrystal. [ 9 ] Furthermore, we identifi ed that only an initial charge transfer from the polymer to the NC domain is capable of producing free charges effi ciently. Excitons quenched by energy transfer have to undergo a subsequent hole back transfer in order to create free charges. This second process takes place on timescales on which competing recombination mechanisms occur, rendering this mechanism ineffi cient for the generation of free charges. [ 9 ] We note that charge and energy-transfer processes operate on different length scales. While charge transfer relies on a close spatial proximity between donor and acceptor, [ 32,33 ] energy transfer is considered a longer-range interaction. [ 9,34 ] Considering the unchanged absorption in both devices (see Supporting Information S5) and the fi ner blend morphology in devices consisting of hexane-treated NCs (see Figure 4 ) we attribute the improvement Adv. Energy Mater. 2014, 4, 1400139   Figure 6. Transient absorption spectra for hexane-treated (red) and non hexane-treated (black) devices measured at 3 ns at an excitation fl uence of 1.6 µJ cm -2 . in charge generation to a larger fraction of excitons on the polymer being quenched by charge transfer. With energy transfer operating over larger length scales than electron transfer, only excitons generated in close proximity to the interface will lead to electron transfer. As we identifi ed larger polymer domains for devices consisting of non-hexane-treated NCs (see Figure 4 ), excitons generated within the polymer domain will on average be further from the interface than in their hexane-treated analogs. As a consequence, exciton quenching will be dominated by energy transfer, making charge generation in devices containing non-hexane-treated NCs ineffi cient. We note that we cannot monitor charge separation within the CdSe-NC network, as this population would not contribute to the polaronic signal measured here.

Excitons and Free Charges on the Nanosecond Time Scale
We therefore conclude that the reduced polymer domain size in devices consisting of hexane-treated NCs most likely affects the balance between energy and electron transfer yields. As only electron transfer from the polymer to the NC domain leads to effi cient free charge generation, we propose that a higher electron transfer yield in devices containing hexane-treated NCs leads to superior charge generation. This improvement in charge generation is the main cause of the marked improvement in J SC , with improved carrier extraction contributing to a lesser degree.

Conclusion
We have presented a CdSe NC ligand exchange method from long and insulating oleic acid capped to shorter and more conductive n -butylamine capped NCs. In contrast to known ligand exchange strategies our approach is capable of removing all native OA molecules from the NC surface. Furthermore, we found that using hexane as a selective solvent in a post-synthetic purifi cation procedure allowed us to expel aggregated and poorly passivated CdSe NCs. To study the effects of impurities and aggregates on charge dynamics, bulk heterojunction solar cells were prepared using either purifi ed or non-purifi ed NCs as the electron acceptor and MDMO-PPV as the donor material. Devices using purifi ed NCs were found to show a signifi cantly improved PV performance. The reduction of NC aggregates in this system resulted in a fi ner morphology and better longrange connectivity of the NC network, which in turn leads to more than twice the charge generation yield in hexane-treated devices. Aided by improved charge extraction, this leads to a dramatic increase in PV performance, yielding a three-fold increase in short-circuit current and a two-fold increase in PCE compared to non-hexane-treated devices.

Experimental Section
Synthesis of CdSe Nanocrystals (NCs) : NCs were synthesized using a modifi cation of previously reported methods. [ 35 ] All chemicals were purchased from Sigma Aldrich, if not stated otherwise, and were anhydrous if available.
Briefl y, CdO (0.748 g, 5.8 mmol), oleic acid (28.7 mL, 90.6 mmol) and 1-octadecene (ODE, 19 mL, 59.4 mmol) were degassed in a three neck fl ask under vacuum (10 −2 mbar or better) at 110 °C for 2 h. For formation of the Cd-oleate complex the fl ask was fl ushed with N 2 gas and the temperature was increased to 230 °C. Water residues present from formation of the Cd 0 -oleate complex were removed by degassing again under vacuum (10 −2 mbar or better) at 110 °C for 1 h. Subsequently, the fl ask was set under N 2 atmosphere and heated to 230 °C. Two identical Se precursor solutions, containing Se (Alfa Aeser, 0.345 mL, 4.4 mmol) dissolved in trioctylphosphine (TOP, 4.6 mLl, 10.0 mmol), were prepared in a nitrogen fi lled glove box. The fi rst Se precursor was injected rapidly into the Cd precursor at 230 °C under nitrogen atmosphere to initiate crystal nucleation. In order to replenish the depleted Se content in the reaction fl ask the second Se precursor was injected after 15 min and the solution was kept at 230 °C for another 45 min. To quench crystal growth, the fl ask was placed into a water bath and 5 ml methanol was injected. The NCs formed were isolated from the reaction mixture by fl occulating to turbidity using 2-propanol and methanol. After redispersion in toluene and a second precipitation the NCs were fi nally dispersed in toluene and fi ltered through a PTFE fi lter (pore size 0.2 µm). The synthesis yielded particles with a diameter of 5.9 nm ± 1.3 nm (see Supporting Information S2).
Butylamine Ligand Exchange : Nanocrystals (600 mg) were transferred into a nitrogen fi lled glovebox and dispersed in anhydrous toluene to yield a concentration of 50 mg mL −1 . n -butylamine (3 mL, 30 mmol) was added and stirred for 11 h at room temperature. The exchange solution was fl occulated with a mixture of 2-propanol and methanol, and the NCs were re-dispersed in chlorobenzene and fi ltered through a syringe fi lter with a pore size of 0.2 µm to yield a fi nal concentration of 50 mg mL −1 . [ 7 ] As will be shown later, these samples contained a large density of aggregates.
For samples containing a reduced density of dispersed aggregates, the exchanged NCs were precipitated with a mixture of 2-propanol and methanol and re-dispersed in hexane. The dispersion was centrifuged in order to exclude aggregated and poorly passivated NCs. NCs that remained in the supernatant were precipitated using a mixture of 2-propanol and methanol, and were fi nally redispersed in chlorobenzene. Prior to device fabrication, the NCs were fi ltered through a syringe fi lter with a pore size of 0.2 µm to yield a concentration of 50 mg mL −1 .
Determination of NC Concentration : NC mass concentrations were measured gravimetrically by drying of 100 µL solution on an electronic scale.
Absorbance Measurements : For absorbance measurements, fi lms were spin-cast on Spectrosil quartz substrates. The spectra were collected using a Hewlett-Packard 8453 UV-vis spectrometer.
Atomic Force Microscopy (AFM) : Samples identical to those fabricated for absorbance spectra were prepared for AFM analysis and annealed at 140 °C for 10 min in a N 2 atmosphere. Imaging was performed using a Veeco Dimension 3100, operated in tapping mode.
Proton Nuclear Magnetic Resonance Spectroscopy ( 1 H NMR) : NC samples were diluted to a concentration of ≈2 mg mL −1 and dispersed in d-chloroform. NMR spectra were recorded using a Bruker AVC-500 spectrometer. δH values are reported relative to the internal standard CDCl 3 (δH = 7.26 ppm).
Transmission Electron Microscopy (TEM) : Pristine QDs were diluted to a concentration of ≈1 mg mL −1 in chlorobenzene and drop cast on a TEM Grid (200 Mesh Cu, Agar Scientifi c) in a nitrogen-fi lled glove box. As an independent control, NC:polymer blends were spun onto PEDOT:PSS (Heraeus Clevios) coated indium tin oxide (ITO) patterned glass substrates and annealed at 140 °C for 10 min in a N 2 atmosphere. Subsequently, the blend fi lm was transferred to a TEM grid by a lift-off technique. TEM images were taken on a FEI Philips Tecnai 20 at an electron acceleration of 200 keV in bright fi eld mode.
X-Ray Photoemission Spectroscopy (XPS) : XPS samples were prepared by spin-coating from 50 mg mL −1 solution of NCs in chlorobenzene onto silicon wafer substrates. The samples were then transferred to an ultrahigh vacuum (UHV) chamber (ESCALAB 250Xi) for XPS measurements. XPS measurements were carried out using a XR6 monochromated AlKα X-ray source (hν = 1486.6 eV) with a 900 µm spot size and 20 eV pass energy.
Preparation and Characterization of Solar Cells : Solar cells were prepared on ITO patterned glass substrates, cleaned in an ultrasonic bath with ethanol, acetone and isopropanol, respectively. The substrates were then treated with oxygen plasma for 10 min at 250 W, after which they were coated with a 50 nm layer of PEDOT:PSS and dried under N 2 fl ow at 230 °C for 30 min. The active layer was spin-coated from a 17.5 mg mL −1 blend solution of MDMO-PPV and CdSe NCs in chlorobenzene, in a 9:1 (NC:PPV) ratio by weight. The devices were then annealed under N 2 at 140 °C for 10 min, before thermally evaporating a 60 nm aluminum electrode under vacuum.
For external quantum effi ciency (EQE) measurements, a 100 W tungsten halogen lamp dispersed through a monochromator was used as the light source and a Keithley 2635 source measure unit (SMU) was used to measure the short-circuit current at various wavelengths. Current-voltage characteristics were measured under AM 1.5G equivalent conditions using an Abet Sun 2000 solar simulator, at an intensity equivalent to 100 mW cm −2 , after correcting for spectral mismatch. Both the dark and light current-voltage characteristics were measured using the Keithley 2635 SMU.
Transient Photovoltage Decay and Transient Photocurrent Decay : For the transient photovoltage and photocurrent measurements, a highbrightness 525 nm green LED (Kingbright, L-7104VGC-H) was used as the light source and a Hewlett Packard (HP) 8116A function generator was used as the power supply for the LED. A ring of six white highbrightness LEDs was used for constant background illumination and a set of lenses was used to focus the incident light (both from the pulse light source and background light source) onto a single pixel of the device. The intensity of the background illumination was calibrated by comparing the short-circuit current density ( J SC ) with the values previously obtained from the solar simulator. Open-circuit voltage transients were recorded by connecting the device under test in series with an Agilent DSO6052A digital oscilloscope with input impedance of 1 MΩ. Short-circuit current transients were recorded by connecting the device under test in series with an Agilent DSO6052A digital oscilloscope with input impedance of 50 Ω.
Transient Absorption Spectroscopy (TA) : In time-resolved transient absorption spectroscopy, the change of absorption after an initial excitation is monitored as a function of time. Here we use a frequencydoubled Q-switched Nd:YVO 4 laser (AOT-YVO-25QSPX, Advanced Optical Technologies) as the excitation source. A narrowband 1 kHz repetition rate Ti:Sapphire amplifi er system (Spectra-Physics, 90 fs pulse width) was used to power a home-built non-parametric amplifi er. The resulting broad band near IR pulse (850-1050 nm) of 300 fs width was used as the probe beam. The probe beam was subsequently split in two, with the second "reference" beam probing a region of the device that is not excited. In this way the fl uctuations of the laser system can be compensated and the overall noise reduced. Pump and probe beam are overlapped on the device. Both probe and reference beam were subsequently dispersed into a Andor Shamrock SR-303i spectrometer and detected using a pair of 16-bit 1024-pixel linear image sensors (Hamamatsu, S8381-1024Q), which were driven and read out by a custom-built board from Stresing Entwicklunsbuero.
The polaronic signal of the hole charge on MDMO-PPV was measured by exciting the sample at 532 nm at a pump fl uence of 1.6 µJ cm −2 and probing in the near IR (850-1050 nm).

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.