semiconductors for enhanced electron extraction from solution processed solar cells using alkali metals

To improve charge carrier injection into or extraction from organic optoelectronic devices, electrically doped layers are often employed. Whereas n-doping of organic semiconductors has been widely used in vacuum processed optoelectronic devices, adequate solution processes to enable future device printing are underdeveloped. In this work, we study n-doping of 1,3,5-tris(1-phenyl-1H-benzimidazol-2-yl) benzene (TPBi) in a solution process, using sodium as the electron donor. Upon addition of elementary sodium to a clear TPBi/toluene solution, we observed a change in color, indicating charge carrier transfer between sodium and TPBi. The optical and electrical properties of doped and undoped TPBi were characterized in solution and in the corresponding thin-films. Electron Paramagnetic Resonance (EPR) measurements revealed an increase of the number of unpaired spins upon doping, indicating the presence of doping-induced charge carriers. Implementing TPBi : Na as electron extraction layers in organic solar cells, we found almost the same device performance as compared to state-of-the-art solar cells comprising zinc oxide electron extraction layers.

In light of the prospect of employing printing and coating techniques for future large-area roll-to-roll device fabrication, electrical doping of organic layers will have to be achieved through solution processing. 13Whereas p-doping of molecular materials and polymers from solution has been reported in the literature before, 14 n-doping of organic molecules using solution-based processes appears more challenging due to the non-solubility of alkali metals in most processing agents.Rare examples of n-doping from solution utilize organometallic dimers or dimers of benzimidazoline radicals. 15,16Alkali metals are soluble in ammonia at a process temperature of À78 C, not being suitable for any relevant deposition processes.
In this work, we dissolved sodium in 1,3,5-tris(1-phenyl-1Hbenzimidazol-2-yl)benzene (TPBi)/toluene solutions at room temperature, enabling the solution deposition of n-doped organic layers.We investigated the n-doped and undoped TPBi solutions by UV-Vis absorption spectroscopy and studied the electrical properties of the respective TPBi : Na bulks by Electron Paramagnetic Resonance (EPR) and conductivity measurements.The n-doped layers were then incorporated into organic solar cells, replacing the widely employed zinc oxide (ZnO) electron extraction layers.

Experimental
Toluene (Sigma-Aldrich, anhydrous, 99.8%) was dried over sodium (Na) to remove water and oxygen.The dopant sodium (Chempur, in oil, 99.95%) was cut under inert atmosphere and cleaned from residual mineral oil.TPBi (Syntec-Sensient, 99.95%) was dissolved (10 g L À1 ) in dry toluene, and the dopant sodium was added to the TPBi/toluene solution.Aer stirring the solution overnight, the solid sodium surplus was removed.
Absorbance spectra of the solutions were measured using a Lambda 1050 UV-Vis-NIR spectrophotometer (Perkin Elmer).
Nuclear magnetic resonance (NMR) spectra were measured by a Bruker Avance III Microbay 400 MHz.Therefore, toluene-D8 solutions of TPBi : Na or, for reference, neat TPBi were lled into NMR tubes under inert atmosphere and sealed using a blowtorch.
Room temperature continuous wave electron paramagnetic resonance (cwEPR) spectra were recorded on a laboratory-built X-band EPR spectrometer to detect unpaired spins in TPBi : Na lms.The respective solutions were lled in EPR tubes, the solvent was removed in vacuo, the tubes were lled with inert gas and sealed using a blowtorch.To control the magnetic eld, a eld controller (Bruker BH15) was used.Microwaves were generated and monitored by a microwave bridge (Bruker ER 048 R).Magnetic-eld modulation in combination with lock-in detection was employed using a lock-in amplier (Stanford Research SR810) and a modulation amplier (Wangine WPA-120), resulting in derivative spectra.To determine the quality factor Q of the sample in a Bruker ER 4122 SHQ microwave resonator, a mode picture was measured before each measurement started.The measured EPR signal was double integrated and compared to a 4-hydroxy-TEMPO-reference sample with a known number of spins.This comparison yielded the absolute number of spins. 17The uncertainty with respect to the absolute number of unpaired spins is less than 20%, which is estimated based on the uncertainty of the sample volume when lling the tubes.
Thin-lm conductivities s were measured in electron-only devices comprising a spin cast 30 nm thick undoped TPBi or doped TPBi : Na layer, respectively, sandwiched between two thermally evaporated aluminum electrodes.The conductivities were calculated from the linear regime of the current densityvoltage (J-V) curves at 0.04 V using Ohm's law.
Amplitude Modulation Kelvin Probe Force Microscopy (AM-KPFM) dual pass experiments were carried out in tapping mode on a Bruker Dimension ICON using a Pt/Ir coated FMV-PT tip to measure the contact potential difference CPD ¼ À(F tip À F sample )/e, F < 0. The CPD was referenced with the work function of highly oriented pyrolytic graphite F HOPG ¼ À4.4 eV. 18olar cells were built on ITO coated glass substrates (R , ¼ 13 U sq À1 ) according to the device architecture depicted in Fig. 1a.The substrates were successively cleaned in acetone and 2-propanol in an ultrasonic bath and treated with oxygen plasma.Aerwards, the samples were transferred into a glovebox and kept under nitrogen atmosphere for the remaining fabrication and characterization process.Electron extraction layers (EELs) from either undoped or sodium doped TPBi (Fig. 1b) or, for reference, ZnO were employed.Therefore, 30 nm thick lms of sodium doped or undoped TPBi were spin cast from toluene solution (10 g L À1 , 1000 rpm, 30 s, 4000 rpm, 3 s).For reference, 30 nm thick lms of ZnO were spin cast from nanoparticle dispersion (Nanograde N-10, Nanograde Ltd., 1 wt% in isopropanol, 4000 rpm, 30 s) and thermally annealed on a hotplate (85 C, 10 min 1c) and [6,6]-phenyl C 61 -butyric acid methyl ester (PC 61 BM, Solenne, 99%, Fig. 1c) were dissolved (1 : 1.5, w : w, 23 g L À1 ) in o-xylene (Sigma-Aldrich, anhydrous, 97%) and stirred overnight at 85 C. The warm solution was spin cast (90 nm, 1500 rpm, 60 s) on top of the electron extraction layer.The molybdenum oxide/silver counter electrode (MoO x /Ag, 10 nm/100 nm) was thermally evaporated through a shadow mask in an evaporation chamber (10 À6 mbar) attached to the glovebox, dening the photo-active area of the solar cell (3 Â 3.5 mm 2 ).
Current density-voltage curves were recorded with a source measure unit (Keithley 238) under illumination from a spectrally monitored solar simulator (Oriel 300 W, 1000 W m À2 , ASTM AM 1.5G), calibrated by a KG5 ltered silicon reference cell (91150-KG5, Newport).The sample edges were masked to avoid light-incoupling through substrate modes.
Film thicknesses were measured by a tactile stylus surface proler (Bruker Dektak XT).
Time-of-ight secondary ion mass spectrometry was performed on a TOF-SIMS 5 instrument (ION-TOF GmbH, Münster, Germany), equipped with a Bi cluster liquid metal primary ion source and a non-linear time-of-ight analyser.The Bi source was operated in "bunched" mode providing 1 ns Bi 3 + ion pulses at 25 keV energy, an analysed area of 100 Â 100 mm 2 and a lateral resolution of approx.4 mm.Negative polarity spectra were calibrated on the C À , C 2 À , C 3 À , and C 4 À peaks.Positive polarity spectra were calibrated on the C + , CH + , CH 2 + , and CH 3 + peaks.Sputter depth proles were performed using a 1 keV Cs + ion beam and a raster size of 500 Â 500 mm 2 .

Results and discussion
In order to avoid losses in the open-circuit voltage (V oc ) of organic solar cells, as a rule of thumb, EELs with Fermi energies higher than or equal to the electron affinity of the fullerene acceptor have to be chosen.TPBi complies with this requirement, having a LUMO energy of about À2.8 eV 19 and, upon ndoping, a fermi level energy close to the LUMO energy.With an ionization energy of about 2.75 eV, 20 3s valence electrons of sodium can transfer to TPBi, forming electron excess on the TPBi matrix.Whereas alkali metal doping is widely used in vacuum deposited organic semiconductor devices, its applicability by wet processing is widely unexplored, since neat alkali metals do not dissolve in common processing agents.However, when adding a piece of sodium to a 10 g L À1 solution of TPBi in toluene, the clear TPBi/toluene solution turned red (Fig. 2, inset).The respective UV-Vis absorbance spectrum depicted in Fig. 2 shows additional absorption bands that hint towards new polaronic states in the reduced TPBi. 21Notably, the reduction of the TPBi molecules is reversible: aer exposing the TPBi À /Na + / toluene solution to air, the solution turns clear again.
To explore changes of the molecular conformation aer reducing the TPBi molecules, we recorded NMR spectra of the TPBi/toluene and the TPBi À /Na + /toluene solution as shown in Fig. 3. Upon doping, we observed a broadening of all peaks, which can be attributed to an oligomerization of the TPBi molecules, 22 initiated by the formation of radicals aer electron transfer from sodium to TPBi.We note that the oligomerization of TPBi in solution does not affect the principle electron excess on the TPBi molecules.
Aer the analysis of the TPBi À /Na + /toluene solution, we investigated the effect of doping on the solid TPBi : Na bulk and, for reference, neat TPBi samples.Therefore we lled 50 ml of the TPBi À /Na + /toluene or TPBi/toluene solution in EPR tubes, removed the solvent under vacuum and sealed the tubes.In order to investigate if the reduced TPBi À molecules prevailed, we measured the number of unpaired electrons within the dry thin-lms by EPR.The Zeeman interaction between the magnetic momentums associated with unpaired electron spins and an external magnetic eld causes a splitting between two states with different magnetic spin quantum number . Transitions between the two energy levels can be induced by resonant microwave absorption.In typical cwEPR experiments, eld modulation in combination with lock-in detection is used, which leads to the typical derivative EPR line shape.The quality factor Q of the resonator loaded with the sample was determined from the mode picture before each EPR Fig. 2 Normalized UV-Vis absorbance spectra of neat TPBi/toluene and TPBi À /Na + /toluene solutions.Upon addition of sodium, absorption bands between 350 and 650 nm occur, indicating the formation of new polaronic states through transfer of the 3s electrons from sodium to TPBi.
Fig. 3 NMR spectra of neat TPBi and TPBi : Na solutions in toluene-D8.The spectrum of undoped TPBi exhibits sharp peaks whereas the peaks of the doped TPBi are much broader.We attribute this broadening to an oligomerization of TPBi molecules induced by radical formation due to the electron transfer from sodium to TPBi.For better visibility, the spectra have been plotted using arbitrary units.
This journal is © The Royal Society of Chemistry 2016 measurement was started.The measured derivative spectra were background-corrected and double-integrated to obtain the integrated EPR signal amplitude.As depicted in Fig. 4, layers of undoped TPBi show weak background signatures of unpaired electrons.However, for thin-lms deposited from TPBi À /Na + / toluene solution, a signicantly stronger EPR signal can be detected, indicating the presence of doping-induced electrons in the TPBi layer.The absolute number of spins per sample was calculated by double-integration of the EPR signal and comparison to a 4-hydroxy-TEMPO-reference sample with a known number of spins.Spin concentrations were calculated by dividing the total amount of measured spins by the sample volume, and the doping efficiencies were determined by neglecting the number of sodium atoms and considering the fraction of n-doped TPBi molecules.The concentration of unpaired spins increases from 1 Â 10 16 cm À3 for undoped TPBi by more than one order of magnitude to 4 Â 10 17 cm À3 for TPBi : Na.This spin concentration corresponds to 1 unpaired electron per 5000 TPBi molecules.The actual number of electrons transferred from Na to TPBi is higher, with the electrons from TPBi radicals that were oligomerized, not being visible in EPR measurements.
To investigate the effect of Na-doping on the physical properties of TPBi thin-lms, we sandwiched 30 nm thick lms between two aluminum electrodes and measured their J-V characteristics.The lms were spin cast from either the TPBi or the TPBi À /Na + solutions.From the J-V curves at 0.04 V, we estimated the conductivity of undoped lms on the order of 10 À9 S cm À1 whereas the conductivity improved by two orders of magnitude to 10 À7 S cm À1 in doped lms.
Ideally, n-doping of TPBi does not only enhance the conductivity of lms but also improves the extraction (solar cells) or injection (OLEDs) of electrons.Therefore, we probed the work functions of TPBi and TPBi : Na thin-lms by KPFM.Upon Na-doping, the work function shied from À4.40 eV We exemplify the merits from solution processed TPBi : Na EELs on organic solar cells featuring a glass/ITO/EEL/PTB7-Th:PC 61 BM/MoO x /Ag device architecture as depicted in Fig. 1a, employing a 30 nm EEL from either neat (undoped) TPBi, sodium doped (n-doped) TPBi or, for reference, nanoparticulate ZnO that allows for comparison with the literature.The 90 nm thick PTB7-Th:PC 61 BM photo-active layer was deposited from oxylene, intentionally omitting processing additives such as 1,8diiodooctane or p-anisaldehyde that would chemically react with the sodium in the underlying EEL.
The integrity of the layer stacks and in particular the presence of an intact TPBi : Na EEL was probed by ToF-SIMS on the same devices used for J-V measurements.To simultaneously investigate sodium diffusion within the solar cells, we deliberately stored the devices for six months aer fabrication and J-V characterization.During ToF-SIMS measurements, the layers were successively ablated with a Cs + ion beam while positively and negatively charged ions were detected.The ToF-SIMS signal and the corresponding layers of the solar cell are depicted in Fig. 5. Starting the analysis from the Ag top electrode (CsAg + ), we sequentially identied the MoO x layer (CsMo + ), the photo-active PTB7-Th:PC 61 BM layer (CS À ), the TPBi : Na EEL (C 2 N 2 À ) and the ITO bottom electrode (In 2 O 2 À ).The CsNa + signal provided information about the sodium content throughout the layer  stack.For better readability, all curves were normalized to their maxima, leading to a somewhat higher (articial) background noise of the C 2 N 2 À signal.Not only did we nd the sodium peak coinciding with the TPBi layer, indicating an intact TPBi : Na EEL.We also found that sodium did not diffuse within the device, which is an oen observed process that is detrimental to the device performance. 23We carried out the same measurement for reference devices comprising undoped TPBi layers and found the TPBi layer preserved at its location in the layer stack, too (data not shown here).Even without additives, the reference PTB7-Th:PC 61 BM cells with ZnO EELs yielded hero efficiencies of almost 7%, thereby outperforming additive-free, literature-known solar cells. 24,25nO is an efficient EEL with a work function F ZnO ¼ 4.1 eV which is oen employed for electron extraction in organic solar cells due to its excellent electronic interface formation with PC 61 BM.With F ZnO ¼ 4.1 eV being close to the LUMO of PC 61 BM, typically, the V oc ¼ 812 mV does not suffer any losses but rather achieves the maximum possible value.Fig. 6 depicts the corresponding current density-voltage curves.Their key performance data open-circuit voltage (V oc ), short-circuit current density (J sc ), ll factor (FF) and power conversion efficiency (PCE) are summarized in Table 1.
Under reverse bias and hence upon eld-assisted charge carrier extraction, all solar cells exhibit the same saturation current density, which indicates about equal photo-generation of charge carriers in all three devices.However, under shortcircuit conditions, the photo current of the solar cell comprising an undoped TPBi layer is reduced as TPBi forms a barrier for electrons that hampers their collection by the cathode and negatively affects the FF ¼ 28%.At the same time, the reduced built-in eld affects the V oc ¼ 290 mV.Altogether, the solar cells with TPBi EEL yield a PCE ¼ 0.7% only.Upon n-doping of TPBi with sodium, the extraction barrier is effectively reduced almost reinstalling the full V oc ¼ 777 mV and a FF ¼ 54%, yielding a hero PCE ¼ 6.1% which is only 10% (rel.)below the performance of the ZnO reference device.
We note that, in preliminary experiments, UV-Vis absorption measurements and solar cell device data hinted towards similar physical properties of other semiconductor/alkali metal combinations, such as potassium doped 2-(4-tert-butylphenyl)-5-(4-biphenylyl)-1,3,4-oxadiazole (BuPBD : K) or lithium doped bathophenanthroline (BPhen : Li).This indicates that doping of organic semiconductors with alkali metals from solution is a more general concept that can be adopted for a variety of material combinations.

Conclusions
n-Doping can be achieved employing wet processes, when solid alkali metals (here: sodium) are added to an organic semiconductor solution.The electrical doping effect endures the thin-lm deposition processes under inert atmosphere.They can be incorporated into organic solar cells where they enhance electron extraction.In contrast to ZnO, the doped organic layers do not require thermal annealing.Beyond solar cells, this n-doping process may be a versatile asset of the organic semiconductor toolbox for future low-temperature solution fabrication of other organic optoelectronic devices such as OLED or thermoelectric generators.

Fig. 1
Fig. 1 (a) Device architecture of the PTB7-Th:PC 61 BM solar cells comprising different electron extraction layers (EELs).Chemical structures of (b) TPBi and (c) the bulk-heterojunction components PTB7-Th and PC 61 BM.

(
undoped TPBi) to À3.55 eV, indicating a Fermi level shi towards the TPBi LUMO.As a consequence, in solar cells, photogenerated electrons on PC 61 BM face a substantial extraction barrier towards neat TPBi, hampering the extraction of charges.Likewise, in OLEDs, undoped TPBi will form a high injection barrier with the cathode.Upon shiing the Fermi level towards the TPBi LUMO, both barriers are reduced, enabling efficient charge carrier extraction or injection, respectively.

Fig. 4
Fig. 4 EPR spectra of neat TPBi and TPBi : Na layers.Whereas undoped TPBi exhibits only weak EPR signals, TPBi : Na layers show a strong signature of unpaired spins which can be attributed to doping-induced electrons and hence to n-doping of the TPBi.

Fig. 5
Fig. 5 ToF-SIMS measurements on a typical solar cell comprising a TPBi : Na layer.Ag (CsAg + ), MoO x (CsMo + ) and Na (CsNa + ) were detected in the positive polarity spectrum.PTB7-Th:PC 61 BM (CS À ), TPBi (C 2 N 2 À ) and ITO (In 2 O 2 À ) were detected in the negative polarity spectrum.The co-location of TPBi and Na proves the integrity of the TPBi : Na electron extraction layer between the photo-active PTB7-Th:PC 61 BM layer and the ITO cathode.