Origin of the p-type characteristics of organic semiconductors


 Organic semiconductors (OSC) are generally considered intrinsic (undoped), an assumption which underpins our understanding of the charge transport in this promising class of materials. However, this premise conflicts with a variety of experimental observations, that suggest the presence of excess holes carriers in OSCs at room temperature. Here, using a low-power plasma de-doping method, we report that trace amounts (~1015 cm-3) of oxygen-induced organic radical cations (OIORCs) are inherent in the lattice of OSCs as innate hole carriers, and that this is the origin of the p-type characteristics exhibited by the majority of these materials. This finding clarifies previously unexplained organic electronics phenomena and provides a foundation upon which to re-understand charge transport in OSCs. Furthermore, the de-doping method can eliminate the trace OIORCs, resulting in the complete disappearance of p-type behavior, while re-doping (under light irradiation in O2), reverses the process. These methods can precisely modulate key electronic characteristics (e.g., conductivity, polarity, and threshold voltage) in a nondestructive way, expanding the explorable charge transport property space for all known OSC materials. Accordingly, we conclude that our tailorable OIORC doping strategy, requiring only off-the-shelf equipment and a glovebox, will become a core technology in the burgeoning organic electronics industry.

Read Full License mismatched energy levels at the metal-semiconductor junctions 20,21 , and the strong electron trapping ability of the acceptor-like localized states of insulator interfaces 22 and oxygen/water 23 .
These interpretations are based on the unconfirmed premise that OSCs are intrinsic (undoped).
However, despite great experimental efforts, very few OSCs have achieved a balanced ambipolarity 1,24 . Another perplexing fact is that OSCs generally exhibit hole carrier densities in the region of 10 12 -10 16 cm -3 6,10 and conductivities of 10 -8 ~10 -4 S·cm -1 2,9 , much higher than what is expected from thermally-activated forbidden transitions at room temperature (the typical bandgap of OSCs of 2-3 eV should yield negligible intrinsic carrier concentrations considering = � (− 2 ) ) 25 . Furthermore, depletion mode transistors are frequently observed 4,5 . Some spectral investigations suggested that the Fermi level is close to the HOMO in OSCs 7,8 , which is similar to the observation of inorganic semiconductors doped by some acceptor impurities. These striking discrepancies suggests the presence of excess hole carriers and imply that previous studies have not revealed the true origin of OSC p-type characteristics, which must originate from acceptor impurities that cause these dominant hole carriers.
Here, we demonstrate that trace oxygen, innately incorporated in the OSC lattice as an acceptor, induces the geminate superoxide anion (SA) and organic radical cation (ORC), and fundamentally endows OSCs with p-type characteristics. This has not been revealed until now due to the lack of a de-doping method to remove the lattice-incorporated trace oxygen (for details, see Supplementary Section 1). Moreover, because, in previous reports, OSCs still showed p-type transporting behavior after the removal of the adsorbed oxygen by vacuum annealing or sublimation 26,27 , these characteristics were considered to be inherent, with oxygen doping merely modulating the p-type transport [26][27][28][29][30] . This is the cause of some perplexing observations and the fundamental reason why some classical physical models used in the field are invalid for organic devices. We developed a dedoping method (using a low-power plasma) to eliminate ORC and SA induced by the innate trace oxygen, and a re-doping method (light irradiation in oxygen) to re-generate them, by which OSCs can be switched between innately doped (p-type) and intrinsic (non-conductive) states reversibly (Fig. 1a), revealing the decisive role of the innate trace oxygen in charge transport. An organic field-effect transistor (OFET, Fig. 1a Despite the drastic change in electrical characteristics, the OFET was not damaged by the lowpower plasma. When 1 atm of oxygen was filled in the system for 5 min, p-type characteristics of the OFET recovered slightly (Fig. 1c). Upon further extending the time of exposure to oxygen, the rate of recovery remained quite slow. However, when the OFET was illuminated by a full-spectrum Xe lamp in an oxygen atmosphere, its p-type characteristics rapidly recovered, almost to the original level after 3 h (output characteristics of this process are shown in Supplementary Section 5). The process is reversible, suggesting a negligible impact of oxygen (de)intercalation on the ordered lattice; a remarkable result given the general difficulty in doping both inorganic and organic semiconductors without affecting their structure. These results also suggest that the plasma treatment is nondestructive for OSCs at the molecular and aggregate levels, and this was verified by X-ray diffraction, atomic force microscopy, Raman, infrared, and UV-Vis absorption measurements (Supplementary Section 6). In addition, p-type characteristics are only recoverable in the presence of oxygen (Supplementary Section 7), indicating that it plays a key role in the disappearance and recovery of the p-type characteristics in the above process.
Oxygen has a tiny molecular radius (about 1.73 Å, and the intermolecular π-π stacking distance of OSCs is generally larger than 3.4 Å) and a large electron affinity 32 (relative to OSCs), so it can penetrate the lattice of OSCs, find an interstitial position, and play an electron-withdrawing role.
Therefore, we deduce that electron transfer may occur between the neutral semiconductor molecule and oxygen, i.e., oxygen works as an interstitial acceptor impurity and induces mobile hole carriers in OSCs. However, the concentration of the inherent oxygen in OSCs is at trace-levels, as confirmed by the fact that X-ray photoelectron spectroscopy (XPS) of the fresh DNTT film (without exposure in air) shows no oxygen peak signal (Supplementary Section 8). While exposed in air, a clear oxygen peak signal is observed, which suggests that the extra oxygen is adsorbed on OSCs (Supplementary Section 8). It is the easily detectable oxygen that has been extensively studied in previous work [26][27][28][29][30] and can be easily removed by vacuum annealing 26,27 . Quite differently, our goal is to gain an insight into the role of the trace, lattice-incorporated, oxygen that is inherent in OSCs. However, most conventional spectroscopy techniques, such as Raman, infrared, and UV-Vis absorption measurements, have limited capacity to identify the trace impurities. Therefore, to explore the role of oxygen in OSCs, we used electron paramagnetic resonance spectroscopy (EPR), a technique that has extremely low detection limits, to track unpaired electrons in the above process.
As shown in Fig. 1d, the fresh DNTT OFET in the static magnetic field exhibits the identifiable EPR signal with g = 2.0034 and peak-to-peak linewidth (△Hpp) ≈ 5G (see Methods). The temperature dependence of the signal shows that the magnetic susceptibility (ꭕ) obeys the Curie law (Supplementary Section 9). The signal was thus assigned to the DNTT radical cation with S = 1/2 spins in the solid state 33 . Furthermore, the spin density Nspin of ORC (carrier concentration) was about 1.9 × 10 15 cm -3 by the Curie law (Supplementary section 9). After confirming the existence of ORC, its putative partner should be SA, but the EPR signal of SA is usually too broad to be directly observed because of fast spin relaxation. However, the presence of SA could be investigated through a spin trapping strategy using 5,5-dimethyl-1-pyrroline N-oxide (DMPO), as the SA radical trapping agent.
Significantly, after the DNTT film was immersed in the DMPO diluent for 30 min (see Methods), the characteristic EPR signal of the adduct DMPO-OOH 34 of DMPO with SA was observed, with six peaks of equal height and hyperfine coupling constants AN ≈ 13.9 G and Ahβ ≈ 8.8 G (Fig. 1e). These EPR signals are direct indicators of the existence of the geminate ORC and SA, revealing that OSCs are innately doped by trace oxygen (10 15 cm -3 ), and are not, as has been long assumed, intrinsic.
To further investigate the correlation between these species and OSC p-type characteristics, the EPR signals of ORC and SA (by spin trapping) in the plasma-treated OFET were examined under nitrogen protection (see Methods). As expected, no EPR signal was observed, indicating that the plasma treatment can efficiently eliminate (reduce to neutral molecules) the ionized acceptors, SA, and hole carriers, ORC. Furthermore, the EPR signals of ORC and SA were almost recovered to their initial levels after illumination in O2 for 1 h ( Fig. 1d and 1e).
From the above electrical and EPR measurements, the underlying mechanism for the key role of oxygen on p-type characteristics is deduced thus: oxygen, acting as an acceptor, withdraws one electron from a neutral OSC molecule, generating SA (i.e., an ionized acceptor) and ORC (i.e., a hole carrier). After plasma treatment, the ORC and SA are eliminated, resulting in the disappearance of ptype characteristics, and upon illumination in oxygen, they are regenerated and thus restore the p- causing the extended tail to shift towards a high energy level (i.e., a reduction in the relative distance between EF and the HOMO), which suggests that oxygen plays the role of an acceptor dopant in OSCs.
To elucidate these results, the energy band model of the OFET is outlined in Fig. 2a off-state current ( Fig. 1b and c). Furthermore, the de-doping makes the pre-emptied donor-like trap states occupied, and the carrier transport is strongly impeded by the trapping effect, causing the subthreshold swing to deteriorate. According to the multiple trapping and release (MTR) model 38 , a decrease in mobility is expected. At this point, even if a large gate voltage is applied, the EF-HOMO energy difference remains large due to the EF pinning effect of the traps, and hence the threshold voltage dramatically increases. The above analyses qualitatively explain the decline and the ultimate disappearance of OSC p-type characteristics after plasma treatment ( Fig. 1b and c, details of the discussions in Supplementary section 11). and Ultraviolet Photoelectron Spectra (UPS). The UPS and TOF-SIMS are connected through a vacuum pipeline (see Methods), whose environmental chamber is integrated with a plasma source.
Benefiting from this, all characterizations of the DNTT film can be performed in vacuum conditions.
Significantly, TOF-SIMS clearly shows the decrease of oxygen concentration after plasma treatment ( Fig. 2b) implies that bandgap absorption produces excitons efficiently, and that the photoexcitation of OSCs is also a key step in the re-doping process (as shown in Fig. 2c). The mechanism of the re-doping can be described as follows: an OSC molecule is photoexcited, i.e., an electron is excited from the HOMO to LUMO, and the electron in the excited state is energetic enough to be transferred to a nearby interstitial oxygen molecule. This process forms SA and ORC (Supplementary Section 13).
To further verify that trace oxygen is innate and stable in OSCs, a bottom-contact OFET was fabricated by vacuum evaporation (i.e., not in the presence of oxygen) and then was measured in situ.
Upon the deposition of the semiconductor layer, p-type transport behavior can be observed under vacuum, as shown in Fig. 2f. Therefore, it can be validated that the trace oxygen is inherent in OSCs and can survive the vacuum sublimation, which is also supported by the EPR signals of the purified materials (Supplementary Section 14). Since oxygen is ubiquitous, spontaneous trace oxygen doping of OSCs may occur in the processes of material synthesis and storage. Therefore, some OSCs are 'born' with OIORC and thus endowed with p-type characteristics. Although some reports have speculated that there are unintentional dopants in OSCs 2 , this is the first study to unambiguously assign and uncover the impact of these dopants.
The universality of the innate OIORC was investigated with other six typical OSCs including small molecules and polymers, with their chemical structures shown in Fig. 3 Fig. 1b and c) in the processes of plasma treatment and illumination in oxygen ( Fig. 3a-f). These results reveal the same origin of p-type characteristics in OSCs (i.e., OIORC) and demonstrate the university of the de-doping and re-doping methods. The incomplete elimination of ptype characteristics for single crystal rubrene and the C8-BTBT crystalline film after plasma treatment may be due to the serious suppression of the short-circuit diffusion for the plasma and oxygen in thick crystals. It is well-known that controlled doping is the core technique enabling the fine control of semiconducting characteristics in inorganic electronics. To demonstrate the potential applications of the reversible de-doping and re-doping methods in organic electronic devices, three kinds of tests for modulating key device parameters were performed: 1) The threshold voltage (Vth) is a vital parameter for transistors, and precise Vth control can be achieved for inorganic semiconductors through ion implantation (a doping technique) 25 , but this has been a challenge in organic electronics 40 . As shown in Fig. 4a and b, 14 OFETs with a negative Vth showed weak dependence on gate voltage (Fig. 4c). Notably, the conductivity (σ) sharply increased over two orders of magnitude to 72 S m -1 , and the field-effect mobility ( ) showed only a slight decrease (5.2 to 3.6 cm 2 ·V -1 ·s -1 ). With these data, the carrier density was calculated as 2×10 18 cm -3 (using = ), which is a high doping level for small molecule OSCs (molecular density about 10 21 cm -3 ). In addition, the I-V curve was transformed from nonlinearity to linearity (Fig. 4d), indicating the obvious improvement of metal-semiconductor junctions under high concentration doping.
3) The polar type of the transport behaviors can be modified by oxygen doping and de-doping.
OSCs are intrinsically ambipolar, but the innate oxygen doping generally compensates for the electrons. However, we hypothesized that p-type materials could appear n-type after eliminating oxygen and, conversely, n-type materials could be converted to be p-type through further oxygen doping. For the former case, we showed that an OFET of tips-pentacene, showing p-type transport behavior both in air and vacuum, was modified to be n-type via de-doping by plasma treatment (Fig.   4e). For the latter case, we showed that P(NDI2OD-T2), a stable and high-performance n-type polymer 41 , was endowed with p-type behavior via doping by simultaneous annealing and illumination in oxygen (Fig. 4f). Although similar re-doping methods have been reported, due to the lack of a dedoping method for the removal of lattice-incorporated oxygen, full control of the vital parameters of OSCs and devices could not be realized by the re-doping method alone. These experiments demonstrate the power and utility of the de-doping and re-doping processes; dramatically widening the possible OSC charge transport property space without introducing any new materials.
With the lack of an effective trace oxygen de-doping strategy, as outlined in this study, it is understandable that the p-type charge transport in OSCs was considered intrinsic in origin, and why oxygen doping was always considered an additional influence on the 'inherent' p-type characteristics [26][27][28][29][30] . Based on this incorrect premise, many phenomena, and properties of OSCs and photo-electronic devices have been studied and reported, which, in fact, might be strongly affected by the innate oxygen doping. In this light, some classical physical models for the quantitative description of organic devices are invalid, such as depletion approximation theory 25,42 and Debye length calculations 34 , both of which require the estimation of doping concentration. We believe this work will help to correctly use these models, clarify previously perplexing key organic electronics phenomena, and push forward OSCs material research: 1) Depletion-mode transistors 4,5 , carrier densities 6,10 and conductivities beyond expectation 2,9,43 , and the small energy gap between the EF and the HOMO 7,8 can be easily interpreted by oxygeninduced excess hole carriers. This also explains the difficulty of inversion of polarity for OSCs 44 .
2) The deterioration of subthreshold swing (S) at low temperature 11,12 was previously unclear. Ideally, S should improve at low temperature due to the relationship: ≡ ( 10) where Ci and Ct are capacitance of dielectric layer and traps, respectively 45 . However, experimentally, the S value always increases at low temperature in OFETs. This can be explained by the freeze-out effect; at low temperature, the carriers (ORCs) are frozen around the impurities (oxygen) (i.e., impurities are not ionized), and thus the trap states passivated by oxygen doping recover the ability of trapping field-induced (gate voltage) carriers, which increases the trap capacitance and thus deteriorates S.
3) These findings may inspire the research on the origin of n-type characteristics in OSCs. We anticipate that an unrevealed impurity is prevalent in n-type OSCs, just like oxygen, and passivates the acceptor-like traps and induces mobile electron carriers in the LUMO.
Our study reveals, through a low-power plasma de-doping strategy, that trace oxygen is the origin of the p-type characteristics of OSCs. This finding explains previously perplexing organic electronics phenomena and will enable a foundation upon which we can re-understand charge transport in OSCs.
Furthermore, our de-doping and re-doping processes will expand the property space for all known OSC materials and, we believe, will become a core technology for the growing organic electronics industry. was spin-coated on the highly doped Si at 800 rpm for 1 min, and then 20 nm Au was deposited on PMMA/Si. The air gap was fabricated by mechanical scraping using a probe. The C8-BTBT crystalline film was prepared by the space-confined self-assembly method 46 , and the Au (80 nm) stripes were stamped on the C8-BTBT crystalline film as the source and drain electrodes 47 .

In-situ Electrical characterizations
The electrical characterizations of the OFETs were carried out in a home-made system integrated with a DC inductively coupled plasma source, an optical window and the environmental control system

EPR characterizations
X-band (9.5 GHz) electron paramagnetic resonance (EPR) measurements were performed with both a Bruker EMXplus-6/1 and JEOL JES-FA200. The microwave power and the modulation magnetic field were carefully adjusted to produce the optimal EPR signal. DNTT films were deposited onto a 3.5 mm wide × 20 mm long quartz substrate. The sample was then inserted into a quartz EPR tube.
The signal of the DNTT film is usually weak due to the larger spin-orbit coupling caused by the sulfur atoms 48 . The signal intensity can be improved by increasing the quantity of samples. For lowtemperature measurements, a continuous liquid nitrogen flow cryostat was used. 2,2-Diphenyl-1picrylhydrazyl (DPPH) was used as a standard spin counting reference. For the detection of the superoxide anion (SA), DMPO (volume ratio of 1:100 in methanol or dimethylsulphoxide) was used as a spin trapping agent. To boost the signal intensity, multiple DNTT films were immersed into DMPO diluent for at least 30 min. The solution was sucked up with a glass capillary and inserted into a quartz EPR tube. To track unpaired electrons in the de-doping process, the plasma-treated DNTT film (not exposed to the air) was inserted into a quartz EPR tube in the N2 glovebox and then the tube was sealed with silicone grease. The samples for SA detection of the plasma-treated DNTT films were prepared and sealed in the glovebox as described above. Note that purified gases (99.999%) for plasma treatment needed to be further purified by the dehydration tube and deoxygenation tube in series. To investigate the re-doping process, the plasma-treated DNTT films were illuminated in O2 with a Xe lamp for 24 h, and then the tests were performed as above. For the detection of raw materials, OSCs powders or suspensions blended into DMPO diluent were loaded into glass capillaries. After being sealed with silicone grease, the glass capillaries were inserted into a quartz EPR tube. Some weak signals were carefully denoised.

Quasi-in-situ UPS and TOF-SIMS characterizations
UPS (PHI 5000 Versaprobe II) and TOF-SIMS (TOF.SIMS5-100) are connected through a vacuum pipeline, whose environmental chamber (which controls the gas type and the atmospheric pressure) is integrated with a plasma source. DNTT films (about 50 nm) were deposited onto the highly doped Si and SiO2/Si for UPS and TOF-SIMS tests, respectively. The excitation source for UPS was He Iα (hν = 21.22 eV). Vacuum level shifts were determined from the low kinetic energy part of UPS spectra with a -5 V sample bias. Secondary ions employed Cs + as the primary ion source. The size of the crater was 100 × 100 um, and the area of acceptance was 25% of the total sputtered area. The UPS and TOF-SIMS tests were performed on the fresh DNTT films in turn, and they were returned to the environmental chamber through a vacuum pipeline to be treated by plasma with purified gas (gas was further purified by the dehydration tube and deoxygenation tube in series). The plasma-treated DNTT films were transferred to the UPS and TOF-SIMS in turn through a vacuum pipeline for testing. All the transfer processes of samples were carried out in vacuum. The in-situ test equipment and technology supports were provided by NANO-X (Vacuum Interconnected Nanotech Workstation) in Suzhou.

Morphology, physical phase, and chemical structure characterizations
Morphology, physical phase, and chemical structure characterizations on the DNTT films were performed before and after plasma treatment. AFM measurements were carried out on a Dimension ICON (Bruker). XRD measurements were carried out on a MiniFlex600 (Rigaku). DNTT films were deposited on a quartz plate for the measurement of the UV-Vis absorption spectrum (a Lambda 750) and deposited on Au-coated Si for the measurements of the Raman spectrum (a DXR Microscope with a 532 nm laser), and IR spectrum (Nicolet IN10, Thermo Fisher Scientific). A 50 nm DNTT film was deposited on highly doped Si in a vacuum thermal evaporation system and transferred under the protection of N2 to an Ar glovebox connected to an ESCALAB-250Xi (Thermo Fisher Scientific).

Modulation of the key device parameters
Threshold voltage: In-situ electrical measurements were performed on 14 DNTT (20-30 nm) OFETs with negative Vth after plasma treatment. The parameters of the plasma treatment were carefully modulated until the Vth was tuned at about 0 V. Correspondingly, 14 DNTT OFETs with positive Vth were repeatedly measured after simultaneous annealing and illumination in oxygen until the Vth was tuned at about 0 V.

Conductivity:
The C10-DNTT OFET was annealed at 80 ºC in a cavity with an inspection window, and O2 was filled into the cavity with a pressure of about 3 atm. Illumination was performed by a Xe lamp.

Polarity:
The Tips-pentacene film was prepared by the drop-casting method (1 mg ml -1 in toluene) onto the PS-coated SiO2/Si and then Au was deposited as electrodes. The P(NDI2OD-T2) film was prepared by spin-coating onto OTS-treated SiO2/Si and Au was also used as electrodes. The Tipspentacene OFET was treated by plasma at 20 W for 1 min and P(NDI2OD-T2) OFET was annealed at 60 ºC and illuminated in O2 for 3 h.
Data availability 22 The data presented in this study are available from the corresponding authors.