The dynamic organic p?n junction

The dynamic organic p?n junction

(Parte 1 de 2)

The dynamic organic p–n junction

Static p–n junctions in inorganic semiconductors are exploited in a wide range of today’s electronic appliances. Here, we demonstrate the in situ formation of a dynamic p–n junction structure within an organic semiconductor through electrochemistry. Specifically, we use scanning kelvin probe microscopy and optical probing on planar light-emitting electrochemical cells (LECs) with a mixture of a conjugated polymer and an electrolyte connecting two electrodes separated by 120µm. We find that a significant portion of the potential drop between the electrodes coincides with the location of a thin anddistinctlight-emissionzonepositioned>30µmawayfromthenegativeelectrode.Theseresultsarerelevantinthecontext of a long-standing scientific debate, as they prove that electrochemical doping can take place in LECs. Moreover, a study on the doping formation and dissipation kinetics provides interesting detail regarding the electronic structure and stability of the dynamic organic p–n junction, which may be useful in future dynamic p–n junction-based devices.

Organic semiconductors are heralded for their simple processing and extraordinary chemical customizability, and emerging low-cost and flexible devices based on these materials are expected to revolutionize the role electronics have in our everyday lives1,2. Some devices are already commercially available—a notable example being the organic light-emitting diode in television and cell-phone displays—but it is clear that further opportunities exist beyond the current state-of-the-art. For instance, the soft nature of organic semiconductors can enable in situ electrochemical tuning of important material properties3–13. One device that exploits this opportunity in an attractive manner is the light-emitting electrochemical cell14–2 (LEC). The nominal difference between an LEC and an organic light-emitting diode is that the former contains mobile ions in the active material23–30. These ions rearrange during operation, which in turn allows for a range of attractive device properties, including low-voltage operationwiththickactivelayersandstableelectrodematerials31–36.

However, the further development of LECs is currently hampered by an inadequate understanding of the device operation. In fact, an active debate regarding the fundamental nature of LEC operation has continued for more than a decade, and two distinct models are competing for acceptance: the electrochemical doping model18,32,37–40 and the electrodynamic model36,41–4. To distinguish them, it is appropriate to establish the electrostatic potential profile in a device during steady-state operation, as the models predict distinctly differing profiles. The electrodynamic model predicts that the entire applied potential will drop over thin electric double layers (EDLs) at the electrode/active material interfaces, whereas the electrochemical doping model predicts that a significant fraction of the applied potential will drop over a light-emitting p–n junction. Thus, to discriminate between the two models, it is essential to record the potential profile within an LEC device where the light-emission zone is positioned away from the electrode/active material interfaces.

Two attempts to measure the electrostatic potential within an

LEC device during operation have been published. Slinker et al.43 used electric force microscopy on a planar LEC, with an ionic organometallic semiconductor as the active material. Although their measurements have been questioned, primarily because light

1The Organic Photonics and Electronics Group, Department of Physics, Umeå University, SE-901 87 Umeå, Sweden, 2Department of Applied Physics, Eindhoven University of Technology, PO Box 513, 5600 MB, Eindhoven, The Netherlands, 3The Transport and Separations Group, Department of Physics, Chemistry and Biology, Linköping University, SE-58183 Linköping, Sweden. *

Au Au O

Au Au

Figure 1 | Schematic diagram illustrating the probing of a planar LEC device with SKPM. The solid line marks the topographic scan and the dashed line indicates the SKPM scan in lift mode. The active material components are included in the inset to the right.

emission and the potential profile were observed under rather differing experimental conditions37, they demonstrated that such direct potential measurements could be carried out in planar devices.Inamorerecentreport,Pingreeetal.45 usedscanningkelvin probe microscopy (SKPM) on planar LECs containing a mixture of aconjugatedpolymerandasolid-stateelectrolyteastheactivematerial. They did not report any light-emission data, and the presented potentialprofileswereunfortunatelyinconclusiveinthatthepotential drop was positioned at the negative electrode interface, and as suchineffectconsistentwithbothproposedmodelsofoperation.

Here, we use the SKPM technique in parallel with lightemission detection on planar LEC devices, comprising an active material mixture of poly[2-methoxy-5-(2-ethylhexyloxy)- 1,4-phenylenevinylene] (MEH–PPV)+poly(ethylene oxide)(PEO)

+ KCF3SO3 positioned between two Au electrodes, where the electrodes define an inter-electrode gap of 120µm. Figure 1 shows a schematicdiagramillustratingtheSKPMprobingofaplanardevice, and the active material constituents are shown in the inset on the


Potential (V) a b c

Figure 2 | Light-emission and potential profiles in a planar LEC device during operation. a, Micrograph showing the light emission from a planar LEC device during steady-state operation at V =5V; the positive and negative electrodes are indicated by the + and − sign, respectively. b, 2D topographic image of an identical planar LEC device. c, Electrostatic potential profile recorded on a pristine device under open-circuit conditions. d, Temporal evolution of the potential profile during the initial operation at V =5V; the arrow indicates increasing time. The time delay between successive potential profiles was ∼20s. e, The subsequent steady-state potential profile recorded at V =5V.

right. As we used a perpendicular tip–electrode geometry, which may lead to artefacts when interpreting SKPM data, we also carried outadeconvolution(seeSupplementaryInformation)thatdemonstrated that the recorded potential profiles are at least qualitatively correct. Importantly, we carried out SKPM measurements and light-emission probing under similar conditions on separate sets of LEC devices (see the Methods section for details), in which the light-emission zone is positioned far away from the electrode/active material interfaces35,39. As mentioned above, measuring the electrostatic potential within the device is critical to distinguish between the two models. It is made possible in these experiments by the use of an appropriate organic semiconductor/electrolyte active material driven with a relatively high voltage, so that undesired electrochemicalsidereactionsattheelectrode/activematerialinterfacesare effectively suppressed46,47.

Figure 2a is an optical micrograph of an LEC device taken at t =300s after an external voltage of V =5V was applied. A distinct light-emission zone is apparent in the bulk of the active material at a distance d ∼35µm away from the negative Au electrode interface. Figure 2b,c shows the two-dimensional (2D) topography image and the potential profile at open circuit, respectively, recorded on an identical device. The structure in the potential profile is due to a minor charge transfer between the Au electrodes and the MEH-PPV polymer. The two interfaces between the electrodes and the active material are clearly distinguishable at d = 0 and 120µm and indicated by the vertical dashed lines. Atomic force microscopy (AFM) data demonstrating the sharpness of this interface are included in Supplementary Information.

Figure 2d,e shows the temporal evolution of the potential profile up to t = 180s (with increasing time indicated by the arrow) and between t = 180s and t = 600s, respectively, for the same device during operation at V = 5V. The potential profile changes from

Potential (V) d e a f

Figure 3 | Experimental data illustrating the formation and reformation of an organic p–n junction. a, 2D topography image of a planar LEC device. b, Electrostatic potential profile during steady-state operation at V =+5V. c, Transient potential profile measured with the device disconnected (open circuit), directly after long-term operation at V =+5V. d, Temporal evolution of the potential profile after a subsequent switch to V =−5V; the arrow indicates increasing time. e, Steady-state potential profile at V =−5V. f, Micrograph showing the light emission during steady-state operation at V =+5V. g, Subsequent micrograph showing the light emission from the same device during steady-state operation at V =−5V.

dropping essentially linearly between the two electrodes at t ∼ 10s (first line scan in Fig. 2d) to reach steady state (potential profile changes very little over several minutes) at t ≥ 180 s (last line scan in Fig. 2d,e), where the steepest potential drop is localized over a limited spatial region centred at d ∼ 35µm. It is notable that the spatial position of this potential drop coincides very well with the location of the light-emission zone in Fig. 2a. The corresponding currentmeasurementisincludedinSupplementaryFig.S2.

At this stage, it is relevant to consider the key steps within the electrochemical doping model. (1) Following the application of an external voltage, thin EDLs form at the electrode/active material interfaces; (2) if the applied voltage is sufficiently large (V ≥ Eg/e, where Eg is the bandgap of the organic semiconductor and e is the elementary charge), holes and electrons are injected into the organic semiconductor through the EDLs at the positive and negative electrodes, respectively; (3) the injected electronic charge carriers attract electrostatically compensating ions, which establish dopedregionswithhighconductivityatthetwoelectrodeinterfaces; (4) a p-type doping front (comprising holes and compensating anions) and an n-type doping front (comprising electrons and cations) grow towards each other and, after a turn-on time, make contact, forming a p–n junction; (5) subsequently injected holes and electrons migrate through the doped regions and recombine within, or in close proximity of, the undoped p–n junction, causing the emission of light32,38.

The electrodynamic model includes the first two steps above, but, notably, does not include any electrochemical doping, and


V = 5 V

Vbi n-doped polymer p-type dopant n-type dopant Positive ion

Negative ionHole Electron

Metal electrodesp¬n junction p-doped polymer

V¬Vbi a b

Figure 4 | Schematic diagrams illustrating the electrostatic profile and the electronic and ionic charge distribution within a p–n junction structure established at V =+5V. The net charge motion is indicated by the white arrows. a, The transient charge distribution and initial ionic motion directly after a shift to the open-circuit condition, where Vbi indicates the built-in potential over the junction. b, The steady-state charge distribution, electronic motion and radiative recombination at V =+5V.

hence no conductivity increase or p–n junction formation. Instead, it claims that the electronic charge carriers are driven by diffusion in the undoped bulk of the active material, and importantly that the potential drops at steady state are localized at the electrode interfaces41,42.

However, our observation of a significant localized potential drop positioned in the bulk of the active material, far away from the electrode interfaces, at steady state (see Figs 2e and 3) directly contradicts the electrodynamic model. In contrast, the observation of a localized potential drop in the bulk that spatially coincides with a distinct light-emitting zone is consistent with the existence of a forward-biased p–n junction (as schematically illustrated in Fig. 4b). Thus, the presented data provide strong evidence that LEC devices can operate in a manner consistent with the electrochemical doping model.

WenotethatwedonotdetecttheformationofEDLsatthemetal electrode interfaces during the initial operation of these devices (as predicted by both models). We propose, on the basis of X-ray photoemission spectroscopy data and the temporal evolution of recorded AFM phase data (see Supplementary Figs S3 and S4 and the corresponding discussion in the Supplementary Information) as well as the absence of an interfacial potential drop between the negative Au electrode and the n-type-doped material at steady state, that this is due to the existence of a thin layer of ioncontaining material on top of the metal electrodes that screens some of the potential.

We also wish to mention a number of interesting features of the probed dynamic organic p–n junction structure. First, both the light-emission data and the potential profiles indicate that the effective p–n junction is broad, with a width of the order of 10µm. Second, the slope in the potential profile at steady state is notably steeper on the n-type side of the junction than on the p-type side (see Fig. 2e). We believe that the latter is an indication of the conductivity difference between the n-doped and the p-doped MEH-PPV. The n-doped material seems to be less conducting, so that a larger potential gradient is necessary on the n-type side to provideaconstantcurrentdensityacrossthedeviceatsteadystate.

To obtain more detailed information on the structure and stability of the dynamic p–n junction and the kinetics and reversibility of the doping process, we have investigated the effects of shifting the voltage applied to a device operating at steady state. Figure 3a shows a 2D topography image that, together with potential images at open circuit (not shown) and micrographs, identifies the positions of the electrode interfaces as indicated by the vertical dashed lines. Figure 3b,f shows the potential profile and a micrograph recorded at steady state at a ‘forward bias’ of V = +5V. A direct correlation between a significant potential drop in the bulk of the active material and the location of the emission zone is once again observed. The device was thereafter left disconnected at open circuit (impedance>100M ) for a brief period of time (∼30 s), and the first recorded potential profile after disconnection is shown in Fig. 3c. Finally, a ‘reverse bias’ of V = −5V was applied. Figure 3d shows the temporal evolution of the potential between t =0 and 1,800s following the application of the reverse bias (the arrow indicates increasing time). Figure 3e,g presents the subsequent steady-state potential profile at t ≥ 1,800 s and a micrograph recorded at t =1,920s, respectively.

To facilitate a discussion of the recorded data in Fig. 3, two schematic diagrams are shown in Fig. 4. First, we note that the steady-state potential profile and the position of the light-emission zone in the inter-electrode gap at V = −5V (recorded at the end of the measurement) are essentially mirror images of those at V = +5V. This implies that all of the relevant processes are highly reversible under the conditions and the timescale of this study, and that the effects of chemical and electrochemical side reactions can be excluded from the coming discussion.

We note with interest that Fig. 3c reveals a potential drop of ∼1.5V at the location of the p–n junction, immediately after the long-term, steady-state operation voltage of V = +5V is disconnected (the circuit is opened). This observation is consistent with the existence of a built-in potential of Vbi ∼ 1.5V (which is slightly lower than the bandgap potential, VBG = Eg/e, of the semiconductor, MEH-PPV; ref. 48) over the p–n junction, with a polarity opposite that of the applied voltage that built the p–n junction. We propose, in analogy with the well-established physics of p–n junctions in inorganic crystalline semiconductors, that Vbi can be attributed to an equilibration of electronic charge carriers over the junction region. More specifically, one can predict the establishment of a diffusion–drift balance for the electrons over a p–n junction at open circuit, resulting in a uniform electrochemical potential for the electrons (or Fermi level) throughout the device, as follows: ‘fast’ electrons diffuse from the n-type side to the p-type sideleaving‘slow’cationsbehind;thisprocessestablishesapotential drop Vbi over the junction, which causes an equal and opposite drift of electrons from the p-type side to the n-type side. (An equivalent set of events involving ‘fast’ holes and ‘slow’ anions will simultaneously take place.)

Figure 4a shows the electrostatic potential (top part) and a schematic diagram of the doping structure and the charge separation over a p–n junction at open circuit (bottom part), which rationalizes the recorded potential profile shown in Fig. 3c. Figure 4a also indicates the onset of net ionic motion over the junction, which motivates the transient character of the potential profile. This relaxation process, which is a manifestation of the

NATUREMATERIALSDOI:10.1038/NMAT2478 ARTICLES non-equilibrated electrochemical potential for the ions over the junction, is further discussed below.

The existence of Vbi over the junction region also explains why the probed potential Vtot over the p–n junction under steady-state conditions is smaller than the applied potential V (see, for example,

Fig. 3b). The potential probed with SKPM corresponds to the total electrostatic potential within the device, which here is a combination of the externally applied potential and the built-in potential, where the latter at steady state has a sign opposing the former; that is, Vtot = V − Vbi (see top part in Fig. 4b). The net charge separation over the junction region, as well as the charge transport and radiative recombination, in a planar LEC device at steady state is shown in Fig. 4b. The existence of Vbi also explains why the probedpotential is larger thanthe applied voltage following a fast switch in the polarity of the applied voltage (from +5 to −5V in Fig. 3d). Here, the applied voltage and the built-in potential are temporarily oriented in the same direction, and the total probed potential is accordingly Vtot =V +Vbi >V. The p–n junctions discussed here are distinctly different from conventional p–n junctions in, for example, Si, in that the compensating, and dopant-stabilizing, counter ions are mobile. Thus, these novel doping structures are appropriately termed dynamic p–n junctions, whereas conventional p–n junctions are static structures. The mobility of the ions enables the dynamic p–n junction to be formed in situ, but also means that the doping and junction will dissipate and/or reform in response to a shift in the external applied voltage, as demonstrated in Fig. 3. Such redistribution processes involve a complicated interplay of electronic and ionic motion, as exemplified in the complex evolution of the potential profiles in Fig. 3d following a shift in the applied voltage.

We further note that a dynamic p–n junction can be stable only above a critical applied voltage exceeding the built-in potential,

V > Vbi, because the ionic concentration gradients and the accompanying doping profiles will dissipate at a lower voltage. A simple example involving the stability of the cationic concentration gradient can illustrate this point (but the same argument holds for the anionic gradient as well): the cations on the n-type side of the p–n junction exhibit a net diffusive motion towards the p-type side (owing to the cationic concentration gradient over the junction), and this diffusive motion can be compensated by a drift motion of cations in the opposite direction only if the total electrostatic potential over the junction (the sum of

V and Vbi) drives the cations in the direction opposing the concentration gradient. Thus, because Vbi opposes V under steady- state conditions, it follows directly that V > Vbi is the prerequisite for a stable dynamic p–n junction structure. One scenario under which the ionic concentration gradients and the junction structure are unstable (that is, V < Vbi) is illustrated in Fig. 4a, whereas

Fig. 4b shows a situation (V = 5V > Vbi) at which the p–n junction structure is stable.

Finally, it is appropriate to mention that the primary differences between the experiments presented here and those of our predecessors is the chemical system used and the conditions under which the measurements were made. We have previously demonstrated that the use of small Li+ cations and/or the application of low voltages in conjugated polymer-based LECs can negatively affect the placement of the p–n junction, leaving it very near the negative metal electrode35,46. This is plausibly the reason why Pingree et al.45 observed a different potential profile than we do. The ruthenium-based ionic transition-metal complex used by Slinker et al.43 in their light-emitting device differs chemicallyfromtheMEH-PPVconjugatedpolymerpresentedhere. It is not clear, for example, that this type of transition-metal complex can be doped into a highly conductive state, which would result in an operational mechanism unlike that observed in conjugated-polymer LECs.


We have used a combination of SKPM and light-emission probing to establish that a p–n junction can form in situ in the bulk of the active material of an LEC device during operation. This observation provides evidence that the so-called electrochemical doping model can describe the operation of LECs, which is relevant in the context of an ongoing debate in the scientific literature. Moreover, the availability of a dynamic organic p–n junction opens the possibility for interesting fundamental physics. For example, we demonstrate the existence of a built-in potential over the junction, which must be compensated by an external voltage to stabilize the junction structure. Finally, we note that the tantalizing subject of dynamic organic p–n junctions is largely unexplored, and that further studies in this field could open a wide range of novel and useful applications.


(Parte 1 de 2)