Probing transport energies and defect states in organic semiconductors using energy resolved electrochemical impedance spectroscopy

Determining the relative energies of transport states in organic semiconductors is critical to understanding the properties of electronic devices and in designing device stacks. Futhermore, defect states are also highly important and can greatly impact material properties and device performance. Recently, energy‐resolved electrochemical impedance spectroscopy (ER‐EIS) is developed to probe both the ionization energy (IE) and electron affinity (EA) as well as sub‐bandgap defect states in organic semiconductors. Herein, ER‐EIS is compared to cyclic voltammetry (CV) and photoemission spectroscopies for extracting IE and EA values, and to photothermal deflection spectroscopy (PDS) for probing defect states in both polymer and molecular organic semiconductors. The results show that ER‐EIS determined IE and EA are in better agreement with photoemission spectroscopy measurements as compared to CV for both polymer and molecular materials. Furthermore, the defect states detected by ER‐EIS agree with sub‐bandgap features detected by PDS. Surprisingly, ER‐EIS measurements of regiorandom and regioregular poly(3‐hexylthiophene) (P3HT) show clear defect bands that occur at significantly different energies. In regioregular P3HT the defect band is near the edge of the occupied states while it is near the edge of the unoccupied states in regiorandom P3HT.


Introduction
Measuring the ionization energy (IE), electron affinity (EA), and the density of states (DOS) with high sensitivity is critical for understanding and predicting the electronic properties of semiconducting materials and devices. [1][2][3][4] For example, at a heterojunction, it is the relative IEs and EAs of the two materials that determine the direction in which charge transfer of an electron of sufficient kinetic energy (IPES). In both cases the measurements are with respect to the kinetic energy of electrons in vacuum, [2] thereby all measurements in principle share a common reference level (i.e., the energy of an electron at rest in a vacuum) and thus can be directly compared. When the material is being used in a solid-state semiconducting device, where no solvent or electrolyte is present, the IE, EA, and work function measured by UPS and IPES are the most relevant energies for understanding material and device properties. However, these methods are often not sensitive enough to detect defect states, except for UPS variations that rely on tunable low-energy photon sources or intense monochromatic light sources. [22][23][24][25][26] Another concern with OSCs is that low photon fluxes are often necessary to minimize measurementinduced sample damage, which further reduces the sensitivity of the measurement and limits the application of high intensity monochromatic light sources. [27][28][29] In contrast to photoemission spectroscopies that are conducted in ultra-high vacuum, electrochemical measurements are performed in electrolyte solution and measure the oxidation or reduction potential relative to a standard redox system with a known electrochemical potential (e.g., ferrocene/ferrocene + , Ag/Ag + , or the normal hydrogen electrode). [30] These electrochemical measurements are then converted to an IE or EA based on the reduction potential of the standard redox system versus vacuum. It is important to pay attention to the conversion factors used, as these differ throughout the literature. [30] During an electrochemical measurement, such as cyclic voltammetry, differential pulse voltammetry, or ER-EIS, the material or compound of interest is reduced or oxidized as the voltage is scanned. For a solid material, oxidation or reduction is accompanied by the influx of ions to balance the injected charge. [31][32][33][34] These ions are often accompanied by solvent molecules, with the quantity of solvent molecules depending on both the electrolyte solvent and the structure of the ion. [31][32][33][34][35] As a result of solvent and ion uptake, the electrochemically determined IE and EA are reflective of the electrolyte solution used, including both the solvent and ions. Such solvent and electrolyte dependence is documented throughout the literature. [36][37][38][39] For example, Baustert, et al. recently showed that large ions are excluded from the crystalline regions of regioregular-poly(3-hexylthiophene), RR-P3HT, thus resulting in a shift of approximately 250 mV in the oxidation potential relative to when a small anion is used in the electrolyte solution. [37] Additionally, in OECTs, which rely on electrochemical doping, the performance parameters are highly dependent on the electrolyte ions. [35,36,40] This dependence on electrolyte chemistry makes electrochemical measurements preferable to photoemission measurements when the intended use of the material is in an electrochemical device.
One downside of commonly employed electrochemical methods, which also applies to UPS and IPES, is that they are not sensitive enough to probe defect states at device-relevant concentrations. On the other hand, pioneering work on ER-EIS by Schauer and coworkers has shown that the density of states near the edge of the HOMO and LUMO can be extracted over a range of five to six orders of magnitude, thus allowing the detection of defect states at densities down to 10 16 or 10 15 cm −3 . [7][8][9][10][11] In an ER-EIS measurement the DC voltage bias is slowly scanned while a small AC voltage perturbation is applied. Through selecting an appropriate AC frequency and fitting to an equivalent circuit model the bulk capacitance can be determined and the DOS extracted as a function of DC voltage. The ER-EIS measurements also contain a second term that can provide information on states closer to the surface, which can be particularly insightful for understanding device behavior where interfacial properties often have an outsized influence on device performance.
There are several additional methods that can be used to probe defect states in OSCs. For example, photothermal deflection spectroscopy (PDS) is a sensitive optical absorbance spectroscopy that can probe sub-bandgap absorption features. [41,42] Sensitive time-dependent charge extraction measurements, such as time-of-flight photocurrent, can provide information on defect states that act as charge traps. [11,43,44] Here, the current decay as a function of time is measured and related to trap state emission. Thermal admittance spectroscopy (TAS) also provides a sensitive probe of defect states, where the temperature-dependent capacitance provides information on the trap depths. [45][46][47][48][49] These techniques are all highly useful for detecting defect states, but they are not without limitations. For example, PDS provides a measurement of sub-bandgap absorbance, but no information on whether the low energy features arise from states near the HOMO (hole traps) or LUMO (electron traps). Both time-dependent charge extraction measurements and TAS are highly useful; however, the data analysis and interpretation can be difficult for low mobility semiconductors, where trap state emission may not be the slowest step in carrier generation and free carrier response can be misinterpreted as a defect band. [45,48,49] An advantage of ER-EIS is that the energy of the trap state distributions relative to a known reference electrode can be directly measured, thereby providing a complementary measurement to probe defect state energies in OSCs.
The main purpose of this work is to provide a comparison between photoemission spectroscopy measurements, cyclic voltammetry, ER-EIS, and PDS for determination of transport state energies and defect states in both π-conjugated polymers and molecular OSCs. In this work we examine three π-conjugated polymers and one prototypical molecular OSC using this suite of characterization techniques. Consistent with previous literature, [50][51][52][53] we find that the extracted IEs and EAs differ between electrochemical methods and photoemission spectroscopies. We show that the combination of ER-EIS and PDS is highly insightful for probing sub-bandgap states and identifying their energies with respect to the HOMO and LUMO bands.

Results and Discussion
The basic principles of UPS, IPES, and ER-EIS are shown schematically in Figure 1, with Figure 1b showing the equivalent circuit used to interpret the ER-EIS data. In the equivalent circuit model adopted from Schauer, et al. [7] R S is the electrolyte resistance, C dl and R CT are the double-layer capacitance and the charge transfer resistance related to the interface between the electrolyte and the OSC, respectively, and C bulk and R bulk are the capacitance and resistance of the bulk, respectively.

www.advmatinterfaces.de
Energy-resolved EIS can be used to extract the DOS in the OSC through the bulk capacitance (C bulk ). Additionally, Schauer proposed that the charge transfer resistance (R CT ) between the electrolyte and the semiconductor can be used to extract the DOS near the interface between the organic semiconductor and the electrolyte. [7,8] In ER-EIS the electrochemical potential (or Fermi level) in the semiconductor is controlled by applying an external DC voltage, hence the desired DOS as a function of instantaneous Fermi level (E F ) can be evaluated by: in which S is the area of the organic semiconductor, d is the thickness of the organic film, and e is the elementary charge.
In this work, we primarily use C bulk to obtain the density of states. Unlike EIS where the AC frequency is scanned over several orders of magnitude to extract the different capacitance and resistance values through fitting of the measured impedance to an equivalent circuit model, ER-EIS measurements are performed at a single frequency that is selected to simplify the circuit model in such a way that R S , R CT , and C bulk are the dominant components in the circuit (as highlighted in red in Figure 1b) and R S << R CT ; therefore, the real part of the measured complex impedance represents the charge transfer resistance and the imaginary part provides the capacitance of the semiconductor. This simplified single-frequency analysis allows C bulk and R CT to be extracted at varying DC biases over the course of a 3-hour measurement. Figure 2 shows the measured C bulk and 1/R CT for a regioregular-poly(3-hexylthiophene), RR-P3HT, film, along with the extracted density of states from these measurements. On a logarithmic scale, there is a clear sub-bandgap distribution of occupied states with a maximum that is shifted by 0.38 eV relative to the center of the HOMO peak. This corresponds exactly with previous reports that used TAS to determine the defect band position in RR-P3HT and RR-P3HT:PCBM blends. [46,47] The intensity of this distribution varies between the C bulk and 1/R CT terms, with the larger magnitude for 1/R CT suggesting that these states are more localized near the film surface. Similar sub-bandgap states were also observed by Schauer and colleagues using ER-EIS, [8][9][10] where the larger intensity for 1/R CT was attributed to differences in crystallinity of the P3HT film at the film surface as compared to the bulk. The same data can also be plotted on a linear scale, as shown in Figure 2b to allow the assignment of the IE and EA through determining the intersection between a linear fit to the leading edge of the ER-EIS signal and the background.
Ultraviolet and inverse photoemission spectra, cyclic voltammograms, and C bulk extracted from the ER-EIS measurements for RR-P3HT are displayed in Figure 3 on semi-log and linear plots, respectively. The greatly enhanced sensitivity of ER-EIS is readily apparent in Figure 3a. Here, the ER-EIS measurements span over six orders of magnitude in intensity, whereas CV spans three to four orders of magnitude and UPS and IPES less than two over the energy ranges presented. Several differences in addition to the noise floor are apparent when comparing ER-EIS and CV. The first is that the slope of the LUMO onset is www.advmatinterfaces.de less for CV than for ER-EIS, which is particularly pronounced when viewing the onsets on the semi-log plot. Additionally, the IE and EA differ for the CV and ER-EIS spectra, with the CV values displaying a higher IE, lower EA, and thereby an increased bandgap relative to ER-EIS. These differences may be partly attributed to the kinetically limited nature of CV versus the more steady-state nature of ER-EIS. For example, in CV the scan rate was 100 times faster than the DC voltage scan rate in the ER-EIS measurements. UPS and IPES measurements have the lowest sensitivity and display even more gradual slopes to the HOMO and LUMO onsets. In part, the more gradual slopes are due to the relatively low instrumental resolution in IPES of ≈400 meV (as detailed in Figure S1, Supporting Information). In addition to the lower sensitivity and decreased resolution, the UPS and IPES measurements also show a different IE and EA relative to those extracted with CV or ER-EIS, as presented in Table 1 and further discussed in proceeding paragraphs.
The amorphous analogue of RR-P3HT, regiorandom-P3HT (RRa-P3HT), was also investigated using the three methods (Figure 4 and Figures S3-S5, Supporting Information). Comparing the ER-EIS data from RR-P3HT to RRa-P3HT, a major difference is that RRa-P3HT does not show the sub-bandgap states at the edge of the HOMO band, but instead shows a welldefined distribution of sub-bandgap states extending beyond the edge of the LUMO band centered at ≈−2.85 eV versus vacuum.
Although both P3HT derivatives clearly show the presence of defect bands, ER-EIS shows that the positions of these defect distributions are completely different. Thus, the nature and origin of the defect states between RR-and RRa-P3HT are different. Our results are in contrast with TAS extracted trap densities, where similar trap state distributions near the HOMO edge were observed for both RR-P3HT and RRa-P3HT, [48] with an even greater trap state density for RRa-P3HT than RR-P3HT. It may be that the RRa-P3HT materials used here and in the work of Munatsir and Chaudhary contain different types of defect states. [48] Another possibility is that in the TAS measurements the defect bands with RRa-P3HT are actually from trapping of electrons in defect states residing near the LUMO, which appears reasonable given the use of a low work function calcium cathode. In future work, direct comparison between TAS and ER-EIS measurements could be performed for more definitive determination of the position of defect bands.
Greater differences between ER-EIS and CV-determined IEs and EAs are observed for RRa-P3HT as compared to RR-P3HT. In the CV of RRa-P3HT both the HOMO and LUMO onsets are shifted by ca. 0.5 and 0.3 eV, respectively, as compared to the ER-EIS onsets. Furthermore, the CV-determined bandgap is 0.21 eV greater than that determined by ER-EIS. To determine if these differences could be attributed to the faster scan rates used with CV, we took CV measurements with varying scan rates  www.advmatinterfaces.de from 10 to 100 mV s −1 , as shown in Figure S2 (Supporting Information). The scan rate-dependent results indicate that the IE extracted from CV shifts as the scan rate is varied. This is in contrast to RR-P3HT, where the IE remains fairly constant over this same range of scan rates, with only a 35 mV shift. We suggest that this discrepancy arises due to faster ion diffusion in RR-P3HT as compared to RRa-P3HT, with the slow ion diffusion kinetically limiting the current to a large extent in RRa-P3HT. Such differences in ion diffusion agree with previous investigations of ionic dopant diffusion, where dopant diffusion is slower in amorphous polymers as compared to semicrystalline polymers. [54,55] The polymer PDPP4T is investigated as a model donoracceptor (D-A) polymer for comparison between CV, ER-EIS, and UPS/IPES (Figure 4b). Here, PDPP4T shows a large deviation between the CV and ER-EIS spectra in the oxidative region. In ER-EIS the oxidation onset is at 5.15 eV, whereas this onset shifts by 0.39 eV to 5.54 eV for CV. At scan rates between 10 and 100 mV s −1 ( Figure S2f, Supporting Information) there is minimal change in the onset of the CV, which is similar to what is observed for RR-P3HT. The ER-EIS spectra show sub-bandgap states extending beyond both the occupied and unoccupied distributions of states; however, the density of these states are 3 to 4 orders of magnitude lower than the primary HOMO and LUMO peaks.
To determine if ER-EIS could also be applied to molecular organic semiconductors, we examined 6,13-bis(triisopropylsil ylethynyl)pentacene, TIPS-pentacene, thin films. The ER-EIS The EA is calculated by determining the linear onset to the IPES signal and adding one-half of the instrumental resolution, as originally proposed by Yoshida. [59]  www.advmatinterfaces.de spectra are depicted in Figure 4c and indeed show that this method can be effectively applied to molecular OSCs, with the caveat that the probing range is less than for the polymers. We suspect that the more limited probing range is associated with the difficulty in electrolyte penetration into the highly crystalline film of TIPS-pentacene. Such an explanation is supported by the large shift in the oxidation onset with varying scan rate. Furthermore, pinholes in the TIPS-pentacene film could also be lowering the accesible range. Similar to the polymers, TIPSpentacene also shows a distribution of sub-bandgap states that extends beyond the HOMO edge.
A comparison of the optical gaps and transport gaps measured with the different methods are displayed in Table 1. Comparing the transport gaps extracted with CV and ER-EIS, the transport gaps measured with CV are between 0.1 and 0.3 eV greater than those measured with ER-EIS. This discrepancy can be in part attributed to the more kinetically limited reduction and oxidation processes when using CV, as supported by the scan rate-dependent CV measurements presented in Figure S2 (Supporting Information). The transport gaps extracted from ER-EIS are more similar to the UPS/IPES extracted gaps than are the CV measurements for the polymers. For example, the transport gaps measured by ER-EIS are between 0.04 eV (PDPP4T) and 0.29 eV larger (RR-P3HT) than the UPS/IPESdetermined gaps for the same materials. Whereas, the transport gaps measured by CV are 0.36 to 0.49 eV larger than the UPS/ IPES determined gaps for the polymers. On the other hand, in TIPS-pentacene UPS/IPES yields a larger gap than the ER-EIS and CV measurements. Linear fits to the low-energy edge of the UV-Vis absorbance data for the films were used to determine the optical gap, [56,57] as displayed in Figure S4 (Supporting Information). In all cases, the transport gaps determined by UPS and IPES are larger than the optical gaps by between 0.11 and 0.51 eV, while the transport gaps determined by ER-EIS are 0.09 to 0.55 greater than the optical gaps. This general trend is in line with recent ER-EIS measurements and in agreement with theory, [58] as the optical gap should be less than the transport gap, with the difference being equal to the exciton binding energy. [21] A well-established method for investigating disorder and subbandgap states in organic semiconductors is through sensitive UV-vis-NIR absorbance spectroscopies, such as PDS. [41,42] A disadvantage of PDS, however, is that the method does not provide any information as to where the sub-bandgap states are located with respect to the HOMO and LUMO, and the measurement depends on the absorbance cross-section of the sub-bandgap states. The DOS extracted from the ER-EIS spectra as well as the corresponding PDS spectra for each material are presented in Figure 5. The PDS measurements probe an absorbance range of approximately four orders of magnitude, which is in the relative intensity range of the sub-bandgap states in the ER-EIS spectra. Comparing the ER-EIS and PDS spectra of RRa-P3HT, there is a pronounced sub-bandgap feature that is centered 0.36 eV away from the LUMO peak in the ER-EIS and 0.51 eV away from the center of a Gaussian peak fit to the primary absorbance edge of the PDS spectra. Furthermore, this pronounced peak has a similar shape and width, 0.09 eV, in both the ER-EIS ( Figure S5, Supporitng Information) and PDS spectra, and it is accompanied by a gradually diminishing distribution of sub-bandgap states in both the ER-EIS and PDS spectra. These strong similarities indicate that the same defect states are appearing in both the ER-EIS and PDS spectra. TIPS-pentacene also displays a relatively intense tail of sub-bandgap states near the HOMO edge in the ER-EIS spectra, which matches well with the low energy shoulder in the PDS spectrum. The sub-bandgap absorbance for RR-P3HT also shows a low-energy tail that appears to correspond with the sub-bandgap distribution near the HOMO edge of the ER-EIS spectra. This comparison clearly shows that ER-EIS can be used to complement PDS or other optical methods to identify both the intensity of sub-gap states as well as the relative positions of these states within the gap.

Conclusion
The discrepancy between CV, ER-EIS, and UPS/IPES-determined IEs, EAs, and bandgaps is a reasonable and previously reported phenomenon. In part, this discrepancy arises because the measurements are measuring fundamentally different processes and the measurements are conducted in different environments. Measured HOMO and LUMO values with ER-EIS are closer to UPS/IPES determined values, but the much greater complexity and time required for the measurement may not warrant performing ER-EIS as opposed to CV with lower scan rates for rough IE and EA approximations. The primary advantage of ER-EIS is that it is sensitive enough to Figure 5. ER-EIS spectra (a) and PDS spectra (b) on semi-log plots with colored arrows highlighting the correspondence between defect states in the ER-EIS spectra and sub-bandap absorption feature in the PDS spectra.
www.advmatinterfaces.de detect sub-bandgap defect states. The ER-EIS measurements performed here indicate that RR-P3HT and RRa-P3HT contain distinctly different defect states, with the primary defect band in RR-P3HT occuring near the HOMO band and the main defect band in RRa-P3HT occuring near the LUMO band. The ER-EIS measurements show good agreement with PDS data, which further confirms the presence of these sub-bandgap states and emphasizes the power of ER-EIS in probing subbandgap defect states. We expect ER-EIS to fill an important role in investigating defect state chemistry and degradation products in OSCs.
Indium-doped tin oxide (ITO) substrates were cleaned following a consecutive series of sonication cycles (10 min each) with diluted detergent (sodium dodecylsulfate), deionized water, acetone, and isopropanol. Finally, the substrates were treated with UV-ozone for 20 min before film deposition. Polymer films were spun cast from the solutions described above onto ITO-coated glass substrates in an inert gas (N 2 ) environment at 500 RPM for 50 s followed by thermal annealing at 110 °C for 7 min. Samples of P3HT were also solvent annealed for 30 min preceding thermal annealing. To prepare a more uniform film from the small molecule, the TIPS-pentacene solution was spin cast at 500 RPM for only 8 s while blowing nitrogen on the film, then the film was annealed on a hot plate at 80 °C for 10 min. Films for PDS measurements were prepared using similar procedures and the films cast on clean quartz substrates. For absorbance measurements RRa-P3HT and RR-P3HT thin films were prepared at 4000 rpm for 40 s followed by annealing at 110 °C for 5 min, PDPP4T thin films were made at 3000 rpm for 50 s followed by annealing at 110 °C for 5 min, and TIPS-pentacene was spun at 3000 rpm for 8 s and annealed at 80 °C for 10 min.
Electrochemical Measurements: A conventional three-electrode setup was used with an EDAQ ET072-1 Ag/AgCl reference electrode and a platinum wire counter electrode held in a bottom mount electrochemical cell (Redoxme) with a nominal sample exposure area of 0.2 cm 2 . A solution of 0.1 m TBAPF 6 in acetonitrile was used as the supporting electrolyte. The Ag/AgCl reference electrode energy was converted to 4.45 eV versus vacuum based on measurements carried out on ferrocene and treating ferrocene as equal to 5.10 eV versus vacuum, [30] and all data were shifted accordingly to obtain the potential versus vacuum.
All electrochemical measurements were performed with an Ametek VersaSTAT 4 potentiostat on freshly made samples. To eliminate artifacts associated with electrolyte or polymer degradation at higher bias voltages, ER-EIS measurements were carried out on two samples for each experiment, one for positive potential (oxidative) and the other for negative potential (reductive) sweeps. Cyclic voltammetry measurements were performed at a scan rate of 0.1 V s −1 starting from zero potential versus the Ag/AgCl reference. The suitable frequency for ER-EIS measurements was chosen to be 0.5 Hz based on work carried out by Schauer and colleagues, and a small AC amplitude of 50 mV was applied over the DC voltage ramp with sweep rate 1 mV s −1 .
Film Characterization: Ultraviolet photoemission spectroscopy measurements were carried out in a PHI 5600 UHV system with a hemispherical electron energy analyzer with a 5.85 eV pass energy, negative 5 V sample bias, and an Excitech H Lyman-α lamp (E-Lux 121, 10.2 eV emission) with a dry oxygen purge of the beam path at ≈8 Torr used as an excitation source. [27] The IPES measurements were collected using a Kimball Physics ELG-2 electron gun with a low-temperature (1150 K) BaO cathode for excitation and a photon detector consisting of a fused silica lens, optical bandpass filter (254 nm, Semrock), and a photomultipler tube (R585, Hamamatsu Photonics). A negative 20 V bias was applied to the sample during IPES measurements.
PDS measurements were carried out using chopped monochromatic light from a 300 W Xe light source coupled with a monochromator with ±4 nm wavelength resolution. The monochromatic light was focused onto the sample surface to cause a periodic temperature change in the focal spot region. The sample was immersed in Fluorinert FC72 and the periodic temperature change of the sample resulting from the absorbance of the chopped monochromatic light caused a corresponding temperature and index of refraction change in the Fluorinert. A CW probe beam from a HeNe laser passed through the Fluorinert parallel to the sample surface, with deflection resulting due to temperatureinduced variations in the index of refraction. The beam deflection was monitored by a quadrant photodiode connected to a lock-in amplifier to measure the periodic deflection. The excitation wavelength was scanned to generate the PDS spectra.
UV-vis absorbance spectroscopy measurements were performed outside of the glove box using an Agilent Technologies Cary 5000 UV-vis NIR spectrophotometer with an integrating sphere. The film thicknesses were measured using an Asylum Dektak profilometer.

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