- Email: [email protected]
Energy band alignment in operando inverted structure P3HT:PCBM organic solar cells
Nano Energy 40 (2017) 454–461
Contents lists available at ScienceDirect
Nano Energy journal homepage: www.elsevier.com/locate/nanoen
Full paper
Energy band alignment in operando inverted structure P3HT:PCBM organic solar cells
MARK
Qi Chena, Fengye Yea,b, Junqi Laia, Pan Daic, Shulong Luc, Changqi Mad, Yanfei Zhaoe, Yi Xieb, ⁎ Liwei Chena,e, a
i-Lab, CAS Center for Excellence in Nanoscience, Suzhou Institute of Nano-Tech and Nano-Bionics, Chinese Academy of Sciences, Suzhou 215123, China Hefei National Laboratory for Physical Sciences at the Microscale and Synergetic Innovation Center of Quantum Information & Quantum Physics, University of Science and Technology of China, Hefei 230026, China c Key Laboratory of Nanodevices and Applications, Suzhou Institute of Nano-Tech and Nano-Bionics, Chinese Academy of Sciences, Suzhou 215123, China d Division of Printed Electronics, Suzhou Institute of Nano-Tech and Nano-Bionics, Chinese Academy of Sciences, Suzhou 215123, China e Vacuum Interconnected Nanotech Workstation (Nano-X), Suzhou Institute of Nano-Tech and Nano-Bionics, Chinese Academy of Sciences, 215123 Suzhou, China b
A R T I C L E I N F O
A B S T R A C T
Keywords: Cross-sectional scanning Kelvin probe microscopy Inverted structure organic solar cells Interlayer Energy band alignment Deconvolution
Inverted structure thin-film organic solar cells (OSCs) are becoming increasingly important as they deliver higher power conversion efficiency and demonstrate better long-term stability than conventional devices. However, the energy band alignment and the built-in field across the device, which are crucial in understanding the device operation, is yet to be directly characterized. Here we present a direct visualization of the energy level alignment in operando inverted structure poly(3-hexylthiophene) (P3HT): [6,6]-phenyl-C61-butyric acid methyl ester (PCBM) OSCs using cross-sectional scanning Kelvin probe microscopy. The raw data of measured energy level alignment appear to be inconsistent with each other, and sometimes can even be contradictory to the device polarity observed in current density-voltage measurements. It is identified to be caused by the tip/cantilever induced convolution effect, which may severely mask abrupt energy level offsets at the thin electrode interlayers. A numerical deconvolution method is devised to quantitatively recover the energy level alignment across the device, and reveals the non-uniform electric field distribution in photoactive layer.
1. Introduction Inverted structure organic solar cells (OSCs), with holes extracted from the top electrode and electrons from the bottom electrode, have the reverse polarity from conventional structure OSCs. In spite of the relatively short history of development, inverted structure devices are becoming increasingly important due to their high power conversion efficiency (PCE), good long-term device stability and the adaptability to tandem device structures [1–5]. The energy band depth profile of inverted structure devices has its unique features compared to that of conventional structure devices. The thin electrode interlayers in the inverted devices are expected to introduce abrupt energy level offsets in the depth profile. Substantial research efforts have been made to analyze energy levels at the active layer/interlayer interface, and the results demonstrated that interfacial doping/dipole is one of the most important factors that lead to non-uniform electric field distribution and determine the charge carrier transport [6–10]. However, characterizations of the energy band depth profile across the entire inverted structure devices are still rarely seen in literature. One such effort ⁎
actually reported an energy band depth profile whose direction of the built-in field was opposite to the device polarity measured from J-V curves [11]. These conflicting results require direct measurements for clarification. Cross-sectional SKPM is a powerful tool in measuring vacuum level (VL) depth profile in thin-film devices [11–15]. We have previously used this technique to visualize the energy level alignment in operando conventional structure OSCs [16]. It was discovered that the finite tip size and the cantilever beam crosstalk result in a convolution effect, which causes an apparent smoothing and averaging in measured surface energy depth profiles [17–19]. Here we further present an improved in-operando cross-sectional SKPM study on inverted structure poly(3-hexylthiophene) (P3HT): [6,6]-phenyl-C61-butyric acid methyl ester (PCBM) OSCs (Fig. 1) to reveal the true energy band alignment and built-in field distribution. It is observed that the same tip/cantilever convolution effect is highly significant in inverted structure device measurements, to a degree that under certain conditions that the abrupt energy level offsets introduced by interlayers could be completely masked and the measured VL profile
Corresponding author at: i-Lab, CAS Center for Excellence in Nanoscience, Suzhou Institute of Nano-Tech and Nano-Bionics, Chinese Academy of Sciences, Suzhou 215123, China. E-mail address: [email protected] (L. Chen).
http://dx.doi.org/10.1016/j.nanoen.2017.08.050 Received 2 July 2017; Received in revised form 24 August 2017; Accepted 27 August 2017 Available online 30 August 2017 2211-2855/ © 2017 Elsevier Ltd. All rights reserved.
Nano Energy 40 (2017) 454–461
Q. Chen et al.
spectra were measured by a Merlin (Model No. 70104) Digital Lock-in Radiometry System using a 300-W xenon lamp and a 74125 Oriel Cornerstone 260 1/4m monochromator with order-sorting filters (Newport). The light intensity at different wavelengths was modulated through an aperture and calibrated using a 70714 UV-enhanced Si photodiode (Newport). All of the measurements were performed under ambient atmosphere at room temperature. 2.2. MBE reference sample GaInP (Ga:In 0.51:0.49 for lattice match)/GaAs heterojunction on p++ GaAs wafer was grown by a Veeco GEN20A dual-chamber all solid-state MBE equipped with valved phosphorous and arsenic cracking cells. The typical growth rate of GaAs and GaInP was 1 µm/h, and the growth temperatures of the GaAs and GaInP were 580 °C and 510 °C respectively. Be was used as p-type doping source and doping concentration was 2.2 × 1018 cm−3 in GaAs and 8.0 × 1017 cm−3 in GaInP. The group-III mole fraction (In and Ga in this case) was calibrated by a combination of beam flux gauge and X-ray diffraction (XRD) measurements to satisfy lattice match. 2.3. Device cross-section preparation and characterization The device cross-section was fabricated using an Ilion+ 693 System (Gatan Inc.). A freshly mechanically cleaved device was mounted into the vacuum chamber (6.4 × 10−5 Torr) and cooled using liquid nitrogen. Then, the device was milled by argon ions using a beam voltage of 5 keV and a beam current of 10 μA for ~ 2 h. Five cross-section samples were prepared for thin interlayer and thick interlayer devices respectively, the error bars of PCE before and after cross section preparation, and after SKPM measurement are listed in Table S1. Stability of typical cross-sectioned devices was tested under continuous light illumination in the glove box. The morphology of the device cross section was characterized using FEI Quanta 400 FEG SEM (FEI Corp., Hillsboro, OR, USA). The SKPM surface potential (SP) profile measurements of device cross-sections were carried out using a Park XE-120 AFM (Park Systems Corp., Suwon, Korea) using Cr/Au-coated conducting tips (NSC18, Mikromasch, Tallinn, Estonia) with a resonance frequency of ~ 80 kHz and a spring constant of ~ 2 N m−1. Amplitude modulation (AM) mode SKPM [23,24] was carried out in a nitrogen gas filled glove box. During the first pass, standard AC mode imaging (typical tip oscillation amplitude 20 nm) was performed to acquire the topography and phase signal of the sample; in the second pass, the tip was lifted up by a certain height (typically 10 nm) and scanned on the basis of the topography line obtained from the first pass. An AC voltage (3 V in amplitude and 10 kHz in frequency) was applied to actuate the cantilever, and the DC voltage applied to the tip that nullifies the tip-sample interaction was collected as the SP signal. The VL profile can be obtained by multiplying the measured SP profile with the absolute electron charge (at least 10 scan lines are included for averaging) [25,26]. Fig. 1a shows the configuration of device-wiring during cross-section characterization. The Al electrode was grounded. The ITO electrode was connected with the Al electrode in short-circuit; open-circuit conditions were simulated by disconnecting the wire between the two electrodes. A full-spectrum optical fiber was used to transmit AM 1.5G solar simulator light to illuminate the device from the ITO glass side. Bias voltage was applied via a tunable voltage source between the electrodes.
Fig. 1. (a) Schematic illustration of SKPM measurements of the vacuum level depth profile of operando cross-sectional devices. (b) The SKPM measured profiles are the convolution of the true profile and the tip transfer function. The true profile can be recovered via numerical deconvolution.
may appear to be contradictory to the device polarity. To resolve this issue, a deconvolution process including a calibration on a molecular beam epitaxy (MBE) grown GaInP/GaAs reference sample and a numerical deconvolution calculation is devised to negate the tip and cantilever induced convolution effects in SKPM. The true energy band alignment in inverted OSCs has thus been recovered for the first time. The results confirm that the work function of the interlayers and interfacial effects are critically important in determining the energy level alignment and leading to a non-uniform electric field across the active layer in inverted structure OSCs.
2. Experimental section 2.1. Device fabrication and characterization The inverted structure OSCs with a stacking of ITO/ZnO/P3HT: PCBM/MoOx/Al were fabricated by sequentially cleaning pre-patterned ITO glass substrates (15 Ωsq−1) using ultrasonication for 10 min in a detergent, deionized water, ethanol, acetone and isopropyl alcohol, followed by treatment with oxygen plasma for 10 min. Then, ZnO nanocrystal solution synthesized according to our previous reports was spin-coated onto ITO substrates and annealed at 150 °C for 20 min [20]. The P3HT (1-Material Inc.): PCBM (American Dye Source Inc.) blend (36 mg/ml, 1:0.8 w/w) was dissolved in 1,2-dichlorobenzene and stirred at 60 °C for 14 h in a glove box. The blend active layer was prepared by spin-coating at 600 r.p.m. for 60 s, and annealed at 130 °C for 10 min. A 20-nm or 80-nm MoOx layer and 100-nm Al layer were subsequently evaporated under a pressure of 4 × 10−4 Pa through a shadow mask to define the active area of the device (0.12 cm2) and to form a top anode. The J–V characteristics of the devices were recorded using a Keithley 2635 A source meter (Keithley Instruments Inc., Cleveland, OH, USA) under one sun, AM 1.5 G irradiation (100 mW cm−2) from a solar simulator (Newport 67005, Newport, Irvine, CA, USA). The illumination intensity was calibrated using a reference cell (Oriel 91150V). The light-soaking effect was checked by applying a UV filter, which was not significant here [21,22]. External quantum efficiency (EQE)
2.4. Numerical convolution and deconvolution simulation Numerical simulations were performed using Wolfram Mathematica 10 with home-written codes. The SKPM measurements were simulated as a convolution of the true SP profiles and the tip transfer functions with noise. Tip transfer functions were obtained by adjusting the σ and A parameters to make the simulated (i.e. convoluted) profile of a 455
Nano Energy 40 (2017) 454–461
Q. Chen et al.
batch (NSC18 Cr/Au, Mikromasch), named Tip 1 and Tip 2, respectively. Comparison of the extracted SP profiles (Fig. 2d) yields stark differences. The profile obtained with Tip 1 exhibits a continuous increase in SP from the ITO electrode to the Al electrode. This behavior is similar to that in conventional structure device [11,13,16], even though the device polarity is completely the opposite. This has been reported in a previous literature [11], but the discrepancy between the cross-sectional SKPM and device polarity was not commented or resolved. Interestingly, the profile obtained with Tip 2 displays a shallow kink in the middle of the active layer. The built-in field of the kinked region is the opposite from that in conventional structure devices, and thus it agrees with the device polarity but contradicts with that obtained with Tip 1. The apparent differences in measured SP profiles could be related with variations in the tip geometry used in SKPM. By imaging an array of sharp Si spikes of a test grating (TGT1, NT-MDT), it is revealed that the radius of curvature of Tip 2 (21 nm) is significantly smaller than that of Tip 1 (35 nm) (Fig. 2e-f). Furthermore, the measured SP profiles are also sensitively dependent on the tip-sample distance in the SKPM measurement. Shown in Fig. 2g is a SP image obtained with Tip 2 at different lift heights. The corresponding SP profiles (Fig. 2h) show a transition from the one with the kink at the lowest lift height to those with monotonic increase at higher lifts. These results agree well with our previous study that the cross-sectional SKPM of thin film devices are severely affected by tip/cantilever convolution effects [16]. The convolution effect due to finite tip size and cantilever crosstalk in SKPM has been extensively documented [17–19,27,28,33]. The effect is mathematically accounted by the convolution of the true potential profile with a tip transfer function, typically in a Gaussian form, f(x) = Aexp(x2/2σ2), in which σ determines the widening and A determines the contrast smoothing in the convolution [17,19,27,28,33]. To recover the true potential profile from the convoluted measurements, accurate tip transfer functions are needed for deconvolution. Previous efforts in determining the tip transfer function usually involved modeling based on vendor specified tip and cantilever geometry. However, since there exists significant tip-to-tip variation in commercial SKPM probes, and as we have shown above that the convolution effect is highly sensitively dependent on measurement parameters, a more precise approach is needed to reliably gauge the transfer function. Here we demonstrate a deconvolution routine involving calibration
reference sample to overlap with the measured profile. Then, the true profiles of samples under study can be recovered by Wiener deconvolution in frequency domain. Here, the system noise is a white Gaussian noise, which is negligible in convolution and will be disproportionately amplified at high frequencies by direct deconvolution. Thus, a Weiner filter in frequency domain (TTF)−1/[1 + (SNR × |TTF|2)−1] (in which TTF is the tip transfer function and SNR is the signal-to-noise ratio) was employed to minimize the impact of deconvolution noise at high frequencies, which have a poor signal-to-noise ratio [27,28].
3. Results and discussion The current density-voltage (J-V) curves and EQE curves of the inverted structure device with optimized interlayer thickness (20 nm ZnO and 10 nm MoOx) are shown in Fig. S1. The PCE before and after crosssection preparation are 3.93% and 3.39%, respectively (device performance parameters detailed in Table S1), and PCE after SKPM measurement is 3.15%, indicating that the cross-sectioned device is relatively stable during the SKPM measurement. The stability of a typical cross-sectioned device under ~ 10 h continuous light illumination is shown in Fig. S1. The Voc is very stable at first (< 1% drop after ~ 3 h), and then it decreases slightly (~ 5% drop after 10 h). The stable Voc in early stage may attribute to the optimized crystallinity during device preparation [29]. With long-term light illumination, the crystallinity decreases gradually and it leads to broadened density of states (DOS), which reduces the splitting of the quasi-Fermi levels (i.e. Voc decreases) [29]. The degradation in PCE (~ 19% drop after 3 h and ~ 36% drop after 10 h) is mainly attributed to the decrease in Jsc (~ 10% drop after 3 h and ~ 24% drop after 10 h) and FF (~ 8% drop after 3 h and ~ 12% drop after 10 h). The decrease in Jsc and FF can be explained by the change in morphology that large domains grow gradually in aging [30–32]. The cross-sectional SKPM measurement was mostly completed within the first two hours of continuous illumination. The device Voc exhibits little change in this period; and thus the energy band alignment measurement is reliable. The scanning electron microscopy (SEM) backscattered electron image displayed in Fig. 2a shows a cross section with clear contrast between different stacking layers. Fig. 2b and c display two cross-sectional SKPM measured SP images of the same device at roughly the same location in dark acquired by two different tips from the same
Fig. 2. (a) SEM image of the cross-sectional device with layer stacking of ITO/ZnO(20 nm)/P3HT:PCBM/MoOx(10 nm)/Al (scale bar: 300 nm). (b-c) SP images of the device in dark state with two tips: Tip 1 (b) and Tip 2 (c) (scale bar: 200 nm). (d) SP profiles extracted from (b) (black square for Tip 1) and (c) (red circle for Tip 2). (e) Schematic illustration of TGT1 for tip geometry calibration. (f) The profile of the spike at its apex. The inset shows the entire profile of the spike. (g-h) The SP image and the extracted SP profiles of the same device in dark state with Tip 2 under different tip-sample distance from 20 nm to 120 nm (scale bar: 200 nm).
456
Nano Energy 40 (2017) 454–461
Q. Chen et al.
Fig. 3. (a) SEM image of GaInP/GaAs heterojunction (scale bar: 1 µm). (b-c) SP images of GaInP/GaAs heterojunction acquired by Tip 1 and Tip 2 (scale bar: 250 nm). (d) SP profiles extracted from (b) (black square) (c) (red circle) and compared with the calculated SP profile. The green triangle is the SP profile acquired by Tip 2 at + 0.25 V bias voltage. All these profiles are acquired under tip-sample distance of 20 nm. (e) The transfer function of Tip 1 and Tip 2 derived from numerical convolution fitting. (f) Deconvoluted SP profile based on the transfer function of Tip 1 and Tip 2. (g) SP profiles of GaInP/GaAs heterojunction acquired by Tip 2 under different tip-sample distances from 20 nm to 120 nm. (h) Tip transfer function of Tip 2 under different tip-sample distances derived from numerical convolution fitting. (i) Deconvoluted SP profiles based on the transfer function of Tip 2 under different tip-sample distances.
the GaAs is visible. The high doping concentration used here is important to get a narrow depletion layer and create a steep VL shift at the interface. Fig. 3b and c show the SP images measured with the Tip 1 and Tip 2 at tip-sample distance of 20 nm, respectively. The profiles displayed in Fig. 3d show flat SP at both the GaInP and the GaAs sides, with a sloped step at the interface. The SP profile acquired by Tip 2 (red open circle) has a greater SP difference between the two sides and a steeper slope at the interface than that acquired by Tip 1 (black open square). Interestingly, when a bias voltage of +0.25 V is applied onto the GaInP side against the GaAs side, the SP profile acquired by Tip 2 appears to be flat across the interface (the open triangles in Fig. 3d), regardless of different lift heights in the measurements (Fig. S3). According to the bias compensation method developed in our previous publication [16], this indicates that the VL shift at the GaInP/GaAs interface is ~ 0.25 eV. A SP profile of the reference sample is thus calculated based on an abrupt isotype heterojunction model [36,37]. The width of the transition region is interpreted as the sum of the band bending zones on both side of the heterojunction, i.e., the width of the sloped step in the SP profile is found to be 32.2 nm (Fig. S2 in Supporting information). Numerical convolution and deconvolution calculations are implemented to derive tip transfer functions based on the measurements on the reference sample. The σ and A of tip transfer functions are
Fig. 4. Deconvoluted SP profiles of the device with thin interlayer in Fig. 2 based on the transfer functions of Tip 1 and Tip 2.
over a reference sample with a computable VL profile (Fig. 1b). A MBE grown GaInP (Ga:In 0.51:0.49)/GaAs heterojunction is used as the reference sample [34,35]. The materials interface in MBE grown heterojunction is typically 1–2 atomic layers wide, therefore, the particular sample composed of a 1 µm thick p++ GaAs epitaxial layer on p++ GaAs substrate, followed with 1 µm thick p++ GaInP epitaxial layer MBE heterojunction is an ideal abrupt junction. Fig. 3a shows the SEM backscattered electron image of the cross-section of the GaInP/GaAs abrupt heterojunction, in which the interface between the GaInP and 457
Nano Energy 40 (2017) 454–461
Q. Chen et al.
Fig. 5. (a) SEM image of the ITO/ZnO(80 nm)/P3HT:PCBM/MoOx(80 nm)/Al device cross-section (scale bar: 300 nm). (b-c) SP images of the device in dark state (b) and under light illumination (c) in open-circuit condition (scale bar: 400 nm). (d) SP profiles extracted from (b) (black square) and (c) (red circle), and the SP profile at +0.6 V bias in dark (blue line). (e) The transfer function of Tip 3 derived from numerical convolution fitting. (f) Deconvoluted SP profiles from (d): black line for the dark and red line for the illuminated state of the device.
adjusted to make the convoluted profile overlap with the measured profile. For the measurements with tip-sample distance of 20 nm, the σ and A for Tips 1 and 2 are yielded as 74 nm and 1.67 µm−1 for Tip 1, and 35 nm and 4.96 µm−1 for Tip 2, respectively (Fig. 3e). With these parameters, the corresponding Gaussian functions are used as the transfer function in the numerical deconvolution calculation. Fig. 3f shows that although the SP profiles of the reference sample measured with Tips 1 and 2 are different, the two deconvoluted SP profiles essentially overlap with each other and agree well with the calculated SP profile. This corroborates the literature report that tip radius is one of the most important factors that determines σ and A of tip transfer function [33], and also confirms the validity of our deconvolution calculation. Furthermore, the convolution effect of tips at different lift height was also investigated. SP profiles of the GaInP/GaAs reference sample measured by Tip 2 under different tip-sample distance from 20 nm to 120 nm are shown in Fig. 3g. As the tip is lifted away from the sample, the measured profiles show less SP difference between the GaInP and GaAs region and become less steep. The convolution fitting thus results in increasing σ and decreasing A in the tip transfer function (Fig. 3h). With these transfer functions; the measured profiles are deconvoluted as shown in Fig. 3i. The deconvoluted SP profiles obtained with tipsample distance of 20 nm and 45 nm show satisfying agreement with the calculated SP profile, which further corroborates the reliability of the deconvolution process. However, if tip-sample distance is greater than 45 nm, the deconvoluted SP profiles exhibit serious oscillation in GaAs and GaInP regions that are expected to be flat in SP. This is due to lack of resolution in the primary signal and the magnification of high frequency noises in the numerical deconvolution algorithm [27,28,38]. Therefore, a tip transfer function with small σ and large A, i.e. a sharp tip and a small tip-sample distance, is preferred in order to recover the true SP profile reliably by using deconvolution. With the tip transfer function calibrated and the deconvolution process validated, the SP depth profiles of inverted structure OSCs could then be obtained. The deconvoluted and the originally measured SP profiles obtained using Tip 1 and Tip 2 are shown in Fig. 4. It is found that the SP differences between Al and ITO after deconvolution are ~ 1.36 V for Tip 1 and ~ 1.37 V for Tip 2, respectively. The values
approximately agree with literature reported 1.6 V (the work functions of ITO and Al electrodes have been separately measured to be 5.0 and 3.4 eV using ultraviolet photon spectroscopy (UPS) and Kelvin probe (KP), respectively) [39]. Importantly, both deconvoluted SP profiles exhibit a shallow kink in the active layer, indicating that energy band alignment and built-in field in the inverted structure device is in opposite direction to conventional structure devices. This result is consistent with the device concept as well as the device polarity displayed in J-V characteristics. The deconvoluted SP profile from Tip 1 shows oscillation in cathode and anode regions while that from Tip 2 shows negligible oscillation. This is again due to magnification of high frequency noise by numerical deconvolution when the primary data measured with Tip 1 lacks spatial resolution. According to the metal-insulator-metal structure (taking interlayer/ electrode as assembled entities) device model [40], the SP drops from ZnO across the active layer to MoOx can be interpreted as upper-limit of Vbi; however, the SP difference in the kinked region in the deconvoluted profiles is actually smaller than device Voc. We suspect that even with the deconvolution method, the ZnO (~ 20 nm) and MoOx (~ 10 nm) interlayers are still too thin to be accurately probed because of the large tip radius of commercial SKPM probes. In order to better quantify the energy band alignment in inverted structure devices, we intentionally increase the thickness of the interlayers to 80 nm ZnO and 80 nm MoOx. The thick interlayers are clearly visible in cross section SEM micrographs (Fig. 5a). The J-V curves and EQE curves of the device with thicker interlayers are shown in Fig. S1. The PCE before and after cross section preparation are 3.09% and 2.79% respectively (the device performance is detailed in Table S1). JSC and FF have decreased slightly due to the high series resistance resulted from the thick interlayer. On the other hand, VOC remains unchanged, suggesting the Vbi and energy band alignment of the thick interlayer device shall be very similar to those of the thin interlayer devices [41–46]. The PCE of the device after SKPM measurement is 2.61%, and the stability of a typical cross-sectioned devices under ~ 10 h continuous light illumination is shown in Fig. S1. The Voc is very stable (~ 1% drop after 3 h and ~ 6% drop after 10 h), which is similar to that of devices with optimized interlayer thickness. Fig. 5b and c are SP images of the device cross-section measured with a new tip denoted as Tip 3 in 458
Nano Energy 40 (2017) 454–461
Q. Chen et al.
compensated by VL shifts at ZnO/BHJ interface and the band bending in the BHJ (qVbi), leading to a non-uniform electric field across the active layer [7,9]. The SP profile (red open circle in Fig. 5d) of the same device in open-circuit under light illumination was extracted from Fig. 5c. A reduction in potential difference from ZnO to MoOx is observed, because the direction of electric field induced by photogenerated charge carriers accumulated at interlayer/BHJ interface is opposite to that of built-in field; therefore, the reduction in build-in field results in decreased overall potential drop from ZnO across BHJ to MoOx. By using the bias voltage compensation method [16], a bias voltage of + 0.6 V is required to overlap the biased SP profile in dark (blue line) with that in open-circuit under light illumination. This bias voltage is the same as the device VOC, which confirms the reliability of the SP profiles. The transfer function of Tip 3 is calibrated as shown in Fig. 5e with σ = 32 nm, A = 3.62 µm−1 (calibration process is detailed in Fig. S5); and the measured SP profiles are deconvoluted. The deconvoluted SP profiles in dark (black line) and under light illumination (red line) are shown in Fig. 5f. Both profiles keep the same line shape as the measured profiles but exhibit much enhanced SP contrast. After deconvolution, the SP difference between Al and ITO in dark is ~ 1.40 V, which is very close to that in thin interlayer devices (1.36–1.37 V). From the deconvoluted SP profile in dark, SP drop across the active layer, i.e. from ZnO to MoOx, is ~ 0.97 V, which approaches the work function difference between ZnO and MoOx of 0.94 V, if taking the work functions of ZnO and MoOx interlayers to be 4.26 eV and 5.20 eV, as measured using the KP technique in ambient, respectively [39,50]. It is known that work function of ZnO and MoOx varies significantly depending on sample preparation and/or measurement conditions [51], so references with sample preparation and measurement environment resembling those in SKPM measurements are deliberately chosen here. The work function of MoOx used in this work is also measured by UPS (Fig. S6), which is 5.13 eV and is close to 5.20 eV that measured by KP. Upon light illumination, the SP drop from ZnO to MoOx across the active layer is reduced to 0.39 V due to quasi-Fermi level splitting. Thus, the Voc obtained from deconvoluted SP profiles is 0.97–0.39 = 0.58 V, which is very close to the 0.6 V Voc from J-V curve measurements (deviation < 5%). Based on these measurements, the energy band alignment of inverted structure device in dark and under illumination is illustrated in Fig. 6. It highlights that the work function of the interlayers and interfacial effects such as contact-induced doping and charge transfer generated dipole at active layer/interlayer interface etc. are important factors leading to the non-uniform electric field within the active layer, which significantly impacts on charge carriers behavior and device performance [7,9]. In order to optimize the device performance parameters such as Voc, special attention needs to be paid to the band offset at interlayer/active layer interfaces in addition to the conventional design rule considering the energy level offset between the donor and acceptor.
Fig. 6. Energy band diagrams of inverted structure OSC device in dark (a) and under light illumination (b) in open-circuit condition.
open-circuit in dark state and under light illumination, respectively. In Fig. 5d, the SP profile in dark (black open square) extracted from Fig. 5b shows decreased potential from ZnO across the bulk heterojunction (BHJ) active layer towards MoOx. This result validates the presence of the kinked region in the thin interlayer device (Fig. 2d). Interestingly, SP changes more rapidly at the ZnO/BHJ interface than the BHJ/MoOx interface, which is also manifested in the more negative value in the derivative of SP profiles (Fig. S4). This indicates that there is a more significant VL shift (qΔ) induced by a greater interfacial dipole at the ZnO/BHJ interface, which probably arises from charge injection from ZnO to localized interface states of BHJ (mostly PCBM due to vertical phase segregation in bottom surface of BHJ) [47–49]. On the other hand, vertical phase segregation in BHJ results in P3HT accumulating towards the top electrode, oxidation and/or MoOx contactinduced p-doping of P3HT may lead to increased work function of P3HT, resulting in reduced work function difference between P3HT and MoOx and thus small interfacial VL shift [6,47,49]. Overall, the large work function difference between ZnO and MoOx is mainly
4. Conclusion In summary, it is demonstrated that the apparent inconsistency in cross-sectional SKPM investigations of inverted OSC devices is due to tip-to-tip variations on the probe geometry. The occasionally observed conflict between the SKPM measured built-in field and the J-V characteristics is the result of the low spatial resolution in combination with the tip/cantilever induced convolution effect, which can mask abrupt energy level offsets caused by thin interlayers. The effects of these artifacts can be minimized by calibrating the transfer function of sharp tips and performing numerical deconvolution. The energy band alignment depth profile of inverted OSCs are thus obtained for the first time via in-operando cross-sectional SKPM measurements and deconvolution calculation. The results clarify that the built-in field direction in inverted devices is consistent with the device polarity; more importantly, the work function of the interlayers and the interfacial states between 459
Nano Energy 40 (2017) 454–461
Q. Chen et al.
6 (2015) 8397. [16] Q. Chen, L. Mao, Y. Li, T. Kong, N. Wu, C. Ma, S. Bai, Y. Jin, D. Wu, W. Lu, B. Wang, L. Chen, Nat. Commun. 6 (2015) 7745. [17] A. Liscio, V. Palermo, P. Samorì, Acc. Chem. Res. 43 (2010) 541–550. [18] D.S.H. Charrier, M. Kemerink, B.E. Smalbrugge, T. de Vries, R.A.J. Janssen, ACS Nano 2 (2008) 622–626. [19] G. Elias, T. Glatzel, E. Meyer, A. Schwarzman, A. Boag, Y. Rosenwaks, Beilstein J. Nanotechnol. 2 (2011) 252–260. [20] X. Liang, Y. Ren, S. Bai, N. Zhang, X. Dai, X. Wang, H. He, C. Jin, Z. Ye, Q. Chen, L. Chen, J. Wang, Y. Jin, Chem. Mater. 26 (2014) 5169–5178. [21] W. Xu, R. Xia, T. Ye, L. Zhao, Z. Kan, Y. Mei, C. Yan, X.-W. Zhang, W.-Y. Lai, P.E. Keivanidis, W. Huang, Adv. Sci. 3 (2016) 1500245. [22] S. Trost, K. Zilberberg, A. Behrendt, A. Polywka, P. Goerrn, P. Reckers, J. Maibach, T. Mayer, T. Riedl, Adv. Energy Mater. 3 (2013) 1437–1444. [23] F. Chen, Q. Chen, L. Mao, Y. Wang, X. Huang, W. Lu, B. Wang, L. Chen, Nanotechnology 24 (2013) 484011. [24] F. Tang, Q. Chen, L. Chen, F. Ye, J. Cai, L. Chen, Appl. Phys. Lett. 109 (2016) 123301. [25] H. Ishii, K. Sugiyama, E. Ito, K. Seki, Adv. Mater. 11 (1999) 605–625. [26] H. Ishii, N. Hayashi, E. Ito, Y. Washizu, K. Sugi, Y. Kimura, M. Niwano, Y. Ouchi, K. Seki, Phys. Status Solidi A 201 (2004) 1075–1094. [27] G. Cohen, E. Halpern, S.U. Nanayakkara, J.M. Luther, C. Held, R. Bennewitz, A. Boag, Y. Rosenwaks, Nanotechnology 24 (2013) 295702. [28] S.U. Nanayakkara, G. Cohen, C.-S. Jiang, M.J. Romero, K. Maturova, M. Al-Jassim, J. van de Lagemaat, Y. Rosenwaks, J.M. Luther, Nano Lett. 13 (2013) 1278–1284. [29] D.M. Gonzalez, C.J. Schaffer, S. Proller, J. Schlipf, L. Song, S. Bernstorff, E.M. Herzig, P. Muller-Buschbaum, ACS Appl. Mater. Interfaces 9 (2017) 3282–3287. [30] W. Wang, C.J. Schaffer, L. Song, V. Koerstgens, S. Proeller, E.D. Indari, T. Wang, A. Abdelsamie, S. Bernstorff, P. Mueller-Buschbaum, J. Mater. Chem. A 3 (2015) 8324–8331. [31] Q. An, F. Zhang, X. Yin, Q. Sun, M. Zhang, J. Zhang, W. Tang, Z. Deng, Nano Energy 30 (2016) 276–282. [32] M. Zhang, F. Zhang, Q. An, Q. Sun, W. Wang, X. Ma, J. Zhang, W. Tang, J. Mater. Chem. A 5 (2017) 3589–3598. [33] E. Strassburg, A. Boag, Y. Rosenwaks, Rev. Sci. Instrum. 76 (2005) 083705. [34] S. Lu, L. Ji, W. He, P. Dai, H. Yang, M. Arimochi, H. Yoshida, S. Uchida, M. Ikeda, Nanoscale Res. Lett. 6 (2011) 1–4. [35] P. Dai, S.L. Lu, Y.Q. Zhu, L. Ji, W. He, M. Tan, H. Yang, M. Arimochi, H. Yoshida, S. Uchida, M. Ikeda, J. Cryst. Growth 378 (2013) 604–606. [36] R.C. Kumar, Solid State Electron. 11 (1968) 543–551. [37] S.I. Cserveny, Int. J. Electron. 25 (1968) 65–80. [38] S.W. Smith, The Scientist and Engineer's Guide to Digital Signal Processing, California Technical Pub., San Diego, 1997. [39] Y. Zhou, C. Fuentes-Hernandez, J. Shim, J. Meyer, A.J. Giordano, H. Li, P. Winget, T. Papadopoulos, H. Cheun, J. Kim, M. Fenoll, A. Dindar, W. Haske, E. Najafabadi, T.M. Khan, H. Sojoudi, S. Barlow, S. Graham, J.-L. Brédas, S.R. Marder, A. Kahn, B. Kippelen, Science 336 (2012) 327–332. [40] I.D. Parker, J. Appl. Phys. 75 (1994) 1656–1666. [41] V.D. Mihailetchi, P.W.M. Blom, J.C. Hummelen, M.T. Rispens, J. Appl. Phys. 94 (2003) 6849–6854. [42] G.G. Malliaras, J.R. Salem, P.J. Brock, J.C. Scott, J. Appl. Phys. 84 (1998) 1583–1587. [43] Z.C. He, C.M. Zhong, X. Huang, W.Y. Wong, H.B. Wu, L.W. Chen, S.J. Su, Y. Cao, Adv. Mater. 23 (2011) 4636–4643. [44] W. Liu, T. Liang, Q. Chen, Z. Yu, Y. Zhang, Y. Liu, W. Fu, F. Tang, L. Chen, H. Chen, ACS Appl. Mater. Interfaces 8 (2016) 9254–9261. [45] L. Mao, Q. Chen, Y. Li, Y. Li, J. Cai, W. Su, S. Bai, Y. Jin, C.-Q. Ma, Z. Cui, L. Chen, Nano Energy 10 (2014) 259–267. [46] Y.X. Li, D.P. Qian, L. Zhong, J.D. Lin, Z.Q. Jiang, Z.G. Zhang, Z.J. Zhang, Y.F. Li, L.S. Liao, F.L. Zhang, Nano Energy 27 (2016) 430–438. [47] A. Guerrero, L.F. Marchesi, P.P. Boix, S. Ruiz-Raga, T. Ripolles-Sanchis, G. GarciaBelmonte, J. Bisquert, ACS Nano 6 (2012) 3453–3460. [48] Y. Vaynzof, D. Kabra, L. Zhao, L.L. Chua, U. Steiner, R.H. Friend, ACS Nano 5 (2011) 329–336. [49] M. Kroger, S. Hamwi, J. Meyer, T. Riedl, W. Kowalsky, A. Kahn, Org. Electron. 10 (2009) 932–938. [50] Y.J. Lee, J. Yi, G.F. Gao, H. Koerner, K. Park, J. Wang, K.Y. Luo, R.A. Vaia, J.W.P. Hsu, Adv. Energy Mater. 2 (2012) 1193–1197. [51] J. Meyer, A. Shu, M. Kroger, A. Kahn, Appl. Phys. Lett. 96 (2010) 133308.
the active layer and the interlayers are critically important in determining the band bending in the active layer. Although the cross-sectional SKPM measurements are here demonstrated in a model device, i.e., P3HT: PCBM inverted OPV device, this method can be used for other solar cells or thin-film optoelectronic devices. The potential challenges of the cross-sectional SKPM method are to keep the device under investigation in operando states and to improve the imaging resolution. It is important to maintain operating status of the device, which is usually hindered by the fact that many thin-film optoelectronic devices suffer from environmental instability. Another challenge is to push the limit in spatial and energy resolution. The deconvolution protocol described in this manuscript helps to alleviate the convolution effect, however, for small interfacial features less than 10 nm, intrinsically sharper tips and low noise level microscope systems are desired. Notes Conflicts of interest None. Acknowledgment Q.C. thanks Yi Zhang at the Electron Microscope Lab of SINANO, CAS for her assistance with fabricating the device cross section and the SEM characterization. Wenxian Yang and Dr. Jinhua Cai at SINANO are acknowledged for helpful discussion on deconvolution process. This work was supported by the Ministry of Science and Technology of China (Grant No. 2016YFA0200703), the CAS Research Equipment Development Program (YZ201654), and the National Natural Science Foundation of China (Grant Nos. 21625304, 51473184 and 11504408). Partial support from Collaborative Innovation Center of Suzhou Nano Science and Technology (CICSNST) is also appreciated. Q.C. acknowledges Collaborative Academic Training Program for Post-doctoral Fellows support from CICSNST. Appendix A. Supplementary material Supplementary data associated with this article can be found in the online version at http://dx.doi.org/10.1016/j.nanoen.2017.08.050. References [1] Z. He, C. Zhong, S. Su, M. Xu, H. Wu, Y. Cao, Nat. Photon. 6 (2012) 591–595. [2] Z. He, B. Xiao, F. Liu, H. Wu, Y. Yang, S. Xiao, C. Wang, T.P. Russell, Y. Cao, Nat. Photon. 9 (2015) 174–179. [3] H.L. Yip, A.K.Y. Jen, Energy Environ. Sci. 5 (2012) 5994–6011. [4] S. Li, L. Ye, W. Zhao, S. Zhang, S. Mukherjee, H. Ade, J. Hou, Adv. Mater. 28 (2016) 9423–9429. [5] W.C. Zhao, D.P. Qian, S.Q. Zhang, S.S. Li, O. Inganas, F. Gao, J.H. Hou, Adv. Mater. 28 (2016) 4734–4739. [6] J. Wang, L. Xu, Y.-J. Lee, M.D.A. Villa, A.V. Malko, J.W.P. Hsu, Nano Lett. 15 (2015) 7627–7632. [7] J.G. Tait, U.W. Paetzold, D. Cheyns, M. Turbiez, P. Heremans, B.P. Rand, ACS Appl. Mater. Interfaces 8 (2016) 2211–2219. [8] P. Büchele, M. Morana, D. Bagnis, S.F. Tedde, D. Hartmann, R. Fischer, O. Schmidt, Org. Electron. 22 (2015) 29–34. [9] G.F.A. Dibb, M.A. Muth, T. Kirchartz, S. Engmann, H. Hoppe, G. Gobsch, M. Thelakkat, N. Blouin, S. Tierney, M. Carrasco-Orozco, J.R. Durrant, J. Nelson, Sci. Rep. 3 (2013) 3335. [10] I. Zonno, A. Martinez-Otero, J.C. Hebig, T. Kirchartz, Phys. Rev. Appl. 7 (2017) 034018. [11] R. Saive, M. Scherer, C. Mueller, D. Daume, J. Schinke, M. Kroeger, W. Kowalsky, Adv. Funct. Mater. 23 (2013) 5854–5860. [12] T. Glatzel, D.F. Marron, T. Schedel-Niedrig, S. Sadewasser, M.C. Lux-Steiner, Appl. Phys. Lett. 81 (2002) 2017–2019. [13] J. Lee, J. Kong, H. Kim, S.-O. Kang, K. Lee, Appl. Phys. Lett. 99 (2011) 243301. [14] V.W. Bergmann, S.A. Weber, F. Javier Ramos, M.K. Nazeeruddin, M. Gratzel, D. Li, A.L. Domanski, I. Lieberwirth, S. Ahmad, R. Berger, Nat. Commun. 5 (2014) 5001. [15] C.-S. Jiang, M. Yang, Y. Zhou, B. To, S.U. Nanayakkara, J.M. Luther, W. Zhou, J.J. Berry, J. van de Lagemaat, N.P. Padture, K. Zhu, M.M. Al-Jassim, Nat. Commun.
460
Nano Energy 40 (2017) 454–461
Q. Chen et al. Qi Chen received his Ph.D. degree in physics from University of Science and Technology of China (USTC) in 2014. He then worked as a postdoctoral researcher at Suzhou Institute of Nano-Tech and Nano-Bionics (SINANO), Chinese Academy of Sciences and University of Washington, Seattle from 2014 to 2017. He is currently an associate professor at SINANO. His research interests focus on interface characterization of thin-film optoelectronic devices by scanning probe microscopy.
Changqi Ma received his Ph.D. degree at the Technical Institute of Physics and Chemistry, Chinese Academy of Sciences with Professor B.-W. Zhang in 2003. After that he was a postdoctoral research assistant at Heriot-Watt University in Edinburgh, UK, until he joined Professor P. Bäuerle's group at the University of Ulm in 2004 as a Humboldt research fellow. From January 2007 till May 2011, he did his Habilitation at the same institute. He was appointed as a professor of chemistry at SINANO in June 2011. His research mainly focuses on solution printable photovoltaic technologies.
Fengye Ye obtained his B.S. degree from USTC in 2013. He is currently a Ph.D. candidate at USTC, co-supervised by Prof. Liwei Chen and Prof. Yi Xie. His research focuses on synthesis and characterization of perovskite solar cells and light-emitting diodes.
Yanfei Zhao received her Ph.D. degree from the International Center for Quantum Materials, Peking University. She is currently an Assistant Professor at Vacuum Interconnected Nanotech Workstation at SINANO. Her research interests focus on the surface and interface of semiconductor and related materials.
Junqi Lai received his B.S. degree in condensed matter physics from USTC in 2015. He is currently a Ph.D. candidate at SINANO. His research focuses on atomic force microscopy characterization and numerical simulation of charge carrier transport in organic photovoltaics devices.
Yi Xie received her B.S. from Xiamen University in 1988 and Ph.D. from USTC in 1996. She is currently a Principal Investigator in Department of Nanomaterials and Nanochemistry, Hefei National Laboratory for Physical Sciences at Microscale, a full professor in Department of Chemistry, USTC, and a member of Chinese Academy of Sciences. Her research interests focus on the design and synthesis of inorganic functional solids with efforts to modulate their electronic and phonon structures.
Pan Dai received her Ph.D. in Microelectronics and Solidstate Electronics from SINANO in 2016, and then worked as a postdoctoral researcher at SINANO for III-V multi-junction solar cell research. Her research interests focus on the MBE growth of III-V compound semiconductor materials and solar cell devices.
Liwei Chen obtained his B.S. from USTC in 1993, MS from Peking University in 1996, and Ph.D. from Harvard University in 2001. After working as a joint postdoctoral research fellow at Columbia University and IBM T.J. Watson Research Center, he started his independent research career as an assistant professor at Department of Chemistry and Biochemistry at Ohio University in 2004. He moved to SINANO as a professor in 2009. His current research interests focus on energy nanotechnology including thin-film solar cells and lithium batteries.
Shulong Lu obtained his Ph.D. degree from the Institute of Semiconductor of Chinese Academy of Sciences in 2003. From 2003–2008, he worked in Paul-Drude institute of Berlin and in Waseda University in Tokyo as a post-doctoral researcher and a visiting lecturer, respectively. He joined SINANO in 2008. His research interests focus on the MBE growth of III-V compound semiconductor material and related devices, such as photovoltaic cells and infrared detectors.
461
Contents lists available at ScienceDirect
Nano Energy journal homepage: www.elsevier.com/locate/nanoen
Full paper
Energy band alignment in operando inverted structure P3HT:PCBM organic solar cells
MARK
Qi Chena, Fengye Yea,b, Junqi Laia, Pan Daic, Shulong Luc, Changqi Mad, Yanfei Zhaoe, Yi Xieb, ⁎ Liwei Chena,e, a
i-Lab, CAS Center for Excellence in Nanoscience, Suzhou Institute of Nano-Tech and Nano-Bionics, Chinese Academy of Sciences, Suzhou 215123, China Hefei National Laboratory for Physical Sciences at the Microscale and Synergetic Innovation Center of Quantum Information & Quantum Physics, University of Science and Technology of China, Hefei 230026, China c Key Laboratory of Nanodevices and Applications, Suzhou Institute of Nano-Tech and Nano-Bionics, Chinese Academy of Sciences, Suzhou 215123, China d Division of Printed Electronics, Suzhou Institute of Nano-Tech and Nano-Bionics, Chinese Academy of Sciences, Suzhou 215123, China e Vacuum Interconnected Nanotech Workstation (Nano-X), Suzhou Institute of Nano-Tech and Nano-Bionics, Chinese Academy of Sciences, 215123 Suzhou, China b
A R T I C L E I N F O
A B S T R A C T
Keywords: Cross-sectional scanning Kelvin probe microscopy Inverted structure organic solar cells Interlayer Energy band alignment Deconvolution
Inverted structure thin-film organic solar cells (OSCs) are becoming increasingly important as they deliver higher power conversion efficiency and demonstrate better long-term stability than conventional devices. However, the energy band alignment and the built-in field across the device, which are crucial in understanding the device operation, is yet to be directly characterized. Here we present a direct visualization of the energy level alignment in operando inverted structure poly(3-hexylthiophene) (P3HT): [6,6]-phenyl-C61-butyric acid methyl ester (PCBM) OSCs using cross-sectional scanning Kelvin probe microscopy. The raw data of measured energy level alignment appear to be inconsistent with each other, and sometimes can even be contradictory to the device polarity observed in current density-voltage measurements. It is identified to be caused by the tip/cantilever induced convolution effect, which may severely mask abrupt energy level offsets at the thin electrode interlayers. A numerical deconvolution method is devised to quantitatively recover the energy level alignment across the device, and reveals the non-uniform electric field distribution in photoactive layer.
1. Introduction Inverted structure organic solar cells (OSCs), with holes extracted from the top electrode and electrons from the bottom electrode, have the reverse polarity from conventional structure OSCs. In spite of the relatively short history of development, inverted structure devices are becoming increasingly important due to their high power conversion efficiency (PCE), good long-term device stability and the adaptability to tandem device structures [1–5]. The energy band depth profile of inverted structure devices has its unique features compared to that of conventional structure devices. The thin electrode interlayers in the inverted devices are expected to introduce abrupt energy level offsets in the depth profile. Substantial research efforts have been made to analyze energy levels at the active layer/interlayer interface, and the results demonstrated that interfacial doping/dipole is one of the most important factors that lead to non-uniform electric field distribution and determine the charge carrier transport [6–10]. However, characterizations of the energy band depth profile across the entire inverted structure devices are still rarely seen in literature. One such effort ⁎
actually reported an energy band depth profile whose direction of the built-in field was opposite to the device polarity measured from J-V curves [11]. These conflicting results require direct measurements for clarification. Cross-sectional SKPM is a powerful tool in measuring vacuum level (VL) depth profile in thin-film devices [11–15]. We have previously used this technique to visualize the energy level alignment in operando conventional structure OSCs [16]. It was discovered that the finite tip size and the cantilever beam crosstalk result in a convolution effect, which causes an apparent smoothing and averaging in measured surface energy depth profiles [17–19]. Here we further present an improved in-operando cross-sectional SKPM study on inverted structure poly(3-hexylthiophene) (P3HT): [6,6]-phenyl-C61-butyric acid methyl ester (PCBM) OSCs (Fig. 1) to reveal the true energy band alignment and built-in field distribution. It is observed that the same tip/cantilever convolution effect is highly significant in inverted structure device measurements, to a degree that under certain conditions that the abrupt energy level offsets introduced by interlayers could be completely masked and the measured VL profile
Corresponding author at: i-Lab, CAS Center for Excellence in Nanoscience, Suzhou Institute of Nano-Tech and Nano-Bionics, Chinese Academy of Sciences, Suzhou 215123, China. E-mail address: [email protected] (L. Chen).
http://dx.doi.org/10.1016/j.nanoen.2017.08.050 Received 2 July 2017; Received in revised form 24 August 2017; Accepted 27 August 2017 Available online 30 August 2017 2211-2855/ © 2017 Elsevier Ltd. All rights reserved.
Nano Energy 40 (2017) 454–461
Q. Chen et al.
spectra were measured by a Merlin (Model No. 70104) Digital Lock-in Radiometry System using a 300-W xenon lamp and a 74125 Oriel Cornerstone 260 1/4m monochromator with order-sorting filters (Newport). The light intensity at different wavelengths was modulated through an aperture and calibrated using a 70714 UV-enhanced Si photodiode (Newport). All of the measurements were performed under ambient atmosphere at room temperature. 2.2. MBE reference sample GaInP (Ga:In 0.51:0.49 for lattice match)/GaAs heterojunction on p++ GaAs wafer was grown by a Veeco GEN20A dual-chamber all solid-state MBE equipped with valved phosphorous and arsenic cracking cells. The typical growth rate of GaAs and GaInP was 1 µm/h, and the growth temperatures of the GaAs and GaInP were 580 °C and 510 °C respectively. Be was used as p-type doping source and doping concentration was 2.2 × 1018 cm−3 in GaAs and 8.0 × 1017 cm−3 in GaInP. The group-III mole fraction (In and Ga in this case) was calibrated by a combination of beam flux gauge and X-ray diffraction (XRD) measurements to satisfy lattice match. 2.3. Device cross-section preparation and characterization The device cross-section was fabricated using an Ilion+ 693 System (Gatan Inc.). A freshly mechanically cleaved device was mounted into the vacuum chamber (6.4 × 10−5 Torr) and cooled using liquid nitrogen. Then, the device was milled by argon ions using a beam voltage of 5 keV and a beam current of 10 μA for ~ 2 h. Five cross-section samples were prepared for thin interlayer and thick interlayer devices respectively, the error bars of PCE before and after cross section preparation, and after SKPM measurement are listed in Table S1. Stability of typical cross-sectioned devices was tested under continuous light illumination in the glove box. The morphology of the device cross section was characterized using FEI Quanta 400 FEG SEM (FEI Corp., Hillsboro, OR, USA). The SKPM surface potential (SP) profile measurements of device cross-sections were carried out using a Park XE-120 AFM (Park Systems Corp., Suwon, Korea) using Cr/Au-coated conducting tips (NSC18, Mikromasch, Tallinn, Estonia) with a resonance frequency of ~ 80 kHz and a spring constant of ~ 2 N m−1. Amplitude modulation (AM) mode SKPM [23,24] was carried out in a nitrogen gas filled glove box. During the first pass, standard AC mode imaging (typical tip oscillation amplitude 20 nm) was performed to acquire the topography and phase signal of the sample; in the second pass, the tip was lifted up by a certain height (typically 10 nm) and scanned on the basis of the topography line obtained from the first pass. An AC voltage (3 V in amplitude and 10 kHz in frequency) was applied to actuate the cantilever, and the DC voltage applied to the tip that nullifies the tip-sample interaction was collected as the SP signal. The VL profile can be obtained by multiplying the measured SP profile with the absolute electron charge (at least 10 scan lines are included for averaging) [25,26]. Fig. 1a shows the configuration of device-wiring during cross-section characterization. The Al electrode was grounded. The ITO electrode was connected with the Al electrode in short-circuit; open-circuit conditions were simulated by disconnecting the wire between the two electrodes. A full-spectrum optical fiber was used to transmit AM 1.5G solar simulator light to illuminate the device from the ITO glass side. Bias voltage was applied via a tunable voltage source between the electrodes.
Fig. 1. (a) Schematic illustration of SKPM measurements of the vacuum level depth profile of operando cross-sectional devices. (b) The SKPM measured profiles are the convolution of the true profile and the tip transfer function. The true profile can be recovered via numerical deconvolution.
may appear to be contradictory to the device polarity. To resolve this issue, a deconvolution process including a calibration on a molecular beam epitaxy (MBE) grown GaInP/GaAs reference sample and a numerical deconvolution calculation is devised to negate the tip and cantilever induced convolution effects in SKPM. The true energy band alignment in inverted OSCs has thus been recovered for the first time. The results confirm that the work function of the interlayers and interfacial effects are critically important in determining the energy level alignment and leading to a non-uniform electric field across the active layer in inverted structure OSCs.
2. Experimental section 2.1. Device fabrication and characterization The inverted structure OSCs with a stacking of ITO/ZnO/P3HT: PCBM/MoOx/Al were fabricated by sequentially cleaning pre-patterned ITO glass substrates (15 Ωsq−1) using ultrasonication for 10 min in a detergent, deionized water, ethanol, acetone and isopropyl alcohol, followed by treatment with oxygen plasma for 10 min. Then, ZnO nanocrystal solution synthesized according to our previous reports was spin-coated onto ITO substrates and annealed at 150 °C for 20 min [20]. The P3HT (1-Material Inc.): PCBM (American Dye Source Inc.) blend (36 mg/ml, 1:0.8 w/w) was dissolved in 1,2-dichlorobenzene and stirred at 60 °C for 14 h in a glove box. The blend active layer was prepared by spin-coating at 600 r.p.m. for 60 s, and annealed at 130 °C for 10 min. A 20-nm or 80-nm MoOx layer and 100-nm Al layer were subsequently evaporated under a pressure of 4 × 10−4 Pa through a shadow mask to define the active area of the device (0.12 cm2) and to form a top anode. The J–V characteristics of the devices were recorded using a Keithley 2635 A source meter (Keithley Instruments Inc., Cleveland, OH, USA) under one sun, AM 1.5 G irradiation (100 mW cm−2) from a solar simulator (Newport 67005, Newport, Irvine, CA, USA). The illumination intensity was calibrated using a reference cell (Oriel 91150V). The light-soaking effect was checked by applying a UV filter, which was not significant here [21,22]. External quantum efficiency (EQE)
2.4. Numerical convolution and deconvolution simulation Numerical simulations were performed using Wolfram Mathematica 10 with home-written codes. The SKPM measurements were simulated as a convolution of the true SP profiles and the tip transfer functions with noise. Tip transfer functions were obtained by adjusting the σ and A parameters to make the simulated (i.e. convoluted) profile of a 455
Nano Energy 40 (2017) 454–461
Q. Chen et al.
batch (NSC18 Cr/Au, Mikromasch), named Tip 1 and Tip 2, respectively. Comparison of the extracted SP profiles (Fig. 2d) yields stark differences. The profile obtained with Tip 1 exhibits a continuous increase in SP from the ITO electrode to the Al electrode. This behavior is similar to that in conventional structure device [11,13,16], even though the device polarity is completely the opposite. This has been reported in a previous literature [11], but the discrepancy between the cross-sectional SKPM and device polarity was not commented or resolved. Interestingly, the profile obtained with Tip 2 displays a shallow kink in the middle of the active layer. The built-in field of the kinked region is the opposite from that in conventional structure devices, and thus it agrees with the device polarity but contradicts with that obtained with Tip 1. The apparent differences in measured SP profiles could be related with variations in the tip geometry used in SKPM. By imaging an array of sharp Si spikes of a test grating (TGT1, NT-MDT), it is revealed that the radius of curvature of Tip 2 (21 nm) is significantly smaller than that of Tip 1 (35 nm) (Fig. 2e-f). Furthermore, the measured SP profiles are also sensitively dependent on the tip-sample distance in the SKPM measurement. Shown in Fig. 2g is a SP image obtained with Tip 2 at different lift heights. The corresponding SP profiles (Fig. 2h) show a transition from the one with the kink at the lowest lift height to those with monotonic increase at higher lifts. These results agree well with our previous study that the cross-sectional SKPM of thin film devices are severely affected by tip/cantilever convolution effects [16]. The convolution effect due to finite tip size and cantilever crosstalk in SKPM has been extensively documented [17–19,27,28,33]. The effect is mathematically accounted by the convolution of the true potential profile with a tip transfer function, typically in a Gaussian form, f(x) = Aexp(x2/2σ2), in which σ determines the widening and A determines the contrast smoothing in the convolution [17,19,27,28,33]. To recover the true potential profile from the convoluted measurements, accurate tip transfer functions are needed for deconvolution. Previous efforts in determining the tip transfer function usually involved modeling based on vendor specified tip and cantilever geometry. However, since there exists significant tip-to-tip variation in commercial SKPM probes, and as we have shown above that the convolution effect is highly sensitively dependent on measurement parameters, a more precise approach is needed to reliably gauge the transfer function. Here we demonstrate a deconvolution routine involving calibration
reference sample to overlap with the measured profile. Then, the true profiles of samples under study can be recovered by Wiener deconvolution in frequency domain. Here, the system noise is a white Gaussian noise, which is negligible in convolution and will be disproportionately amplified at high frequencies by direct deconvolution. Thus, a Weiner filter in frequency domain (TTF)−1/[1 + (SNR × |TTF|2)−1] (in which TTF is the tip transfer function and SNR is the signal-to-noise ratio) was employed to minimize the impact of deconvolution noise at high frequencies, which have a poor signal-to-noise ratio [27,28].
3. Results and discussion The current density-voltage (J-V) curves and EQE curves of the inverted structure device with optimized interlayer thickness (20 nm ZnO and 10 nm MoOx) are shown in Fig. S1. The PCE before and after crosssection preparation are 3.93% and 3.39%, respectively (device performance parameters detailed in Table S1), and PCE after SKPM measurement is 3.15%, indicating that the cross-sectioned device is relatively stable during the SKPM measurement. The stability of a typical cross-sectioned device under ~ 10 h continuous light illumination is shown in Fig. S1. The Voc is very stable at first (< 1% drop after ~ 3 h), and then it decreases slightly (~ 5% drop after 10 h). The stable Voc in early stage may attribute to the optimized crystallinity during device preparation [29]. With long-term light illumination, the crystallinity decreases gradually and it leads to broadened density of states (DOS), which reduces the splitting of the quasi-Fermi levels (i.e. Voc decreases) [29]. The degradation in PCE (~ 19% drop after 3 h and ~ 36% drop after 10 h) is mainly attributed to the decrease in Jsc (~ 10% drop after 3 h and ~ 24% drop after 10 h) and FF (~ 8% drop after 3 h and ~ 12% drop after 10 h). The decrease in Jsc and FF can be explained by the change in morphology that large domains grow gradually in aging [30–32]. The cross-sectional SKPM measurement was mostly completed within the first two hours of continuous illumination. The device Voc exhibits little change in this period; and thus the energy band alignment measurement is reliable. The scanning electron microscopy (SEM) backscattered electron image displayed in Fig. 2a shows a cross section with clear contrast between different stacking layers. Fig. 2b and c display two cross-sectional SKPM measured SP images of the same device at roughly the same location in dark acquired by two different tips from the same
Fig. 2. (a) SEM image of the cross-sectional device with layer stacking of ITO/ZnO(20 nm)/P3HT:PCBM/MoOx(10 nm)/Al (scale bar: 300 nm). (b-c) SP images of the device in dark state with two tips: Tip 1 (b) and Tip 2 (c) (scale bar: 200 nm). (d) SP profiles extracted from (b) (black square for Tip 1) and (c) (red circle for Tip 2). (e) Schematic illustration of TGT1 for tip geometry calibration. (f) The profile of the spike at its apex. The inset shows the entire profile of the spike. (g-h) The SP image and the extracted SP profiles of the same device in dark state with Tip 2 under different tip-sample distance from 20 nm to 120 nm (scale bar: 200 nm).
456
Nano Energy 40 (2017) 454–461
Q. Chen et al.
Fig. 3. (a) SEM image of GaInP/GaAs heterojunction (scale bar: 1 µm). (b-c) SP images of GaInP/GaAs heterojunction acquired by Tip 1 and Tip 2 (scale bar: 250 nm). (d) SP profiles extracted from (b) (black square) (c) (red circle) and compared with the calculated SP profile. The green triangle is the SP profile acquired by Tip 2 at + 0.25 V bias voltage. All these profiles are acquired under tip-sample distance of 20 nm. (e) The transfer function of Tip 1 and Tip 2 derived from numerical convolution fitting. (f) Deconvoluted SP profile based on the transfer function of Tip 1 and Tip 2. (g) SP profiles of GaInP/GaAs heterojunction acquired by Tip 2 under different tip-sample distances from 20 nm to 120 nm. (h) Tip transfer function of Tip 2 under different tip-sample distances derived from numerical convolution fitting. (i) Deconvoluted SP profiles based on the transfer function of Tip 2 under different tip-sample distances.
the GaAs is visible. The high doping concentration used here is important to get a narrow depletion layer and create a steep VL shift at the interface. Fig. 3b and c show the SP images measured with the Tip 1 and Tip 2 at tip-sample distance of 20 nm, respectively. The profiles displayed in Fig. 3d show flat SP at both the GaInP and the GaAs sides, with a sloped step at the interface. The SP profile acquired by Tip 2 (red open circle) has a greater SP difference between the two sides and a steeper slope at the interface than that acquired by Tip 1 (black open square). Interestingly, when a bias voltage of +0.25 V is applied onto the GaInP side against the GaAs side, the SP profile acquired by Tip 2 appears to be flat across the interface (the open triangles in Fig. 3d), regardless of different lift heights in the measurements (Fig. S3). According to the bias compensation method developed in our previous publication [16], this indicates that the VL shift at the GaInP/GaAs interface is ~ 0.25 eV. A SP profile of the reference sample is thus calculated based on an abrupt isotype heterojunction model [36,37]. The width of the transition region is interpreted as the sum of the band bending zones on both side of the heterojunction, i.e., the width of the sloped step in the SP profile is found to be 32.2 nm (Fig. S2 in Supporting information). Numerical convolution and deconvolution calculations are implemented to derive tip transfer functions based on the measurements on the reference sample. The σ and A of tip transfer functions are
Fig. 4. Deconvoluted SP profiles of the device with thin interlayer in Fig. 2 based on the transfer functions of Tip 1 and Tip 2.
over a reference sample with a computable VL profile (Fig. 1b). A MBE grown GaInP (Ga:In 0.51:0.49)/GaAs heterojunction is used as the reference sample [34,35]. The materials interface in MBE grown heterojunction is typically 1–2 atomic layers wide, therefore, the particular sample composed of a 1 µm thick p++ GaAs epitaxial layer on p++ GaAs substrate, followed with 1 µm thick p++ GaInP epitaxial layer MBE heterojunction is an ideal abrupt junction. Fig. 3a shows the SEM backscattered electron image of the cross-section of the GaInP/GaAs abrupt heterojunction, in which the interface between the GaInP and 457
Nano Energy 40 (2017) 454–461
Q. Chen et al.
Fig. 5. (a) SEM image of the ITO/ZnO(80 nm)/P3HT:PCBM/MoOx(80 nm)/Al device cross-section (scale bar: 300 nm). (b-c) SP images of the device in dark state (b) and under light illumination (c) in open-circuit condition (scale bar: 400 nm). (d) SP profiles extracted from (b) (black square) and (c) (red circle), and the SP profile at +0.6 V bias in dark (blue line). (e) The transfer function of Tip 3 derived from numerical convolution fitting. (f) Deconvoluted SP profiles from (d): black line for the dark and red line for the illuminated state of the device.
adjusted to make the convoluted profile overlap with the measured profile. For the measurements with tip-sample distance of 20 nm, the σ and A for Tips 1 and 2 are yielded as 74 nm and 1.67 µm−1 for Tip 1, and 35 nm and 4.96 µm−1 for Tip 2, respectively (Fig. 3e). With these parameters, the corresponding Gaussian functions are used as the transfer function in the numerical deconvolution calculation. Fig. 3f shows that although the SP profiles of the reference sample measured with Tips 1 and 2 are different, the two deconvoluted SP profiles essentially overlap with each other and agree well with the calculated SP profile. This corroborates the literature report that tip radius is one of the most important factors that determines σ and A of tip transfer function [33], and also confirms the validity of our deconvolution calculation. Furthermore, the convolution effect of tips at different lift height was also investigated. SP profiles of the GaInP/GaAs reference sample measured by Tip 2 under different tip-sample distance from 20 nm to 120 nm are shown in Fig. 3g. As the tip is lifted away from the sample, the measured profiles show less SP difference between the GaInP and GaAs region and become less steep. The convolution fitting thus results in increasing σ and decreasing A in the tip transfer function (Fig. 3h). With these transfer functions; the measured profiles are deconvoluted as shown in Fig. 3i. The deconvoluted SP profiles obtained with tipsample distance of 20 nm and 45 nm show satisfying agreement with the calculated SP profile, which further corroborates the reliability of the deconvolution process. However, if tip-sample distance is greater than 45 nm, the deconvoluted SP profiles exhibit serious oscillation in GaAs and GaInP regions that are expected to be flat in SP. This is due to lack of resolution in the primary signal and the magnification of high frequency noises in the numerical deconvolution algorithm [27,28,38]. Therefore, a tip transfer function with small σ and large A, i.e. a sharp tip and a small tip-sample distance, is preferred in order to recover the true SP profile reliably by using deconvolution. With the tip transfer function calibrated and the deconvolution process validated, the SP depth profiles of inverted structure OSCs could then be obtained. The deconvoluted and the originally measured SP profiles obtained using Tip 1 and Tip 2 are shown in Fig. 4. It is found that the SP differences between Al and ITO after deconvolution are ~ 1.36 V for Tip 1 and ~ 1.37 V for Tip 2, respectively. The values
approximately agree with literature reported 1.6 V (the work functions of ITO and Al electrodes have been separately measured to be 5.0 and 3.4 eV using ultraviolet photon spectroscopy (UPS) and Kelvin probe (KP), respectively) [39]. Importantly, both deconvoluted SP profiles exhibit a shallow kink in the active layer, indicating that energy band alignment and built-in field in the inverted structure device is in opposite direction to conventional structure devices. This result is consistent with the device concept as well as the device polarity displayed in J-V characteristics. The deconvoluted SP profile from Tip 1 shows oscillation in cathode and anode regions while that from Tip 2 shows negligible oscillation. This is again due to magnification of high frequency noise by numerical deconvolution when the primary data measured with Tip 1 lacks spatial resolution. According to the metal-insulator-metal structure (taking interlayer/ electrode as assembled entities) device model [40], the SP drops from ZnO across the active layer to MoOx can be interpreted as upper-limit of Vbi; however, the SP difference in the kinked region in the deconvoluted profiles is actually smaller than device Voc. We suspect that even with the deconvolution method, the ZnO (~ 20 nm) and MoOx (~ 10 nm) interlayers are still too thin to be accurately probed because of the large tip radius of commercial SKPM probes. In order to better quantify the energy band alignment in inverted structure devices, we intentionally increase the thickness of the interlayers to 80 nm ZnO and 80 nm MoOx. The thick interlayers are clearly visible in cross section SEM micrographs (Fig. 5a). The J-V curves and EQE curves of the device with thicker interlayers are shown in Fig. S1. The PCE before and after cross section preparation are 3.09% and 2.79% respectively (the device performance is detailed in Table S1). JSC and FF have decreased slightly due to the high series resistance resulted from the thick interlayer. On the other hand, VOC remains unchanged, suggesting the Vbi and energy band alignment of the thick interlayer device shall be very similar to those of the thin interlayer devices [41–46]. The PCE of the device after SKPM measurement is 2.61%, and the stability of a typical cross-sectioned devices under ~ 10 h continuous light illumination is shown in Fig. S1. The Voc is very stable (~ 1% drop after 3 h and ~ 6% drop after 10 h), which is similar to that of devices with optimized interlayer thickness. Fig. 5b and c are SP images of the device cross-section measured with a new tip denoted as Tip 3 in 458
Nano Energy 40 (2017) 454–461
Q. Chen et al.
compensated by VL shifts at ZnO/BHJ interface and the band bending in the BHJ (qVbi), leading to a non-uniform electric field across the active layer [7,9]. The SP profile (red open circle in Fig. 5d) of the same device in open-circuit under light illumination was extracted from Fig. 5c. A reduction in potential difference from ZnO to MoOx is observed, because the direction of electric field induced by photogenerated charge carriers accumulated at interlayer/BHJ interface is opposite to that of built-in field; therefore, the reduction in build-in field results in decreased overall potential drop from ZnO across BHJ to MoOx. By using the bias voltage compensation method [16], a bias voltage of + 0.6 V is required to overlap the biased SP profile in dark (blue line) with that in open-circuit under light illumination. This bias voltage is the same as the device VOC, which confirms the reliability of the SP profiles. The transfer function of Tip 3 is calibrated as shown in Fig. 5e with σ = 32 nm, A = 3.62 µm−1 (calibration process is detailed in Fig. S5); and the measured SP profiles are deconvoluted. The deconvoluted SP profiles in dark (black line) and under light illumination (red line) are shown in Fig. 5f. Both profiles keep the same line shape as the measured profiles but exhibit much enhanced SP contrast. After deconvolution, the SP difference between Al and ITO in dark is ~ 1.40 V, which is very close to that in thin interlayer devices (1.36–1.37 V). From the deconvoluted SP profile in dark, SP drop across the active layer, i.e. from ZnO to MoOx, is ~ 0.97 V, which approaches the work function difference between ZnO and MoOx of 0.94 V, if taking the work functions of ZnO and MoOx interlayers to be 4.26 eV and 5.20 eV, as measured using the KP technique in ambient, respectively [39,50]. It is known that work function of ZnO and MoOx varies significantly depending on sample preparation and/or measurement conditions [51], so references with sample preparation and measurement environment resembling those in SKPM measurements are deliberately chosen here. The work function of MoOx used in this work is also measured by UPS (Fig. S6), which is 5.13 eV and is close to 5.20 eV that measured by KP. Upon light illumination, the SP drop from ZnO to MoOx across the active layer is reduced to 0.39 V due to quasi-Fermi level splitting. Thus, the Voc obtained from deconvoluted SP profiles is 0.97–0.39 = 0.58 V, which is very close to the 0.6 V Voc from J-V curve measurements (deviation < 5%). Based on these measurements, the energy band alignment of inverted structure device in dark and under illumination is illustrated in Fig. 6. It highlights that the work function of the interlayers and interfacial effects such as contact-induced doping and charge transfer generated dipole at active layer/interlayer interface etc. are important factors leading to the non-uniform electric field within the active layer, which significantly impacts on charge carriers behavior and device performance [7,9]. In order to optimize the device performance parameters such as Voc, special attention needs to be paid to the band offset at interlayer/active layer interfaces in addition to the conventional design rule considering the energy level offset between the donor and acceptor.
Fig. 6. Energy band diagrams of inverted structure OSC device in dark (a) and under light illumination (b) in open-circuit condition.
open-circuit in dark state and under light illumination, respectively. In Fig. 5d, the SP profile in dark (black open square) extracted from Fig. 5b shows decreased potential from ZnO across the bulk heterojunction (BHJ) active layer towards MoOx. This result validates the presence of the kinked region in the thin interlayer device (Fig. 2d). Interestingly, SP changes more rapidly at the ZnO/BHJ interface than the BHJ/MoOx interface, which is also manifested in the more negative value in the derivative of SP profiles (Fig. S4). This indicates that there is a more significant VL shift (qΔ) induced by a greater interfacial dipole at the ZnO/BHJ interface, which probably arises from charge injection from ZnO to localized interface states of BHJ (mostly PCBM due to vertical phase segregation in bottom surface of BHJ) [47–49]. On the other hand, vertical phase segregation in BHJ results in P3HT accumulating towards the top electrode, oxidation and/or MoOx contactinduced p-doping of P3HT may lead to increased work function of P3HT, resulting in reduced work function difference between P3HT and MoOx and thus small interfacial VL shift [6,47,49]. Overall, the large work function difference between ZnO and MoOx is mainly
4. Conclusion In summary, it is demonstrated that the apparent inconsistency in cross-sectional SKPM investigations of inverted OSC devices is due to tip-to-tip variations on the probe geometry. The occasionally observed conflict between the SKPM measured built-in field and the J-V characteristics is the result of the low spatial resolution in combination with the tip/cantilever induced convolution effect, which can mask abrupt energy level offsets caused by thin interlayers. The effects of these artifacts can be minimized by calibrating the transfer function of sharp tips and performing numerical deconvolution. The energy band alignment depth profile of inverted OSCs are thus obtained for the first time via in-operando cross-sectional SKPM measurements and deconvolution calculation. The results clarify that the built-in field direction in inverted devices is consistent with the device polarity; more importantly, the work function of the interlayers and the interfacial states between 459
Nano Energy 40 (2017) 454–461
Q. Chen et al.
6 (2015) 8397. [16] Q. Chen, L. Mao, Y. Li, T. Kong, N. Wu, C. Ma, S. Bai, Y. Jin, D. Wu, W. Lu, B. Wang, L. Chen, Nat. Commun. 6 (2015) 7745. [17] A. Liscio, V. Palermo, P. Samorì, Acc. Chem. Res. 43 (2010) 541–550. [18] D.S.H. Charrier, M. Kemerink, B.E. Smalbrugge, T. de Vries, R.A.J. Janssen, ACS Nano 2 (2008) 622–626. [19] G. Elias, T. Glatzel, E. Meyer, A. Schwarzman, A. Boag, Y. Rosenwaks, Beilstein J. Nanotechnol. 2 (2011) 252–260. [20] X. Liang, Y. Ren, S. Bai, N. Zhang, X. Dai, X. Wang, H. He, C. Jin, Z. Ye, Q. Chen, L. Chen, J. Wang, Y. Jin, Chem. Mater. 26 (2014) 5169–5178. [21] W. Xu, R. Xia, T. Ye, L. Zhao, Z. Kan, Y. Mei, C. Yan, X.-W. Zhang, W.-Y. Lai, P.E. Keivanidis, W. Huang, Adv. Sci. 3 (2016) 1500245. [22] S. Trost, K. Zilberberg, A. Behrendt, A. Polywka, P. Goerrn, P. Reckers, J. Maibach, T. Mayer, T. Riedl, Adv. Energy Mater. 3 (2013) 1437–1444. [23] F. Chen, Q. Chen, L. Mao, Y. Wang, X. Huang, W. Lu, B. Wang, L. Chen, Nanotechnology 24 (2013) 484011. [24] F. Tang, Q. Chen, L. Chen, F. Ye, J. Cai, L. Chen, Appl. Phys. Lett. 109 (2016) 123301. [25] H. Ishii, K. Sugiyama, E. Ito, K. Seki, Adv. Mater. 11 (1999) 605–625. [26] H. Ishii, N. Hayashi, E. Ito, Y. Washizu, K. Sugi, Y. Kimura, M. Niwano, Y. Ouchi, K. Seki, Phys. Status Solidi A 201 (2004) 1075–1094. [27] G. Cohen, E. Halpern, S.U. Nanayakkara, J.M. Luther, C. Held, R. Bennewitz, A. Boag, Y. Rosenwaks, Nanotechnology 24 (2013) 295702. [28] S.U. Nanayakkara, G. Cohen, C.-S. Jiang, M.J. Romero, K. Maturova, M. Al-Jassim, J. van de Lagemaat, Y. Rosenwaks, J.M. Luther, Nano Lett. 13 (2013) 1278–1284. [29] D.M. Gonzalez, C.J. Schaffer, S. Proller, J. Schlipf, L. Song, S. Bernstorff, E.M. Herzig, P. Muller-Buschbaum, ACS Appl. Mater. Interfaces 9 (2017) 3282–3287. [30] W. Wang, C.J. Schaffer, L. Song, V. Koerstgens, S. Proeller, E.D. Indari, T. Wang, A. Abdelsamie, S. Bernstorff, P. Mueller-Buschbaum, J. Mater. Chem. A 3 (2015) 8324–8331. [31] Q. An, F. Zhang, X. Yin, Q. Sun, M. Zhang, J. Zhang, W. Tang, Z. Deng, Nano Energy 30 (2016) 276–282. [32] M. Zhang, F. Zhang, Q. An, Q. Sun, W. Wang, X. Ma, J. Zhang, W. Tang, J. Mater. Chem. A 5 (2017) 3589–3598. [33] E. Strassburg, A. Boag, Y. Rosenwaks, Rev. Sci. Instrum. 76 (2005) 083705. [34] S. Lu, L. Ji, W. He, P. Dai, H. Yang, M. Arimochi, H. Yoshida, S. Uchida, M. Ikeda, Nanoscale Res. Lett. 6 (2011) 1–4. [35] P. Dai, S.L. Lu, Y.Q. Zhu, L. Ji, W. He, M. Tan, H. Yang, M. Arimochi, H. Yoshida, S. Uchida, M. Ikeda, J. Cryst. Growth 378 (2013) 604–606. [36] R.C. Kumar, Solid State Electron. 11 (1968) 543–551. [37] S.I. Cserveny, Int. J. Electron. 25 (1968) 65–80. [38] S.W. Smith, The Scientist and Engineer's Guide to Digital Signal Processing, California Technical Pub., San Diego, 1997. [39] Y. Zhou, C. Fuentes-Hernandez, J. Shim, J. Meyer, A.J. Giordano, H. Li, P. Winget, T. Papadopoulos, H. Cheun, J. Kim, M. Fenoll, A. Dindar, W. Haske, E. Najafabadi, T.M. Khan, H. Sojoudi, S. Barlow, S. Graham, J.-L. Brédas, S.R. Marder, A. Kahn, B. Kippelen, Science 336 (2012) 327–332. [40] I.D. Parker, J. Appl. Phys. 75 (1994) 1656–1666. [41] V.D. Mihailetchi, P.W.M. Blom, J.C. Hummelen, M.T. Rispens, J. Appl. Phys. 94 (2003) 6849–6854. [42] G.G. Malliaras, J.R. Salem, P.J. Brock, J.C. Scott, J. Appl. Phys. 84 (1998) 1583–1587. [43] Z.C. He, C.M. Zhong, X. Huang, W.Y. Wong, H.B. Wu, L.W. Chen, S.J. Su, Y. Cao, Adv. Mater. 23 (2011) 4636–4643. [44] W. Liu, T. Liang, Q. Chen, Z. Yu, Y. Zhang, Y. Liu, W. Fu, F. Tang, L. Chen, H. Chen, ACS Appl. Mater. Interfaces 8 (2016) 9254–9261. [45] L. Mao, Q. Chen, Y. Li, Y. Li, J. Cai, W. Su, S. Bai, Y. Jin, C.-Q. Ma, Z. Cui, L. Chen, Nano Energy 10 (2014) 259–267. [46] Y.X. Li, D.P. Qian, L. Zhong, J.D. Lin, Z.Q. Jiang, Z.G. Zhang, Z.J. Zhang, Y.F. Li, L.S. Liao, F.L. Zhang, Nano Energy 27 (2016) 430–438. [47] A. Guerrero, L.F. Marchesi, P.P. Boix, S. Ruiz-Raga, T. Ripolles-Sanchis, G. GarciaBelmonte, J. Bisquert, ACS Nano 6 (2012) 3453–3460. [48] Y. Vaynzof, D. Kabra, L. Zhao, L.L. Chua, U. Steiner, R.H. Friend, ACS Nano 5 (2011) 329–336. [49] M. Kroger, S. Hamwi, J. Meyer, T. Riedl, W. Kowalsky, A. Kahn, Org. Electron. 10 (2009) 932–938. [50] Y.J. Lee, J. Yi, G.F. Gao, H. Koerner, K. Park, J. Wang, K.Y. Luo, R.A. Vaia, J.W.P. Hsu, Adv. Energy Mater. 2 (2012) 1193–1197. [51] J. Meyer, A. Shu, M. Kroger, A. Kahn, Appl. Phys. Lett. 96 (2010) 133308.
the active layer and the interlayers are critically important in determining the band bending in the active layer. Although the cross-sectional SKPM measurements are here demonstrated in a model device, i.e., P3HT: PCBM inverted OPV device, this method can be used for other solar cells or thin-film optoelectronic devices. The potential challenges of the cross-sectional SKPM method are to keep the device under investigation in operando states and to improve the imaging resolution. It is important to maintain operating status of the device, which is usually hindered by the fact that many thin-film optoelectronic devices suffer from environmental instability. Another challenge is to push the limit in spatial and energy resolution. The deconvolution protocol described in this manuscript helps to alleviate the convolution effect, however, for small interfacial features less than 10 nm, intrinsically sharper tips and low noise level microscope systems are desired. Notes Conflicts of interest None. Acknowledgment Q.C. thanks Yi Zhang at the Electron Microscope Lab of SINANO, CAS for her assistance with fabricating the device cross section and the SEM characterization. Wenxian Yang and Dr. Jinhua Cai at SINANO are acknowledged for helpful discussion on deconvolution process. This work was supported by the Ministry of Science and Technology of China (Grant No. 2016YFA0200703), the CAS Research Equipment Development Program (YZ201654), and the National Natural Science Foundation of China (Grant Nos. 21625304, 51473184 and 11504408). Partial support from Collaborative Innovation Center of Suzhou Nano Science and Technology (CICSNST) is also appreciated. Q.C. acknowledges Collaborative Academic Training Program for Post-doctoral Fellows support from CICSNST. Appendix A. Supplementary material Supplementary data associated with this article can be found in the online version at http://dx.doi.org/10.1016/j.nanoen.2017.08.050. References [1] Z. He, C. Zhong, S. Su, M. Xu, H. Wu, Y. Cao, Nat. Photon. 6 (2012) 591–595. [2] Z. He, B. Xiao, F. Liu, H. Wu, Y. Yang, S. Xiao, C. Wang, T.P. Russell, Y. Cao, Nat. Photon. 9 (2015) 174–179. [3] H.L. Yip, A.K.Y. Jen, Energy Environ. Sci. 5 (2012) 5994–6011. [4] S. Li, L. Ye, W. Zhao, S. Zhang, S. Mukherjee, H. Ade, J. Hou, Adv. Mater. 28 (2016) 9423–9429. [5] W.C. Zhao, D.P. Qian, S.Q. Zhang, S.S. Li, O. Inganas, F. Gao, J.H. Hou, Adv. Mater. 28 (2016) 4734–4739. [6] J. Wang, L. Xu, Y.-J. Lee, M.D.A. Villa, A.V. Malko, J.W.P. Hsu, Nano Lett. 15 (2015) 7627–7632. [7] J.G. Tait, U.W. Paetzold, D. Cheyns, M. Turbiez, P. Heremans, B.P. Rand, ACS Appl. Mater. Interfaces 8 (2016) 2211–2219. [8] P. Büchele, M. Morana, D. Bagnis, S.F. Tedde, D. Hartmann, R. Fischer, O. Schmidt, Org. Electron. 22 (2015) 29–34. [9] G.F.A. Dibb, M.A. Muth, T. Kirchartz, S. Engmann, H. Hoppe, G. Gobsch, M. Thelakkat, N. Blouin, S. Tierney, M. Carrasco-Orozco, J.R. Durrant, J. Nelson, Sci. Rep. 3 (2013) 3335. [10] I. Zonno, A. Martinez-Otero, J.C. Hebig, T. Kirchartz, Phys. Rev. Appl. 7 (2017) 034018. [11] R. Saive, M. Scherer, C. Mueller, D. Daume, J. Schinke, M. Kroeger, W. Kowalsky, Adv. Funct. Mater. 23 (2013) 5854–5860. [12] T. Glatzel, D.F. Marron, T. Schedel-Niedrig, S. Sadewasser, M.C. Lux-Steiner, Appl. Phys. Lett. 81 (2002) 2017–2019. [13] J. Lee, J. Kong, H. Kim, S.-O. Kang, K. Lee, Appl. Phys. Lett. 99 (2011) 243301. [14] V.W. Bergmann, S.A. Weber, F. Javier Ramos, M.K. Nazeeruddin, M. Gratzel, D. Li, A.L. Domanski, I. Lieberwirth, S. Ahmad, R. Berger, Nat. Commun. 5 (2014) 5001. [15] C.-S. Jiang, M. Yang, Y. Zhou, B. To, S.U. Nanayakkara, J.M. Luther, W. Zhou, J.J. Berry, J. van de Lagemaat, N.P. Padture, K. Zhu, M.M. Al-Jassim, Nat. Commun.
460
Nano Energy 40 (2017) 454–461
Q. Chen et al. Qi Chen received his Ph.D. degree in physics from University of Science and Technology of China (USTC) in 2014. He then worked as a postdoctoral researcher at Suzhou Institute of Nano-Tech and Nano-Bionics (SINANO), Chinese Academy of Sciences and University of Washington, Seattle from 2014 to 2017. He is currently an associate professor at SINANO. His research interests focus on interface characterization of thin-film optoelectronic devices by scanning probe microscopy.
Changqi Ma received his Ph.D. degree at the Technical Institute of Physics and Chemistry, Chinese Academy of Sciences with Professor B.-W. Zhang in 2003. After that he was a postdoctoral research assistant at Heriot-Watt University in Edinburgh, UK, until he joined Professor P. Bäuerle's group at the University of Ulm in 2004 as a Humboldt research fellow. From January 2007 till May 2011, he did his Habilitation at the same institute. He was appointed as a professor of chemistry at SINANO in June 2011. His research mainly focuses on solution printable photovoltaic technologies.
Fengye Ye obtained his B.S. degree from USTC in 2013. He is currently a Ph.D. candidate at USTC, co-supervised by Prof. Liwei Chen and Prof. Yi Xie. His research focuses on synthesis and characterization of perovskite solar cells and light-emitting diodes.
Yanfei Zhao received her Ph.D. degree from the International Center for Quantum Materials, Peking University. She is currently an Assistant Professor at Vacuum Interconnected Nanotech Workstation at SINANO. Her research interests focus on the surface and interface of semiconductor and related materials.
Junqi Lai received his B.S. degree in condensed matter physics from USTC in 2015. He is currently a Ph.D. candidate at SINANO. His research focuses on atomic force microscopy characterization and numerical simulation of charge carrier transport in organic photovoltaics devices.
Yi Xie received her B.S. from Xiamen University in 1988 and Ph.D. from USTC in 1996. She is currently a Principal Investigator in Department of Nanomaterials and Nanochemistry, Hefei National Laboratory for Physical Sciences at Microscale, a full professor in Department of Chemistry, USTC, and a member of Chinese Academy of Sciences. Her research interests focus on the design and synthesis of inorganic functional solids with efforts to modulate their electronic and phonon structures.
Pan Dai received her Ph.D. in Microelectronics and Solidstate Electronics from SINANO in 2016, and then worked as a postdoctoral researcher at SINANO for III-V multi-junction solar cell research. Her research interests focus on the MBE growth of III-V compound semiconductor materials and solar cell devices.
Liwei Chen obtained his B.S. from USTC in 1993, MS from Peking University in 1996, and Ph.D. from Harvard University in 2001. After working as a joint postdoctoral research fellow at Columbia University and IBM T.J. Watson Research Center, he started his independent research career as an assistant professor at Department of Chemistry and Biochemistry at Ohio University in 2004. He moved to SINANO as a professor in 2009. His current research interests focus on energy nanotechnology including thin-film solar cells and lithium batteries.
Shulong Lu obtained his Ph.D. degree from the Institute of Semiconductor of Chinese Academy of Sciences in 2003. From 2003–2008, he worked in Paul-Drude institute of Berlin and in Waseda University in Tokyo as a post-doctoral researcher and a visiting lecturer, respectively. He joined SINANO in 2008. His research interests focus on the MBE growth of III-V compound semiconductor material and related devices, such as photovoltaic cells and infrared detectors.
461