Intrinsic and Extrinsic Incorporation of Indium and Single‐Walled Carbon Nanotubes for Improved ZnO‐Based DSSCs

A common goal in renewable energy research is the production of semiconductor layers with matching additives to gain desirable effects and advantages. Important examples are adjustment of the band gap, enhancing the conductivity, tailoring roughness of the semiconductor, or reducing the recombination rate of charge carriers. In this work a twofold approach is demonstrated. First, the intrinsic incorporation of indium into zinc oxide tunes the band gap of the semiconductor layer of the DSSCs. Second, single‐walled carbon nanotubes (SWCNT) as an extrinsic additive improves the charge transport capabilities. To better combine the SWCNTs with the semiconductor layer a protective ZnO shell is grown on the SWCNTs. A combination of current–voltage, incident photon to current efficiency, and electrochemical impedance spectroscopic assays corroborates the positive effects of both additives on the cell performance. The effective charge‐carrier diffusion lengths are enhanced from 0.52 µm for pure ZnO to 5.12 µm for low SWCNT concentrations and 15 mol% of indium. This, as well as the increase in collection efficiency from 30.3% to 62.3% enables enhanced transport in these novel devices. A final device utilizing 0.1 wt% SWCNT@ZnO20nm and In15Zn85O yields a current density of 16.49 mA cm−2 and an efficiency of 4.46%.

nanotubes (CNT) (η: 2.70%), [29] copper (η: 2.03%), [30] and magnesium (η: 4.20%) [24] have been explored in an effort to reduce charge recombination losses. Especially, carbon nanotubes (CNTs) have emerged as the doping material of choice to increase the conductivity and enhance charge transport in semiconductors. [31][32][33] CNTs feature good conductivity and high chemical stability, which is a key incentive for their application in solar cell technologies. [32] To the best of our knowledge, simultaneous intrinsically and extrinsically concentrationdependent doping of ZnO has not been shown yet.
In this work, we report the incorporation of single-walled carbon nanotubes (SWCNTs) into the semiconductor layer of ZnO-based DSSCs. To optimize the electron injection kinetics as well as the charge collection efficiencies, on one hand, and to match the energy levels of the dye and semiconductor, on the other hand, ZnO was intrinsically modified with indium. [25] The corresponding binary indium-zinc oxide (IZO) was combined with SWCNTs, coated with a ZnO protective shell, and incorporated into DSSCs (Figure 1). IZO featuring an indium content ranging from 0 to 80 mol% was prepared via flame spray pyrolysis (see the Experimental Section), which decreases the band gap from 3.3 to 2.1 eV. [25] For the ZnO-based devices, single-layers were used and the devices were investigated by means of photophysical, electrochemical, and microscopic techniques. For the IZO-based devices, different CNT concentrations and different layer thicknesses were applied, with the objective to optimize cell performance. Important to note is that IZO-devices featuring small amounts of SWCNTs gave rise to the best figures-of-merit among all devices.

Optimizing ZnO using SWCNTs Featuring ZnO Shells
In the first step, we optimized the concentration of SWCNTs within ZnO films. We employed SWCNTs protected by ZnO shells of either 10 nm SWCNT@ZnO 10nm or 20 nm SWCNT@ZnO 20nm thickness, which were prepared as reported in literature. [34] To verify the presence of the ZnO shell on the SWCNT bundles and determine the layer thickness thereof we turned to scanning electron microscopy (SEM). Figure 2 corroborates that the ZnO shell covers the SWCNT bundles segmentally with a thickness of ≈10 nm for SWCNT@ZnO 10nm and 20 nm for SWCNT@ZnO 20nm (for additional images and details on the identification of ZnO see Figures S1 and S2 in the Supporting Information). Following these results, pastes with five different concentrations of SWCNT@ZnO 10nm or SWCNT@ZnO 20nm in the range from 0 to 1.5 wt% were doctor bladed onto FTO substrates. These electrodes (4 µm) were tested in a single-layer device setup featuring N719, an iodine-based electrolyte, and a platinum counter electrode (see the Experimental Section for details). Devices fabricated from pure ZnO were used as reference. Figures-of-merit and corresponding current density-voltage (J-V) curves are presented in Figure 3, as well as, Figures S3 and S4 in the Supporting Information, respectively.
The addition of 1.5 wt% SWCNT@ZnO 10nm reduces the open circuit voltage (V oc ) from 0.76 V found for the reference device to 0.7 V. At the same time, the short-circuit current density (J sc ) increases from 2.34 for the ZnO reference to 3.27 mA cm −2 for 0.1 wt% of SWCNT@ZnO 10nm . Beyond this point, J sc steadily decreases to reach 2.92 mA cm −2 for 1.5 wt% SWCNT@ZnO 10nm . Independent from that, the fill factor (FF) initially drops from 60.8% for the ZnO reference to 51.9% for devices featuring 0.5 wt% SWCNT@ZnO 10nm s, and then grows to 59.6% in the case of the highest SWCNT@ZnO 10nm concentration, namely 1.5 wt%. When turning to the photocurrent efficiency (η), a similar trend as noted for the J sc is observed; η increases from 1.08% for the ZnO reference to 1.37% for 0.1 wt% SWCNT@ZnO 10nm .
For SWCNT@ZnO 20nm a different trend is concluded. V oc of these devices remains nearly constant at around 0.77 V for all SWCNT concentrations. J sc rises from 2.34 mA cm −2 for the ZnO reference to 2.83 and 3.60 mA cm −2 for the 0.1 and 1.5 wt% SWCNT@ZnO 20nm devices, respectively. FFs for the SWCNT@ZnO 20nm devices show a more subtle trend compared to the SWCNT@ZnO 10nm devices with values between 58% and 61%. The efficiency η follows the trend of the J sc with values of 1.08% for the ZnO reference and 1.28% for 0.1 wt% SWCNT@ZnO 20nm as well as 1.67% for 1.5 wt% SWCNT@ ZnO 20nm . These findings suggest that the highest efficiencies are obtained with 0.1 wt% SWCNT@ZnO 10nm , on the one hand, and 1.5 wt% SWCNT@ZnO 20nm , on the other hand. To relate the decrease in V oc to the acquired data it is important to understand what influences the open circuit voltage of a cell. V oc is defined as the difference between the quasi fermi-energy level of the semiconductor's CB and the redox potential of the utilized electrolyte. [35] Therefore, changes regarding the redox potential of the electrolyte [36] or the CB energy level of the semiconductor [25] will ultimately change V oc . In the current case, the decrease is explained by incorporating SWCNTs into the ZnO semiconductor. By virtue of the lower CB energy of SWCNTs relative to ZnO, [37,38] V oc is expected to drop as the SWCNT concentration is increased, which is exactly what we see. Due to the thicker ZnO shells for the SWCNT@ZnO 20nm devices the amount of pristine SWCNT is lower, which in turn reduces the drop in V oc for this series.
J sc of a cell correlates with the number of injected and transported charges, which is reduced by recombination processes  across the different interfaces within the cell. [39,40] Therefore, any attempts to enhance charge injection and facilitate transport or to hamper charge recombination should increase J sc . In the case of SWCNT@ZnO, the noted increase in J sc is likely to be a combination of both alterations. First, the higher conductivity of the SWCNTs and the lower lying CB should boost the charge injection and transport. [41][42][43] Second, the ZnO should hamper the recombination with the electrolyte due to a higher CB of ZnO. The results observed for J sc corroborate this hypothesis. For FF and η of measured devices the values are calculated according to the following equations where V MPP is the voltage at the maximum power point, J MPP is the current density at the maximum power point, and P Light is the power of incident light, which is 100 mW cm −2 . [44] Large increases in only V oc or J sc will lead to a reduction in FF. Additionally, shunt and series resistances play an important role in the case of FF but will only be mentioned briefly here. In the case of η, it is entirely dependent on the maximum power point. As such, changes in V oc and J sc govern η, although FF is known to play an important role as well. In terms of FF a slight drop is noted after SWCNT addition, which correlates with the increase in J sc . η follows the same trend as J sc , which is not surprising looking at Equation 2 and the way the maximum power point shifts to larger values as J sc is increased ( Figure S5, Supporting Information). To fully understand the trend of the J sc electrochemical impedance spectroscopy (EIS) measurements were conducted and the obtained key parameters are presented in Figure 4. In the case of SWCNT@ZnO 10nm , both light resistance (R w ), which is assigned to the charge injection and charge transport, and dark resistance (R k ), which is designated to the charge recombination, follow similar trends. Most importantly, R w slightly reduces from 202 Ω for the ZnO reference to 199 Ω for 0.1 wt% SWCNT@ZnO 10nm . R k increases from 290 to 312 Ω for the same concentrations. The electron lifetime (τ ) is elongated to 0.91 ms for all concentrations of SWCNT@ZnO 10nm . Similarly, the collection efficiency (η coll ) is boosted by ≈20% after the addition of SWCNTs. Finally, the effective chargecarrier diffusion length (L eff ) is reduced in the presence of small SWCNT@ZnO 10nm concentrations before starting to increase and reaching a maximum as the SWCNT@ZnO 10nm concentration approaches 1.5wt%. The increase in R k and decrease in R w at the lowest SWCNT concentration nicely corroborates the increase in J sc . On one hand, from a lower R w we postulate a better charge transport through the layers via highly conductive channels, which is supported by the increase in collected charges η coll . [45] On the other hand, the increase in R k indicates a reduction in charge recombination. This is most probable due to the ZnO shells surrounding the SWCNTs, shielding them from the electrolyte. With ZnO having a higher CB edge, the back reaction becomes less favorable. Independent support for this assumption comes from an increase in electron lifetime within the semiconductor layer together with a rise in η coll . The impact in the case of the SWCNT@ZnO 20nm is somehow different. Relative to the ZnO reference, R w and R k rise at 0.1 wt% SWCNT@ZnO 20nm . From here on, R w and R k drop steadily to values that are lower than those seen for the ZnO reference. τ stays constant at around 0.75 ms up to a SWCNT@ZnO 20nm concentration of 1.0 wt%, but rises afterwards to 0.91 ms. L eff follows the same -albeit weaker -tendency as noted for SWCNT@ZnO 10nm . For 0.1 wt% SWCNT@ZnO 20nm , it initially decreases, yet for concentrations exceeding 0.1 wt% L eff is enlarged until it reaches 0.73 µm at the highest concentration. η coll and J sc go hand-in-hand, that is, revealing two maxima at 0.1 and 1.5 wt% SWCNT@ZnO 20nm with 32%, and 36%, respectively. The subtle increase in R w at low SWCNT@ ZnO 20nm concentrations seems to be out of line at first. However, the increase in R k and the fact that it is significantly larger than R w prompts to the unlikelihood of recombination. The subsequent reduction of both resistances with increasing concentration keeps R k larger than R w and, in turn, improves charge transport. This is reflected in the rise in η coll and J sc . In the case of the thicker ZnO shells, which surround the SWCNTs, the improvement in L eff as well as τ is only observable for higher concentrations. Due to the nature of the sample, the pure CNT weight cannot be precisely determined. Therefore, less SWCNT is present in these layers, meaning higher concentrations are needed to see the same effects as for SWCNT@ZnO 10nm .
Both series clearly indicate that η coll for ZnO-based devices is far from optimum. An immediate consequence is the moderate efficiency, which is typically reported for such devices. [46] In order to overcome the "η coll " bottleneck, we revisited our previous study of incorporating indium into ZnO, with the goal to match the band gap of the semiconductor with that of N719. [25] Our previous studies have documented that a systematic lowering of the quasi Fermi energy level by means of variable indium-containing ZnOs (IZOs) is a successful approach to match the energy levels within DSSCs.
With the spray pyrolysed IZO, a batch-to-batch optimization is necessary to accurately match the dye and semiconductor energy levels. In addition, we checked the IZO and compared our results with values already published in the literature. [25] The obtained data matched the expected results nicely -Figures S15 and S16 in the Supporting Information.

Optimizing ZnO using Indium and SWCNTs
In the second step, previous measurements were taken into account to disclose any synergetic effects: SWCNT@ ZnO 10nm and SWCNT@ZnO 20nm were employed at the best performing concentration, namely 0.1 and 1.5 wt%, respectively, and incorporated into single-(3 µm), double-(5.8 µm), and triple-layer (9 µm) devices featuring In 15 Zn 85 O. To verify the presence of SWCNTs within the semiconducting IZO we turned to microscopic studies and recorded both atomic force   Figure 5. SWCNTs, which merged into the semiconductor layer, are clearly discernable in both images. Additional images (Figure 6 and Table 1) revealed the surface roughness at different concentrations of SWCNT@ZnO 20nm , that is, 0, 0.1, and 1.5 wt%. From these images it is evident that the layer roughness decreases with the SWCNT@ZnO 20nm concentration. It is likely that this relates to the sintering process, during which IZO not only attaches to the FTO surface, but also to the ZnO shells of SWCNTs. [34,47] To further prove the presence and incorporation of SWCNTs into IZO we turned to SEM studies. For both SWCNT@ZnO 10nm and SWCNT@ZnO 20nm the images in Figure 2c,f, respectively, confirm the presence of both types of SWCNTs. In accordance with the AFM images, IZO appears as a homogeneous nanoporous layer featuring good connectivity between the ZnO shells and the bulk material.
Next, we continued with the construction of DSSC devices featuring both IZO and SWCNT. Figure 7 shows the J-V curves measured for the SWCNT@ ZnO 10nm and SWCNT@ZnO 20nm devices. For the SWCNT@ ZnO 10nm and SWCNT@ZnO 20nm cells featuring 0.1 wt% SWCNTs, similar trends are observed for the figures-of-merit presented - Figure 8. A V oc decrease from 0.76 V for the ZnO reference is present upon SWCNT addition and increasing layer thickness. For example, the values for SWCNT@ZnO 10nm and SWCNT@ZnO 20nm are 0.59 and 0.60 V, respectively. At the same time, J sc is augmented to 13.54 mA cm −2 for the SWCNT@ZnO 10nm triple-layer cells and 16.59 mA cm −2 in the case of SWCNT@ZnO 20nm .
Noteworthy is that the J sc of the triple-layer device featuring SWCNT@ZnO 10nm is slightly lower than the double-layer devices with a value of 13.94 mA cm −2 . The FF shows a slight decline for all devices, especially noticeable for thicker layers. The only exception is the SWCNT@ZnO 10nm triple-layer cell yielding the highest FF of 53%. A significant jump is seen for the η of the cells when going from single-to double-layer cells. This is followed by a more elusive jump for the triple-layer cells.
To be exact, η values reach 4.21% for SWCNT@ZnO 10nm and 4.46% for SWCNT@ZnO 20nm . Changes in V oc are explained through incorporation of SWCNTs and the increase in surface trap states, especially within thicker layers. Such surface trap states lead to an increase in recombination with the electrolyte  and, therefore, to a lowering of V oc . [48] The boost in J sc stems mainly from the SWCNT incorporation as discussed above as well as the thicker layers. Due to the mesoporous structure of the semiconductor layers an increase in thickness results in a magnification of absorption sites within the layers. This results in an overall higher dye loading, which directly increases J sc of the device. [21] For the FF a slight decline is observed, which stems from the J sc increases as proven in Equation 1. η unsurprisingly follows a very similar trend as the J sc . The increase in J sc leads to up-shifting of the maximum power point. According to Equation 2, this increases η. In a direct comparison, triple-layer cells utilizing SWCNT@ ZnO 20nm give rise to a 6% improved power-conversionefficiency compared to those with SWCNT@ZnO 10nm . Here, the corresponding ηs are 4.46% and 4.21%.
Similar trends evolve at higher concentrations of 1.5 wt% - Figure 9. With increasing layer thickness, V oc decreases, which is more pronounced for SWCNT@ZnO 10nm than for SWCNT@ ZnO 20nm . The thinner ZnO shell is likely to be responsible here. As the semiconductor layers get thicker and thicker, higher J sc s are recorded. Mono-layer devices show a similar 10.7 mA cm −2 , while thicker layers of SWCNT@ZnO 20nm yield 16.21 mA cm −2 compared to 15.27 mA cm −2 for layers of SWCNT@ZnO 10nm . For the higher SWCNT concentration FFs remain nearly constant throughout our assays: 50% for SWCNT@ZnO 10nm and 46% for SWCNT@ZnO 20nm . For the triple-layer devices the highest ηs are 4.13% and 4.54% for SWCNT@ZnO 10nm and SWCNT@ZnO 20nm , respectively.
To gain further insights into the effects that the incorporation of SWCNTs into IZO has, we probed the devices by means of EIS. All measured and calculated data for the J-V and EIS measurements are shown in Figure 8 and Table S1 (Supporting Information) for 0.1 wt% SWCNT and in Figure 9 and Table S2 (Supporting Information) for 1.5 wt% SWCNT. Comparing the EIS data for 0.1 and 1.5 wt% SWCNTs, the only significant difference, which is discernable, is that for τ. At first glance, both concentrations exhibit similar trends. R k as well as R w decrease for thicker layers. In the low SWCNT concentration regime R k drops from 106 to 56.  for SWCNT@ZnO 20nm . For the final two parameters, L eff and η coll , SWCNT@ZnO 10nm devices show slightly larger values than SWCNT@ZnO 20nm devices.
Considering our EIS findings the R k decrease implies less hindered charge recombination for thicker layered-devices. This is not surprising as thicker layers result in more surface trap states, which, in turn, increase the probability of recombination with the electrolyte. However, the simultaneous decrease in R w implies a better charge transfer. This mainly stems from the SWCNT incorporation into the IZO semiconductor as it generates highly conductive channels. This is verified by the increase in J sc as well as η. More layers result in an increase in both L eff and τ due to the need for electrons to travel longer distances to reach the electrode back contacts, which is observed. Finally, better η coll indicates a superior charge collection and, as such, documents the success of our work. All of our devices perform much better than devices made with conventional ZnO, which yielded 1.08% as a reference point. By modifying ZnO intrinsically with 15 mol% indium and extrinsically with 0.1 wt% SWCNT efficiencies of 4.13% for SWCNT@ZnO 10nm and 4.46% for SWCNT@ ZnO 20nm were determined. Considering that the largest J sc emerged in the case of the triple-layer devices featuring 0.1 wt% of SWCNT@ZnO 20nm , we, thus, conclude that low SWCNT concentrations are already the best choice in terms of thick layer performance.

Conclusions
In the first step, the incorporation of SWCNTs into layers of different ZnO-based semiconductors offers desirable advantages in terms of improving DSSCs. This work utilized SWCNTs outfitted with a 10 nm (SWCNT@ZnO 10nm ) and 20 nm (SWCNT@ZnO 20nm ) thick shell of ZnO. The incorporation of these SWCNTs into ZnO-based electrodes resulted in longer τ and higher η coll due to reduced recombination and better charge transport. This yielded boosted J sc s with only a slightly negative impact on the V oc owing to the lower CB level of the SWCNTs. Overall ηs for these devices ranged from 1.08% for just plain ZnO to 1.37% and 1.67% in the presence of 0.1 wt% of either SWCNT@ZnO 10nm or SWCNT@ZnO 20nm , respectively.
In the next step zinc in ZnO was replaced by indium to improve the J sc s and η coll s as well as to adjust the energy levels of dye and semiconductor. A ratio of 15 mol% indium to 85 mol% zinc in In 15 Zn 85 O was found to be the optimum, yielding a J sc of 9.81 mA cm −2 and η of 3.66% relative to 2.34 mA cm −2 and 1.08% for bare ZnO. In the final step, SWCNT@ZnO 10nm and SWCNT@ZnO 20nm , were integrated into IZO-based DSSCs featuring In 15 Zn 85 O. Here, the effects from the presence of SWCNT shows their full potential as the charge transport and collection efficiency were largely facilitated in the In 15
IZO Nanopowder Synthesis: Stock solution A (zinc acetylacetonate hydrate, acetic acid in 2-methoxyethanol) and B (indium acetate, ethanolamine in 2-methoxyethanol) were sonicated for 15 min before mixing. The precursor solution was heated to 70 °C and stirred for 15 min until a clear solution was obtained. The solution was then spray ignited (5 mL min −1 ; dispersion gas: O 2 , 5 L min −1 ; 1.5 bar) and evaporated by a stoichiometric CH 4 (3 L min −1 )/O 2 (1.5 L min −1 ) flame. The resulting nanopowders were filtered and vacuum dried on a polyimide membrane filter. A detailed description of this standard process is given by Kunzmann et al. [25] and a description of the burner by Engel et al. [49] Device Fabrication and Characterization: FTO substrates were cleaned by a standard procedure immersing the slides subsequently into solutions of acetone, detergent (deconex FPD 120, 1% vol. solution in 150 mL deionized water), pure deionized water, and isopropanol. The solutions were sonicated for 15 min each (Elma, Elmasonic P), and finally the substrates were dried under a nitrogen flow. To cleanse the electrode surface of remaining organic waste the slides were treated with a UV-ozone cleaner for 18 min (model 42-220, Jelight Company).
For preparation of the IZO-based electrodes a standard paste recipe published by Ito et al. was used. [16] The SWCNT@ZnOs were dissolved in ethanol (1.25 mg mL −1 for SWCNT@ZnO 10nm and 2.5 mg mL −1 for SWCNT@ZnO 20nm ) and sonicated for 30 min. Following this procedure the SWCNT@ZnO solution was added accordingly to the prepared ZnO and IZO pastes. Utilizing a circular scotch tape with a diameter of 5 mm and thickness of 50 µm, the IZO pastes were doctor bladed onto the conducting side of FTO substrates. FTOs were heated from room temperature up to 150 °C with a ramp of 10 °C min −1 , holding the temperature for 10 min. Then, utilizing a ramp of 15 °C min −1 , the temperature was increased to 325 °C and kept constant for 5 min. Next, the slides were heated up to 375 °C with a ramp of 5 °C min −1 , holding the temperature for 5 min. Afterwards the temperature was increased up to 450 °C at 7°C min −1 and kept for 30 min. Finally, a temperature of 500 °C was obtained by a ramp of 5 °C min −1 . The samples were kept at this temperature for 15 min, after which they were slowly cooled to 80 °C. The slides were immersed into a N719 dye solution (5 × 10 −4 m), consisting of a 50:50 mixture of acetonitrile and tert-butanol, for 240 min in order to achieve optimal coverage of the semiconducting surface. For the counter electrodes, FTO slides with two 1 mm diameter holes at the edge of the semiconductor layer were prepared. The slides were cleaned with the aforementioned procedure. Next, a thin film of 26 µL H 2 PtCl 6 (0.5 mmol in isopropanol), prepared from chloroplatinic acid hydrate was drop casted onto the counter electrode slides. The FTO substrates were dried in air prior to baking at 400 °C for 20 min with a ramp of 40 °C min −1 . For the final step, both electrodes were sealed together, utilizing a transparent film of Surlyn (DuPont Ltd., UK), featuring a 6 mm hole for the active layer. An electrolyte solution, in a 85:15 v/v mixture of acetonitrile and valeronitrile, consisting of 0.6 m 1,2-dimethyl-3-propylimidazolium iodide 99%, 0.5 m 4-tert-butylpyridine 96%, 0.1 m iodine, and 0.1 m lithium iodine 98%, was added into the cells by capillary force using the drilled holes. The final cell was sealed immediately after, using Surlyn and a piece of microscope glass.
Photocurrent measurements were conducted under dark and AM 1.5 illumination utilizing a custom made solar simulator, featuring an adjustable Xe lamp source (from 350 to 1000 Watt, LOT) and fitting AM 1.5 filter. Current voltage characteristics were recorded using a potentiostat/galvanostat (PGSTAT30, Autolab) in a range from −0.8 to 0.2 V. The measurements were conducted utilizing a black mask after calibrating the aforementioned apparatus with a reference silicon solar cell (SRC 1000 TC K KG5 N, VLSI standards) at room temperature. Electrochemical impedance measurements were performed using the same potentiostat under dark and AM 1.5 conditions with varying frequencies (from 0.01 Hz to 100 kHz). To ensure a linear response of the measured cell, the AC amplitude was set to 10 mV and measurements were conducted under V oc conditions. Following each impedance measurement, a single current voltage measurement was recorded to corroborate the stability of the device figure-of-merit. The acquired Nyquist plots were fitted with an appropriate electric circuit model utilizing the NOVA 1.11 software. Incident photon to current efficiency (IPCE) was determined using a model 70 104 apparatus (Newport). In order to determine the thickness of the semiconductor layer, the cells were opened after measurement and investigated with a profilometer (Dektak XT, Bruker). With the Vision 64 software the height profile of each electrode was obtained.
Microscopic Investigation: Atomic force microscopy measurements were conducted with a JPK Nanowizard 4 Nanoscience microscope holding a Multi75-G cantilever from Budget-Sensors with a resonance frequency of 75 kHz and a tip radius below 10 nm. Samples were prepared on FTO slides according to the described procedure above.
For the scanning electron microscopy measurements SWCNT@ ZnO 10nm and SWCNT@ZnO 20nm were dispersed in isopropanol by handshaking to minimize the mechanical peel-off of ZnO shell and then drop-casted on silicon wafer substrates to obtain the SEM images except for Figure 2c,f, by which the samples were prepared according to the described procedure above. The SEM experiments were conducted on a Delta-SEM (Demonstrator, Carl Zeiss Microscopy, Germany) with 1 keV primary beam energy and a working distance of 3 mm.

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