Planar-defect-rich zinc oxide nanoparticles assembled on carbon nanotube films as ultraviolet emitters and photocatalysts

Structural defects in zinc oxide (ZnO) nanoparticles are complex and hard to be controlled during the synthesis, however, diversifying the chemical and physical performances. Here we report a rapid and low-temperature deposition method to fabricate planar-defect-rich ZnO nanoparticles on freestanding and aligned carbon nanotube films, different from common treatments which remove structural defects as many as possible. The defect energy states are very close to the valence band of ZnO and serve as recombination centers for a nearly monochromatic ultraviolet luminescence within a wavelength range of 373–376 nm. The absence of point defects, especially of oxygen vacancies whose energy level is <1 eV below the conduction band, allows photoinduced electrons and holes to take parts in possible photocatalytic reactions rather than to recombine at the shallow energy levels of planar defects.

Z inc oxide (ZnO) is a wide band gap semiconductor with potential applications in optoelectronics, transparent electronics, and light emitting devices. Due to the inevitable defects or low degree of crystallization, its luminescence usually exhibits a ultraviolet (UV) emission together with a wide band in the visible region 1,2 . Efforts to enhance either the UV or visible emission of ZnO include the coating of other materials to adjust the band structure [3][4][5][6] and the improvement of crystallinity to tailor the emission behavior 7 . Similar modulation on the electronic structure is also applied to enhance the photocatalytic performance of ZnO, among which metal ions doping is widely used [8][9][10] . However, there exist complicated factors like the electronic structure, dopant concentration, energy levels of dopants, and d electronic configuration of the ions that contribute to the photocatalytic activities 9 . This makes it rather challenging to fabricate high performance photocatalysts based on pure ZnO nanostructures.
Recently, attempts on ZnO/carbon nanotube (CNT) hybrid structures have also attracted considerable attention and shown that the presence of CNT can influence the growth and thus the structure of ZnO nanoparticles. For example, various morphologies including ZnO nanowires 11 or branched nanorods 12 on CNT surfaces, ZnObeaded CNTs 13 , and coaxial heterostructures 14 were synthesized by using atomic layer deposition 12,14,15 , RF magnetron sputtering and substrate heating 13,[16][17][18] , or a solution based treatment 19,20 . These structures yielded an enhanced UV to visible emission ratio 12,15,[17][18][19] owing to the improved surface-plasmon mediated emission, new defect emission, or a good crystalline quality of ZnO. The presence of CNTs also enhanced the photocatalytic performance of ZnO by the efficient separation of photoinduced charge carriers [21][22][23] . Today, with the development of mechanically strong, flexible, and freestanding super-aligned CNT films [24][25][26] , especially through the socalled method of array spinning [27][28][29][30] , new ZnO/CNT hybrid structures can be developed by taking advantages from the super alignment. Such new hybrid is also expected to possess new luminescent and photocatalytic properties.
Here, we show that ZnO nanoparticles exhibit a large number of basal plane stacking faults rather than other structural defects when being deposited on super-aligned CNT films. The deposition can be rapidly performed at ambient air condition under a temperature of 170uC. The planar defects result in a nearly monochromatic UV emission at 373-376 nm owing to their shallow energy levels as recombination centers of photoinduced electrons and holes. The absence of point defects like oxygen vacancies allows a long lifetime of photoinduced charge carriers, corresponding to a high photocatalytic ability. The hybrid film also exhibits flexibility and superior stability. An annealing treatment in the presence of O 2 can convert the planar defects into oxygen vacancies, and thus shift the luminescence into blue and decrease significantly the photocatalytic performance. The rapid and low-temperature deposition of planar-defect-rich ZnO can be well applied on other freestanding films which contain a large area of carbon sp 2 network and high air permeability.

Results and discussion
Deposition and crystal structure of ZnO nanoparticles. ZnO nanoparticles were deposited at 170uC on freestanding CNT films supported in a semi-opened ceramic boat by a thermal deposition using zinc acetylacetonate (Zn(C 5 H 7 O 2 ) 2 ) as source materials (Fig. 1a). The deposition temperature was determined by the thermogravimetric analysis of Zn(C 5 H 7 O 2 ) 2 ( Supplementary Fig.  S1). The as-produced ZnO/CNT hybrid film can be transferred onto a silicon wafer or a flexible polyethylene terephthalate (PET) substrate with maintaining the flexibility (Fig. 1b,c). Despite the uncoated regions owing to the limited source area at the bottom of the boat, the ZnO deposition was uniform in the coated region. Scanning electron microscopy (SEM) revealed that the ZnO-coated   CNTs had a diameter of ,60 nm ( Fig. 1e and Supplementary Fig.  S2a,b). The CNTs were ,6 nm in diameter and had 2-3 walls 31 , see Supplementary Fig. S5a-c. Here, the ZnO thickness was controlled mainly by the mass of Zn(C 5 H 7 O 2 ) 2 (0.1 g as a default value) and slightly by the deposition time. Energy-dispersive X-ray spectroscopy (EDX) quantitatively showed a larger atom amount of zinc than oxygen, indicating that the defects were formed under the lack of oxygen (Fig. 1f).
Transmission electron microscopy (TEM) showed a certain level of crystallinity for the as-produced ZnO nanoparticles ( Fig. 1g and Supplementary Fig. S2c,d). On one hand, the crystal planes were separated by 0.29 nm, a spacing close to but slightly larger than the idea value between the (002) planes. On the other hand, the crystal planes were not exactly planar as many curved segments were observed. This suggests that the ZnO nanoparticles were formed by stacking the imperfective (002) planes. The stacking faults were detected by TEM, where the numbers of crystal planes entering and leaving a small segment were not equal, corresponding to the missing or appearing of one plane (see the red circles in Fig. 1g and Supplementary Fig. S2c,d). An annealing treatment at 400uC was used to improve the crystallinity by healing the planar structural defects. This resulted in the decrease in the inter-plane distance (0.26 nm) and the flatness of the planes (Fig. 1h). Notice that the annealing did not heal all the defects; there were still a small number of stacking faults left ( Supplementary Fig. S2e,f).
The crystal structure was further examined by X-ray diffraction (XRD) (Supplementary Fig. S3). The strongest XRD peak was observed at 2h < 34.4u, corresponding to the (002) plane, while the other two peaks for the (100) and (101) planes (2h < 31.7u and 36.2u) were weaker. This confirms that the preferential stacking direction was along the (002) plane. However, the (100), (002), and (101) peaks of the as-produced film appeared at 2h about 0.1u smaller than their standards of 31.82u, 34.48u, and 36.30u, respectively, and was not sharp, corresponding to a high level of lattice disorder. This also agreed with the expanded distance of 0.29 nm between the (002) planes. Nevertheless, the XRD confirmed that the ZnO crystals were hexagonal wurtzite structure with lattice constants of a 5 0.325 nm and c 5 0.52 nm (JCPDS card No. . After the annealing, as a large number of stacking faults were healed, the XRD peaks became higher and sharper, and reached the standards much more. Luminescent property of ZnO/CNT hybrid films. The presence of stacking faults had a strong influence on the luminescent property of ZnO. Cathodoluminescence (CL) measurement exhibited a single UV emission at 376 nm for the as-produced film at room temperature (Fig. 2a). This observation differed greatly from recent studies on ZnO nanostructures, where a multi-wavelength emission in the UV and visible regions were often observed 12,32-37 and only few cases were concerning with the nearly pure UV emissions (wavelength ranging from 379 to ,400 nm) 15,18,19 . The nearly monochromatic UV emission was maintained without any change after the film was kept at ambient conditions with relative humidity ,10% for 24 days. However, when the film was annealed, the CL spectrum became dominant blue where the emission peak was at 462 nm, indicating that the types of defect in ZnO had changed (Fig. 2a).
Slightly different from CL, photoluminescence (PL) measurement using a 325 nm laser exhibited a new emission peak at 490 nm for the as-produced film (Fig. 2b). This emission peak was attributed to another type of defect inside the ZnO nanoparticles, and was excited because photons travelled into ZnO deeper than a low voltage (5 kV) electron beam. After the annealing, this emission peak became much stronger and shifted to the wavelength of 485 nm, close to the emission at 462 nm in CL. Therefore we suspect that the annealing treatment converted the stacking faults into other types of defect, probably into those already existed inside the ZnO. There appeared  also a new emission at 632 nm after the annealing, indicating that a third type of defect should exist and that this emission could arise from electron jumps between different defect energy states.
It was found that the luminescence had small dependence on the amount of ZnO deposited on a CNT film. By increasing the mass of zinc source from 0.1 g up to 0.2-0.35 g and using the same holding time of 30 min (see Methods), more ZnO nanoparticles were deposited around the CNTs. Fortunately, owing to the rapid deposition, there always existed planar structural defects in ZnO and the luminescence was still nearly monochromatic UV (Supplementary Fig. S4). Furthermore, as the ZnO particle size was much larger than the CNT diameter and the ZnO was deposited on isolated CNTs rather than bundled CNTs, the quenching effect by a metallic substrate was nearly negligible, corresponding to the high UV emission we have observed.
Free-to-bound transition for UV emission. To identify the electronic nature of the UV emission and therefore to characterize the types of defect in ZnO, PL measurements were performed at temperatures ranging from 5 to 260 K (Fig. 2c,d). For the asproduced and annealed samples, the UV peaks both shifted to a smaller wavelength with decreasing the temperature. By fitting the temperature dependence of the transition energy, it is found that the UV emission was for a free-to-bound transition, i.e., the recombination of an electron from the conduction band with a hole bound to an acceptor state, an eA 0 transition. The eA 0 transition energy is given by where E g (T) 5 E g (0) 2 aT 2 / (T 1 b) is the band gap for ZnO described by the Varshni formula 38 , and E A is the acceptor binding energy above the energy state of valence band. Here E g (0) is the band gap at T 5 0, and a and b are two Varshni coefficients. The measurements were well fitted using equation (1) (see the right panels in Fig. 2c,d), giving E g (0) 2 E A 5 3.322 eV, a 5 3.56 3 10 24 eV K 21 and b 5 220 K for the as-produced film, and E 0 g 0 ð Þ{E 0 A~3 :343 eV, a9 5 6.44 3 10 24 eV K 21 and b9 5 85.3 K for the annealed one. Notice that the change in E A with temperature was ignored as E A =E g , and also that the present measurements were not sufficient to estimate the value for either E g (0) or E A , while their difference was calculated. The increase in a means that the overall influence of thermal expansion and electron-phonon interaction on the fundamental band gap became much stronger after the annealing, as a result of the remarkable change in types of defect. As reflected in the UV emission, the change in a caused the different temperature dependence of the emission wavelength. For example, from 5 to 260 K, the UV emission redshifted from 373 to 376 nm for the as-produced film, but from 370 to 379 nm for the annealed one (Fig. 2c,d). The decrease in b, which is closely related to the Debye temperature, agreed with the observation that the low temperature fitting on E eA 0 was valid at temperatures below 230 and 180 K for the as-produced and annealed films, respectively.
Annealing-induced defect transition. The enhanced blue emission after the annealing arose from the newly formed defects. An energy analysis suggested the existence of another two possible defects of oxygen vacancy (V O ) and zinc vacancy (V Zn ), and one defect transition The transition took place as it was difficult to fully heal the stacking faults into perfect crystallinity by annealing, and thus a large number of V O were formed after the treatment and served as centers of blue luminescence. It was confirmed by the enhanced blue emission in the annealed films. However, there should exist a small number of V O deep inside the as-produced ZnO nanoparticles, as PL detected a small peak in the blue region while CL did not (Fig. 2a,b). The energy level of V O was calculated to be 2.531 and 2.557 eV above the valence band, based on the wavelength of 490 and 485 nm at 260 K for the as-produced and annealed films. The existence of V Zn was owing to the 632 nm emission for the annealed films, where electron jumps from V O to V Zn became possible owing to the increased number of energy states for V O . Therefore, all the observed PL emissions can be represented by a schematic electronic level diagram of the ZnO nanoparticles (Fig. 3).
As the defect types changed after the annealing, the light absorption might also be affected. From the UV-Vis spectra (Supplementary Fig. S5) one can find that the annealed ZnO/CNT hybrid films showed more absorption in the wavelength range of about 380-500 nm, owing to the increased V O . However, both the as-produced and annealed films showed intensive absorption in the UV band of about 230-365 nm, as such absorption was mainly caused by the ZnO bulk itself. The intensive UV absorption obviously has benefits to enhance the output luminescence signals.
Photocatalytic degradation of acetaldehyde. Besides the luminescent property, the photocatalytic degradation of acetaldehyde (CH 3 CHO) was investigated to show another possible application of the hybrid film. The degradation in gas phase was conducted in a tubular vessel reactor operating in a continuous flow mode 39 and all the film samples were sized into 1.8 3 2 cm 2 and transfered onto silicon substrates. By measuring the concentrations of CH 3 CHO and CO 2 , the conversion ratio and rate can be well estimated. For the hybrid film produced by using 0.2 g Zn(C 5 H 7 O 2 ) 2 , the CH 3 CHO concentration decreased very quickly within 20 min from ,370 ppmv to 100-120 ppmv after turning on a 300 W Xe lamp (Fig. 4), corresponding to a conversion ratio up to 67-73%. By increasing the mass of Zn(C 5 H 7 O 2 ) 2 up to 0.35 g to increase the ZnO deposition, the photocatalytic reaction was further enhanced. In the meantime, the CO 2 concentration increased by an amount nearly double the change in CH 3 CHO concentration. Similar to the stability of the luminescence, the photocatalytic ability did not decrease after being kept at ambient condition for weeks.
The photocatalytic ability of the hybrid films was compared with the commercial TiO 2 nanoparticles (P25). Under the same test conditions, the CH 3 CHO concentration just decreased by ,150 ppmv, much less than the decreases by the hybrid films. More interestingly, although the sample area was nearly the same, the total mass of P25 we used was about 25 mg (the mass density of P25 is ,3.8 g cm 23 ), and was one order of magnitude larger than the masses of the hybrid  films (both ,2 mg). The small area density of the hybrid film (0.2-0.6 mg cm 22 ) was owing to the small number density of the superaligned CNT of 5-20 mm 21 (Fig. 1d). This means that the stacking-fault-rich ZnO nanostructures exhibited the much higher photocatalytic ability per unit particle mass than P25. Intuitively, this result seems strange because the shallow eA 0 acceptor levels (Fig. 3) should reduce the catalytic activity when acting as the trapping or recombination centers for photoinduced electrons and holes. Fortunately, the eA 0 acceptor levels were very close to the valence band and far from the conduction band. The large energy gap (E g 2 E A . 3.3 eV, more than 0.1 eV larger than the band gap of TiO 2 ) allowed efficient reactions between the electrons with O 2 and between the holes with organic compound, rather than the direct recombination via the hopping of electrons over such energy gap. Another result of the large energy gap was the significantly longer electron lifetime in ZnO than in TiO 2 40 . In other words, the high photocatalytic ability thanks the lack of V O in the as-produced samples, because the energy level of V O was only ,1 eV below the conduction band (Fig. 3) and preferred the recombination of electrons and holes. It was confirmed by the photocatalytic ability of the annealed hybrid films which only decreased the CH 3 CHO concentration by ,100 ppmv (Fig. 4).
Deposition mechanism of ZnO on sp 2 surfaces. To analyze the deposition of ZnO nanoparticles, a procedure with zero holding time and 0.1 g zinc source was used (there was still a heating stage and cooling stage to allow a certain ZnO deposition, see Methods). The particle size was up to 20 nm (Fig. 5a,b,d,e and Supplementary  Fig. S6). For most cases, the (002) crystal planes were perpendicular to the CNT surfaces (Fig. 5a,b and Supplementary Fig. S6c-f), while very few different cases were observed where the (002) planes had an angle of ,60u with the CNT surfaces. As the deposition method was seed layer free, the lattice matching between ZnO crystal and CNT sp 2 network played the key role in determining the deposition procedure. The perpendicular deposition was suggested to be initialized by lattice matching between the (100) layer with the CNT surface (Fig. 5c). The ZnO lattice constant a 5 0.325 nm is nearly 3/4 (76.3%) of the sp 2 armchair period (3 3 a CC 5 0.426 nm, a CC the C-C bond length), and the constant c 5 0.521 nm is only 5% larger than the double zigzag period (2 ffiffi ffi 3 p |a CC~0 :492 nm). The perfect commensurability can maximize the van de Waals interactions between the (100) layer and the sp 2 surface. For the inclined deposition, the formation of a (101) atom layer can be the most possible solution (Fig. 5f). The ideal lattice constants for the (101) plane are a 5 0.325 nm and ffiffiffiffiffiffiffiffiffiffiffiffiffi a 2 zc 2 p~0 :614 nm, ,3/4 and ,5/2 of the armchair and zigzag periods, respectively. However, as the period along the zigzag direction becomes larger, the possibility of such inclined deposition is much smaller than the perpendicular one.
Effect of deposition substrate. The rapid thermal deposition requires the freestanding substrate to be highly permeable to air. When a silicon substrate was used, ZnO nanoparticles was only    Fig. S7). The particle size was even one order of magnitude larger than the result on the aligned CNT films. In the second example, 4 layers of CNT films were stacked together as a substrate and ZnO nanoparticles were still easily deposited around the CNTs (Supplementary Fig. S7). This means that the deposition method can be widely applied on many different substrates. Furthermore, CL characterization exhibited that the nearly UV emission was maintained for these two examples. Of great interest, the wavelength was 366 nm for the ZnO/GO hybrid structures, of 7-10 nm smaller than the ZnO/CNT hybrid films. It means that the level of defect on the sp 2 surface also played an important role in determining the electronic level of the planar defects in ZnO. Notice that, when non-sp 2 substrates were used, the luminescence differed greatly and monochromatic UV behavior no longer existed as many types of defects were formed in the as-produced structure.
In summary, the superaligned CNT films can be an exciting template to deposit planar-defect-rich ZnO nanostructures. The shallow eA 0 acceptor levels of the planar defects, very close to the valence band, serve as recombination centers for a nearly monochromatic UV emission. The lack of V O resulted in the high photocatalytic ability as the photoinduced carrier charges might have a long lifetime to allow them to react efficiently. The intelligent usage of planar structural defects and avoid of point defects open a new way in the design of multifunctional metal oxides.

Methods
Fabrication of hybrid films. The super-aligned CNT films were drawn out from vertically aligned CNT arrays synthesized by a chemical vapor deposition method 30 . Parameters concerning the growth, including the carbon source, catalyst, and growth temperature and time, were reported by our previous studies 31,41 . The film width ranged from 1 to 2 cm. The film was supported by a copper frame and was placed on a semi-opened ceramic boat. The zinc source Zn(C 5 H 7 O 2 ) 2 was placed inside the ceramic boat and below the CNT film. The default mass of Zn(C 5 H 7 O 2 ) 2 was 0.1 g, or others if specified. ZnO nanoparticles were coated on the CNT films in air by thermal deposition. The furnace temperature was increased from room temperature to 170uC at a heat rate of 8uC min 21 , maintained at 170uC for 30 min (a holding time), and then automatically cooled down. To anneal the as-produced ZnO/CNT hybrid film, a gas flow of 300 sccm Ar was first introduced in to the furnace (3 inch in diameter) for 3 hour. Then the furnace temperature was increased to 400uC at a heating rate of 10uC min 21 , maintained for 90 min, and cooled down to room temperature.
Characterization. The structural characterization was performed with a Hitachi S-4800 field emission SEM (Hitachi Ltd., Tokyo, Japan) and an FEI Tecnai G2 F20 S-Twin TEM (FEI Corp., Hillsboro, USA). CL and PL were applied with a Gatan Mono CL3 system (Gatan Inc., Pleasanton, USA) and PI-PLE-2355/2558 1 PIXIS-256E system (Princeton Instruments, Trenton, USA). A Bruker AXS D8 Advance XRD system (Bruker Corp., Karlsruhe, Germany) and an Oxford EDX detector (Oxford Instruments, Concord, USA) were also used to analyze the hybrid structure. The photocatalysis was performed with a CEL-GPPC continuous flow gas-phase reactor (Beijing CHN EDU AuLight Co. Ltd., Beijing, China) which contained a 300 W Xe lamp. The UV-vis absorption spectra were recorded in a Perkin-Elmer Lambda 950 spectrometer (PerkinElmer Inc., Waltham, USA). The gas concentrations were determined with an online gas chromatograph (SP7800, Keruida Technology Co. Ltd., Beijing, China). All of the photocatalyst samples were sized into 3.5-3.6 cm 2 . A mixture of air and CH 3 CHO was injected into a tubular vessel reactor. The flow rate of air and CH 3 CHO were 12.5 and 5 ml min 21 , respectively. The gas pressure inside the vessel was 0.08 MPa.