Universal model for defect-related visible luminescence in ZnO nanorods

Abstract

We study the optical properties of ZnO nanorods (NRs) fabricated by chemical bath deposition, hydrothermal, and the vapour–liquid–solid method (VLS). Scanning electron microscopy demonstrates differences in the structural properties for the various samples. The optical emission properties are studied by photoluminescence (PL) spectroscopy where all samples are characterized by a UV and visible emission band. The visible emission band is due to defects in the nanorods. VLS samples show a blue shift in the visible region of the PL spectra with respect to the other samples, however, all three samples are fitted with the same three visible Gaussian components under varying percent contributions due to the structural differences. The visible defect components are characterized by blue (B), green (G), and orange (O) states with energies at 2.52 eV, 2.23 eV, and 2.03 eV, respectively. The applicability of this universal model for visible B–G–O defects is tested against the literature and successfully fitted regardless of the fabrication method. Differences in the percentage contribution for the visible B–G–O defects is explained by variations in the fabrication method. This model indicates how defects can be controlled based on the fabrication method. Furthermore, the B state, which is associated with the ‘green luminescence band’, results from a transition from the defect level to the valence band, or possibly a shallow-acceptor, according to photoluminescence excitation measurements. The role of the B state in sensing applications is discussed.

---

1 Introduction

ZnO is a highly investigated transition metal oxide, because of its desirable optical properties^1,2 and surface reactivity.^2–4 This material contains a proportionate number of native defects,⁵ whereby in nanostructured ZnO these defects are central to device function,^6,7 in particular for sensing applications.^8–11 To this aim, if the relationship between optically active defects and the sensing mechanism is understood, then a clear route to high sensitivity optical-based sensors can be realised. While native ZnO defects are central to modelling sensor operation,¹² it is often argued that an ideally responsive sensor will have a high optical quality or a low defect density.^13–15 Two defects highlighted in the literature are Zn vacancies (V_Zn)¹⁶ and O vacancies (V_O).¹⁷ V_O have often been attributed to the ‘green luminescence band’ centred at 2.5 eV,^18,19 and to UV sensing.^12,15 After annealing in an Ar environment an enhancement in the optical quality and the responsivity was measured in ref. 20 and cited as due to a reduction of V_O. A reduction in V_O states²¹ effectively reduces the number of sites for atmospheric O₂ to bond at,²² which is essential for device performance.¹² Ar and O₂ annealing are both reported to reduce V_O and the overall visible photoluminescence (PL) spectra,²¹ but this picture is not consistent across all reports.²³ Recently, it has been shown that the ‘green luminescence band’ is associated with V_Zn,^16,24 while it is also considered to consist of both an V_O and V_Zn component with a molecular type character.^25–27 Therefore, if the ‘green luminescence band’ is related to sensing performance, it is reasonable to assume that the V_Zn should be included in a sensing model.

Hence, clear identification of defect-related visible luminescence will aid in the understanding of the sensing mechanism in order to build better sensors. Often defect-related visible luminescent states are identified by observing general changes in the PL spectra after a post-growth treatment while specific visible luminescent components are not considered in detail.^28–32 In a previous work,² we identified three principle photoluminescent defect states in ZnO nanorods (NRs) fabricated by chemical bath deposition (CBD). These states were labelled as blue (B), green (G), and orange (O), with luminescence with respect to the valence bands at 2.52 eV, 2.23 eV, and 2.03 eV, respectively. The defects related to these states were identified as , and O: V_O. Therefore, from the discussion above, the G and O states are important for sensing applications.

In this work, we show that the B–G–O defect-related visible components can be fitted to the luminescence spectrum of ZnO NRs fabricated by a variety of methods. We fabricated ZnO NRs by the CBD, hydrothermal, and the vapour–liquid–solid (VLS) methods, which were fitted with the above mentioned states under varying percent contributions. Furthermore, we demonstrate that our B–G–O model can be extended to several fabrication methods discussed in the literature. Variations in the contributions from each component are understood in terms of differences in the fabrication method. Therefore, we discuss how individual colour components can be controlled during the fabrication or post-growth treatment process. PL excitation (PLE) measurements indicated that the B state is optically excited at sub-band gap energies, which indicates that this state is a donor-type level and thus can also be included in a sensing model. Thus, our B–G–O model is universal to ZnO NRs and intended to aid in the development of improved sensors; oxidation at the ZnO surface defect sites changes the electronic properties for UV or gas sensing applications.³³ Oxidation of ZnO reduces the electron population in the conduction band, thus reducing the conductivity. Meanwhile, H₂ absorbed at surface O₂ contributes electrons for increased conductivity. Therefore, it is essential to understand how to control defects in ZnO.

2 Experiment

In this work, we consider ZnO NRs grown by chemical bath deposition (CBD), hydrothermal, and the vapour–liquid–solid (VLS) method. CBD samples were fabricated by seeding an SiO₂ substrate with a ZnO nanoparticle solution (0.1 wt% in ethanol) via spin-coating (1000 rpm for 60 s). The seeded substrate was then placed in a solution of 25 mM zinc nitrate hexahydrate [Zn(NO₃)₂·6H₂O] and 25 mM hexamethylenetetramine [C₆H₁₂N₄] (HMTA) at a temperature of 90 °C for 1 h. See ref. 23 for more details.

Under the hydrothermal growth method, a cleaned Si substrate was spin-coated with zinc acetate dihydrate Zn(CH₃COO)₂·2H₂O (∼5 mmol in ethanol) at 3000 rpm for 30 s, then baked at ≈150 °C on a hotplate for 2–3 min. The spin-coating procedure was repeated up to 10 times for seed layer uniformity. A final hotplate annealing step (200–350 °C for 30–40 min) was performed to decompose zinc acetate into ZnO nanocrystals. For the growth of ZnO NWs, substrates with the seed layer were then immersed face down in a Teflon flask containing a solution of 1 : 1 molar ratio of Zn(NO₃)₂, HMTA (25 mM) and 5 mM PEI. The flask was subsequently sealed inside a stainless steel autoclave reactor and placed in a preheated convection oven (Memmert 100 universal) at ≈95–100 °C for ≈12–17 h. See ref. 34 for more details.

The VLS growth of ZnO NWs was performed inside a horizontal quartz tube furnace by carbothermal reduction of ZnO nanopowder on c-plane sapphire substrates. Prior to NW synthesis, cleaned substrates were coated with a Au film (2 ± 1 nm) using a electron-beam evaporator. Coated substrates and the source material (ZnO and C at 1 : 1 weight ratio) were placed on top of an alumina ‘boat’, which was inserted inside a furnace. The samples were annealed in an Ar ambient at 850 °C for 120 min to promote the formation of ZnO NWs. See ref. 35 for more details.

The UV transition and visible emission band for all the samples were measured using photoluminescence (PL) spectroscopy at room temperature. PL measurements were performed by pumping with the 325 nm (3.82 eV) line of a He–Cd laser chopped through an acousto-optic modulator at a frequency of 55 Hz and a fixed power of 1.5 mW. The PL signal was analyzed by a single grating monochromator, detected with a Hamamatsu visible photomultiplier, and recorded with a lock-in amplifier using the acousto-optic modulator frequency as a reference. It is important to note that all PL spectra reported here were measured under the same laser power, as the relative intensity of the visible to UV emission can change as a function of laser power.³⁶

The PL excitation (PLE) instrument was equipped with a 450 W Xe lamp and a double grating monochromator as excitation source. The bandpass was set to ≈300 nm. The excitation light intensity was monitored continuously before reaching the sample by a silicon photodiode. The sample was illuminated at normal incidence and the emitted light was collected in near backscattering at room temperature. After passing through a single monochromator with a 1200 grooves per mm grating, the light intensity was measured by a high sensitivity photomultiplier in photon counting mode. All the spectra have been corrected for the varying excitation light intensity, the dark signal of the detector, and the efficiency of the detector and grating.

3 Results and discussion

Scanning electron micrographs (SEM) images of the CBD, hydrothermal, and VLS samples are shown in Table 1 with the relevant structural data. The CBD and hydrothermal samples are structurally very similar. The clearest difference between the two samples is that the hydrothermal samples are taller on average. Both hydrothermal and CBD samples demonstrate fairly uniform coverage and are roughly perpendicular to the sample surface. On the other hand, the VLS samples are quite different structurally. The density of NRs in these samples is lower, while the average NR is longer. Furthermore, due to the nature of the fabrication process the NRs are not well-oriented with respect to the substrate.

Table 1 The SEM image for the CBD, hydrothermal, and VLS sample are shown with their corresponding structural parameters

	CBD
	Diameter: 80 ± 20 nm
	Height: 600 ± 50 nm
	Density: ≈100 NRs per μm^2
	Hydrothermal
	Diameter: 59 ± 44 nm
	Height: 1.8 ± 0.5 μm
	Density: ≈100 NRs per μm^2
	VLS
	Diameter: 90 ± 50 nm
	Height: 3 ± 1 μm
	Density: ≈25 NRs per μm^2

---

Fig. 1 plots the full PL spectra for the CBD, hydrothermal, and VLS samples. It is important to mention that it is difficult to compare the absolute PL intensity between samples, because the PL intensity is not corrected according to the number of luminescent centres. However, the relative intensity between the UV and visible PL emission bands for a particular fabrication method can be compared. All samples demonstrate a UV transition at ≈3.25 eV. The VLS sample demonstrates a large UV transition probability, where the ratio of the UV to visible intensity is I_UV/I_visible = 2412.55. We find the opposite behaviour in the hydrothermal sample with I_UV/I_visible = 0.11. The CBD samples demonstrate a moderate transition probability with respect to the UV and visible emission with I_UV/I_visible = 43.21. According to these results, the VLS sample has a very high optical quality, whereas, the hydrothermal samples are highly defective. Despite these differences in the PL spectrum between samples, each visible PL spectrum can be modelled with the same visible defect states.

Fig. 1 UV and visible PL spectra for the various growth methods as indicated in the Figure.

Previously,² we identified, in CBD grown ZnO NRs, four defect-related visible luminescence states as blue (B), green (G), orange (O), and red (R) states with respective energies of 2.52 eV, 2.23 eV, 2.03 eV, and 1.92 eV. R was found to only exist when the sample was annealed in O₂ at 600 °C, and was assumed to be related to O_i.² The B(V_Zn), , and O(V_O) states were found to be intrinsic NR defects, where we only measured a change in the relative peak intensity when the annealing condition was varied. In this work, we extend our visible defect model to variations in the fabrication method. In Fig. 2(a) we plot the normalized visible PL spectra for the samples fabricated by CBD, hydrothermal, and VLS. CBD and hydrothermal have a nearly identical visible spectra centred at ≈2.18 eV, while the VLS sample is respectively blue-shifted at ≈2.38 eV. All visible spectra were successfully fitted by three defect-related visible luminescence states. Fig. 2(b) through (d) plot the Gaussian fits for the B, G, and O states, which are coloured accordingly to emphasise their relative differences. The VLS sample has a larger B contribution compared to the other samples. Only the Gaussian peak energy was fixed to the corresponding B, G, and O energies, while all other parameters were allowed to vary during the fit. Importantly, the B, G, and O states demonstrate a full width half maximum (FWHM) of ≈0.5 eV, ≈0.5 eV, and ≈0.4 eV, respectively, for all of the samples. The B state in the hydrothermal sample does exhibit a FWHM of 0.9 eV, but this peak can be neglected without effecting the quality of the overall fit. This agreement in the FWHM demonstrates the quality of the fit and thus our choice in the B, G, and O states. Furthermore, the PL spectrum at low temperature was measured, whereby, we find the same defect-related visible luminescence states slightly red-shifted: B – 2.45 eV, G – 2.13 eV, O – 1.95 eV. The FWHM of the G, and O states decrease to 0.4 eV and 0.3 eV, respectively, while the contribution from the B peak is small, see Fig. S1.†

Fig. 2 (a) Comparison of the normalized visible PL spectra for the CBD, hydrothermal, and VLS samples. The Gaussian fit for the B, G, and O defect states are, respectively, coloured for (b) the CBD, (c) the hydrothermal, and (d) the VLS sample.

The differences in the visible PL spectra between the samples can be attributed to differences in the fabrication temperature. Hydrothermal samples (Fig. 2(c)) demonstrate the same dominant contribution from the G state, which thus contributes to the fact that the CBD and hydrothermal approaches are both low-temperature chemical-based methods. On the other hand, the VLS sample (Fig. 2(d)) demonstrates a relatively large B contribution. It has been shown that under high temperature annealing, H is removed from the structures thus increasing the radiative probability from the B state.^2,42,43 Furthermore, high temperature VLS fabrication produces a reduced O peak, which is believed to come from V_O in the bulk and may reflect the fact that these samples are highly crystalline.⁴⁴

We have fitted the B, G, O, and R states to the PL spectra of several other works.^15,37–39 The results of these fits are shown in Table 2, with the PL spectra fits given in Fig. S2 to S5.† We expect there to be differences between the sample fabrication methods, yet we find good agreement with our model. The FWHM values in Table 2 are consistent with our results above, while any variation can be attributed to differences in the size-distribution or sample quality. This confirms that, despite the fabrication method, visible luminescence in ZnO NRs is always composed of three defect-related visible luminescence states. There are some important conclusions that we can derive from the fits in Table 2. In the work of Ahn et al.,³⁷ sputtering in an Ar atmosphere yields a PL spectrum more blue-shifted compared with an O₂ atmosphere. A similar feature was seen in our CBD samples when annealed under similar conditions² and in Kushwaha and Aslam,¹⁵ Table 2. High temperature fabrication produces a blue signal as in the case of the thermally evaporated samples,³⁷ as reported above for VLS (Fig. 2(d)), and by Dhara and Giri²⁰ in Table 2. It is interesting to note that linear sweep voltammetry (LSV)³⁸ produces a similar result to high temperature fabrication while the sample was fabricated at 65 °C. Meanwhile, electroless deposition³⁹ produces no B states, and the sol–gel method¹⁵ is similar to the CBD and hydrothermal approaches²⁵ reported here and in Table 2. Therefore, we may draw the general conclusion that either using high-temperature fabrication or a (post-)treatment procedure in an inert environment actualizes samples with a higher concentration of B states. On the other hand O₂ treatment yields samples with a higher concentration of R states. This conclusion holds for un-doped samples, where doping creates a more diverse defect landscape that can result in a blue spectral shift with O₂ treatment, as shown in the results of Mamat et al.⁴⁰ in Table 2. It is important to note that in the case of thermal evaporation³⁷ and Zn-annealed hydrothermal NRs,²⁵ we find a slightly red-shifted B peak at ≈2.48 eV. This result indicates that high temperature processing in a reactive Zn environment produces Zn-related defect complexes that are slightly red-shifted, in agreement with Pal et al.⁴⁵ Finally, we briefly mention that the B–G–O states can also be applied to ZnO nanowalls,⁴¹ shown in Table 2 and Fig. S6.†

Table 2 Percentage area contribution for the blue (B), green (G), orange (O), and red (R) defect states for the references listed in the Table. The peak position gives the maximum of the visible spectrum. The FWHM for each peak is given in parenthesis after the percent area in units of eV

	Fabrication	Peak position/eV	B	G	O	R
Ahn et al.^37	Sputter: O2-atm	1.94	57% (0.95)	—	—	43% (0.48)
	Sputter: Ar-atm	2.49	63.5% (0.41)	36.5% (0.56)	—	—
	Thermal evaporation	2.47	(2.47 eV) 100% (0.39)	—	—	—
Atourki et al.^38	LSV	2.47	65.1% (0.41)	34.9% (0.47)	—	—
Florica et al.^39	Electroless deposition	2.15	—	80% (0.56)	20% (0.37)	—
Kushwaha et al.^15	Sol–gel (As-Prep)	2.20	23% (0.70)	36.6% (0.45)	40.4% (0.63)	—
	Sol–gel (O2 400 °C)	2.03	—	—	100% (0.53)	—
Dhara et al.^20	VLS (As-Prep)	2.33	36% (0.41)	64% (0.48)	—	—
	VLS (Ar 700 °C)	2.52	100% (0.64)	—	—	—
Čížek et al.^25	Hydrothermal (As-Prep)	2.25	33% (0.58)	63.2% (0.42)	3.8% (0.19)	—
	Hydrothermal (Zn 1000 °C)	2.47	(2.48 eV) 84.7% (0.28)	15.3% (0.37)	—	—
Mamat et al.^40	Al-doped sol–gel (As-Prep)	2.09	6.4% (0.25)	11.6% (0.31)	82% (0.52)	—
	Al-doped sol–gel (O2 500 °C)	2.23	39.7% (0.65)	36.7% (0.46)	—	23.6% (0.45)
Iwu et al.^41	Nanowalls	2.21	47.9% (0.83)	37.7% (0.38)	14.4% (0.40)	—

---

To gain further insight into the radiative mechanisms in our samples we performed PLE measurements, as reported in Fig. 3. The PLE spectra were normalized under the assumption that all defect states were populated equally for excitation energies above the E_G. Fig. 3 clearly demonstrates that the G and O states yield approximately zero PL intensity for excitation energies below 3.25 eV, while the B is optically active down to 2.8 eV. Typically, the B state is associated with what has been coined as the ‘green luminescence band’ (see the Introduction). The radiative transition of the 2.52 eV (‘green luminescence band’) is cited as occurring from the conduction band (CB) to a V_Zn acceptor, as depicted in Fig. 4(a). However, our PLE measurements (Fig. 3) indicate that the B state cannot be an acceptor, because this state is optically active below the gap excitation energies. Furthermore, in a previous study² we found that PL measurements at an excitation of 3.02 eV produce a dominant B signal in all of our samples. There is a small probability that optical excitation could occur from the VB to an acceptor and finally to the CB, and thus emission could be between the CB and the acceptor. However, as this event is a second-order process, whereby the joint density of states (JDOS) for this absorption process is low,⁴⁶ we believe it is not probable in our samples. Population of the CB via excitation of V_Zn was only observed in ionized donor level V_Zn.⁴⁷ In addition, second order absorption is only observed under high intensity optical excitation conditions,⁴⁸ which are not studied in this work. Furthermore, considering that the acceptor state is on the order of 0.2 to 0.9 eV above the VB,⁵ it is not likely to be thermally coupled with the VB. Therefore, the B state results from a transition from the defect level to the VB, and thus is not an acceptor state. Theoretically, it has been pointed out that the neutral V_Zn states are acceptors,⁵ while lies at a higher energy within the gap.^6,47,49 It is possible that our measured transition is from , which lies 2.52 eV above the VB (or above a shallow acceptor), in agreement with ref. 6 and 28, as depicted in Fig. 4(b). Therefore, if the B state is related to , then this can play a role in sensing under the same mechanism as the V_O.²³

Fig. 3 PLE spectra of the AsPrep CBD sample for the B, G, and O states.

Fig. 4 Schematic of the transition for the ‘green luminescence band’: (a) represents the model, reported previously in the literature, between the CB and the V_Zn acceptor state; (b) illustrates our model where the transition is between the state and the VB or possibly a shallow acceptor state. The grey region just below the CBM represents an Urbach tail of defect states, as previously measured in our samples.² The dotted arrows indicate the non-radiative transitions. The top of the grey region is at 3.25 eV, corresponding to the UV transition in our samples. The bottom of the grey region represents the extent of the Urbach tail, where only the B state can be populated.

4 Conclusion

In this work, we applied universally a model of visible defects based on , and O(V_O) states with energies at 2.52 eV, 2.23 eV, and 2.03 eV, respectively, across several fabrication methods for ZnO NRs. In a few special cases a R(O_i) state (1.92 eV) appears under O₂-rich fabrication conditions or post-growth treatment. We demonstrated that under high temperature fabrication one can increase the concentration of B states, while low temperature fabrication increases the G contribution. Post growth treatment in an inert environment increases the B component. In the case of Al-doped NRs, the post-growth treatment creates a contrary shift in the PL energy from un-doped NRs, while the initial defect states are the same. The formation of defect complexes manifests in a energetic shift from the initial defect energy. These results indicate how we can control the defects in ZnO NRs. PLE measurements demonstrate that the B state results from a transition between an and the VB. Therefore, B, cited as the ‘green luminescence band’, is likely to be important for sensing applications. Further investigation is needed to understand precisely how the B and G state work, separately or in tandem, at the surface in a nanosensor.

Author contribution

E. G. B. performed the data analysis and interpretation, and drafted the manuscript. V. S. prepared the CBD samples. G. F. performed the PL measurements. R. R. performed the PLE. A. S. D., G. P. V., and D. A. prepared VLS and hydrothermal samples. S. M. revised the manuscript. All authors read and approved the final manuscript.

References

1. Y. Kozuka, a. Tsukazaki and M. Kawasaki, Appl. Phys. Rev., 2014, 1, 011303.
2. E. G. Barbagiovanni, R. Reitano, G. Franzò, V. Strano, A. Terrasi and S. Mirabella, Nanoscale, 2016, 8, 995.
3. X. Zhang, J. Qin, Y. Xue, P. Yu, B. Zhang, L. Wang and R. Liu, Sci. Rep., 2014, 4, 4596.
4. A. Janotti and C. G. Van de Walle, Rep. Prog. Phys., 2009, 72, 126501.
5. A. Janotti and C. G. Van de Walle, Phys. Rev. B: Condens. Matter Mater. Phys., 2007, 76, 165202.
6. M. Willander, O. Nur, J. R. Sadaf, M. I. Qadir, S. Zaman, A. Zainelabdin, N. Bano and I. Hussain, Materials, 2010, 3, 2643.
7. M. Graetzel, R. A. J. Janssen, D. B. Mitzi and E. H. Sargent, Nature, 2012, 488, 304.
8. A. Klini, S. Pissadakis, R. N. Das, E. P. Giannelis, S. H. Anastasiadis and D. Anglos, J. Phys. Chem. C, 2015, 119, 623.
9. S. Anantachaisilp, S. M. Smith, C. Ton-That, T. Osotchan, A. R. Moon and M. R. Phillips, J. Phys. Chem. C, 2014, 118, 27150.
10. M. R. Alenezi, A. S. Alshammari, K. D. G. I. Jayawardena, M. J. Beliatis, S. J. Henley and S. R. P. Silva, J. Phys. Chem. C, 2013, 117, 17850.
11. A. Menzel, K. Subannajui, F. Güder, D. Moser, O. Paul and M. Zacharias, Adv. Funct. Mater., 2011, 21, 4342.
12. C. Soci, A. Zhang, B. Xiang, S. A. Dayeh, D. P. R. Aplin, J. Park, X. Y. Bao, Y. H. Lo and D. Wang, Nano Lett., 2007, 7, 1003.
13. L. Sang, M. Liao and M. Sumiya, Sensors, 2013, 13, 10482.
14. K. Liu, M. Sakurai and M. Aono, Sensors, 2010, 10, 8604.
15. A. Kushwaha and M. Aslam, J. Appl. Phys., 2012, 112, 054316.
16. F. Fabbri, M. Villani, A. Catellani, A. Calzolari, G. Cicero, D. Calestani, G. Calestani, A. Zappettini, B. Dierre, T. Sekiguchi and G. Salviati, Sci. Rep., 2014, 4, 5158.
17. P. K. Sharma, A. C. Pandey, G. Zolnierkiewicz, N. Guskos and C. Rudowicz, J. Appl. Phys., 2009, 106, 094314.
18. K. Bouzid, A. Djelloul, N. Bouzid and J. Bougdira, Phys. Status Solidi A, 2009, 206, 106.
19. Y. Gong, T. Andelman, G. F. Neumark, S. O’Brien and I. L. Kuskovsky, Nanoscale Res. Lett., 2007, 2, 297.
20. S. Dhara and P. Giri, Nanoscale Res. Lett., 2011, 6, 504.
21. B. Ha, H. Ham and C. J. Lee, J. Phys. Chem. C, 2008, 69, 2453.
22. S. Bayan and D. Mohanta, Philos. Mag., 2012, 92, 3909.
23. E. G. Barbagiovanni, V. Strano, G. Franzò, I. Crupi and S. Mirabella, Appl. Phys. Lett., 2015, 106, 093108.
24. R. Gurwitz, R. Cohen and I. Shalish, J. Appl. Phys., 2014, 115, 033701.
25. J. Čížek, J. Valenta, P. Hruška, O. Melikhova, I. Procházka, M. Novotný and J. Bulíř, Appl. Phys. Lett., 2015, 106, 251902.
26. K. Kodama and T. Uchino, J. Appl. Phys., 2012, 111, 093525.
27. C. Ton-That, L. Weston and M. R. Phillips, Phys. Rev. B: Condens. Matter Mater. Phys., 2012, 86, 115205.
28. K. H. Tam, C. K. Cheung, Y. H. Leung, A. B. Djurišić, C. C. Ling, C. D. Beling, S. Fung, W. M. Kwok, W. K. Chan, D. L. Phillips, L. Ding and W. K. Ge, J. Phys. Chem. B, 2006, 110, 20865.
29. J. Fan, Y. Hao, C. Munuera, M. García-Hernández, F. Güell, E. M. J. Johansson, G. Boschloo, A. Hagfeldt and A. Cabot, J. Phys. Chem. C, 2013, 117, 16349.
30. J. Chung, J. Lee and S. Lim, Phys. B, 2010, 405, 2593.
31. Y. H. Leung, A. B. Djurisić, Z. T. Liu, D. Li, M. H. Xie and W. K. Chan, J. Phys. Chem. Solids, 2008, 69, 353.
32. W. M. Kwok, A. B. DjurisÌŇicÌĄ, Y. H. Leung, D. Li, K. H. Tam, D. L. Phillips and W. K. Chan, Appl. Phys. Lett., 2006, 89, 183112.
33. L. Schmidt-Mende and J. L. MacManus-Driscoll, Mater. Today, 2007, 10, 40.
34. C. Opoku, A. S. Dahiya, C. Oshman, C. Daumont, F. Cayrel, G. Poulin-Vittrant, D. Alquier and N. Camara, Nanotechnology, 2015, 26, 355704.
35. A. S. Dahiya, C. Opoku, D. Alquier, G. Poulin-Vittrant, F. Cayrel, O. Graton, L.-P. T. H. Hue and N. Camara, Nanoscale Res. Lett., 2014, 9, 379.
36. D. Hou, T. Voss, C. Ronning, A. Menzel and M. Zacharias, J. Appl. Phys., 2014, 115, 233516.
37. C. H. Ahn, Y. Y. Kim, S. K. M. D. C. Kim and H. K. Cho, J. Appl. Phys., 2009, 105, 013502.
38. L. Atourki, E. H. Ihalane, H. Kirou, K. Bouabid, A. Elfanaoui, L. Laanab, X. Portier and A. Ihlal, Sol. Energy Mater. Sol. Cells, 2016, 148, 20.
39. C. Florica, N. Preda, M. Enculescu and I. Enculescu, Phys. Status Solidi RRL, 2014, 8, 648.
40. M. H. Mamat, Z. Khusaimi, M. M. Zahidi and M. R. Mahmood, Jpn. J. Appl. Phys., 2011, 50, 06GF05.
41. K. O. Iwu, V. Strano, A. Di Mauro, G. Impellizzeri and S. Mirabella, Cryst. Growth Des., 2015, 15, 4206.
42. K. Senthilkumar, M. Tokunaga, H. Okamoto, O. Senthilkumar and Y. Fujita, Appl. Phys. Lett., 2010, 97, 091907.
43. S. G. Koch, E. V. Lavrov and J. Weber, Phys. Rev. B: Condens. Matter Mater. Phys., 2014, 90, 205212.
44. P. Yang, H. Yan, S. Mao, R. Russo, J. Johnson, R. Saykally, N. Morris, J. Pham, R. He and H. Choi, Adv. Funct. Mater., 2002, 12, 323.
45. S. Pal, A. Sarkar, P. Kumar, D. Kanjilal, T. Rakshit, S. K. Ray and D. Jana, J. Lumin., 2016, 169, 326.
46. D. Mora-Fonz, J. Buckeridge, A. J. Logsdail, D. O. Scanlon, A. A. Sokol, S. M. Woodley and C. R. A. Catlow, J. Phys. Chem. C, 2015, 119, 11598.
47. M. K. Kavitha, K. B. Jinesh, R. Philip, P. Gopinath and H. John, Phys. Chem. Chem. Phys., 2014, 16, 25093.
48. R. Prasanth, L. K. Van Vugt, D. A. M. Vanmaekelbergh and H. C. Gerritsen, Appl. Phys. Lett., 2006, 88, 181501.
49. D. Das and P. Mondal, RSC Adv., 2014, 4, 35735.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra14453e

---