Understanding and removing surface states limiting charge transport in TiO2 nanowire arrays for enhanced optoelectronic device performance

An effective wet-chemistry approach is demonstrated to minimize the trap states that limit electron transport in rutile TiO2 nanowire arrays, this leads to an over 20-fold enhancement in the electron diffusion coefficient.


Introduction
Nanoscale semiconducting metal oxides have become promising low-cost electrode materials for solar cells, solar fuels and electric energy storage applications. [1][2][3][4][5] Charge transport within electrode materials is a major determinant of device performance. It has been generally accepted that electrons undergo a random walk through electrode networks and are impeded mainly by surface trap states, grain boundaries and structural disorder. [6][7][8][9] Compared to randomly packed rutile NP lms, ordered single-crystal (grain boundaries free) TiO 2 nanowire (NW) arrays are generally expected to have higher electron mobility, and have been the subject of extensive research. [10][11][12][13][14] Unfortunately, measurements have shown that their electron mobility is not superior to that in NP lms with the same phase ( Fig. S1 ESI †). 15 This implies that the inuence of material architectures on electron transport is less evident in the presence of a large density of surface trap states. 8,9 Thus, it is critical to understand the nature of the surface trap states and minimize them in order to exert the expected high electron mobility and device performance of NW arrays. In this study we reveal and demonstrate an effective way to remove the trap states that limit the electron transport in single-crystal rutile TiO 2 NW arrays and their device performance.

Results and discussion
Aligned singe-crystal rutile TiO 2 NW arrays were prepared via a conventional hydrothermal method. [10][11][12][13][14] As shown in Fig. 1a, the as-prepared TiO 2 NWs grow almost vertically from the substrate with an average diameter and length of about 100 nm and 3 mm, respectively. According to the high-resolution transmission electron microscopy (HR-TEM) image (Fig. 1b) and the selected area electron diffraction (SAED) pattern (the inset of   1b) analysis, the NW is highly crystallized and grows along the [001] direction with side surfaces of {110} crystal plane. Surface treatment is commonly used to boost the performance of semiconductors. [16][17][18] In this report, the as-prepared TiO 2 NW arrays were then processed with a wet-chemistry treatment by immersing in a H 2 O 2 -NH 3 (aq) (10 : 1 v/v) solution at room temperature for different times, then rinsing with a copious amount of distilled water. Subsequently, both the treated and untreated NWs were annealed at 723 K for 30 min in an oxygenrich environment.
Electron transport in these NWs was probed by using intensity modulated photocurrent spectroscopy (IMPS) and the results are shown in Fig. 2. The values of the electron diffusion coefficient (D) and photoelectron densities (n) are determined from the transport time constants (s c ) and lm thickness (d) using procedures described elsewhere. 19 Compared to the NWs without treatment, the D value in treated NWs was enhanced by over 20 times, for example, at a given n of 1 Â 10 17 cm À3 . According to the previous study, the transient photocurrent response revealed from IMPS measurement is dominated by electron transport within electrode materials and can be explained by a trap-assisted diffusion model. The relation between D and the total trap density (N T ) can be described using the following equation: 19 where C 1 is a constant, and a is related to the shape of the distribution of the sub-bandgap trap states. Since 1/a is larger than 1, the existence of traps with a distribution of various energies is detrimental to the transport. Best ts to the data shows that a ¼ 0.37 and 0.26 for NWs with and without treatment, respectively, suggesting that the shapes of the distribution of sub-bandgap trap states are different. A smaller a value indicates a longer tail and a relatively deeper distribution in trap energy levels, which is normally associated with the existence of deeper level traps that not only affect the magnitude but also the slope of the mobility-light intensity curve. 6,20,21 Fig. S2 † shows the dependence of n on voltage for the two cells based on NWs with and without surface treatment. The n value of untreated NWs is about 1.5-fold higher than that of treated ones at the same voltage, suggesting the total trap density is larger for untreated NW samples. Based on the analysis described above, one can conclude that (1) the slower electron transport in the untreated rutile TiO 2 NWs is attributable to surface trap states with relatively deeper energy levels; (2) the treatment can passivate/remove the surface traps especially those with relatively deeper energy levels, leading to shallower distribution trapping energy (larger a value) and faster electron transport for surface-treated NW samples.
Surface trap states are commonly associated with surface defects. N-type rutile TiO 2 is usually in a nonstoichiometric reduced form that has intrinsic defects including oxygen vacancy and titanium interstitial (Ti int 3+ ). 22,23 Oxygen vacancies are thermodynamically unstable and can be readily eliminated via simple oxygen annealing at elevated temperature. 24,25 Regarding the Ti int 3+ , it prefers to have a high coordination and is mainly located in the bulk. However, recent studies on rutile TiO 2 (110) surface using atomic-resolution scanning tunneling microscopy have shown that Ti int 3+ defects can diffuse from the bulk to the surface at temperature higher than 400 K and result in surface reconstruction. [25][26][27][28] Photoluminescence (PL) is a highly sensitive technique for investigating surface characteristics of semiconductors. As shown in Fig. 3 (black line), a strong near-infrared (NIR) PL peak centered at around 835 nm was observed from untreated NWs. Such a PL peak (835 nm) of  rutile TiO 2 has been indexed to surface Ti interstitials, specically to the Ti int 4+ . [29][30][31] Since the rutile TiO 2 NWs were prepared at a temperature of 453 K, and that their side surfaces are {110} crystal plane, Ti int 3+ can diffuse to the {110} crystal plane. When the NWs were further annealed in an oxygen-rich environment, these outward-diffused Ti interstitials will be oxidized and form added islands that occupy preferentially at the surface interstitial sites, which results in the strong NIR PL spectrum. Such interstitial defects will likely lead to the formation of trap states in relatively deeper energy levels that extend virtually all the way to the conduction band edge. It is worth noting that, unlike oxygen vacancies, surface Ti interstitial defects cannot be easily removed via conventional oxygen annealing treatment; in contrast, oxygen annealing will promote their outward-diffusion and surface reconstruction process. Fig. 3 (red line) shows the PL spectrum of TiO 2 NWs aer 10 min wet-chemistry treatment in H 2 O 2 -NH 3 (aq) solution followed by 30 min oxygen annealing at 725 K. The peak intensity is signicantly reduced, implying that the vast majority of surface Ti interstitial defects were removed aer the treatment. Based on the differences of PL intensity and electron diffusion coefficient of NW arrays with and without treatment, it can be concluded that the surface Ti interstitial defects related trap states limit the electron transport of ordered single-crystal TiO 2 NW arrays.
During In the meanwhile, NH 3 (aq) will neutralize the H + and promote the forward reaction. The process of removing surface interstitials is rather fast as conrmed via PL studies. Fig. S3 † shows the PL spectra of TiO 2 NWs as a function of treatment time. The PL peak intensity was signicantly lowered with only 30 s treatment. Upon the treatment, a Ti interstitial depletion layer near the surface will be formed, and consequently, their related trap states can be minimized. Crystal structures analysis including X-ray photoemission spectroscopy (XPS) (Fig. S4 †), X-ray diffraction patterns (XRD) (Fig. S5 †) and Raman spectra (Fig. S6 †) of NW arrays with and without treatment remain unchanged, which conrms that no new phase or impurity was introduced during the treatment. Thus, the wetchemistry treatment that is presented here is an effective approach to remove surface trap states in deeper energy levels, as exemplied by the observed over 20-fold enhancement in electron transport.
Fast electron transport is expected to lead to high electron collection efficiency and enhanced optoelectronic device performance. Fig. 4a compares the recombination times (s r ) as a function of photoelectron density n of solar cells based on the two samples. Base on the data in Fig. 2 and 4a, we determined that the electron diffusion length (L n ) for treated NW is approximately 20 mm, 4 times longer than that for NWs without treatment ($5 mm) using the relation of L n ¼ (Ds r ) 1/2 . The charge collection efficiency (h cc ) of treated NW-based cells, as described by the relation: h cc ¼ 1/[1+(d/L n ) 2 ], where d is the electrode thickness, can be calculated to be 98%, which is 23% larger than that of untreated NW counterpart. The cell based on NWs with faster electron transport rate exhibits a short-circuit photocurrent density (J sc ) of 6.01 mA cm À2 , 25% higher than that obtained on untreated NW-based cell (4.81 mA cm À2 ). In general, J sc is determined by light-harvesting efficiency (h lh ), charge-injection efficiency (h inj ), and h cc . It worth noting that the surface areas of NW arrays with and without treatment are almost identical according to dye desorption results (Fig. S7 †), suggesting that both h lh and h inj values of these devices are similar. Thus, the 25% improvement in J sc of the treated NW-based cells should be mainly ascribed to the enhancement in h cc of the electrode materials as described above. Besides, the treated NW-based cell exhibits a higher ll factor (FF), which can also be ascribed to the much enhanced electron transport property. Under similar surface areas, the faster electron transport leads to higher values of J sc and FF, and 62% enhancement in solar to electricity conversion efficiency. In the future, if the growth technique could be extended to very long 1D ordered nanostructures then one would have a superior electrode material for various kinds of sensitized and heterojunction solar cells.

Conclusions
In conclusion, we have revealed that the surface trap states of rutile TiO 2 NWs in relatively deeper energy levels result in slower-than-expected electron transport. Moreover, we have demonstrated an effective wet-chemistry approach to remove these trap states, leading to over 20-fold enhancement in electron transport and 62% improvement in solar conversion efficiency. Considering the wide range of studies and applications of 1D single crystal TiO 2 nanowire arrays, the signicant charge transport enhancement achieved in this work will make it a real enabling technology for future solar cells, water splitting and electric energy storage device applications.