Role of point defects on the reactivity of reconstructed anatase titanium dioxide (001) surface

The chemical reactivity of different surfaces of titanium dioxide (TiO2) has been the subject of extensive studies in recent decades. The anatase TiO2(001) and its (1 × 4) reconstructed surfaces were theoretically considered to be the most reactive and have been heavily pursued by synthetic chemists. However, the lack of direct experimental verification or determination of the active sites on these surfaces has caused controversy and debate. Here we report a systematic study on an anatase TiO2(001)-(1 × 4) surface by means of microscopic and spectroscopic techniques in combination with first-principles calculations. Two types of intrinsic point defects are identified, among which only the Ti3+ defect site on the reduced surface demonstrates considerable chemical activity. The perfect surface itself can be fully oxidized, but shows no obvious activity. Our findings suggest that the reactivity of the anatase TiO2(001) surface should depend on its reduction status, similar to that of rutile TiO2 surfaces.

Images acquired at room temperature and in situ at 773 K, respectively for the as-grown sample. c, d, Images acquired in situ at 500 K and 773 K for the Ar + sputtered sample. e, f, g, Images of the sample: as-grown, after electron bombardment (150 eV, emission current of 1.1 mA, 10 min), after annealing at 873 K for 20 min, respectively. h, i, Images within the same area for the sample after electron bombardment and followed 873 K annealing, acquired at 1.5 V and 10 pA and at 1.0 V and 200 pA, respectively for h and i. All of the annealing treatments were performed under pressure better than 5.5×10 −9 Pa. Scale bars: 10 nm for a-g, and 2 nm for h and i.  K, and g-i: 0.6 V and 5 pA, at room temperature. Scale bars: 10 nm for images a-f and for images g-i, 2 nm for images j-l.  One may concern that the observed dark defects at the ridges could be due to the that the Sr contamination due to the Sr out-diffusion from the substrate should not be the main contribution to the dark spots, nor to the bright spots. Therefore, we believe that both of the dark and bright spots are most likely to be intrinsic point defects, rather than the Sr contamination because of no Sr signal on the surface when the annealing or the growing temperature is lower than 973 K.

Supplementary Note 2: Comparing the samples by Ar + sputtering and by electron bombardment
Supplementary Figure S3a and b show the images of an as-grow sample acquire at room temperature and an annealing sample acquired in situ at 773 K, respectively.
The annealing sample was maintained at 773 K for 12 h. It suggests that it does not cause bright spots for an as-grown sample by annealing treatment at 773 K under UHV condition. For an Ar + sputtered sample, the reconstructed (1×4) terraces were destroyed. At annealing temperatures lower than 500 K, the images just show the disordered surface structure (Supplementary Figure S3c). After the sample was 11 annealed at 873 K, the bright spots already appeared. Supplementary Figure S3d shows the image acquired at 773 K in situ, where the sample had been annealed at 873 K for 8 h. The image gives the feature similar to the one obtained at room temperature ( Fig. 1b in the main text) We also used the electron bombardment to treat the sample surface.
Supplementary Figure S3e    conditions of 0.6 V and 5 pA, as mild as it could be. However, we still did not observe any adsorption features as the ones we observed at low temperature (Fig. 4 a-c in the main text). Our observations strongly suggest that the water molecules do not tend to adsorb at room temperature.

Supplementary Methods
Details of structural models.    Supplementary   Table S1. The simulated images are all in good agreement with our experimental observations ( Fig. 6 in main text).

Calculated properties.
Based on the oxidized structure, we have calculated the adsorption behavior of tip has not been well described in our calculations, possibly due to some uncertainty due to the invovlement of quite complicated tunneling electron assisted processes.
We have also simulated the images of the structures obtained from different manipulation processes. The structural model and the corresponding images given in Supplementary Figure S8i and j can well describe the processes II and III mentioned in the main text. One can see that the simulated image well resembles the side-positioned paired spot in the process II and the shoulder feature in the process III.
All our theoretical results provide strong supports for the conclusions drawn in the main text.