Single hydrogen atom manipulation for reversible deprotonation of water on a rutile TiO2 (110) surface

The discovery of hydrogen atoms on the TiO2 surface is crucial for many practical applications, including photocatalytic water splitting. Electronically activating interfacial hydrogen atoms on the TiO2 surface is a common way to control their reactivity. Modulating the potential landscape is another way, but dedicated studies for such an activation are limited. Here we show the single hydrogen atom manipulation, and on-surface facilitated water deprotonation process on a rutile TiO2 (110) surface using low temperature atomic force microscopy and Kelvin probe force spectroscopy. The configuration of the hydrogen atom is manipulated on this surface step by step using the local field. Furthermore, we quantify the force needed to relocate the hydrogen atom on this surface using force spectroscopy and density functional theory. Reliable control of hydrogen atoms provides a new mechanistic insight of the water molecules on a metal oxide surface.

T he hydrogen atom detection on a rutile TiO 2 surface is an important topic owing to its intriguing chemical and physical properties related to atomistic water splitting and hydroxyl production on this surface [1][2][3][4][5][6][7][8] . Detail understanding of this adsorbate and its control at atomistic level is essential for fully elucidating the nature of a deprotonated water reaction on a TiO 2 surface  . Moreover, oxidation of hydrogen atoms by oxygen molecule on the rutile TiO 2 surface results in the reactive oxygen species (ROS), such as reaction intermediates of water species H 2 O, HO 2 , H 2 O 2 , and H 3 O 2 24,25 . Exploring these species requires the fundamental understanding and detailed description of the interaction between oxygen and hydrogen atoms. Ideally, attacking these problems require to access the atomic scale study of a hydrogen atom on this surface, which remains a great challenge owing to the light mass and small size of the hydrogen atom. In particular, hydrogen species on a rutile TiO 2 surface were investigated using various experimental techniques, including scanning tunneling microscopy (STM) [1][2][3][4][5][6][10][11][12][13][14][15][16][17][18][19][20][21][22]25,26 . STM provides a unique opportunity for electronically inducing the reactions of a hydrogen atom on this surfaces 3,10,18,21 . However, STM easily induces the stochastic behavior of molecules owing to its flowing current; therefore, it might be difficult to precisely control in the desired configuration at atomic level 3,18 . Moreover, the contrast mechanism of STM is related to the density of electronic states, which is still obscure to investigate the real-space of the atomic configuration 1,12,26 .
Atomic force microscopy (AFM), as a viable alternative, has been used to provide a precise measurement of the surface configuration, also manipulating atoms and molecules based on its force modulation mechanism [27][28][29][30][31][32][33][34][35][36][37][38][39][40][41][42][43][44] . Another fascinating capability of AFM is the possibility to measure directly the interaction forces that induce the adsorbate's motion and thus, to provide a detailed insight into the interaction force for the target species [35][36][37][38][39][40][41][42][43] . Kelvin probe force spectroscopy (KPFS), owing to its force modulation mechanism, allows us to precisely control the different states of on-surface species signaled by the appearance of a jump in the frequency shift (Δf) vs. bias voltage (V bias ) parabola 9,27,28,[31][32][33] . The vertical shift between the two parabolas is strongly influenced by the different local electric fields automatically formed between the tunneling junction 9,27,28,[31][32][33] . Especially on the metallic surface, controllable on-surface lateral manipulation of the single molecule [45][46][47] has been well established and the adsorbate can also be activated 48 by means of electric field 49,50 . Moreover, on the rutile TiO 2 surface, the vertical desorption of hydrogen atom 9,10,16,18,21 , reversible migration of hydrogen atom 23 , the stochastic motion of hydrogen atom induced by inelastic tunneling electrons 3 , manipulation of oxygen adatom 27 and lateral tip induced excitation of the single molecule 35 have been clarified. However, the lateral manipulation of the single hydrogen atom to the desired position on rutile TiO 2 surface has been poorly reported experimentally up to now 10 , although it could provide an efficient means of the arrangement of the on-surface water splitting reactions, dissociation dynamics, which critically control the efficiency of heterogeneous catalytic reactions on this surface  . Especially, elucidating the interaction force between tip and water configuration on the TiO 2 surface is important for answering the current controversial question of whether a proton atom should remain as one water molecule or two distinct hydroxyls from the mechanical point of view [1][2][3][4][5][6][7][8] . For local chemistry, the mechanical sensitivity and stability of hydrogen bond [40][41][42][43] are also crucial properties, but studies of these properties remain elusive on this surface.
Here, we use AFM and KPFS to present the manipulation of hydrogen atom and reversible water reaction on a rutile TiO 2 (110)−(1 × 1) surface, and investigate its mechanical properties by the force field mapping. We analysis the manipulation outcome with the local electric field and the density functional theory (DFT). Our results demonstrate that the hydrogen atom can be manipulated along the oxygen row. The force field on the top of these configurations quantifies the possible tilting of the hydrogen atom on these configurations.

Results and discussion
Control of reversible deprotonation of water on a rutile TiO 2 (110) surface. First, we show an experiment for the reversible water reaction on the TiO 2 surface. By using the KPFS technique, we experimentally demonstrated the water reaction as shown in Fig. 1 (a-i). A Rutile TiO 2 (110)−(1 × 1) surface contains twofoldcoordinated surface bridging oxygen (O s ) atoms and fivefoldcoordinated titanium (Ti) atoms that are alternatively aligned 1 . Moreover, the practical sample preparation in ultrahigh vacuum (UHV) will induce point defects, such as oxygen vacancies (O V ) and hydroxyl defects (O s H) 1 . When the TiO 2 surface is exposed to oxygen at room temperature, oxygen will dissociate on this surface and will be adsorbed as an adatom (O ad ) on Ti row 28,29 . We previously discovered that the O ad has two stable charge states, namely, singly charged (O ad − ) and doubly charged (O ad 2− ). Moreover, we previously found that O ad 2− is the most stable on the rutile TiO 2 surface 28,29 . Figure 1 The characteristics of a net positively charged hydrogen atom were previously well accepted by various experimental techniques 26,30,34,44 . Two black spots that correspond to bistable O s H defects can be observed in Fig. 1(a) 9,30,34,44 . Notably, the oxygen adatom is spontaneously charged to O ad 2− and adsorbed between two O s H defects 28,29 . After AFM imaging ( Fig. 1(a)), the tip was moved on top of the O s H defect and the bias was ramped from zero to a certain negative voltage and then back to zero ( Fig. 1 (f)). Figure 1(b) shows the AFM image of the same scan area at 0 V obtained immediately after the bias voltage back to zero, showing that the black spot corresponding to the O s H defect disappeared and O ad 2− became less bright. The characterization of this species is based on the movement of the target hydrogen atom toward O ad 2− , inducing the formation of (O ad H) − , while the symmetric configuration is initiated by positioning the tip on top of the hydrogen atom and ramping the bias from zero to a certain negative voltage and then back to zero ( Fig. 1(b) and (g)). Notably, this manipulation cycle is based on the double hopping of the hydrogen atom along the ½1 10 direction as shown in Fig. 1(j). Hence, the manipulation cycle can be performed repeatedly as shown in the order of Fig. 1 One has to keep in mind that for a non-zero local contact potential difference (LCPD) between tip and sample, even at zero bias voltage there is a finite electric field across the junction [31][32][33] . Moreover, this LCPD also depends on the tipsample distance owing to the average effect 31 . Only at compensated LCPD, that is, at V bias = V LCPD the local electric field vanishes. Although this is very well known from KPFS, it is usually less considered in pure STM experiments. The quantitative assignment of V jump is further shown in Fig. 2(c) with more tip-sample distances including V LCPD , where V jump − V LCPD linearly becomes large by reducing the tip height. Moreover, during the negative bias KPFS, the current was in the noise limit of our current amplifier (see Supplementary Fig. S2). Additionally, the repulsive interaction does not induce manipulation (see Supplementary Fig. S3). These observation leads us to conclude that the dominant role of the manipulation is the local electric field. Based on Fig. 2(c), the local electric field for manipulating hydrogen was estimated using the following equation where, V jump is the bias voltage where the frequency shift jump occurs, V LCPD is the local contact potential difference of the upper parabola in each KPFS [31][32][33] and z is the relative tip-sample distance. Here, we note that the V LCPD is obtained by fitting the parabolic function to the upper parabola of each KPFS obtained at different tip height. The observed tunneling current with the positive bias KPFS is contrary to that of negligible tunneling with negative bias KPFS (see Supplementary Fig. S2), thus the dominant contribution to the hydrogen desorption may be induced by the tunneling current 3,10 . Hence, suggesting that the local electric field has a dominant role in the lateral manipulation of the hydrogen atom on this surface.
Manipulation mechanism of hydrogen atoms on a rutile TiO 2 (110) surface. The reaction among three atomic structures can be commonly described by a multiple umbrella potential well, as shown in Fig. 4(a) (Fig. 4(a) and (b)), the tip was placed on top of the O s H. As is discussed in Fig. 1    The real potential energy for describing the manipulation is much more complicated, and many factors may need to be considered, such as the local adsorbate charge, subsurface effect and the precise shape of the tip. The tip position is indicated by a blue circle in the AFM images.
( Fig. 4(a)→4(c)). However, the sufficient tunneling current leads to the desorption of the hydrogen atom, not the lateral hopping of hydrogen atom on this surface as again shown in Fig. 3(a)-(d). This desorption mechanism might be related to the vibrational excitation by inelastic tunneling, which is previously reported by Acharya et al. 10 . Consider now applying a negative bias to the oxygen atom corresponding to the tip positioned at the center of the O s H−(O ad H) − −O s (Fig. 4(a) →4(d)). It is seen after the bias was swiped in the negative direction, that one of the hydrogen atoms directly underneath the tip moved toward the oxygen row, result in O s H−O ad 2− −O s H ( Fig. 1(d) → 1(i) → 1(e)). As shown in Fig. 4 Table S1).  on the right (red arrow) has the same position as the oxygen row (the local maxima position of Z in the oxygen row is generally the same as the local minima position of the dip). However, the hydrogen atom on the left (black arrow) is slightly tilted to the left from the center of O ad 2− . The distances from the center of O ad 2− to these dips were 320 pm and 116 pm, respectively. Notably, the distance of 320 pm is the half distance between the oxygen row 1 , which indicates the reliable lateral resolution of the contrast in the AFM images Fig. 5(g-i). From these observations, we expect that the hydrogen atom on left will strongly be perturbed from the top and orient to the left owing to the creation of a hydrogen bond with the nearest neighbor oxygen row (as shown by the blue dotted arrow in Fig. 5(j)). On the other hand, the hydrogen atom on the right has similar interactions from (O ad H) − and the nearest neighbor oxygen row; therefore, a too small tilt was observed (limited by the convolution of the tip apex shape and sample structure). This assertion is also supported by the similar height of (O ad H) − and the nearest neighbor oxygen row. The equivalent net charge of (O ad H) − and the nearest neighbor oxygen row result in a similar topographical height owing to the contrast mechanism of the tip in hole mode 9,30,34 . Moreover, this off-center orientation of the hydrogen bond in Fig. 5(j) are nicely demonstrated in our DFT calculation shown in Supplementary Figure S4(d), and also the theoretical prediction in Tan et al. 3 and Du et al. 5 studies. Hence, give us extra confidence to conclude that the hydrogen atom on the right has similar interactions from (O ad H) − and the nearest neighbor oxygen row. Here, we should note that one would expect a small influence of the possible tilting of the hydrogen atom under the attractive force interaction. Because, for example, the AFM imaging shows different bond lengths and atom positions in C 60 when changing the Z distance 39 . Therefore, we perform the frequency shift mapping as shown in Fig. 5(a-c). The overall values of Δf(z) decrease by reducing the tip height. To allow easy visualization, we draw a constant frequency shift value of −145 Hz and −345 Hz by the circle in Fig. 5(a-c), respectively. Importantly, the value of −145 Hz is the conventional value for stable AFM imaging during the experiment. Focusing on the constant frequency shift value of −145 Hz in Fig. 5(b, c), we also found a similar characteristic of the off-center orientation of the hydrogen bond as shown in Fig. 5 (j). The two hydrogen atoms identified in Fig. 5(j) were identically indicated by the black and red solid arrows in Fig. 5(b, c). This finding is in perfect agreement with the contrast mechanism of the tip in the hole mode, which a positively charged tip gives rise to additional attractive interaction with the negatively charged oxygen atom on the surface, causing a larger negative frequency shift, and the oxygen atom appears bright 9,30,34 . On the other hand, the position of the hydrogen atom is "less negative" than the oxygen atom, which gives them a slight dark contrast. Remarkably, when the tip approaches very close to the surface, the local maximum of Δf(z) appears on top of the hydrogen atom, which is indicated by the black dotted arrow of the constant frequency of −345 Hz in Fig. 5(b, c). This finding also perfectly agree with the contrast inversion expected in the small tip-sample distance, which, the strong attractive force field induce a displacement of the hydrogen atom, gives rise to the screening of the underlying oxygen atoms, and the overall interaction results in a convolution between the positive tip apex, positive hydrogen, and the negative oxygen atom 30,34 as shown in Fig. 5(k, l). The displaced hydrogen atom tends to relocate at the center of the oxygen atom, which is nicely demonstrated by the local minima of the Δf(z) mapping in Fig. 5(b, c) as indicated by the purple dotted arrow. Therefore, the AFM contrast obtained at −145 Hz has a small influence on the possible tilting of the hydrogen atom under the AFM imaging. Figure. 5(d-f) show the F(z) mapping obtained from Fig. 5(a-c). The overall values of F(z) decrease by reducing the tip height, owing to the increasement of the attractive force. In Fig. 5(m) and (n), the component of the force curve between tip and sample at different locations are further evaluated from the short range force. The positions of the curves are indicated by the solid arrows, and the dotted lines inside Fig. 5 (e) and (j) by different colors. The short range force commonly offers an atomic resolution in AFM. On the other hand, longrange force offers background force acting on the tip at a relatively far distance from the surface, such as long-range van der Waals force and long-range electrostatic force 58,59 . The longrange dominant region of the Δf (z) curves, which we defined as z > 0.3 nm, was fitted into the inverse-power function of z −s 58,59 . The short range part of Δf (z) was obtained by subtracting the long-range part of Δf (z) from the raw Δf (z) curve. Then, the short range part of Δf (z) curve was numerically converted to F SR (z) using the Sader Jarvis method [57][58][59] . In Fig. 5(m), the three curves of F (z) generally decrease by reducing tip-sample distance, which indicates the attractive tip-sample interaction. Especially in Fig. 5(n), we found that F SR (z) has a similar tendency with F (z). Hence, the atomic contrast of Fig. 5(g-i) and F (x, z) in Fig. 5(d-f) are dominantly governed by the short range forces. When the contrast inversion occur in the F SR , the attractive short range force acting on the tip was about F SR = −0.6 nN, indicated by the pink arrow in Fig. 5(n). Suppose that the hydrogen relocation occur nearly at the tip-sample distance where the contrast inversion occur in the F SR , the results shown in Fig. 5(n) indicate that the hydrogen bond can be stabilized in this configuration without rearrangement, at the range of −0.6 nN < F SR . In the DFT calculation shown in Supplementary Fig. S5, the force required for the rearrangement of O ad H from tilted geometry to upright is estimated to be around 0.440 nN, which is smaller than experimentally measured 0.6 nN. This difference additionally propose that the rearrangement of the hydrogen atom is presumably induced by the attractive force of tip background short range van der Waals interaction or tip dipole, dominates the tip-sample interaction [60][61][62] . Compared to the other works, the magnitude of F SR = −0.6 nN is quantitatively larger than the lateral or vertical force for displacing a physisorbed CO molecule on metal surface 38 . This is generally in line with the physical aspect of the hydrogen-oxygen atom interaction that the hydrogen bond is generally stronger than the van der Waals interaction 40 H defect is also the most fundamental atomic feature on this surface 1 and known to provide a critical role in photocatalysis at tremendous condition  . Figure 6(a) shows an atomically resolved AFM image of a rutile TiO 2 (110) surface partially exposed to oxygen at room temperature obtained using a hole mode tip 9,30,34 . The black spot can be observed in Fig. 6(a) that corresponds to the O s H defect 9,30,34 . Notably, an oxygen adatom can be observed on the lower left side as a bright spot. Next, we demonstrate that this O s H defect can be precisely manipulated along the [001] direction by KPFS. After AFM imaging ( Fig. 6(a)), the tip was brought slightly to the upper side of the O s H defect, and the bias was ramped from zero to a certain negative voltage and then back to zero (Figure 6(i)). Fig. 6(b) shows the AFM image of the same scan area obtained at 0 V immediately after the bias back to zero, showing that the black spots that correspond to O s H defect moved one-lattice distance toward the [001] direction from its initial position. As shown in Fig. 6(j), Li et al. 14 has reported the H diffusion along the [001] direction with an energy barrier of 1.29 eV. This sufficient large energy barrier is supposed to prevent spontaneous diffusion under the 78 K. Hence, inducing the local electric field near to the hydrogen atom deforms the potential energy landscape, and result in the lateral hopping of the hydrogen along the [001] direction, as shown by the black dotted line in Fig. 6(j). Interestingly, in this process, we also found that the bias voltage of about V bias ≦ −3.0 V is required to hop the hydrogen along the [001] direction ( Fig. 6(i)). This finding nicely agrees with the theoretical aspects that the larger bias voltage is required to overcome such a barrier,~1.29 eV 14 . Notably, we previously found that the positive bias will easily induce the desorption of the hydrogen atom on this surface 9 . Figure 6(c) and (d) show the reproducibility of this manipulation, which indicates that the hydrogen atom can be fully manipulated along the [001] direction without the atom being desorbed. The quantitative assignment of these manipulations can be performed by measuring the line profile on top of the hydrogen atom as shown in Fig. 6(e-h). We found that the displacement of each hydrogen atom is 0.3 nm, which is in perfect agreement with the nearest neighbor distance between the two oxygen atoms in the oxygen row 1 . Therefore, the hydrogen atom can be precisely controlled at the single atomic level using negative bias KPFS.
In summary, we have demonstrated the lateral manipulation of hydrogen atom on the rutile TiO 2 (110) surface by low-temperature AFM and KPFS. We succeeded in the reliable control and characterization of a hydrogen atom on top of the three different outcomes of O s H−O ad 2− −O s H species by using a functionalized tip in hole mode with the KPFS manipulation. The force mapping with an atomic resolution allowing us to preciously determine the hydrogen position; interestingly, one hydrogen atom was tilted forward and another was straight. We believe that the achievement of our large body of work intrinsically provides the opportunity to understand the mechanochemical process of reactive oxygen species, the hydrogen atom and water species, naturally the world's most important chemical species, on an oxide surface.

Methods
Experimental details. The experiments were carried out using a low-temperature ultrahigh vacuum AFM system. The deflection of the cantilever was measured using the optical beam deflection method. The base pressure was lower than 5.0 × 10 −11 Torr. The temperature of the AFM unit was kept at liquid nitrogen temperature (78 K). The AFM measurements were performed in the frequency modulation (FM) detection mode. The atom tracking method was used to compensate for the thermal drift between the tip and surface during the measurements. The dc bias voltage was applied to the sample. The AFM imaging was performed using constant Δf mode at V bias = 0 V to avoid the tunneling current to flow. The cantilever was oscillated at resonance frequency keeping the oscillation amplitude constant. We used iridium (Ir)-coated Si cantilever (Nanosensors SD-T10L100, f 0 = 800 kHz, A = 500 pm, k = 1500 N/m). Metal Ir tips provide stable AFM imaging compared to the bare Si tip. The tip was initially annealed to 600 K and then cleaned by Ar + sputtering to remove the contamination before experiments. The rutile TiO 2 (110)-(1 × 1) sample was prepared by sputtering and annealing to 900 K in several cycles. The sample was exposed to oxygen at room temperature for~0.5 L and then transferred to the measurement chamber precooled to 78 K. The O s H groups on the TiO 2 (110) surface were spontaneously created from the dissociation of water molecules (from a background residual vacuum) over oxygen vacancy sites by transferring hydrogen atoms to neighboring oxygen bridge sites 1 . We modify the tip apex by gently poking the tip into the surface using the controlled force distance spectroscopy, until reaching to the sharp tip in hole mode (positively terminated tip). Notably, we can distinguish O s H from the oxygen vacancies using the previous results 9 .
Calculation method. The DFT calculations were carried out using the CP2K Quickstep package 65 and the PBE0-TC-LRC-ADMM hybrid density functional 66,67 , containing 20% HFX exchange, which was truncated at a distance of 2.5 Å. The cutoff of the finest real-space integration grid is 400 Ry. The primary basis set is MOLOPT 68 , of DZVP quality for Ti and TZV2P quality for O, with corresponding GTH pseudopotentials 69,70 . The auxiliary Gaussian basis for the ADMM method was cFIT11 for Ti and cFIT3 for O. The dispersion interactions were considered within the Grimme D3 method 71 . The force convergence criterion used for geometry relaxations was 0.01 eV/Å. The TiO 2 (110) surface was constructed using slab model consisting of six atomic layers and cell parameters from bulk calculation. The external electric field was simulated by introducing a saw-shaped periodic electrostatic potential along z-axis. In the evaluation of surface geometry under external load, external force was applied on the H atom along z-axis.