Effect of the Hydrogen Bond in Photoinduced Water Dissociation: A

Jan 26, 2016 - Photoinduced water dissociation on rutile-TiO2 was investigated using various methods. Experimental results reveal that the water disso...
8 downloads 11 Views 3MB Size
Download PDF
Letter pubs.acs.org/JPCL

Effect of the Hydrogen Bond in Photoinduced Water Dissociation: A Double-Edged Sword Wenshao Yang,† Dong Wei,† Xianchi Jin,† Chenbiao Xu,† Zhenhua Geng, Qing Guo,* Zhibo Ma,* Dongxu Dai, Hongjun Fan,* and Xueming Yang* State Key Laboratory of Molecular Reaction Dynamics, Dalian Institute of Chemical Physics, 457 Zhongshan Road, Dalian 116023, Liaoning, P. R. China S Supporting Information *

ABSTRACT: Photoinduced water dissociation on rutile-TiO2 was investigated using various methods. Experimental results reveal that the water dissociation occurs via transferring an H atom to a bridge bonded oxygen site and ejecting an OH radical to the gas phase during irradiation. The reaction is strongly suppressed as the water coverage increases. Further scanning tunneling microscopy study demonstrates that hydrogen bonds between water molecules have a dramatic effect on the reaction. Interestingly, a single hydrogen bond in water dimer enhances the water dissociation reaction, while one-dimensional hydrogen bonds in water chains inhibit the reaction. Density functional theory calculations indicate that the effect of hydrogen bonds on the OH dissociation energy is likely the origin of this remarkable behavior. The results suggest that avoiding a strong hydrogen bond network between water molecules is crucial for water splitting. ince the first demonstration of hydrogen production from photocatalytic water (H2O) splitting on TiO2 electrodes by Fujishima and Honda in 1972,1 tremendous efforts have been given to the study of the photocatalytic H2O splitting reaction,2−8 with the hope that eventually clean, affordable, and efficient hydrogen production from H2O can be achieved. Although intensive studies have been focused on its reaction in which hydroxyl radical (·OH) is an important intermediate,9−12 many fundamental questions regarding the elementary processes of H2O splitting on TiO2 remain unresolved. For example, what is the exact molecular mechanism of the H2O splitting reaction on TiO2, and do hydrogen bonds affect this reaction? The clarification of these important questions will certainly help us to understand the key factors in the H2O splitting reaction. The hydrogen bond is a ubiquitous feature in nature. It is essential in folding reaction of proteins,13,14 proton transfer,15,16 and in molecular self-assembly.17,18 It is also a major determinant of specificity in enzyme catalysis and in biological information transfer, and it can influence directly the rate of enzymatic reactions.19−21 Especially for enzymatic reactions,22 the low barrier hydrogen bonds (LBHBs) can provide at least 5 orders of magnitude in rate acceleration for some enzymatic reactions with proper strength of the LBHBs. Shiotari and coworkers23 also reveal that hydrogen bond coupling between H2O and NO on a well-defined Cu(110) surface can induce back-donation and plays a crucial role in the catalytic reduction of NO by weakening the N−O bond. Recently, a series of twodimensional infrared (2D-IR) spectroscopy results24,25 demonstrated that vibrational energy relaxation can occur efficiently because of hydrogen bonds in liquid mixture in picosecond

S

© XXXX American Chemical Society

time scale. Therefore, it is interesting to learn whether the hydrogen bond can enhance or inhibit the photocatalytic H2O splitting process. Although the interaction of H2O with the rutile(R)TiO2(110) surface26−32 has been extensively investigated, only a few experimental studies with emphasis on the hydrogen bonds between H2O molecules on R-TiO2(110) have been carried out.33−35 Previous scanning tunneling microscopy (STM)33 and infrared reflection−absorption spectroscopy (IRRAS)34 investigations show that H2O molecules adsorbed on the five-coordinated Ti4+ (Ti5c) sites start to form onedimensional chain structures with hydrogen bonds along the Ti5c row direction as the coverage of H2O increases on RTiO2(110). We have recently discovered that H2O on RTiO2(110) cannot be dissociated at 400 nm.36 Further experimental studies in our laboratory show that H2O can be clearly dissociated on the same surface at 266 nm, consistent with the STM result in ref 12 at this wavelength. To understand the role of hydrogen bonds in the H2O dissociation reaction at the fundamental level, we have systematically investigated the H2O dissociation reaction on the TiO2(110) surface under 266 nm radiation using temperature-programmed desorption (TPD), time-of-flight (TOF) spectrometry, and STM methods in combination with density functional theory (DFT) calculations. Our results show unequivocally that the hydrogen bond between surface H2O molecules has a dramatic effect on the H2O dissociation reaction on R-TiO2(110). Received: January 4, 2016 Accepted: January 25, 2016

603

DOI: 10.1021/acs.jpclett.6b00015 J. Phys. Chem. Lett. 2016, 7, 603−608

Letter

The Journal of Physical Chemistry Letters Figure 1A−D shows the TPD spectra acquired at a mass-tocharge ratio (m/z) of 18 (H2O+) after the R-TiO2(110)

Figure 1. TPD and TOF spectra as a function of the coverage of H2O. Panels A−D show TPD spectra acquired at m/z = 18 (H2O+) after different coverages of H2O were adsorbed on R-TiO2(110) at 100 K and irradiated for 0 min (blue) and 60 min (red) at 100 K by 266 nm, 70 mW/cm2. Panels E−H show TOF spectra collected at m/z = 18 (H2O+) and m/z = 17 (OH+) after different coverages of H2O were adsorbed on R-TiO2(110) at 100 K and irradiated for 60 min at 100 K by 266 nm, 70 mW/cm2.

Figure 2. (A) Relative dissociation efficiency of surface H2O molecules versus surface H2O coverage, obtained from TPD (solid circles) and TOF (solid squares) results from Figure 1. (B) Percentage of surface H2O molecules photodissociated as H2O monomer, dimer, trimer, tetramer, and long cluster, obtained from STM experiment. The data point of n > 5 is an average value for all long H2O clusters on the surface.

surfaces (at 100 K) with different surface coverages (0.12, 0.25, 0.50, and 0.75 ML) (1 ML ≈ 5.2 × 1014/cm2) of H2O were irradiated for 60 min with 266 nm light, and Figure 1E−H shows the TOF spectra of the photodesorbed products at m/z = 18 (H2O+) and 17 (OH+) directly ejected from these surfaces during laser irradiation. Before irradiation, two peaks are observed in the TPD spectra at m/z = 18. The low-temperature (LT) peak (∼300 K) is assigned to the molecularly adsorbed H2O at the Ti5c sites (H2OTi), while the TPD peak (∼500 K) is due to the recombination of hydroxyls on the bridge bonded oxygen (BBO) rows. The high-temperature (HT) TPD peak is a measurement of the H atom coverage on the BBO rows and thus is an indicator of the amount of H2O dissociation because H atoms from H2O molecules will always dissociate to the BBO sites. After the surface is irradiated with 266 nm for 60 min, the LT TPD peak in these spectra at m/z = 18 all decreases but with considerably different amount of reduction for different H2O coverages. At 0.12 ML coverage, the LT TPD peak after 60 min irradiation by 266 nm light is nearly disappeared, while at 0.75 ML coverage, the peak is reduced only slightly after 60 min of irradiation. This suggests that H2O is photodissociated or photodesorbed by 266 nm light and that the reaction is strongly dependent on the coverage of H2O on R-TiO2(110). Evidence of H2O dissociation can also be seen clearly in the HT TPD peak (Figure 1A) which increases considerably after laser irradiation for the 0.12 ML H2O covered surface, implying that a large amount of H atoms on the BBO sites (BBO-H) are produced by the laser photolysis of H2O at 266 nm. However, the increase of the HT TPD peak is gradually reduced as the surface coverage increases (Figure 1B−D). Using the increase of the HT TPD peak as the indicator of H2O dissociation, divided by the initial surface coverage, a relative H2O dissociation efficiency is obtained (Figure 2A). This strongly suggests that the efficiency of H2O dissociation is greatly reduced as the H2O coverage increases.

Direct evidence of H2O dissociation on the surface, gas-phase OH radicals, were detected clearly by TOF measurement for the 0.12 ML H2O covered R-TiO2(110) surface during laser irradiation (Figure 1E). In the TOF spectra at 0.12 ML coverage, the m/z = 17 TOF signal is considerably larger than the m/z = 18 TOF signal. We have measured the mass cracking pattern of H2O in the detector; the m/z = 18 signal is much bigger than the m/z = 17 signal (Figure S1). Therefore, the larger TOF signal at m/z = 17 in Figure 1E implies that gasphase OH radicals are produced in the photolysis of H2O on RTiO2(110). This indicates that H2O dissociation occurs likely via transferring an H atom to the BBO site and ejecting an OH radical from the surface (Figure S2). Later STM results also support this picture. As the H2O coverage increases, the ratio of the TOF intensity of m/z = 17 and 18 changes quite dramatically from 2.5:1 at 0.12 ML coverage to 0.52:1 at 0.75 ML coverage (Figure 1E−H). The amount of OH radicals ejected from the surface is also an indicator of the amount of H2O molecules photodissociated. We therefore use the TOF signals (Figure 1E−H) to estimate the relative H2O splitting efficiency for different H2O coverages. The OH radical signal can be obtained by subtracting the H2O cracking contribution (OH+) from the m/z = 17 TOF signal, using the cracking pattern measured in this experiment (Figure S1). The obtained OH signal was then divided by the surface coverage to obtain the relative H2O dissociation efficiency, which is also shown in Figure 2A. The result is in rather good agreement with that estimated from the HT TPD peak. Both results suggest high surface H2O coverage inhibits the H2O dissociation reaction. The dynamical origin of this remarkable phenomenon, however, is not immediately clear. A very recent STM investigation of H2O covered R-TiO2(110) surface by Lee and co-workers33 has provided some clues. In their study, they show that H2O molecules adsorb on the Ti5c sites mainly in the form of monomer at very low surface coverage. As the coverage of H2O increases, the adsorbed H2O molecules start to form 604

DOI: 10.1021/acs.jpclett.6b00015 J. Phys. Chem. Lett. 2016, 7, 603−608

Letter

The Journal of Physical Chemistry Letters various lengths of one-dimensional chain structures with hydrogen bonds along the Ti5c row direction, while the number of adsorbed H2O monomers decreases. This naturally led us to postulate that the hydrogen bond between H2O molecules on R-TiO2(110) might be the reason for the greatly reduced photoreactivity of H2O on the surface. To test this conjecture, we performed a high-resolution STM experiment to investigate the photoinduced H2O dissociation reaction at a single molecular level. As shown in Figure 3A, after

Figure 4. Photodissociation of long chain under high coverage of H2O on R-TiO2(110). (A) Image of H2O chain after H2O adsorption. (B) The same area after irradiation. Four dissociated BBO-Hs were observed. (C) After magnification of the area of blue square zone, two heights of new feature on BBO row are marked by blue and green lines. (D) Cut profile of OH and OH pair; the heights are consistent with panel C.

Figure 3. STM images of adsorbed H2O species, monomer, and H2O clusters on R-TiO2(110) at 80 K and the dissociation of these species (acquired at bias voltage of +1.25 V and set point current of 100 pA). (A) Images of adsorbed H2O species on the R-TiO2(110) surface (size of 5.2 × 7.6 nm2), from H2O monomer to tetramer. (B) STM images of photoinduced dissociation of different H2O species, from monomer (I), dimer (II), trimer (III), to tetramer (IV). The STM images of these species were taken before and after UV light irradiation, with bare surface images in the same surface areas. Note that the BBO vacancy on the surface under UV light irradiation could also move.

the blue square area were observed in Figure 4B. In Figure 4C, we magnified the blue square area and measured the cut profile of the two bright spots. The height of the two spots are ∼70− 80 pm on BBO site and ∼110−120 pm on BBOv site, which are consistent with the height of the established OH and OH pair in Figure 4D. The single OH from the photodissociation of H2O molecule and OH pair from spontaneous dissociation of H2O at the BBOv site could be resolved. Four new H atoms on BBO sites were observed in Figure 4B after irradiation by 266 nm light for 60 min (marked by blue arrows). Comparing with the image before light irradiation (Figure 4A), we found that all the dissociations took place at one end of long chains. According to the crystal lattice parameters, there are ∼1170 Ti5c sites in the image and the coverage of H2O on the surface is 0.8 ML; the dissociation probability is roughly estimated to be about 0.3%. From the above STM experiments, the numbers of H2O molecules dissociated in the form of monomer, dimer, trimer, tetramer, and long cluster chain can be counted directly in a larger STM imaged surface area (Table 1). Using these numbers, we can calculate the efficiency of photoinduced dissociation of H2O monomer and clusters. It is clearly shown that about 3.5% of H2O monomers on the R-TiO2(110) surface

the R-TiO2(110) surface was exposed with 0.08 ML H2O, four types of images are observed on the surface, which can be assigned to H2O monomer, dimer, trimer, and tetramer on the Ti5c sites. We then irradiated the surface with 266 nm light for 60 min. The STM images of all identified H2O monomers and clusters were then obtained again after light irradiation. Clearly, some of the monomers adsorbed on the Ti5c sites are dissociated, with one H atom transferred to the adjacent BBO site and the OH radical left on the Ti5c site disappeared (Figure 3B I), which is consistent with the OH ejection picture. For the dimer (Figure 3B,II), one end of the image is tilted toward the BBO row after light irradiation. We then manipulated the undissociated H2O molecule on the Ti5c site away from the original adsorption site using the STM tip, and the BBO-H was clearly exposed. This implies that one of the H2O molecules in the dimer is dissociated by 266 nm light. Similarly, one end of the images of the trimer (Figure 3B,III) and tetramer (Figure 3B,IV) was also tilted toward the BBO rows after light irradiation, indicating that H2O molecules at one end of these clusters are photodissociated. We have also performed photocatalysis studies on R-TiO2(110) surface with much higher H2O coverage, in which most of the H2O molecules exist in the form of long chains, as observed in ref 33. Experimental results show that only a few H2O molecules at the end of some long chains are dissociated in a large surface area (Figure 4). As shown in Figure 4A, lots of long H2O chains were observed on the R-TiO2(110) surface with a high coverage (0.8 ML) of H2O; BBO vacancy (BBOv) and BBO sites are marked by green and blue arrows, respectively. After laser irradiation, two bright spots on both marked positions in

Table 1. Dissociation Probability of H2O Monomer and Clusters in a Low H2O Coverage Performed by STM Experimenta

total number dissociation events reaction probability (per cluster) reaction probability (per water molecule)

monomer (H2O)

dimer (H2O)2

trimer (H2O)3

tetramer (H2O)4

13129 459 3.5%

1332 240 18.0%

114 9 7.2%

27 1 3.6%

3.5%

9.0%

2.6%

0.9%

a The total number of H2O monomers and clusters is counted in the whole surface area monitored by STM.

605

DOI: 10.1021/acs.jpclett.6b00015 J. Phys. Chem. Lett. 2016, 7, 603−608

Letter

The Journal of Physical Chemistry Letters is dissociated, which is similar to previous observations.12 However, H2O dimers show a significantly higher photoreactivity; about 18.0% of dimers (9% per H2O molecule in all dimers) has one H2O molecule dissociated. This is significantly higher than that of the H2O monomer on the surface. It seems that intermolecular hydrogen bonds have considerably enhanced the photoreactivity for dimer. For trimers and tetramers, the probability of photodissociation drops to 2.6% and 0.9% per H2O molecule, respectively (Table 1). The dissociation efficiency of longer H2O chains on a high-coverage surface (0.8 ML) is about 0.3% (Figure 4), which is more than 10 times lower than that of the monomer. Figure 2B plots the probability of H2O dissociation per H2O molecule in the form of monomer, dimer, trimer, tetramer, and longer H2O chains, demonstrating the profound influence of the intermolecular hydrogen bond in the H2O dissociation reaction. The low photoreactivity of long H2O clusters observed at the molecular level is consistent with the above TPD/TOF results, which shows that the H2O dissociation reaction is depressed by increasing the H2O coverage. This clearly suggests that onedimensional (1-D) hydrogen bonds between H2O molecules greatly suppress the H2O dissociation reaction. A big surprise in the experimental result is that the H2O dissociation probability for the dimer is significantly larger than that for the monomer. To gain further insights into the intermolecular hydrogen bond effect, we have carried out theoretical calculations (PBE method, six O−Ti−O layers model, and VASP program; see Supporting Information for computational details) on the H2O dissociation process on R-TiO2(110) on the ground electronic state. The H2O clusters (H2O)n (n = 1−6) covered surfaces were modeled by a 8 × 1 surface unit cell. Similar to previous study,33,34 we found that there are two types of H2O clusters (Figure S3). In the first type, all hydrogen bonds between H2O molecules and BBO sites are on the same side, while in the other type, the hydrogen bonds between H2O molecules and BBO atoms alternate on opposite sides of the chain, forming a zigzag arrangement. The energetics for the two structures are quite similar. Here, we choose the first type structure for the detailed analysis of the intermolecular hydrogen bond effect on the OH bond dissociation energy. The OH bond dissociation energy was defined as the energy difference between the dissociative state (where one H2O molecule dissociates to Ti5c−OH and BBO-H) and the adsorbed state (where all H2O molecules adsorb on Ti5c). In the adsorbed H2O clusters, different H2O molecules have different hydrogen bond environment. The H2O molecule at one end of a linear H2O cluster (for example, H2O 4 in the tetramer in Figure 5A,IV) is a hydrogen bond donor (H2Odonor), while the H2O molecule at the other end of the cluster (for example, H2O 1 in the tetramer) is a hydrogen bond acceptor (H2Oacceptor). Using the DFT method described above, we have calculated the OH bond dissociation energies for both H2Odonor and H2Oacceptor, as well as the average dissociation energy for all H2O molecules in (H2O)n (n = 1−6). The calculated results (Figure 5B) show that the H2Oacceptor has the lowest OH dissociation energy, while the H2Odonor has the highest OH dissociation energy. Thus, the calculated result clearly suggests that H2Oacceptor is the easiest molecule to be dissociated in the adsorbed H2O chain, in agreement with the experimental observation that only one end H2O molecule is photodissociated in all different H2O clusters. Upon dissociation, H2O changes to the more negatively charged (Ti5c) OH, which is a better hydrogen bond acceptor. This is why

Figure 5. (A) Top view of the calculated structures of adsorbed H2O monomer and clusters (from dimer to tetramer) with hydrogen bond to one side of BBO rows on R-TiO2(110). At one end of these linear surface H2O clusters is the hydrogen bond acceptor H2O molecule (1), at the other end is the hydrogen bond donor H2O molecule (2, 3, 4). (B) Calculated OH bond dissociation energies of these adsorbed H2O molecules. Calculated values of the OH dissociation energy for the hydrogen bond acceptor H2O molecules as well as for the hydrogen bond donor molecules are shown, along with the averaged OH dissociation energy for each cluster. The OH bond dissociation energy in the H2O molecules as hydrogen bond acceptor is significantly lower than that as hydrogen bond donor.

H2Oacceptor has the largest dissociation probability because its dissociative state obtained additional stability from the strengthened hydrogen bond. For the same reason, the H2Odonor has the lowest dissociation probability because (Ti5c) OH in the dissociative state is a poor hydrogen bond donor. The calculated dissociation energies also explain why H2O dimer on the surface has the highest photoreaction probability because its OH dissociation energy is shown to be the lowest among all the clusters for the H2Oacceptor molecule, and it is much lower than that of the monomer. As the number of H2O molecules increases in the cluster, the OH dissociation energy of H2Oacceptor increases, and the same is true for the average OH bond dissociation energy. This means that OH dissociation becomes more difficult as the cluster size increases, which is also consistent with the experimental observation. The dissociation energies of H2O monomer and dimer were also calculated on an 8 × 1 slab with a BBOv. (The H2O molecule is not close to the vacancy.) In the monomer, the dissociation energy is 0.074 eV. In the H2O dimer, the dissociation energies are 0.031 eV for H2Oacceptor and 0.157 eV for H2Odonor. The dissociation is a little bit more endothermic on the surface with BBOv, but the change trend for the dissociation energy of H2O clusters (dimer is easier to be dissociated than monomer) is the same as that for the stoichiometric surface. The dissociation barriers of the H2O monomer and dimer were also calculated. In the monomer, the barrier is 0.183 eV. In the H2O dimer, the dissociation barriers 606

DOI: 10.1021/acs.jpclett.6b00015 J. Phys. Chem. Lett. 2016, 7, 603−608

Letter

The Journal of Physical Chemistry Letters

wave function was expanded by plane wave with kinetic cutoff of 400 eV and density cutoff of 650 eV. The generalized gradient approximation (GGA) with the spin-polarized Perdew−Burke−Ernzerh (PBE) functional was used for all of the calculations. The surface was modeled with a six-layer slab cut out of a TiO2 crystal to expose the (110) surface. All Ti5c atoms on the bottom layer were saturated by H2O molecules to maintain the coordination environment in bulk. The periodically repeated slabs were decoupled by 15 Å vacuum gaps. A Monkhorst−Pack grid of 2 × 1 × 1 k points was used for the 8 × 1 surface unit cell. H2O molecules were adsorbed on the top layer (Figure S3). The chemisorbed species and the top four TiO2 layers were allowed to relax until the residual forces were less than 0.03 eV/Å, while the remaining atoms were fixed at their bulk truncated positions. Isolated gas-phase molecules were optimized in a (15 × 15 × 15) unit cell with a single kpoint.

are 0.092 eV for H2Oacceptor and 0.219 eV for H2Odonor. Therefore, in these test systems, a higher barrier associates with a more endothermic (or less exothermic) process, and vice versa. This means the dissociation energies and the dissociation barriers predict similar dissociation probability for H2O in different H2O clusters. We have also calculated the dissociation energies of staggered H2O chains, and the change trends for the dissociation probability of H2O in (H2O)n (n = 1−6) are exactly the same as the eclipsed one shown above. In the present study, pulsed UV light is used to investigate photochemistry on a well-characterized surface in ultrahigh vacuum, whereas most photocatalysis studies use particle photocatalysts in bulk condition and discharge lamps that also contain UV light. Thus, we also carried out a highresolution STM experiment to investigate the photoinduced H2O dissociation reaction with UV light obtained from a Hg− Xe lamp. Because the Hg lamp is a wide spectral range light source, the power intensity for spectrum at 200−400 nm range (