Highly mobile and reactive state of hydrogen in metal oxide semiconductors at room temperature

Hydrogen in metal oxides usually strongly associates with a neighboring oxygen ion through an O-H bond and thus displays a high stability. Here we report a novel state of hydrogen with unusually high mobility and reactivity in metal oxides at room temperature. We show that freshly doped hydrogen in Nb2O5 and WO3 polycrystals via electrochemical hydrogenation can reduce Cu2+ ions into Cu0 if the polycrystals are immersed in a CuSO4 solution, while this would not happen if the hydrogenated polycrystals have been placed in air for several hours before the immersion. Time-dependent studies of electrochemically hydrogenated rutile single crystals reveal two distinct states of hydrogen: one as protons covalently bonded to oxygen ions, while the other one is highly unstable with a lifetime of just a few hours. Observation of this mobile and reactive state of hydrogen will provide new insight into numerous moderate and low temperature interactions between metal oxides and hydrogen.

oxygen ions, while the other one being highly mobile and reactive, probably as isolated protons ''free'' to move in the oxide lattices.

Results
As illustrated in Fig. 1a, we electrochemically hydrogenated an Nb 2 O 5 ceramic pellet for 24 h; the white color turned into deep blue due to the incorporation of hydrogen in the oxide lattices 15 . After that, we immediately immersed the pellet in a 0.5 M CuSO 4 solution for 10 min at room temperature. Surprisingly, we observed a reddish layer appeared on the pellet surface, as shown in Fig. 1b. In contrast, if the hydrogenated pellet (another sample) was placed in air for 10 h at room temperature and then immersed in the CuSO 4 solution, no such layer was deposited though it was still deep blue after 10 h of aging. XRD analysis reveals that the reddish layer is copper as shown in Fig. 1c, and the diffraction peaks of the Cu are much stronger than those of the Nb 2 O 5 ceramics. Since Nb 2 O 5 by itself could not reduce Cu 21 ions as shown in Fig. 1b, the only possible route of the formation of Cu is the reduction of Cu 21 by hydrogen from the hydrogenated Nb 2 O 5 . What should be emphasized is that in just 10 min the pellet surface was covered by a thick copper layer, indicating that a relatively large amount of doped hydrogen exists in a highly mobile and reactive state, which could diffuse out of the Nb 2 O 5 and react with Cu 21 at the surface. The high mobility and reactivity also imply a short lifetime of such hydrogen, and after 10 h of aging they almost disappeared from the oxide lattices so no copper was deposited on the aged pellet during the immersion. Nb 2 O 5 is not the only oxide that can accommodate such highly mobile and reactive hydrogen. Indeed copper was also found to be deposited on a 24 h electrochemically hydrogenated WO 3 pellet when it was immersed in 0.5 M CuSO 4 solution for 10 min immediately after the hydrogenation, but the amount was much less than that formed on similarly hydrogenated Nb 2 O 5 (see the XRD pattern in Fig. 1c). As a large amount of hydrogen can be electrochemically inserted into WO 3 , it is strange that only a little portion of inserted hydrogen seems to exist in the mobile and reactive state so the amount of deposited copper was very small. These results suggest that this special state of hydrogen may be not rare among metal oxides, while its concentration can be dramatically different among different metal oxides.  We expect that it should be possible to reveal a more detailed evolution of this unstable hydrogen in metal oxide single crystals. As Nb 2 O 5 and WO 3 single crystals are not available, we turned to TiO 2 rutile single crystals to conduct time-dependent study. These single crystals have nearly perfect crystal lattices that can eliminate the influence of crystal defects to doped hydrogen. For some metal oxides, phase transformation or crystal reformation occurs upon the hydrogen doping 16 . In this study, although we have tried to insert as much hydrogen as possible into the rutile single crystals, no new phases could be detected through XRD. Instead, the lattice of the rutile single crystals was found to be distorted to an extent beyond expectation when much hydrogen was inserted. Figure 2a shows XRD patterns of a hydrogenated rutile single crystal measured at different aging times, t, after 200 h of electrochemical hydrogen insertion. The (002) peak of as-received single crystals is sharp and highly symmetric (see the inset). In contrast, the (002) diffraction peak of the hydrogenated crystal becomes broad and is distorted to a very irregular shape, indicating non-uniform plane spacing or hydrogen distribution. An important observation is that there exists a wide step at the low angle side of the peak for short aging times, e.g., at 62.42u as indicated by the triangle for t 5 0.3 h. We note that while the specific shape of the (002) peak could vary depending on the hydrogen insertion time, the broadening of the peak and the arising of a wide step always occur (see Supplementary Information, Fig. S1). Chester and Bradhurst have electrochemically inserted hydrogen into vacuum-reduced rutile single crystals and found that the hydrogen concentration gradient was much higher in the first few microns of the surface layer than in the bulk region (,100 mm deep) 19 . The wide step revealed in our study can be well explained in term of this layered structure: the concentration of hydrogen is particularly high at the surface, where the plane spacing is expanded by 0.54% at t 5 0.3 h compared to that of the undoped (002) plane, while at the same time the expansion of the inner layer in the bulk region is only 0.16% as indicated by the distorted peak.
The XRD patterns further reveal that the distortion evolved with the aging time by placing the hydrogenated crystals in air at room temperature, indicating that the distorted crystal lattice as well as the layered structure subsequently experiences a spontaneous and systematic restoration as shown in Fig. 2b. The lattice expansion, Dd, of the hydrogenated crystal decreases exponentially by where Dd 0 is the initial lattice expansion and t is a decay constant. For the case of the wide step, Dd 0 5 0.58% and t 5 3 h, indicating that the layered structure exists only for a few hours in our sample after the hydrogen insertion, and thus the hydrogen in the layer is highly mobile. In contrast, it takes longer aging time for the distorted peak to restore to the original state. The distorted (002) peak in Fig. 2a gradually decreases in width, and increases in height and symmetry following the eq. 1 with Dd 0 5 0.17% and t 5 39 h. After an aging time of t 5 150 h, the peak has almost restored to its original highly sharp and symmetric shape. Spontaneous recovery was also observed for other initial distortions ( Supplementary  Information, Fig. S2). These two dramatically different decay constants exactly reflect the obvious difference in mobility observed for the step and the peak. The difference suggests that hydrogen in the surface layer is not only of a higher concentration but also much more mobile than that in the bulk region, so there exist two distinct states of hydrogen with different mobilities in these electrochemically hydrogenated rutile single crystals. This is consistent with the observations that only the ''freshly'' hydrogenated Nb 2 O 5 and WO 3 polycrystals could reduce Cu 21 ions but those aged samples could not. In other words, highly unstable hydrogen has been observed in all these three metal oxides that were investigated in this study, although its concentration is dramatically different in different metal oxides. Figure 3 shows infrared absorption spectra measured at 300 K for a rutile single crystal that has been electrochemically hydrogenated for 120 h and an as-received rutile single crystal. The as-received rutile single crystals are colorless and almost transparent without clear absorption peak in the range of 3100-3400 cm 21 that is common for as-grown rutile single crystals. In contrast, the hydrogenated sample was deep-blue in color and its transparency was greatly decreased. A strong and broad absorption peak centered at 3279 cm 21 can be clearly observed. This peak is continuously weakened with increasing aging time and meanwhile the transmittance of the hydrogen-inserted rutile single crystal is increased. It is well known that the infrared absorption peak at 3279 cm 21 is due to the stretch mode of O-H bond at room temperature in rutile. So this peak can be assigned to the stretch mode of O-H bond 20 . As a matter of fact, electrochemically doped hydrogen has been found to greatly  increase the conductivity and act as a shallow donor in rutile single crystals 21 . No other hydrogen-related structures can be observed over a wide frequency range of 500-4000 cm 21 ( Supplementary  Information, Fig. S3), and it is necessary to point out that hydrogen can exist in metal oxides in some forms that are not seen by IR spectroscopy 22,23 . Obviously, for the two distinct states of hydrogen revealed by XRD, the one with a bigger decay constant can be identified as protons covalently bound to oxygen ions according to the infrared absorption spectra analysis, which can be expressed as (OH) .

24
; while the highly mobile one is still largely unknown, e.g., its charge state that can be either positive, neutral or negative 25 . For simplicity, we tentatively assume a positive charge state for it and term it as H .
im (i for interstitial and m for mobile) hereinafter.

Discussion
Hydrogen existing as (OH) . O is well-known for metal oxides and is found in many applications of metal oxides. For example, for electrochromic metal oxide WO 3 , (OH) .
O is charged/discharged reversibly through controlling applied voltage 10 , which can be expressed using the notation of Kröger and Vink in the following way: where W 0 W represents a W 51 at the lattice site of W 61 in WO 3 and H 1 (a) a proton in a solution. The electrochromic effect is directly related the formation of W 0 W , namely W 51 in the lattice. It should be pointed out that the formation of O-H bond tends to weaken the ferroelectric distortion in WO 3 10 and with increasing content of (OH) .
O , the crystal structure is modified from monoclinic (WO 3 ) to tetragonal (H 0.1 WO 3 and H 0.33 WO 3 ), and to cubic (H 0.5 WO 3 ) 16 . Through thermal treatment, it leaves from the lattice of WO 3 as water 26 : O , H 2 should be formed from H . im , which exists in electrochemically hydrogenated Nb 2 O 5 , upon thermal treatment in the following way: where the electron is weakly-bonded or free in the metal oxide and the out-diffusion of hydrogen decreases the concentration of charge carriers and thus the conductivity of metal oxides. As for the deposition of copper on ''freshly'' hydrogenated Nb 2 O 5 and WO 3 pellets when being immersed in the CuSO 4 solution, it reveals the following reaction between H . im and Cu 21 at room temperature: As H . im can be of a high concentration in hydrogenated Nb 2 O 5 , a large portion of H 2 is formed besides H 2 O upon thermal treatment, and a large amount of copper is formed when immersed in CuSO 4 solution at room temperature. On the other hand, since the content of H .
im is very small in electrochemically hydrogenated WO 3 , it is difficult to collect H 2 upon heat-treatment and only H 2 O can be detected, and the amount of copper deposition is very small when immersed in CuSO 4 solution.
It is clear that with such surprisingly high mobility and reactivity at room temperature, this special state of hydrogen will not be observed or studied in samples treated through hydrogen-doping at elevated temperatures. Our finding of the highly mobile and reactive state of hydrogen, and this room-temperature hydrogen doping method as well, are thus especially valuable for studying those interactions between metal oxides and hydrogen at moderate and low temperatures. For example, the peculiar aging behavior of electrochemically hydrogenated TiO 2 rutile single crystals 14,21 , and some other metal oxides as well 12,18,27 , can now be better understood: the mobile hydrogen H .
im should play an important role in the initial stage of aging, while (OH) .
O should be mainly responsible for the subsequent aging 28 . For hydrogen-sensing of metal oxide semiconductors, it is generally considered that the change of electrical resistance with hydrogen exposure is only related to the spilt hydrogen atoms on the surface 9,29,30 , in this way electron transfer from hydrogen to metal oxides occurs while hydrogen is still highly mobile. According to our present results, however, ionized hydrogen, or proton, can actually exist inside metal oxides while being highly mobile. So it is necessary to examine whether hydrogen will enter into the lattice of metal oxides in the course of hydrogen-sensing. Similarly, hydrogen entrance into the lattice of metal oxides should also be considered for other chemisorption of atomic hydrogen on the surface of metal oxides 31,32 . Nb 2 O 5 is so far the most effective oxide promoter for accelerating adsorption/desorption of hydrogen in magnesium hydride storage materials. It has been assumed that Nb 2 O 5 may play the role as a reservoir loading a certain amount of split hydrogen in its lattice in the course of catalysis 8,11,26 . Obviously, our results provide a direct support for this assumption and more attention should be paid to it when the catalysis mechanism of Nb 2 O 5 is studied.
In summary, by studying electrochemically hydrogenated Nb 2 O 5 and WO 3 polycrystals, and TiO 2 rutile (001) single crystals, we found the coexistence of hydrogen in two distinct states inside the lattice of these metal oxides: one is as protons attached to oxygen ions by strong O-H bonds while the other one is highly mobile and reactive, but exists only for several hours in the lattice.  Methods Nb 2 O 5 samples were prepared through 2 h sintering at 1250uC in air of pellets (15 mm in diameter and 2 mm thick) pressed from analytical grade (99%) Nb 2 O 5 powders. Analytical grade (99%) WO 3 powder was first ball milled with de-ionized water for 24 h. After drying, the milled powder was pressed into pellets of 15 mm in diameter and 2 mm thick and the pellets were sintered at 1150uC for 2 h in air. The TiO 2 rutile single crystals used in this work were (001)-oriented 10 3 10 3 0.5 mm 3 commercial wafers purchased from HEFEI KEJING (China). Two 2 mm diameter silver electrodes were fired on the two sides of a corner of the single crystals. Electrochemical hydrogen insertion was conducted by immersing a sample in a 0.01 M NaOH solution. The Nb 2 O 5 or the WO 3 pellets or the silver electrodes of the rutile single crystals were connected to dc voltages (6 V) to act as the cathode to electrolyze water. Infrared absorption spectra were recorded with a Nicolet iS-10 Fourier transform spectrometer by measuring the transmittance of as-received and hydrogenated rutile single crystals at 300 K. The spectral resolution was 0.5 cm 21 . X-ray diffraction patterns were taken through an X-ray diffractometer (Philips PW 3719) with Cu Ka radiation.