Reactions of Hydrazoic Acid on TiO2 Nanoparticles: an Experimental

J. Phys. Chem. B 2005, 109, 5133-5142

5133

Reactions of Hydrazoic Acid on TiO2 Nanoparticles: an Experimental and Computational Study Jeng-Han Wang and M. C. Lin* Department of Chemistry, Emory UniVersity, Atlanta, Georgia 30322

Ying-Chieh Sun Department of Chemistry, National Taiwan Normal UniVersity, 88 TingChow Road, Sec. 4, Taipei 11718, Taiwan ReceiVed: September 16, 2004; In Final Form: December 16, 2004

This article reports the results of a computational and experimental study on the reaction of hydrazoic acid, HN3, adsorbed on 15-20 nm TiO2 particle films. Experimentally, FTIR spectra of HN3(a) have been measured by varying HN3 dosage, UV irradiation time and surface annealing temperature. Three sharp peaks, related to Va(NNN) of HN3(a) and N3(a) with different configurations in the 2000-2200 cm-1 region, and a broad band absorption, related to associated and isolated HN(a) and HO(a) adsorptions in the 3000-3800 cm-1 region, have been detected. Computationally, molecular structures, vibrational frequencies and adsorption energies of possible adsorbates including HN3 and its derivatives, N3, N2, NH, and H, have been predicted by first-principles calculations based on the density functional theory (DFT) and the pseudopotential method. On the basis of the experimental and computational results, the peak appeared at 2075 cm-1, which increases at a faster rate with HN3 exposure time, is attributed to a stable adsorbate, N3-Ti(a), with the predicted adsorption energy, Eads ) 13 kcal/mol. The peak at 2118 cm-1, which survives at the highest surface temperature in the heating experiment, is attributable to the most stable adsorbate, Ti-N2N(H)-O(a) with Eads ) 36 kcal/mol. The peak at 2170 cm-1, which vanishes most readily in all of the aforementioned experiments, is related to less stable molecular adsorbates, end-on HN3-Ti(a) with Eads ) 5 kcal/mol and side-on HN(N2)Ti(a) with Eads ) 8 kcal/mol. A potential energy diagram for the formation of various absorbates with their transition states has been established for the HN3/TiO2 system. On the basis of the predicted desorption energies, the four most stable products of the HN3 reaction on TiO2 are H-O(a), 118 kcal/mol; HN-O(a), 85 kcal/ mol; Ti-N2N(H)-O(a), 36 kcal/mol; and N3-O(a), 19 kcal/mol.

1. Introduction TiO2, considered to be a promising material for photochemical applications spanning from photocatalysis to wet solar cells to photoelectrochemical water splitting, has been widely investigated.1-3 Due to the narrowed band gap resulting from the p-state of nitrogen mixing with O 2p states, the films and powders of N-doped TiO2 (TiO2-xNx) had been shown to have a noticeable improvement over pure TiO2 under visible light in their optical absorption and the level of photocatalytic activities in the visible region.3-6 Among the N-containing species with different bond dissociation energies, such as N2 (9.8 eV), NH3 (4.5 eV) and hydrazoic acid, HN3 (0.5 eV), HN3 is considered to be the most efficient nitrogen source for N-doping of TiO2. We have recently employed it for the low-pressure organometallic chemical vapor deposition (OMCVD) of InN on TiO2 nanoparticle films;7 the deposited films over 15-20 nm TiO2 particles show a very broad UV/visible absorption between 390 and 800 nm, quite similar to Graetzel’s “black” dye.8 Similar OMCVD of GaN9 and InN10 have been made previously with HN3 as N-precursor for optoelectronic applications. Various studies of HN3 on semiconductors, metal surfaces, and insulators such as GaAs single crystal,11 Si(100)-2×1,12,13 * To whom correspondence [email protected].

should

be

addressed.

E-mail:

Si(111)-7×7,14 Ge(100)-2×1,15 C(100),16 polycrystalline gold,17 amorphous ice,17 Al(111)18 and NaCl(100),19 had been done experimentally by using high-resolution electron energy loss spectroscopy (HREELS) and infrared spectroscopy techniques for vibrational frequency measurements, temperature-programmed desorption (TPD) for surface reaction studies, X-ray photoelectron spectroscopy (XPS), Ultraviolet photoelectron spectroscopy (UPS), and Auger electron spectroscopy (AES) for binding energy detections, and low-energy electron diffraction (LEED) for surface morphology determinations. Furthermore, studies of HN3 on the Si(100)-2×1,20,21 Ge(100)-2×1,22,23 and C(100)-2×124 surfaces had also been done computationally with cluster model calculations. On the single crystalline silicon surfaces,12,13,25,26 HN3 chemically adsorbed on the surface at 120 K under a low dosage and formed dimers under higher dosages. Upon heating, HN3 dissociated at around 270 K, forming N2 and NH, ultimately leading to the production of silicon nitride at temperatures over 1200 K. On the single crystalline GaAs,11 308 nm excimer laser was used to photodissociate the chemically adsorbed HN3 into a mixture of NHx, N2 and N3 species at 120 K. Subsequent surface annealing led to nitridation of the GaAs crystalline. On the Ge(100) surface,15 HN3 chemically adsorbed on the surface at low temperatures, and dissociatively desorbed into N2 and H2 from the surface. The nitride began to decompose at about

10.1021/jp0458046 CCC: $30.25 © 2005 American Chemical Society Published on Web 02/22/2005

5134 J. Phys. Chem. B, Vol. 109, No. 11, 2005

Figure 1. (a) Design of the IR cell for the simultaneous photochemistry and IR spectroscopy. (b) Film on tungsten grid mounted on the stainless steel clamp.

750 K and was completely desorbed from the surface at 875 K. On the Al(111) surface,18 HN3 physically and chemically adsorbed on the surface at 100 K. N2 and NH dissociatively adsorbed on the surface around 125 K and N2 and H2 desorbed from the surface at 295 and 615 K, respectively. On the NaCl(100), polycrystalline gold and amorphous ice,17,19 the adsorbed HN3 formed ammonia azide, NH4N3 by UV or X-ray irradiation and surface annealing treatments. HN3 breaks at the N-H bond, instead of the HN-NN bond, forming C-N3 and C-H on the C(100) surface at 100 K,16 resulting from a stronger bonding between hydrogen and the diamond surface. In this paper, we have investigated the adsorption and reaction of HN3 on 15-20 nm TiO2 particle films experimentally and computationally. The objective of the present study lies in elucidating the mechanism of the HN3 adsorption and reactions on TiO2 under similar conditions employed in our InN OMCVD process7 cited in the Introduction. Sections 2 and 3 describe current experimental setup and the computational method, respectively. In section 4.1, the results of FTIR experiments carried out to systematically examine the effects of HN3 dosage, UV irradiation time and surface annealing temperature have been reported. In section 4.2, the computational results, which include the information on the optimized geometrical structures, adsorption energies and vibrational frequencies, are compared with the experimental observations reported in section 4.1. This paper is concluded in section 5 with a brief summary. 2. Experimental Setup The experiment was carried out in a stainless steel IR cell with two CaF2 windows sealed by Viton O-rings and one UV grade sapphire window, as shown in Figure 1(a). The IR cell is connected to a high-vacuum chamber with a base-pressure of 1 × 10-7 Torr evacuated with a Leybold turbomolecular pump (TURBOVAC 360). The sample, TiO2 nanoparticle film supported on a tungsten grid, was positioned in such a way that both the IR beam and the UV light are at a 45 ° incidence angle to the normal of the grid and was held rigidly by a power/ thermocouple feedthrough via a pair of stainless steel clamps, as shown in Figure 1(b). The sample could be resistively heated to 900 K measured by a K-type thermocouple spot-welded on the top-central part of the tungsten grid. This design was similar to that employed by Basu et al.,27 and had been widely applied elsewhere.27-30 FTIR spectra were measured by Bruker IFS66 FTIR spectrometer equipped with a MCT detector. The entire optical path

Wang et al. was purged with dry air passing through H2O- and CO2-free filters (Ballston). The spectra presented in this report have been subtracted with the spectrum of a clean TiO2 nanoparticle film with 4 cm-1 resolution and averaged by 2000 scans. The UV light source used for the present photochemical experiments was a 1000 W high-pressure Hg(Xe) arc lamp (Oriel). The intensity of the light beam could be varied up to ∼2.0 W/cm2 as measured in air inside the IR cell with a power meter. TiO2 nanoparticles were prepared by a method similar to that reported by Zaban and co-workers.31 The synthesis for the TiO2 sol-gel solution of nanoparticles was carried out by the controlled hydrolysis of titanium (IV) isopropoxide (Ti(i-OC3H7)4, Aldrich 97%) in a mixture of glacial acetic acid water solution at 273 K. The resulting solution was heated to 353 K for 8 h and then autoclaved at 503 K for 12 h. The resulting TiO2 sol-gel solution was spread onto a tungsten grid (Alfa Aesar) and baked in an oven at 723 K to form TiO2 nanoparticle films with specific sizes. The film in the IR cell was then heated to 900 K under vacuum for 24 h to increase the adhesion between TiO2 nanoparticles and tungsten grids. The size of TiO2 nanoparticles lies in the range of 15-20 nm as measured by SEM and AFM images.7 Prior to every experiment, the film was annealed at 673 K in a vacuum with 2 Torr oxygen for 2 h to remove the small amount of organic contamination and to fill oxygen vacancies in TiO2. HN3 was prepared by acidification of NaN3 with 50% H3PO4 and purified with a train of cold traps maintained at dry ice and liquid nitrogen as described by Bu et al.32,33 The sample has to be handled with extreme care and safety precaution to avoid detonation which may occur unexpectedly. The purified HN3 was stored in a Pyrex tube at dry ice temperature as a pure and clear liquid. The vapor over the liquid HN3 at dry ice temperature was used as the effusive beam source and was introduced into the system through a 1/8 in. stainless steel tube above the sample. The gas flow rate was controlled by a combined needle and shut-off valves and monitored by a lowpressure transducer (MSK Baratron). 3. Computational Method The TiO2 nanoparticle film is a polycrystalline material with different TiO2 surface structures. The 101 and 110 surfaces of anatase and rutile, respectively, have lower energies with similar characteristics;34 they may coexist in the nanoparticle film. The annealing treatment to clean the film before each experiment as aforementioned may transform more anatase to rutile because of the small particle size.35,36 Besides, the TiO2(110) surface of rutile is the most studied structure computationally and has abundant information to verify our surface modeling discussed in section 4; it was therefore employed in the present simulation. The periodic surface is simulated by the repeated super cells in three directions. Repetition of the in-plane super cells creates an infinite slab, whereas periodicity in the direction perpendicular to the slab creates an infinite stack of slabs. The super cell geometry used in this study consisting of 16 [TiO2] units with the bond lengths and angles, initially defined according to the experimental result,37 as shown in Figure 2. To minimize the interaction between distinct slab surfaces in this infinitely periodic model system, dangling bonds in the lower surface of the super cell were saturated with hydrogen atoms to avoid unphysical low coordination numbers, and a vacuum region was used to separate the top and bottom surfaces of the slabs. The geometrical structures were optimized by Vienna ab initio simulation package (VASP),38-41 implementing the density functional theory. The exchange-correlation function was treated

Hydrazoic Acid Reactions on TiO2 Nanoparticles

J. Phys. Chem. B, Vol. 109, No. 11, 2005 5135

Figure 3. FTIR spectra of HN3 on TiO2 treated by varying HN3 exposure time from 5 to 25 min in the range 1800-3900 cm-1.

Figure 2. Perspective view of the TiO2(110) surface. The dashed lines indicate the super cell used in the slab calculations.

with the local-density approximation (LDA).42 The generalized gradient approximation (GGA)43,44 used for the total energy calculations was that of the Perdew-Wang 1991 (PW91) formulation,43 which has been shown to work well for surfaces.45 The core pseudopotentials supplied with VASP were used for the present calculation. The 10 3p, 3d, and 4s electrons of each Ti atom and the six 2s and 2p electrons of each O atom were explicitly considered. For the periodic condition, the electronic orbitals were expanded in a plane-wave basis. The plane-wave expansion includes all plane waves with their kinetic energies smaller than the chosen cutoff energy, pK2/2m < Ecut, which ensures the convergence with respect to the basis set. The Brillouin zone was sampled with the chosen Monkhorst-Pack46 k-points, which also ensures the convergence of the whole system. The vibrational frequencies of the optimized structures were calculated by Gaussian03.47 A hybrid Hartree-Fock/density functional theory (HF/DFT) method, B3LPY, which includes Becke’s three-parameter nonlocal-exchange function48 with the correlation functional of Lee-Yang-Parr,49 and the standard all-electron split-valance basis set 6-311G(d,p),50 were used for the vibrational frequency analysis. 4. Results and Discussion 4.1. Experimental Section. HN3 can dissociate in the gas phase and adsorbed state12-19 by the following two paths:

HN3 f N2 + NH

(1)

HN3 f N3 + H

(2)

Although the formation of N3 in reaction 2 was shown to be a secondary reaction (formed by NH abstraction)52-54 in the gas

phase, it has been detected in the present and earlier experiments on different surfaces.14,16,17 From these reactions, possible adsorbed species are HN3(a), N2(a), NH(a), N3(a) and H(a). The FTIR spectra presented in this section are related to these adsorbed species, and the three effects: HN3 dosage, UV irradiation time and surface annealing temperature, have been thoroughly examined. 4.1.1. HN3 Dosage Experiment. Figure 3 shows the FTIR spectra of a TiO2 nanoparticle film exposed to 0.1 Torr HN3 with the indicated time and followed by evacuation for one minute at room temperature. These spectra will be separately discussed in the region of 3000-3800 cm-1, which is related to the stretching vibrations of HN(a) and HO(a) adsorbates, and in the region of 2000-2200 cm-1 attributed to the asymmetric stretch, Vs(NNN), of HN3(a) and N3(a) species. The vibrations with weak IR intensities in the fingerprint region, e.g., NH deformation δa(N-H) of HN3(a) and symmetric stretching Vs(NNN) of HN3(a) and N3(a), are not shown in these spectra. All the assignments will be confirmed with the computational results described in section 4.2. The absorption in the region of 3000-3800 cm-1 is contributed from different types of IR absorptions: isolated and associated NH and OH vibrations. The NH may be from HN3 and dissociated HN, analogous to reaction 1. The OH may be formed from dissociated H adsorbed on surface O-sites, OH radicals and H2O on the surface. The associated HN(a) and HO(a) absorptions may be resulted from hydrogen bonding interaction with other HN(a), HO(a) or HN3(a). These hydrogen bonds strongly couple with each other and form a broad IR absorption. This broad band absorption, which was not observed in previous experiments under low HN3 dosages,14,17,18 is attributed to the higher HN3 dosage (0.1 Torr) and the higher surface temperature (room temperature) employed in the present experiment. Similar broad band absorptions were observed in the other high dosage experiments.28,29 Unlike the complex coupling in 3000-3800 cm-1, the three sharp peaks in the 2000-2200 cm-1 region are better resolved and can be deconvoluted by three Gaussian functions located at ca. 2075, 2118 and 2170 cm-1 as shown in Figure 4(a)-(e). The peaks at 2170 and 2075 cm-1 may be attributed to the N3 asymmetric stretching, Va(NNN), from the molecularly adsorbed HN3(a) and dissociatively adsorbed N3(a), respectively. The red shift of the Va(NNN) in N3(a) from that in HN3(a) can be expected, because electrons in the π bonds delocalized across the whole N3 radical after the dissociation of the N-H bond. The assignment of these two peaks is straightforward and similar to previous experimental observations.12-19 However, the assignment of the other peak at 2118 cm-1 is not immediately clear. It is probably derived from HN3(a) or N3(a) with different


Wang et al.

Figure 4. (a)-(e) FTIR spectra in Figure 3 (solid line) with peak fitting results using three Gaussian functions (dot lines). (f) Plot of the peak areas as a function of the HN3 dosing time.

adsorption structures. The specific assignment of this peak can be assisted by the computational results to be presented below. Finally, Figure 4(f) shows the changes of the peak areas varying with HN3 dosage. The peak at 2075 cm-1 shows the fastest increasing rate and the peak at 2170 cm-1 shows the slowest rate. This qualitative plot will be compared with the results obtained by varying with UV irradiation time and annealing temperature in the next two subsections. 4.1.2. UV Irradiation Experiment. To have a deeper insight on these peaks, the effect of UV irradiation on the surface is systematically examined. The infrared spectra in the photoreaction experiments are shown in Figure 5(a). A continuous UV lamp is applied as the light source during the whole set of experiments. To avoid collecting the scattered infrared from the UV lamp, the lamp was blocked during every one-minute IR measurement. In the UV irradiation, a slight temperature jump (∼20 K) from room temperature occurred. No additional peaks were found in the UV irradiation experiments; however, the peak intensities change at different rates during these experi-

ments. Using the deconvolution method employed in Figure 4, the plot of the three obvious peaks at 2075, 2118 and 2170 cm-1 in the UV irradiation experiment are shown in Figure 5(b). Among those three peaks, the one at 2075 cm-1 decays at the slowest rate under UV irradiation (Figure 5(b)), whereas it increases at the fastest rate in the HN3 exposure experiment (Figure 4(f)). This indicates that N3(a) is very stable on the surface. On the other hand, the peak at 2170 cm-1 dropped quickly under UV irradiation, indicating that the molecularly adsorbed HN3(a) easily desorbs from the surface or dissociates into other species by UV photons. This result is consistent with previous experiments, in which the adsorbed HN3(a) disappeared from other surfaces photolyzed by 308 nm excimer laser11,14 or X-ray.17 The peak at 2118 cm-1 decreases faster than the peak at 2075 cm-1 during the 60-minute UV irradiation. This implies that the absorption at 2118 cm-1 may be attributable to another form of HN3(a), which can dissociate to N3(a) by UV photons. 4.1.3. Surface Annealing Experiment. Figure 6(a) shows the infrared spectra of the thermally treated experiment by directly



Figure 5. (a) FTIR spectra of HN3 on TiO2 treated with UV irradiation for 15, 30, 45 and 60 min in the range 1800-3900 cm-1. (b) Plot of the peak areas as a function of the UV irradiation time.

Figure 6. (a) FTIR spectra of HN3 on TiO2 taken after surface annealing from room temperature to 523 K. (b) Plot of the peak areas as a function of surface temperature.

annealing the surface. The infrared spectra are taken after the surface is heated to the indicated temperatures for 5 min and then cooled to room temperature for spectral measurements to avoid excited phonon emitting from the surface. The plot related to the annealing temperatures is shown in Figure 6(b). Similar to the UV irradiation results, all the peaks decrease as the surface temperature increases. The peak at 2170 cm-1 disappears first at 373 K, which supports the same conclusion as above that molecularly adsorbed HN3(a) weakly bonded on the surface. However, the peak at 2118 cm-1 shows a slower decreasing rate than that of the peak at 2075 cm-1 as the surface temperature increases. This phenomenon indicates that the peak at 2118 cm-1 is attributable to the adsorption by a species with the greatest stability. Combining with the results from the UV irradiation experiment, this adsorbate is likely to be a stable form of HN3(a). It is worth noting that the surface annealing process may produce N(a) through the dissociating processes such as HN(a) f N(a) + H(a) and N3(a) f N(a) + N2(a) (or N2(g)), which cannot be detected from the IR spectra. These nitridation processes were observed in previous annealing experiments of HN3 on different surfaces14,15,18 and showed that HN3 is an efficient nitrogen source for the OMCVD experiments,9,10 as mentioned in the Introduction. To conclude, the experimental results show that HN3 can be molecularly and dissociatively adsorbed on the surface forming HN3(a) (V ) 2118 and 2170 cm-1) and N3(a) (V ) 2075 cm-1). No N2(a) and H-Ti(a) absorptions are found in the IR spectra because of small IR intensity and lower vibrational frequency,

respectively. With additional energy supplied by UV irradiation or heating, the molecularly adsorbed HN3(a) dissociates into adsorbed N3(a), NH(a) and H(a). This result agrees with previous experiments on other single-crystal surfaces.14-19 However, no energetic detail nor the surface reaction mechanisms had been studied before. In section 4.2, the surface reaction will be elucidated by the computational studies which provide adsorption energies and vibrational frequencies of these adsorbed species. 4.2. Computational Section. To ensure the reliability of the computational results, the method and parameters employed in the current study have first been examined by optimizing the bulk TiO2, the bare surface and the isolated gas molecules of HN3 and its dissociation products, which may be involved in the surface reactions. With acceptable accuracy, the computed condition is then applied on the surface reaction calculations, which can be used to rationalize experimental results. 4.2.1. Computational Condition Tests. The computational parameters, cutoff energy and k-point sampling regime, employed here have been tested for the accuracy by examining the optimization of the bulk of TiO2, which is tetragonal, and by characterizing its lattice constants, a and c. It is desirable to use a large cutoff energy and k-point to ensure that the total energy is converged, but in practice this must be balanced against the computational cost. The parameters with 4 × 5 × 4 Monkhorst-Pack46 k-points and 600 eV cutoff energy are used in the present calculations. At the cutoff energy and k-point settings, the calculated lattice parameters, a ) 4.593 Å and c


Wang et al.

TABLE 1: (a) Optimized Bond Lengths (Å) and the Heat of Reaction (kcal/mol), ∆Ereaction, for the HN3(g) Molecule in the Gas Phase Calculated by VASP and (b) Vibrational Frequencies (cm-1) Calculated at the B3LYP/6-311G(d,p) Level by Gaussian 03 with a Scaling Factor58,a (a)

r(N1-N2)

r(N2-N3)

r(N3-H)

∆Ereaction

HN3(g) N2(g) + NH(g) H(g) + N3(g)

1.146 (1.133)59 1.112 (1.098)61 1.193 (1.184)60

1.248 (1.237)59

1.034 (0.975)59 1.054 (1.034)61

0 70b 112c

(b) HN3(g) N3(g) + H(g) N2(g) + NH(g)

Vs(NN)

1.193 (1.184)60 Va(NNN)

Vs(NNN)

δa(NH)

Vs(NH)

2187 (2160)61 1652 (1645)60

1131 (1151)61 1318 (1320)60

1248 (1264)61

3346 (3336)61

2349 (2359)61

3149 (3282)61

a The numbers in parentheses are previous results. b ∆Ereaction(HN3 f N2 + NH) ) E(N2) + E(NH) - E(HN3). c ∆Ereaction(HN3 f N3 + H) ) E(N3) + E(H) - E(HN3).

) 2.933 Å, and internal coordinate, 0.296, have about 1% deviation from the experimental data.37 As shown in Figure 2, the TiO2(110) surface is modeled as an infinite slab and the periodic boundary conditions are applied. This model generates an infinite stack of two-dimensional slabs, each separates from its neighbors by a certain a vacuum layer. The surface super cell has dimensions of x2a x c along the (1h10) and (001) directions, where a and c are the lattice constants of the bulk TiO2 unit cell. The slab thickness is related to the number of layers, where each layer is defined as a (110) plane that contains both Ti and O atoms. In present calculations, a slab contained four layers (16 [TiO2] units) with all the bridged oxygen atoms and the Ti atoms on top of the surface as active sites. Dangling bonds in the bottom surface of the slab are saturated with hydrogen atoms to avoid unphysical interactions between slabs perpendicular to the surface. The vacuum separation is chosen equal to 10.4 Å, which is greater than pervious studies51-54 and guarantees no significant interactions between the neighboring slabs. Furthermore, we have tested the reliability of this surface model by computing the adsorption energy of H2O on the surface. The difference between the fully relaxed result, -12 kcal/mol, and the bottom layer fixed result, -20 kcal/mol (which is in close agreement with previous results55,56) is consistent with the tendency of the infinite-slab estimation.57 However; this type of difference is not noted in the current HN3/TiO2 system. As shown in Table 1, the bond lengths, energies and vibrational frequencies of these isolated gas molecules, HN3, N2, NH, N3 and H, which may be involved in the surface reactions, are also initially tested with the same parameters used for the following surface calculations. The corresponding errors of the bond lengths are found to be less than 4% and the vibrational frequencies predicted with the scaling factor,58 0.96, deviating only by about 1% from those of high level theoretical calculations and the experimental values.59-61 In these calculations, we have put one gas molecule isolated in a box of different lengths which reflect the intermolecular distances. For these gas molecules, the interactions between species are found to be negligible when the intermolecular distance is larger than 3 Å. 4.2.2. Surface Reaction Calculations. The computational results performed in the bulk TiO2, the water adsorption on TiO2 surface and the gas phase molecules imply that the present computational approach can provide a good description for gassurface reactions. First, the geometrical structures and adsorption energies of the adsorbates are optimized and calculated. Second, the adsorbate effect and hydrogen bond interaction are evaluated by comparing the results of coverage ) 0.5 and 1. Third, the frequencies of the stable adsorptions are calculated and compared with the experimental result. Finally, the potential energy

Figure 7. Optimized adsorbate structures presented with two layers of TiO2(110) planes. The coverage of all adsorptions is 0.5 (i.e., the adsorbates occupy one of the two surface Ti or O atoms in the super cell).

surface of the surface reactions is computed to account for the experimental observations. a. Adsorbate Structures. Possible adsorptions of HN3, N3, N2, NH and H on either the 5-fold-coordinate Ti atoms or on the bridged oxygen atoms, labeled as adsorbate-Ti or adsorbate-O, are optimized by calculations with the VASP code. The adsorption of HN3 can take place with different orientations, giving an end-on or a side-on structures, as proposed in the previous experiments of HN3 on the Si(100)-2×1 and Al(111) surfaces.11,14,18 The optimized structures presented with the partial surface model for brevity are shown in Figure 7 and the related bond lengths and adsorption energies are listed in Table 2. For HN3(a), the end-on structure, end-on HN3-Ti(a), with a 5-kcal/mol adsorption energy, has a more stable structure than end-on HN3-O(a) (0 kcal/mol). The negligibly small adsorption energy of the latter can be understood from the structure that the bond lengths of the physisorbed HN3(a) are the same as HN3(g) in the gas phase and the length of the N-O bond, 2.746



TABLE 2: Optimized Bond Lengths and Adsorption Energiesa (kcal/mol), Eads, for HN3 and Its Fragments on TiO2(110) Predicted with the Slab Method Using VASPb end-on HN3-Ti(a) end-on HN3-O(a) side-on HN(N2)-Ti(a) side-on HN(N2)-O(a) Ti-N2N(H)-O(a) Ti-N3H‚‚‚O(a) N3-Ti(a) N3-O(a) N2-Ti(a) N2-O(a) HN-Ti(a) HN-O(a) H-Ti(a) H-O(a) a

r(Ti(O)-N(H))

r(N1-N2)

r(N2-N3)

r(N3-H)

Eads

2.370 2.746 2.379 3.568 2.384 2.501 2.278 1.377 2.518 2.777 2.093 1.330 1.874 0.965

1.141 1.145 1.141 1.146 1.141 1.143 1.183 1.261 1.112 1.113

1.233 1.247 1.254 1.248 1.204 1.224 1.184 1.156

1.035 1.034 1.034 1.034 1.034 1.034

5 0 8 0 36 8 7 19 2 0 14 85 6 118

1.033 1.035

Eads ) -(Etotal - Emolecule - Esurface) ) -(heat of reaction). b Optimized structures are shown in Figure 7.

Figure 8. Optimized adsorbate structures and adsorption energies. aEads ) -(Etotal - 2Emolecule - Esurface) and bEads ) -(Etotal - EHN3 - Esurface), for adsorptions with coverage ) 1.

Å, is too long for N and O orbitals forming a stable chemical bond. These results, which indicate that HN3 does not adsorb on the bridged oxygen atom, may be attributed to the high repulsive energy between lone pair electrons of O and N atoms. Because of the partial negatively charged N(1) and N(3) in HN3 resonance structures (see the numbering of N atoms in Figure 7), similar results are obtained for the side-on adsorbate, sideon HN(N2)-Ti(a), with Eads ) 8 kcal/mol, whereas side-on HN(N2)-O(a) does not exist on the TiO2 surface. On the other hand, when N(1) connects with a surface Ti atom, N(3) becomes electron deficient and is able to form a chemical bond with a bridged O atom. As a result, this doubly bonded species, TiN2N(H)-O(a), has the most stable structure with Eads ) 36 kcal/ mol. The hydrogen bonded species, Ti-N3H‚‚‚O(a), which is similar to the end-on HN3-Ti(a) except the additional hydrogen bonding with a bridged O atom, is more stable by 3 kcal/mol. The N2(a) species was predicted to have a low N2-Ti(a) binding energy, 2 kcal/mol, whereas the N2-O(a) adsorbate does not exist. Therefore, N2 desorbs readily and is hard to undergo dissociative adsorption on the TiO2 surface. This is consistent with the absence of the N2(a) stretching vibration in the IR spectra.

The radical adsorbates, N3(a), HN(a) and H(a), all have more stable structures bonding with a bridged O atom forming N3O(a), HN-O(a) and H-O(a) than to bond with a surface Ti atom forming N3-Ti(a), HN-Ti(a) and H-Ti(a). This can be attributed to the better overlaps between the p orbitals of the O atom and the available p orbitals of the N atoms and the s orbitals of the H atom, forming stronger N-O and H-O bonds than N-Ti and H-Ti bonds. b. Adsorbate Effect and Hydrogen Bond Interaction. To calculate the adsorbate effect, several selected adsorbed species with coverage ) 1 are calculated and compared with the adsorptions with coverage ) 0.5 (Table 2 and Figure 7). The optimized structures presented with the partial surface model and the related adsorption energies with coverage ) 1 are shown in Figure 8. For molecular adsorptions, N3-Ti(a), N3-O(a) and N2-Ti(a), the adsorption energies with coverage ) 1 are about twice of those with coverage ) 0.5. This result shows that negligible interactions between the adsorbates and can be understood from the large separation, >3.0 Å, between the two active sites on the surface. The dissociative adsorptions forming N3-Ti(a) + H-O(a) and N2-Ti + HN-O(a) also show the same results,


Wang et al.

TABLE 3: Vibrational Frequencies Calculated at the B3LYP/6-311G(d,p) Level by Gaussian 03 with a Scaling Factor58 end-on HN3-Ti(a) 2 x end-on HN3-Ti(a) side-on HN(N2)-Ti(a) Ti-N2N(H)-O(a) Ti-N3H‚‚‚O(a) N3-Ti(a) 2 × N3-Ti(a) N3-O(a) 2 × N3-O(a) N2-Ti(a) HN-O(a) N2-Ti(a) + HN-O(a) Ti-N2‚‚‚HN-O(a) a

Va(NNN)

Vs(NH)

Va(NNN) of exp

2170 2171, 2161 2180 2130 2175 2040 2040, 2040 2100 2100, 2100 2205 (w)a

3315 3308, 3217 3326 3301 3189

2170 2170 2170 2118 2170 2075 2075 2118 2118

2203 (w) 2204 (w)

3435 3438 3290

w ) weak IR intensity.

in which no interactions occur between N3(a) and H(a) nor N2(a) and NH(a). These are confirmed by their adsorption energies: Eads ) 13 kcal/mol for N3-Ti(a) + H-O(a) is close to Eads(N3-Ti(a)) + Eads(H-O(a)) - E(HN3 f N3 + H), 14 kcal/ mol and Eads ) 18 kcal/mol for N2-Ti(a) + HN-O(a) is similar to Eads(N2-Ti(a)) + Eads(HN-O(a)) - E(HN3 f N2 + NH), 17 kcal/mol. On the other hand, the adsorption energy of the end-on HN3Ti(a) with coverage ) 1 is slightly larger than twice of that with coverage ) 0.5. This difference, 2 kcal/mol, is contributed from hydrogen bonding between the two HN3(a) on the surface. The same results are also found in the dissociated products such as Ti-N2‚‚‚HN-O(a). As expected, hydrogen bonding has a strong influence on the vibrational frequencies observed in the IR spectra in the 3000-3800 cm-1 region experimentally and supported by the results of frequency calculations below. c. Frequency Calculations. To reduce the computational time, two layers of TiO2 (8 [TiO2] units) were employed for frequency calculations; this has been proven to be sufficient for simulating the force field of the TiO2 surface. The frequencies of HN3(a) computed have ∼0.5% difference by using two, three and four layers of TiO2. Frequencies of the strongest absorption, Va(NNN), and of the hydrogen stretching vibration, Vs(NH), were calculated at the B3LYP/6-311G(d,p) with the scaling factor,58 0.96, as shown in Table 3. First, to check the effect of hydrogen bond interactions, the differences in vibration frequencies between 1 × end-on HN3Ti(a) and 2 × end-on HN3-Ti(a) are compared. A small shift is noted in Va(NNN), 2170 f 2171 and 2161 cm-1 but about a 100-cm-1 shift is seen in Vs(NH), 3315 f 3308 and 3217 cm-1. Similar results are also found in the vibrations in Ti-N3H‚‚‚ O(a) and Ti-N2‚‚‚HN-O(a). Therefore, hydrogen bond interactions show a strong influence on Vs(NH), but not on Va(NNN). This result is consistent with the experimental observation, in which a broad band absorption shows in the range of 30003800 cm-1 and three sharp peaks related to Va(NNN) vibrations appear in the range of 2000-2200 cm-1. Furthermore, these three strong peaks appear in the 20002200 cm-1 range can be resolved from the computational result. As described in the experimental result above, the Va(NNN) of the end-on HN3-Ti(a) and side-on HN(N2)-Ti(a) with higher frequencies and that of the N3-Ti(a) with a lower frequency can be expected. These results are consistent with the N1-N2 bond lengths in Table 2, related to the π bond delocalization as alluded to above. The calculated frequencies of the two adsorbates, Ti-N2N(H)-O(a) and N3-O(a), are in between. It is worth noting that N3-O(a), which may be readily transformed

from N3-Ti(a), has very different N(1)-N(2) and N(2)-N(3) bonds from those of N3-Ti(a). From the geometrical structures given in Table 2, the N3 geometry in N3-O(a) is more similar to that of the HN3(g) than that of the N3(g) radical, whereas the structure of N3 in N3-Ti(a) is closer to that of N3(g). Therefore, the frequency of Va(NNN) in N3-O(a) is predicted to be different from that of N3-Ti(a). Finally, comparing the computed results with experimental observations, the end-on HN3-Ti(a) and side-on HN(N2)-Ti(a) are related to the peak observed at 2170 cm-1. N3-O(a) and Ti-N2N(H)-O(a) are corresponding to the peak at 2118 cm-1. N3-Ti(a) is attributed to the peak at 2075 cm-1. These assignments will be further elucidated by the potential energy surface prediction described later. d. Potential Energy Surface of Surface Reactions. The potential energy diagram for the surface reactions, forming N3Ti(a) + H-O(a) and N2-Ti(a) + HN-O(a) with adsorption energies lower than the initial reactants, HN3(g) + TiO2(110), is presented in Figure 9 on the basis of the predicted transition states that connect their corresponding reactants and products. The transition states for dissociative adsorptions via end-on HN3-Ti(a) and side-on HN(N2)Ti(a) giving N3(a) + HO(a) and N2(a) + HN-O(a), respectively, were located by a series of elastic bands using the NEB method,62 connecting the reactants and products with 16 intermediate points. The hydrogen bonded intermediate, Ti-N3H‚‚‚O(a), can be formed via TS1 by a slight bending of the angle ∠N2N1Ti in the molecular adsorbate, end-on HN3-Ti(a). The hydrogen bonded intermediate can isomerize through a 11 kcal/mol barrier at TS2, in which the N3-H bond is partially broken and H-O bond is partially formed, to produce the dissociated adsorption, N3-Ti(a) + H-O(a). Alternatively, the end-on HN3-Ti(a) can cyclize to give the most stable intermediate Ti-N2N(H)-O(a) via TS3 with a small 1 kcal/mol barrier. The direct dissociation of Ti-N2N(H)-O(a) leads to N2-Ti(a) + HN-O(a) with an overall exothermicity of 18 kcal/mol from the reactants. No intrinsic transition state was found by an automatic optimization with NEB nor by manually scanning the N(2)-N(3) coordinate for the dissociation of Ti-N2N(H)-O(a). Starting from the other molecular adsorbate, side-on HN(N2)-Ti(a), N3-Ti(a) + H-O(a) can be formed via TS4 with a low barrier of 1 kcal/ mol with an exothermicity of 13 kcal/mol relative to the reactants. Thus, the formation of both product pairs N2-Ti(a) + HN-O(a) and N3-Ti(a) + H-O(a) can take place readily even at room temperature. The IR spectra obtained in the three series experiments can be rationalized by the results of frequency and potential energy surface calculations. In the HN3 dosage experiment, the peak detected at 2075 cm-1 due to N3-Ti(a) with the strongest intensity can be understood by the low predicted barrier at TS4 to form N3-Ti(a) + H-O(a). The peak at 2170 cm-1 attributable to HN3-Ti(a) with the weakest intensity is consistent with the lower stabilities of the end-on HN3-Ti(a) and side-on HN(N2)-Ti(a). In the UV irradiation experiment, the peak at 2118 cm-1 with a slightly higher decreasing rate than that at 2075 cm-1 may be attributed to the UV breaking of the N(2)-N(3) bond in the cyclic intermediate Ti-N2N(H)-O(a), similar to the most efficient photodissociation process of HN3(g) because of the weak HN-N2 bond,63,64 although Ti-N2N(H)-O(a) has a higher thermal stability than N3-Ti(a). Therefore, the most stable Ti-N2N(H)-O(a) species is naturally linked to the slowest decreasing rate of the peak at 2118 cm-1 in the surface annealing experiment.



Figure 9. Potential energy diagram of the HN3/TiO2(110) system.

5. Summary In this work, the adsorption and reaction of HN3 on TiO2 nanoparticle films have been experimentally studied by FTIR spectroscopy, monitoring the effects of dosage, UV irradiation and heating. Computationally, the adsorption energies and vibrational frequencies of possible adsorbates have been predicted by the first-principles calculations. Four stable forms of HN3 (a) adsobates have been predicted and some of them supported by the experiments: end-on HN3-Ti(a), side-on HN(N2)-Ti(a), the hydrogen bonded Ti-N3H‚‚‚O(a) and the doubly bonded cyclic Ti-N2N(H)-O(a). These adsorbates can be readily desorbed or dissociated by UV irradiation or heating, producing H-O(a), HN-Ti(a), N2-Ti(a), N3-Ti(a), N3-O(a) and HN-O(a). The cyclic Ti-N2N(H)-O(a) with the highest adsorption energy (36 kcal/mol) can survive to higher temperatures on the surface during the annealing experiment, as one would expect. On the basis of the results of frequency calculations, three Va(NNN) sharp peaks from deconvoluted FTIR spectra were assigned to the adsorptions by the end-on HN3-Ti(a) and the side-on HN(N2)-Ti(a) at 2170 cm-1, by Ti-N2N(H)-O(a) and possibly N3-O(a) at 2118 cm-1, and by N3-Ti(a) at 2075 cm-1. From the assignment, the changes in the peak intensities under different experimental conditions could also be rationalized by the computed PES, which suggests that the fast increasing peak at 2075 cm-1 in the HN3 dosage experiment can be attributed to the fast nonthermally activated dissociation process, HN3Ti(a) f N3-Ti(a) + H-O(a), the slowest decreasing peak at 2118 cm-1 in the surface annealing experiment is attributable to the most stable adsorbate, Ti-N2N(H)-O(a), and the fast decaying peak at 2170 cm-1 in all the experiments is most likely related to the weak adsorbates, end-on HN3-Ti(a) and side-on HN(N2)-Ti(a). Acknowledgment. J.H.W. is grateful for the support from the Graduate School of Emory University and M.C.L. acknowledges the support from the R. W. Woodruff Professorship at Emory University and from Taiwan’s National Science Council for a Distinguished Visiting Professorship at the Center for Interdisciplinary Molecular Science, National Chiao Tung University, Hsinchu, Taiwan. We are also grateful to Taiwan’s National Center for High-performance Computing for the extensive use of its computational facility. We particularly thank Dr. C. M. Wei. of Academia Sinica, Taiwan, for valuable

suggestions on the application of the VASP code and Dr. Tim Lian and his group at Emory University on the preparation of TiO2 nanoparticle films. References and Notes (1) Fujishima, A.; Honda, K. Nature 1972, 238, 37. (2) Graetzel, M. Nature 2001, 414, 338. (3) Asahi, R.; Morikawa, T.; Ohwaki, T.; Aoki, K.; Taga, Y. Science 2001, 293, 269. (4) Irie, H.; Watanabe, Y.; Hashimoto, K. J. Phys. Chem. B 2003, 107, 5483. (5) Burda, C.; Lou, Y.; Chen, X.; Samia, A. C. S.; Stout, J.; Gole, J. L. Nano Lett. 2003, 3, 1049. (6) Diwald, O.; Thompson, T. L.; Groralski, E. G.; Walck, S. D.; Yates, J. T. J. Phys. Chem. B 2004, 108, 52. (7) Wang, J.-H.; Lin, M. C. ChemPhysChem 2004, 5, 1615. (8) Nazeeruddin, M. K.; Pechy, P.; Graetzel, M. Chem. Commun. 1997, 18, 1705. (9) Bu, Y.; Lin, M. C.; Fu, L. P.; Chtchekine, D. G.; Gilliland, G. D.; Chen, Y.; Ralph, E. E.; Stock, S. R. Appl. Phys. Lett. 1995, 66, 2433. (10) Bu, Y.; Ma, L.; Lin, M. C. J. Vac. Sci. Technol., A: Vac., Surf. Films 1993, 11, 2931. (11) Bu, Y.; Lin, M. C. Surf. Sci. 1994, 317, 152. (12) Chu, J. C. S.; Bu, Y.; Lin, M. C. Mater. Lett. 1992, 14, 207. (13) Chu, J. C. S.; Bu, Y.; Lin, M. C. Surf. Sci. 1993, 284, 281. (14) Bu, Y.; Lin, M. C. Surf. Sci. 1994, 301, 118. (15) Tindall, C.; Hemminger, J. C. Surf. Sci. 1995, 330, 67. (16) Thoms, B. D.; Russell, J. N., Jr. Surf. Sci. 1995, 337, L807. (17) Carlo, S. R.; Torres, J.; Fairbrother, D. H. J. Phys. Chem. B 2001, 105, 6148. (18) Russell, J. N., Jr.; Bermudez, V. M.; Leming, A. Langmuir 1996, 12, 6492. (19) Heidberg, J.; Hustedt, M.; Oppermann, J.; Paszkiewicz, P. Surf. Sci. 1996, 352-354, 447. (20) Bacalzo-Gladden, F. Ph.D. Dessertation, Emory University, 1999. (21) Lu, X.; Lin, M. C. Int. ReV. Phys. Chem. 2002, 21, 137. (22) Lu, X.; Su, Z.; Xu, X.; Wang, N.; Zhang, Q. Chem. Phys. Lett. 2003, 371, 172. (23) Lu, X.; Su, Z.; Xu, X.; Wang, N.; Zhang, Q. Abstr. Pap., 225th ACS Natl. Mtg. 2003, COLL. (24) Lu, X.; Fu, G.; Wang, N.; Zhang, Q.; Lin, M. C. Chem. Phys. Lett. 2001, 343, 212. (25) Bu, Y.; Chu, J. C. S.; Lin, M. C. Surf. Sci. 1992, 264, l151. (26) Bu, Y.; Lin, M. C. Surf. Sci. 1994, 311, 384. (27) Basu, P.; Ballinger, T. H.; Yates, J. T., Jr. ReV. Sci. Instrum. 1988, 59, 1321. (28) Rusu, C. N.; Yates, J. T., Jr. J. Phys. Chem. B 2000, 104, 12299. (29) Wu, W.-C.; Chuang, C.-C.; Lin, J.-L. J. Phys. Chem. B 2000, 104, 8719. (30) Liao, L.-F.; Wu, W.-C.; Chuang, C.-C.; Lin, J.-L. J. Phys. Chem. B 2001, 105, 5928. (31) Zaban, A.; Ferrere, S.; Sprague, J.; Gregg, B. A. J. Phys. Chem. B 1997, 101, 55. (32) Bu, Y.; Ma, L.; Lin, M. C. J. Vac. Sci. Technol. A 1993, 11, 2931. (33) Bu, Y.; Lin, M. C. Mater. Res. Soc. Symp. Proc. 1993, 335, 21.

5142 J. Phys. Chem. B, Vol. 109, No. 11, 2005 (34) Herman, G. S.; Dohnalek, Z.; Ruzycki, N.; Diebold, U. J. Phys. Chem. B 2003, 107, 2788. (35) Li, W.; Ni, C.; Lin, H. C.; P., H. C.; Shah, S. I. J. Appl. Phys. 2004, 96, 6663. (36) McCormick, J. R.; Zhao, B.; Rykov, S. A.; Wang, H.; Chen, G. J. Phys. Chem. B 2004, 108, 17398. (37) Wyckoff, R. W. G. Crystal Structure, 2nd ed.; Wiley: New York, 1964; Vol. 1. (38) Kresse, G.; Hafner, J. Phys. ReV. B 1993, 47, 558. (39) Kresse, G.; Hafner, J. Phys. ReV. B 1994, 49, 14251. (40) Kresse, G.; Furthmuller, J. Comput. Mater. Sci. 1996, 6, 15. (41) Kresse, G.; Furthmuller, J. Phys. ReV. B 1996, 54, 11169. (42) Cleperley, D. M.; Alder, B. J. Phys. ReV. Lett. 1980, 45, 566. (43) Perdew, J. P.; Yang, Y. Phys. ReV. B 1992, 45, 244. (44) Lee, C.; Yang, W.; Parr, R. G. Phys. ReV. B 1988, 37, 785. (45) Perdew, J. P.; Chevary, J. A.; Vosko, S. H.; Jackson, K. A.; Penderson, M. R.; Singh, D. J.; Fiolhais, C. Phys. ReV. B 1992, 46, 6671. (46) Monkhorst, H.; Pack, J. Phys. ReV. B 1976, 13, 5188. (47) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.; Robb, M. A.; Cheeseman, J. R.; Zakrzewski, V. G.; Montgomery, J. A., Jr.; Stratmann, R. E.; Burant, J. C.; Dapprich, S.; Millam, J. M.; Daniels, A. D.; Kudin, K. N.; Strain, M. C.; Farkas, O.; Tomasi, J.; Barone, V.; Cossi, M.; Cammi, R.; Mennucci, B.; Pomelli, C.; Adamo, C.; Clifford, S.; Ochterski, J.; Petersson, G. A.; Ayala, P. Y.; Cui, Q.; Morokuma, K.; Malick, D. K.; Rabuck, A. D.; Raghavachari, K.; Foresman, J. B.; Cioslowski, J.; Ortiz, J. V.; Stefanov, B. B.; Liu, G.; Liashenko, A.; Piskorz, P.; Komaromi, I.; Gomperts, R.; Martin, R. L.; Fox, D. J.; Keith, T.; Al-Laham, M. A.;

Wang et al. Peng, C. Y.; Nanayakkara, A.; Gonzalez, C.; Challacombe, M.; Gill, P. M. W.; Johnson, B.; Chen, W.; Wong, M. W.; Andres, J. L.; Gonzalez, C.; Head-Gordon, M.; Replogle, E. S.; Pople, J. A. Gaussian03; Gaussian, Inc.: Pittsburgh, 2003. (48) Becke, A. D. J. Chem. Phys. 1993, 98, 5648. (49) Lee, C.; Yang, W.; Parr, R. G. Phys. ReV. B 1989, 37, 785. (50) Dunning, J., T. H.; Hay, P. J. In Modern Theoretical Chemistry; Plenum: New York, 1977. (51) Rasmussen, M. D.; Molina, L. M.; Hammer, B. J. Chem. Phys. 2004, 120, 988. (52) Rego, L. G. C.; Batista, V. S. J. Am. Chem. Soc. 2003, 125, 7989. (53) Menetrey, M.; Markovits, A.; Minot, C. Surf. Sci. 2003, 524, 49. (54) Sorescu, D. C.; Rusu, C. N.; Yates, J. T., Jr. J. Phys. Chem. B 2000, 104, 4408. (55) Lindan, P. J. D.; Harrison, N. M. Phys. ReV. Lett. 1998, 80, 762. (56) Goniakowski, J.; Gillan, M. J. Surf. Sci. 1996, 350, 145. (57) Harris, L. A.; Quong, A. A. Phys. ReV. Lett. 2004, 93, 086105. (58) Ostmark, H.; Launila, O.; Wallin, S.; Tyrman, R. J. Raman Spectrosc. 2001, 32, 195. (59) Winnewisser, M.; Cook, R. L. J. Chem. Phys. 1964, 41, 999. (60) Tian, R.; Facelli, J. C.; Michl, J. J. Phys. Chem. 1988, 92, 4073. (61) Shimanouchi, T. J. Phys. Chem. Ref. Data 1977, 6, 993. (62) Mills, G.; Jonsson, H.; Schenter, G. K. Surf. Sci. 1995, 324, 305. (63) Gericke, K.-H.; Lock, M.; Comes, F. J. Chem. Phys. Lett. 1991, 186, 427. (64) Gericke, K.-H.; Hass, T.; Lock, M.; Theinl, R.; Comes, F. J. J. Phys. Chem. 1991, 95, 6104.

Reactions of Hydrazoic Acid on TiO2 Nanoparticles: an Experimental

Recommend Documents