Article pubs.acs.org/JPCB
Adsorption, Desorption, and Displacement Kinetics of H2O and CO2 on TiO2(110) R. Scott Smith,* Zhenjun Li, Long Chen, Zdenek Dohnálek, and Bruce D. Kay* Fundamental and Computational Sciences Directorate and Institute for Integrated Catalysis, Pacific Northwest National Laboratory, Richland, Washington 99352, United States S Supporting Information *
ABSTRACT: The adsorption, desorption, and displacement kinetics of H2O and CO2 on TiO2(110) are investigated using temperature programmed desorption (TPD) and molecular beam techniques. The TPD spectra for both H2O and CO2 have well-resolved peaks corresponding to desorption from bridge-bonded oxygen (Ob), Ti5c, and defect sites in order of increasing peak temperature. Analysis of the saturated surface spectrum for both species reveals that the corresponding adsorption energies on all sites are greater for H2O than for CO2. Sequential dosing of H2O and CO2 reveals that, independent of the dose order, H2O molecules will displace CO2 in order to occupy the highest energy binding sites available. Isothermal experiments show that the displacement of CO2 by H2O occurs between 75 and 80 K.
I. INTRODUCTION
II. EXPERIMENTAL SECTION All experiments were conducted utilizing an ultrahigh vacuum system (UHV) with a base pressure of 18 M Ωcm), were deposited using separate effusive molecular beam sources impinging onto the TiO2(110) at normal incidence. The backing pressure for both species was 1 Torr and the fluxes were 0.22 ML/s for CO2 and 0.252 ML/s for H2O. One “ML” was defined as the coverage needed to saturate the surface Ti sites (including the 4% of VO’s) which is ∼5.4 × 1014 sites/cm2. The definition was the same for both CO2 and H2O, i.e., 1 ML of either species contains the same number of molecules. This definition translates to an absolute flux of 1.19 × 1014 molecules/cm2·s for CO2 and 1.36 × 1014 molecules/cm2·s for H2O. All TPD spectra were acquired using a quadrupole mass spectrometer (UTI 100C) in a line-of-sight geometry and at a constant heating rate of 1 K/s.
III. RESULTS AND DISCUSSION A. Desorption Kinetics of CO 2 and H 2 O from TiO2(110). Figure 1 displays TPD spectra for CO2 and H2O deposited at 50 K on TiO2(110) and heated at a rate of 1 K/s. The top panel, Figure 1a, contains CO2 desorption spectra with coverages from 0.06 to 2.9 ML. The spectra fill from high to low temperature meaning that CO2 molecules occupy the highest binding energy sites first. There is a small peak at ∼163 K which is related to CO2 desorption from/near bridge-bonded oxygen vacancies (VO),9 followed by peaks at 131 K, 115 K, and 84 K. The peak at 131 K is due to CO2 desorption from Ti5c surface sites (blue curves), the peak at 115 K is from bridgebonded oxygen (Ob) surface sites (red curves), and the peak at 84 K is due to desorption of multilayer CO2 (black curves). These peak assignments are consistent with previous work.6,8,9 The combined VO, Ti5c, and Ob sites are saturated with CO2 at a coverage of ∼1.76 ML. Figure 1b contains H2O desorption spectra with coverages from 0.13 to 2.3 ML. As was seen with CO2, the spectra fill from high to low temperature meaning that H2O adsorbates occupy the highest binding energy sites first. For H2O, the peak for recombinative desorption from bridging hydroxl species is at ∼500 K and is therefore offscale in Figure 1b. The peaks for H2O desorption from Ti5c (blue curves) and Ob (red curves) sites are at 265 and 177 K, respectively.4,5,7,10 Multilayer desorption (black curve) is not well-resolved from the Ob desorption peak and manifests itself as a low-temperature shoulder on the Ob peak. For H2O, the combined VO, Ti5c, and Ob sites are saturated at a coverage of ∼2.0 ML.12 The reason for the lower number of CO2 than H2O needed to saturate the TiO2(110) surface (1.76 CO2 vs 2.0 H2O) has been explored in-depth in a combined experimental (scanning tunneling microscopy, infrared reflection adsorption spectroscopy, molecular beam scattering, TPD) and theoretical (density functional theory, ab initio molecular dynamics) study.9 Briefly, on an ideal TiO2(110) surface, there are an equal number of Ti5c and Ob sites (5.2 × 1014 sites/cm2 for each), and the same number of CO2 and H2O molecules is needed to saturate the Ti5c sites (defined as 1 ML; see Figure 1). For H2O, an additional 1 ML can be adsorbed on the Ob sites, whereas for CO2 only an additional 0.76 ML is enough to saturate the Ob sites. The explanation for this is that CO2 molecules adsorbed on Ti5c sites tilt toward nearby Ob sites, alternating the tilt
Figure 1. TPD spectra for CO2 and H2O deposited on TiO2(110) at 50 K and heated at 1 K/s. (a) TPD spectra for CO2 coverages of 0.06, 0.11, 0.17, 0.22, 0.33, 0.44, 0.66, 0.88, and 1.1 ML (blue curves); 1.3, 1.5, 1.76, and 2.0 ML (red curves); and 2.2, 2.4, and 2.6 ML (black curves). The peaks at 163, 131, and 115 K correspond to desorption from VO, Ti5c, and Ob sites, respectively. The peak at 85 K is due to multilayer desorption. (b) TPD spectra for H2O coverages of 0.13, 0.25, 0.38, 0.50, 0.75, and 1.0 ML (blue curves); 1.25, 1.5, 1.75, and 2.0 ML (red curves); and 2.3 ML (black curve). The peaks at 265 and 177 K correspond to desorption from Ti5c and Ob sites, respectively.
direction to form a zigzag pattern that effectively blocks about half of the Ob sites.9 In real experiments, slightly more than 0.5 ML of CO2 (0.76 ML in Figure 1) is observed to absorb, and this additional amount is likely due to compression effects described previously.13 The TPD spectra with surface saturation coverages for CO2 (1.76 ML) and H2O (2.0 ML) were analyzed to determine the site binding energies and distributions. The procedure used the TPD inversion method described in detail elsewhere.14−16 Briefly, the desorption rate is described by the Polanyi−Wigner equation, dθ/dt = νθn exp(−E/RT), where θ is the coverage, T is the temperature, E is the desorption activation energy, R is the gas constant, ν is the prefactor, and n is the desorption order. For the analysis, one assumes first-order desorption kinetics (n = 1) and a constant prefactor. The Polanyi−Wigner equation is rearranged to give the coverage dependent desorption energy, E(θ) = −RT ln((−dθ/dt)/vθ), and then solved using the saturated coverage spectrum for the desorption rate. In our analysis here, we perform the inversion analysis 8055
dx.doi.org/10.1021/jp501131v | J. Phys. Chem. B 2014, 118, 8054−8061
The Journal of Physical Chemistry B
Article
lines for v2Dgas and solid lines for vMax). The desorption energy versus coverage curves for both CO2 and H2O have two plateau regions corresponding to desorption from Ob (above ∼1 ML) and Ti5c (below 1 ML) sites. There is a step increase in energy for desorption from Ti5c compared to desorption from Ob sites. For both adsorbates, below ∼0.10 ML, there is a steep increase in energy corresponding to desorption from very higher energy sites such as VO’s. For CO2 (Figure 2a), the E(θ) curves calculated with the two extreme case prefactors (v2Dgas = 1.8 × 1013 s−1 and vMax = 4.6 × 1016 s−1) differ by about 25%. For H2O (Figure 2b), the E(θ) curves calculated with the two extreme case prefactors (v2Dgas = 2.8 × 1013 s−1 and vMax = 1.4 × 1016 s−1) differ by about 17%. In principle, one can simulate the subsaturation desorption spectra by numerically integrating the Polanyi−Wigner equation using the calculated E(θ) and the assumed prefactor. The prefactor could be used as a variational parameter to find the value that minimizes the error between the experimental and simulated spectra. However, the simulated spectra calculated with the E(θ) curves obtained with the two extreme case prefactors (v2Dgas and vMax in Figure 2) were not sufficiently distinct so as to make a determination of the “best” fit to the experiment within the uncertainty of the data. We have observed this situation previously for cases where the binding energy varies widely as a function of coverage.11 Therefore, the values in between the E(θ) curves in Figure 2 are our best estimates for the range of possible CO2 and H2O coverage dependent binding energies on TiO2(110). The simulated spectra for CO2 and H2O, obtained using the E(θ) curves and corresponding prefactors in Figure 2, are displayed in Figure S1 in the Supporting Information section. The probability distribution of binding site energies was obtained by differentiating the desorption energy versus coverage curve, i.e., P(E) = −dθ/dE. The distributions are shown as insets in Figure 2a for CO2 and Figure 2b for H2O using the E(θ) curves calculated with the vMax prefactors. Analogous distribution plots can be obtained using the E(θ) curves calculated with the v2DGas prefactor, but these would be qualitatively similar albeit shifted in energy by the amounts described above and are not shown for clarity. Both distributions have two peaks corresponding to desorption from Ti5c and Ob sites. The peak desorption energies for CO2 from Ti5c and Ob sites are at 44 and 39 kJ/mol, respectively. These values are considerably lower than the desorption energies for H2O, which are 89 and 59 kJ/mol for desorption from Ti5c and Ob sites respectively. That is, the binding energy of H2O, on both Ti5c and Ob sites, is greater than that of CO2 on either of the sites. B. Competitive Adsorption and Displacement Kinetics for CO2 and H2O on TiO2(110). In this section we explore the competitive adsorption and displacement kinetics between CO2 and H2O, where both species are deposited together in sequential doses at 50 K. Figure 3 displays a series of TPD spectra where 0.2 ML of CO2 was deposited after the preadsorption of various amounts of H2O (0 to 2.0 ML). The bottom trace, where CO 2 is deposited without preadsorbed water, has a peak near 144 K. With increasing amounts of predosed water, the CO2 desorption peak shifts to lower temperature eventually shifting to 103 K when 2.0 ML of H2O was predosed. The corresponding H2O TPD spectra (not shown) are unaffected by the presence of CO2, and the TPD set is similar to that shown in Figure 1b. The results here suggest that preabsorbed H2O molecules adsorb, find, and preferentially bind to the highest energy binding sites. This
using both the minimum and maximum plausible prefactors.17 The minimum prefactor, v2Dgas, is calculated assuming that the adsorbate has the translational and rotational degrees of freedom of a two-dimensional gas on the surface. This corresponds to the minimum entropy change going from the surface to the gas phase. The maximum prefactor, vMax, is calculated by considering the lower limit of the adsorbate partition function to be fully hindered in all translational and rotational modes, i.e., we take the partition function to be one. This corresponds to the maximum entropy change going from the surface to the gas phase. A constant prefactor for both cases was calculated at a temperature in the middle of the desorption range (140 K for CO2 and 210 K for H2O). Inversion using these two prefactors will yield the lowest (v2Dgas) and highest (vMax) possible binding energy curves. Figure 2 displays the results of the inversion analysis for CO2 (Figure 2a) and H2O (Figure 2b) using both prefactors (dashed
Figure 2. (a) The CO2 desorption binding energy versus coverage (E(θ)) curves obtained by inverting the 1.76 ML TPD spectrum in Figure 1(a) with a prefactors of vMax = 4.6 × 1016 s−1 (solid line) and v2DGas = 1.8 × 1013 s−1 (dashed line). Inset: The CO2 desorption energy probability distribution, P(E) = −dθ/dE obtained by differentiating the E(θ) curve obtained with the vMax prefactor. (b) The H2O desorption binding energy versus coverage (E(θ)) curves obtained by inverting the 2.0 ML TPD spectrum in Figure 1b with prefactors of vMax = 1.4 × 1016 s−1 (solid line) and v2DGas = 2.8 × 1013 s−1 (solid line). Inset: The H2O desorption energy probability distribution obtained by differentiating the E(θ) curve obtained with the vMax prefactor. 8056
dx.doi.org/10.1021/jp501131v | J. Phys. Chem. B 2014, 118, 8054−8061
The Journal of Physical Chemistry B
Article
Figure 3. A series of TPD spectra where 0.2 ML of CO2 was deposited after the preadsorption of various amounts of H2O. Both H2O and CO2 were deposited at 50 K, and the heating rate was 1 K/s. The H2O predoses were (from bottom to top) 0, 0.2, 0.4, 0.6, 0.8, 1.0, 1.2, 1.4, 1.6, 1.8, and 2.0 ML.
Figure 4. A series of TPD spectra where 1.7 ML of CO2 was deposited after the preadsorption of various amounts of H2O. Both H2O and CO2 were deposited at 50 K and the heating rate was 1 K/s. The H2O predoses were (from bottom to top) 0, 0.25, 0.50, 0.75, 1.0, 1.25, 1.50, 1.75, 2.0, and 2.25 ML. The vertical dashed lines at 131, and 115 K demark the peak temperatures observed in Figure 1a for CO2 desorption from Ti5c and Ob sites, respectively. The vertical line at 85 K is assigned to CO2 desorption from the second layer.
leaves only lower energy sites for subsequently dosed CO2 molecules, and this results in the CO2 peak shifting to lower temperature with increasing amounts of predosed H2O. The idea that H2O preferentially binds to the highest energy sites was further tested using a larger dose of CO2. Figure 4 displays a series of CO2 TPD spectra where 1.7 ML of CO2 is deposited after a predose of various amounts of H2O (0 to 2.25 ML). The bottom spectrum is for a 1.7 ML dose of CO2 without preadsorbed H2O. The spectrum has the expected desorption peaks at 131 and 115 K characteristic of desorption from Ti5c and Ob sites, respectively. Vertical dashed lines mark these peak positions. As H2O is added in 0.25 ML increments, the Ti5c desorption peak decreases in intensity and eventually disappears after the addition of 1 ML of H2O. Meanwhile, the Ob peak appears to shift to lower temperature with the addition of low doses of water (
