18236
J. Phys. Chem. C 2007, 111, 18236-18242
Alcohol Chemistry on Rutile TiO2(110): The Influence of Alkyl Substituents on Reactivity and Selectivity Yu Kwon Kim,‡ Bruce D. Kay,*,§ J. M. White,†,‡,§ and Z. Dohna´ lek*,§ Department of Chemistry and Biochemistry, Center for Materials Chemistry, UniVersity of Texas at Austin, Austin, Texas 78712, and Fundamental Sciences Directorate and Institute for Interfacial Catalysis, Pacific Northwest National Laboratory, Richland, Washington 99352 ReceiVed: July 17, 2007; In Final Form: September 13, 2007
Product yields and selectivities, based on ultrahigh vacuum temperature-programmed desorption, are compared for 10 C2 to C8 aliphatic alcohols dosed at 100 K on highly ordered TiO2(110) with a 3.5% concentration of surface oxygen vacancies. Dehydration to form an alkene and water typically dominates, while two other channels, dehydrogenation to form an aldehyde and re-formation of alcohol, make detectable contributions for primary alcohols. Depending on the alcohol, there are two distinct dehydration pathways: one operative at low temperature (LT, 300-425 K) and the other at high temperature (HT, 480-650 K). The HT dehydration pathway is common to all, while the LT channel is not observed for tertiary butanol and 3- and 4-octanol. The observed trends are accounted for in terms of the inductive and steric effects of the alkyl substituents.
1. Introduction Many reactions on solid surfaces (e.g., oxides) occur on defect sites. On prototypical rutile TiO2(110) surfaces, there is direct scanning tunneling microscopy (STM) evidence that bridgebonded oxygen vacancies (BBOV’s) are sites for dissociation of oxygen,1-4 water,1,5-8 and alcohols.9,10 In the surface science and catalysis literature involving alcohol chemistry on TiO2, dehydration to alkenes above 500 K is well-established11-14 while dehydrogenation to aldehydes or ketones and re-formation of the dosed alcohols are less well-established. The chemical pathways leading to these products are typically described in terms of paths requiring BBOV’s. However, the inability to control surface order and BBOV concentration may account for reported variations in temperature-programmed desorption (TPD) profiles of alcohol dehydration12,13 that make it difficult to establish unambiguous connections between catalytic dehydration activity and the local surface structure of TiO2(110).12,13,15,16 As part of a research program to establish direct atomic level descriptions by combining STM and reaction rate measurements for oxygenates on oxides, we have recently reported on the ensemble average surface chemistry of three different isotopically labeled 2-propanols dosed at 100 K on TiO2(110).14 Three surface conditions were compared: 3% BBOV, 7% BBOV, and Ne+-sputtered. In TPD, there was no evidence for dehydrogenation; only 2-propanol, propene, and water were observed. For the first time, a low-temperature (LT) dehydrogenation path (300-450 K) was found to accompany the well-established high-temperature (HT) path (T > 450 K). Illustrating the importance of controlling the details of the surface, the intensity of the LT path was higher for 7% than that for 3% BBOV but was completely suppressed for the Ne+ bombarded surface. The HT path was interpreted in terms of decomposition of 2-propoxy * Corresponding authors. E-mails:
[email protected]; Bruce.
[email protected]. † Deceased August 31, 2007. ‡ University of Texas at Austin. § Pacific Northwest National Laboratory.
species located on BBO rows, while the LT path was attributed to a reaction between a pair of C3 species bound to Ti4+ rows. Both paths exhibit isotope effects; the HT path favors the hydrogenated isomer and the LT path favors the deuterated isomer. The proposed LT path differs from typical models5,17-22 in requiring nondefective Ti4+ rows exceeding some minimum length. While surface structure is one key, variations with the alcohol are also expected and have been reported; while dehydration typically dominates, dehydrogenation contributes more for primary alcohols.11,12 The present study was motivated by a desire to determine the detailed roles played by the alkyl substituents on the accessibility of the LT and HT channels and on the dehydration selectivity. The inductive and steric effects of alkyl substituents are the concepts commonly used to understand the trends in the reactivity of similar molecules in homogeneous organic chemistry (e.g., in liquids).23 Here, we show that similar concepts can be successfully applied to understand the heterogeneous reactions at gas-solid interfaces.24,25 In a communication,24 we recently reported correlation of the HT activation energy for dehydrogenation with tabulated values of the Taft induction parameter, σ*,23 suggesting that the HT reaction transition state lying between the reactant, adsorbed alkoxide, and the gas-phase product, alkene, was a structure involving concerted elongation of the C-O bond of the alkoxide and a C-H bond on the β-carbon of one of its alkyl groups. In this article, we report the details of the LT and HT reaction paths for primary, secondary, and tertiary alcohols for 100 K doses on well-ordered TiO2(110) with a reproducible, relatively small BBOV concentration (3.5 ( 0.5%). In one case, we blocked all the vacancies before dosing 2-propanol. To ensure reproducibility of the surface order and the BBOV concentration, we relied, as described in the Experimental Section, on H2O TPD spectra (e.g., Henderson26,27) that exhibit an HT peak that quantitatively determines the BBOV concentration and a reproducibly shaped monolayer molecular H2O desorption profile that reflects the order.
10.1021/jp075608+ CCC: $37.00 © 2007 American Chemical Society Published on Web 11/08/2007
Alcohol Chemistry on TiO2(100)
J. Phys. Chem. C, Vol. 111, No. 49, 2007 18237
The peak temperature of the HT dehydration channel trends downward in passing from primary to secondary to tertiary alcohols and to a lesser extent also downward as the length of the alkyl chain increases among primary or secondary alcohols. These trends are interpreted in terms of stabilization of the transition state of the alkoxy bound to BBOV through the inductiVe effect of alkyl substituents. While the peak temperature of the LT channel does not vary significantly with the alcohol, the relative intensity tends to drop as the chain length increases and is undetectable for tertiary alcohols. These LT variations are interpreted in terms of steric effects in a quasi-bimolecular reaction of two alcohol-derived species that migrate along Ti4+ rows (i.e., adsorbate-adsorbate interactions among alkoxy and/ or alcohol groups on Ti4+ rows are required to access the LT dehydration channel). 2. Experimental Section A rutile TiO2(110) crystal (10 × 10 × 1 mm3, Princeton Scientific) bonded with ceramic adhesive (Aremco, 503) to a Ta plate was introduced into a previously described UHV molecular beam-surface scattering apparatus28 with a base pressure of 5 × 10-11 Torr. The substrate was heatable to 1000 K and coolable to 60 K. After repeated cycles of Ne+-sputtering (1.5 kV, 10 µA) at 300 K and annealing at 850-900 K, the surface gave a sharp 1 × 1 low-energy electron diffraction (LEED) pattern and Auger electron spectroscopy (AES) indicated no detectable impurities. As noted above, H2O TPD spectra were used to indicate surface order and to quantitatively assess the BBOV concentration.26,27 The surface BBOV concentration was 3.5 ( 0.5% throughout, except for one case where we blocked the vacancies with OH groups before dosing 2-propanol.1,5-8 The error bar of the BBOV concentration is obtained from the variations of the 510 ( 10 K H2O TPD peak area26,29 in the H2O TPDs routinely performed before each experimental series with the alcohols. For a 1 ML 100 K dose of H2O, the strong molecular desorption from Ti4+ rows was reproducible and the rising and falling edges were characteristic of a well-ordered, stable surface.26,27 These TPD characteristics may be more sensitive than LEED in assessing the ensemble average local structure relevant for most chemical reactions.30,31 Alcohols were purified with freeze-pump-thaw cycles and dosed using an effusive gas-phase molecular beam source that directs a uniform 1.3 cm diam flux onto the TiO2 crystal and completely covers the surface. After dosing alcohols at 100 K, TPD spectra with a ramp rate of 1.8 K/s to 700 K were taken in line-of-sight geometry with a UTI mass spectrometer. Numerous ion fragments were tracked in exploratory TPD experiments to search for potential reaction products. Additionally, alcohol cracking patterns were determined in the absence of dissociation using their multilayer desorption peaks. This analysis allowed for unambiguous assignments of all contributing desorbing species as reported in the Results. Only alcohol, alkene, water, and aldehyde were detectable. There was no mass spectrometric evidence for desorption of molecular hydrogen. Reflecting a clean, stable surface after TPD to 700 K, no carbon was detected in AES, the BBOV concentration remained at 3.5 ( 0.5% based on H2O TPD, and repeated alcohol doseTPD cycles altered neither the TPD profile nor the dehydration yield. 3. Results To frame these results, we begin with an overall model that captures the global picture, but not the details of the surface chemistry probed here. Alcohols, ROD (or ROH), dosed on
Figure 1. TPD profiles of dehydration products from 0.9 ML monodeuterated 2-propanol, (CH3)2CHOD, dosed on TiO2(110) at 100 K. Contributions from fragmentation of (CH3)2CHOD (traced by 46 amu) have been subtracted from the 18, 19, 20, 41, and 45 amu profiles to obtain the net desorption profiles of H2O, HDO, D2O, C3H6, and (CH3)2CHOH, respectively. Shaded regions correspond to the desorption of reactants and products from the TiO2(110). The unshaded regions in the HDO and H2O spectra are affected by experimental artifacts and therefore cannot be unambiguously assigned (see the text for discussion). The LT and HT regions are identified.
TiO2(110) below room temperature adsorb dissociatively on BBOV’s and both dissociatively and molecularly on Ti5c,9,10 where Ti5c denotes five-coordinate Ti4+. In terms of reaction equations, using Obb and Ov to designate BBO and BBOV, the dissociative adsorption on BBOV’s can be written as:
ROD + Ov + Obb f RObb + DObb
(1)
The molecular and dissociative channels on Ti4+ (Ti5c) can be written as:
ROD + Ti5c f ROTiD
(2a)
ROTiD f ROTi + DObb
(2b)
respectively. Here ROD is the dosed alcohol, and ROTiD is that alcohol functioning as a Lewis base (electron donor) to Ti5c, a Lewis acid (electron acceptor). Deprotonation (eq 2b) involves oriented and mobile ROTiD and occurs in competition with LT alcohol desorption. The proton attaches to a neighboring BBO and can be mobilized when assisted by a mobile alcohol.9,10 The other product, ROTi, increases the coordination of the Ti from 5 to 6, and is likely immobile. For doses exceeding saturation of the Ov and Ti5c sites, alcohols fill first the Obb sites and then multilayer films where the added alcohol is not in contact with the surface.14 In the experiments reported here, doses were all done at 100 K and did not exceed one monolayer (ML), the latter defined as the dose required to saturate the Ov and Ti5c sites. Using TPD, we track the reactions of RObb, DObb, ROTi, and ROTiD as the temperature increases. As a typical example, Figure 1 shows all detectable decomposition products obtained from TPD of 0.9 ML of monodeuterated 2-propanol, (CH3)2CHOD. The products are limited to
18238 J. Phys. Chem. C, Vol. 111, No. 49, 2007 2-propanol, propene, and water, and consistent with our prior report,14 there are two distinct desorption regions: LT from 300 to 425 K and HT from 450 to 650 K. There is no evidence for other products (e.g., no molecular hydrogen, acetone, or other hydrocarbons), and after TPD to 700 K, the AES measurements show that the surface is free of carbon. In the LT region, the 18, 19, and 20 amu TPD signals rise before any C-containing signals, but the detailed analysis in this region is not possible since water from the chamber background, from impurity in the alcohol source, and from isotope exchange processes in the QCM ionizer cannot be quantitatively disentangled from water formed by reactions of the dosed alcohol.14 Qualitatively, there is marginal evidence, gray area in the 20 amu profile between 300 and 400 K, for some LT D2O desorption exceeding that attributed to an impurity in the source. Otherwise, the LT 18, 19, and 20 amu signals are ambiguous and not necessarily attributable to reactions of (CH3)2CHOD as already discussed in detail in our prior study of 2-propanol.14 Beginning at 250 K, the 46 amu signal, tracking (CH3)2CHOD desorption, begins to rise, and the rate maximizes in the LT zone at 310 K before dropping sharply and becoming negligible for any T > 425 K (i.e., there is no HT (CH3)2CHOD desorption). In agreement with prior studies,11-14 we assign the LT (CH3)2CHOD desorption to the reverse of eqs 2a and 2b (i.e., desorption of 2-propanol and/or 2-propoxy chemisorbed on Ti4+ rows, respectively). Typical of all the alcohols we studied, 90 ( 5% of 1 ML desorbs in the LT alcohol peak. Two other C3 products, CH3CHdCH2 (propene, measured at 41 amu) and (CH3)2CHOH (measured at 45 amu), account for the remaining 10%. Both exhibit local rate maxima at the same temperatures in both the LT and HT regions. In the HT region, water also desorbs. In the LT region, as the intense (CH3)2CHOD desorption (46 amu) begins to decay, the 45 amu intensity rises and then falls. This is neither an impurity nor a fragment of (CH3)2CHOD; rather, it is desorption of (CH3)2CHOH where the H comes from a C-H bond of a neighboring species (discussed later). The potential effect of background H2O was investigated by purposely dosing a few percent H2O with the alcohol. This surplus H2O does enhance the overall (CH3)2CHOH desorption rate proportional to the intense (CH3)2CHOD desorption (46 amu) profile because of random scrambling of hydroxyl hydrogen from H2O with deuterium from (CH3)2CHOD, but does not change the difference spectral shape between 45 and 46 amu, which is the net (CH3)2CHOH desorption as shown in Figure 1. Neither the LT nor the HT (CH3)2CHOH yield is altered as already discussed in our previous study.14 In the HT region, the propene and 2-propanol rates maximize at 575 K, 60 K higher than the first water peak at 515 K that is dominated by D2O, and 5 K lower than the second water peak that is dominated by H2O. The D2O desorption is a result of recombination of two DObb’s, that is, 2 DObb f D2O + Ov, a reaction that has been characterized in considerable detail in TPD of water dosed on TiO2(110).18,26,29,32 This reaction consumes the D involved in the dissociation of (CH3)2CHOD, and thus, negligible D2O appears at 580 K. The H2O peak at 580 K in Figure 1 is a result of thermally activated C-H bond breaking in RObb, that is, the reaction, (CH3)2CHObb f CH3CHdCH2(g) + HObb. The CH3CHdCH2 species desorbs, and the mobile HObb either reacts with the remaining (CH3)2CHObb, that is, (CH3)2CHObb + HObb f (CH3)2CHOH(g) + Ov, to form and desorb (CH3)2CHOH, or reacts with another HObb, that is, HObb + HObbf H2O(g) + Ov, to form H2O and Ov. The
Kim et al.
Figure 2. TPD spectra of alkenes from 10 alcohols dosed on TiO2(110) at 100 K grouped as primary, secondary, and tertiary. The net alkene desorption spectra (shown) were obtained by subtracting the fragmentation contributions of molecular alcohol. The 27 and 41 amu were used for ethylene and propene, respectively, and 56 amu was used for butenes and octenes. The alcohol dose was set to 1 ML, and all spectra were normalized to the HT alkene desorption peaks.
reaction to re-form (CH3)2CHOH accounts for the D-free 2-propanol desorption at HT (compare 45 and 46 amu). With the above picture in mind, the alkene TPD spectra for 1 ML of 10 alcohols were measured and are presented in Figure 2. In the LT region, seven of the 10 (ethanol, 1- and 2-propanol, 1- and 2-butanol, and 1- and 2-octanol) exhibit intensity, whereas the other three (3- and 4-octanol, and t-butanol) do not. When the LT channel is present, the peak temperature shows no systematic variation with alcohol structure. On the other hand, the HT channel is observed for all 10 alcohols and, as described in our earlier communication,24 varies systematically with two properties of the dosed alcohol. First, the HT peak positions group together; primary alcohols are systematically higher by 50-60 K than the secondary alcohols which are, in turn, higher by another 50-60 K than the tertiary alcohol. Within each grouping, the production rate maximum drops 10-20 K as the chain lengths increase. The HT shifts are discussed elsewhere24 in terms of inductive electronic effects of the alkyl groups;25,33-35 the latter are described empirically using the Taft parameters of organic reaction chemistry.23 The variations in LT alkene production confirm effects related to the number of carbon atoms in the alkyl chains, R1, R2, and R3, of the alcohol, denoted as R1R2R3CR-OH. For primary alcohols, R1 ) R2 ) H, and the low-temperature channel is active for all of R3 chain lengths studied (up to C8). For t-butanol, R1 ) R2 ) R3 ) CH3, there is no evidence for LT alkene production. For secondary alcohols, R1 ) H, and R2 and R3 vary from CH3 to C6H13. If R2 ) CH3, then LT alkene production is observed, but if neither R2 nor R3 ) CH3, then LT alkenes are not observed. Summarizing, LT alkene desorption is observed when one of the R groups is H and when one of the other R groups is either H or CH3. Otherwise, the LT channel is undetectable. To assess the role of BBOV’s in the LT alkene channel, we compare in Figure 3 (upper panel) propene produced from 1
Alcohol Chemistry on TiO2(100)
J. Phys. Chem. C, Vol. 111, No. 49, 2007 18239
Figure 4. Alcohol TPD after dosing 1 ML of 1- and 2-propanol on TiO2(110) at 100 K. The HT intensity for 1-propanol is 2% of the LT intensity and 4× that of HT alcohol desorption from 2-propanol.
Figure 3. LT propene desorption spectra from 2-propanol dosed on TiO2(110) with and without BBOV’s (upper portion). The TiO2(110) surface without BBOV’s was prepared by predosing H2O at 100 K followed by a rapid flash to 450 K and recooling to 100 K before doing the (CH3)2CHOH. The HT, but not the LT, intensity is lowered by filling the BBOV’s with OH derived from H2O. The LT propene yield varies with (CH3)2CHOH dose as shown in the lower plot for the two surfaces.
ML of 2-propanol dosed on surfaces with two different BBOV concentrations, 0.0 and 3.5%. The former was realized by filling the surface vacancies with OH groups formed by dosing 1 ML of H2O and annealing to 450 K. This creates a surface with the BBOV’s filled with HObb and no chemisorbed H2O. Strikingly, filling the BBOV’s with HObb has no effect on the LT propene TPD profiles. The LT profiles are indistinguishable, showing that BBOV’s are not a prerequisite for the LT reaction path. As depicted in the lower panel of Figure 3, the LT propene yield rises superlinearly with coverage but filling the vacancies with HObb has no effect. This fact, in turn, indicates that the alkoxides (2-propoxides) bound to BBOV’s are not related to the LT propene formation. Also, the alcohols (2-propanol) adsorbed on BBOV’s desorb at
