3555
J . Phys. Chem. 1993, 97, 3555-3564
Surface Reactions of Acrolein and Propionaldehyde on Cu20( 100): Nonselective Oxidation and Enolate-Mediated Side Reactions to CJ Products Kirk H. Scbulzt and David F. Cox' Department of Chemical Engineering, Virginia Polytechnic Institute and State University, Blacksburg, Virginia 24061 Received: March 24, 1992; I n Final Form: December 29, 1992
The adsorption of acrolein and propionaldehyde on the polar, Cu+-terminated Cu20( 100) surface was investigated using X-ray photoelectron spectroscopy (XPS) and thermal desorption spectroscopy (TDS).The major reaction products observed for both aldehydes were nonselective oxidation products (CO, C02, and H20). Propene and acrolein were also detected as minor reaction products during propionaldehyde thermal desorption. Additionally, propene and propionaldehyde were detected as minor products during the reaction of acrolein with preadsorbed hydrogen. The coincident desorption of the propionaldehyde, acrolein, and propene from either acrolein and preadsorbed hydrogen or propionaldehyde adsorption suggests that a common surface intermediate is involved in these minor side reactions. A reaction pathway involving a surface enolate intermediate (CHjCH=CHO-) is proposed to explain the side reactions to C3 products. The surface enolate intermediate is believed to be formed from propionaldehyde via a Lewis acid catalyzed reaction similar to Bronsted acid catalyzed keto-enol tautomerization in solution. A process similar to solution-phase (1,4) addition is postulated to give rise to the enolate intermediate during the reaction of preadsorbed hydrogen and a,B unsaturated acrolein. The enolatemediated reactions to C3 compounds generate only a minor amount of the total reaction products. As with the reaction of other oxygenates over Cu20( loo), the nonselective oxidation pathway involves deoxygenation of the adsorbate, subsequent dehydrogenation of hydrocarbon fragments, and the burn-off of surface carbon with lattice oxygen,
Introduction The partial oxidation of propene (CH2=CHCH3) to acrolein (CH2=CHCHO) is a useful model for theclassofallylicoxidation reactions of olefins for which Cu20 is the only reported singlecomponent oxide catalyst to exhibit significant activity and ~electivity.'-~The basic steps in the propene oxidation pathway have been widely studied, but the form of the oxygen-containing a-allyl surface species is not well understood.'-5 In this study, the reactivity of acrolein with Cu20 has been studied to provide information on the chemistry of the postulated u-allylspecies obtained if hydrogen abstraction from a a-allyl occurs prior to oxygen insertion.' A study of acrolein reactivity also gives information on products formed from subsequent reactions (e.g., nonselective oxidation) of the acrolein product. Also, since propionaldehyde (CH3CH2CHO) has been detected as a minor product in studiesof propene oxidation to acrolein on C u 2 0powder catalysts: the chemistry of both acrolein and propionaldehyde have been investigated on CuzO( 100). There is one reported study of the adsorption of acrolein and propionaldehyde on metal oxide single-crystal surfaces under ultrahigh-vacuum (uhv) condition^.^ Vohs and Barteau studied propionaldehyde and acrolein decomposition on ZnO(0001) polar surfaces7and demonstrated that the oxidation of aldehydes occurs by the nucleophilicattackof latticeoxygen at the carbonyl carbon. The resulting C j surface species decompose through parallel pathways by either alkyl or vinyl elimination to give surface formate or hydride elimination to the corresponding surface carboxylate species. CO, COz, and H2O were detected as reaction products from both acrolein and propionaldehydeover ZnO, while acrolein was also detected as reaction product from proionaldehyde. The acrolein formation from propionaldehyde was attributed to the unimolecular decompositionof propionate species by comparison with propionic acid results on the same ZnO ~urface.~ To whom correspondence should be addressed. Current address: Department of Chemical Engineering, University of North Dakota, Box 8101. University Station, Grand Forks, N D 58202. +
The C u 2 0 (100) surface used in this stud is a reconstructed, &+-terminated surface which displays a (3 2Xd2)R4So LEED pattern with many missing spots.8 No definitive model of the structure of the reconstructed surface is available. However, the periodicity of the reconstruction suggests a relaxation of top atomic layer Cu+ cations, possibly associated with a weak Cu+-Cu+ bonding interaction.8 No lattice oxygen is available in the top atomic layer of the ideal CuzO( 100) surface, and the top-layer Cu+ cations are singly coordinated to oxygen with one vacant coordination site. A more complete description of the characterization of the clean Cu20(100) surface has been reported previously.8
9
Experimental Section All experiments were performed in a dual-chamber, ultrahighvacuum system equipped with X-ray photoelectron spectroscopy (XPS), low-energy electron diffraction (LEED), and thermal desorption spectroscopy (TDS). Dosing was accomplished by back-filling through a variable leak valve. For all TDS experiments the sample was heated at a linear rate of 2 K/s. An Inficon Quadrex 200 was used to monitor up to six mass numbers simultaneously during the thermal desorption experiments. The mass spectrometer was equipped with a quartz skimmer to minimize the sampling of desorption products from the sample support hardware. All acrolein and propionaldehyde doses have been corrected for ion gauge s e n s i t i ~ i t y , and ~ ~ ' ~all TDS traces have been corrected for mass spectrometer sensitivity. The background pressure was less than 1.5 X 1O-Io Torr throughout the thermal desorption experiments. The Cu20 crystal was oriented by Laue back-reflection and mechanically polished to within f O . 5 O of the (100) surface. Sample dimensions were approximately 7 X 5 X 1 "3. The sample was mechanically attached to a tantalum holder which acted as an indirect heating and cooling source. A type K thermocouple was held in contact with the back of the sample through a hole in the sample holder with Aremco type 569 ceramic
0022-3654/93/2091-3555~~4.00/00 1993 American Chemical Society
Schulz and Cox
3556 The Journal of Physical Chemistry, Vol. 97, No. 14, 1993
I
In
n
0.4 L Propionaldehyde
CH3CH $HO CH2=CHCHO CHCHO ((x30) ~30)
-El
Cl+ =CHCH3 ( ~ 3 0 )
P
w
i?
L-----, 100
.-
200
300
400
- -L-
500
Temperature (K)
co I
I 100
co2
---200
-
300
400
---
500 Temperature (K)
H20 -
-
600
700
800
Figure 1. Thermal desorption spectra of 0.2-langmuir doses of acrolein and propionaldehyde from Cu*O( 100).
Figure 2. Thermal desorption spectra of 0.4-langmuir dose of propionaldehyde from CQO( 100).
cement. Thus, direct measurements of the crystal temperature were possible. All XPS spectra were collected using Mg K a radiation (hv = 1253.6 eV) exclusively and run a t a resolution of 1.03 eV full width at half-maximum (fwhm) on Ag 3 d s p Because of the width and asymmetry of the C 1s XPS spectra obtained from acrolein and propionaldehyde, peak fitting was necessary. A Gauss-Newton optimization routine employing Gaussian peaks was used." The variation in fitted peak positions which gave acceptable fits were f0.1 eV for the lowest and highest binding energy peaks and f0.2 eV for peaks at intermediate binding energies. For all fits, the minimum number of peaks which gave an acceptable fit was used. The peak-fitting routine was used with a fwhm of 1.9 eV for the C 1s signals. All spectra were referenced to the Cu+ 2~312binding energy of 932.4 eV.I2 Aldrich propionaldehyde (99+%) and acrolein (97%) were used in these studies and were purified by repeated f r e e z e p u m p thaw cycles prior to dosing. MSD Isotopes CH3CD2CH0(98% isotopic purity) was used after repeated freezepumpthaw cycles. Matheson Research grade H2 (99.9995%) was used as received. Since H2 does not dissociatively adsorb on the Cu20( 100) surface under uhv conditions,I3 hydrogen was predissociated using a hot platinum filament located 15 mm from the sample surface. All hydrogen doses are reported as the equivalent H2 dose.
desorption trace in Figure 1bshows only a small desorption feature between 300 and 400 K in addition to the primary desorption peak at 240 K. No difference in conversion or selectivity to products was observed for propionaldehyde as a function of sample history. This difference in desorption behavior is likely due to the presence of as yet unidentified surfacedefects associated with the highest-temperature desorption features observed in Figure la. Differences in desorption behavior due to surface defects have been observed previously for propene thermal desorption from oxygen-deficient Cu20( 1 11) surfaces.I4 All further propionaldehyde results reported here are for a "defective" surface with desorption behavior corresponding to that of Figure la. Propionaldehyde Thermal Desorption. Propionaldehyde reactivity was studied as a function of coverage for adsorption at 100 K. At low surface coverages (doses less than 0.05 langmuir) the primary propionaldehyde desorption channel appears as a single peak a t 260 K and shifts to 205 K by 0.4 langmuir (not shown). Additional desorption features are observed a t 335 and 450 K as noted above. Larger doses result in the growth of a peak at 145 K by 1.1 langmuirs which did not saturate with increasing coverage, indicating that this peak was due to a propionaldehyde multilayer. No bimolecular or C2 reaction products were observed during propionaldehyde TDS. The operational definition of a 1 M L coverage is taken as the coverage where a propionaldehyde multilayer first appears in TDS and where saturation of the product peaks occurs. The dose corresponding to a monolayer coverage of propionaldehyde at 100 K is 1.1 langmuirs. Propene, acrolein, CO, C02, and H20 were all detected as reaction products during propionaldehyde thermal desorption. Thermal desorption traces of the reaction productsdetected after a 0.4-langmuir (0.4-ML) dose of propionaldehyde are shown in Figure 2. The data from an 0.4-ML propionaldehyde coverage are shown because the desorption feature near 300 K is obscured by the main desorption channel at significantly higher coverages. Propene desorbs in a broad feature from 200 to 500 K, with maxima at 205 K and between 300 and 400 K. Acrolein also desorbs over a wide temperature range extending from about 250 to 550 K. The widths of the acrolein, propionaldehyde, and propene peaks are such that no definitive assignment of the reaction order of the rate-limiting step can be made from the TDS data. Other C3 productschecked for and not detected include allyl alcohol, propanol, acrylic acid, and propionic acid. No temperature shift in the HzO,CO, and COZdesorption signals is observed as a function of coverage, implying first-order kinetic processes.15 Water desorbs in a peak at 575 K, with a lower temperature shoulder at 500 K. Water TDS studies have shown that recombination of dissociated water occurs a t 465 K,16 and TDS studies of predissociated hydrogen adsorption have shown that atomic hydrogen extracts latticeoxygen to form water at about 500 K.I3 Thus, the rate-limiting step for the water
Results
Thermal Desorption. Figure 1compares the thermal desorption spectra of acrolein and propionaldehyde following 0.2-langmuir doseson theCu20(100)surfaceat 100K. Theprimarydesorption channel for both propionaldehyde and acrolein appears at 240 K, and the similarity in desorption behavior in this temperature range suggests similar bonding modes on the C u 2 0 (100) surface. However, clear differences in the desorption behavior of propionaldehyde (Figure la,b) and acrolein (Figure IC)wereobserved. Little acrolein desorption is observed above 300 K, while propionaldehyde desorption is clearly observed between 300 and 400 K in Figure l b and out to 500 K in Figure la. For propionaldehyde, differences in the number of thermal desorption features were observed dependent on sample history, as illustrated in Figure la,b. The propionaldehyde TDS trace in Figure l a was taken after that in Figure lb, following several months of studies of the adsorption and reaction of hydrocarbon oxygenates (alcohols and organic acids) on this (100) surface. The clean starting surfaces which gave rise to Figure la,b both exhibited the same XPS Cu/O ratio (1.6) and the same (3d2Xd2)R4So LEED pattern. No surface contamination was evident in either case with XPS. In addition to the primary desorption peakat 240 K, the propionaldehyde thermal desorption trace in Figure l a shows a distinct shoulder near 335 K as well as a second peak at 450 K. In contrast, the propionaldehyde
Acrolein and Propionaldehyde on Cu20( 100)
The Journal of Physical Chemistry, Vol. 97, No. 14, 1993 3551
TABLE I: Relative Yields for hopionaldehyde TDS on Cu*Ot100)
CHjCH2CHO TDS (0.4 ML) product
relative yield
C H 3C H 2C HO CH?=CHCHO C H :=CHCH 3
3.60 0.04 0.03 0.70 I .oo 1.20
co co2
CHyCHCHO (~90)
H2 0
i. 1 -.-.~
100
200
.
_
I
300 400 500 Temperature (K)
-
~
600
i
-
_
700
-
-
800
Figure 3. Thermal desorption spectra of the reaction of 0.4 langmuir of propionaldehyde with preadsorbed hydrogen (200-langmuir equivalent Hl dose) on Cu20(100).
produced at 575 K is the dehydrogenation of surface hydrocarbons. The water desorptions at 500 and 575 K have corresponding first-order activation energies of 33.9 and 39.1 kcal/mol, respectively, assuming a preexponential of 10'3 s-1. Reaction-limited CO2I7desorbs in a broad feature with a peak maximum near 625 K and a lower-temperature shoulder at 575 K, with corresponding activation energies of 42.6 and 39.1 kcal/ mol. CO desorbs in one reaction-limited peak18 at 650 K, corresponding to an activation energy of 44.4 kcal/mol. H2 was not detected as a desorption product; however, the background of hydrogen in the vacuum system was high enough that some H2 might have gone undetected. A change in the conversion of propionaldehyde was observed as a function of coverage and varied from 65% at low coverage (0.06 ML) to 10% at 1-ML coverage. The selectivity to reaction products remained constant at the coverages studied. The selectivity (on a C3 basis) to CO was 37 f 5%, to C02 was 52 f 5%, to propene was 5 f 3%, and to acrolein was 6 f 3%. The relative yields of the products from a 0.4-langmuir propionaldehyde dose are summarized in Table I. From the amounts of oxygen-containing products formed it is clear that lattice oxygen is extracted during the TDS experiments. No C3 products were observed above 450 K from propionaldehyde decomposition on the Ynondefectivensurface characterized by Figure 1b. Thus, the desorption of all C3 species above 450 K is attributed to unidentified surface defects. This observation suggests that the different temperatures for product formation may be due to different surface sites and not necessarily different reaction intermediates. H-Propionaldehyde Thermal Desorption. Figure 3 shows thermal desorption spectra of 0.4 langmuir of propionaldehyde from a hydrogen-predosed (200-langmuir equivalent H2 dose) surface. No new products were detected besides those reported previously for propionaldehyde adsorption on the clean (100) surface. Propionaldehyde desorbs at 190 K, 15 K lower than observed for a 0.4-langmuir dose on the clean (100) surface. Propene desorbs in a temperature range extending from 160 to 500 K, with clear desorption features near 195 and 300 K and a small feature centered near 450 K. Acrolein desorption is also observed at 300 K with a small contribution near 450 K. The range of desorption temperatures for propene and acrolein is similar to that observed for the clean surface with two exceptions. The desorption of propene at 190 K is 15 K lower than on the clean (100) surface, and the high-temperature desorption of acrolein has decreased by 50 K. C02 desorbs in two peaks at 480 and 585 K which are both reaction limited.I7 The C02 desorption peak at 480 K was not observed on the clean (100) surface, and the 585 K peak is 40 K lower than from the clean surface. Significant amounts of background water and CO were generated by the procedure used
u
H/CH,CHzCHO TDS (0.4 product CH3CH2CHO CHl=CHCHO CH~SCHCH 3
c02
ML)
relative yield 4.00
0.05 0.12
1.oo
to predissociate hydrogen, thus preventing quantification of the CO and H2O formed as surface reaction products. 1 Since the C O product signal could not be quantified, no calculation of conversion or selectivity for the hydrogenpropionaldehyde reaction could be made. However, about 3 times more propene was detected from the hydrogen-predosed surface than from the clean surface. The increase in the propene signal due to the presence of adsorbed hydrogen demonstrates the importance of surface hydrogen in the formation of propene via the hydrogenation of some surface intermediate. The relative yields of the products from the reaction of propionaldehyde with preadsorbed hydrogen are shown in Table I. CHjCD2CHO Thermal Desorption. The adsorption of deuterated propionaldehyde was examined to help elucidate the reaction pathways. The deuterium content of product propene and acrolein was specifically examined. The assignment of desorption products from CHjCD2CHO was made using the parent mass numbers of the expected products. Thus, singlydeuterated acrolein (C3H3DO)was identified using an m / z of 57, and labeled propionaldehyde (CH3CD2CHO) was identified using an m/zof 60. Different types of propene (CjHsD, C3H4D2) were identified by examination of cracking fragments using several mass numbers of 40 to 44. While it was possible to identify products of a particular mass, the overlap in the cracking patterns made it impossible to separate fragments which identified the positions of the deuterium label(s) in the C3 products. The thermal desorption behavior of the CH3CD2CHOwas the same as for the nonlabeled propionaldehyde discussed previously. Similar trends in selectivity and conversion to acrolein, propene, and the nonselective oxidation products were observed for CH3CD2CHO as for the nonlabeled propionaldehyde. Singly-labeled acrolein (CjH3DO), doubly-labeled propene (C3H4D2),and CHjCD2CHO were the only C3 compounds detected during thermal desorption, and all desorbed in the same temperature ranges observed with unlabeled propionaldehyde. The possibility of a reaction involving an intramolecular hydrogen transfer to form the doubly-deuterated propene (e.g., CHjCD2CHO+CH3CD=CHD + 0)waschecked by repeating CH3CD2CHO thermal desorption on a H-predosed surface. For these experiments, doubly-labeled propene ( C ~ H ~ D Zsingly), labeled propene (C,H5D), and singly-labeled acrolein (C3H3DO) were detected as products. The singly-labeled (CjHsD) and doubly-labeled (C3H4D2)propene were detected in a ratio of 5:4. The presence of the singly-labeled propene demonstrates that propene formation occursvia the reaction of a surfacespecies with surface hydrogen. No singly-labeledpropionaldehyde (C3HsDO) was detected; however, separation of the m / z 59 signal for a few percent of this product from the cracking fragments of the large CH3CD2CHO contribution was not possible. Thus, some singly-labeled propionaldehyde may have been produced but not detected in the TDS experiments. Acrolein Thermal Desorption. Acrolein reactivity was studied on the Cu20(100) surface as a function of coverage after adsorption at 100 K. Acrolein desorbs in one peak which shifts with coverage from 260 K for low doses (
