J . Phys. Chem. 1991, 95, 274-282
274
of the recombination process. The constants also depend strongly on the pretreatment of the samples but less on the surface loading. Conclusions
Laser cxcitation of distyrylbenzenes adsorbed on high-surface-area silica and alumina surfaces is shown to give rise to short-lived triplet states and long-lived radical cations that can be studied by time-resolved diffuse reflectance spectroscopy. The decrease in reflectivity due to triplet-triplet absorption by the adsorbate increases with increasing laser fluence until the laser dose is sufficient to create a conversion layer that is thick enough to shield the rest of the sample from spectroscopic detection. The observed variation, which rises quickly to a saturation value, follows expectations for a "one-photon" process involving absorption by ground-state adsorbate to give SI followed by intersystem crossing to give T,. In contrast, the absorption by the radical cation shows a much slower initial rise with laser fluence compatible with a sequential "two-photon" absorption process, i.e. So + hu -,SI hu S, R+ + e-.
-
+
-+
The radical-cation decay is extremely nonexponential with a proportion decaying over a few microseconds but with complete recovery taking several minutes. However, the decay is reversible; Le., all the radical cations decay to give the original adsorbate. Thus decay is due to radical ion-electron recombination. The amount of transient produced is independent of the number of times the sample has been exposed, but the decay shows a large temperature dependence, so large that cooling to 77 K virtually eliminates decay and ESR spectra of the stabilized radical ions can be measured. The decay is independent of the loading of the adsorbate and of laser fluence, which proves that the decay involves the recombination of the originally produced radical ion-electron pair. At long times, pair recombination theory predicts that pair decay will be a diffusional process and the amount of radical ion remaining will have a dependence, as is found experimentally. Acknowledgment. This work was supported by the European Community (Contract No. ST2-0069) and by the Deutsche Forschungsgemeinschaft (Oe 57/12) to whom we are grateful.
Dinitrogen Photoreduction to Ammonia over Titanium Dioxide Powders Doped with Ferric Ions J. Soria, J. C. Conesa, Instituto de Catcilisis y Petroleoquimica, CSIC, 28006 Madrid, Spain
V. Augugliaro,* L. Palmisano, M. Schiavello, and A. Sclafani Dipartimento di Ingegneria Chimica dei Processi e dei Materiali, University of Palermo, 901 28 Palermo, Italy (Received: April 2, 1990)
The production of ammonia from dinitrogen and water vapor at mild conditions of temperature and pressure on Fe3+-doped titanium dioxide powders under UV radiation has been studied in a continuous photoreactor working in the gassolid regime. A net activity decline was observed after a few hours of irradiation; this decline did not depend on the reactor temperature, the powder composition, or the amounts of produced ammonia. The highest activity is found when no excess iron is segregated at the surface; overall turnover numbers for dinitrogen reduction as high as six electrons per iron atom can then be reached before powder deactivation, showing the catalytic character of the participation of Fe in this process. An IR investigation of the active and spent specimens revealed that the irradiation determines the almost complete disappearance of the OH groups from the powder surface. Furthermore, an ESR study of all the powders showed that bulk Fe3+ions are better electron traps than Ti4+ ions so that, when a UV photon generates a hole-electron pair, the electron can be stabilized on the iron ions. The charge separation, thus favored by the latter, makes possible the adsorption of dinitrogen by its reaction with surface species activated by holes. From these data, the deactivation process is ascribed to the decrease in the amount of the reactive hydroxyls which make possible the regeneration of the bulk Fe3+traps necessary for the ammonia photoproduction; the activity, as well as the Fe3+ and OH- groups, is recovered by treatment of the powders in air at 823 K followed by slow cooling.
Introduction
The dinitrogen reduction to ammonia in the presence of water over irradiated semiconductor powders in mild conditions of temperature and pressure constitutes a new route for the solar energy conversion into chemical energy. This process is even more difficult than that of H 2 0 photodecomposition. Several factors contributing to this are the following: (a) the reduction of one molecule of N2 involves transfer of six electrons compared to two electrons in the H20reduction; (b) although the energy requirements of H20decomposition and N2 reduction are quite similar, the activation energy for N2 reduction is very high; (c) N, adsorption is a necessary requirement for catalytic activity, but most semiconductor oxides do not adsorb N2 easily; (d) in addition, when batch systems are used, any photogenerated ammonia accumulates in the reacting system and can be involved in back-reactions which ultimately block reaction progress. Schrauzer and Guth' were the first authors to report on the dinitrogen photoreduction; by testing various iron-doped and ( I ) Schrauzer, N.; Guth, T. D. J. Am. Chem. SOC.1977, 99,7189-7193.
0022-3654/9 1 /2095-0274$02.50/0
iron-free Ti02 (rutile) specimens, they found that Fe ions improve the activity, which however declines for Fe contents above 0.2 wt %. After that, many studies have been performed on this photoreaction, which have definitively proved its feasibility in heterogeneous conditions, and exhaustive reviews have been reported in the Among the several semiconductor oxides t e ~ t e d ,titanium ~,~ dioxide doped with various cations is found to be one of the most effective.'V6-l5 A very recent paperI6 reports (2) Schiavello, M.; Sclafani, A. Photoelectrochemistry, Photocatalysis and Photoreactors. Fundamentals and Developments; D. Reidel Publishing Co.: Dordrecht, 1985; pp 503-519. (3) Augugliaro, V.; Palmisano, L. Photocatalysis and Environment Trends and Applications; Kluwer Academic Publishers: Dordrecht, 1988; pp 425-444. (4) Endoh, E.; Leland, J. K.; Bard, A. J. J. Phys. Chem. 1986, 90, 6223-6226. (5) Khader, M. M.; Lichtin, N. N.; Vurens, G.H.; Salmeron, M.; Somorjai, G. A. Langmuir 1987, 3, 303-304. (6)Augugliaro, V.; Lauricella, A.; Rizzuti, L.;Schiavello, M.; Sclafani, A. Int. J. Hydrogen Energy 1982, 7, 845-850. (7) Augugliaro, V.; DAlba, F.; Rizzuti, L.; Schiavello, M.; Sclafani, A. Int. J . Hydrogen Energy 1982, 7, 851-855.
0 1991 American Chemical Society
Dinitrogen Photoreduction to Ammonia that pure Ti02 powders, normally inert for ammonia generation in either the rutile or anatase form, acquire catalytic activity for this photochemical reaction after being annealed in air at 1300 K, this being ascribed to the generation of defect states in the band gap of the semiconductor, Le., donor centers for the photochemical reaction. The powders, after being subjected to successive cycles of irradiation and temperature-programmed desorption, showed a progressively decreasing activity, due to the consumption of certain hydroxyl groups generated on the Ti02 surface by the annealing treatment. The question of the catalytic nature of this system and, in a general way, the meaning of “catalytic” regime in a process that involves a semiconductor solid stimulated by the radiation have been widely disc~ssed.”-~~ The formation of ammonia amounts that correspond to only few OH- layers, the significant decline of reactivity after several hours, and the relatively low turnover number were all reasons presented for casting doubts on the catalytic nature of this system. The present paper reports the results of an investigation studying the formation of ammonia over a series of iron-doped Ti02 specimens, UV-irradiated in a continuous fixed-bed laboratory reactor. The investigation allows one to draw the operative conditions for obtaining ammonia, the role played in the reaction mechanism by ferric ions and by the O H groups present on the surfaces of the specimens, and the reasons for the observed decline of activity. The question of the catalytic nature of the photoreaction is also addressed. ESR, TR, XRD, and SEM were used, besides the activity tests, to achieve a more detailed understanding of the process. The information obtained allows to draw a likely picture of several key aspects of the process.
Experimental Section Powders Preparation. The coprecipitation and wet impregnation methods were used for preparing two series of iron-doped titanium dioxide powders; these specimens are hereafter designated as (CP)Xand (IM)X, respectively, where X refers to the nominal iron concentration (0.2,0.5, 1, and 2% of molar cationic contents). A (CP)O.l sample was also prepared for performing only an IR investigation. For preparing the C P powders, iron and titanium hydroxides were coprecipitated by reacting an aqueous solution of TiC1, (15 wt %, Carlo Erba) containing the required nominal quantity of Fe3+ ions (from Fe(N03)3.9H20,Merck) with an aqueous ammonia solution (25 wt %, Merck). This was done by adding the (8) Schiavello, M.; Rizzuti, L.;Sclafani, A.; Majo, I.; Augugliaro, V.; Yue, P. L. Hydrogen Energy Progress IV; Pergamon Press: Oxford, 1982; pp 821-826. (9) Schiavello, M.; Rizzuti, L.;Bickley, R. 1.; Navio, J. A.; Yue, P. L. In Proceedings of the 8th International Congress on Catalysis; Verlag Chemie: Basel, 1984; pp 383-392. (IO) Bickley, R. 1.; Gonzilez-Carreno, T.; Palmisano, L. Preparation of Cafalysts I V ; Elsevier Science Publishers: Amsterdam, 1987; pp 297-306. ( I I ) Augugliaro, V.;Palmisano, L.; Schiavello, M.; Sclafani, A,; Conesa, J . C.; Soria, J. In Proceedings of the XI Iberoamerican Symposium on Catalysis; 1988; pp 217-224. (I2) Palmisano, L.;Augugliaro, V.; Sclafani, A,; Schiavello, M. J . Phys. Chem. 1988, 92, 6710-6713. ( I 3) Conesa, J . C.; Soria, J.; Augugliaro, V.; Palmisano, L. Structure and Reactiuify of Surfaces; Elsevier Science Publishers B.V.: Amsterdam, 1989; DD 307-3 17. (14) Tennakone, K.;Punchihewa, S.;Tantrigoda, R. Sol. Energy Mater. 1989, 18, 217-221. ( I S ) Endoh, E.; Bard, A. J. Nouu. J . Chim. 1987, 1 1 , 217-219. (16) Bourgeois, S.;Diakite, D.; Perdereau, M. React. Solids 1988, 6,
..
_-
w-ind .-..
( I 7) Formenti, M.; Teichner, S.J. Catalysis; The Royal Chemistry Society: London, 1978; pp 87-106. (18) van Damme, H.; Hall, W. K. J . Am. Chem. SOC.1979, 101, 4373-4374. ( 19) Pichat, P. Phoroelectrochemistry, Photocatalysis and Photoreactors. Fundamentals and Deuelopments; D. Reidel Publishing Co.: Dordrecht, 1985; pp 425-455. (20) Teichner, S. J.; Formenti, M . Photoelectrochemistry, Photocatalysis and Photoreactors. Fundamentals and Deuelopments; D. Reidel Publishing Co.: Dordrecht, 1985; pp 457-489. (21) Childs, L. P.; Ollis, D. F. J . Catal. 1980, 66, 383-390.
The Journal of Physical Chemistry, Vol. 95, NO. 1 , 1991 275 latter dropwise to the metal solution at room temperature, with vigorous stirring owing to the exothermicity of the reaction. After standing for 24 h at room temperature, the solid was filtered and repeatedly washed with bidistilled water for 15-20 days until complete disappearance of CI- ions in the effluent liquid. The resulting solid was dried at 393 K for 24 h and then fired in air at 823 K for 24 h. A few CP specimens were prepared similarly except for the use of a quick washing procedure, which left a substantial amount of C1- ions both in the bulk and on the surface of the specimens. A “home-prepared” Ti02 (hereafter named “TD hp”) was obtained exactly in the same way as the C P samples cited, but with no Fe added. For preparing the IM powders, a portion of TD hp sample was added, at room temperature with stirring, to a minimum volume of an aqueous solution of ferric nitrate containing the required amounts of iron; after standing at room temperature for 48 h, the liquid phase was evaporated at 393 K for 24 h and the dried solid fired in air at 823 K for 24 h. Pure Fe2TiOSwas prepared by mixing the required amounts of FeCI, and TiCI, solutions followed by reoxidation of Fe2+ with HNO,; the resulting acid solution was then heated to incipient boiling and neutralized with ammonia solution (8 M). The dark precipitate was dried at 370 K, ground, and heated in air at 970 K for 5 h, at 1570 K for 12 h, and at 1370 K for 120 h, with careful grinding between each thermal treatment. Some reactivity runs were also made on commercial specimens of pure TiO, (BDH, Degussa P25, and Tioxide CLDD 1601/2) or Fe203(Carlo Erba, reagent grade) which were used as received, by only performing on them, before the run, the usual firing treatment in air. After the firing treatment all the solids were crushed and sieved, and those powders with particles size in the 37-90-pm range were tested in the reactivity experiments. Before each reactivity run was performed, the powder was conditioned in air a t 823 K for few hours. Characterizationand Spectroscopic Methods. X-ray diffraction patterns were obtained at room temperature with a Philips diffractometer using Ni-filtered Cu K a radiation. SEM observations were made on gold-coated samples using a Philips 505 microscope equipped with a KEVEX analytical apparatus. Particle and pore sizes were evaluated from these micrographs, the approximate lower limit being q5 = 100 nm. Mesopores (4 < 50 nm) were therefore outside the resolution of the instrument. BET surface areas were determined with a Micromeritics Flowsorb 2300 instrument, using N2 at 77 K. ESR spectra were recorded at 77 K with a Bruker spectrometer (Model ER 200D SRE, X band); the samples were placed in silica cells with greaseless stopcocks allowing thermal treatments under vacuum and admission of gases when required. For IR spectra (recorded with a Nicolet SZDX system), the specimens were previously pressed to self-supporting disks and outgassed for 1 h at different temperatures up to 623 K; the spectra were recorded at the temperature (ca. 313 K) resulting from direct heating of the sample by the infrared beam. Photoreactivity Tests. All reactivity runs were performed in a continuous flow, fixed-bed photoreactor made of a Pyrex tube (i.d. = 0.4 cm, 0.d. = 0.6 cm) filled with 1 g of the powder to a height of ca. 10 cm. It was vertically positioned inside a water bath made of Pyrex and thermostated at 295,303,313,333, and 353 K. For UV irradiation, a 160-W high-pressure Hg lamp (OSRAM HWL) with bulb height of 10 cm was positioned parallel to the reactor at a distance of ca. 8 cm. Dinitrogen (99.95% purity) was bubbled at 0.4 cm3 s-I through distilled water at room temperature and fed to the reactor. (Several runs were performed with the water saturator at different temperatures, but no difference was observed.) For a few runs a dry N2-H, mixture, 10% in H2. was used. The gas exiting the reactor was bubbled through a 0.01 N HCl aqueous solution in order to fix the evolved ammonia; quantitative determination of the thus trapped NH3 was performed by the standard indophenol colorimetric method able to detect 1 ppb. Irradiation of the reactor was started when steady-state conditions for powder temperature, gas composition, and flow rate
Soria et al.
276 The Journal of Physical Chemistry, Vol. 95, No. 1, 1991 TABLE I: Total Amounts of Ammonia, NH,(tot), Produced during the 1and 3-h Reactivity Runs at Various Reactor Temperatures, Tn@ surf. run NH,(tot), IO-* g m-2, at TR= area, time, specimen m2g-' h 295 K 303 K 313 K 333 K 353 K 40.2 3 4 (1) 68.7 3 117 102 195 (74) (37) (81) 66 198 3 113.1 (27) (87) 33 264 189 80.6 3 (16) (238) (142) 26 33 90.3 3 44.4
3
44.4
1
69.5
3
69.5
1
48.7
3
48.7
1
73.1
3
(196) 126 (20) 120 (16)
0 (0)
(173) 246 (188) 297 (178) 240 (198) 327 (203) 225 (192)
(162)
(127)
297 (170)
327 (127)
384 (260)
327 (156)
(128) 260 ( 140) 441 (120) 237 (131) 375 (113) 198 (116) 0 (0)
0 (0) @Thevalues in parentheses indicate the amounts of ammonia desorbed from the powder after the reactivity run.
were achieved. Two series of runs were performed. With the first ones, the initial rates of ammonia production were determined for all the powders at the cited reactor temperatures; these runs lasted 1 or 3 h, the ammonia trap solution being replaced and analyzed each hour in the latter case. The second series of runs was performed for determining the possible decline in reactivity; in these runs, lasting between 20 and 28 h, only the CP powders were tested at the reactor temperatures of 303 and 353 K. In all cases, at the end of the photoreactivity run the reactor was purged with N 2 at 573 K in order to recover and analyze the ammonia that remained adsorbed on the powder; temperatures less than 573 K were not effective in desorbing all the ammonia. Some experiments of ammonia desorption from the powder were also performed by using He instead of N2 as carrier gas, but no difference was observed. Before the reactivity runs, blank tests were carried out on the system either without powder or without irradiation or by feeding only H, and/or N2 to the reactor, other reaction conditions being unchanged. These tests confirmed that ammonia was produced only on the irradiated powder and that dinitrogen and water were the reagents.
Experimental Results XRD Results. The TiO, modifications identified by X-ray analysis were anatase and rutile. The thermal treatment at 823 K of samples prepared by coprecipitation, including also the TD hp, yielded powders consisting practically of pure anatase, while the samples prepared by wet impregnation consisted of anatase with small percentages of rutile that increased with the iron content. For the (CP)2 and (IM)2 specimens the presence of highly disordered pseudobrookite, Fe2Ti05,and Fe203was detected. SEA4 Results. For all the specimens (TD hp included) a very large distribution of particle shapes and dimensions was observed. Macropores (diameter >50 nm) were not observed in T D hp but did appear in CP and IM specimens, the latter being possibly a little more macroporous than the C P samples. The surface microanalyses showed for IM samples a very inhomogeneous distribution of iron, not only by comparing different particles but also within a single particle: moreover, the values of iron content measured at various points of the particle surface were always higher than the nominal one. For the C P specimens the surface microanalyses always showed a homogeneous dis-
l5
-
t
1
A
1
1:
0
10 I
( h)
20
30
Figure 1. Amounts of ammonia released to the gas phase per unit time and unit surface area vs the run time. Reactor temperature = 303 K: A, (CP)0.2; 0, (CP)0.5; 0, (CP)I. Reactor temperature = 353 K: A and V, (CP)0.2; 0 , (CP)O.5; m, (CP) 1. TABLE 11: Experimental Results of the Total Amounts of Ammonia Produced during Long Time Reactivity Runs, NH3(tot), and of the Amounts of Ammonia Desorbed from the Powder after the Reactivity Run, NHJdes) NH3(tot), NH,(des), specimen run time, h TR, K g m-' IO-* g m-2 (CP)0.2 20 303 667.6 212 18 353 1720 136 (CP)0.2 28 353 1852.8 148 (CP)0.2 583.1 195 (CP)0.5 25 303 (CP)0.5 23 353 1096.8 I40 519.5 185 (CP) 1 26 303 23 353 945.7 108 (CP) 1
tribution of iron content with values close to the nominal ones. Reactivity Results. No ammonia production was detected when pure Ti02, Fe2Ti05,or Fe203was used for reactivity tests. The reactivity results obtained by testing the T D hp, IM, and C P samples in 1- and 3-h runs are reported in Table I. All the figures reported in this table are normalized with respect to the total surface area of the powder used; this parameter, and not the powder mass, is considered the most correct for specifying the reactivity data, as the main phenomena necessary for the occurrence of this photoreaction (light absorption, subsequent e-h pair generation, adsorption of reactive molecules, and desorption of products) involve the powder surface. The expressed total amount of ammonia produced during the run was obtained by adding the amount of ammonia released to the gas phase during the run and that desorbed at high temperature from the powder after completion of the run; these latter values are also included in parentheses. For the 3-h runs it was observed that the amount of ammonia was always almost the same in the three traps removed every hour; on the other hand, no substantial difference was found in the amount of ammonia remaining adsorbed after 1- or 3-h runs. The data obtained indicate that, in general, the activity increases by increasing the temperature, although this trend is less marked for the C P specimens. Table 1 shows that a small amount of ammonia is produced by the TD hp sample while no activity was shown by the (CP)2 specimen; the highest activity is exhibited by the 0.2-0.5 specimens. Ammonia remains adsorbed in larger amounts at lower temperatures; this behavior is more evident for the CP specimens (which, as shown by the SEM results, are more homogeneous) while for the IM specimens there is larger scatter in the data; compare e.g. the (1M)l specimen with other samples in its series. It must be also reported that all CP specimens, prepared by using a quick washing procedure, exhibited a very poor activity; this can be traced to the presence of chloride ions. These samples acquired activity comparable to that of the other specimens after a drastic treatment involving heating in H2 for 90 h at 823 K and subsequent calcination in air; tests for checking the presence of
Dinitrogen Photoreduction to Ammonia
The Journal of Physical Chemistry, Vol. 95, No. 1 , 1991 277
n
1000 G H
-,k
2.004
4DPPH
L--A
xl
d L d
Figure 2. ESR spectra of CP samples outgassed at 298 K: (a) (CP)0.2; (b) (CP)0.5; (c) (CP)I; (d) (CP)2.
bulk CI- ions indicated their total absence after this treatment. The results of ammonia production obtained with long-time runs are reported in Figure 1 and in Table 11. In Figure 1 the amounts of ammonia released to the gas phase during the run (expressed as average hourly yields since the precedent point) are displayed versus time. Table I1 reports, for the same runs of Figure I , the total amounts of ammonia produced together with those desorbed from the powders after the reactivity runs; the latter values do not differ substantially from those in Table I. These data show that, in all cases, the hourly amounts of ammonia released to the gas phase are practically constant for the first 5-7 h; after that time a net decline of the powder activity is evident. This powder deactivation was not irreversible; a complete recovery of the activity was achieved by heating again the specimen at 823 K in air for several hours followed by slow cooling, temperatures lower than that not being effective. Performing the previous treatment in flow of dinitrogen had no effect in recovering the activity. ESR Results. The 77 K spectra of the C P and IM samples, outgassed a t 295 K, are shown in Figures 2 and 3, respectively. The spectrum of C P samples (Figure 2) can be described as a superposition of several Fe3+signals. The peaks at g = 8.18,5.64, 3.43, and 2.6 (signal A) have been assignedZZto Fe3+ ions substituting for Ti4+in the TiOt rutile lattice. In order to confirm this assignment, some C P samples were calcined at 1073 K, whereupon most of the Ti02was transformed to the rutile phase; the resulting spectra showed the same peaks, with sharper characteristics, of the signal A reported in Figure 2. Therefore, although XRD only shows anatase peaks in these samples, small rutile nuclei must exist which retain iron ions in their lattice. As reported below, 1R results indirectly confirm the presence of rutile phase in the C P specimens. Signal B at g = 4.26 has been assignedz3to Fe3+ ions in rhombic symmetry; it is observed with smaller intensity for the mentioned C P samples calcined at high temperature, suggesting that these centers are mainly located in the anatase phase. The peaks with g values closer to 2 can be tentatively interpreted as a superpition of an axial signal C, with g, = 2.004 and unresolved g,,< g,, together with another signal D displaying a split structure centered at g N 2, probably due to Fe3+ in octahedral symmetry. The intensity of these peaks decreases after calcination at high temperature, supporting their (22) Maksimov, N . G.; Mikhailova, 1. L.; Anufrienko, V. F. Kinef. Card. 1973, 13, 1162-1 167. (23) Thorp, J. S.;Eggleston, H. S. J . Murer. Sci. Lerr. 1985,4, 1140-1 142.
Figure 3. ESR spectra of IM samples outgassed at 298 K: (a) (1M)O.Z; (b) (1M)OS; (c) (1M)l; (d) (IM)2.
1
,l000G,
4 x25
Figure 4. Effect of the outgassing temperature on the ESR spectra of (CP)0.2 sample: (a) 298, (b) 373, (c) 473, (d) 573, (e) 673, and (f) 773 K.
assignment24to Fe3+ ions in anatase. The effect of a higher iron content in the CP samples consists mostly in an increase in intensity and line width of the lines C and D, which coalesce in a single peak. The spectrum of IM samples (Figure 3) shows essentially the same signals as CP samples, but here all of them appear broader and, with the exception of signal B, show a lower amplitude. In the case of the (IM)0.2 sample, the larger spectrometer gain used for recording the spectrum allows the observation of a broad underlying component, centered around g = 2, which is probably due to aggregated Fe3+ ions with mutual magnetic coupling. Also for the IM powders a higher amount of iron leads mostly to an increase in lines C-D, while the other lines appear with lower amplitudes and intermediate line widths. (24) Cordischi, D.; Burriesci, N.; DAlba, F.; Petrera, M.; Polizzotti, G.; Schiavello, M. J . Solid Stare Chem. 1985, 56, 182-190.
218
The Journal of Physical Chemistry, Vol. 95, No. 1. 1991
Soria et al. 1000 G
1000 G
I
e
I
xl f
-
xl
xl
Figure 6. ESR spectra of (CP)0.2 sample UV-irradiated in a special ESR cell under N2-H20flow for (a) 0, (b) 1, (c) 2, (d) 3, (e) 4, (0 5, and (8) 6 h.
1 0 0 0; ; 1
xl
Figure 7. ESR spectra of (CP)0.2 sample UV-irradiated in a special ESR cell under N2 flow for (a) 0, (b) 2, and (c) 8 h.
After irradiation, the initial intensity of the Fe3+ signals was recovered after a prolonged treatment under air or dry O2at 823 K; no such recovery occurred by heating under flowing N2. IR Results. IR spectra were obtained on several samples after outgassing at T = 298-623 K. It is worthwhile to recall that all the CP and IM specimens used in this work underwent the same thermal treatment before reactivity or spectroscopic experiments. Figure 8 reports, for a (CP)0.5 sample, typical changes of the intensities of the OH infrared bands as a result of thermal dehydroxylation under vacuum. The stretching frequencies of the OH groups observed after evacuation a t 423 K have the values of ca. 3725, 3690, 3670, 3650, and 3420 cm-l, which almost coincide with those reported by Primet et and by Munuera et aL2* The presence of the 3650- and 3420-cm-’ bands, typical of OH groups on rutile, suggests, as the ESR results do, that this phase is also present in the CP specimens. Isolated OH groups are responsible for the band at 3725 cm-’ (anatase) and for that at 3690 cm-’ (rutile); the bands at 3670 cm-I (anatase) and at 3650 and 3420 cm-’ (rutile) originate from OH groups bonded to each other by hydrogen bridges and situated in adjacent unit cells.26 Similar spectra are obtained for other specimens (Figure 9). The dehydroxylation pattern in all cases is the same: the hydrogen-bonded OH groups, responsible for the bands at 3420, (26) Primet, M.; Pichat, P.; Mathieu, M.-V. J . Phys. Chem. 1971, 75, 1216-1220.
The Journal of Physical Chemistry, VoL 95, No. 1, 1991 279
Dinitrogen Photoreduction to Ammonia
W V C
V
-m
C
m
-.-
Y
e
E
E
(1
c
v)
m
C
m
L I
L I
4000
3000
4000 wavenumber
(cm ' )
Figure 8. Effect of the outgassing temperature on the IR spectra of a (CP)0.5 sample: (a) 298, (b) 323, (c) 373, (d) 423, (e) 473, (f) 523, (9) 573. and ( h ) 623 K .
wavenumber
I cm-' 1
3000
Figure 10. 1R spectra of a (CP)0.2 sample irradiated for (a) 0, (b) 2, (c) 4, and (d) 6 h. (A) Spectra of samples evacuated at 423 K. (B) Spectra of samples evacuated at 573 K.
W 0 C
m Y c
.-
E v)
c
m E v)
c m L d
I V I '
1
1
-
3000
4000 wavenumber 1
1
'
1
'
4000
1
I 3000
wavenumber
(cm"')
Figure 9. IR spectra of specimens outgassed at 423 K (a) (1M)O.S; (b) TD hp; (c) (CP)O.l; (d) (CP)0.2; (e) (CP)O.5; (f) (CP)l; (g) (CP)2.
3650,and 3670 cm-I, are removed at temperatures below 473 K, while the isolated OH groups disappear at higher temperatures. In order to check the influence (if any) of the iron ions content on the surface hydroxyl groups, the OH bands in these spectra
(cm-')
Figure 11. IR spectra of a (CP)O.5 sample irradiated for (a) 0, (b) 2, (c) 4, and (d) 6 h. (A) Spectra of samples evacuated at 423 K. (B) Spectra of samples evacuated at 573 K.
were integrated and normalized with respect to the surface area; almost the same values were obtained for all the specimens irrespective of their preparation method and iron content, thus indicating that all the samples contain about the same number of OH groups per unit surface area. Figures 10 and 11 report the IR spectra of (CP)0.2 and (CP)O.5 samples after being subjected in the ESR cell to photoreactivity
280
The Journal of Physical Chemistry, Vol. 95, No. I, 1991
a V C
m
-
c I
E v)
C
m L I
4000
3000 wavenumber
(cm-’)
Figure 12. Effect of the outgassing temperature on the IR spectra of a (CP)O.l sample irradiated for 2 h: (a) 298, (b) 323, (c) 373, (d) 423, (e) 523, and (f) 623 K .
runs of 2,4, and 6 h under flow of wet dinitrogen and subsequent outgassing at 423 and 573 K. By increasing the irradiation time, a net decrease of the OH bands intensity is observed. Figure 12 reports, for a (CP)O.l specimen irradiated for 2 h under reaction conditions in the ESR flow cell, the 1R spectra recorded after successive thermal treatment under vacuum. These spectra show, besides the decrease of the O H bands with respect to those of the nonirradiated specimen (reported in Figure 9), a broad band at ca. 3250 cm-I; this band is assignable2’ to ammonia adsorbed on the OH groups with the formation of 0-H-N hydrogen bonds.
Discussion Bulk and Surface Properties of the Powders. The structural investigation of the powders shows that most of the titanium dioxide exists in the anatase phase. The solubility of iron in anatase24is about 1 atom % so that, for samples with nominal iron concentration below this limit, a homogeneous distribution of iron in the Ti02is, in principle, possible. SEM observations confirm the homogeneity of the CP samples; for the IM ones, on the contrary, some enrichment in iron of the external surface is found, due obviously to its imperfect diffusion to the bulk of the Ti02 grains at the low temperature of preparation. Such an enrichment must be indeed present in the (CP)2 and (IM)2 samples, for which an average surface concentration of excess iron of (at least) 1 .O and 0.8 ions/nm2, respectively, can be calculated by considering the quoted solubility limit. The ESR spectra of Figures 2 and 3 also confirm this difference in the iron distribution. The spectrum of (CP)2 sample differs from the others reported in Figure 2 mainly in the (expected) higher intensity and in the larger line width of the signals due to isolated Fe3+ ions; this broadening is almost certainly due to the dipolar magnetic interactions induced by the higher concentration of Fe3+ inside the TiOz. For the other IM samples, the signals found are broader and smaller than for the corresponding CP ones. These features indicate that in the powders prepared (27) Primet. M.; Pichat, P.; Mathieu, M.-V. J . Phys. Chem. 1971, 75, 1221-1226. ( 2 8 ) Munuera, G.; Rives-Arnau, V.; Saucedo, A. J . Chem. Soc., Faraday
Trans. I 1979, 75, 736-747.
Soria et al. by impregnation Fe3+ ions have penetrated into the Ti02 matrix to a relatively low depth, so that their signals are smaller and the proximity of most of such ions to the surface determines distortions of their coordination environment and broadening of the peaks. Even for the (IM)0.2 sample a broad underlying signal is detected which can correspond to agglomerates of excess Fe3+ ions remaining at the surface. In the case of the (IM)2 sample, the presence of larger amounts of iron ions may allow their penetration in the Ti02 grains to a larger depth; the sharpness of the signal detected here suggests that the saturation level has not been reached everywhere in this sample, allowing a smaller dipolar line broadening in part of the dissolved iron ions. It must be mentioned in any case that, even in the most concentrated samples, no ESR signal was found which could be ascribed to the Fe2TiOs phase. Photoreactivity Results. The salient indications of the reactivity results, obtained in short- and long-time runs, can be summarized as follows. Dinitrogen photoreduction occurs only when iron is present in the Ti02 samples, pure Ti02 and Fe203 being totally inactive. The reactivity, however, does not parallel the iron content; on the contrary, the highest activity is exhibited by samples with the lowest iron content. By increasing the reactor temperature, the ammonia production rate increases while the amount of adsorbed ammonia decreases. As the composition of the N2-H20 reagent mixture does not affect the reaction rate and the activation of the semiconductor surface by the radiation is a process independent of the temperature, it may be concluded that the ratedetermining step of the process is not the reactants adsorption and activation but, more likely, some surface chemical step following the previous ones or the desorption of ammonia from the powder surface. The results of ammonia production rates, reported in Figure 1 versus time, together with the observed constancy with time of the amounts of adsorbed ammonia clearly show that the process of dinitrogen photoreduction proceeds at constant rate for the first 5-7 h of irradiation. After that time (which the data show to be independent of the reactor temperature, the powder composition, and the total amount of ammonia produced until that time), a net decrease of activity occurs. As the main feature of a continuous reactor is that of keeping in steady-state conditions its operative parameters (some of which may act as driving forces for the physical and chemical processes occurring in it), the decline of activity can be explained only by a modification of the physicochemical properties of the powder during the course of the irradiation. Furthermore, the finding that the decline occurs always after the same time suggested that this modification is linked to a photoreaction independent of the dinitrogen photoreduction process. Therefore, two phenomena in parallel may be envisaged to occur under irradiation: dinitrogen photoreduction and physicochemical modification of the powder via an independent photoreaction, with the feature that the second process is able to stop the first one. Role of Iron Ions. The reactivity tests show that the presence of iron ions in the Ti02 matrix is essential for the occurrence of the dinitrogen reduction. The activity shown by the TD hp sample, the ESR spectrum of which revealed the presence of a small amount of iron, and the lack of activity in the Fe-free samples confirm the previous consideration. An excess of iron is however detrimental, even if the Ti02surface is not fully covered by iron compounds. In fact, the highest activity is shown by the (CP)0.2 and (CP)0.5 samples, for which a surface enrichment in Fe is very unlikely to exist. The favorable effect of iron must arise therefore from the Fe3+ ions inside the TiOZmatrix, and the decrease in activity with the increase of the surface enrichment in Fe indicates that surface iron negatively affects particular mechanism steps and/or blocks specifically some particular surface sites. Since only nonsurface Fe is active, its role must be linked to the photogeneration and/or the evolution of electrons and holes in the Ti02 grains. The observed influence of these iron ions in the dinitrogen photoreduction can be explained by the hypothesis that they trap the electrons, thus helping their separation from holes and allowing the latter to reach the surface, react with hydroxyl groups, and give rise to highly reactive OH’ radicals which ori-
The Journal of Physical Chemistry, Vol. 95, No. 1, 1991 281
Dinitrogen Photoreduction to Ammonia ginate from the chemical reactions at the surface:
+
Ti02 hv Fe3+
-.
+ e-
OH-surf+ h+
-
Ti02(e-h+)
(1)
Fe2+
(2)
OH*surf
(3)
The electron-accepting role for the Fe3+ ions is supported by the ESR results here reported. In fact, outgassing of the powders at temperatures less than or equal to 573 K does not produce Ti3+ as in Fe-free samples, but instead the Fe3+ signals decrease and, only after most of the Fe3+ has disappeared, can Ti3+ions be found. This shows clearly that the bulk Fe3+ ions trap excess electrons (generated by desorption of O2from previously existing surface 02,-) in order to give Fez+ ions (undetectable by ESR at 77 K) and that they are better electron traps than Ti4+ions. Furthermore, the spectra in Figure 6 of (CP)0.2 sample, irradiated for long time under reaction conditions in the special ESR flow cell, do show the almost complete disappearance of the Fe3+ signals when the photoreactivity almost disappears. This fact strongly suggests that this redox state of iron is essential for photoactivity and also reveals a photoredox process that must correspond to a reduction of Fe3+, since it can be reversed by treating at 823 K under air but not under N2. The finding that in the presence of hydrogen the reactivity tests reveal no ammonia production and the iron ESR signals do not decrease may also be explained by supposing that the hydrogenation of the sample surface determines the formation of centers which favor the electron-hole pairs recombination. The role of bulk Fe3+ ions is therefore that of helping the separation of electrons and holes by (temporarily) trapping the former and allowing the thus generated surface OH' to have an increased lifetime, making it possible for these highly reactive radicals to give at the surface some type of activated [N*--OH]* complex:
Nz
+ OH'
[N2--OH]*
(4)
The hypothesis of the existence of a specific N2--OH interaction is supported by some experimental observations done in this work. The ESR spectra reported in Figure 7 show a decrease of the iron signals for a (CP)0.2 sample irradiated under a flow of pure dinitrogen. On the contrary, in the absence of dinitrogen, longtime irradiation of the same sample under flow of wet He does not lead to a decrease in the Fe3+ ESR signal. This finding suggests that, when the only surface radical species formed is OH', the recombination reaction Fe2+
+ OH'
-
Fe3+
+ OH-
(5)
dominates. Furthermore, the need of the presence of surface OHspecies for the dinitrogen activation is also supported by the observed inactivity of samples containing chloride ions which, as is well-known,29are better hole traps than OH- groups. In conclusion, all these findings suggest that, for the setting up of activity in T i 0 2 samples, two conditions must be verified: the presence of bulk Fe3+ions, necessary for separating the charge carriers by temporarily trapping the electrons, and the presence of surface hydroxyl groups, needed for trapping holes with subsequent activation of the dinitrogen molecule. Subsequently the activated surface dinitrogen species evolve toward NH3 formation by a series of unknown reduction steps. The low or null activity of samples containing Fe2TiOs can be explained, as previously reported9 by considering that this semiconductor has a quite small electron mobility. Finally, a comment can be given to the effect of excess Fe at the sample surface. Fe2Ti05and Fe203have no activity;' other Fe species at the surface seemingly act as poisons, but their specific role cannot be ascertained yet with certainty. It can however be observed that some change in the reaction mechanism may be induced by them, since in the IM samples, where these species (29) Boonstra, A. H.; Mutsaers, C. A. H. A. J . Phys. Chem. 1975, 79, 1694-1698.
are most likely to exist, the total activity increases with reaction temperature, while for CP samples the activity is less affected by an increase in reactor temperature. Possibly, some intermediate species in the reaction can become stabilized by the Fe ions at the surface, and its further reaction requires a sizable activation energy, while the latter would be negligible in the absence of such surface iron. Catalytic Nature of the System. The observed decline in activity can now be discussed. The parallel decrease in Fe3+ concentration, revealed by ESR, suggests that this could be due to consumption of these ions (which are necessary for activity) through reaction 2. However, the explanation cannot be this simple, since there is an initial period (5-7 h) in which both the amount of Fe3+and the activity remain unchanged. Comparison of the total amount of iron in the catalyst (36 pmol/g for (CP)0.2 sample), of which at most a small fraction would be transformed from Fe3+ to Fe2+ in the initial period according to ESR, with the amount of NH3 which is generated during the same period (30 pnol/g in the most favorable case, Le., the same sample at 353 K), shows clearly that the role of iron in the reaction is catalytic, not that of a reactant. This implies that, after Fe3+ initially captures an electron (reaction 2), the latter is able to return to the surface, probably reacting with one of the species which appear as intermediates in the NH! photogeneration process. Consequently, during this first period, some other property of the powder must be changing; otherwise, there would be no explanation to the fact that, after several hours of seemingly stable operation, the activity suddenly declines instead of remaining unchanged indefinitely. The length of the first period is roughly independent of iron content and amount of N, transformed, so that this change is more likely to be due to transformations involving only the UV light-TiO, interaction. Since IR shows a continuous decrease in the amount of surface O H groups of several kinds, we consider surface dehydroxylation as the most probable primary cause for the observed deactivation. Recombination of photoproduced surface OH' radicals to give H 2 0 2 has been proposed to occur in Ti02-based systems generating reduced products such as H2,30,31Hole-assisted deprotonation of surface OH- by water32or ammonia according to \ -TWH
+ H20 + h+
\ /
+ NH3 + h+
/
and -TiH
-
-C
-
\ * -Ti0
+
H30+ (6)
\ * /
+
NH:
/
- T i
(7)
followed in both cases by 2
1 Ti' /
-
\
0
0
\
-Tta-Ti
has been also suggested. A process involving two neighboring OH groups, i.e.
might be considered as well. Surface OH- groups would be thus eliminated in the h+-consuming part of the mechanism, and the overall process would change when either these OH- groups are exhausted (in which case the reaction would continue by using other types of hydroxyls) or their concentration falls below a certain limit. Since the IR results do not show that any particular type of OH- group is consumed at rates significantly higher than (30) Duonghong, D.; Gratzel, M. J . Chem. SOC.,Chem. Commun. 1984, 1597-1598. (31) Munuera, G.; Soria, J.; Conesa, J. C.; Sanz, J.; Gonz4lez-Elipq A. R.; Navio, A.; Lopez-Molina, E. J.; Mufioz, A.; Fernandez, A.; Espinos, J. P. Stud. Surf. Sci. Carol. 1984, 19, 335-346. (32) Ulmann, M.; de Tacconi, N . R.; Augustynski, J. J. Phys. Chem. 1986, 90, 6523-6530.
Soria et al.
282 The Journal of Physical Chemistry, Vol. 95, No. I , 1991
the other one(s), the second alternative seems more likely. Therefore, during the first hours of reaction, in which there is no net consumption of Fe3+, the overall stoichiometry could be N2
+ 3H20 2 2NH3 + 7 2 0 2
(10)
or a similar one. The reason for the change in the kinetics of the reaction would be then that, when the steady-state concentration of the key surface intermediate species (generated from the mentioned OH- groups) falls below a certain level, the process of back-transfer of electrons to them from Fe2+ions would become less efficient than the capture of photogenerated electrons by Fe3+, this leading to a progressive decrease in the concentration of the latter and, consequently, in the activity for NH3 photoproduction, which would become zero as soon as either Fe3+ or active OHspecies disappear. The overall process would then be describable as a superposition of reaction 10 and the following one: 2Fe3+
+
\
2 -TWH
+ 2H20
0
2FeB
+
hv
0
\ -T-Ti-
/
\
+
2H30*
(11)
It is not clear whether reaction 11 simply runs in parallel to ( I O ) or is linked mechanistically to it so that a stoichiometric relationship exists between NH,.produced and Fe3+ consumed in this stage of the reaction. During the latter, the total amount of NH, produced before complete deactivation (ca. 20 pmol/g for (CP)0.2 catalyst at 353 K), compared to the amount of iron in the sample, implies that, if the second hypothesis were true, the overall stoichiometry for the coupled mechanism could be
2Fe2+
+
\
0
/
\
2NH3 + -T-Ti-
+ 3/202 + 2H30+
(12)
This last point, however, is difficult to ascertain, especially since the two-reaction regimes may overlap somewhat and a net repartition of the amounts of NH, produced cannot be made. Conclusions The iron ions in the Ti02samples can be dispersed in the bulk of either the anatase or rutile forms as isolated paramagnetic species or can form ESR-inactive species, probably in mixed oxides, depending on the preparation method. The presence in the Fe-
Ti02 samples of reticular Fe3+ ions, which show an easier reducibility than that of Ti4+ ions, favors the charge separation of the carriers generated in Ti02by UV irradiation. The trapping of electrons by iron thus makes possible the reaction of N z molecules with surface OH species activated by holes. The UV irradiation of Fe-Ti02 samples produces the contemporary occurrence of two processes: the reduction of bulk Fe3+ ions, which becomes irreversible when the surface species activated by holes originate very stable species, and the irreversible consumption of surface hydroxyl groups. Both the disappearance of OH groups acting as hole traps and the extensive reduction of Fe3+ to Fe2+ can eventually determine the deactivation of the powder for the dinitrogen photoreduction. The results of this investigation indicate that the process of dinitrogen photoreduction to ammonia begins with the interaction of the dinitrogen molecule with highly reactive radicals generated on the surface by hole trapping, the role of iron being that of trapping electrons; very recently, some authors33have also hypothesized an oxidative pathway for the dinitrogen fixation over pure rutile. On the contrary, for explaining the dinitrogen photoreduction occurring on Ti02rutile, Schrauzer and Guth’ describe the electrons generated in the conduction band as excited titanium atoms in lower oxidation states at which the dinitrogen reduction occurs while Bourgeois et a1.I6 invoke the presence of defect states (as Ti3+ ions located at the bottom of the conduction band) generated by thermal treatment at 1300 K. The presence of such defect states in the powders used in this work is very unlikely as the heating treatment only favored the amorphous-anatase transformation. Work is in progress for detecting any such change in the reaction mechanism induced by the anatase-rutile phase transformation. Acknowledgment. V.A. thanks the “Consiglio Nazionale delle Ricerche” (Rome, Italy) for the award of a NATO Senior Fellowship to enable part of this wprk to be carried out at the “Instituto de Catdlisis y Petroleoquimica” CSIC (Madrid, Spain). He is grateful to all the researchers and the technical staff of the ‘Instituto de Cadlisis y Petroleoquimica” for the kind collaboration given to him during his stay. Financial support by M.P.I. (Rome, Italy) is also gratefully acknowledged. Registry No. N2, 7727-37-9; NH3, 7664-41-7; H20, 7732-18-5; TiO,, 13463-67-7; Fe, 7439-89-6. (33) Bickley, R. I.; Day, R.; Jayanty, R. K. M.; Navio, A. Proc. 9th Int. Congress on Catalysis; Chemical Institute of Canada: Ottawa, 1988; pp 15os-1s12.