J . Phys. Chem. 1990, 94, 5275-5282
5215
Chemlstry of Phosphomolybdate Adsorption on Alumina Surfaces. 1. The Molybdate/Alumlna System J . A. R. van Veen,* P. A. J. M. Hendriks, E. J . G . M. Romers, and R. R. Andrea KoninklijkelShell- Laborarorium, Amsterdam (Shell Research B. V.), Badhuisweg 3, 1031 CM Amsterdam, The Netherlands (Received: February 21. 1989; In Final Form: December 7 , 1989)
The adsorption of ammonium heptamolybdate (AHM) on y-A1203and a-AIOOH has been studied by using FTIR and Raman spectroscopies, TPR, and EXAFS. It is shown that molybdate adsorption on alumina cannot be explained in terms of electrostatic binding to protonated surface hydroxyls but can be rationalized by assuming (i) a reaction between the molybdate and basic surface OH groups, leading to the decomposition of the adsorbing molecule/ion, and (ii) physisorption on coordinatively unsaturated (cus) AI” sites. In the AHM/a-AIOOH case, surface precipitation, presumably of (NH4)2M03010, takes place at surprisinglylow initial AHM concentrations. The presence of (divalent) cations can influence the shape of the AHM/y-AI2O3 isotherm but does not affect the ultimate loadings that can be achieved through processes (i) and (ii).
1. Introduction In the context of alkene metathesis and the hydrotreatment of petroleum liquids there is a continuing interest in the detailed characterization of Mo/AI2O3 catalysts. Samples studied are often prepared from 7-AI2O3and aqueous ammonium heptamolybdate (AHM) solution, via, e.g., pore-volume impregnation or equilibrium adsorption, but surprisingly little is known about the chemical interactions taking place between them in the preparation step. The present work was undertaken in an effort to improve this situation. Over the years a variety of AHM/y-AI2O3adsorption isotherms have been published; the adsorption step was either thought to consist of a molybdate/surface O H condensation reaction’s6 or presumed to be due to an electrostatic interaction consequent on the surface charging by protonation of hydroxyl groups at pH < IEP (yA1203)(IEP = isoelectric p~int)!.~-l~ We have shown previously that the very high Mo loadings sometimes reportedz.3J1 are due not to AHM adsorption but to the formation of a (surface) precipitate.I2 The true adsorption isotherm, determined under near-neutral conditions, turns out to consist of two branches.12 On the basis of IR, Raman, and EXAFS data it could be shownI3 that the first branch is to be ascribed to an irreversible reaction of A H M with the basic hydroxyl groups of the alumina 8Al;OH
-.
+ M070246-
4(A1J2Mo04 + 3MoOd2-
+ 4H20 (1)
where subscript s denotes surfaces species. Adsorption in the second branch was proposed to involve physisorption of AHM on coordinatively unsaturated (cus) AI3+ sites13-a claim we should like to substantiate in the present paper. Obviously, the above mechanism, deduced, let it be repeated, (1) Iannibello, A.; Mitchell, P. C. H. Preparation of Catalysts II, Elsevier:
New York, 1979; p 469.
(2) Aulmann, M. A.; Siri, G.J.; Blanco, M. N.; Caceres, C. V.; Thomas, H. J. Appl. Coral. 1983, 7, 139. (3) Tsigdinos, G. A.; Chen, H. Y.; Streusand, B. J. Ind. Eng. Chem. Prod. Res. Deo. 1981, 20, 619. (4) Wang, L.; Hall, W. K. J . Catal. 1982. 77. 232. (5) Martinez, N . P.; Mitchell, P. C. H.; Chiplunker, P. J . Less-Common Mer. 1977, 54, 33. (6) Ratnasamy, P.;Sivasanker, S. Card Rev.-Sci. Eng. 1980, 22, 401. (7) Wang, L.; Hall, W. K. J. Caral. 1980, 66. 251. (8). Kasztelan, S.;Grimblot, J.; Bonnelle, J. P.; Payen, E.; Toulhoat, H.; Jacquin, Y. Appl. Caral. 1983, 7, 91. (9) Houalla, M.; Kibby, C. L.; Petrakis, L.; Hercules, D. M. J . Card. 1983, 83, 50. (IO) Luthra, N . P.; Cheng, W.-C. J . Catal. 1987, 107, 154. ( 1 1) Ciceres. C. V.: Fierro. J. L. G.:L h.e z Aaudo. - A.: Blanco. M. N.: Thomas, H. J. J. Catal. 1985, 95, 501. (12) (a) van Veen, J. A. R.; Hendriks, P. A. J. M. Polyhedron 1986, 5, 7 5 . (b) van Veen, J. A. R.; de Wit, H.; Emeis, C. A.; Hendriks, P. A. J. M. J . Catal. 1987, 107. 519. (13) Mensch, C. T. J.; van Veen, J. A. R.; van Wingerden, B.; van Dijk, M. P. J. Phys. Chem. 1988, 92, 4961.
0022-3654/90/2094-5275$02.50/0
TABLE I: The Number of cus AI3+ Sites and Basic OH Groups Present in the Aluminas Employed in This Study BET surface basic OH, cus AI3+, alumina area. m 2 d umo1.m-2 rrmo1.m-2 AI,O,-A 250 3.0 0.80 A1203-B 250 3.5 0.92 scs-79 95 3.2 1 .o boehmite 265 4.5 0.25
for AHM adsorption from near-neutral solutions, differs substantially from those proposed before. Consequently, it seems pertinent to inquire whether it is valid in general. In particular, the following questions need to be addressed: (i) does the same chemistry obtain at low pH, (ii) does the presence of Coz+or Ni2+ influence AHM adsorption, as observed in continuous adsorption experiment^,'^ and, if so, can the effect be understood in terms of our mechanism, (iii) how is the well-known interfering effect of phosphate on molybdate adsorption15J6to be interpreted, and (iv) are the surface sites involved in the adsorption of phosphomolybdates, recently described in the literature,17the same as those mentioned above, as expected on the ground that they are basesensitive polyanions just as heptamolybdate. The former two questions will be discussed in the present paper, while we leave the discussion of the latter two to a companion paper. The techniques employed here include FTIR and Raman spectroscopies, EXAFS, and TPR.
2. Experimental Techniques Supported Mo catalysts were prepared either via pore-volume impregnation or via equilibrium adsorption. For the determination of an adsorption isotherm, 2 g of AZO3was contacted with aqueous Mo-containing solutions (100 mL), and the mixture was shaken occasionally. The standard adsorption time was 2 days. The Mo uptake was determined by measuring the Mo concentration left in solution (atomic absorption spectrometry) and/or the Mo content of the solid (X-ray fluorescence). The desorbable fraction was determined by contacting the dried sample three times with demineralized water, the desorption time being 2 days in each case, and measuring the amount of Mo left in the solid. Three yAI203)s were employed: one from Rh6ne-Poulenc, SCS 79, of 9 5 m2.g-I, and two from Shell, denoted A1203-A and A1203-B,both having a specific surface area of 250 m2@. The concentrations of basic O H groups and cus AI3+ sites were determined via M o O , ( a c a ~ )(acac ~ = acetylacetonate) and Pd(acac),/Fe(acac), adsorption, respectively, as previously deFor comparative purposes, we also included a (14) Iannibello, A.; TrifirB, F. Z . Anorg. Allg. Chem. 1975, 413, 293. ( 1 5 ) Parfitt, G. D. Pure Appl. Chem. 1976, 48, 415. ( I 6) Gishti, K.; Iannibello, A.; Marengo, S.; Morelli, G.;Tittarelli, P. Appl. Catal. 1984, 12, 381. (17) Cheng, W.-C.; Luthra, N. P. J . Coral. 1988, 109, 163.
0 1990 American Chemical Society
5276
van Veen et al.
The Journal of Physical Chemistry, Vol. 94, No. 13, 1990 BASIC
BASIC
OH
3900
'
3700 I>.
0
i
2
0
1
3
4
3
io1 4 (Ai AHM.mmoI
c
L-'
Figure 1 . AHM/y-AI2O3 adsorption isotherms: (O,U)AI,O,-A; (A) A1203-B;(0)SCS-79. (O,A,O) near-neutral conditions, (R) pH = 2. Adsorption according to eq I i s indicated by the full curves.
boehmite, a - A 1 0 0 H sample (Catapal Condea, 265 m 2 g 1 ) . Alumina samples were calcined at 550 OC just before characterization or use in an adsorption experiment; in other words, the surface we are probing is that which is formed during cooling of the alumina from 550 OC to room temperature in ambient air. The boehmite sample is only dried at 120 O C . iR spectra were recorded on a Digilab FT instrument, using self-supporting disks of the sample (-8 m p " ) . Raman spectra were determined with a Spex Ramalog spectrometer. TPR profiles were measured with a laboratory-built instrument, the standard conditions k i n g as follows: 0.2-g sample, heating rate 5 K-min-I, 5% H2 in N2 at 6 mlmin-'. In the standard procedure samples were calcined in situ at 500 OC prior to the TPR run. For details on the EXAFS measurements we refer to an earlier paper.I3
3. Results and Discussion The amount of basic hydroxyl groups and the concentration of cus A13+ sites determined for our four alumina samples are collected in Table I. It is seen that the figures for the three yAIZO3's are similar, though not equal, while the a-AIOOH contains more basic OH, but less cus AI3" than a y-A1203,as expected on the basis of its structure.20 3. I . Molybdate Adsorption under Near-Neutral Conditions. The AHM/y-AI,O, adsorption isotherms, determined by using solutions which were neither acidified nor alkalified, are given in Figure 1. The part of the Mo uptake due to eq I , taking into account the experimentally determined amounts of basic OH groups, is indicated in the figure. That really only the basic hydroxyls are involved is evidenced by the top two IR spectra in Figure 2A: adsorption of 1.7% Mo on A1203-A,which should account for 45% of the basic groups present (Figure l ) , does in fact lead to a reduction in absorbance of the corresponding (Le., high frequency)I8 band in the IR spectrum by about 40%, while the other bands remain virtually unchanged. From a chemical point of view, indeed, one should expect that any heteropolyanion would primarily react with (and be depolymerized by) the basic hydroxyls of the alumina. For example, ammonium dimolybdate (ADM) should react with A120,-B according to 2A1,OH
-
+ Mo202-
(AI,),MoO,
+ MOO,*- + H,O
(2)
to lead to the same maximum Mo loading as in the A H M case (Figure I ) , and again the IR spectra fully support this expectation (Figure 2B): at 1% loading the absorbance of the high-frequency band has decreased by about 20% (expected: 24%). and at 5% (18) van Veen, J. A . R. J . Colloid InterfaceSci. 1988, 121, 214. (19) van Veen, J. A. R.; Jonkers, G.; Hesselink, W. H. J. Chem. SOC., Faraday Trans. i 1989, 85, 389. (20) Lippens, B. C.; Steggerda, J. J. In Physical and Chemical Aspects of Adsorbents and Catalysts, Linsen B . G.. Ed.; Academic Press: New York, 1970; Chapter 4
35t
cm-
A : AHM/A1203-A
3900
3700
3500 v , cm-'
E : ADM/Al,O,-E
Figure 2. FTIR spectra of Mo/Al20,. Spectra acquired after heating at 450 OC in vacuo.
loading the band has entirely disappeared, while the other two band systems are still relatively unaffected. Probably due to the relatively short adsorption time, the first plateau previously observed for adsorption times of the order of 2 weeks', has not developed. The adsorption over and above that of eq 1 in the region of the first branch is initiated by the dissociation of the basic OH still present on the surface, as discussed before,I2 and this is apparently a slow process. Consistently, the pH rise occasioned by these dissociating hydroxyls is not observed either. Samples corresponding to the first branch of the adsorption isotherms, described by eq 1, contain only tetrahedrally coordinated Mo, "Mo, as verified by EXAFS,') and are characterized by a rather weak Raman band at 910 cm-I (samples dried at 120 O C in vacuo). The Mo that adsorbs at higher loadings, i.e., in the second branch, which begins as soon as adsorption due to eq 1 has been completed, gives rise to a stronger Raman band at 950-960 cm-I, which can be ascribed to a polymeric aggregate of molybdenum-oxygen octahedra,,I the octahedral coordination having been verified by EXAFS.I3 The questions are now (i) which adsorption sites are responsible for the second branch, and (ii) can we deduce the precise structure of the adsorbed phase? ( i ) With the basic hydroxyl groups already accounted for, we are left with the neutral and acidic OH'S and the cus AI3+ sites as the possible adsorption centers for the second branch. It is well-known that the IR absorbance due to the nonbasic hydroxyls decreases strongly at higher Mo loadings; 22 as far as A1,03-A is concerned, at 6.7 wt % Mo the absorbance is only about one-third of that before Mo adsorption (Figure 2A, bottom spectrum). However, IR spectra are taken only after heating the sample at T b 450 O C , and in view of the known spreading of molybdenum oxides over the alumina surface,z3the decrease in absorbance may be due to a high-temperature condensation between adsorbed polymeric molybdate and the nonbasic hydroxyl groups rather than indicating the latter to be the primary adsorption sites. With reference to the two adsorption mechanisms mentioned in the Introduction, neither a molybdate/surface O H condensation nor a protonation of the surface hydroxyls to any great extentz4 would appear to be very likely in view of their nonbasic character. This would leave the cus AI3+ sites as the primary adsorption centers (21) Knozinger, H.; Jeziorowski, H. J. Phys. Chem. 1978, 82, 2002. (22) Millman, W. S.; Segawa, K.; Smrz, D.; Hall, W. K. Polyhedron 1986, S, 169.
(23) Stampfl, S. R.; Chen. Y.; Dumesic, J. A.; Niu, C.; Hill, C. G. J. Caral. 1987, 105, 445. Leyrer, J.; Zaki, M. I.; Knozinger, H. J. Phys. Chem. 1986, 90, 4775. Margraf, R.; Leyrer, J.; Knozinger, H.; Taglauer, E. SurJ Sei. 1987, 189, 842.
(24) Protonation of surface groups preferentially occurs with the basic ones,'* and these have already been reacted away in the first branch. A decrease in IEP is therefore to be expected, and experimentally observed: Llambias. F. J. G.; Castro, A. M. E.; Agudo, A. L.; Fierro, J. L. G. J . Coral. 1984, YO. 3 2 3 .
Phosphomolybdate Adsorption on Alumina
The Journal of Physical Chemistry, Vol. 94, No. 13, I990
5277
Mo(ads.)”/pmol. m-’
r
0
0.5
1 .o C.U.S. AI3+/fimol. m-’
Figure 3. Plot of amount of Mo, adsorbed in the second branch of the AHM/alumina isotherm, vs the number of cus A13+ ions in the alumina surface. Error bars correspond to 110% relative error in both quantities.
for the second branch. Since we can determine the concentration of these sites (Table I) and also the amount of Mo adsorbed in the second branch, Mo”(ads) (isotherms were extrapolated by plotting 1/C’vs l/Mo”), we can construct the required plot to test this idea. As Figure 3 shows, it is fully corroborated by the experimental results (the AHM/a-AIOOH isotherm is discussed in section 3.4). We have previously shown that MoO,(acac), adsorption involves exactly the same adsorption sites as those now identified for AHM adsorption: basic OH groups and cus AI3+sites.I9 We should expect, therefore, that no or, seeing that Mo” adsorption is not quite complete, very little MoO(acac), will adsorb (from MeOH solution) on, e.g., the AHM/AI2O3-A sample with the highest loading. In the event, a slight increase of only 0.5 wt % in Mo loading is observed, again corroborating the AHM-adsorption mechanism proposed here.25 (ii) As to the structure of Mo”(ads), taking into account the nature of the adsorption sites and the absence of any substantial pH change consequent on A H M adsorption,12 it would appear reasonable to suppose that the second branch is simply due to physisorption of AHM molecules, so that the latter as such constitute the Mo” adsorbed phase. This is in agreement with the observation that the Raman spectra of wet and dried samples are quite comparable to the one of the AHM solution,8 though EXAFS indicates that after drying the structure of Mo”(ads) is not quite that of crystalline AHM.” If the present hypothesis is correct we should expect the adsorption in the second branch to be reversible. The result of the pertinent experiment, involving contacting 1 g of Mo/AI20,, dried at 120 O C in air, three times with 100 mL of H 2 0 (cf. section 2), is shown in Figure 4. It is seen that adsorption in the first branch is totally irreversbile and that adsorption in the second branch is largely, though not totally, reversible. We surmise that the latter would have been entirely reversible but for the drying step, the temperature of 120 O C apparently being already high enough to induce some chemical bonding between the AHM molecules and the subjacent y-A120326 Thus, the second branch in the AHM/y-Al2O3 adsorption isotherm can be accounted for by physisorption of AHM molecules on cus AI3+ surface sites. Note that the present scheme implies that monomolybdate does not adsorb on these sites! ( 2 5 ) It may of course be asked whether cus AI3+sites can still exist when alumina is suspended in an aqueous medium. In our opinion, the answer is in the affirmative, since the first branch of the M002(acac)~/y-A1,0~ adsorption isotherm, Le., the one corresponding to the Mo02(acac)2/cusAI3+ r e a ~ t i o n , ’is~ entirely .~~ unaffected by having the pores of the freshly calcined alumina prefilled with water. This is, although water molecules will in all probability coordinate to the cus AI3+sites, this apparently does not lead, at least in the short term, to (re)hydroxylation of the alumina surface. (26) Jeziorowski, H.; Knozinger, H. J . Phys. Chem. 1979, 83, 1166.
”
0
~
2
4
6 8 C’AHM/mmol.1-’
Figure 4. AHM/AI,O,-A adsorption isotherms: (-) as such, (A)after washing. TABLE 11: Mo Adsorotion Levels at Neutral DH. and DH = 2 Mo’(ads), Mo”(ads), Mo(ads), mgg-’ mg-g-’ at pH N 5 at p H = 2 mgg-’ -y-AI2O3 A1203-A 38 32 70 93 42 38 A1203-B 80 107 scs-79 14.6 14.3 29 ~~~~
~~~
~~
3.2. Molybdate Adsorption at p H = 2. The AHM/A1203-A adsorption isotherm, determined by using AHM solutions acidified to pH = 2 with concentrated HN03, is shown in Figure 1 (AHM M have been used to avoid concentrations lower than 9 X precipitate formation). Only one plateau is now observed, while the isotherm is much steeper than that for neutral conditions. Similar results have been obtained for A1203-B. The ultimate adsorption level at pH = 2 is slightly over 30% higher than that at pH N 6 (Table 11) in agreement with the literature!v8 Taking into account that Mo exists at low pH in the form of M080264ions:’ this modest increase in adsorption capacity already indicates that surface-charge compensation does not play a major role in molybdate adsorption: at pH = 2 the alumina surface should be expected to be more highly charged than under near-neutral conditions, and with the charge per Mo in octamolybdate being almost twice as low as in heptamolybdate, the adsorption level would be predicted to increase by at least a factor of 2. Strong (co)adsorption of NO3- would, of course, vitiate this conclusion, but the IR spectra given in Figure 5, showing that at high Mo loadings hardly any nitrate is present in the sample, indicate that this does not occur.28 To study the adsorbed species we have applied TPR. The important parameters for our purpose are Tmax,the temperature at which the reduction rate is maximal, and R, the extent of reduction, here defined, arbitrarily, as the number of moles of H2 consumed per mole of Mo in the temperature range 200-665 OC. R is low for monomeric Mo species and increases with increasing aggregate size of the adsorbed phase,29reaching 1.0 or somewhat ~
~~
~~~~~
(27) Tytko, K.-H.; Glemser, 0. Adu. h o g . Chem. Radiochem. 1976.19, 239. (28) The nitrate ion can be considered to coordinate to (i) protonated basic OH group^,^' and (ii) cus AI3+ ions. This is in quantitative agreement with the results reported in Figure 5 . The absorbancesat 1390 cm-’ of the samples A-D are, respectively, 0.18, 0.085, 0.050, and 0.015. On the basis of the interpretation of the AHM/y-AI2O3adsorption results at pH = 2, to be given in section 3.2, in sample B 25% of the basic OH groups and all of the cus AI3+ sites are still present (Le., not covered by molybdate), in sam le C only the latter remain, and in sample D no basic OH groups or cus AI3 sites are left. Thus, we deduce for the number of cus AI3+sites (0.0500.0l5)/0.18 = 0.19 m m o l g l , and for the number of basic OH groups 4 (0.085-0.050)/0.18 = 0.78 mmol-g-I, in excellent agreement with the figures quoted in Table I. (29) Thomas, R.; De Beer, V. H. J.; Moulijn, J. A. Bull. Soc. Chim. Belg. 1981, 90, 1349. Arnoldy, P. Ph.D. Thesis, Amsterdam, The Netherlands, 1985; Chapter 4.
P
5278 The Journal of Physical Chemistry, Vol. 94, No. 13, I990
van Veen et al. REDUCTION RATE/a.u 300
400
500
I
I
5
60
1
1500
1400
1300 uicm-'
Figure 5. FTIR spectra of adsorbed NO,-: (A) KNO,/AI,O, containing 1 mmo1.g-I nitrate: (B-D)Mo/AI,O, ex AHM adsorption at pH = 2. Mo loadings ( m g - g ' ) : B, 40; C, 60: D, 93 (max).
higher for ( S O 2 supported) MOO, crystallite^.^^ TPR profiles and a plot of R vs Mo(ads) for some AHM/AI2O3-A catalysts prepared at pH N 6 are shown in Figures 6A and 7, respectively. We make the following remarks: (i) At Mo loadings of 20 and 30 mg-g-' we have only adsorbed MOO^^- and R is indeed low (about 0.2). From Figure 6A we may derive for T,,, a value of about 550 OC for this species. (ii) At 40 mg@ we still have virtually only adsorbed and, sure enough, R is still low. The TPR pattern, however, shows some additional structure, perhaps due to slight sintering during calcination. (iii) At 60 mgg-l only part of the adsorbed Mo is due to MOO^^-, Mo7O2?- coming into play as well. At 68 m g g ' this latter species, or, rather, the species that results from heating AHM/A1203 to 500 O C 3 ' will become even more prominent. That is, R will increase continuously for Mo loadings >38 mg-g-I, as observed. T,,, for the AHM species would seem to be about 435 "C, and its R value can be estimated to be about 0.45. The TPR results for the pH = 2 catalysts are summarized in Figures 6B and 7 . At 20 m g g ] we observe a species having a T,,, of approximately 550 O C with R = 0.2, which should therefore be adsorbed monomolybdate. At intermediate loadings, 40 and 60 mgg-l, another species enters the picture, characterized by T,, = 435 O C and R = 0.44,which can therefore be identified as adsorbed HM. At the highest loadings a third species is observed, for which T,,, appears to be about 450 O C and R is estimated to be 0.72, and which we presume to be OM (octamolybdate, M%OZ6&).This interpretation is corroborated by some Raman spectroscopic results. Raman spectra of AHM solutions of pH = 5 and pH = 2 show that HM is characterized by a band (30) de Jong, K. P.; Glezer, J. H. E. Unpublished results. (31) It is observed that the shape of the TPR profile and T,,, depend on the calcination temperature, but, fortunately, R does not. We note in passing that the present interpretation of the TPR results is close to that given in ref 29 as against: Burch, R : Collins, A Appl. Cuful 1985. i8. 389.
80
100
T. OC 600 I
120
140 160 TIME, min
Figure 6. TPR profiles of Mo/A1203 catalysts: (A) A H M adsorption at pH N 5-6; (B) adsorption at pH = 2. Numbers in parentheses indicate the Mo loading in mg-g-'.
=
0.6
0.4
-
0.2 -
0 0
20
40
60
80 100 Mo(ads.)/mg. g-'
Figure 7. The H 2 / M o ratio R as a function of Mo loading: (A)adsorption under near-neutral conditions: (0)adsorption at pH = 2.
at 950 cm-I, while the main band of OM also occurs at 950 cm-I, in agreement with the observation of Kasztelan et a1.* (Murata and Ikeda quote a value of 970 cm-l, however32), but a second band is observed at 900 cm-I. The Raman spectra of the OHM/A1203 (pH = 2) samples with 60 and 93 mgg-I, respectively, are given in Figure 8. Only the latter sample shows a well-developed band/shoulder at 900 cm-I, in agreement with the above. It appears, then, that even in the presence of HNO, the basic O H groups on the Al,03 surface react preferentially with AOM, (32) Murata. K.: Ikeda, S. Polyhedron 1983, 10, 1005.
The Journal of Physical Chemistry, Vol. 94, No. 13, 1990 5279
Phosphomolybdate Adsorption on Alumina COUNTS1a.u.
Mo(ads.)/mg. g-' 942 I
"
0
1
3
2
4
C'AHM/mmol. 1-l
Figure 9. AHM/AI2O3-A adsorption isotherms in the presence of M2' ions: (A)Co2+/Mo2' = 0.2 mol/mol; (0)Mg2'/Mo2+ = 1.4 mol/mol; ( 0 )Ni2+/Mo2' = 1.4 mol/mol; (-) isotherm in the absence of M2'. M(ads.)/mg.g-
'
L
6 - 5
0 ~
1100
4
8
1
2
I 900
700
vicm-' Figure 8. Raman spectra of (A) 60 m g g ' Mo/A1203(pH = 2), (B) 93 mgg-' Mo/A1,03 (pH = 2), and (C) 56 m g g ' Mo/AI2O3:HNO3(pH = 2). Samples were dried in vacuo before spectra were taken.
depolymerizing it all the way to monomolybdate at low initial AOM concentrations, and to a mixture of MOO^^- and H M at intermediate ones. From the present TPR and Raman results we deduce that one half of the basic O H groups (of A1203-A)leads to adsorbed monomolybdate, giving Mo,'(ads) = 19 mg@, and the other half to adsorbed heptamolybdate, giving Mo2'(ads) = 44 mg-g-I. This leaves 30 m g g ' to be accounted for, in the absence of surface-charge compensation, by adsorption of AOM on cus AI3+ sites, which is about the same amount as observed under near-neutral conditions (Table 11). In agreement, on Ti02, where Mo adsorption takes place mainly on the cus (Ti4+) sites,'2b 5 and pH = 2 are virtually the maximum Mo loadings at pH Upon preconditioning the alumina in aqueous H N 0 3 of pH = 2, which will remove a substantial fraction of the basic O H groups through protonation,Is the Mo loading decreases to 56 mgg-' (and S A to 210 m2-g-I), again indicating that surface-charge com-
pensation is not an important Mo-adsorption mechanism. The TPR profile, in showing an asymmetric peak with a maximum at T = 450 OC and an R value of 0.59, indicates that the adsorbed phase is a fifty-fifty mixture of AM and OM. This can be rationalized by assuming that 30% of the original basic OH groups are still capable of reacting with OM to form adsorbed HM, giving Mo,'(ads) = 26 mgg-I, and that O M adsorbs on the original cus AI3+sites, expected to be still there, giving Mo"(ads) = 30 mgg-l. The Raman spectrum of this sample is shown in Figure 8. The pronounced shoulder at -900 cm-' indicates the presence of O M (vide supra), although we would perhaps have expected it to be somewhat more intense. Another way to remove the basic O H groups is through exchange with F ion^.'^,^^ The observed decrease in IEP,about 2 pH is less than expected on the basis of our proposition (33) Yamagata, N.; Owada, Y.; Okazaki, S.; Tanabe. K. J. Caraf. 1987, 47, 358.
0
4
8 12 M ~ CONC., + mmol.
L-'
Figure 10. Influence of molybdate on the adsorption of M2': ( 0 , O ) C O ( N O ~ )(A,A) ~; Ni(NO,),; neutral conditions. Open symbols: no A H M , filled symbols: A H M present (Co/Mo molar ratio in the adsorption solution is initially 0.2). Inset: Final pH as a function of initial [Co2'] for the system C O ( N O ~ ) ~ / ~ - A I ~ O ~ .
that the basic hydroxyls are the primary protonation sites of the alumina.ls We have previously found, however, that treatment of AI2O3 with NH4F, while removing all basic O H groups, as evidenced by IR spectroscopy, creates other basic sites, capable of decomposing such base-sensitive complexes as M ~ O ~ ( a c a c ) ~ . ' ~ Adsorption of AHM at pH = 2 on a fluorided sample leads to a Mo loading of about 65 mg-g-l. The TPR profile shows a relatively symmetric peak, but T,,, is rather high at 495 OC; apparently, the presence of F has a negative effect on the reduction rate of the adsorbed Mo species. Under the circumstances, it is unclear whether the relatively low value of R, 0.48, implies that little OM(ads) is present. A last experiment to verify that the cus AI3' sites are in fact covered by adsorbed molybdate upon AHM adsorption at pH = 2 involved contacting a Mo/AI2O3 sample of maximum loading with a MeOH solution of Mo02(acac),. As expected, no adsorption of the complex was observed. We conclude that molybdate adsorption at pH = 2 involves the same surface sites as under near-neutral conditions, viz., basic hydroxyls and cus AI3+. 3.3. A H M Adsorption in the Presence of Divalent Cations. As shown in Figure 9, the presence of a modest amount of Co(NO,), in the AHM adsorption solution, Le., at a Co/Mo atomic ratio of 0.2, does not appear to influence the AHM adsorption process. On the other hand, cation adsorption is slightly greater in the presence of AHM than in its absence, Figure 10. However, upon increasing the Co or Ni(N03)2concentration to a cation/Mo atomic ratio of 1.4, a change in the shape of the AHM adsorption (34) Vordonis, L.; Koutsoukos P. G.: Lycourghiotis, A. J . Chem. SOC., Chem. Commun. 1984, 1309. (35) Mulcahy, F. M.; Houalla, M.; Hercules, D. M. J . Card 1987, 106, 210.
5280 The Journal of Physical Chemistry, Vol. 94, No. 13, 1990 C OU N TSr a .u
van Veen et al. lose out, however, and AHM adsorption is but little disturbed, but at high Ni concentrations competition becomes effective. So, as far as the first branch is concerned, we then have, apart from reaction 1, also
918 945
+ Ni2+
AIs-OH
-
AI,+
+ Ni(OH)+
(3) The AI: sites created in this way subsequently accept Moo4*ions formed in reaction 1, according to
+
2AIs+
(iii)
(ii) 1
I
1000
900
800
700 Y,
cm-'
Figure 11. Raman spectra of coadsorbed NiMo/AI2O3samples. Mo loadings (mgg-l): ( i ) 27, (ii) 40, (iii) 64, (iv) 7 3 . Ni/Mo atomic ratio in the adsorption solutions was 1.4.
isotherm is observed, although the Mo(ads) plateau value is not greatly affected, Figure 9. While the presence of Ni2+ and Co2+ has much the same effect, addition of Mg(N03)2leads to a much less important change of shape, cf. Figure 9. Since Mg2+is less hydrolysis-prone than Co2+ or Ni2+ (see ref 36), this finding indicates that it is the formation of species such as Co(OH)+ and Ni(OH)+ that plays the essential role here. The data reported in the inset of Figure 10 indicate the formation of Co(OH)+ species to take place at pH around 6 (cf. ref 36). To construct a plausible coadsorption mechanism we need some information on the structure of the adsorbed phase. Raman spectra of some NiMo/AI2O3 samples, dried at 110 O C , are given in Figure 1 1. At Mo loadings of 27 and 40 mgg-] a band at 918 cm-' is observed, and we conclude that monomolybdate ions constitute the adsorbed phase, just as in the no-Ni case at comparable Mo-adsorption levels. At higher Mo loadings, a band at 945 cm-I appears, which may be ascribed to adsorbed HM, or perhaps, in view of the broad shoulder at about 890 cm-' and the (small) band at about 550 cm-I (not shown), to some NiMo heter~polyanion~'(which would also account for the observed increased Ni2+adsorption). In short, the Mo-adsorption mechanism in the presence of Coz+and Ni2+appears to be quite similar to the one obtaining in their absence. The steepness of the AHM adsorption isotherm in the presence of relatively large amounts of Co or Ni(NO,), can now be explained as follows: Ni2+ions can compete with AHM for the basic O H groups on the A1203surface. At low Ni concentrations they ( 3 6 ) Kragten, J. Ailas of Metal-Ligand Equilibria in Aqueous Solution; Ellis-Horwood: Chichester, U.K., 1978. (37) Jeziorowski, H.; Knozinger, H. Appl. Surfi Sci. 1980, 5. 335
-
(AI,),MoO,
(4)
Thus, while the limiting monomolybdate loading remains unaffected, as suggested by the Raman results discussed above, the isotherm is much steeper, as less MOO^^- remains in solution. The second branch, however, is also considerably steeper in the presence of large amounts of Ni2+ (or Co2+). This is perhaps due to a decreased AHM solubility, in analogy with the situation at pH = 2, where AHM solubility is much lower than at near-neutral conditions, and the adsorption isotherm very steep (Figure I , and ref 12b). 3.4. AHM Adsorption on a-AIOOH. The AHM/a-AIOOH adsorption isotherm, obtained under near-neutral conditions, is shown in Figure 12. It appears to consist of three branches. Branch (i) is, as in the 7-A1203case, due to the reaction between AHM and the basic O H groups on the surface of the boehmite, eq I , to form adsorbed monomolybdate (Raman band at -920 cm-l; see inset in Figure 12). Remarkably, the ultimate adsorption level predicted by eq 1 is not reached. Apparently, the geometric distribution of the basic O H groups differs from that obtaining for 7-A1203in such a way that they do not all form pairs that can be exchanged for MOO^^-. Branch (ii), in analogy with the 7-A1203case, is taken to be due to physisorption of AHM on cus AI3+sites. The EXAFS spectra of the 4.5 and 4.7 Mo samples, corresponding to the first two points of branch (ii), are compared with those of 3.5% Mo/A120, (tetrahedral Mo onlyI3)and of 5.6% Mo/AI2O3 (ratio tetrahedral-to-octahedral Mo of 213) in Figure 13a. Qualitatively, a comparison of Figure 13a (A) and 13a (D) shows that the transition from pure tetrahedral Mo to a mixture of tetrahedral and octahedral Mo manifests itself most clearly in a doubling of the maximum at -80 nm-' in the k 3 x ( k ) oscillatory pattern and in the appearance of a peak at -0.3 nm in the raw distribution function. It is exactly these features that are also present, albeit to a lesser extent, in Figure 13a (B) and (C), reporting the Mo/AIOOH data, from which it can be concluded that some octahedral Mo (deriving from adsorbed AHMI3) is present in these samples. A quantification was not attempted. The steplike character of branch (iii) implies that an adsorbate-adsorbate interaction is involved,38i.e., in the present system, the formation of a (surface) precipitate. It is known that, in concentrated A H M solutions, on the time scale of the present adsorption experiments, a precipitate of (NH,), Mog02,develops.39 We have found that in the presence of 7-A1203precipitate formation occurs at lower AHM concentrations, the structure of the precipitate being, moreover, not quite that of the octamolybdate quoted above.'2b Branch (iii) of the AHM/a-AlOOH adsorption isotherms occurs at still lower AHM concentrations, so apparently boehmite catalyzes the formation of a precipitate even more efficiently than 7-A1203. The precipitate formed in the presence of y-Al,O, has Raman bands at 950 and 910 em-', it does not readily dissolve in H 2 0 , and its formation is accompanied by a rise in pH. The one formed in the presence of (Y-AIOOH,on the other hand, has a single Raman band at -950 cm-', cf. inset in Figure 12; its formation is not associated with a shift in pH, and it can be (partially) removed by washing with H 2 0 , Figure 12. To convince ourselves that we really have a precipitate, and that it is not a question of AHM being somehow entrained by the boehmite, we have conducted an EXAFS study of some highloading Mo/a-AlOOH samples. From a comparison of the EXAFS oscillating patterns and corresponding raw radial distribution (38) Davis, B. W.; Pierce, C. J . Phys. Chem. 1966, 70, 1051. (39) Glemser, 0.;Wagner, G.; Krebs, B. Angew. Chem. 1970, 82, 639. Tytko. K. H. Z . Naturforsch. B 1973, 28, 272.
The Journal of Physical Chemistry, Vol. 94, No. 13, 1990 5281
Phosphomolybdate Adsorption on Alumina
[ j, Moladr.l/%w
INTENSITY,
i’xlkl ~
I.“.
!L I
500
INTENSlTY
r;
,,b---o-
A922cm-1
4
12
1
0 2 -4 -6 20
1000
40
60
80
1W
120 140
160
0 1 0 2 0 3 0 4 05 06 07 08
A
R . nm
k nm-’
4
2
200
0
t/
100
-2
W500
-4
0 0
1
I
I
1
1
2
3
4
30
50
70
90
I-’
functions, Figure 13b, it is clear that the branch (iii) adsorbed phase in no way resembles AHM but that its local structure is quite similar to that of the yA1203precipitate: cf. especially the succession of small and large maxima which occurs in the 75140-11m-~region of the oscillatory patterns (B)-(D), but not in (A), and the single peak at 0.3 nm in the raw distribution functions (9)-(D) as against the double peak observed for AHM ((A)). The present data are not sufficient to unequivocally assign a structure to the a-AIOOH precipitate, but from Tytko and Glemser’s review of the solution chemistry of oxomolybdates (cf. Pope, ref 40), a suitable candidate would appear to be the species denoted by (NH4)20-3M002(of unknown structure, unfortunately). The adsorption process related to branch (iii) could then be written as
-
2(NH4)2M030101+
l;:p,
-6 30
50
70
90
110 130 150 170 t nm-’
I 20
40
60
80 100 120
~
1
0 4 05 06 07 08 R , nm
50 0 01 02 03 0 4 05 06 0 7 08
140 180 k
R nm
nrrl
;;E INTENSITY a u
A
1w 50
-8
0
20
60
40
80
1M) 120 140
160
0
t , nm-’
0 1 02 03 0 4 05 06 0 7 08 R nm
4
3 2 1
0 -1
-2 -3 -4
30
50
70
90
110 130 150
170
4. Conclusions
(40) Pope, M. T.Heteropoly and Isopoly Oxometallates;Springer: Berlin,
,
0 1 02 03
D
k , nm-’
1. Adsorption of AHM on y-A1203 from aqueous solutions takes place via two processes: (i) irreversible reaction with basic surface O H groups to form adsorbed monomolybdate, and (ii) largely reversible physisorption on cus AI3+ sites. 2. Adsorption of AOM on 7-AI2O3 at pH(initia1) = 2 is not due to electrostatic adsorption on protonated surface hydroxyls but involves the same surface sites as above. The reaction with basic OH groups with O M leads to a mixture of H M and monomolybdate. This implies that O M effectively competes with H+ for these groups. 3. The first two branches of the AHM/boehmite adsorption isotherm correspond to those observed for y-A1203. The third one is not due to a regular adsorption process but stems from the largely reversible formation of a surface precipitate, which is tentatively identified as (NH4)2M03010. 4. The presence of M2+ ( M = Co, Ni) in the AHM solution does not affect the AHM adsorption process at relatively low concentration, M/Mo = 0.2-0.3 mol/mol, but changes the shape
‘0
100
(5)
The presence of Co- or Ni(N03)2in the adsorption solution at a Co(Ni)/Mo molar ratio of 0.3 has little influence on the shape of the A H M adsorption isotherm. The isotherm determined at pH = 2 is also shown in Figure 12. The EXAFS spectrum of a sample containing 10.4 wt % Mo is virtually identical with the one of the corresponding sample obtained under near-neutral conditions, implying that at pH = 2 the regular adsorption and (surface) precipitation processes are not well separated. In contrast to the y A 1 2 0 3case, the TPR profiles of Mo/aAIOOH samples, after calcination at 500 O C , which of course destroys the boehmite lattice, do not contain well-defined peaks.
1983.
8
k. nm-’
1
5 c’A,,/mmoi.
Figure 12. AHM/a-AIOOH adsorption isotherms. Neutral conditions: (o,o,@,O) (different series); (W) after washing; the solid line indicates adsorption according to eq 1. (A)pH = 2. Inset: Raman spectra of Mo/a-AIOOH (pH N 6). (a) 4.2, (b) 11.5 wt % Mo.
4NH4+ + Mo70246-
01 02 03 04 E 0 5 0 6 0R7 nm 08
110 130 150 170
5 4 3
2
;E 0
01 02
03
0 4 05
06 07 08 R nm
C
1
0
100
-1
-2
50
-3 -4
30
50
70
90
110 130 150 l;0 k , nm-
0
0 01 0 2 03 0 4 05 06 07 0 8 R nm
D
-6 20
40
60
80
100 120 140 k nm
160 R Im
Figure 13. EXAFS oscillatory patterns k 3 x ( k ) ,with k reciprocal wave vector (nm-I) and corresponding raw radial distribution functions: (a, upper) (A) 3.4 wt % Mo/AI2O3, (B) 4.5 wt % Mo/a-AIOOH, (C) 4.7 wt % Mo/a-AIOOH, (D) 5.6 wt % Mo/AI20,. (b, lower) (A) A H M , (NH.J,MO,O~,*~H~O, (B) 10.6 wt % Mo/a-AIOOH, (C) 11.7 wt % Mo/a-AIOOH, (D) 21.4 wt % Mo/AI,03 (mainly precipitate).
5282
J . Phys. Chem. 1990, 94, 5282-5285
of the isotherm at relatively high concentration, M/Mo = 1.4 mol/mol. This is attributed primarily to M2+ at high concentrations being able to compete for the basic surface O H groups. Acknowledgment. We thank Dr. M. P. van Dijk and Mr. C.
T. J. Mensch for the EXAFS measurements, and Mr. R. de Ruiter for the data reported in Figure 5. Registry No. A H M , 12027-67-7; A120,, 1344-28-1; Mo. 7439-98-7: A I O O H , 24623-77-6.
Chemistry of Phosphomolybdate Adsorption on Alumina Surfaces. 2. The Molybdate/Phosphated Alumina and Phosphomolybdate/Alumina Systems J. A. R. van Veen,* P. A . J. M. Hendriks, R. R. AndrCa, E. J. G. M. Romers, and A. E. Wilson Koninklijke/Shell- Laboralorium, Amsterdam (Shell Research B. V.), Badhuisweg 3, 1031 CM Amsterdam, The Netherlands (Received: February 21, 1989: I n Final Form: December 7 , 1989)
The adsorption of ammonium heptamolybdate (AHM) on phosphated alumina and of PzM05023~and PMo,,0m3- on y-A1,03 has been studied by using FTIR, Raman, and 31Psolid-state NMR spectroscopies, and TPR. Adsorption of phosphate prior to contacting of y-Al,03 with AHM leads to a deactivation of the alumina surface still remaining. Phosphate reacts with y-Al,O, to form an AIPO4-type surface phase, which itself is capable of adsorbing molybdate, with the formation of an irreducible surface molybdophosphate, its adsorption capacity being lower, however, than that of the original alumina. The primary phosphomolybdate adsorption reaction on alumina is shown to be the same as that observed in the case of AHM, viz., a reaction with the basic surface OH groups, leading to the decomposition of the adsorbing species. Thiophene hydrodesulfurization (HDS) activity data of variously prepared Mo(P)/A1203 catalysts are in agreement with the idea that HDS activity increases with increasing Mo reducibility.
Introduction In the previous paper,' it has been shown that the adsorption of poly(oxomolybdates) on yA1203(and a - A I 0 0 H ) involves two processes: in the first instance, we have a reaction with the basic surface hydroxyl groups, leading to depolymerization of the sorbing poly(oxomolybdate) ion and adsorption of (part of) the resulting monomolybdate ions, eventually followed by physisorption on the coordinatively unsaturated (cus) AI3+sites. In the present paper, we discuss the mechanism of the well-known interfering effect of phosphate on molybdate ad~orption,~J and determine whether the same surface sites, basic O H groups and cus AI3+ sites, are involved in the adsorption of phosphomolybdates, as expected on the ground that they are base-sensitive polyanions just as heptaand ~ctamolybdate.~ Since a plethora of phosphomolybdate species exists in solutions containing both phosphate and molybdate at pH 5 7,4-7 and since it is known that phosphate interacts more strongly with y-A1203than m ~ l y b d a t e , it~ ~was ~ ~decided * to study in the first instance the adsorption of AHM on alumina samples which had previously been exposed to phosphate ((NH4)H2P04), rather than the coadsorption of phosphate and molybdate. The interfering effect of phosphate on molybdate adsorption could simply mean that the former species interacts with the same alumina surface sites as the latter, but more strongly. In current adsorption models, indeed, phosphate is pictured as adsorbing either covalently or electrostatically on protonated (and, therefore, On the other hand, originally basic9) surface hydroxyl (I! van Veen, J. A. R.; Hendriks, P. A. J. M.; Romers, E. J. G. M.; Andrea, R. R. J . Phys. Chem., preceding paper in this issue. (2) Parfitt, G. D. Pure Appl. Chem. 1976, 48, 415. (3) Gishti, K.; Iannibello, A.; Marengo, S . ; Morelli, G.;Tittarelli, P. Appl. Catal. 1984, 12, 381. (4) Pope, M. T. Heteropoly and Isopoly Oxometallates; Springer: Berlin, 1983. ( 5 ) Souchay, P. folyanionr and Polycationr; Gauthier-Villars: Paris, 1963. (6) Van Veen, J. A. R.; Sudmeijer, 0.;Emeis, C. A,; de Wit, H. J . Chem. SOC.,Dalton Trans. 1986, 1825. (7) Pettersson, L.; Andersson, I.; Ohman, L.Inorg. Chem. 1986, 25, 4726. (8) Hsu, P. H. Minerals in Soil Enuironments; Soil Science Society of America, Inc.: Madison, WI, 1977; Chapter 4. (9) Van Veen, J. A. R. J. Colloid Interface Sci. 1988, 121, 214. ( I O ) Muijadi, D.; Posner. A . M.; Quirk, J. P. J . SoilSci. 1966, 17, 212.
0022-3654/90/2094-5282$02.50/0
it has been reported that phosphate eliminates the strongly acidic sites of alumina3.I2and that treatment of ZSM-5 zeolites with, e.g., H3P04 or (NH4)2 HPO,, decreases the concentration of strong Bransted ~ i t e s , ' which ~ J ~ suggests that phosphate primarily reacts with acidic hydroxyls. Our own results concerning (NH4)H2P04 adsorption on AI2O3-Aindicate that in fact both basic and acidic O H groups are involved. The adsorption of some phosphomolybdates has been recently described in the 1iterat~re.l~In that paper, 12-molybdophosphate, PMO,~O~;-,has been reported to adsorb intact on an alumina. This would be very surprising, if true, since PMolz is not a very stable compound6 and should be at least as susceptible to depolymerization by surface hydroxyl groups as heptamolybdate. To ascertain whether the various adsorbed phases encountered in this study differ at all in their catalytic properties, we have measured their specific activity in the hydrodesulfurization of thiophene. Experimental Section To study the alumina-phosphate interaction, 2-g samples of freshly calcined y-A1203(denoted A1,03-A in the previous paper') were contacted with aqueous (NH4)H2P04solutions (100 mL), and the mixtures were shaken occasionally. Standard adsorption time was 1 day. The P uptake was determined by measuring the P content of the solid after drying at 120 O C in vacuo with X-ray fluorescence. An A1PO4/AI20, sample was prepared by contacting y-A120, with 1 M H3P04 for 2 days. Adsorption of ammonium heptamolybdate (AHM) on phosphated AI203 was effected as ( I I ) Mikami, N.; Sasaki, M.; Hachiya, K.; Astumian, R. D.; Ikeda, T.; Yasunaga, T. J. Phys. Chem. 1983, 87, 1454 and references therein. (12) Stanislaus, A.; Absi-Halabi, M.; AI-Dolama, K. Appl. Catal. 1988, 39, 239. ( I 3) Lercher, J. A.; Rumplmayr, G.; N o h , H. Acta Phys. Chem. 1985, 31, 71. (14) Ashton, A. G.; Dwyer, J.; Elliott, I.
S . ; Fitch, F. R.; Qin, G.; Greenwood, M.; Speakman, J. Proceedings of the 6th International Zeolite Conference Butterworths: London, 1984; p 704. (IS) Cheng, W.-C.; Luthra, N. P. J. Catal. 1988. 109, 163. (16) Van Veen, J. A. R.;Jonkers, G.; Hesselink, W. H. J . Chem. SOC., Faraday Trans. 1 1989, 85, 389.
0 1990 American Chemical Society