695
STRUCTURE AND PROPERTIES OF AMORPHOUS SILICOALUMINAS
Structure and Properties of Amorphous Silicoaluminas. 111. Hydrated Aluminas and Transition Aluminas
by A. J. L60nard,la F. Van Cauwelaert,lb and J. J. FripiatlO Labmatoire de Physico-Chimie Mintale, Agronomic Institute of the University of Louvain, HdverlCLouvain, Belgium (Received October 17, 1966)
The structure of amorphous aluminas and of some transition aluminas has been studied using the generalized Fourier synthesis. The comparison of electron density distribution functions obtained for hydrated gels and gels calcined a t various temperatures up to 700' with similar distribution functions computed from structure models has led to interesting conclusions. Amorphous hydrated gels obtained in acid or alkaline solutions behave differently on dehydration. Gels obtained in acid solutions evolve a t 300" toward a partially hydrated phase in which a spinel-like structure is realized in contrast with those precipitated in alkaline solution which acquire a boehmite-like structure at 300". At higher temperature, both series have a spinel-like structure but with an excess of tetrahedral aluminum cations in the case of the samples obtained initially in acid conditions. This suggests a differentiation between the 7 and y transition forms. The structural characteristics of hydrated alumina gels reveal a disorganization in the distribution of aluminum cations, the degree of disorganization being higher for the samples obtained by acid precipitation. This and the thermal behavior mentioned above are believed to result from the prolonged presence of aquo groups. The significance of the estimation of the relative content of tetrahedral aluminum cations from the shift of the A1 K a fluorescence line has been checked with respect t o the Al(1V) content in spinel-like structures. The agreement between both sets of measurements was reasonable when gibbsite and sodium zeolite were used as reference materials.
I. Introduction Transition aluminas are obtained from the dehydration of aluminum trihydrates (gibbsite, bayerite, and nordstrandite) or monohydrate (boehmite) according to various decomposition sequences. A great deal of effort has been devoted to the study of these sequences as shown in Table I.2-14 More details may be found in the excellent review published by Newsome, Heiser, Russell, and Stumpf.15 For the sake of clarity, the Alcoa nomenclature will be used in this paper. Among the reasons for the broad interest in this field are the high surface areas and the catalytic properties displayed by transition aluminas. Activation is produced by heating aluminum trihydrates or monohydrate (boehmite) a t rather high temperatures (500-900"). The two main transition aluminas used in catalysis are the 7 and y species. Maciver, et a1.,16
have studied particularly their surface properties and have concluded that 7 alumina differs from y ~~
~~
(1) (a) The University of Louvain; (b) Aspirant van het N.F.W.O.;
(c) The University of Louvain and M.R.A.C. (Tervurent). (2) (a) G. W.Brindley and J. 0. Choe, Am. Mineralogist, 46, 771 (1961); (b) J. F. Brown, D. Clark, and W. W. Elliott, J. Chem. Soc., 84 (1953). (3) H. C. Stumpf, A. 5. Russell, J. W. Newsome, and C. M. Tucker, Ind. Eng. Chem., B42, 1398 (1950). (4) H. Saalfeld, Clay Minerals Bull., 3, 249 (1958). (5) B. C. Lippens, Doctoral Thesis, Technische Hogeschool, Delft, The Netherlands, 1961. (6) U. Hauschild, 2. Anorg. Allgem. Chem., 324, 15 (1963). (7) E. J. W. Verwey and E. L. Heilmann, J . Chem. Phys., 15, 174 (1947). (8) E.Kordes, 2.Krist., 91, 193 (1935). (9) H. Ginsberg, W.Htlttig, and G. Strunk-Lichtenberg, 2. Anorg. Allgem. Chem., 293, 204 (1957). (10) H. P.Rooksby and C. J. M. Rooymans, Clay Minerals Bull., 4, 234 (1961).
Volume 71, Number 9 February 1967
A. J. L~~ONARD, F. VANCAUWELAERT, AND J. J. FRIPIAT
696
Table I A. Decomposition Sequences of Alumina Hydrates (h, Hydrothermal Treatment) Starting m 8t e ri81
Ref 270°
970°
Gibbsite Gibbsite
+ boehmite + y +
Gibbsite
+boehmite
+
X
+
K
-
400°
300°,h
Gibbsite Bayerite
e
+
K
+
e+
(Y
e
85O0
750’
-, s
4 3 5 6 3, 5
Same as bayerite +
3
llOOo
+ t l + e + a
450’
Boehmite
1150°
1060°
y 9000
+
-+ boehmite+ y + + boehmite + 7 4 + (Y
230°
Bayerite Nordstrandite
750°
x
2a 2b
e +Q
9oOo +
e+a
B. Structure of Transition Aluminas Form
Structure
v
Spinel cubic or tetragonal (cubic close packing of oxygen atoms with tetrahedral and octahedral aluminum sites) Spinel cubic tetragonal Spinel orthorhombic tetragonal Spinel cubic hexagonal Spinel hexagonal monoclinic
Y
s X K
e
alumina, not only in its pore-size distribution but also in the fact that the water content of y alumina is greater than that of q alumina. Both types develop surface acidity as they are heated up to 900” but the strength of the acid sites is greater in the case of q alumina. These observations, associated with many others not reviewed here, suggest structural differentiations between the various transition forms. Unfortunately our knowledge of these structures is rather poor. With the exception of the spinel form (a), the species are mainly characterized by the dimensions of the unit cell and by some indications on their crystallographic system (Table I). The microcrystalline state of these substances and the poor quality of their X-ray diffraction patterns preclude the application of refined methods of structure determination. Moreover, the earlier works by Brown, et aL12and by Tertian and Papbe’’ have shown that the combination of heating rate, particle size, and pressure produces various decomposition sequences and thus complicated mixtures of transition forms. For instance, the two reaction paths given for gibbsite in Table I correspond ,respectively, to the decomposition of a fine-grained mineral for the first and of coarse particles for the second. I n the case of a broad range of particle size, dehydration would give possibly a mixture of x and y aluminas. The Journal of P h y s M Chemistry
Ref
7, 8 9 4 3 10 3 2a, 11 2a, 11 12, 13, 14
Brindley and Choe,28 using electron diffraction by single crystals, have suggested an interesting way to obtain more detailed information. It is possible to follow the decomposition mechanisms by recording the electron diffraction pattern of one specified particle and to determine subsequently the corresponding structure evolution. However, because of preferential orientations of the reciprocal lattice with respect to the electron beam, the number of parameters obtained in this manner is too restricted for complete structure determinations. Moreover in the case of extremely well-dispersed materials, as those used for catalysts, electron diffraction patterns cannot be obtained. I n view of the limitations of the standard X-ray and electron diffraction techniques, we have approached this problem from a completely different viewpoint. The transition aluminas will no longer be considered as (11) G. W. Brindley, A m . Mineralogist, 46, 1187 (1961). (12) J. A. Kohn, G. Kats, and J. D. Broder, ibid., 42, 398 (1957). (13) R. F. Geller, J. Chem. Phys., 33, 676 (1960). (14) H. Saalfeld, Neues Jahrb. Mineral. Abhandl., 95, 1 (1960). (15) J. W. Newsome, H. W. Heiser, A. S. Russell, and H. C. Stumpf, “Alumina Properties,” Technical Paper No. 10, Aluminum Co. of America, Pittsburgh, Pa., 1960. (16) D. 8. Maciver, H. H. Tobin, and R. T. Barth, J . Catalysis, 2, 485 (1963). (17) R. Tertian and D. Papbe, J. Chim. Phys., 55, 341 (1958).
STRUCTURE AND PROPERTIES OF AMORPHOUS SILICOALUMINAS
crystalline but as amorphous materials and a method especially suited for amorphous substances will be applied. From the diffuse scattering of a monochromatized X-ray beam, the radial electron density is computed with respect to an arbitrarily chosen origin. The successive maxima in this distribution correspond to the various interatomic vectors. This technique, initially proposed by Warren18-20 for studying the structure of glasses, gives the mutual arrangements of atoms within a sphere of about a 5-A radius; for longer distances from the center, the successive maxima corresponding to various interatomic distances overlap. Fripiat, LBonard, and B a r a k F have applied such a method to hydrated silica gels, but as far as we know, neither the amorphous alumina hydrates nor the transition aluminas obtained from their dehydration process have been studied in that manner. This will be the main topic of this contribution. To obtain detailed information, reference data are required : these will be obtained by computing the electron density distribution from known comparable structures, namely, gibbsite, bayerite, boehmite, or the 7 spinel structure. As in the latter type of oxygen atom arrangement, the distribution of aluminum in tetrahedral or octahedral sites is an important feature; the method used previouslyz2 and based on the shift of the aluminum K a emission line will also be applied. Infrared spectra in the range where the A1 octahedra and tetrahedra vibrations arc’ observed will also give some useful information.
11. Procedures (a) Generalized Fourier Synthesis. The method, developed originally by Warren for studying structures of amorphous materials, is based on the relationship between the radial electron density and the length of interatomic vectors. In the simplest case of a monoatomic substance the basic relationship is 47rr2p(r) = 4ar2po
+ (2r/a)
si(s) sin srds
(1)
where p is the electron density, r the distance from an arbitrary atom taken as origin, and s = (4a sin e)/A. I n eq 1
i(s)
=
I ( s ) / N f ( s )- 1
(2)
where I ( s ) is the intensity of the X-ray radiation diffracted by an indifferently oriented lattice, N is the effective number of atoms in the sample, and f(s) is a function of the diffraction angle 0 such that e4
+ cos2 20)
f(s) = I02m2C4R2fA2(1
(3)
697
where IO is the intensity of the radiation reaching the sample and fA the atomic scattering factor. R, e, m, and c are the distances from the scattering electron, the electron charge and mass, and the light velocity, respectively. For a substance containing more than one atomic species, eq 1becomes xK,4ar2pm = xK,4ar2po m
m
+
( 2 r / ~ ) ~ ~ s isin ( s srds ) (4) the summation being extended to all the atoms contained in the stoichiometric unit. K, is the effective electron number of atomic species m defined as the ratio fm/fe where f m and fe are the atomic scattering coefficient and the mean electron scattering coefficient, m respectively. The latter is equal to ~ f m / x Z where m
m
is the atomic number. The electron density, p,, is thus a function of the mean electron density (po) per cubic angstrom, of s, and of i(s). Hence
2,
Nd X 10-24xK, m
PO =
M
(5)
where d is the specific weight and M the molecular weight. I n eq 4, i(s) has the same meaning as in eq 2 and becomes lcoh
i(s) =
-
x f m 2
m
fez
(6)
- Iincoh
(7)
where Icoh
=
Iexp
I n this work the atomic scattering coefficients and the contribution of the incoherent intensity were taken from the “International Tables for Cry~tallography.”~~ The calculations were made as shown previously for silica gels by Fripiat, LBonard, and Barak6,21using an IBM 1620 computer. I n the calculation of K,, 16 electrons were attributed to aluminum and 7 electrons to oxygen. This mathematical artifice has no inflbence on the final results provided that the total number of electrons is constant in the stochiometric unit, but it (18) B. E. Warren and N. S. Gingrich, Phys. Ibev., 46, 368 (1934). (19) B. E. Warren, H. Krutter, and 0. Morningstar, J. Am. Ceram. SOC.,19, 202 (1936). (20) B. E. Warren, J . Appl. Phys., 8 , 645 (1937). (21) J. J. Fripiat, A. LQonard, and N. BarakQ, Bull. SOC.Chim. France, 122 (1963). (22) A. LQonard,S. Suzuki, J. J. Fripiat, and C. De Kimpe, J. Phys. Chem., 68, 2608 (1964). (23) “International Tables for X-Ray Crystallography,” Vol. 111, The Kynoch Press, Birmingham, England, 1962, p 202, 250.
Volume 71, Number 8 FebruarV 1967
698
A. J. LkONARD, F. VANCAUWELAERT, AND J. J. FRIPIAT
diffraction spectrum of a thin sheet of muscovite, using a narrow receiving slit of 0.8 mm (5 of Figure 1) as compared with the 6-mm asymmetrical aperture used for the amorphous substances. The following corrections were made on the experimental data. Let x be the sample thickness and x. the “infinite thickness” defined by CullityZ6
I
=
I,,
[
1-
exp(-2px,/sin e) - exp(-Zpx/sin 8) 1 - exp(-Zpx/sin 8)
1 (8)
Figiirc 1. Iitstrttmcnt, used for recording X-ray scattering ewvc nird l o compote bhe radial d i s t r i b u h r of electron density.
produces a better resolution in the electron density distribution. The instrument used in this work is shown in Figure 1; the X-ray beam was first monochromatized by means of a curved asymmetrical quartz crystal ( l ) , the goniometer (2) being adjusted for fitting the focalization conditions; the distanrc hctmeen the monochromator and the sample (3) was 325 mm and the distance between the sample and the scintillation detector (4) was 180 mm. A molybdenum (A 0.709 A) “fine-focus’’ target was used at 54 kv and 14 ma. ~ ratio has been reduced from The I < c ~ : I < cintensity one-half to one-eighth. As proposed by McKinstry and Short,24the contribution of the harmonics was suppressed by discriminating a t 44 f 12 v. In the diffracted beam the number of pulses was between 150 and 1500 counts/sec; this is appreciably below the maximum speed of discrimination, i.e., 4500 pulses/ sec. The diffraction spectra werc recorded by scanning the 28 domain between 4 and 162” by steps of 28 = l”,using the technique of fixed counts (16 X lo3pulses). The maintenance of a correct geometry was checked before and after every determination by recording the The Journal of Physical Chemistry
where p is the linear absorption coefficient. The important contribution of the X-ray scattering by the air has been measured and subtracted from I,,,, as proposed by Zarzycki.26 The experimental values have also been corrected for the double polarization effect undergone by the diffracted beam according to Azaroff.m It is well known that neglecting the end terms of a Fourier series creates a perturbation“ and raises a series of oscillations in the electron density d i s t r i b u t i ~ n . ~ ~ Taking into account the wavelength of the 110 KCI radiation, the upper limit of the Fourier integral in eq 4 was s = 17.5 for 28 = 162”. These oscillations have the strongest intensity near the origin of the radial distribution. Their amplitude decreases progressively as the radius increases and finally disappears at ahout 4 A. These so-called “ghosts” were ohserved mainly below 2.0 A. They modify the shape of the hand appearing in the electron density distribution forradiibetween 1,5and2,1A(bandI),whichisohviously due to the AI-0 interatomic vectors. Above 2 A, the weak peak a t 2.3 A, which appears in some cases, is probably a ghost too. Rloreover, according to Becherer, et a1.,30if the limit of the Fourier integral is extended to higher s values as by Richter, et the lack of convergence of Ieoh and in eq 6 may also
-
cfm2 m
modify the relative intensities of the electron density maxima since fc 0 for s -+ m . It is believed that the limit adopted in this work represents a compromise which minimizes the errors due to the end terms of the (24) H. A. MeKinstry and M. A. Short. Z. Kiist., 114, 278 (1960). (25) B. D. Cullity. “Elements of X-Ray DiRraetion.” AddisonWesley, London. 1959, P 188. (26) J. Znraycki. J . Phys. Radium. 115, 44A (1956). (27) L. 1’. Amroff. Acta Civsl.. 8, 701 (1955). (28) H. Krebs and F. Sohultie-Gebhardt. ibid., 8 , 4 1 2 (1955). (29) H. Liuson and W. Cochran, “The Determination of Crystal Stmeturns.” Bell and Sons. London. 1953. p 291. (30) G. Beeherer. 0. Brummer, and G. Herrns. Silikot Tech., 13. 339 (1962). (31) H. Richter, G . Breitling. and F. Herre, Z. Naturforach.. 9 , 390 (1954).
699
STRUCTURE AND PROPERTIES OF AMORPHOUS SILICOALUMINAS
Fourier integral and the effect of the lack of convergence of I c o h and Efn12. m
(b) X - R a y Fluorescence Spectroscopy and Infrared Spectroscopy. The method for determining the coordination numbers of aluminum and silicon from the shift of the A1 K a and Si K a fluorescence lines has been described previously. 2 2 The samples prepared for recording the diffraction spectra and computing the radial electron density distribution were also used for X-ray fluorescence. The X-ray spectrometer was equipped with a chromium target, a pentaerythritol analyzer, and a flow counter, the settings being as follows: X-ray generator, 50 kv, 20 ma; collimator aperture, 160 p ; discrimination, 39 f 12 v; scanning speed, l / S o min-l. KBr pellets (5"/,,) were used for recording the infrared spectra in the 400-1200-~m-~region with a Beckman IR12 spectrometer.
111. Materials The hydrated alumina gels were prepared from a 0.25 N A1C13.6Hz0 solution by precipitation with a 4 N XaOH solution. Two series of gels were studied according to the pH at which the precipitation was carried out. The G series was obtained from gels precipitated at pH 4.5 and purified by dialysis at 50" for 3 (GI), 4 (GII), 5 (GIII), and 7 (GIV) days, respectively. The B series was prepared from gels precipitated at pH 8 and purified by dialysis at the same temperature and for the same periods of time (BI, BII, BIII,and BIV). The samples were then dried under vacuum at 20" J!
Table I1 : Number of Water Molecules per A1203 Unit (WC) and Specific Surface Area (SO, m2/g) Pretreatment temp, OC
20
G aeries
mC SO
300
IYC
550
SO WC
so
I
I1
111
IV
3.67 4 2.00 7 1.18 27
3.04 9 1.39 8 0.86 37
3.23 4 1.38 8 1.02 29
3.68 5 1.51 107 0.99 305
3.39
3.14 235 1.39 286 0.09 208
3.41 224 1.48 296 0.40 276
2.72 243 1.14 287 0.14 240
B series
20 300 550
wc so
wc SO wc S0
158 0.91 301 0.50 247
V
Figure 2. Some examples of radial distribution functions of electron density.
(GI,~o-GIv,~o or B I , ~ ~ - B I Vor, ~heated ~ ) at 300 or 5.50" to constant weight ( G I , ~ ~ - G I vand , ~ ~&,3OO-BIV,j50). o In summary, four different samples of each series, dried or calcined a t 20, 300, and 550", were studied. Their water contents, obtained by measuring the weight loss a t lOOO", and their BET (Nz at -196") surface areas are shown in Table 11. I n addition, two samples belonging to the G and B series were calcined at 700" (G700 and B700). For the X-ray diffraction and fluorescence methods, the samples were prepared by molding a pellet of 5-mm thickness and 30-mm diameter under a pressure of 500 kg/cm2. As shown by Gastuche and H e r b i l l ~ n ,the ~ ~ con(32) M. C. Gastuche and A. Herbillon, Bull. Soe. Chin. France, 1404 (1962).
Volume 71, Number S
February 1967
700
A. J. L~ONARD, F. VAN CAUWELAERT, AND J. J. FRIPIAT
A
X1
Figure 3. Projections of the characteristic framework of gibbsite, boehmite, and W-A~ZOS showing the interatomic vectors contributing to bands I1 (upper part A ) and I11 (lower part B) of the radial distribution function of the electron density. Gibbsite: zy projection of the octahedral layer; A, second coordination shell of a specified aluminum atom (small circle a t the center). The AI-0 and AI-A1 interatomic vectors contributing to band I1 involve the various atoms shown here. The atoms represented by dashed lines belong to the upper or lower hexagonal rings shifted with respect to the reference plane on the left or on the right, respectively. Boehmite: yz projection; % height: z = 0.75, --; x = 0.25, - - A, oxygen and aluminum atoms in the second coordination sphere of the aluminum atom a t the center. The differentiation of atoms exactly superposed is obtained by using small (aluminum) and larger circles (oxygen or hydroxyl) of slightly different diameters. B, atoms contributing to band I11 are shown with respect to aluminum a t the center. q-Al203: the heights in eighths of the cubic cell are represented by figures against each atom. For the sake of clarity, half the environment of each kind of aluminum is represented: on the left, 0, Al(V1) and on the right, 0,Al(1V). Al(1V) and Al(V1) atoms lie in even or odd shells, respectively.
-.
tinuous dialysis under the conditions explained above leads to the formation of gibbsite for the G series and to the formation of bayerite for the B series. These crystalline species are obtained after about 10-12 days. According to Fripiat and P e n n e q ~ i n , ~the 3 dialysis process results a t first in a depolymerization of the initial coarse particles. The molecular weights of the fully depolymerized particles obtained after 2-3 days are of the order of magnitude of 10.000 in the G series and of 50.000 in the B series. After this first step, a polymerization process occurs which leads to the progressive formation of crystalline trihydrates. The first intense diffraction lines typical for gibbsite or bayerite are observed after 6-7 days. Therefore the The Journal of Physical Chemistry
periods of time chosen in the G as well as the B series cover the whole polymerization process preceding crystallization. The G samples dried a t 20" or calcined at 300" are amorphous. The samples calcined at 550" show Xray diffraction spectra similar to those obtained by Tertian and Papbe" and attributed to x alumina. After the thermal treatment at 700°, a mixture of y and x aluminas is found. I n the B series the BI,zoB I I I , samples ~~ are amorphous while the B I V ,sample ~~ contains some pseudoboehmite. The same crystalline form is found for the B ~ Osamples O while the four B550 (33) J. J. Fripiat and
(1965).
M. Pennequin, Bull.
SOC.Chim. France, 1655
STRUCTURE AND PROPERTIES OF AMORPHOUS SILICOALUMINAS
701
samples exhibit diffraction patterns similar to that of BToo contains the same transition species.
q alumina.
IV. Results of thz Generalized Fourier Analysis As shown by some examples in Figure 2, the radial distribution functions obtained for the various samples may be split into three main “bands,” Le., band I from 1.5 to 2.1 A, band I1 from 2.4 to 3.7 A, and band I11 from 3.9 to 5.2 A. They will be discussed separately. This division is founded on the distribution of the interatomic A1-0, 0-0, or A1-A1 vectors in gibbsite, 34 b a ~ e r i t eboehmite,36 ,~~ and q alumina? I n order to decompose one band into its various components and to compute the respective contributions of the corresponding interatomic vectors, a gaussian distribution of the electron density is assumed. The amplitude “y” of a component is related to the distance z from the maximum of the peak by the relation = ke--h2z2 (9) where k is the maximum amplitude; h is related to half band width, h,, by the relationship h =
