ELSEVlER Applied Catalysis A: General 118 ( 1994) 73-86

Control of porosity and surface area in Ti02-A1203 mixed oxides catalytic supports. A statistical

approach

Jorge Ramirez a$*, Tatyana Klimova a, Yadira Huerta a, Jose Aracil b ’ Deparfamenfo de Ingenierfa Quimica, Fact&ad de Quimica, I/NAM, Cd. Vniversitatia, Mhico D.F. (04510),

Mexico

b Departamento de Ingenierfa Quimica, Facultaa’ de Quimica, Vniversidad Complutense, Madrid (28040). Spain

Received 4 April 1994; revised 23 June 1994; accepted 27 June 1994

Abstract

The synthesis of TiOs-Al,O, mixed oxides catalytic supports with a TiOs/ (TiOs + Also,) molar ratio equal to 0.5 was made, at various experimental conditions, by the co-precipitation of the corresponding metallic isopropoxides using ammonium carbonate as pore regulating agent. The influence of the preparation parameters (amount of ammonium carbonate, water and the procedure for the precipitation of the isopropoxides) on the surface area and pore structure of the final solid were studied using a full 23 factorial experimental design. Within the experimental conditions used here, the model response equations for surface area, cumulative pore volume and mean pore diameter showed good agreement with the experimental results. It was found that the amount of ammonium carbonate was the most important parameter, affecting the three responses in a positive way. The amount of water resulted in a positive influence on surface area and a negative influence on cumulative pore volume and mean pore diameter. The manner in which the precipitation was made also had a significant influence on pore diameter and pore volume. The differences in the responses may be mainly due to an interaction between the carbonate ions and the aluminium or titanium hydroxide gels which favours the link between particles leading to high surface areas and different porosities depending on the carbonate concentration.

Keywords: Experimental design; Mixed oxides; Porosity control; Surface area; Titania-alumins

1. Introduction

In many catalytic systems, the performance is known to depend not only on the inherent catalytic activity of the active phase, but also on the textural and physi-

* Corresponding author. Tel., fax. ( + 52-.5)6225366.

0926-860X/94/$07.00 0 1994 Elsevier Science B.V. All rights reserved SSDIO926-860X(94)00146-4

14 J. Ramirez et al. /Applied Catalysis A: General 118 (1994) 73-86

cochemical properties of the support. In the attempt to obtain better catalysts, the use of some new supports such as Ti-Al mixed oxides has been tried with promising results in some reaction systems such as hydrodesulphurization reactions [ l-31. In this case, greater catalytic activities have been found due to the role of the support.

In general, sol-gel methods have been preferred to produce this mixed oxide system. However, the control of the surface area and porosity of the catalysts remains a problem since small diameter pores, which can cause diffusion intrusions in the case of large reactant molecules, are normally obtained [ 41.

Numerous methods have been used in order to create materials with defined specific surface properties [ 51, and most commercial supports are already available in a variety of surface area-pore size combinations which are somewhat interrelated in such a way that large pore supports typically have a reduced surface area. In heterogeneous catalytic reactions, it is important to have control over both the surface area and the pore size in order to achieve the highest possible catalytic activity.

Several methods have been reported in the literature for preparing refractory support materials with variable surface characteristics. In the case of alumina, careful control of the specific gelation and aging conditions of pseudo-boehmite have been found to have an effect on the surface properties of the final support [ 61. The control of pore structure in alumina supports has also been achieved by rapidly ‘swinging’ the pH of the solution during the gelation process [ 71.

The mean pore size of the support can be further increased by adding a pore- regulating agent (or additive) to the stock solutions from which the gel is prepared. In the preparation of porous metal oxides ( A1203, SiOZ, A1203Si02) various types of additives have been used [ 51. In general, they are organic materials which can be evaporated or cornbusted during calcination of the gel [ 81. Also, the use of template molecules, usually quaternary ammonium ions, has been described in the literature [ 5,9,10].

The effect of carbonate and other ammonium salts as additives upon the physical properties of A1203 and the A1203Si02 mixed oxide has been reported previously [9,10], however, no systematic study of the effects of each variable and of the interactions between variables has been done and therefore it is difficult to make an a priori prediction of the resulting textural properties of these solids when the preparation variables are changed.

In the present work we test the effectiveness of the use of ammonium carbonate in controlling the porosity and surface area of a Ti02-A1203 mixed oxide, prepared by the sol-gel method. To our knowledge, no such study has been reported in the past. During this study, the amount of ammonium carbonate, water, and the method of addition of the reactants was varied in order to see the effect that these variables have on the surface area, pore volume and pore size distribution of the Ti02-A1203 support.

J. Ramirez et al. /Applied Catalysis A: General 118 (1994) 73-86 15

In this paper, emphasis is placed on the use of factorial design of experiments in order to show that the statistical methods allow us to obtain a useful model for the control of the textural properties with a reduced cost and number of experiments [ 11,121.

The experimental study, which followed a 2k experimental design, was under- taken in order to provide a relationship for the surface area, pore volume and pore diameter as a mathematical function of the amount of ammonium carbonate and water added for hydrolysis, as well as the method used to prepare the TiO*-A1203 supports.

From preliminary trials, appropriate values for the two levels of the factors (amount of ammonium carbonate, amount of water and method of preparation), were selected to carry out the final experiments. The results from these final exper- iments were used to build a statistical model which uses coded values ( + 1, - 1) of the experimental factors or parameters. With this statistical model, the real influence of each factor is calculated without being affected by its relative magni- tude with respect to magnitude of the other factors. Finally, the statistical model is transformed into an experimental model, which uses real values of the different parameters expressed in their normal units.

2. Experimental

2.1. Support preparation and characterization

The mixed oxide samples were prepared with a TiOJ (TiOz + A1203) molar ratio equal to 0.5, using titanium and aluminium isopropoxides as precursors and n-propyl alcohol as a solvent. In the experiments, an aqueous solution of ammonium carbonate, with pH = 9.0, was used to produce the precipitation of the metallic hydroxides, The resulting precipitates were aged, with slow stirring, for 24 h and then filtered under vacuum and washed with water. The solids formed, after drying at 100°C for 24 h and calcining for 24 h at 5OO”C, were characterized by surface area and porosity using a commercial BET physisorption apparatus. In general, the errors found in repeated measurements of surface area determination were within 3-5% of the total surface area.

2.2. Statistical analysis

Experimental designs are frequently performed in the study of empirical rela- tionships in terms of a mathematical model between one or more measured responses and a number of factors. If the levels of factors are equally spaced, the orthogonal polynomials may be used. The statistical design chosen for this study was a full two level factorial design 23 (three factors at two levels). The selection of responses was made considering the most important textural properties of the

76 J. Ramirez et al. /Applied Catalysis A: General I18 (1994) 73-86

Table 1

23 Factorial design: experiment matrix and experimental results

Run Order of run AC

(g)

Water Method X,, Xw X,,, Us, YPV (ml) (m*/g) (cm3/g)

1 5 2.88 40 A - - - 248 0.373 41

2 7 7.50 40 A + - - 343 0.874 76

3 8 2.88 150 A - + - 293 0.35 5 34

4 1 7.50 150 A + + - 345 0.76 5 64

5 2 2.88 40 B - - + 226 0.642 78

6 4 7.50 40 B + - + 326 0.99 9 90

7 6 2.88 150 B - + + 343 0.51 4 40

8 3 7.50 150 B + + + 394 0.99 5 70

For the standard sample prepared with no ammonium carbonate: Ys* = 233 m*/g, Y, = 0.234 cm3/g, Y,, = 30

A.

catalytic supports: surface area ( YsA) , cumulative pore volume ( YPv) and average pore diameter ( YPD). The factors (k) chosen were the amounts of ammonium carbonate (XAc), water (X,) and the type of method of hydrolysis (X,) used (adding the ammonium carbonate solution to the isopropoxides, method A, or adding the isopropoxides to the ammonium carbonate solution, method B) .

The selection of levels was carried out on the basis of results obtained in a pre!iminary study, keeping in mind that our main interest was to obtain solids with both large surface areas and pore diameters. These considerations made it possible to fix the upper ammonium carbonate level at 15.0 g, in order to avoid the loss of surface area caused by the presence of exceedingly large pores. On the other hand, for the lower limit it was found that ammonium carbonate amounts under 2.88 g led to high surface areas but with the presence of small pores of 25 A in diameter or less.

An amount of water equal to 40 ml was selected as the minimum level in order to achieve complete hydrolysis of titanium and aluminium isopropoxides, the upper water level was fixed at 750 ml in order to study the influence of a sufficiently large excess of water upon the textural properties of the support.

The way in which the reactants are mixed to achieve precipitation of the hydrox- ides has a marked influence on the physical properties of the resulting supports. In this work, two different methods were studied - method A: slow addition of the ammonium carbonate solution to the titanium and aluminium isopropoxides-n- propyl alcohol mixture, and method B: rapid addition of the isopropoxides dissolved in n-propyl alcohol to the aqueous ammonium carbonate solution.

Once the maximum and minimum factor values were selected, the statistical analysis was applied. The experimental matrix for the factorial design was built according to standard procedures [ 121 and is presented in Table 1. In this table, the third, fourth and fifth columns of data give the factor levels for ammonium carbonate and water on a natural scale and the method used, respectively, the sixth to eighth columns give the coded factor levels ( + ) or ( - ), on a dimensionless

J. Ramirez et al. / Applied Catalysis A: General lIB (1994) 73-86 77

(1)

scale. All the experimental runs were performed at random and the experimentalerror estimation was made by the Daniel's method [13].

The coded values Xk were obtained by calculating:

X - (Xk -:tk)k- d

where Xk is a natural variable for factor k, Xk is the mean value of xi; and d is theabsolute difference between x and X.

The response can be written as a function of the factors in the following way:

Y=ao +atXAC +a2XW+a3XM +a12XACXw +a13XACXM

(2)

3. Results

In addition to the runs required for the experimental design, a standard sampleprepared using method A, 40 ml of water, and no ammonium carbonate wasprepared. This sample had a surface area of 233 m2

/ g, pore volume of 0.234 cm3 /

g, and an average pore diameter of 30 A.The results obtained for the responses are also reported in Table I. From this

table it is possible to see that, in general, both methods of preparation lead to surfaceareas above 200 m2

/ g. However, the pore volume and pore diameter of the samplesvary over a wide range of values.

From the experimental results, after the mathematical analysis was performed,the main effects and second and third order interactions were calculated. The resultsfrom this calculation are shown in Table 2.

The effect of a factor on the response is simply the difference between the averagevalue of the response ofthe runs at high level and the average value of the responseof the runs at low level.

Figs. l a, band c show Daniel's plots for error estimation. From these plots, byeliminating the points with an effect contribution close to zero, it is possible todetect which effects have significant influence on the response.

The form of the best fitting response functions to the significant factors andinteractions are also presented in Table 2.

3.1. Statistical model

In the statistical model which uses coded values of the factors, Eqs. (3), (4) and(5) show the response equations for surface area, pore volume and average porediameter in terms of the significant factors obtained from the statistical analysis.The correlation coefficient (r), is also shown for each case.

.I. Ramirez et al. /Applied Catalysis A: General 118 (1994) 73-86

l-

O.l-, 1 -23 -3 17 37 57 7’

PARAMETER STANDARDIZED EFFE

99.9-

99-

!z $ 95- -,’ . hi 50

fj 20- 0. .

s 5

5 l-

.

o-‘h7Hed% -8 Ix

r!d

0.01) PARAMETER STANDARDIZED EFFECTS

“:;v

PARAMETER STANDARDIZED EFFECTS

Fig. I. Error estimation: Daniel’s method: (a) surface area, (b) cumulative pore volume and (C) mean pore

diameter.

Y SA = 314.75 + 37.25X,, +29.00X, +7.50X, - 11.5OX~~rw

+ 17.25XweM (I= 1.000) (3)

Table 2

J. Ramirez et al. /Applied Catalysis A: General 118 (1994) 73-86 79

2’ Factorial design: statistical analysis

Response YSA YQV YFIl

No. experiment 8 Degrees of freedom 7 Y 314.8

Main effects and interactions:

AC

W

AC*W

M

AC*M

W*M

AC*W*M

Significant test

Significant main effects and

interactions

Response equation

8 a 7 I 0.68 5 61.6

14.5 0.42 I

58.0 - 0.07 5

- 23.0 -0.002

15.0 0.11 8

1.0 -0.02 8

34.5 -0.01 I - 1.5 -0.044

Daniel’s method Daniel’s method

26.8

- 19.3

3.3

15.8

- 5.6

- 9.8

5.8

Daniel’s method

AC(+LW(+LM(+), AC(+LW(-hM(+) AC(+),W(-),M

AC*W(-),W*M(+) (+LW*M(-)

Y=o,+a,X,,+ Y=ao+olXAc-alXw+ Y=aO+a,X,,-

GhhGh- GM a,Xw + GM - a,& .M

a,2X.0w+a23L.M

Y,, = 0.685 + 0.214X,c - 0.037x, + 0.093x, ( r = 0.989) (4)

Y,, = 61.63 + 13.38XAc -9.63X, +7.88X, -4.88XweM (r=0.939)

(5)

3.2. Experimental model

To be useful, the statistical model has to be transformed into a model which uses the real values of the experimental parameters. This type of model is called here Experimental Model. Also, because the method of precipitation cannot be expressed as a numerical variable, the responses for the experimental model have been written in terms of 2 factors.

Eqs. (6) to ( 11) present the experimental model for each of the responses, obtained for the two methods of preparation used in this study in a separate way.

80 J. Ramirez et al. /Applied Catalysis A: General 118 (1994) 73-86

Method A

Us, = 162.67+23.95AC+0.65W-O.O8AC*W (r= 1.000) (6)

YPV =0.135 + 0.099AC - O.OOlW (r= 0.985) (7)

YPr, = 25.44 + 7.03AC - 0.09W (r= 0.992) (8)

Method B

YsA=110.01+25.50AC+1.34W-O.lOAC*W (r=l.OOO) (9)

Y Pv =0.404+0.086AC-O.OOlW (r=0.984) (10)

Y r,, =70.95+4.55AC-0.26W (r=O.907) (11)

3.3. Influence of variables (factors)

The influence of three factors, ammonium carbonate mass, amount of water and method, as well as the interactions on the responses can be obtained from Eqs. (3)) (4) and (5). The unit percent influence of a parameter on a given response is the ratio of the calculated influence of that parameter to the calculated response obtained with the average value of the same parameter, divided by the difference between the high and low experimental values of that parameter, For example, for ammo- nium carbonate, the influence on surface area calculated by the model was + 74.5 m2/g, the calculated response with the average value of ammonium carbonate amount (5.19 g) was 314.75 m2/g, which corresponds to the value of a0 in the statistical model and the difference between the experimental values of ammonium carbonate amount was 7.5-2.88 = 4.62 g. Therefore, the unit percent influence of ammonium carbonate on surface area can be calculated as: ((74.5/314.75)/ (7.5 - 2.88)) X 100 = 5.12%. This last calculated value is the influence that one gram of ammonium carbonate will have on the surface area response.

3.4. Sulfate area response

For the experimental range of factors studied, the most significant main effect is the amount of ammonium carbonate. An increase of 1 gram of ammonium carbonate leads to a 5.12% increase in the surface area of the mixed oxide. A change in the method of addition of the reactants from A to B also has a positive influence and contributes with 4.77% of the increase in surface area. The amount of water added for hydrolysis has a positive influence on surface area. An increase of one millilitre in the initial amount of water used for hydrolysis corresponds to an increase of 0.17% in surface area. The influences of the W * M (water-method) and AC * W

J. Ramirez et al. /Applied Catalysis A: General I18 (19%) 73-86 81

(ammonium carbonate-water) interactions are equal to 0.10% per ml of water and - 0.01% per g of ammonium carbonate-ml of water respectively and they are lower than the main effects produced by ammonium carbonate or water. Contributions of other cross-variable interactions can be neglected.

3.5. Cumulative pore volume response

In this case, the most significant main effect is the amount of additive used, this effect has a positive influence and 1 g of ammonium carbonate produces an increase of 13.5% in the cumulative pore volume. An increase in the amount of water produces a negative effect in pore volume response which is equal to - 0.1% per ml of water. The change of method of addition of the reactants from A to B produces a 27.1% increase in the pore volume.

3.6. Average pore diameter response

Regarding the average pore diameter, the mass of ammonium carbonate has a positive influence. An increase of one gram in the additive mass produces an increase of 9.4% in this parameter. On the other hand, the amount of water has a negative effect of - 0.3% per ml of water. The change of method of preparation from A to B produces a 25.6% increase in pore size. The influence of the W * M interaction is negative and equal to - 0.14% per ml of water.

4. Discussion

The differences between the values of the responses for the experiments reported in Table 1 and the standard sample indicate a clear effect of the factors considered. A comparison of the experimental and calculated responses for the models (Fig. 2), shows the agreement between the observed and predicted values. From this figure it is possible to see that the mathematical model used for the prediction of the responses gives a good fit of the experimental data.

It is interesting to note the very accurate prediction that the model makes of the surface area of the support ( r = 1 .OOO) . Pore volume and average pore diameter are also predicted with good accuracy, r = 0.989 and r = 0.939, respectively. The random distribution of the residuals (Fig. 3) show the absence of a trend indicating also that the mathematical model is adequate since it does not detect any inconsis- tency between experimental and calculated values.

However, given the complexity of the process that takes place during co-precip- itation and later growing of the chains of particles, care should be taken when attempting to use this type of models outside the range of the tested experimental conditions. Clearly, the limitations in the use of the proposed model are those inherent to the statistical approach used here.

J. Ramirez et al. /Applied Catalysis A: General 118 (19%) 73-86

3,~ , , 1 , , ,I 220 250 200 310 340 370 400

PREDICTED x,, (m2/g) [x 0.01)

114-

34 54 74 94 11 J

[x 0.041] PREDICTED Ypv (cm31g] . .

PREDICTED ‘6D (iij

Fig. 2. Comparison of experimental and calculated responses: (a) surface area, (b) cumulative pore volume and (c) mean pore diameter. Some experiments have been repeated to evaluate the reproducibility of the data.

J. Ramirez et al. /Applied Catalysis A: General 118 (I 994) 73-86 83

1.2. = .

0.7-

0.2 - . .

-0.3 - . .

-0.8 -

-1.3-,‘, , ( , , , , . , , , _ 220 250 200 310 340 370 401

PREDICTED Y,[m2/g]

[x 1 E-3)

57 -

t 37- .

1 .

? 17 -.

2 B . _3-

.

en

;-23 - ’ ,

-43 -, .# I I , I , ) I 34 54 74 94 114

[x 0.01)

PREDICTED ‘$,, [cm3&l

8 . ICI

.

35 55 75 95

PREDICTED +D [A]

Fig. 3. Residual analysis for estimated models: (a) surface area, (b) cumulative pore volutt~ and (c) mean POE diameter.

84 J. Ramirez et al. /Applied Catalysis A: General 118 (1994) 73-86

The results from the statistical analysis show that an increase in the amount of ammonium carbonate has a positive influence on the three responses, increasing the surface area, pore volume and pore diameter.

The amount of water also has a positive influence on the surface area but a negative influence on pore volume and pore diameter. In this case, a compromise should be reached between achieving a large surface area with reasonable values of the pore diameters and pore volume. The decision will clearly depend on which kind of support is desired.

The change of method of preparation from A to B also gives a positive effect on the three responses. However, the most important influence of this factor is on the pore volume and pore diameter responses.

The increase in surface area induced by ammonium carbonate has been related to the presence of a metal ammonium carbonate hydroxy-hydrate intermediate phase [ 91, the formation of which is promoted by the use of larger concentrations of this additive, when the pH is maintained above 8. The increase in pore volume and pore diameter while maintaining high surface areas indicates that the increase in the concentration of ammonium carbonate promotes the link between small particles in such a way that during the drying and calcination steps the collapse of the porous structure is avoided and therefore high values of the pore volume, pore diameter and surface area are preserved. This effect is enhanced when method B is used due to the higher possibility of reaction between the ammonium carbonate and the metallic hydroxides.

An increase in the amount of water, on the other hand, produces a complete hydrolysis of the precipitate and at the same time a dilution effect on the ammonium carbonate concentration. This combination of effects is reflected in a decrease in the volume and diameter of the pores and in an increase of the surface area due to the formation of isolated, small precipitated particles. These effects are more pro- nounced in the case of method B, where a large amount of water is contacted with a small amount of the titanium and aluminium isopropoxides.

5. Conclusions

The synthesis of a titanium-aluminium mixed oxide with a TiOz/ ( TiOz + Al,O,) molar ratio equal to 0.5 has been studied using a 23 full factorial design. The response equations for surface area, cumulative pore volume and mean pore diameter have been established in terms of the amounts of ammonium car- bonate and water used during the synthesis. Also, the effect of a change in the way in which the reactants are contacted during the preparation has been explored.

From the results of this study, the following conclusions can be drawn: 1. The proposed statistical model can predict the three responses, surface area,

cumulative pore volume and average pore diameter, with reasonable accuracy within the range of the experimental conditions used in this study.

J. Ramirez et al. /Applied Catalysis A: General I18 (1994) 73-86 85

2. A study of the factors which affect the three responses shows that the amount of the ammonium carbonate used is the most important of them.

3. The synthesis of the supports by method B leads to greater pore volume and mean pore diameters.

4. The amount of water has a positive influence on surface area but a negative influence on pore diameter and pore volume.

Notation

AC Amount of ammonium carbonate w AC*M Ammonium carbonate-method interaction AC*W Ammonium carbonate-water interaction AC * M * W Ammonium carbonate-method-water interaction d k

M

M*W PD PV

:A W

xk

xk

xk

Y PD

Y PV

Y SA

Absolute difference between x and X Factor, either AC (ammonium carbonate mass), W (amount of water), M (method) Method of addition of reactants: Method A = addition of water to the Ti and Al isopropoxides, Method B = addition of the Ti and Al isopropoxi- des solution to water Method-water interaction Mean pore diameter (A) Cumulative pore volume (cm3/g) Correlation coefficient Surface area (m*/g) Amount of water used for hydrolysis of Ti and Al (ml) alkoxides Natural variable for factor k Mean value of xk Coded level for factor k Experimental response for mean pore diameter (A) Experimental response for cumulative pore volume ( cm3/g) Experimental response for surface area (m*/g)

Acknowledgements

Financial support for this work by DGAPA and DGIA (UNAM, Mexico) and the EEC is gratefully acknowledged.

86 J. Ramirez et al. /Applied Catalysis A: General I18 (1994) 73-86

References

11 I M. Bmysse, J.L Portefaix and M. Vrinat, Catal. Today, 10 (1991) 489-505. 121 A. Nishijima, H. Shimada, T. Sato. Y. Yoshimura and J. Hiraishi, Polyhedron, 5 ( 1986) 243-247. 131 W. Zaobin, X. Qin, G. Xiexian, E.L. Sham, P. Grange and B. Delmon, Appl. Catal., 63 (1990) 305-317. 141 J. Ramircz, L. Ruiz-Ramfrez, L. Cedeflo, V. Harle, M. Vrinat and M. Breysse, Appl. Catal., 93 ( 1993) 163-

180. 151 D.L. Trimm and A. Stanislaus, Appl. Catal., 21 (1986) 215-238. [61 B.E. Yoldas, J. Appl. Chem. Biotechnol., 23 (1973) 803-809. 171 T. Ono, Y. Ohguchi and 0. Togari, Control of the Pore Structure of Porous Alumina (Preparation of catalysts

III), Elsevier, Amsterdam, 1983, pp. 631-641. [81 D. Basmadjian, G.N. Fulford, B.I. Parsons and D.S. Montgomery, J. Catal., 1 ( 1962) 547-563. 191 R.F. Vogel, G. Marcelin and W.L. Kehl. Appl. Catal., 12 (1984) 237-248.

[ 101 R. Snel, Appl. Catal., 12 (1984) 347-357. [ 11 I M. Martinez and J. Aracil, Ing. Quim., 11 ( 1989) 163-168. [ 121 G. Box, J. Hunter and J. Hunter, Statistics for experimenter, John Wiley and Sons, New York, 1978. 1131 C. Daniel, Technometrics, 1 (1959) 311-341.