ChemicalEngineering Science, 1972, Vol. 27, pp. 763-778. Pergamon Press.
Printed in Great Britain
A structural model for gas-solid reactions with a moving boundary-III A general dimensionless representation of the irreversible reaction between a porous solid and a reactant gas H. Y. SOHN and J. SZEKELY Department of Chemical Engineering and Center for Process Metallurgy, State University of New York at But&lo, Buffalo, New York 14214, U.S.A. (Firs? received 13 May 1971; in revisedform 16 J uly 1971) Abstract-A general structural model is developed for the description of non-catalytic gas-solid reactions; which allows for spherical and flat plate like pellets, made up of spherical or flat plate-like grains. The numerical solution of the dimensionless form of the governing equations results in a plot of the dimensionless time to achieve complete reaction against a gas-solid reaction modulus, which is valid for all the geometries considered. The use of this plot could be useful for the planning of experiments or for the interpretation of experimental data, in terms of structural models, without the need for digital computation. It is shown that for extreme values of the gas-solid reaction modulus, the results of the model reduce to the asymptotic solutions corresponding to the shrinking core and the uniform reacting pellet models respectively. Thus the gas-solid reaction modulus is helpful for the quantitative definition of the range of these asymptotic regimes, in terms of the chemical and structural parameters of the system. 1. INTRODUCTION IN RECENT years a great deal of work has been done on the development of alternatives to the shrinking core model for non-catalytic gas-solid reactions[l-71. Much of the earlier work was purely analytical, and only in two very recent studies was there an attempt made to compare the predictions of the models with actual experimental measurements [6,73. A common feature of all these models is the postulate that the reaction takes place on a distributed surface, within the porous body, which is represented as an assembly of individual particles or channels. In the majority of cases the mathematical representation of these systems led to a set of simultaneous differential equations which had to be solved numerically. The computed results shown in the previous publications served primarily an illustrative purpose, indicating certain trends, for given specific situations. It follows that in their present form the struc- tural models available at present are not immediately accessible for the interpretation of experimental data or for the prediction of the behavior of a given system. Therefore it would be desirable to consolidate ideas currently available on structural models into a general dimensionless representation. Ideally such treatment should provide both an overview of the interaction of the various parameters and a convenient framework for prediction and interpretation of experimental data, without recourse to numerical computation in each instance. The purpose of the paper is to attempt the development of such a general model, at least for systems undergoing first order irreversible reactions and where structural changes occurring in course of the reaction may be neglected. 2. FORMULATION Let us consider a porous pellet of volume VP and of superficial area Ap, made up of individual 763 particles (grains) having a volume and surface area V, and A, respectively. The shapes to be considered for the pellets and the grains in the subsequent treatment will include spheres, long cylinders and slabs, the behavior of which should provide good guidelines for most practical systems. of the small size of the grains. Recent work by Calvelo and Smith[7] has shown that the importance of intragrain diffusion is negligible, provided the grain Biot number, r,k/D, is smaller than 1. Let us consider that the solid material forming the pellet reacts with a gas, irreversibly, according to the scheme: Within the framework of these assumptions the problem may be stated by combining an expression for the conservation of the gaseous reactant with a mass balance for the reacting solid. Thus we have A(gas) + bB(solid) + Products. (1) D,QC, - VA =0 (2) The following additional assumptions are made: (1) the pseudo-steady state approximation is appropriate for describing the concentration of A within the pellet where D, is the effective diffusion coefficient within the porous medium, CA is the reactant concentration, and VA is the local rate of con- sumption of A. (2) The resistance due to external mass transfer is negligible. (This simplilication will be relaxed in Section 4 where its effect will be studied.) (3) Diffusion within the pellet is either equi- molar counter-diffusion or at low concentration of the diffusing species. Regarding the conservation of the solid, we shall assume that within each grain the reaction front retains its original geometrical shape (i.e. concentric spheres, or coaxial cylinders, or parallel planes) for successive times. Then the local rate of reaction at a solid surface may be expressed as (4) The system is isothermal. (5) The solid structure is macroscopically uniform and is unaffected by the reaction. p,,,g = bkc, (3) (6) Diffusion of the gaseous reactants through the product layer of the individual grains is not rate-limiting. The consequences of and the limitations imposed by these assumptions will be discussed subsequently. At this stage it is noted that assumption (1) is generally acceptable for gas-solid systems& 93. The effect of external mass transfer (i.e. 2) may be readily incorporated into the analysis, but at this stage this would have introduced one additional parameter. Ishida and Wen[2] have shown that the effect of external mass transfer is negligible if the pellet Sherwood number, k,,,RJ De, is greater than 100. While assumptions (3)-(5) are restrictive, they are thought to be appropriate to a range of problems of practical interest. Finally, it is thought that it would be safe to neglect the effect of diffusion through the product layer, because where pm is the molar density of the solid, r is the distance coordinate perpendic- ular to the moving reaction front, b is the stoichiometric factor, and k is the reaction rate constant. The boundary conditions for Eqs. (2) and (3) have to express the stipulations that initially all the grains are unreacted, and that the reactant COnCeMiXtiOII iS maintained at CA0 at the OUter surface of the pellet, and the gaseous reactant concentration profile is symmetrical about the centerline (center) of the pellet. These con- straints are readily expressed mathematically but, for the sake of brevity, will be only given in their dimensionless form. It is noted that a further assumption is implicit in Eq. (3) through the form chosen for the rate expression. A relationship is readily established between H. Y. SOHN and J. SZEKELY 764 A structural model for gas-solid reactions the local rate of reaction and the consumption r) is the length coordinate in the V operator; of A, provided the geometry of the grain is since radial or cylindrical symmetry has been specified. For flat plates, cylinders and spheres, assumed, the angular terms are not con- aA is given as sidered, and R is the dimensional length coordinate within nA, VA = (I --E)eA jyg (4) the pellet. Finally, Here, E is the porosity of the pellet, and Fg is the shape factor, which has the values of 1,2 and r* = bkC% A, t ( > Pm FJg (11) 3 for flat plates, cylinders and spheres, res- pectively. is the dimensionless time. Equations (2) and (3) may be expressed in The boundary conditions for Eqs. (5) and (6) dimensionless forms given by Eqs. (5) and (6) are readily written down as follows: below : v*2,)- ,3,y5Fg-l = 0 e= l,att* =0 (12) and +I= l,atr)= 1 (13) where (7) i.e. the dimensionless concentration, normalized with respect to the fixed value at the outer surface of the pellet, Ar 5,--8- FJ , (8) i.e. the dimensionless position of the reaction front within the grain, and $$ = 0, at r) = 0. The system of Eqs. (5), (6) and (12)-(14) is analogous to that derived earlier[5] in a dimen- sional form. It does, however, provide a more general treatment in that it allows the considera- tion of spherical, cylindrical and flat pellets, made up of spherical, cylindrical or flat grains. It is noted that the dimensionless representa- tion of the governing equations shows that the dependent variables 5 and I/J are related to t* and 7 and the single parameter, which appears in this formulation as cr. FJ , This quantity u incorporates both structural a=A, J ( 1 - l ) kc& A, D, v, (9) and kinetic parameters and, on the analogy of the Thiele modulus in heterogeneous catalysis, should be helpful in the characterization of the Here, u is a parameter which has a role akin to system. that of the Thiele modulus in heterogeneous In general the governing equations have to catalysis, and F, is the shape factor for the be solved numerically. For most practical pellet, which has the values of 1, 2 and 3 for purposes the results sought are not the time plates, cylinders and spheres respectively. and spatial dependencies of I,!J and 6, but rather the extent to which the pellet has reacted at (10) a given time. In terms of the parameters used in the formula- 765 CX.5 Vol. 27 No. 4-H tion, the overall extent of reaction is defined as I 1 X= O +Fp-l(l -p) drl s : (15) $FP- &-, 3. RESULTS In this section we shall present solutions for the governing equations, obtained generally by numerical techniques, together with some analytical solutions available for certain special cases. Before the presentation of the results it is worthwhile to examine the asymptotic behavior of the system as this may provide useful guide- lines, both for the range to be covered by com- putation and for the most convenient form in which the results may be presented. 3.1 Asymptotic behavior (i) When u approaches zero, the overall rate is controlled by chemical kinetics, and diffusion within the pellet is rapid compared with the rate of the chemical reaction. Under these conditions the reactant concentration is uniform throughout the pellet (9 = 1). Furthermore, 5 is independent of 7, although still a function oft*. Under these conditions Eq. (6) is readily integrated to obtain 5 = 1 - t*, for 0 d t* d 1 andt=O, fort* 2 1 * (16) Upon using Eq. (15) we obtain the following relationship between X and t*: t* = l- (l-x)F~. (17) For subsequent use, let us define &?QJX) = l- (l-XX)lFs (18) where the quantity gFg (X) refers to the value of t* within the asymptotic regime where u --j 0. It is noted that Eq. (17) provides a convenient means for determining F,, i.e. the grain or H. Y. SOHN and J. SZEKELY particle shape factor from experimental data obtained under conditions where diffusion within the pellet was not rate-controlling. (ii) When u approaches in$nity, the overall rate is controlled solely by the diffusion of the gaseous reactant within the pellet; this case corresponds to that accurately represented by the shrinking core model, thus the solution may be written as: pFp(X) = 2FF* (19) where pFp(X) = X2, for Fp = 1 (20) pFz,(X) =X+(1-X)ln(l-X),forF,=2 (21) p+(X) = 1-3(1-X)23+2(1-X), forF,=3. (22) The form of Eq. (19) suggests the following generalized definition for the parameter CT: Vp = d2iQFp = A, J (23) In teIIm of &, p&(X) may now be written as pFp(X) = t*@. (24) Equation (24) provides a convenient yardstick to which other overall extent vs. dimensionless time curves may be compared. 3.2 Analytical solutions Equations (17) and (24) correspond to asymptotic solutions for pure chemical control and pure diffusion control respectively. Analytical solutions may be obtained for certain special geometries of the general case, where both diffusion and chemical kinetics play a role in determining the overall rate. Thus for a spherical pellet (Fp = 3) containing 766 grains shaped as flat plates (F, = l)*, and for of the pellet, f2C1, is given by the following ex- a first order reaction, Ishida and Wen[2] ob- pression for both the above cases (i.e. spherical tained a solution, which may be put in the pellet, plate type grains, and slab, with plate following form. type grains): X=&[&coth (&&)-lJt*,O dldl (25) andforr L 1, t;=, = l+Y. (33) The solution for a long, cylindrical pellet with t* = 1+a(l--q,)2(1+2?jm)+ (l--r)& x [&&qm coth (<?&/,) - l] (26) flat plate-type grains is expected to be an inter- mediate between that for a sphere and a slab; it is thought likely, therefore, that Eq. (33) should also hold valid for this case. 3.3 The general case- numerical solutions and X = 1- qm3 + %[tiirn coth (tibm) -11 (27) where q,,, is the position of the boundary between the reaction zone and the completely reacted layer. A solution may also be generated for a slab, made up of flat plate shaped grains, i.e. F,, = 1, and FB = 1. For a first order reaction, i.e. n = 1, we have the following: forOGt*G 1 5= 1_coSh (--$* cash (fib> (28) and x,tanh (*h)t* v% * (29) Fort* 2 1, t* = 1+*-2(1-7jm)2+V%(1-rJm) x tanh (fibm) (30) 5 = 1 _ cash <fiM cash (fiti,) (31) and X= l-,,+A& -tanh (fiti,). (32) The time required to achieve complete reaction A structural model for gas-solid reactions In the preceding Sections, 3.1 and 3.2, we presented asymptotic solutions for & -+ 0 and u += w and analytical solutions for the particular cases of flat-plate-type grains, for which the interfacial area available within a grain does not change in course of the reaction. In the general case, the system of equations (5)-(6) with the boundary conditions (12)-(14) has to be solved numerically. These equations are identical in form to those solved previously; indeed with minor adaptation, the same com- puter program was used in order to generate the solutions. The computational details are available elsewhere [6], therefore we shall confine ourselves to the presentation and discussion of the computed results, which are shown in Figs. l-8. Figures 1 and 2 show computed results for small and large values of 6, respectively. The coordinates for these plots are so chosen as to allow the convenient presentation of the appropriate asymptotic solutions, which were derived in 3.1. *The shape of the grains is an interpretation supplied in the present paper; in their original article Ishida and Wen[21 merely postulated that the rate of reaction is independent of the solid reactant concentration. Thus Fig. 1 shows a plot of gFg(X) against t* +th 6 as a parameter, for various values of the two shape factors, Fp and F,. Also shown is the line for & = 0, i.e. gFg(X) = t* which, according to Eq. (18) is valid for all the geomet- ries. It is readily seen from Fig. 1 that the 6 = 0 asymptote should be appropriate, provided 6 > 0.3; this information is helpful, since it obviates the need for any detailed computation within this region of & values. Another important feature of Fig. 1 is that 767 H. Y. SOHN and J. SZEKELY 0.0 0 2 0.4 0.6 0.6 1.0 1.2 1.4 1.6 1.6 20 i Fig. 1. Conversion function vs. reduced time for small values of B. 0.0 0.2 0.4 0.6 0.6 1.0 1.2 1.4 1.6 I.6 2.0 t/ 6 Fig. 2. Conversion function vs. reduced time for large values of $. Fig. 3. The dependence of time for complete conversion on &. the straight-line relationship between gFg (X) and t* persists for t* < 1. For small values of u (say & < 0*7), the straight-line relationship exists until the conversion becomes 90 per cent or higher. This shows that the existence of a Stkght-fine relatiOnShip between gFg(x) and t does not necessarily mean that the system is under a chemical-reaction control; even when there is a substantial effect of intrapellet diffusion, one may obtain a linear relationship. This is consistent with the findings reported by St. Clair[lO]. For pellets made up of flat grains (F, = 1) the linear relationship is given by Eq. (29) for slablike pellets and by Eq. (25) for spherical pellets. Approximate equations may be obtained for small values of B: ForF,= FQ= 1, 768 A structural model for gas-solid reactions Fig. 4. The dependence of time for 90 per cent conversion on 6. Fig. 5. The dependence of time for 75 per cent conversion on B. x= 1+++L.. . . ( > t* (34) andforF,=3andF,= 1 Fig. 6. The dependence of time for 50 per cent conversion on B. Fig. 7. The dependence of time for 30 per cent conversion on 6. x= 21 8, l-y++&-*. (35) As diffusion becomes important (& becomes large), the slope of the straight line becomes 769 Fig. 8. Generalized plot for reaction time vs. conversion. smaller than would be in the absence of the diffusional effect. Thus, for the reduction of Fe& by hydrogen, St. Clair[lOl, obtained an activation energy of 25 Kcal/g mole consider- ing the diffusional effect, compared with a value of 15 Kcal/g mole obtained by McKewan 1111 assuming the reaction to be controlled by chemical reaction alone. Figure 2 shows a plot of pFP(X) against 1*/? with & as a parameter, for large values of &. As in the previous graph, the choice of the coordinates was dictated by the form of the asymptotic solution, i.e. Eq. (24). It is seen on inspection of Fig. 2 that the & -+ 00 asymptote is approached when 6 > 3.0 for all the geometries considered. This finding is again helpful, because it provides a definition for the range of validity of the shrinking core model, in terms of the struc- tural and kinetic parameters of the system. Again, it is of interest to note that the straight- line relationship between pFP(X) and t persists even when the value of + does not correspond to infinity (& > 2). This shows that the existence of a straight-line relationship betweenp,JX) and t does not necessarily prove that the system is H. Y. SOHN and J. SZEKELY controlled by intrapellet diRusion alone, unless the straight line passes through the origin and extends up to X = 1. When 2 < 6 < m, a straight line will be obtained which upon extrapolation does not pass through the origin. The data on the reduction of iron oxide obtained by Kawasaki et al. [ 121 clearly show this effect, indicating that chemical reaction played a significant role, in contrast to their conclusion that the system was controlled by diffusion alone. For pellets made up of flat grains (F, = 1) the approximate equation for the linear relation- ship can be obtained from Eqs. (30) and (32), or Eqs. (26) and (27) to give $ = PFB(X) +& In case of such straight line relationship the form of Eq. (36) provides a convenient way of obtaining the reaction parameters for the system. Thus the extrapolation of the linear portion of the curve to &,(X) = 0 will give the kinetic parameters, whereas the effective diffusion coefficient may be deduced from the slope. Except for the asymptotic regimes, in Figs. 1 and 2 the different geometries had to be re- presented by different curves. A more unified representation is possible, if we consider the time required to achieve complete reaction, or a given degree of reaction, as the dependent variable. Such plots are shown in Figs. 3-7. Figure 3 shows a plot of l/f$_1 against 6. Also shown, with the broken line, is the asymptote corresponding to cr 4 00, i.e. the relationship given by Eq. (24). As indicated by the single continuous line, the curves for Fp = 1 and FB = 1, Fp = 3 and F,=3, and F,=l and F,=3 all coincide, thus it appears that the quantity l/& may be represented as a unique function of 6 irrespective of the pellet or grain geometry, at least for the geometries considered in the formulation. The actual equation describing the continuous line in Fig. 3 has been previously given as Eq. (33): t& = 1+@. (33) 770 A structural model for gas-solid reactions Since the ordinate may be regarded as the ratio of the time taken to complete the reaction for i? = 0 to the time taken to complete the reaction for the actual value of 5, the quantity may be regarded as the overall effectiveness factor for the complete reaction. This behavior shows a marked analogy to that found in heterogeneous catalysis, with regard to the role played by the Thiele modulus. Figures 4-7 show similar plots to that given ;n Fig. 3, but for the times required to achieve a given fractional completion of the reaction. As seen on inspection of these graphs, the quantity t* can no longer be represented as a unique function of C although the results for the different geometries may be brought quite close by the appropriate choice of the co- ordinates, in particular, for the range, X > 0.5. The coordinates used in this instance were & tx* , i.e. the ratio: time required to achieve a given degree of reaction with & = 0 time required to achieve a given degree of reaction for the actual value of 6 and The results shown in the preceding Figs. 3-7 are summarized in Fig. 8, with X as a parameter. In these curves we have averaged out the scatter observed for the various geometrical factors, which as noticed would become signifi- cant only at low levels of fractional reaction. It is thus noted that the following general relationship may be used for predicting the extent of reaction: (37) or t* = gFg (x) + ?pFp(x). Thus we may conclude from our computation that the time required to achieve a given frac- tional reaction may be predicted through the use of Eq. (37). Figure 9 shows the comparison between the exact solution and Eq. (37) for Fg = 1 and Fp = 3, and Fg = Fp = 3. It is seen that Eq. (37) gives a satisfactory representation of the exact solution, especially for X > 0.5. The comparison is made for the intermediate region (a = 1 and 2), where the difference is largest. For smaller and larger values of & the agreement is better than shown in this figure. Furthermore, Eq. (37) is asymptotically correct as I? + 0 or 6 + m. 4. THE EFFECT OF EXTERNAL MASS TRANSFER In the preceding discussion we assumed that external mass transfer was not rate controlling. Fig. 9. Comparison of Eq. (37) with exact solutions for (T = 1 and 2. 771 In laboratory scale experimental studies such conditions are readily realized by using sufficient- ly high gas velocities; however, external mass transfer may be fully or partially rate controlling in practical systems involving packed or fluid- ized beds of solid particles. When the effect of external mass transfer cannot be neglected, the governing Eqs. (5) and (6) and the boundary conditions (12) and (14) still apply but boundary condition (13) must be replaced by the following: ~=ml-6), atr)= 1 (38) where (39) Figure 10 shows the effect of the modified Sher- wood number, Nib on the plot of the conversion vs. time for three different values of 6. ---F.*3, Fp*3 Parameter: ; -* p 0.0 1.0 2.0 t/ BL Fig. 10. Effect of external mass transfer on the relationship between conversion and time. H. Y. SOHN and J. SZEKELY The solution for a spherical pellet made up of flat grains may be obtained analytically[2] whereas the case of a spherical pellet containing spherical grains was solved numerically. It is seen in Fig. 10 that when 6 is small, variations in Nib will play only a minor role, which is to be expected, because in this case the system is chemically controlled. While N& may play an important role in deter- mining the overall rate for larger values of u, the effect of external mass transfer becomes unimportant when NL,, > 30. Let us now generalize the results obtained in the preceding sections for Nib + CO to allow for finite values of the Sherwood number. The rate of reaction can be expressed in terms of the overall rate of conversion X or the rate of diffusion through the gas film. dN..+ _ --- (l-E)PJ* dX dt b dt (40) dNA _ - - --Apkm(C~o--AS) dt (41) where CA8 is the gaseous IWAant concentration at the external pellet surface. Upon assuming quasi-steady state, the term dX/dt in Eq. (40) can be obtained from Eq. (37) with CA, replacing CA, in the definition of t *, to give 1 g;,(x) + ?&J(x) * (42) Substituting Eq. (42) into Eq. (40) and equating with Eq. (4 1) we obtain 23 + N;/,{&,((x) +&$,(x)} (43) Finally, on combining Eqs. (42) and (43) we obtain an approximation for t *. t* = &W) +s (4) Figure 11 shows the comparison between the 772 A structural model for gas-solid reactions 5.3 5=3 Exact solution - -a- f*OF*w+&*(PFp(XI+~) _-__ -+- 0.0 0.5 1.0 1.5 2.0 f Fig. 11. Comparison of Eq. (44) with exact solutions. exact solution and Eq. (44) for 6 = 1-O and NLh = 3. It is seen that Eq. (44) gives a satis- factory representation of the exact solution. For smaller and larger values of 6 the agreement is better than shown in this figure. Furthermore, Eq. (44) converges to the exact solution as 6 + 0 or & + 03. The asymptotic relation for large 6 can be obtained analytically from the solution of Ishida and Wen[2]. For X = 1 Eq. (44) again becomes identical to the exact solution. This was also checked for the case of FB = Fp = 1 for which an analytical solution is possible [ 171. The implications of the above findings are discussed in the subsequent section. 5. DISCUSSION In the paper a dimensionless representation was developed for the reaction between a porous solid and a reactant gas. The formulation was quite general, in that it allowed the pellet to be a slab or a sphere (cylinder by interpolation) and the grains making up the pellet could also assume a spherical, or slablike shape. By the appropriate manipulation of the governing equations it was possible to obtain a general set of solutions on the plot of the extent of reaction against a suitably chosen dimension- less time, with a single quantity & as the para- meter. The dimensionless time required to attain complete reaction of the pellet could be re- presented as a unique function of a generalized reaction modulus, 6 through a relationship which was valid for all the geometries considered. The potential practical application of these results appears to be threefold. (i) Prediction of the reaction characteristics of porous solids. (ii) Providing insight into the general behavior of gas-solid reaction systems. (iii) Providing a framework for the interpreta- tion and planning of experimental studies involv- ing gas-solid reactions. (i) prediction of the behavior of gas-solid systems. If information is available on the quantities that make up &, the time taken to achieve complete reaction, or a given fractional reaction, may be calculated with aid of Eq. (44), without recourse to any digital computation. If the overall rate of reaction is desired, this may be readily obtained by the differentiation of Eq. (44): dx 1 dt = g;,(X) + S-Q&(X) +2/N:,,} * (45) Another useful feature of the dimensionless representation, given here is that the value of &, the generalized gas-solid reaction modulus, provides quantitative guidance whether the system falls within one of the asymptotic regimes. It has been shown earlier that when & < O-3 the pellet will react at a uniform rate throughout, and when 6 > 3-O chemical kinetics are not rate controlling and the use of the shrinking core 773 H. Y. SOHN and J . SZEKELY asymptote would be appropriate. As discussed in the paper, these considerations apply to both spherical and slablike geometries, and are likely to hold, at least as a reasonable approximation for most other geometries to be encountered in usual practice. (ii) Some general characteristics of gas-solid reaction systems. In a previous publication[5] it was shown that one attractiveness of the grain model was its ability to suggest certain optimal geometric configurations, which would give the fastest overall reaction rate for a given situation. The dimensionless representation given in this paper allows a more general statement of this problem. In order to find the optimum values of the structural parameters giving the maximum average rate of reaction up to the complete conversion, we write this quantity, R,,,, as follows with the aid of Eq. (37): For a given value of k and structural parameters, u is specified provided D, the effective diffusivity is known. The effective diffusivity is independent of the grain size, if diffusion within the pellet in the molecular regime; however, D, will be a function of the grain size, if diffusion is in the Knudsen or in the transition regime. Let us consider a pellet made up of spherical grains; then the Knudsen diffusion coefficient may be estimated with the aid of the dusty gas model proposed by Mason et al. [ 131: where DeK is the effective Knudson diffusivity r, is the radius of the grain, E is the porosity of the pellet, T is the tortuosity, MA is the molecular weight of the species A, R, is the universal gas constant, and T is the absolute temperature. The effective diffusivity may then be obtained with the aid of the Bosanquet interpolation formula: 1 _1+1 D,- De+, D~K (48) where DeM = ED,&- is the effective molecular ditfusivity. We may now proceed to obtain the conditions for the maximum R,,, in terms of the grain size; thus by using Eqs. (23), (47) and (48) we have F kCn-Kr rAol3t= (1-e) p J 2: (49) where r,lOPt is the optimum grain size for a given constant porosity. It is noted from Eq. (49) that for a given pellet porosity the optimum grain size is proportional to the pellet size, and is independent of the molecular diffusivity; this behavior is consistent with physical reasoning, since the grain size will not have an effect on the performance of the system, at constant porosity and under conditions where the overall rate is controlled by molecular diffusion. By analogous considerations we may also obtain an expression for the optimum porosity, for a fixed value of r,, which is given as E -1 =!NJ~) (50) 1 - l opt r o where K= (l+$)($$$z (51) (iii) The use of the model for the planning of experimental studies and the interpretation of experimental data. As discussed earlier, all the previously available structural models had the inherent drawback of requiring extensive com- putation before they could be used for the interpretation of experimental data. The model described in this paper could be useful for both the planning of experimental studies and the interpretation of experimental data, without recourse to digital computation. 774 (a) Planning of experimental studies In the study of a given system it is desirable to define the range of variables over which a given mechanism controls the overall rate. Ideally, one would wish to carry out a series of rate measurements under conditions, where the process is chemically controlled, i.e. when & + 0. Within this regime a straight line plot, according to Eq. (17) should enable one to obtain a suitable average value for F,. While it is desirable to corroborate such find- ings through the use of electron microscopic techniques, this latter procedure may become quite cumbersome, especially in the case of irregularly-shaped grains. The intrinsic reaction rate constant, k could also be obtained from such data, through the use of Eq. (17) and the definition of t* given in Eq.(ll). Once information is available on F,, the other grain parameters, and on the intrinsic chemical rate constant, the model would enable one to predict the behavior of the system over the whole spectrum of possible operating conditions - within the limitations contained in the simplifying assumptions made in the derivation. The procedure outlined above should be rather more satisfactory than experimental studies working exclusively with pellets, where the principal variable would be the operating temperature. Factors other than the reaction rate constant may be strongly temperature- dependent (sintering, etc.); thus the apparent activation energies obtained from reaction studies at different temperatures could be quite misleading. (b) I nterpretation of experimental data The model can be used for the interpretation of experimental data obtained using pellets or larger particles. Typically, these data would consist of information on conversion vs. time, and possibly data on the structural parameters Under these conditions, in the absence of separately measured kinetic constant and diffusion coefficient, Eq. (37) enables one to determine these parameters from the data of A structural model for gas-solid reactions time taken for complete reaction. Thus Eq. (37) may be put in the following form: bP tx=l_ 1 - T - k,,e-ERgT A, R,P, The use of Eq. (52) should enable one to estimate the value of the group [ (1 - l )FJ/2D, from experiments carried out with pellets of different sizes; it is noted, however, that such procedure is unsatisfactory means of estimating D, when diffusion is not rate controlling, although this state of affairs would be readily deduced from the small values obtained for the above group under such conditions. In principle the use of Eq. (52) would allow us to obtain information on the V,F,/A&, and the E/R0 groups from experiments conducted at different temperatures, although this procedure may not be entirely satisfactory because at extreme temperatures other effects may come into play. Thus at very low temperatures one may encounter induction periods [ 141 requir- ing the nucleation of a new phase, whereas at higher temperatures sintering may occur. The representation of these phenomena would require rather more involved kinetic expressions than what were given in this model. It is noted, however, that the form of Eq. (52) conveniently separates the diffusional and the chemical kinetic terms, so that the relative importance of these variables may be readily assessed, even under conditions where the accurate evaluation of the individual groups is not possible. Having the approximate expression given by Eq. (37), one might be tempted to put it in the following form: gFgo_ bG, A, t PFp (X) ---(ye pm FJ , PFJ X) (53) A linear plot of gFg (X)/pFp(X) against t/pFp(X) for a single experiment would give the values of k and 6 (hence D,). Unfortunately, however, this does not work, due to the approximate nature of Eq. (37), especially at a low conversion, 775 H. Y. SOHN and J. SZEKELY where the term on the left hand side is the ratio of two small numbers, and thus is very sensitive to errors. However, as shown in Section 3.3 of the text, when 6 approaches 0, or infinity, the use of Eqs. (34) or (35) and (36) respectively, would provide a reasonable estimate of these para- meters. 6. CONCLUDING REMARKS In the paper a structural model was proposed for describing the first order, irreversible reaction between a porous solid and a gas. The model was stated in quite general terms, so that allowances could be made for both the grains and the pellets having slablike and spherical shapes. The behavior of cylindrical grains may be estimated by interpolation. The formulation was developed in a dimension- less form and all the systems considered could be represented on a curve, relating the time required to achieve complete reaction of the pellet to a general reaction modulus. The time required to achieve a given fractional conversion of the solid could not be uniquely described on a single curve, but rather as a family of curves with a geometric factor as a parameter. The model is thought to have potential for both the interpretation of experimental data and for the prediction of the behavior of gas-solid reaction systems, on the basis of a limited number of measurements. The simplifying assumptions made in the derivation of the model will clearly limit its applicability and it may be worthwhile to comment on the most critical of these. (i) The assumption of first-order kinetics for the chemical reaction may not be appropriate to many practical systems, nonetheless, judging from prior experience on hetergeneous catalysis [ 15, 161 the net effect of higher order rate expressions is unlikely to alter drastically the general nature of the relationships found in this investigation. (ii) Perhaps the most critical assumption has been the neglect of structural changes in course of reaction; such structural changes could either increase the diffisional resistance (sintering) or alternatively could lead to a more open pore structure. Appropriate allowance for these factors would be desirable, but any such work has to be pre- ceded by investigations concerning the kinetics of these structural changes. (iii) The isothermality assumed for the system may also be an oversimplification for many cases, and corrections for nonisothermality should be made in future studies. Nonetheless, it is felt that the model given here will apply to a number of practical systems and, at the same time, it should provide a useful starting point for further work in this area. Acknowledgements-The authors wish to thank the A. E. Anderson Foundation for financial support, and Dr. J. W. Evans for providing the computer program for the numerical solution. &, 4 b C* c A8 C A0 Dt? D t?K D au D&7 DM E Fw Fp &?Fg (x) k NOTATION external surface area of the individual grains and the pellet respectively, cm2 stoichiometric coefficient in Eq. (1) gaseous-reactant concentration, g moles/cm3 gaseous-reactant concentration at the outer surface of the pellet, g mole/cm3 gaseous-reactant concentration in the bulk, g moles/cm3 effective difIusivity of A in the porous pellet, cm2/sec effective Knudsen ditfusivity of A, cm2/sec effective molecular ditfusivity of A (ED&), cm2/sec effective diffusivity of A in the product layer of individual grains, cm2/sec molecular diffusivity ofA, cm2/sec activation energy, Cal/g mole shape factor for the grains and the pellet respectively (= 1, 2 and 3 for flat plates, cylinders and spheres, respec- tively) conversion function for small 6 defined by Eq. (1% reaction-rate constant, (cm3/g mole)- cmlsec 776 ko km K n dN,lclt %t PFp (x) HI 121 131 141 VI El 171 181 191 1101 [ill WI r131 t::; V51 u71 A structural model for gas-solid reactions pre-exponential factor, (cm3/g mole)-l T absolute temperature, K cm/set external-mass-transfer coefficient, set VA reaction rate for the species A, g moles/ cm/ cm2/sec V,, VP volume of the individual grains and the pellet respectively, cm3 the X conversion of solid reactant rate Greek symbols factor defined by Eq. (44) reaction order with respect to gaseous reactant concentration rate of formation or mass transfer per pellet, g moleslsec modified Sherwood number defined by Eq. (39) conversion function for larger $ defined by Eqs. (20)-(22) total pressure, atm E 7) %?I porosity of the pellet dimensionless distance coordinate de- fined by Eq. (10) dimensionless position of the boundary between the reaction zone and the completely reacted layer dimensionless position of the reaction front within the grain defined by Eq. (8) distance coordinate perpendicular to the moving reaction front in the t individual grains, cm radius of the grain, cm distance coordinate in the pellet, cm Pm average reaction rate defined by Eq. (39), g moles/cm3/sec u universal gas constant time, set 6 dimensionless time defined by Eq. (11) time for complete conversion, set dimensionless time for a given degree of ; conversion REFERENCES molecular density of the solid, g moles/ cm3 gas-solid reaction modulus defined by Eq. (9) generalized gas-solid reaction modulus defined by Eq. (23) tortuosity factor dimensionless concentration of gaseous reactant AUSMAN J. M. and WATSON C. C.. Chem. Enann Sci. 1962 17 323. ISHIDA M. and WEN C. Y.,A.Z.Ch.bJ I 1968 14311. CALVELO A. and CUNNINGHAM R. E.. J. Caral. 1970 17 1. SZEKELY J. and EVANS J. W., Chem. E&g Sci. 1970 25 1091. SZEKELY J. andEVANS J. W.,Chem. EngngSci. 197126 1901. SZEKELY J. and EVANS J. W., Met. Trans. 19712 1691; 1698. CALVELO A. and SMITH J. M., Proc. Chemeca In press. BISCHOFF K. B., Chem. Engng Sci. 1963 18 7 11. LUSS D., Can. J. Chem. Engng 1968 46 154. ST. CLAIR H. W., Trans. metall. Sot. A.I.M.E. 1965 233 1145. MCKEWAN W. M., Trans. metall. Sot. A.I.M.E. 1960 218 2. KAWASAKI E., SANSCRAINTE J. and WALSH T. J., A.1.Ch.E J l. 1962 8 48. MASON E. A., MALINAUSKAS A. P. and EVANS R. B., III, J . them. Phys. 1967 46 3 199. EVANS J. W., Ph.D. Thesis, State University of New York at B&do, 1970. PETERSEN E. E., Chemical Reaction Analysis, Chap. 4. Prentice-Hall 1965. SOHN H. Y., MERRILL R. P. and PETERSEN E. E., Chem. Engng Sci . 197025 399. SOHN H. Y. and SZEKELY J., Met. Trans. submitted for publication. RQum6-Les auteurs dkveloppent un modble structural general pour d&ire des reactions non- catalytiques gaz-solide; ce modble tient compte des pastilles spheriques ou plates faites de grains spheriques ou plats. La solution numerique sans dimension des equations gouvemantes donne une courbe qui montre la relation entre le temps, sans dimension, necessaire a laccomplissement de la reaction et le module de reaction gaz-solide; cette courbe est valable pour toutes les geometries considerees. 777 H. Y. SOHN and J. SZEKELY Lutilisation de ce graphique pourrait &tre utile au planning dexperiences ou a linterpretation de don&es experimentales, par rapport a des modbles structuraux et sans la necessite de calculs numer- iques. Les auteurs montrent que pour des valeurs extr8mes du module de reaction gaz-solide, les resultats du modele se rtduisent aux solutions asymptotiques correspondant respectivement aux modeles du noyau retrecissant et des grains reagissant uniformement. Ainsi le module de reaction gaz- solide est utile a la definition quantitative de la gamme de ces regimes asymptotiques par rapport aux parambtres chimiques et structuraux du systeme. Zusammenfassung-Es wird ein allgemeines Strukturmodell zur Beschreibung von nichtkatalytischen Gas-Feststoffreaktionen entwickelt, das kugelfiirmige sowie flache, plattenfiirmige Platzchen, die aus kugelfiirmigen oder flachen, plattenfdrmigen Kiimem bestehen, beriicksichtigt. Die numerische Losung der dimensionslosen Form der massgebenden Gleichungen ergibt ein Diagramm der dimensionslosen Zeit zur Erzielung vollstandiger Reaktion gegen einen Gas-Feststoff Reaktionsmodulus, gtiltig fur alle in Betracht gezogenen Geometrien. Die Verwendung dieses Diagramms konnte sich bei der Planung von Versuchen oder bei der Auswertung von Versuchsdaten im Sinne von Strukturmodellen als niitzlich erweisen, ohne digitale Datenverarbeitung erforderlich zu machen. Es wird gezeigt, dass fur Extremwerte des Gas-Feststoff Reaktionsmodulus, die Ergebnisse des Modells sich auf die asymptotischen Lijsungen entsprechend den Modellen mit schrumpfenden Kern bzw. gleichfi%rnig reagierendem Pfatzchen reduzieren. Der Gas-Feststoff Rwaktions- modulus ist daher brauchbar fiir die quantitative Bestimmung des Bereichs dieser asymptotischen Systeme, ausgedriickt durch die chemischen und strukturellen Parameter. 778