25 GAS-PHASE DIFFUSION COEFFICIENT: MEASUREMENT AND PREDICTION . sa yen 10 The experimental values of binary diffusion coefficients for several gas mixtures a arn Table 2.1. It is seen that for many small gas molecules, the values lie between ea (10° mis or 0.4 fi°%h) and 1 cm/s (10% m/s or 4 f1°/h) at atmospheric pressure raised temperature. The gas-phase diffusivity generally varies with the absolute temperature 1°, 5, the power 1.5 to 1.75, and inversely to the total pressure up to about 10 atm (Ds Dap & (1/P)]. It depends significantly also on the molecular weight of a species. As expected hinders intuitively, a large molecule diffuses slowly compared to a small molecule, and also cients ; 5 = cefficiem! * In an interesting article, Kraajjeveld and Wesselingh (1993) showed that in some cases the diffusion © of ions in a mixture of electrolytes or in an ion-exchange can even be negative.Table 2 Experimental tus values for sete, Gas palr Temperature Dyn 105 aa 6 (ols) Ar-H, ae 2 2762 142 2982 26 2982 03 2001 087 28 675 258 22 28 846 2 139 075 2.02 297 779 2982 "a2 307.4 815 263 io 278 185 He-Benxene 2068.2 sot diffusion of other molecules. For example, the diffusivi cm’/s at 0°C and 1 atm compared to 0.674 em2/s in nitros upon the intermolecular forces in a mixture and is also g molecules with others present in the mixture. The rate of ci decreases. This is favoured at a higher pressure, increasing pressure. There are a few methods of experimental determi Two simple methods are described here, ty of hydrogen in n-butane is 0.272 gen. The diffusivity strongly depends overned by collisions of the diffusing ollision increases as the mean free path and hence the diffusivity of a gas decreases with ination of the gas-phase diffusion coefficient. 2.5.1 Twin-bulb Method The apparatus used in this method consists of two reasonably large bulbs or chambers of volumes V; and V>, connected by a narrow tube fitted with a plug-type valve or a stopcock. There should be a suitable arrangement for stirring the contents of the bulbs in order to keep the concentrations in them uniform (the bulbs are said to be ‘well-mixed’). The entire assembly should be maintained at a constant temperature. A schematic sketch of the apparatus is shown in Figure 2.11. Initially, the valve in the connecting tube is kept closed. The two bulbs are evacuated. One of the bulbs is repeatedly flushed with pure A and the other with pure B and then filled with the gases at the same pressure, P. The valve is then opened to allow diffusion to start and continue undisturbed for some time, at the end of which the valve is closed. Samples of the gases from the bulbs are taken and analyzed for their composition. F Because the total pressures in the bulbs remain constant and equal, equimolar counter diffusion through the connecting tube occurs. The bulbs being large, the concentrations or partial Pressures of the components in the bulbs will change rather slowly. A theoretical analysis of38 Chapter 2 Molecular Diffusion Bulb | Figure2.11. The twin-bulb apparatus—a schematic sketch, diffusion in the system can be made with the pseudo-steady state approximation, 7 that, at any instant, diffusion through the connecting tube occurs at steady seo This meg centrations in the bulbs change a little, a new steady state of diffusion is attained sin aa Con. ‘The assumption is correct ifthe ‘time scale’ forthe change of the concentrations 4a" is considerably larger than the time scale for the attainment of steady-state diffusion ae bulbs ecting tube. The working equation for the calculation of the diffusion coefficient Re 2 ie conn concentration changes in the bulbs is derived below. Ifa is the inner cross-section of the connecting tube, / is its length, and ; : + and pay and p, (pas > psx) are the partial pressures of A in the bulbs at any time 4, the steady-state transport of A from bulb 1 to bulb 2 can be written as [see Eq. (2.30)] ale of — GDpw Pai = Par) Ng = ergy ae SI (2.46) If C4y and Cp are the instantaneous concentrations of A in the bulbs, the rates of change ofthese concentrations are given by the following equations: dC, yd “SG aay aah ma am aCe Vy dpa and V, —*2 =aN, a es 2.48) ani /, aN, => gp = aNa (2.48) Here Cy) decreases with time but C,p increases. Hence the negative sign in Eq. (2.46). Alo, wwe have used the relations Cyy = pay/RT and C4p = Pya/RT applicable to ideal gases. Combining Eqs. (2.47) and (2.48), aDyp (Pai ~ Pa2) ( dey ) d = art|t a eM pn) = ent(t + 1 “iw 2.49) dn = e/ (pai ~ Pa2) Lay We use the following conditions: = 1=0pu-po=P-0=P and f= 1, par par = Pai ~ Pho (say) + The pseudo-steady state approximation is used in a number of exercises and examples in this chapterGus-Phase Diffusion Coefficient 39 tion of Eq. (2.49) withi snegation a n the above limits gives the working equation for the calculation of Dan That is. InP = 2Pan (1, 1) M=maaaaia (Raval 2.50) ‘The s to be measured in this experiment are the initial pressure in the vessels and the partial pressures of one of the components (say A) in the vessels at the end of the experiment, ‘The mutual diffusion coefficient Dyy can be directly determined from Eq. (2.50). 2.5.2 Use of the Stefan Tube ‘This method is suitable if, under the given set of experimental conditions, one of the components (say A) is available as a volatile liquid and the other component (B) is a gas which is not soluble in A. The apparatus is very simple. A vertical glass tube, sealed at the bottom, is joined to a larger diameter horizontal tube to form a tee (T) as shown in Figure 2.12. The liquid A is taken in the narrow vertical tube and the gas B is forced through the horizontal tube. Evaporated A diffuses through the mixture of A and B in the vertical tube, reaches the top and is swept away by the flowing stream of B. As B is insoluble in A, it will not diffuse and the situation will conform to diffusion of A through non-diffusing B. The liquid level in the vertical tube will drop very slowly and pseudo-steady state assumption (i.e. A diffuses through the tube virtually at steady-state at all time) is reasonable. This means that as the liquid level falls by a small amount, a new steady- state rate of diffusion is established simultaneously. The drop in the liquid level over a period of time is noted. Gas B A+B yy = Vapour pressure of liquid 4 Volatile liquid LJ Figure2.12 A schematic ofthe Stefan tube. Let, at any time 1, the liquid level be at a distance z from the top of the vertical tube, pay be the partial pressure of A at the liquid surface and py that at the top. The diffusional flux of A through this distance z is given as follows [see Eq. (2.27)] Paw Play = Par) 2.51) RT=Pa MeChapter 2 Molecular Diffusion 40 If the fall in the liquid level is dz in a small time dr, the number of Moles of 4 ce ove ime dt, that gi is adz(p,/M,). By a material balance over the time di, iz Paw PUP = Pap %y adzpy _ anat= 4 ui MP2) oy, My Past = ction of the vertical tube, py and M, are the de e a is the inner cross-section o! : fi Si (ogy. Pa weight of A respectively. If at time ¢ ites liquid level j AL 29 from he lng atime? .at the end of the experiment) the liquid level is at z’, integration Of the Above. "°D, ay, and rearrangement gives “ation Day = RE Pom Pale” = 25) 8 2PM (Pu — Paadt 2x iquid surface, pay, is equal to its vapour pressure artial pressure of A at the liquid sur ‘ ° al the ; eee At the open top of the tube, the partial pressure of A is Virtually zero ring because 4 is greatly diluted by the gas B flowing ata high rate. All the quanti being = Dag can be calculated om 2.5.3 Predictive Equations for the Gas-phase Diffusivity Although experimental diffusivity values fora large numberof binary gas mixtures are availa \we often come across mixtures for which no experimental data have been reported, In such age” we take the help of a suitable predictive equation or correlation for the estimation Of diffusvng Many such predictive equations—theoretical, semi-empirical orempirical are ain’ A detaig aecount of the more important equations and their suitability is given by Poling et ah 2001, A useful and reasonably accurate theoretical equation based on the Kinetic theory of gar Wwas suggested independently by Chapman and by Enskog. The diffusion coefficien Dan strong depends upon binary interaction parameters of the A-B pair. Chapman and Enskog used the Lennard-Jones potential function (Chapman and Cowling, 1970) given below te calculate te interaction parameters 2 rove Or) = «((2) -(2) ] 3) FOr where (7) = potential energy = distance between the centres of two molecules €and 6 = Lennard-Jones potential parameters ‘The Chapman-Enskog equation is given by 7 pe z v2 Du = SSSI? TRUM + My)" rie 58) POA Qy Here solute temperature, in K May My = molecular weights of the components A and BGas-Phase Diffusion Coefficient 4) total pressure, in atm acteristic length parameter of the binary, in A Pp Onn = Cl Qp = collision integral which is a function of k7/eqy (where k is the Boltzmann’s constant and €4y is another characteristic binary parameter). “the following equations can be used to calculate Gay and Exp. Gay = (Oy + Op2: Egy = Lene)? (2.55) ‘The values of and £ are listed for anumber of common substances in Table 2.2 and the collision Jntegral, 2p. is given in Table 2.3. The values of 6 and € for substances not listed in Table 2.2 lated from the following approximate relations: can be calcul: chk =0.75T, and = (5/6)V-1° (2.56) Table 2.2 Lennard-Jones potential parameters for selected compounds Compound o ek Compound a oh (A) (K) A) « He 2551 1022 Air 3m 7a8 CCl 5947 327 CHCl; 5369 3402 CHCl 4.182 3500 CH,OH 3626 4818 cH, 3758 1486 cs, 4489 4670 CHa 4168 247 CHs 449 2187 CoHsCl 4898 3000 | C.HsOH 4500 626 Propylene 4678 2389 ‘Acetone 4600 5602 Butane 4687 5314 Di-ethylether 5678 338 mHexane 5.949 3993 Cl, 4217 60 Ethyl acetate 5205 5213 HCI 3309 MAT HBr 3353 449.0 He 2827 597 H,0 2641 209.1 HS 3 wo NHy 2900 5583 NO 3492 1167 Ne 3798 nd Op ‘3467 106.7 SO, 4i2 (3354 Benzene 5349 4123 where 7, is the critical temperature (K) and V, is the critical volume (cm?/gmol) of a component. Use of the Chapman-Enskog equation has been illustrated in Example 2.11 ‘An empirical equation suggested by Fuller, Schettler and Giddings (1966) is not only simple to use but also reasonably accurate in predicting binary gas-phase diffusivity up to moderate pressures: 1.0133 1077 71° [ 7 Wa = Seale mt | m/s (2.57) PLENe +(e P ue M Here T= temperature, in K My and My = molecular weights of A and B respectively P = total pressure, in bar.44° Chapter 2_ Molecular Diffusion A diffuses towards the surface — Cy = Oat the surface of the catalyst particle since the reaction is instantaneous \ > Liquid-film Cy = Cyyat the edge ofthe liquid Figure 2.13 Diffusion through the fim on a spherical pellet. x10 m: D=7 x 10°! m/s; C4 = 1.0 kmol/m’; and the rate of Given: r,=0.5 mm othe flux of the reactant to the surface of the catalyst particle, Ny = 3.15 x 10-® mop on lim. > Ws dar Ny = 4nDCy, Ete 8) . ; mS a 5 pate = Nate, _ GISx1O*VSx10“) 5 DCs (7107 )(1.0) = 1 + (5 x 10-4)/5 = 2.25. The liquid-film thickness, 5 = 0.0004 m = [0.4mm] eel 2.6.2 Experimental Determination of Liquid-phase Diffusion Coefficient There are a number of methods for the determination of the liquid-phase diffusion coefficient Here we describe a common method that uses a “diaphragm cell” (Dullien and Shemilt, 1961). The diaphragm cel is a two-compartment cell, as shown in Figure 2.14, separated by a porous diaphragm usually of sintered glass. Solutions of the species A in the solvent B at wo different concentrations are taken in the two compartments. Diffusion is allowed to occur through the narrow passageways of the pores from the higher concentration cell to the lower concentration cell for some time. At the end of a run, samples of the solutions are taken from the cells and analyzed for the Diaphragm Feed/sampling Concentrations of A. The contents of the compartments \ are always kept well-stirred. If the area of cross-section of the diaphragm is a and € is its porosity, then the effective area available for diffusion is ae. However, the length of the diffusion path is not equal to the thickness of the phragm since the pores are not straight. Many of the O pores are interconnected and some may even have dead ends, A simple method to take into account this characteristic of the pores is to introduce an empirical magnet parameter called “tortuosity factor’, 7 Itis the ratio of the 7 line average length of the diffusion path (through the Feadsop diaphragm) to the thickness of the diaphragm. Figure 2.14 The diaphragm cellMolecular Diffusion in Liquids ‘The working equation for the calculation o| with a diaphragm cell can be derived follows pulb method (Section 2.5.1). For diffusion in fsce Eq. (2.17)), it is reasonable to neglect t form of Fick’s law is It Fan > Car (2.62) Here 1 is the thickness of the diaphragm, and Iris the effective lengih of the dil Vand Ysa the volunes of hesluionsinhetwocelsand Cy ame ae en Ea therein at time ¢, then by invoking the arguments given in Section 2.5.1 tthe Pseudo-steady stat approximation), vata dCy, dy iar = AEN, and Vy “C42 gen, (2.63) Adding the above two equations and substituting Ny from Eq, (2.62), = ae Pan Ge) (1 It \ d a Ga Ge (2.64) If Caio and C429 are the initial concentrations and Car. and C42, pare the final concentrations in the compartments at time fy integration and rearrangement of Eg, (2.64) yields te (11) ).(Gis—Gey Dag = — — — oy i. Gey (re Tei Cay aa Equation [2.65(a)] can be used to calculate the diffusion coefficient from the measured concen- trations and time. However, itis not easy to determine the effective porosity e and the tortuosity factor T of the diaphragm. It is more convenient to carry out separate measurements for a solute of known diffusion coefficient in the given solvent to calibrate the cell. Then the quantity ae ea 2 wet } B (say) (2.65b) can be calculated for a particular diaphragm cell. The quantity Bis called the “cell constant’, Once the cell constant is determined, the diffusivity of any species in any solvent can be measured. Potassium chloride—water is a common system used to determine the cell constant. Sample calculations are illustrated in Example 2.13, An apparatus based on similar principles, called the “Wilke-Kallanbach diffusion cell’, is used for the determination of effective diffusivity of a gas in a porous solid (for example, a catalyst pellet).Chapter 2_ Molecular Diffusion japhragr two-com . EXAMPLE 2.13 (Diffusivity determination —tiaphray cel?) Pa wo-con patent dig, cell is used to measure the diffusion ca ist a Pont neta The cay es Sa ia er Seine 4 calibration experiment is done by filling yg nets I eae KCl and the smaller compartment with m. The concent cormpartment with 0. molar agucous station Te é Molecular diffusion of the solute occu ra a iy of RCS non & a Own 315 mola nd of 5 of KCI redues 10 0.315 meat J temperature of 18.5°C. Calculate the cen, cons tan to be 1.51 x 10° em*/s at the experimental 46 from these data. — ents fille cashed and the larger compartment is filled with 0,4 The cell is then thoroughly washed an re i Olle oi aqueous propionic ‘acid, The smaller compartment is filled iat ie Beto AUthe ena 56-4 h, a sample from the larger compartment is found to con ain PSE molar acid. Caley the diffusivity of propionic acid in water. Ifthe average porosity of the diaphragm is 0.39, 5% is the tortuosity of the pores? working equations for this problem. The end. concentrati 2 on Solution Equations [2.65(a) and (b)] are the c ae compartment ias to be obtained by material balance in each case, Diffusion of KCI: ViCaso + V2Ca20 = MiCang + V2Cane V, = 60.2 em’; V2 = 59.3 cm’; Cx = 0:3; Caro = 0: Cary = 0.215 Given: yxy = 0.0863 molar Other quantities given are: D = 1.51 x 10% em//s; time, ty = 55.2 h = 55.2 x 3600 5 The cell constant can be calculated from Eq. [2.65(a)]. z 1 in 03-9 =n Tavs aon; = [0282 en] (151x107 )(55.2 x 3600) => Diffusion of propionic acid: Given: Cuyo= 04: Cyz0= 0: Cars = 0.32; then by material balance as above, Cya,, = 0.0812. also the time allowed is f= 56.6 h = 56.4 x 3600 s. Substituting these values in Eq. [2.65y, 1, Lao ~ Ca2.0 1 04-0 — h — = fo ——_—_ > 6 one 14% 3600) "032 — 0.0812 [901x10""em™s] D= ab jy =A A200 S By Cary -Crag — (0.282)C (This value compares well with the literature value, 8.74 x 10-6 em Cateulation of the tortuosity factor: Cell constamt, p = ££ (+ + i) Ty ¥, 0.18 cm; B = 0.282 Given: area of the diaphragm, a = (m/4)(3,5)? = 9,621 em* € = 0.3 = Tortuosity factor, ¢ = nae u fel [ (0.282)(0.18) \ 60.2” 593) = =>.Molecular Diff min Liquids 47 2.63 predictive Equations for Liquid-phase Diffusivity rye magnitude of (he liquid-phase diffusivity is much smaller than that of the gas-phase diffusivity, The ns bref list of experimental iguid-phase diffusivity values of several common subsanecg From in Table 2. itis apparent that for diffusion of small solute molecules in water, the site generally Hes between 0.5 10-* cm/s and 2 x 10 ems. A larger molecule has a smaller value diffusivity Table25 _Liquid-phase diffusivities of selected compounds atinfrite dilution at 25°C Solvent Dg x 10 (mls) Solute ‘Solvent Dig % 10° (més) Caibondiowde Wile 192 Benzvicacid Wer 100 Chlorine Weler i=) scotoh, Water 116 iti oxide Water 260 Acetone (Chloroform 235 onal Water 210 Benzene (Chloroform 289 ‘ammonia Water 16 Acetic acid Benzene 209 Hydrogensulphide Water 1a Benzoicacid Benzene 1B Sulphuric acid Water 4173 Water Acetone 456 Nitric acid Water 260 Benzene try alcohol 181 Methanal Water 16 logine try! alcoho! 12 ‘Ethanol Water 128 Water Ethy! alcohol 124 Formic acid Water 150 Water mButylaicohol 8a Aeatic acid Water 121 Waker Ethyl acetate 320 al equations and correlations are available for the prediction of the liquid-phase diffusivity. But generally the accuracy of prediction is not as good as in the case of gases. This is because the theory of the liquid state is less clearly understood. The more important equations for the estimation of the liquid-phase coefficient are described by Poling et al. (2001). Hayduk and Minhas (1982) proposed a simple and useful correlation, Here we cite the Wilke— Chang equation (1955) which is simple to use but not always good in prediction. ‘A number of semi-empiri 1.173 107" (@ Mg)!?T wy? Die (2.66) Here Dy is the diffusivity of the solute A in solvent B in m°/s; the superscript *0° refers to an infinitely dilute solution. The different quantities involved are: Mg = molecular weight of B (the solvent); g = association factor for the solvent [water: g= 2.26; methanol: g = 1.5; ethanol: P= 1.5; non-associated solvents: p= 1.0}; T= absolute temperature, in K; 1 = solution viscosity, in kg/m-s; vy = solute molar volume at the normal boiling point, in m'/kmol Let us estimate the diffusivity of acetic acid(a) in water(B) using the Wilke-Chang equation at 293 K. Molar volume of the solute (acetic acid) at its normal boiling point, vy = 0.0641 m'/kmol (Poling et al., p. 54); viscosity of the solvent (water) at 293 K (Poling et al. p. 441), = 1.0121 x 10° kg/m-s; @ = 2.26 for water, My = 18. Substituting the values of different quantities in Eq. (2.66), Day = 1.116 10° cm*/s. This value would be a little larger at 298 K.Molecular Diffusion 48 Chaprer 2 ct perimental se EY sn eriain characteristics of ty may be nod i B, ae cern vert with the absolute temperature gia in 18 a He SHY OF the iii, Daw varies Tine ¢ for many sfute-solvent pairs. The Well-known ge . is roximately true for m kes tiyg it, which is app! ly tion given below also shows this type of dependence, equation g “ Din ence of ven in Eq. (2.66), it again apne, From the eee tea porn diffusivity of isopropanol in cine I ig dara eetanel A nore vitsons medium offers greater Hindcunce {0 the movement an ee rapa 1uid-phase diffusivity decreases as the Liscosity oF the solvent ingle at arclacs eect cp lacie eorelniona tad ere Suabiy i given gg ’ : et al. (2001) and in Perry's Handbook (7th edition; Chapter 5), 2.6.4 Concentration Dependence of the Liquid Diffusivity y sometimes be substantially differen a concentrated solution may eas The reaton (ane afconcen ated solution is more non-ideal mre have a more complex solute-solvent interaction. The change in the viscosity with Concentration has als some effect on te difsiviy. Sometimes the difusvity in a binary solution ange ; maximum or minimum with change inthe concentration Here we cite an importa’ equa proposed by Leer and Cullinan (1970) fr the determination of diffusivity ina conceit solution. z diny, Dap H= (Da Ma) (Dy He)! [ I+ al a where Dfy is the diffusivity of A in an infinitely dilute solution in B, and Dia is the same for p in an infinitely dilute solution in A. The activity coefficient of A (i.e. 74) is expressed by the following equation at ordinary pressure (where the vapour phase is rather ideal), i.e. YP = eee. (2.69) 4h ) Here is the vapour pressure ofA at the given temperature, The derivative (dl In Yale Inx4) can be obtained from the above equation. Details of the concentration dependence of diffusivity and So far we have analyzed Problems in which the cross-s Clional area of diffusion remains constant. This is true for a flat Scomeiry. However, there are many situations where the area of diffusion changes along the direction of diffusion, or changes with time even, Two common Scometries that involve diffusion through a variable area are the cylindrical and the sphericalDiffsion Through a Variable Area 49 Diffusion in ata geometric Apical eaves are analyred ered or conical geometry is also e red in some cases. A few ometry is also : ncountered in some 2.7.1 Diffusion in Spherical Geometry m of a drop of water in stagna Evaporation ¢ Al stagnant air will be anal an evaporating drop that has radius r, aeated eectaceeeg Cer conside at any insta ner radius rand thickness Ar around the drop as ahaa inh pas eee nce aeel igure 2.15. This is a binary system volving diffusion of water vapour(A) through air(B). Then Rate of input of A into the thin shell (at r = ry : (xr Ny, Rate of output of A from the thin shell (at r= r+ Ar)» camry), Concentric spherical ‘ shell Figure 2.15 Asketch indicating shell balance for mass transfer from a sphere, ‘The notation |, means that the quantity is evaluated at the position r. By a steady-state mass balance (note that the rate of accumulation = 0), (AMP Ng, — ATP Nylon = 0 Input Ouiput Accumulation Dividing both sides by Ar and taking the limit Ar — 0, sim, SP ah Ne 205 -Lemmy =0 = Anr?N, = constant = W (say) (2.70) ation (2.70) i n be generalized as very important result for steady stare diffusion through a variable area and (Area)(Flux) = Constant (2.71)