Diffusion Chapter 9 (old book) General process of flow Heat as the example Fick’s First law Use of diffusion Coefficients personal passive samplers, SO2 accommodation coef. Particles Probability distribution using random movement Diffusion distances Fick’s Second Law Diffusion in a GC column Diffusion in a sphere Estimating Diffusion Coefficients gases water; liquids Turbulent Diffusion Lake system Atmospheric System 1 Heat Templow2 Temphigh1 x The heat that flows thru a slab of material is proportional to the cross sectional area, A, of the slab and the time, t, for a given Temp Heat flow is also to Temp/x for a given A and time if Temp/x is small Δq ΔT A Δt Δx if we think about really small thickness of x dq dT kA dt dx dq/dt = the rate of heat transfer with respect to time dT/dx = the temperature gradient k = thermal conductivity 2 dq dT kA dt dx k has the units of Kcal / sec meter 2 oC / meter material k Al Steel Pb air glass 4.9x10-2 1.1x10-2 8.3x10-3 5.7x10-6 2.0x10-4 Temphigh1 Templow2 L at steady state for a const. temp gradient across the rod Δq T T k 1 2 tA L 3 Diffusion 1b. cool 1a. hot 2b. low voltage 2a. high voltage 3b. low mass 3a. high mass 4b. low pressure 4a. high hydrostatic pressure flux = flow area-1 time-1 a gradient drives the flow 4 page 184 table 9.1 Page 184 Table 9.1 5 Let us consider a gas diffusing in into a zone where it is constantly collected or removed O3 O3 O3 O3 O3 O3 [O3] x FluxO D 3 inlet x O3 (moles O3 ) = ; (time ) area x (moles O 3 ) [O 3 ] D ( time)area x If we measure the # of moles of O3 collected over a period of time; know the diffusion coef. for O3 in air, [O3] can be calculated 6 Diffusion and sticking coefficients the average speed of gas molecules is given by c 8RT Mw The rate of collisions per unit time with a wall of surface A in a given volume is rate = 1/4 c x area x concgas rate RT A Cgas 2 Mw rate/A = # molecules time-1 area-1 = flux if we think about the # of effective collisions, i.e. the ones that actually stick to the wall, a factor is introduced called “sticking” factor surface recombination accommodation coefficient removal rate RT A Cgas 2 Mw 7 Judeikis et al. were interested in the effectiveness of coal surfaces in the uptake of SO2 gas. rate/A = # molecules time-1 area-1 = flux flux = radial velocity Cgas D C/r r SO2 measure SO2 coal soot coating on surface -D d [SO2]/d r = rad. vel x [SO2] ln {[SO2]/[SO2,o]} = rad. vel x r /D ln {[SO2]/SO2,o]} = krate t 8 Removal of SO2 along a tube reactor coated with fly ash. The accommodation coef. = 4.4x10-4. The total pressure was 55 torr, with O2 = 6 torr and SO2 = 9m torr; % RH= 0 9 The Stokes-Einstein Equation (particles) Let us now think of diffusion in terms of chemical potential We can think of the driving force of diffusion as the negative gradient of the chemical potential, i.e. d/dx = free energy/mol /dx The frictional force resisting the flow, due to an imbalance in chemical potential, is the frictional coef. f (force/velocity) on each molecule x the velocity, v, of the flow, for a mole this force is f vNo f v No= -d/dx Flux has the units of moles, molecules or mass per area per time Flux = moles/(cm2 time) Conc x velocity = moles/cm3- x cm/time = moles/(cm2 time) We can define a diffusion velocity caused by a driving force or chemical potential, or the concentration gradient such that: C v = Flux recall 10 RT d dp P substituting for P from PV=nRT, and n/V = C P= CRT RT d dC C f Nov = -d/dx RT v dC / dx f N0 C C v = Flux = RT dC / dx f N0 11 RT Flux dC / dx f N0 diffusion coef D= RT/(f No) Stokes (including Cunningham’s slip factor) showed that for unit spheres and nonturbulent viscous flow that the resisting force on a particle flowing through a fluid is f = 6r/Cc where = is the viscosity of the medium (poise) air(20oC) = 1.83x10-4 g/(cm sec) r is the radius of the particles and Cc = 1+/r(A+ Qe-rb/) where is the mean free path of air = 0.067 m particle Cc size (m) 0.01 22.2 0.05 4.97 0.1 2.87 0.25 1.69 0.5 1.33 1 1.16 5 1.03 12 The Randomness of Diffusion Consider 17 boxes arranged in a row PAGE 185 FIGURE 9.2 13 page 186 figure 9.3 fit to random walk distribution 2 pn (m) n 1/ 2 m/ 2 2 exp n / 2 Normal Gaussian Distribution 2 x p( x) 1/ 2 exp 2 2 2 1 14 Where we would like to go with this is relate the sigma, , which is a basic feature of the normal distribution to the diffusion coefficient D x 2 1 p( x) 1/ 2 exp 2 2 2 2 pn (m) n 1/ 2 m/ 2 2 exp n / 2 22= n/2 n / 2 if we multiply this by x, the distance across a box, can be related to an actual distance We will then calculate the flux across from one box to another The concentration gradient which caused the flux Substitute into Flux = -D dC/dt 15 BOX # 3 4 x 5 6 40 0 8 60 0 24 0 0 42 0 14 0 28 0 5t 4 6t 7t 7 8t If we look at 5th box in seventh time step (7t), 12 particles are entering box #5 from the left and 2 from the right at the eighth (8t) step, 7 particles leave box #5 and go back to box #4 this gives a net flux of 5 particles between the 7th and 8th steps or Flux = 5/2t We now define an average change in concentration per length between adjacent boxes because for every step one box is emptied and the adjacent one filled C = N/x the spatial conc. gradient dC/dx is dC/dx = C/x = N/x2 16 If we look at time step 6, the gradient dC/dx driving the diffusion between boxes for time step 7 is dC/dx = -N/x2 = -(24-4)/x2 recalling that F = -D dC/dx and F= 5/(2t) 5 /(2t ) ( x ) 2 D 2 20 /(x ) 8t t= nt and x = n1/2/2 x x = (2Dt)1/2 for three dimensional movement s = (4/Dt)1/2 17 Example How long does it take for a gas molecule of biphenyl and 0.25 m particle to diffuse from the center of a 5 cm sphere to the a walls of the sphere? Assume a diffusion coef of 0.06 cm2/sec for biphenyl and 1.6x10-6 cm2/sec for the particle. 2.5 cm for biphenyl s = (4/Dt)1/2 S2 3.14x 2.5 cm2 t 4D 4x0.06 cm2sec 1 t = 82 seconds for the aerosol, D= 1.62x10-6 cm2/sec (d= 0.25 m) t = 35.5 days in water biphenyl diffusion is much slower than in air D10-6 so t = tens of days 18 Fick’s second law Fick’s second law attempts to express the change in concentration with respect to time with the change in Flux dC/dt = f (flux) Consider an elemental volume (box) with a flux of material in and out FX FO x A mass balance on the elemental volume per unit time (both in and out) mass= conc x vol); flux = mass/(area time) flux x area= mass/time 19 mass= conc x vol); flux = mass/(area time) mass/time = (conc x vol)/time d(mass) d(conc vol) d(time) d (time) = -area flux V dC/dt = -area flux V = area x division by V dC/dt = -flux/x; as x --> zero 1 dflux flux lim x dx x o it appears that dC/dt = - dflux/dx Fx C t x cons tan t x t const 20 Fx C t x cons tan t x t const C C 2C ( D ) D 2 t x x x If we think of three dimensions Fx Fy Fz C t x y z C C C C (D x ) (D y ) (D z ) t x x y y z z 2C 2C 2C C D t x 2 y 2 z 2 21 1. diffusion in the x direction A solution in the x direction for a long tube, where diffusion in the y and z direction is insignificant, is x 2 M C( x,t) exp 1/ 2 2 2 ( Dt ) 4 Dt where = (2Dt)1/2 Figure 9.5 Page 193 22 What kind of diffusion can we expect for a compound traveling down a 30 m fused silica column (0.25mm id); assume a flow of 1 cm3/min? id vol of the col = (0.025cm/2)2x 30m x100cm/m; vol =1.47cm3 ; Flow = ?? The flow time = 1.47cm3/ 1cm3 per min = 1.47 min A typical diffusion coef= 0.07 cm2/sec, (2Dt)1/2= so our peak would broaden by 4 x = 4(2x0.07x1.47*60)1/2 = 14.06 cm; why 4 the carrier travels 30 m or 30x100 cm in 1.47 min this equals 3,000 cm/1.47 min = 34 cm/sec our peak would broaden in this time 14.06 cm /34cm/sec = ~0.4 sec 23 2. Diffusion in the Radial Direction PAH 2C 2C 2C C D ) 2 2 2 t y x z converting to polar (radial) coordinates x= r sincos, y = r sinsin , z = r cos 2C C C 2C 1 C 1 2C D 2 2 cot 2 2 2 2 t rr r sin r r r if diffusion is only in the radial direction 2C C C D 2r t r r 2 2 (C r ) (C r ) D 2 t r 24 PAH dr These types of systems can be solved with numerical techniques to calculate the C at successive depths of dr into the particle over time U= Cr U(p) t t U(p) t D t U(p 1)2U(p)U(p 1) 2 r Figure 3. Modeling the uptake of deuterated pyrene by diesel exhaust aerosol at 298K. 25 Diffusion between two parallel plates Cout W L Cin H Let’s say that we wanted to strip a gas and not particles Solutions to the partial differential equations take on the form C out An exp n C in n 1 C/Co= 1 - 1.52652/3 +1.5 +0.0342 4/3 where = 8 x D x L x W/(H x flow) 26 Large Stripper System Transmission Properties 100 % Transmission 80 particles D=.00005 60 40 organic gas D=.05 20 0 0 10 20 30 40 Flow Rate (cfm) Organic Gases (D=0.05cm2/s) 50 60 Particles (D=0.0005cm2/s) 27 Estimating Diffusion Coefficients Factors that influence diffusion average distance traveled between collisions, i.e. mean free path, more collisions for a given distance translates into a lower mean free path organic air org,air where 1 (1 z)1/ 2 Nair 2org,air N= # air molecules/vol = collision diameter of air and organic z= molecular wt ratio of organic/air We would predict that diffusion coefficients would decrease with increasing (molecular weight)1/2 and effective collision diameters squared. Diameter2 radius2 cross sect. area of molecule V r3or r V1/3; so r2 or area V2/3 If we assume the volume of a molecule is molar vol V 28 and molar volume = molecular weight in cm/mol liquiddensity Figure 9.6 top page 195 See Figure 9.6 page 195 29 Molecular weight vs Diffusion Coefficients Page 195 Fig 9.6 bottom 30 Diffusivities in water page 196 Figure 9.7 31 Estimating Gas Phase Diffusion Coefficients (page 197) equation on page 197 with definitions 32 Estimating molar volumes 1. molar volume, V = molecular weight in cm3 /mol liquiddensity for benzene Mw= 78, density =0.88gcm-3=89 cm3/mole 2. sum of atom size---> diffusion for benzene V= 6(C) +6(H) +ring V= 6x16.5+6x2.0-20.2 = 90.8 cm3mol-1 33 35 Calculating liquid Diffusion Coefficients Dw 13.26x10 5 2 1 1.14 ( cm sec ) 0.589 xV V = molar volume = solution viscosity in centipoise (10-2g cm1 sec-1) at the temperature of interest. The units of poise refer to the property of fluids that requires a shearing force of one dyne (g cm/sec2) of two parallel layers of one cm2 at velocity of 1 cm/sec over a gradient of one cm log = A / T+B For other liquids Wilke-Chang (1955) give Dliq 7.4x10 8 ( Mw )1/ 2 T 2 1 ( cm sec ) 0.6 xV where is the solvent association term, Mw is the molecular weight of the solvent, is the solvent viscosity, T is the temp. water CH3-OH ethanol Heptane Benzene 2.6 1.9 1.5 1 1 36 page 409 In Barrow Physical Chemistry 1962 Table 12.5 37 Reynolds Numbers A body moving through a fluid creates 1. Inertial forces, i.e. forces due to the acceleration or deceleration of small fluid masses near the body 2. Viscous frictional forces due to the viscosity of the medium Inertial forces/viscous forces = Re# Re# = m x vel x d/ /m = = (kinomatic for air) = Re# = vel x d/ = 0.00121 g/cm-1sec-1 /1.82x10-4 g/cm3= 0.151cm2sec-1 for flow in a pipe laminar region intermediate region turbulent flow 1-2000 2100 - 4000 >4000 A 1 m aerosol is flowing in a 16” duct with a velocity of 3500 ft/min. What is the Re# in the duct? Re# = duct diameter x rel. velocity of air to duct/v Re# = vel x d/ = 3500x12x2.54/60 x 15x2.54/.151= 479000 38 Turbulent Diffusion for molecular diffusion x = (2Dt)1/2 L2 td 2D If we look a molecular diffusion times (page 201 Table 9.5) 39 For transport by advection Flux = C v Fad = Cv (m L-2 t-1) C C C C divF v v ) v x adv y z t adv x y z (M L-3 T-1) For advection the time scale is simply L2 td (diffusion) 2D L t a (advection) v L L2 ; v 2D L crit 2D v For distances larger than L advection is more important than diffusion. In your homework, you will calculate the critical distances for typical air and water advection and then explain what it means. An expression for Turbulent Flux It is possible to develop a turbulent diffusion coef. analogous to a molecular diffusion coef. Let’s assume that turbulent fluctuations cause a change in concentration along with an associated flow Qex of volume between C1 and C2 over some distance Lx. 40 Lx C2 C1 x If Lx is small, the concentration difference C1 -C2 can be described by the product of Lx and the slope of the curve at the point it intercepts the plane between C1 and C2: C1-C2 = - Lx C/x Since this moves across some area, a, at a flow of Qex the velocity vx = Qex/ a Since Flux = C vx= Qex/ a ( - Lx C/x) Fturb,x L xQ x C a x Fturb,x Ex C C vx x and Ex Cvx C/ x This is property of the fluid motion and not the substance described by C 41 Effect of eddies on dispersion Consider a patch of ink on a water surface that is turbulent. It is characterized by the size of its concentration variance 2 about its center of mass. The patch after time t will grow and be displaced = 2Et and 2/t = 2E Figure 9.9 page 206 42 The growth of 2 becomes faster with increasing 2. Figure 9.10 page 207 43 Table 9.6 page 206 44 How do we determine turbulent diffusion coefs in vertical mixing? 1) from the change in heat content Tz dz ; from Fickian flux t z Cp = heat capacity c p Az Ez ( T / z ) zB c p Az substituting and solving for Ez Tz dz Az t z Ez c p Az ( T / z ) Az ( T / z ) zB 45 2) Radioactive tracers Radon-222 gas is released from Radium-226 in sediments the radon diffuses into the water, where it is transported upward by turbulence Broecker sets up a Fick’s second law flux expression for excess radon, i.e. radon beyond the “conc.” of that which should be in equilibrium with existing Ra (hence your book uses the word activity) Rn/t Ez 2Rn h 2 Rn if there is little change in Rn/t after some time Rn/t Ez 2Rn h 2 Rn = zero a solution to this at Rn/t = zero Rn (h) = Rn(h=0) exp[-(/Ez )1/2 h] so if we plot ln Rn(h) vs h we get a slope of ........ 46 Figure 9.13 page 212 47 Parameters used to characterize vertical mixing in Lakes In a lake water is usually stratified with the most dense water at the bottom. A stability time N, is introduced which describes how how long it will take to “restore” a water parcel which is moved vertically to a different density regime. g d N dz 1/ 2 g= gravity acceleration = density 48 the coef. of thermal expansion is related to the change in density with temperature by d/dT = - This gives N2 = -g dT/dz Finally N2 is related to the vertical eddy diffusion coef. Ez = a (N2)-q 49 Turbulent Diffusion in the Atmosphere 50 Fisk’s 2nd law Solutions in 1,2,3 dimensions If we just think in two dimensions; ie there is no diffusion in the x direction compared to y and z, and substitute an emissions strength for X (emissions in g/time normalized fro velocity) 51 Historically Ky and Kz have been related to a Gausian such that substituting 52