24.12.2014 Views

Miscellaneous notes on mass transfer coefficient models

Miscellaneous notes on mass transfer coefficient models

Miscellaneous notes on mass transfer coefficient models

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

P.C. Chau (UCSD, 1999)<br />

Chapter 6 Supplementary Notes<br />

We provide here the derivati<strong>on</strong>s needed to clarify Secti<strong>on</strong> 6.4 in Middleman. Example 6.4.1 makes use of<br />

<strong>mass</strong> <strong>transfer</strong> <strong>coefficient</strong>s in a channel flow. The c<strong>on</strong>centrati<strong>on</strong> is “understood” to be the cup-mixing<br />

average, but the story is l<strong>on</strong>ger than that. We illustrate here with flow through a tube. After reading these<br />

<str<strong>on</strong>g>notes</str<strong>on</strong>g>, you should repeat as an exercise the derivati<strong>on</strong>s with a channel flow.<br />

❏ Mass <strong>transfer</strong> <strong>coefficient</strong> model of flow through a tube<br />

If we have a fluid flow through a tube where <strong>mass</strong> <strong>transfer</strong> occurs at the wall, it is not<br />

uncomm<strong>on</strong> that we find people write the steady-state <strong>mass</strong> balance equati<strong>on</strong> of species A as<br />

dC A<br />

dz = 2k<br />

U R (C s –C A ) (1)<br />

where z is the axial distance down the tube, R is the radius, and U – is the average velocity. There is<br />

no chemical reacti<strong>on</strong>, and the <strong>mass</strong> transport is described by the average or overall <strong>mass</strong> <strong>transfer</strong><br />

<strong>coefficient</strong> – k . As written, the surface c<strong>on</strong>centrati<strong>on</strong> of the species A at the wall, Cs, is a c<strong>on</strong>stant<br />

and is presumed to be higher than that in the fluid, C A = C A (z). Thus Eq. (1) describes how the<br />

c<strong>on</strong>centrati<strong>on</strong> C A increases al<strong>on</strong>g the tube.<br />

The curious (and c<strong>on</strong>fusing) thing is that the same general form holds whether the flow is<br />

laminar or turbulent, and whether the tube is empty or filled with a packing material. The<br />

derivati<strong>on</strong> details are in the Appendix. We just point out here that Eq. (1) is not entirely correct,<br />

although we can still use it. The meaning of each quantity is very different under different<br />

circumstances, and we have to be careful with their interpretati<strong>on</strong>s. One more time: read the<br />

Appendix.<br />

With the two boundary c<strong>on</strong>diti<strong>on</strong>s C A (0) = C Ao , and C A (L) = C AL , we can integrate Eq. (1) to<br />

obtain<br />

ln C s –C AL<br />

=– 2kL C s –C AL<br />

or<br />

= exp – 2kL<br />

(2)<br />

C s –C Ao UR C s –C Ao UR<br />

The handling of the negative sign from the integrati<strong>on</strong> is a bit odd; we simply follow how most<br />

textbooks do the integrati<strong>on</strong>. What is even odder is how the RHS of Eq. (2) is rearranged to<br />

C s –C AL<br />

4Sh<br />

= exp –<br />

D<br />

C s –C Ao Sc Re D (D/L)<br />

(3)<br />

There is no trick here. It is just a matter of multiplying a fracti<strong>on</strong> top and bottom with the same<br />

quantities until we re-mold them into dimensi<strong>on</strong>less groups:<br />

2kL<br />

U R = k D D AB<br />

D AB<br />

ν<br />

ν 2L<br />

U D R =<br />

4Sh D<br />

Sc Re D (D/L)<br />

where we have introduced the tube diameter D = 2R, binary diffusivity D AB , and kinematic<br />

viscosity ν. Am<strong>on</strong>g the fracti<strong>on</strong>s in the middle, the first <strong>on</strong>e is the Sherwood number (based <strong>on</strong> k –<br />

and D), the sec<strong>on</strong>d and the third are the reciprocal of the Schmidt and Reynolds numbers. The<br />

fourth becomes the D/L ratio in the denominator. The factor of 4 comes from the substituti<strong>on</strong> of<br />

R = D/2.<br />

Because of the <strong>mass</strong> c<strong>on</strong>fusi<strong>on</strong> that we may create with three probable choices of length scales<br />

in R, D, and L, we use the subscript D in the Sherwood and Reynolds numbers to denote that they<br />

are defined <strong>on</strong> the basis of using the tube diameter D.<br />

Back to Eq. (3). One reas<strong>on</strong> why we choose to use this form in dimensi<strong>on</strong>less groups is that<br />

most <strong>mass</strong> <strong>transfer</strong> data are published in correlati<strong>on</strong>s using dimensi<strong>on</strong>less groups. It certainly is<br />

arguable if Eq. (3) is easier for us to grasp the physical picture, at least at the learning stage.<br />

1


P.C. Chau (UCSD, 1999)<br />

_____________<br />

Appendix<br />

Derivati<strong>on</strong> 1. Turbulent flow in a tube<br />

In a fully developed turbulent flow, the velocity profile is "flat," and at steady state, we have<br />

U – = U – (z), and C A = C A (z). A steady state differential <strong>mass</strong> balance would lead to the step<br />

0= (UC A ) z –(UC A ) z+dz πR 2 +k(C s –C A )2πRdz (A1)<br />

where the first two terms are the c<strong>on</strong>vective transport in and out of the differential volume, and the<br />

last term is the <strong>mass</strong> transport from the wall with some local <strong>mass</strong> <strong>transfer</strong> <strong>coefficient</strong> k. Recall<br />

that the laminar boundary layer "grows" with distance and hence k = k(z).<br />

We next do a Taylor expansi<strong>on</strong> <strong>on</strong> the z+dz term and if we further assume that U – = c<strong>on</strong>stant,<br />

the <strong>mass</strong> balance becomes<br />

U dC A<br />

dz = 2k<br />

R (C s –C A )<br />

(A2)<br />

We now integrate this equati<strong>on</strong> between the two boundaries C A (0) = C Ao , and C A (L) = C AL :<br />

C L dC A<br />

= 2<br />

(C s –C A ) UR<br />

C o<br />

0<br />

L<br />

k dz<br />

(A3)<br />

The result is<br />

ln C s –C AL<br />

C s –C Ao<br />

=– 2kL<br />

UR<br />

(A4)<br />

where we have introduced the definiti<strong>on</strong><br />

k = 1 L<br />

0<br />

L<br />

k dz<br />

(A5)<br />

as the average (or overall) <strong>mass</strong> <strong>transfer</strong> <strong>coefficient</strong>.<br />

With this result, we could have written Eq. (A2) with the overall <strong>mass</strong> <strong>transfer</strong> <strong>coefficient</strong> as<br />

in Eq. (1), and the integrati<strong>on</strong> would still lead to (A4) since k – is a c<strong>on</strong>stant. Because we could get<br />

away with this sloppiness in the final result, it is <strong>on</strong>e reas<strong>on</strong> why we see people write Eq. (1) even<br />

in research papers. In other words, Eq. (1) is really wr<strong>on</strong>g; (A2) is the proper form. This comment<br />

applies to derivati<strong>on</strong>s below and we will not repeat it.<br />

Derivati<strong>on</strong> 2. Laminar flow in a tube<br />

The derivati<strong>on</strong> for the case of laminar flow is not as straightforward. Now, the axial velocity<br />

and c<strong>on</strong>centrati<strong>on</strong> of species A are functi<strong>on</strong>s of radial positi<strong>on</strong>, u z = u z (z) and C A = C A (r, z). At<br />

steady state, an exercise of differential <strong>mass</strong> balance leads to<br />

0= u z (r)C A (r,z) z –u z (r)C A (r,z) z+dz πR 2 +n A,w (z) 2πRdz (A6)<br />

In c<strong>on</strong>trast to Eq. (A1), we have written the radial and axial dependence explicitly. We also have<br />

used the notati<strong>on</strong> n A,w (z) to denote the local wall flux; the introducti<strong>on</strong> of the <strong>mass</strong> <strong>transfer</strong> model<br />

will be delayed. As analogous to (A2), the next step after (A6) is<br />

d<br />

dz u z(r)C A (r,z) = 2 R n A,w(z)<br />

(A7)<br />

2


P.C. Chau (UCSD, 1999)<br />

However, we cannot integrate (A7) as we did with (A2). We need to do an area average by<br />

integrating (A7) across the tube cross-secti<strong>on</strong><br />

d<br />

dz<br />

0<br />

R<br />

u z (r)C A (r,z) r dr<br />

R<br />

= 2 n<br />

R A,w (z) r dr<br />

0<br />

=Rn A,w (z)<br />

(A8)<br />

We cannot go further with the LHS since C A (r,z) remains an unknown functi<strong>on</strong>. We work around<br />

that by defining the cup-mixing average<br />

C A =<br />

0<br />

R<br />

u z (r)C A (r,z) r dr<br />

0<br />

R<br />

u z (r) r dr<br />

(A9)<br />

With the average velocity evaluated as<br />

U = 2π<br />

πR 2<br />

0<br />

R<br />

u z (r) r dr<br />

(A10)<br />

we can rewrite (A8), or rather (A7), in terms of the cup-mixing average as<br />

or<br />

d<br />

dz U C R 2<br />

A<br />

2<br />

U dC A<br />

dz = 2 R n A,w(z)<br />

=Rn A,w (z)<br />

(A11)<br />

We now finally can introduce the definiti<strong>on</strong> n A,w = k(C – s – C – A) in terms of the local <strong>mass</strong> <strong>transfer</strong><br />

<strong>coefficient</strong> k = k(z). Eq. (A11) becomes<br />

dC A<br />

dz = 2k<br />

UR (C s –C A )<br />

(A12)<br />

and its integrati<strong>on</strong>, following (A3) and (a5), should <strong>on</strong>ce again arrive at the form<br />

ln C s –C AL<br />

C s –C Ao<br />

=– 2kL<br />

UR<br />

(A13)<br />

These last two steps are equivalent in form to Eqs. (1) and (2). Texts or research papers<br />

generally do not write the equati<strong>on</strong>s with the overhead bar <strong>on</strong> the c<strong>on</strong>centrati<strong>on</strong>s, but we should<br />

know better that they represent the cup-mixing average.<br />

Furthermore, reflect <strong>on</strong> the derivati<strong>on</strong> steps, and we should observe that we can <strong>on</strong>ly arrive at<br />

(A12) if we use the steady state balance and if we delay introducing the <strong>mass</strong> <strong>transfer</strong> <strong>coefficient</strong><br />

model. You'll find that the c<strong>on</strong>cept of cup-mixing average is also used in heat <strong>transfer</strong>.<br />

3


P.C. Chau (UCSD, 1999)<br />

❏ Height of a theoretical unit and the log-mean c<strong>on</strong>centrati<strong>on</strong> difference<br />

There are two more definiti<strong>on</strong>s that we can derive based <strong>on</strong> the result in Eq. (2). The first is to<br />

write the argument of the exp<strong>on</strong>ential functi<strong>on</strong> as<br />

2kL<br />

UR = L<br />

(4)<br />

H TU<br />

where H TU is the height of a theoretical <strong>transfer</strong> unit defined as<br />

H TU = UR<br />

2k<br />

(5)<br />

The actual definiti<strong>on</strong> depends <strong>on</strong> different problems. The general idea is to interpret the argument in<br />

Eq. (4) as a ratio of two length scales: the tube length L and the "<strong>transfer</strong> length" H TU . Loosely<br />

speaking, the <strong>mass</strong> transport process is more effective if the H TU is small.<br />

The sec<strong>on</strong>d definiti<strong>on</strong> comes from the need to calculate the overall rate of <strong>mass</strong> <strong>transfer</strong> in the<br />

tube as in<br />

R =U(C A,L –C A,o ) πR 2 (6)<br />

We may ask how the overall <strong>mass</strong> <strong>transfer</strong> <strong>coefficient</strong> may be used or may be related to this<br />

calculati<strong>on</strong>. One easy rearrangement is to substitute for the average velocity U – in (6) using Eq. (2).<br />

We also need to manipulate the simple c<strong>on</strong>centrati<strong>on</strong> difference in (6) to become a messier looking<br />

form involving the surface c<strong>on</strong>centrati<strong>on</strong>, and the result is<br />

R = πR 2 2kL<br />

R<br />

(C s –C AL )–(C s –C Ao )<br />

ln C (7)<br />

s –C AL<br />

C s –C Ao<br />

If we define the big ugly c<strong>on</strong>centrati<strong>on</strong> mess as ∆C LM , the "logarithmic mean," or simply the logmean<br />

c<strong>on</strong>centrati<strong>on</strong> difference,<br />

∆C LM = (C s –C AL )–(C s –C Ao )<br />

ln C (8)<br />

s –C AL<br />

C s –C Ao<br />

Eq. (7) becomes the deceptively simple looking<br />

R =(2πRL)k∆C LM (9)<br />

We can certainly argue if we need Eq. (9) if we can use (6). We find better use of log-mean<br />

difference in heat <strong>transfer</strong> design calculati<strong>on</strong>s.<br />

❏ Mass <strong>transfer</strong> correlati<strong>on</strong>s in laminar flow<br />

One big c<strong>on</strong>fusi<strong>on</strong> is the use of <strong>mass</strong> <strong>transfer</strong> <strong>coefficient</strong>s in laminar flow problems. We can<br />

solve a problem using <strong>on</strong>ly Chapter 5, or we can rearrange everything as in Secti<strong>on</strong> 6.4. What it<br />

comes down to is style and pers<strong>on</strong>al preference. Either approach should lead us to the correct<br />

answer. What follows are the derivati<strong>on</strong>s to the two important correlati<strong>on</strong>s.<br />

4


P.C. Chau (UCSD, 1999)<br />

• Equati<strong>on</strong> (6.4.6)<br />

This equati<strong>on</strong> is derived from the short c<strong>on</strong>tact time approximati<strong>on</strong> at the gas/liquid interface<br />

in a laminar liquid film (Secti<strong>on</strong> 5.1). We begin with Eq. (5.1.25), the local flux, 1<br />

N y (x) = – (C o –C a )<br />

D AB v m<br />

πx<br />

(10)<br />

We now state that Eq. (10) is to be written in terms of a local <strong>mass</strong> <strong>transfer</strong> <strong>coefficient</strong> k = k(x):<br />

It is clear that we need to define<br />

N y (x) = k (C a –C o ) (11)<br />

k=<br />

D AB v m<br />

πx<br />

(12)<br />

The next step is to evaluate the average <strong>mass</strong> <strong>transfer</strong> flux over some length L. In other words, we<br />

want to find the average <strong>mass</strong> <strong>transfer</strong> <strong>coefficient</strong><br />

k = 1 L<br />

0<br />

L<br />

k dx<br />

(13)<br />

and the result is 2 k = D ABv m<br />

π<br />

1/2 1<br />

L<br />

0<br />

L<br />

x –1/2 dx<br />

=2 D ABv m<br />

πL<br />

1/2<br />

(14)<br />

We are ready to define the Sherwood number based <strong>on</strong> this average <strong>mass</strong> <strong>transfer</strong> <strong>coefficient</strong> and the<br />

liquid film thickness δ:<br />

Sh δ = kδ<br />

D AB<br />

(15)<br />

The final result involves substituting for the <strong>mass</strong> <strong>transfer</strong> <strong>coefficient</strong> with Eq. (14) and the typical<br />

multiplying a fracti<strong>on</strong> top and bottom with the same quantities until we re-mold them into<br />

dimensi<strong>on</strong>less groups:<br />

Sh δ =2 D AB 3<br />

πL 2 U 1/2 δ<br />

= π 4 3 D AB<br />

2 L<br />

U δ2<br />

2<br />

D AB<br />

= π 6 Uδ ν δ<br />

ν D AB L<br />

= 6 π Re δ Sc δ L<br />

D AB<br />

1/2<br />

1/2<br />

1/2<br />

(16)<br />

which is Eq. (6.4.6). Note that in the derivati<strong>on</strong>, we have to use, for film flow, v m = 3/2 U – . The<br />

product of the Reynolds number and the Schmidt number is the Peclet number:<br />

1 Recall that C o is the c<strong>on</strong>centrati<strong>on</strong> of species A “deep” in the liquid and C a is the liquid<br />

c<strong>on</strong>centrati<strong>on</strong> at the gas/liquid interface. The sign of the flux is dictated by the coordinate system.<br />

As derived in Secti<strong>on</strong> 5.1, the y-directi<strong>on</strong> points into the liquid film. So a negative flux means the<br />

transport of the species A is out of the liquid film into the gas phase. And vice versa.<br />

2 Middleman defines first the c<strong>on</strong>tact (or exposure) times t = x/v m and T = L/v m . The average<br />

<strong>mass</strong> <strong>transfer</strong> <strong>coefficient</strong> is evaluated as a time average over the “age” T, and we can arrive at the<br />

identical result in Eq. (14). This is actually a very comm<strong>on</strong> approach.<br />

5


P.C. Chau (UCSD, 1999)<br />

Pe = Re δ Sc = Uδ<br />

D AB<br />

(17)<br />

• Equati<strong>on</strong> (6.4.9)<br />

This equati<strong>on</strong> is derived from the short c<strong>on</strong>tact time approximati<strong>on</strong> at the solid/liquid interface<br />

in a laminar liquid film (Secti<strong>on</strong> 5.2). We begin by stating that the local flux in Eq. (5.2.13) is to<br />

be put in terms of the local <strong>mass</strong> <strong>transfer</strong> <strong>coefficient</strong>:<br />

N y (x) = D bC s<br />

Γ(4/3)<br />

α<br />

9D b x<br />

1/3 =kC s (18)<br />

where now the coordinate y is based <strong>on</strong> the solid surface. The local <strong>mass</strong> <strong>transfer</strong> <strong>coefficient</strong> has to<br />

be defined as<br />

k= D b α 1/3 x<br />

Γ(4/3) 9D –1/3 (19)<br />

b<br />

and the average <strong>mass</strong> <strong>transfer</strong> <strong>coefficient</strong> can be evaluated as<br />

k = 1 L<br />

0<br />

L<br />

k dx<br />

= 3 2<br />

D b<br />

Γ(4/3)<br />

α<br />

9D b 1/3 L –1/3 (20)<br />

We now need to get a handle <strong>on</strong> α. For film flow, the velocity profile is usually derived with<br />

the coordinate based <strong>on</strong> the gas/liquid interface, z = δ – y,<br />

u x (z) = 3 2 U 1– z δ<br />

2<br />

(21)<br />

Thus the velocity gradient at the solid surface is<br />

α = du x<br />

=– du x<br />

dy y=0 dz<br />

z=δ<br />

= 3U δ<br />

(22)<br />

We finally are ready to find the Sherwood number as defined in Eq. (15):<br />

Sh δ =<br />

δ 3<br />

D AB 2<br />

= 1.165 Uδ<br />

ν<br />

D b<br />

Γ(4/3)<br />

ν<br />

D b<br />

δ<br />

L<br />

1/3<br />

3U 1<br />

δ 9D b L<br />

(23)<br />

1/3<br />

which is Eq. (6.4.9) with the substituti<strong>on</strong> of the definiti<strong>on</strong>s of dimensi<strong>on</strong>less groups.<br />

• Equati<strong>on</strong> (6.4.12)<br />

This equati<strong>on</strong> is derived from the l<strong>on</strong>g c<strong>on</strong>tact time approximati<strong>on</strong> at the solid/liquid interface<br />

in a laminar liquid film (Secti<strong>on</strong> 5.3). The basis is the result in Eq. (5.3.32). If we take the P*<br />

terms to be large, the P* will cancel out and leave us with the factor 1.6.<br />

6

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!