05diffsep07 PDF

Chapter 5
Diffusion
5-1. Diffusion: thermal motion in viscous media
5-1.1. Brownian movement and diffusion
5-1.2. Diffusion coefficients
5-1.3. Influences of molecular shape on diffusion coefficients
5-2. Fick’s “Laws”
5-2.1. Fick’s First Law of motion
5-2.2. Fick’s Second Law of diffusion
5-3. Measuring diffusion coefficients
5-3.1. Diffusion across a thick membrane
5-3.2. Diffusion across an interface between unstirred regions
5-4. Applications to biological situations
5-4.1. Flux by diffusion across a uniform membrane depends on solubility
5-4.2. Diffusion through pores
5-4.3. Diffusion across walls and pores in parallel
5-4.4. Diffusion from a stirred infinitely large source into a non-consuming stagnant region
5-4.5. Diffusion into a region with solute consumption
5-5. Diffusion in Heterogeneous Media
5-5.1. Diffusion through ISF and cells in parallel and in series
5-5.2. Diffusion across an uneven slab
5-5.3. Diffusion through a hindering matrix
5-6. Diffusion of solutes which can be bound to absorbing sites
5-6.1. General aspects of diffusion of solute in the presence of binding sites
5-6.2. Diffusion in the presence of immobile binding sites (DB = DSB = 0)
5-6.3. Diffusional Facilitation due to binding to a mobile site
5-6.4. Effect of slow binding kinetics on diffusional facilitation
5-6.5. Particle transport in a nerve axon: motor transport proteins
5-7. Chapter summary
5-8. Problems
5-9. Further readings
5-10. References
5-1. Diffusion: thermal motion in viscous media

Diffusion is movement of solute and water molecules by random thermal, Brownian motion;
the motion results from the impact of one molecule hitting another, imparting momentum. The
random motions of a spherical molecule in a uniform medium have the same statistical properties
in all directions; diffusion is normally isotropic. When the medium is structured, by a set of
aligned macromolecules with parallel interstices between them, for example, a given energy
produces higher-velocity motion parallel to the fibers than in the perpendicular direction; the
diffusion is anisotropic. Diffusion is due to thermal motion: the molecular motion is proportional
54 Diffusion
to the temperature in degrees Kelvin: a 10° rise in temperature from 300° K to 310° K would
cause only a 3% rise in diffusivity if other things remained constant. Diffusion is opposed by the
viscosity of the medium: the more important effect of raising temperature is reducing viscosity
and lowering the resistance to particle motion. Raising the temperature of water from 20 C to
30 C lowers its viscosity by almost 20%. See Table 5-1. The viscosity of water, 1 cP or 0.01
Table 5-1: Viscosity of water and air at 1 atmos pressure: υ = η / ρ (from Bird et al., 1960)
WATERa AIRb
Viscosity: Kinematic Viscosity Viscosity: Kinematic Viscosity

Temperature
η, cP υ . 102 cm2s-1 η, cP υ . 102
0 1.787 1.787 0.01716 13.27
20 1.0019 1.0037 0.01813 15.05
40 0.6530 0.6581 0.01908 16.92
60 0.4665 0.4744 0.01999 18.86
80 0.3548 0.3651 0.02087 20.88
100 0.2821 0.2944 0.02173 22.98
a. Calculated from the results of R.C. Hardy and R. L. Cottington, J. Research Nat. Bur.
Standards, 42, 573-578 (1949), and J. F. Swindells, J. R. Coe, Jr. and T. B. Godfrey, J.
Research Nat Bur. Standards, 48, 1-31, (1952)
b. Calculated from “Tables of Thermal Properties of Gases,” Nat. Bur. Standards Circ. 464
(1955), Chapter 2.
g cm−1 s−1, is about 100 times that of air. The kinematic viscosity, υ = η/ρ ⋅ cm2s−1, viscosity per
unit density, is about 10 times higher for air than for water, simply because the density of air is so
low, about 1.205 g/liter at 20° C. Kinematic viscosity has the same units as a diffusion coefficient.
Note that the viscosity of fluids decreases with increasing temperature, but that gasses do the
reverse.
Net diffusional fluxes occur by the net movement of particles from a locale of high
concentration to one of lower concentration by this random movement of particles. Diffusion
dissipates gradients, so concentrations tend toward equilibrium. (Entropy is raised by the
dissipation of concentration gradients.)
Diffusion is thermal motion, usually random, and hindered by collisions with other
molecules. It is slower in high-viscosity fluid. Water density, ρ, is 1 g/ml, and its viscosity is 1 cP.
Diffusion in gases is much more rapid than in fluids since the density of molecules is much less:
molecular spread of an aromatic substance can be smelled across a room in seconds, even if given
little convective assistance. The density of dry air at 273° K is 1.29 kg/m3 or 1.29 × 10−3 g/cm3,
one one-thousandth of that of water. The viscosity of dry air is 1.72 × 10−4 g cm−1 s−1 or
0.0172 cPoise (CRC Handbook of Tables for Applied Engineering Science, Second Edition,
1973), about one-sixtieth of that of water.
A common laboratory demonstration of diffusion involves layering solvent water carefully
over a solution of deep blue copper sulfate, which has a higher density than water so the layers do
07 September 2007, 5:28 pm /userA/jbb/writing/903/2005/05diff7sep07.fm

Diffusion 55
not mix (Fig. 5-1). Initially the layering gives a sharp boundary, across which the solute
concentration changes abruptly from zero above to a value Co below the interface. Immediately,
random Brownian motion of the molecules begins and gradually blurs the boundary so that the
gradation in color spreads. Only after a very long time in a graduated cylinder sitting on the
classroom shelf for months does this lead to complete mixing and a uniform blue color.
Thermodynamics says that mixing will eventually occur: the initial state is certainly not one of
equilibrium; free energy will be lower, and entropy higher, when mixing is complete. Neglecting
small contributions from the heat of mixing, the free energy change results entirely from the
higher entropy of the mixed state.
Figure 5-1: Diffusion of solute and solvent. Open circles are solvent; filled circles are
solute. At the start, solute is in the bottom half of the fluid, but eventually the
concentration becomes uniform throughout, and no concentration gradients remain.
Solvent diffuses into the solution, net from above to below; net solute diffusion moves
in the opposite direction.
5-1.1. Brownian movement and diffusion

Measurement of the change of position of a hard spherical particle in any one dimension will
allow calculation of the diffusion coefficient and of the particle radius. The dominating condition
is that the particle concentration be very low, which means that the effective viscosity is the
viscosity of the pure solvent. The mean value of the square of the distance (∆x)2 travelled along
the x-axis in a chosen time interval, ∆t, is obtained and the effective diffusion coefficient
D = (∆x)2/2∆t. Substituting into this the Stokes-Einstein relation, D = RT/ 6πaηℵ, gives an
experimental approach to estimating the effective molecular radius, a, when the viscosity of the
medium, η, is measured separately:
2 2 RT ∆t
( ∆x ) ⁄ ∆t = 2D ; or ( ∆x ) = -------- ⋅ ------------- . (5-1)
ℵ 3πηa
The frictional coefficient is fs = 6πaη, for the molecule with the medium, assuming sphericity.
Saxby (1923??) used this method, as in Fig. 5-2, with Staph. aureus (diameter = 1.12 µ), to
find ℵ, Avogadro’s number, obtaining a value of 6.08 × 1023 (correct value = 6.023 × 1023).
5-1.2. Diffusion coefficients

The diffusion coefficient, D, is not a physical constant but an observed phenomenological
coefficient. Its dependence on particular features is indicated by the Einstein-Stokes expression,
/userA/jbb/writing/903/2005/05diff7sep07.fm 07 September 2007, 5:28 pm

56 Diffusion
Figure 5-2: Calculation of D from Brownian motion. Saxby observed the x-positions of
a spherical bacterium every 30 seconds, and calculated D and ℵ, Avogadro’s number,
using Eq. 5-1.
RT RT
D = ------------------- or --------- .
6πaηℵ fℵ
In this expression the energy per molecule is RT/ℵ, and the friction per molecule is 6πaη, or f,
and is equivalent to a force divided by a velocity, (g cm s-2)/ (cm s-1).
Because the denominator contains the molecular radius a, the equation suggests that D might vary
inversely with the reciprocal of (molecular weight)1/3 but over a large range a square root
relationship is closer for small molecules with MW < 4000 Daltons, but a cubic relationship is
better for larger molecules. Some diffusion coefficients for important biological molecules are
listed in Table 5-2.
Correction to T = 20° C (293.2° K) was made by correcting for T and η from the value
observed at T° K:
293.2 η T
D 20 = ------------- -------- D T . (5-2)
T η 20
This does not correct for changes in molecular size such as occur with unfolding of the molecule
or change in the degree of hydration. The viscosity η is the most important variable influencing D.
Proteins vary in their density and density’s reciprocal measure, specific fractional volume,
ml/g protein over only a small range, most being around 0.72 ml/g or a density of 1.39 g/ml.
Eq. 5-1 can be used to estimate the diffusion coefficients for substances when the data are not
available, using inference from observed substances. For example the diffusion coefficient for
CO2 in air at 20° C is 2.45 cm2s−1. To estimate the diffusion coefficient in air for methyl salicylate
(Wintergreen, used in liniment), one could calculate the Stokes radii for both CO2 and
methylsalicylate. Alternatively, simply take the square root of the ratio of molecular weights, as
an approximation to the ratio of Stokes radii, a(MeS)/a(CO2), and substitute into the
Stokes-Einstein relationship directly:
MW C O 44 2 –1
D MeS = D C O2 ⋅ -------------------2 = 2.45 ⋅ --------- = 1.32cm s , (5-3)
MW MeS 152

Diffusion 57
Table 5-2: Diffusion Coefficients in Water, Dw (in aqueous solution at 20°C)
Molecular
Molecular Weight Dw × 106 Specific Volume
Substance Radius
(g/mole) (cm2/sec) v (cm3/g)
a or a,b (Å)
H2 2 52
H2O 18 20
O2 32 19.8
CO2 44 17.7
KCl 76.5
NaCl 58.5 13.9
Urea 60 11.8
Glycine 75 9.335
Glucose 180 6
Sucrose 342 4.586
Inulin 5,000 1.0
Lysozyme 14,400 1.12 0.703
Bovine serum albumin 66,500 0.603 0.734
Hemoglobin 65,485 0.6
Albumin 68,000 0.6 70 by 30 0.733
Tropomyosin 93,000 0.224
Fibrinogen 330,000 0.202 0.723
assuming that the viscosity of the air is not changed by the presence of the odor. This method is
pretty crude, but will still be quite good if the comparison can be made with a molecule of closely
similar molecular weight. Attempting to calculate gaseous diffusion coefficients from aqueous
ones using substitution for the viscosities of gas versus liquid doesn’t work. The estimates for
gaseous diffusion coefficients are 100-fold too low. Bird, Stewart and Lightfoot (1960, 2001)
present an accurate theoretical approach, the Chapman-Enskog theory, which works well over a
wide range of temperatures and pressures.
5-1.3. Influences of molecular shape on diffusion coefficients

The computation of frictional coefficients has been achieved by analysis of continuous systems,
for example, Stoke’s law:
f = 6πηa , (5-4)

58 Diffusion
where f is the coefficient for friction between solute and solvent, a is the radius of a spherical
particle and η is viscosity of the medium. For simple nonspherical shapes other estimates of f
were obtained from hydrodynamic expressions. Examples of the effects of molecular asymmetry
Table 5-3: Frictional coefficientsa
Shape Frictional Coefficient Explanation
Sphere f = 6πηR R = sphere radius

2a = major axis,
2 2 1⁄2 2b = minor axis,
(1 – b ⁄ a )
Prolate ellipsoid f = 6πηR p 2⁄3 2 1⁄2
-
------------------------------------------------------------------------------------------------- Rp = radius of sphere
( b ⁄ a ) ln { [ 1 + ( 1 – b ⁄ a ) ] ⁄ (b ⁄ a ) } of equal volume =
2
(ab2)1/3
2b = major axis,
2 2 1⁄2 2a = minor axis,
(b ⁄ a – 1)
Oblate ellipsoid f = 6πηR 0 ---------------------------------------------------------------------
1⁄2
- R0 = radius of sphere
2⁄3 –1 2 2
( a ⁄ b ) tan ( a ⁄ b – 1 ) of equal volume =
(a2b)1/3
a = half-length,
1⁄2 b = radius,
(b ⁄ a)
Long rod f = 6πηR r ----------------------------------------------------------------------------
1⁄3
- Rr = radius of sphere
( 3 ⁄ 2 ) { 2 ln [ 2 ( a ⁄ b ) ] – 0.11 } of equal volume =
(3b2a/2)1/3
a. See Tanford 1961
are shown in Fig. 5-3. Although such analyses cannot take into account the asymmetric and often
f / f0
Figure 5-3: The dependence of frictional coefficient on particle shape. The ratio f/f0 is
the frictional coefficient of an ellipsoid of the given axial ratio divided by the frictional
coefficient of a sphere of the same volume as the ellipsoid. (From Van Holde, 1971.)

Diffusion 59
very irregular shapes of molecules, and completely ignore the effects of random collisions or of
molecular orientation in a field, the estimates are certainly useful. Deviations from sphericity
increase f since the radius of rotation is necessarily increased by any deviation from spherical.
Elongate molecules have much reduced diffusion coefficients in free solution. (Long change
molecules, however, can penetrate fiber matrices and small pores by “reputation”, a phrase chosen
by De Gennes who envisaged the snake-like head-forward motion of a molecule through a
hindered medium.) Of some relevance are the computations of Levitt (1975), who showed that
diffusion modeling by the Markov processes representing the quantum mechanical expressions
gave results essentially indistinguishable from those of hydrodynamic modeling for porous
transport using continuum mechanics.
5-2. Fick’s “Laws”

5-2.1. Fick’s First Law of motion
Fick’s First Law, for the flux due to diffusion across a plane in one dimension, is
dq dC
Net flux per unit area = J D = ( 1 ⁄ A ) ------ = – D ------- , (5-5)
dt dx
where JD = diffusional flux per unit area, moles/(sec cm2); q = amount, moles; t = time, sec;
D = diffusion coefficient in the region, cm2/sec; A = area available for diffusion, cm2. The driving
force is the spatial gradient in concentration or, more properly, in activity.
Consider a steady-state situation where there are two oceans of instantaneously stirred fluids
with concentrations C1 and C2, separated by a stagnant region of thickness ∆x, as in Fig. 5-4. Then
in steady state the net flux = ∆q/∆t = −DA(C2 − C1)/∆x. Here a conductance, L, equals DA and the
driving force Ψ = ∆C/∆x. Fick’s First Law says ∆C tends toward zero.
∆x
C1
C2
0 x
Figure 5-4: Fick’s First Law of diffusion through a stagnant region in steady state.
As an aside, when the membrane is thin, ∆x goes to 0, and the system can be represented by
ordinary differential equations in which the two chambers have volume V1 and V2:
dC 1 Dm Am Pm Am
--------- = ----------------- ⋅ ( C 2 – C 1 ) = -------------- ⋅ ( C 2 – C 1 ), (5-6a)
dt ∆x ⋅ V 1 V1
defining the membrane permeability Pm = Dm/ ∆x, cm/s. Likewise the equation for the second
chamber is:

60 Diffusion
dC 2 Pm Am
- ⋅ (C2 – C1)
--------- = – ------------- (5-6b)
dt V2
To derive the flux equation, Eq. 5-5, from the general statement, a flux equals a conductance
times a driving force, J = LΨ, one goes back to the principle that the gradient in activity (chemical
potential), ∂µs/∂x, is the driving force for solute:
J s = – L s ∂µ s ⁄ ∂x , (5-7)
where Ls is the conductance for the solute. The chemical potential of the solute depends on
temperature, pressure, and the concentration. T and P are assumed to be uniform, so there is no
change in the activity coefficient with position along the gradient, under which
∂µ ∂µ ∂C
------ =  ------- ------- . (5-8)
∂x ∂C T,P ∂x
conditions the concentration gradient determines the flow. When the activity coefficient, φs, is
concentration-dependent as described in the previous chapter, then ∂µ/∂C must be expanded to
account for changes along the gradient:
( ∂µ ⁄ ∂C ) T,P = ( RT ⁄ C ) [ 1 + C ( ∂ ln φ s ⁄ ∂C ) ] ; (5-9)
L s RT  ∂ ln φ s ∂C
-  1 + C ------------- ------- .
J s = – ------------ (5-10)
C ∂C ∂x
The conductance per unit concentration, Ls/C in Eq. 5-10 can be replaced by 1/ℵfs, the reciprocal
of the molecular frictional coefficient times the number of molecules:
∂ ln φ s ∂C
J s = – -----------  1 + C ------------- ------- ,
RT
(5-11a)
ℵ f s ∂C  ∂x
∂C
or when the activity is constant, J s = – D ------- . (5-11b)
∂x
This defines D, the diffusion coefficient, as
∂ ln φ s
D = -----------  1 + C ------------- = ----------- in dilute solution.
RT RT
(5-12)
ℵ f s ∂C  ℵ f s
From an experimental point of view, D is defined by Eq. 5-11b, called Fick’s First Law. Eq. 5-11b
is in accord with our intuitive expectation that the flux will cease only when the concentration
gradient has vanished. Eq. 5-12 demonstrates that D depends on three factors: RT, which may be
taken as a measure of the kinetic energy of the molecules, a correction term, [1 + C(∂lnφs/∂C)],

Diffusion 61
expressing the fact that the chemical potential depends on solute-solute interaction, and thirdly, on
the size and shape of the molecule, reflected in the frictional coefficient, fs.
For ideal solutions we may neglect the activity coefficient factor and obtain D = RT/ℵfs
which reduces to RT/ℵ6πaη for spheres. If, in addition, η does not vary significantly with local
differences in solute concentration, D is a constant.
5-2.2. Fick’s Second Law of diffusion

At a given point, x, this law expresses the rate of change of concentration with respect to time as a
function of the concentration gradients adjacent to this point, as in Fig. 5-5. By differentiating
Eq. 5-11b (Fick’s First Law) with respect to distance, and substituting from the equations of
continuity, we obtain
2
∂C ( x,t ) ∂ C
------------------ = D ---------2 . (5-13a)
∂t ∂x
Figure 5-5: Fick’s Second Law. The graph depicts the diffusion from an ocean of fixed
concentration at x < 0 into a medium where C(x) = 0 initially and the front advances
with time.
Over a small finite distance x2 − x1 the rate of concentration change is approximated by
∆C ⁄ ∆t = D [ ( ∆C ⁄ ∆x ) 2 – ( ∆C ⁄ ∆x ) 1 ] ⁄ ( x 2 – x 1 ) . (5-13b)
Fick’s Second Law indicates that nonuniform gradients tend to become uniform (when D in
the medium is uniform). This occurs at early times within a membrane with fixed concentrations
on either side, eventually resulting in the straight profile seen in Fig. 5-6, considered next.
5-3. Measuring diffusion coefficients

5-3.1. Diffusion across a thick membrane
The Barrer time lag method for estimating D. In the experiment described by Barrer (1953)
there is a thick membrane separating two well-mixed fluid regions. Solute is added to the left
region at t = 0. Since the membrane contains no solute initially there must be a time lag before
solute reaches the right region, and more time until there is the straight steady-state profile seen in
Fig. 5-4. This is diagrammed in the left panel of Fig. 5-6, taken from the study of Safford et al.

62 Diffusion
x10-2
Figure 5-6: Transient in diffusion across a uniform slab of finite thickness. The
experimental setup at left is used for example by Safford and Bassingthwaighte (1978)
where stream of bubbles rising across the face of the sheet of tissue circulate fluid.
Inside the tissue the gradients become uniform when changes in concentration on both
the donor side CD and the recipient side CR are small prior to the time to reach a
psuedo-steady state of linearly rising concentration. In the right panel is shown
C′R = CR/CD for the condition C′R = 0 at t = 0. The pseudo-steady state is approached at
Dt/l2 of about 0.5, diagrammed as t5, where the profile is a staight line between Cm= CD
at x = 0 and Cm= CR at x = L and the actual recipient concentration is still less than
0.4% of that of the source in this particular case. (The linear region is the first part of an
almost monoexponential rise to the equilibrium C′R = CD.
(1978) on the diffusional transport of tracer-labeled water through a sheet of heart tissue. The
profiles C(x) within the sheet of thickness l are curved until steady state has been reached at time
t5. The curvature takes time to dissipate, in accord with Fick’s Second Law (Eq. 5-13a).
Observations of the time course of solute concentration in the right region show a delay before
any solute appears, then a gradually steepening rate of rise until the pseudo steady state is reached
at about t5. The time lag T introduced by Barrer is the intercept of the straight line fitted to the
steady-state data and extrapolated back to the baseline. For a tissue of uniform properties and
thickness, l, D = l 2/6T. The concentration in the right region, C′R = CR/CD, is a small fraction of
that in the source region on the left side and at time T is still only 0.1% of CD so that the backflux
from right to left is still negligible. The slope, dCR/dt, also gives a measure of D when the
membrane surface area and thickness are known: VRdCR/dt = PAm ∆C = Am(D/l) ∆C, where Am is
the membrane surface area, P is a permeability equal to D/l, and ∆C = CD is the driving force, the
concentration on the source side. The approach assumes that the volume, VR, on right side will be
large enough that CR remains so much less than CD that back diffusion from the right to the left
side is negligible. Note the dimensionless time scale used for generality in the right panel of
Fig. 5-6.
This method has been extended by Safford and Bassingthwaighte (1977) and Safford et al.
(1978) to account for additional complexities commonly found in tissues: (1) heterogeneity of
path length across the sheet of tissue, (2) the presence of immobile binding sites for solutes such

Diffusion 63
as calcium, and (3) for diffusion through and around cells distributed in interstitial fluid, with
solute penetrating the cell membranes.
The Barrer equation for diffusion across a planar slab of tissue of thickness l (also given by
Crank, 1975) is provided by a convergent series:
 – Dm 2 π 2 t
m =M m
DAt lA 2lC D A d ( –1 )
C′R = ---------- – ------- – ------------------
Vl 6V π
2
- ∑ m2 exp  ----------------------
-------------
-
λ l 
2 2 , (5-14)
m=1
where D is the apparent diffusion coefficient, cm2/s; A is area, cm2; t is time, seconds; V is volume
of the right or recipient chamber, ml; l is slab thickness, cm. At t = T, the time intercept,
2
l
D = ------- . (5-15)
6T
When a pseudo steady state is reached, that is, when the transient terms have vanished and the
slope dC’R/dt, becomes constant, another measure of D is obtained:
Vl d
D = ----- ⋅ C′R . (5-16)
A dt
5-3.2. Diffusion across an interface between unstirred regions

This is the basis of a technique commonly used for measuring diffusion coefficients in gels, cells,
or tissues. The usual approach is to label one region uniformly with tracer and to allow time for
the tracer to diffuse into the neighboring region; in Fig. 5-7 the region where x < 0 is initially free
of tracer.
The random nature of the diffusion process is re-emphasized by the fact that the solution for
Eq. 5-13a gives the Gaussian error curve for dC/dx, the curve to the right in Fig. 5-7:
∂C Co 2
– x ⁄ 4 Dt
------- = ------------------------
- ⋅ e . (5-17)
∂x 2 ( πDt )
1⁄2
The standard deviation, SD, of ∂C/∂x is 2 Dt and its height, ∂C at x = 0, is Co/2(πDt)1/2 or

0.3998 Co/SD. The SD increases with the square root of time. The area under ∂C/∂x from −∞ to
+∞ is Co, so that as the solute disperses the curve shape remains Gaussian and the area constant,
but the ratio (area/height)2 increases linearly as in Fig. 5-8.
area 1⁄2
--------------- = 2 ( πDt ) = 2.507 × SD . (5-18)
height
This provides a direct estimate of D = (slope, (area/height)2/dt)/4π.

An approach similar to that in Fig. 5-7 is to label a thin cross-section at x = 0. In the ideal
situation, an infinitely thin section, C(x), will have the form that ∂C/∂x has in Fig. 5-7; that is,
impulse labeling (with respect to distance x) has a response which is the derivative of step

64 Diffusion
Figure 5-7: Diffusion across an interface into a stagnant region. The diffusion
coefficients are assumed the same in the source region as in the stagnant region.
Figure 5-8: Determination of diffusion coefficient from plot of (area/height)2 of ∂C/∂x

from an experiment such as that in Eq. 5-7. There is a slightly positive ordinate intercept
due to a small amount of dispersion present at t = 0 due to imperfections at the interface.
function labeling. (This is completely analogous to the relationship between system responses
following impulse inputs and step inputs.)
5-4. Applications to biological situations

5-4.1. Flux by diffusion across a uniform membrane depends on solubility
Solubility in a lipid membrane is increased by alkyl groups and decreased by polar groups (−OH,
−COOH, NH2). The partition coefficient, λ, is the ratio of its lipid solubility to water solubility:
λ = C membrane ⁄ C medium . (5-19)

Diffusion 65
In the situation diagrammed in Fig. 5-9 the gradient across the membrane is not provided by
the concentrations at the surfaces, but by those within the membrane, ∆Cm = λC2 − λC1:
∆C m
J s = – D m ----------- , (5-20a)
l
where Js is the net flux of solute from side 1 to side 2. To be more explicit, it might be written as
λ(C2 – C1)
J s = J S net12 = J 12 – J 21 = – D m --------------------------- . (5-20b)
l
In Eq. 5-20b the driving force is rewritten in terms of the observed solution concentrations times
the membrane-to-solution partition coefficient. In many situations where Dm, λ, and l are
unknown (as with most biological membranes), then an observed flux with a known concentration
difference provides an estimate of the combination of factors lumped together as a permeability:
J s = – P∆C where P = Dm λ ⁄ l . (5-21)
Examples of membrane/water partition coefficients for biological membranes are glycerol, 10−4;
urea, 10−4; ethanol, 10−2; triethylcitrate, 1; dimethylsulfoxide, 200.
Js
lipid
membrane
C1
C2
0 l
x
Figure 5-9: Diffusion across a nonporous membrane. C1 and C2 are concentrations in

well-stirred aqueous oceans. λ = 1.5. The gradient dCm/dx is greater than (C1-C2)/l
when λ > 1, thus high membrane solubility augments permeation.
5-4.2. Diffusion through pores
1. Large pores in an otherwise impermeable membrane can be treated as if there were free
diffusion through a fraction of the membrane. This generally applies to smaller solutes and
membranes with large water contents. The flux of solute is given by
A p ∆c
J s = – D w  ------- ------ , (5-22)
 A m l

66 Diffusion
where Dw is the diffusion coefficient in water, Ap/Am is the pore area as a fraction of the
membrane area, and l is the membrane thickness. (Again Js is moles s−1 cm−2 and ∆c is the
difference in concentrations on the two sides of the membrane.)
2. Small pores have additional effects due to (i) entrance effects, (ii) friction between pore wall
and solute, and (iii) solute-solute and solute-solvent interactions. The entrance effect in its
simplest form can be regarded as exclusion of the center of the molecule from any position
closer to the wall than the molecular radius, as in Fig. 5-10:
2
effective pore area = A p ( 1 – a ⁄ r p ) .
rp
Figure 5-10: Cross-sectional area of pore available to solute is less than the pore area.
The effective pore area is Ap (1- a/rp)2.
This reduction in available volume for solute within a pore has the effect of reducing its
average concentration relative to that in free solution outside the pore. This is equivalent to
other mechanisms of molecular exclusion, and is the same thing causing reduction in
hematocrit in capillaries relative to that in large vessels, that is, due to the fact that red blood
cells cannot be centered at the wall.
If a cell is dumped into medium containing D2O then the concentration inside rises
rapidly. The transport rate of D2O is about 1% of that through a similar thickness of water
and suggests that only 1% of the surface is available for D2O transport. Therefore, consider
aqueous pores totalling 1% of the area of surface, or solubility of D2O in membrane being
1% of solubility in H2O. These are probably mainly aquaporin channels (Agre et al., 1993).
Good reviews on porous transport are those of Bean (1972), Curry (1984), and Deen (1987).
For reviews of solute transport through hindered matrices see the excellent chapter in the
Handbook of Physiology by Curry (1984), the two volume book by Comper (1996a, b) on the
extracellular matrix, and articles by Nugent and Jain (1984) and Tong and Anderson (1996).
Porous transport will be covered in greater detail in Chapter 6.
5-4.3. Diffusion across walls and pores in parallel

Now consider the membrane to have a fraction of the area that has the area of a set of pore, Ap/Am,
and the other fraction being a normal lipid bilayer with fractional area 1 - Ap/Am, The latter
fraction is Npore. πr2/Am. Then Js is the sum of the two fluxes in parallel, per unit membrane area.

Diffusion 67
J s = J pores + J membrane (5-23a)
A p ∆C A ∆C
= – D pore  ------- -------- – D m  1 – ------p- λ --------
 A m l  Am  l
D pore A p Dm A
= – -------------  ------- + λ -------  1 – ------p- ∆C
l  A m l  A m
= – ( P pore + P membrane ) ∆C , (5-23b)
where Ppore and Pmembrane are defined by the expressions on the line preceding.
Since Ap/Am is usually very small for cell membranes, the product of the partition coefficient
λ times Dm is the major determinant of the flux. The Dpore/l is determined by the nature of the
pores and the solutes, as discussed in Chapter 6-7, and is here written as if there is no steric
hindrance.
5-4.4. Diffusion from a stirred infinitely large source into a non-consuming stagnant
region
First consider diffusion into a plane sheet of unstirred isotropic material of thickness 2l and
containing no obstructions or binding sites. The initial concentration in the plane sheet is C0; at
t = 0 the concentrations at the two surfaces are set to C1. Fig. 5-11 depicts the concentration
profiles within the slab at a succession of normalized times, Dt/l2, as solute diffuses in and
eventually equilibrates with the external concentration on each side. At early times these curves
are the same shape as those for low values of Dt/l2 in Fig. 5-6.
Patlak and Fenstermacher (1975) used a similar method for measuring diffusion into brain
tissue which was dependent on maintaining a constant concentration within the cerebral
ventricles, which they perfused, and, after sectioning the brain and measuring concentration as a
function of distance from the surface, observed the profiles shown in Fig. 5-12. Since the profiles
were all close to the surface this is effectively one-dimensional diffusion from a constant source
into an infinite region. The profiles do not become constant since solute continues to enter.
“Infinitely large source” implies that there is no apparent depletion of the solute concentration
within the source. Observation of the concentration profile within the brain was interpreted in
terms of the expression for one-dimensional diffusion, giving C(x) where x is the distance from
the surface:
C ( x) x
------------ = erfc ---------------------
1⁄2
-. (5-24)
Co 2 ( Dt )
(Note that the ordinate scale in Fig. 5-12 is erfc, not exponential.) The complementary error
function, erfc, is 1.0 minus the integral of the Gaussian pdf (probability density function). that is,
is 1.0 minus the integral of (1/((2π)0.5σ) ⋅ exp(−x2/2σ2), and σ = (Dt)2. The profiles in Fig. 5-12
have the shape of the integral of C(x) in Fig. 5-7. (Extend the analogy to time responses: this is
analogous to a ramp input function, the integral of the step input function.) The profile continues
to change with time, an unreal situation, since Eq. 5-24 assumes the stagnant region is also
infinite, and x may increase without limit.

68 Diffusion
1.0
1.5
1.0
0.8
0.6
-
C1 – C0
------------------
C – C0
0.4
0.5
0.3
0.2
0.15
0. 0.0
03 0.04 0.0 0.1
0.00
0.0
0.0
6 8
2
1
5
0.0
−1.0 −0.5 0.0
x/l
Figure 5-11: Diffusion into a plane sheet of thickness 2 l at various times after exposure
to Co at both surfaces. Concentration distributions at various times in the sheet −l < x < l
with initial uniform concentration Co and surface concentration C1. Numbers on curves
are values of Dt/l2. No consumption of solute. (After Crank, 1956.)
5-4.5. Diffusion into a region with solute consumption
The concentration profiles do become constant if there is consumption within the tissue, or if
the tissue thickness x is finite. If the consumption is first-order (that is, in proportion to the
concentration), then Eq. 5-13a has an additional term:
2
∂C ∂ C
------- = D ---------2 – KC , (5-25)
∂t ∂x
where K is a fractional clearance, ml · (ml tissue)−1 s−1. The clearance K can be due to loss by
permeation from the tissue into capillaries or by a first order metabolic consumption. At high K
the concentrations exponentially approach zero.
When the consumption is zero-order, that is, uniform throughout the tissue, calculations can
only be made for regions where sufficient solute is available. For a thin sheet of muscle with
uniform O2 consumption, at steady state the conditions are

Diffusion 69
Figure 5-12: One-dimensional diffusion from a constant source into an infinite region.
The profiles do not become constant since solute continues to enter. Inverse
complimentary error function graph of 14C-creatinine tissue concentration profiles in
dog caudate nucleus after two different methods of perfusion. (From Patlak and
Fenstermacher, 1975.) The complementary error function, erfc, Eq. 5-24, is 1.0 minus
the integral of the Gaussian pdf given in Eq. 5-17. (Note that the graph ordinate scale is
erfc, not exponential.)
2
∂ C K′
---------2 = ----- , (5-26)
∂x D
where K′ is the zero-order consumption in moles · (ml tissue)−1 · s−1 independent of the
concentration, C(x). The resultant profiles, Fig. 5-13, are parabolas. The maximum depth that can
be supplied is dependent on the concentration at the surface relative to the consumption. In this
situation the concentrations become truly zero in the regions where utilization exceeds influx.
Patlak and Fenstermacher (1975) observed that the profiles became stable when the cerebral
ventricles were perfused with 14C-urea solution, shown in Fig. 5-14, in contrast to the situation for
creatinine shown in Fig. 5-12. The difference is due to the loss of urea from the tissue into the
blood perfusing the brain, through the capillary membranes, while creatinine does not cross the
blood-brain barrier and therefore continues to diffuse deeper into the tissue.
5-5. Diffusion in Heterogeneous Media

5-5.1. Diffusion through ISF and cells in parallel and in series
Diffusion of gases and moderately hydrophilic solutes can occur both through and around cells in
the ISF (interstitial fluid space). Heterogeneous diffusion occurs in gel matrices, especially for
molecules large enough to be excluded from some of the water, in cellular tissues, in packed cell
columns and in heterogeneous, multicellular tissues. In 1873, Maxwell worked out an equation

70 Diffusion
consumption
C0 0 C0
−l/2 0 l/2
x→
Figure 5-13: Steady-state profiles of concentration with zero-order consumption in a

uniform sheet of thickness l are parabolic. Parameters for Eq. 5-26 were K′, the
zero-order consumption in moles · (ml tissue)−1 · s−1 = 1, 2, and 3, the thickness l was x
cm, D was y cm2/s, and the surface concentration was z mM. (PARAMS)
for the effective bulk diffusion coefficient in the tissue, Db, for the situation where diffusion
occurs through both:
D 2D e + D i – 2φ ( D e – D i )
------b = ---------------------------------------------------------
- , (5-27)
De 2D e + D i + φ ( D e – D i )
where De and Di are extracellular and intracellular diffusion coefficients and φ is the cell volume
fraction of the tissue. When φ = 0, no cells, the right hand side is unity, so Db = De. When φ = 1,
no interstitial space, Db = Di.
A modern example is diffusion across a slab of unperfused tissue, as seen in an experiment
by Safford et al. (1978) in which they looked at the transport of tracer 3HHO water (permeating
and diffusing across cells as well as ISF) and sucrose (unable to permeate cell membranes and
therefore restricted to the ISF). The formulation of the problem is similar to Maxwell’s but
incorporates a finite permeability of the cell membrane retarding the exchange between cells and
interstitium. (In Fig. 5-15 the cells are diagrammed as square beams of infinite length.)
The calculation is for one-dimensional diffusion via two paths continuously connected: one
path is via extracellular ISF only, and the other is through cells and ISF alternately. The
cell-to-cell spacing Lo cm is the same laterally as axially, and is determined by the cell size L cm
and the cytocrit, Cct, (the cell volume fraction, analogous to the hematocrit for RBC fraction of

Diffusion 71
Figure 5-14: Diffusion profiles at steady state when solute is being uniformly
consumed or removed from the tissue at various rates (Eq. 5-25). Semilogarithmic plots
of 14C-urea tissue concentration profiles in dog caudate nucleus at two durations of
ventriculo-cisternal perfusion with urea solution. (Patlak and Fenstermacher, 1975.)
blood) such that Lo = L(1/ Cct – 1 ), as diagrammed in Fig. 5-15 (from Safford et al., 1978).For
cardiomyocytes L is about 10 microns and with a cell volume fraction of 59%, Lo = 3 microns.
The experiment is diagrammed in Fig. 5-6 and the data on the tracer concentrations, CR(t), of
3
HHO and 14C-sucrose are fitted using Barrer’s Equation 5-14 to give an overall bulk diffusion
coefficient, Db, of 2.2 × 10−6 cm2/s for each solute. From this result one estimates intracellular, Di,
and extracellular, De, diffusion coefficients and the cell membrane permeability, P cm s−1, from
anatomic data by translating the cell volume fraction φ into values of L and Lo, leaving only the
D’s as free parameters. Having data from a pair of tracers, one of which does not enter cells, the
sucrose, further constrains the parameter estimates by defining diffusivity in the extracellular
path. Db is given by:
–1 –1
  L o - Di 
D b = D e   -------------- + L  ------ ( WαK – Q ) – L o ( αK + Q )  , (5-28)
L + L o  D e
 

72 Diffusion
Figure 5-15: Diffusion in parallel through and around cells embedded in interstitial
matrix. While Eq. 5-28 is specific for the square array of cells in this figure, the
numerical solutions for the equations using hexagonal arrays of circular cells are
virtually indistinguishable. (From Safford et al., 1978, their figure 4.)
1⁄2
where α = 2P  ------------- + ----------------
1 1
 L ⋅ D i L o ⋅ D e
E = L + Di ⁄ P
F = [ 1 – cosh ( αL ) ] ⁄ L + P ( 1 + W ) ⁄ D i
G = sinh ( αL ) ⁄ L + Wα
H = ( 1 + Y ) ⁄ E + P ( 1 + W ) ⁄ Di
I = X ⁄ E – Wα
GH
J = --------- + I
F
1 1 H
K = --- --- – -------
J E FL
n = Di α ⁄ P
U (1 + W )
Q = α KW – -------------------------
n
1 1
U = ---- --- + GK
F L

Diffusion 73
Lo De
W =  -----  ------
 L   Di 
X = W [ sinh ( αL ) + n cosh ( αL ) ]
Y = W [ cosh ( αL ) + n sinh ( αL ) ]
Using these expressions for the analysis, they found that the bulk diffusion coefficient for
water in heart muscle was 2.5 × 10-6cm2 s-1 or about 10% of the free diffusion coefficient in water,
2.38 × 10−5 cm2 s−1 at the 23° C temperature of the experiments. The data suggested a cell surface
permeability of 2 cm/s and intraregional diffusion coefficients of about 25% of the free aqueous
diffusion coefficient both inside the cells and in the interstitial space. At the same time the sucrose
(MW = 342 Daltons) diffusion coefficient in the interstitial space was 1.5 × 10−6cm2 s−1, or 22.6%
of the aqueous diffusion coefficient. Thus the tortuosity and steric hindrance of the interstitial
matrix reduces the effective diffusion coefficient for small solutes in the ISF to one quarter of the
free diffusion coefficient, and the intracellular diffusion of water is about as rapid as is
extracellular, but low membrane permeability is a factor in reducing the rate of movement of
tracer water to 10% of that in water.
In whole organ studies we observe (Yipintsoi and Bassingthwaighte 1970 #67;
Bassingthwaighte and Beard 1995 #432) that water exchange is so fast in the heart that it is
essentially flow-limited in its capillary-tissue exchange, which means that slowing of radial
diffusion into the nearby cells cannot be detected. This is exactly what one would predict from the
estimated P of 2 cm/s, since though measurable in the diffusion experiment is far higher than
needed to explain very rapid exchange in an organ with high capillary density and short radial
diffusion distances.
The extreme cases for Eq. 5-28 occur with the cell permeability P of either zero or infinity:
Lo ( L + Lo )
D b ( P = 0 ) = D e ⋅ -----------------------------------2- , and (5-29)
L ( L + Lo ) + Lo
1
D b ( P = ∞ ) = ---------------------------------------------------------------
L Lo
---------------------------- + -------------------------- -
LD i + L o D e ( L + L o )D e . (5-30)
Note that in Eq. 5-29 the L(L + Lo) is the area for diffusion and is divided by the total area of the
plane across which diffusion occurs. With other values being the same, Db (P = 0) would be
1.1 × 10−6 cm2/s, and Db (P = ∞) would be 5.6 × 10−6 cm2/s.
This computation improves upon a formula given by Redwood et al. (1974) because it
accounts for exchange across the whole surface of the cells instead of only the part of the surface
perpendicular to the diffusion front. When P = 0, Eq. 5-29 can be further reduced to illustrate that
the result differs from Maxwell’s formula by accounting for the absence of useful vertical
diffusion in the horizontal spaces between the cells: this is simply a stagnant region, equivalent to
a dead-end pore, when P = 0 so that:

74 Diffusion
2
φ – LL 0 ⁄ ( L + L 0 )
D b ( P = 0) = D e ⋅ --------------------------------------------2- , (5-31)
1 – LL 0 ⁄ ( L + L 0 )
rather than simply De φ. The same reduction with P = 0 would occur with the Redwood model.
This cell-interstitium model analysis is also applicable to the diffusion and exchange of
oxygen in blood, where there is diffusion inside and outside RBC. Inside the RBC most of the
oxygen is bound and diffuses as Hb(O2)4, where the high concentration gives a high intracellular
transfer rate.
5-5.2. Diffusion across an uneven slab of tissue:

Copy in here the theory section of Suenson and ...JB #109 pages 1118 onward as edited by hand.
use figs 1,5,6 of suenson 1974
5-5.3. Diffusion through tissues with dead-end pores:

This is a situation analogous to one with fixed binding sites. Given a Barrer-type experiment, as in
Fig. 5-6, instead of examining intratissue concentration profiles one has only the rising time
course of concentrations in the unlabeled chamber, CR(t), to use to estimate the parameters of the
diffusional system. Tissue sheets are usually non-uniform; the equations need therefore account
for: heterogeneity of path lengths, diffusion into sequestered regions, volumes or binding affinities
in those regions, and both intracellular and extracellular space. Following the derivation from
Goodknight and Fatt (1961 #2251) for oil shales.........??
Here use the figure in Suenson 1974 jbb#109 fig 4
5-5.4. Diffusion through a hindering matrix

Rview sections of Curry and Michel and BSL re pointers for this.
5-6. Diffusion of solutes which can be bound to absorbing sites

Many small solutes bind to proteins; this is physiologically advantageous in a wondrous variety of
circumstances. Fatty acids binds tightly to plasma albumin, occupying as many as seven binding
sites and usually about three, and is about 99.94% bound. Free fatty acids are noxious, form
soaps, dissolve membranes, oxidize becoming rancid, none of which occurs when they are tightly
bound to albumin. Retinone, a hormone ...... and testosterone, an anabolic steroid, have
undesirable effects on many cells, but can be safely and selectively delivered to their target tissues
by having a receptor for the carrier protein - hormone compolex on the cell surfaces in the target
tissues, and allowing for its selective release. Not only do such mechanisms protect the rest of the
body from these agents, but much smaller amounts of the agent need to be produced and delivered
in order to elicit the physiologically desired response. As a broad generality, most of these solutes
which bind to plasma proteins are lipid soluble, and could readily permeate cell membranes
everywhere to cause damage, so the binding prevents the noxious effects. The binding also raises
the amount carried by blood. The total fatty acid concentration in blood is over 100 times the
solubility limit in the absence of binding proteins, so the blood’s carrying capacity is hugely
increased.
From the situations described in the earlier sections of this chapter it is apparent that most
processes influencing diffusion do so by retarding molecular mobility, and it is clear that binding

Diffusion 75
at a diffusion front always retards the rate of entry of the solute into the region. How then can
binding facilitate diffusion? It is not enough just to point out that binding of a solute also provides
for its buffering, stabilizing its availability in fluctuating states; that is not diffusional facilitation.
But if the concentration of binding sites is so high relative to the free solute concentration, that the
diffusional flux of the complexed solute if faster than the diffusional flux of the uncomplexed or
free solute, then one has facilitated diffusion. the facilitation is thus in terms of total flux, not in
the velocity of the individual molecules.
Consider the general situation for the diffusion of a solute, S, with a concentration s, in the
presence of a solute B to which S can bind. The concentrations are functions of time and space, so
we designate the concentrations of free solute S to be s = s(x, t), of uncomplexed solute B to be
b = b(x, t), and of the complex SB to be sb = sb(x, t). The reaction to form SB is
k1
S+B — \ SB ,
\
k –1
k b = k –1 ⁄ k 1 .
Returning to the basic one dimensional flux equation, Eq. 5-5, JD = - D dC/dx, first consider
the analogous steady state situation for fluxes of free and complexed forms of a solute diffusing in
parallel. At any particular point in a plane between a planar source and a planar sink, the steady
state flux must be the sum of the fluxes of the two species:
ds dsb
J total = J s + J sb = – D S ⋅ ------ – D SB ⋅ --------- (5-32)
dx dx
In this equations DSB will ordinarily be much smaller than DS since B is a large molecule
compared to S. The fraction bound is determined by the concentrations and the affinity, in accord
with the equilibrium relationship s . b/ sb = k-1/k1 = kb, where the latter is the dissociation
constant, Molar. For a high affinity binding kb is small, so s / sb is small, and the bulk of S is in
the bound form. If the gradients of s and sb are proportional to their concentrations, then the
effective diffusion coefficient for the flux of S, D′S , is the concentration-weighted average for the
two diffusing species:
s ⋅ D S + sb ⋅ D SB
D′S = ---------------------------------------
-. (5-33)
s + sb
This expression is correct for the diffusion at any point, and can reexpressed in more refined form
to relate to particular circumstances of equilibrium binding or for slower rates of reaction of
substrate with binding site, or to more explicitly account for the total concentration of binding
sites, bT.
5-6.1. General aspects of diffusion of solute in the presence of binding sites

Having the perspective provided by the steady-state expressions, now consider the transients
in order to reconcile the ideas that binding a diffusing solute always retards a diffusion front even

76 Diffusion
though the same binding site may foster facilitated diffusion. The expressions for the
one-dimensional diffusion of the three solutes, S, B, and SB are
2 2
∂s ⁄ ∂t = – k 1 ⋅ s ⋅ b + k –1 sb – G ( s ) ⋅ s + D S ∂ s ⁄ ∂ x , (5-34)
2 2
∂sb ⁄ ∂t = k 1 ⋅ s ⋅ b – k –1 ⋅ sb + D SB ∂ sb ⁄ ∂ x , (5-35)
2 2
∂B ⁄ ∂t = – k 1 ⋅ s ⋅ b + k –1 ⋅ sb + D B ∂ b ⁄ ∂ x , (5-36)
where G(s) ⋅ s is a generic term for concentration-dependent consumption of S, but in the cases
which follow we assume G(s) = 0. In these general equations DSB may differ from DB; but often
where B is a large molecule compared to S, DSB = DB. The equations can then be simplified if at
each point in space:
b ( x, t ) + sb ( x, t ) = b T ( x, t ) , (5-37)
where bT is the total concentration of B and SB. In the case where DSB = DB then bT is constant.
5-6.2. Diffusion in the presence of immobile binding sites (DB = DSB = 0)

Situation 1. Transients in solute concentrations s.
Only S diffuses, but the binding of S by B retards the flux of S. We assume that S is not
consumed [G(s) = 0]. We make the additional assumption that the total binding site concentration
bT is uniform in space. The local concentrations of S and SB are changing only through the
diffusional flux of S, as is seen by summing Eq. 5-34 and Eq. 5-35:
2 2
∂s ⁄ ∂t + ∂sb ⁄ ∂t = D S ∂ s ⁄ ∂ x . (5-38)
When the binding rate, k1, s−1 and unbinding rate, k−1, moles ⋅ s−1, are both fast relative to the
diffusional flux, one can assume instantaneous equilibrium binding:
s ⋅ b ⁄ sb = k b = k –1 ⁄ k 1 . (5-39)
Following Safford and Bassingthwaighte (1977), using the chain rule to estimate ∂sb ⁄ ∂t :
∂sb ∂sb ∂s ∂s k b ⋅ b T
-------- = -------- ⋅ ----- = ----- ⋅ --------------------2- . (5-40)
∂t ∂ s ∂t ∂t ( k + s )
b
Substituting Eq. 5-40 into Eq. 5-38 gives:
2
∂s DS ∂s
----- = ---------------------------------------------2- ⋅ -------2- , (5-41)
∂t 1 + k b bT ⁄ ( k b + s ) ∂x

Diffusion 77
which thereby defines the effective diffusion coefficient, D′S , as a function of bT, and of the local
solute concentration, s:
DS
-.
D′S = ------------------------------------------------------------
2
(5-42)
1 + ( bT ⁄ k b ) ⁄ ( 1 + s ⁄ k b )
The D′S applies at each local point along a diffusion front. As s → ∞, D′S /DS → 1. When s = kb ,
D′S /DS = 4/(4 + bT/kb). The more fixed binding sites, the slower the diffusion. See Fig. 5-16.
1.0
0.1
D′ 1
------S- 0.5
DS
10
100 = bT/kb
0.0
0.001 0.01 0.1 1 10 100 1000
s ⁄ kb
Figure 5-16: Relative effective diffusion coefficient D′S /DS versus concentration for S in
an invading front when there are immobile binding sites B of uniform total
concentration bT (Eq. 5-42). Numbers on curves are values of bT/kb. Note that at s/kb =
1, the values for D′S /DS = 4/(4 + bT/kb) are 0.9756, 0.8, 0.2857, and 0.0099 for bT/kb
= 0.1 to 100. Compare these values of D′S /DS for native solute entering a field of
immobile free binding sites with those for tracer invading a field of pre-equilibrated
sites, Fig. 5-19.
Consider the case of an invading front, for example one spreading from a constant source so
at x = 0 when t = 0, into a region initially containing no S but containing a uniform concentration
of binding sites, b. Then D′ is high at x < 0, low at x > 0 because D′S /DS is << 1 when most S is
bound. For a high affinity site (low kb/bT), there is a low D′S ahead of the front, slowing it because
of the adsorption of S onto the immobile B. If we call the midpoint of the front the point at which
s = 0.5s0, then behind the midpoint D′S is high. The front is therefore steeper at its mid-level then
would be seen with simple diffusion of S in the absence of B.
The steepness of the diffusion front is dependent mainly on the concentration of B in the
medium relative to that of S. In Fig. 5-17, left, are shown profiles of s(x) versus x at one time for

78 Diffusion
three different concentrations bT/s0, where s0 is the concentration at a source of uniform

concentration at x = 0.
1.0 1.0
Fast binding
t = 0.8 s t = 0.8 s
0.8 0.8
bT = 100
100 10 0.001
0.6 bT = 10 0.6
sb/bT
Slow binding
s/s0
bT = 0.001
0.4 0.4
0.001
100
0.2 Slow 0.2
Fast 10
Fast Slow
0.0 0.0
0 1 2 3 4 0 1 2 3 4
x ⁄ Ds ⋅ t x ⁄ Ds ⋅ t
Figure 5-17: Diffusion into a region with immobile binding sites. Concentration
profiles for free solute concentration, s, at t = 0.8 s after joining a constant source of s
with concentration s0 at concentration s0 = 10kb to the left of x = 0 at t = 0. Left panel.
The effects of total binding site concentration, bT, and of the rate of solute binding.
Higher concentrations, bT = 100 or 10 mM, retard the diffusion front compared to that
with bT = 0.001 mM (the effects of which are indistinguishable from those of bT = 0).
Fast binding (with dissociation rate k−1 of 1 s−1) retards the front more than does slow
binding (k−1 = 0.1 s−1) because the binding reduces the fraction of s free to diffuse.
Right panel. The ordinate sb/bT is the fraction of binding sites occupied. With slow
binding (k−1 = 0.1 s −1) the balance between binding and diffusive movement of solute
leaves almost half the sites unoccupied, while with fast binding the fractional site
occupancy, sb/bT, higher. Note that by comparing the profiles in the left and right panels
that the profile of sb appears to be ahead of the profile for free unbound s, even though it
is only free s that moves. Parameters for both panels. s0 = 10 mM, Ds = 10−6 cm2/s,
Dsb = 0, the dissociation constant kb = 1 mM (and kb = k1/k−1, the off and on rates). All
profiles at t = 0.8 s.
An invading concentration front is always steepened by binding to an immobile site. Profiles

at a succession of times as s diffuses into a stagnant front are shown in Fig. 5-18. The dotted lines
show the invasion of s for the same times at a low concentration of b.
Situation 2. Tracer *s in a medium with s and sb constant and uniform.
Both *s and s diffuse, but b and sb do not. For the non-tracer s, ∂s/∂t = ∂sb/∂t = ∂s/∂x = 0, but
for an added tracer, for example with a step front −U(x) ⋅ so , then ∂*s/∂x > 0. Using the fact that
*sb/*s = sb/s = b/kb for equilibrium binding, and that for the non-tracer sb/b = b/kb = bT/(s + kb),
then
* * 2*
∂ s ∂ sb ∂ s
------- + ---------- = D S ---------2- , (5-43)
∂t ∂t ∂x

Diffusion 79
1.0
bT = 0.1 mM
fractional saturation, sb / bT
0.8
bT = 20 mM
0.6
0.4
10 50 100 ms at bT = 0.1 mM
0.2
10 50 100 ms at bT = 20 mM
0.0
0 0.2 0.4 0.6 0.8 1.0
10−6
bT = 0.1 mM
Apparent Diffusion Coefficient, D’s cm2s-1
100
10
t = 50 ms
0.5 × 10−6
bT = 20 mM
10 50 100 ms
0.0
0 0.2 0.4 0.6 0.8 1

Distance from source
Figure 5-18: Diffusion into a stagnant region containing immobile binding sites.
Diffusion front positions (upper) and apparent diffusion coefficients (lower) at three
times (10, 50 and 100 ms) after initiating entry of solute s into the region. Parameters:
s0 = 10 mM, Ds = 10−5 cm2/s; kb = 1 mM; k−1 = 1000 s−1 (very fast binding); bT =
0.1 mM, only 0.1 s0. When bT is low the effective diffusion coefficient at the advancing
front is reduced by less than 10%, whereas with 20 mM bT the front advances much
more slowly with the effective diffusion coefficient reduced to less than 10% of the free
solute Ds.
* 2* * 2*
∂ s bT  ∂ s ∂s Ds ∂ s
------- 1 + -------------
- = D S ---------2- or ------- = ------------------------------------------ ⋅ ---------2- , (5-44)
∂t  s + k b ∂x ∂t ( 1 + bT ⁄ ( s + k b ) ) ∂x
which defines
1 s + kb
D′s ⁄ D s = -------------------------------------- = -------------------------
-. (5-45)
1 + bT ⁄ ( s + k b ) s + k b + bT

80 Diffusion
At very low s, at the low-concentration limit as s → 0, the ratio D′s /Ds approaches the same limits
as given in Eq. 5-42 and Fig. 5-16.
kb 1
D′s ⁄ D s = ----------------
- = ------------------------ . (5-46)
k b + bT 1 + bT ⁄ k b
The relationships shown in Fig. 5-19 between the effective D′s for tracer and the concentration of
nontracer s do not have the same shape as those of Fig. 5-16, being less steep at the inflection
points (Fig. 5-19). At s /kb = 1, D′s /Ds = 2 / (2 + kb / bT). At high values of s >> kb, D′s /Ds still goes
to 1.0.
1.0
0.1
1
D′
--------S 0.5
DS
10
100 = bT ⁄ k b
0.0
0.001 0.01 0.1 1 10 100 1000
s ⁄ kb
Figure 5-19: Relative effective diffusion coefficient D′s /Ds for tracer solute *s in a
medium with constant equilibrium concentrations of mother solute s and immobile
binding site b (Eq. 5-45). Note differences from Fig. 5-16. At s /kb = 1, D′s /Ds = 2 / (2 +
bT / kb), or 0.952, 0.667, 0.166, 0.0196, all of which are lower than for the invading front
of the non-tracer mother solute. These curves are all shifted to the right, higher
concentrations, compared to those of Fig. 5-16.
Equation 5-44 and Fig. 5-19 show that D′s /Ds is dependent only on the concentration of
mother solute S and not on the tracer concentration *s. The diffusion front for tracer therefore has
a shape defined solely by the random molecular motion of the tracer, so that the fronts are
Gaussian, and the effect of the binding sites is to reduce the rate of diffusion, since D′s /Ds < 0.

Diffusion 81
5-6.3. Diffusion in presence of mobile binding sites (both Db and Dsb finite)
5-6.3.1. Fast binding or equilibrium binding:

Situation 1. Transients in solute concentrations. The situation begins with Eqs. 5-34 to5-31.
With equilibrium binding, the total flux of s is given by the sum of fluxes of free and bound s:
2 2
∂s ∂sb s ∂ s sb ∂ sb
----- + --------- = -------------- D s -------2- + -------------- D sb ----------
-. (5-47)
∂t ∂t s + sb ∂x s + sb ∂x
2
Given instantaneous equilibration, kb = s ⋅ b/sb, and given that b + sb = bT, then this reduces to
2 2
∂s ∂sb s + kb ∂ s bT ∂ sb
----- + --------- = -------------------------- D s -------2- + -------------------------- D sb ----------
-. (5-48)
∂t ∂t s + k b + b T ∂x s + k b + b T ∂x
2
The flux by diffusion of sb + s is increased over that by diffusion of the free solute only, since the
effective diffusion coefficient is the weighted sum of the two forms in Eq. 5-48.
Situation 2. Tracer transients with s and sb constant. The diffusion of tracer does not affect the
free or bound concentrations of solute and therefore, with equilibrium binding, the diffusion
coefficient for tracer is the same as that for non-tracer solute (free and bound) at the same location
in the solution.
5-6.4. Slow binding kinetics and mobile binding sites

Preceding sections considered the influences of binding on solute diffusion; in case A we saw that
immobile binding sites retard solute diffusion, while in case B the mobility of the binding sites
could lead to either facilitation or inhibition of diffusion depending on the relative diffusivities,
affinities, and concentrations of the solute and binding protein. In both cases we assumed
infinitely fast equilibration, that is, equilibrium binding.
When there is slow attachment to the binding site, an invading solute front (a wave of high
concentration) advances more quickly than would occur with instantaneous equilibrium binding.
For any particular value of the equilibrium dissociation constant, a slow association rate k1 must
be matched by a slow dissociation rate k−1 to maintain the same kb = k−1/k1. Thus, if the off-rate
k−1 is low, the degree of diffusional facilitation is partially offset by the retardation in release.
In this section we reconsider the factors leading to diffusional facilitation or retardation in
the light of slow binding reactions. The relevant partial differential equations are Eqs. 5-34 to
5-37; in this case the solutions are obtained by numerical methods, as described by Barta et al.
(2000).
An exemplary situation is the facilitation by albumin of the flux of fatty acid from a constant
source across a stagnant layer to a membrane through which the fatty acid permeates and is
consumed on the other side. The results portrayed in Fig. 5-20 show diffusional transients in three
cases, all leading to the same steady state. In Case 1, left column, the stagnant layer, here given as
L = 50 µm thick, contains none of the three reacting species, fatty acid, albumin, or the fatty
acid-albumin complex, so that all three must diffuse in from the left boundary. In Case 2, all three
were in equilibrium in the stagnant layer 0 < x < L at t < 0, and at t = 0 the permeability P of the
membrane was switched from zero to its finite value, P1. In Case 3, albumin, without any fatty

82 Diffusion
acid, was uniformly distributed throughout the stagnant layer and at t = 0 the fatty acid was
introduced at the left boundary so that the fatty acid was reacting with the albumin as it diffused.
In all cases, the arrows indicate the sequence of concentration profiles at a succession of
times, from 0.1 s to 100 s, by which time a steady state was reached in all cases. The steady states
are identical in all columns, though the scales differ from column to column.
Case 1 (Left column): With the region being empty prior to the entry of S, B, and SB from
the source region at x < 0, the diffusion and the binding of S and B and the unbinding of S from B
occur simultaneously. Since S has a much higher diffusion coefficient than B it invades the empty
region more quickly. At x = L the solute S permeates the membrane, so that even at 100 seconds
the concentration s at x = L is lower than at the source at x = 0. The diffusion fronts for B and SB
(middle and bottom panels of Fig. 5-20), but their steady-state positions are notably different: the
concentration sb remains lower at x = L than at x = 0, such that sb(L) / sb(x = 0) = 0.984, because
of the steady loss of S across the membrane. The corollary is that the concentration of free binding
sites, b, must be higher at x = L than at the source, being about 10% higher. The reason that free B
can be 10% higher while SB is only 1.6% lower than at the source is that the Kd is lower than the
free ambient concentration of S, so most B is bound.
Case 2 (Middle column): After pre-equilibration throughout the region, the sudden change
in P allowing S to leave the region through the membrane at x = L diminishes its local
concentration to the same steady-state value as in the left column. Likewise the concentration sb
of the complex diminishes (middle panel), reaching the same steady-state value of sb / sb0 =
0.984. And b rises. Note that significant changes occurred within 0.1 s even though the
permeability is not very high.
Higher permeabilities drag the concentration s down to zero at the membrane if there is no
return flux across the membrane.
Case 3 (Right column): Having unfilled binding sites distributed across the region 0 < x < L
already at t = 0 retards the diffusion front for S dramatically compared to that in Case 1 with the
region empty. At 10 s in Case 3 the profile is not much ahead of that at 0.1 s in Case 1, due simply
to the fact that most of the diffusing S is captured by binding to B. This is seen by the profiles of sb
being farther to the right of those of free s at the same times, since these sites fill before there is
much free S to diffuse to the right. The concentration of B relative to that in equilibration with S at
the source is initially about 10-fold and then diminishes to the same steady-state values as in the
other cases.
To summarize the conditions under which front steepening occurs: (1) Ds >> Dsb so that
retardation occurs when S is bound, (2) concentrations of s < kb at the leading part of the
advancing front but s > kb near to the source, and (3) the concentration of binding sites bT should
be of the same order as the concentration s so that a substantial fraction of S is bound, and (4) the
binding/unbinding transformation fluxes should be fast compared to the diffusive fluxes so that
the substrate is captured before it goes by.
5-7. Descriptions of biological situations involving these phenomena

5-7.1. Hemoglobin facilition of oxygen transport
The three cases in Fig. 5-20 represent a variety of in vivo situations. The third is the
commonest: solute enters a region with unfilled binding sites. For example, with sudden
oxygenation of RBC previously depleted of oxygen, the diffusion of the first oxygen to enter is
retarded by the binding to reduced hemoglobin HHb, forming oxyhemoglobin, Hb(O2)4,

Diffusion 83
Case 1: Empty region Case 2: Pre-equilibrated Case 3: B only in region

1 10 s 1.0 1 at t < 0
ss at t = 100
ss
s 1s ss
----
s0
0.1 s P → P1 at t = 0
0 0.7 0
1 100 s = ss 1.00 1 0.984
10 s
sb
--------
sb 0
1s
0.1 s
0 0.96 0
1.5 1.2 10
ss
1.0 1.1
b
----- 5
b0
1.0
1
0 0.9 0
0 1 0 1 0 1
x/L x/L x/L
Figure 5-20: Diffusion front formation at successions of times: diffusion of a small

solute into or out of a solution with mobile binding sites. Concentrations are plotted
versus position, x/L, at four times after starting, at t = 0.1, 1.0, 10 and 100 seconds, the
succession of times being indicated by the arrows. For all panels Ds = 5 × 10−6 cm2s−1,
Db = Dsb = 9.35−8 cm2s−1, kon = 4.73 × 109 s−1M-1, koff = 0.142 s−1, (so that Kd = 3 ×
10−11 M) and at the source, x < 0, bT(total) = 6 × 10−7 M, s(total) = 5.4 × 10−7 M. At
x = L the membrane permeability, P, was 0.0083 cm s−1. Left Panel, Case 1: Solute and
binding protein diffuse in together into a solution containing neither at t = 0. Middle
panel, Case 2: A stagnant region contains equilibrated solutes and binding sites at t = 0.
Membrane permeability was zero at x = L until t = 0, then P ⋅ Am = 1 ml ⋅ s−1, as it was
in the other panels. Right panel, Case 3: Binding protein B distributed uniformly over
0 < x < L at t < 0. At t = 0, solute S was added at the left boundary. Note the steepness of
the profiles in s at early times. In all cases: Steady state was reached by t = 100 s. The
concentrations at x/L = 1 were s / s0 = ??, sb/sb0 = 0.984, and b/b0 = ??.[Eric, take these
from the plot files, the last points for the series at t= 100 sec]

84 Diffusion
preventing its penetration into the interior of the cell. Oxygen entering later encounters
hemoglobin Hb(O2)4 with filled binding sites just inside the surface layer, and so diffuses quickly
as the free O2, until it encounters HHb with empty binding sites. This causes “front steepening”,
the creation of a much steeper diffusion front than occurs with simple diffusion into an empty
region, as is seen in the right panel of Fig. 5-20. The result is that the front advances almost as a
square wave, with the bottom of the front moving at a speed lowered toward that of HHb
diffusion, but the top of the front almost catches up to the bottom because in highly saturated
Hb(O2)4, there is much free O2 whose diffusion is not retarded.
5-7.2. Calcium diffusion and exxcitation-contraction coupling:

Another kinetically relevant diffusional situation is the entry of calcium into myofilament
bundles after release from the sarcoplasmic reticulum, SR. In muscle fibers the contractile
proteins are arrayed in cylindrical bundles a few microns in diameter. These bundles are
surrounded by the network of the calcium-storing sarcoplasmic reticulum, SR. With each
electrical excitation of the cell, calcium enters the cell at the T-tubular-to-SR junctional region,
triggering the ryanodine-sensitive calcium release channels in the SR to open and to flood the
neighboring cytoplasm with free calcium, which then diffuses into the myofilament bundles.
Within the bundles the Ca2+ binds with high affinity to the thin filament protein troponin C,
retarding Ca2+ diffusion and preventing deeper penetration until the troponin binding sites are
filled. Since troponin-Ca2+ initiates a sequence of events leading to actin-myosin binding and
contraction, the outer filaments of actin and myosin can contract before Ca2+ reaches the inner
fibers because the Ca2+ front is so steep. The result observed by Taylor and Rudel (1970) is that
the outer parts of the bundles shorten, reducing sarcomere length by increasing overlap between
thick and thin filaments, while the myofilaments in the inner part of the bundle become wavy,
compressed as it were, from end to end before undergoing any shortening themselves. Thus the
steep radial diffusion front actually resulted in a reduction in efficiency of contraction under the
conditions of these particular observations. The authors (Taylor and Rudel, 1970; Costantin and
Taylor, 1973) originally attributed the appearance of waviness at short sarcomere lengths to an
“inactivation process” occurring selectively in the central part of these large cells of just over
100 µm diameter.
Steep diffusion fronts play a role in many phenomena where there is spread in 2 or 3
dimensions. They exist in oscillating chemical systems such as the Belousov-Zhabitinsky reaction
where reactants are removed by chemical reaction, giving sharp concentration profiles. Calcium
waves inside cells are closely related. Whether or not signaling cascades, with their high
amplification of product formation rates, result in intracellular waves of reactants can only be
surmised at this point, but the situation lends itself to steep diffusion fronts whenever there is
release of substances which must subsequently be bound or reacted to produce their effect.
5-7.3. Diffusion of calcium through tissues with binding sites :

Use fig from Safford and JB 1977 bioph j #137 fig 1 and fig 7
This is a situation analogous to one with fixed binding sites. Given a Barrer-type experiment, as in
Fig. 5-6, instead of examining intratissue concentration profiles one has only the rising time
course of concentrations in the unlabeled chamber, CR(t), for estimating the parameters of the
diffusional system. Tissue sheets are usually non-uniform; the equations need therefore account
for: heterogeneity of path lengths, diffusion into sequestered regions, volumes or binding affinities

Diffusion 85
Figure 5-21: Delayed activation of central myofilaments of a myofilament bundle in

frog skeletal muscle, Rana temporaria, causing waviness of the central fibrils. (From
Figure 2 of SR Taylor, 1974, reprinted from J. Physiol. with permission.)
in those regions, and both intracellular and extracellular space. Following the derivation from
Goodknight and Fatt (1961 #2251) for oil shales.........??
The key to the experiments and analysis is the simultaaneous use of a set of tracers that give
complimentary information: e.g. a neutral solute that undergoes no binding and does not enter
cells, e.g. sucrose, a solute that enters the total water space and is not bound, and the solute of
interest which is being tested for binding and the affinity of any binding sites.
5-7.4. Transport of organelles along nerve axons by motor proteins plus diffusion:
Motor proteins move large particles, organelles and vesicles, from the body of the neuron toward
the end of the nerve up to a meter away, a remarkable process and a long enduring mystery. Now
we know that similar events occur in all cells. There is directionality to it; most kinesins move
items centrifugally, dynein moves them centripetally. ATP is used to drive the motion, one ATP
per step, along a microtubule. A microtubule is a multistrand fibre where each strand is a tubulin
polymer; a thirteen strand microtubule is about 25 nm diameter, and the tubulin α/β dimers are 8
nm units. The microtubules are up to 500 µm long, rather stiff, and highly labile, growing and
shrinking dynamically with a half life of about 10 minutes. They have direction, the “minus end”
being nearer the cell center or neuronal body, the “plus” end going toward the cell membrane or
the axonal terminal synapse. The plus end grows at about three times the rate of the minus end.
The general process of translocation along the microtubule is: (1) a kinesin with a special
affinity for a particular type of vesicle or organelle binds to it; (2) the kinesin binds to the tubulin,
thus linking the cargo to the conveyor belt; hydrolysis of one ATP to ADP moves the kinesin and
cargo along by 1 dimer, 8 nm distance; (3) the free particle can diffuse randomly; (4) in cells,
unless the microtubules are stabilized by a drug from the Pacific Yew tree, taxol, they
disassemble, losing dimers from the ends of the polymers, with an average lifetime of 10 minutes,
so that the kinesin and its cargo are turned lose to diffuse again. Friedman and Craciun (2005)
define a model for this process in an axon considering 3-dimensional diffusion.
Problem: Write a set of partial differential equations for axonal transport of vesicles of
acetylcholine being carried to a synaptic junction at the end of a motor nerve. Given a free
diffusion coefficient for vesicles of 0.1 µm2 s-1 , a rate of kinesin binding to the vesicle of 2 s-1 and

86 Diffusion
a release rate of 0.5 s-1 , a rate of cargo-loaded kinesin binding to the tubulin microfilament of 1
s-1 and a release rate of 0.1 s-1 , and a forward velocity of 0.5 µm s-1 for kinesin along the
microtubule: (1) How many ATPs are used per µm? (2) Given a square wave labeling of vesicles
from x = 0 to x = 100 µm at t = 0, what are the concentration profiles for bound and for free
vesicles at 1, 10 and 100 seconds? After 10 seconds, what fraction of vesicles are linked to
microtubules and what fraction are diffusing, with or without kinesin attached? (4) How much
ATP was used by 100 seconds? (5) Given that no taxol was used, what was the average number of
microtubules encountered by a vesicle being carried 100 µm?
5-8. Chapter summary

Diffusion is due to random thermal molecular motion. It is enhanced in bulk media by raising
temperature and lowering fluid viscosities. Frictional forces retard diffusion; these are larger for
larger molecules, in higher viscosity solutions, and in hindering media such as pores or gel
matrices.
Transmembrane diffusion of non-electrolytes. To penetrate cell membranes, hydrophilic
molecules must be small. In general they don’t get across without either facilitating transporters
or specialized ion-selective channels. Even for water there is a specialized transmembrane protein
facilitating its flux: aquaporin. However, L-glucose (3.6 Å radius) and other inert
monosaccharides can slowly enter cells. The cell substrate D-glucose enters much faster, but via a
facilitating transporter. Large molecules must be lipid soluble to penetrate. Fastest penetration
occurs for molecules which are both small and lipid soluble (CO2). There is much variation with
cell type and species.
Transmembrane diffusion of electrolytes. Electroneutrality must be maintained. A positive ion
will not pass through a membrane without a negative ion, unless it exchanges for another positive
ion. Membranes are usually charged, and pore edges may repel an ion or accelerate it through the
pore. Transmembrane potential differences provide another driving force. See Ch.7.
Diffusion with binding. Diffusion in the presence of binding sites is complicated, depending on
the concentrations of sites and on their mobility. With high fractional binding, diffusional
facilitation can occur even when the sites are less mobile than the free solute, as with hemoglobin
facilitation of oxygen diffusion. Immobile binding sites tend to cause “front steepening” and
retardation of solute penetration. The binding sites act as buffers of solute concentration and
enlarge the effective volumes of distribution of solute.
5-9. Problems
1. Given the composition of air as 21% O2 and 79% N, calculate the density of dry air at the
freezing point of water and 1 atmospheric pressure (760 mmHg), that is, STPD, standard
temperature and pressure, dry. (Hint: Use molecular weights, and the volume of 1 mole of gas at
STPD, 22.4 liters.)
2. Using the viscosity of air from a reference source (please provide), estimate a diffusion
coefficient for water in air and for methyl salicylate or levomenthone in air.
3. A solute has an apparent permeability of 5 cm/s in the RBC membrane. Estimate the
effective diffusion coefficient in the lipid portion of the bilayer, assuming that it is the impeding
part of the barrier to exchange. How do you interpret the result?

Diffusion 87
4. The diffusion coefficient for a solute in a thick membrane in a dialyser system is

0.1 × 10−6 cm2s−1. Given a membrane thickness of 100 microns and a partition coefficient of 5 for
the solute in the membrane compared to the plasma, what is the membrane permeability in cm/s?
5. Diffusion of solute from a source into a medium containing binding sites results in steep
diffusion fronts under some circumstances. Define the circumstances. Write equations for the
system. Sketch a relationship between steepness of the diffusion front and each of some
parameters which govern the steepness.
5-10. Further readings

Classic textbooks are those of Crank (1975), or Carslaw and Jaeger (1959), particularly for
heat diffusion, of Jost (1960), and of Bird, Stewart and Lightfoot (1960, 2001). These have no
coverage of biology but describe a wide variety of physical situations. The Bird, Stewart and
Lightfoot texts are particularly good on deriving the ideas from first principles.Cussler
More biologically oriented are works by Fall et al. (2002) and James D. Murray (1990).
Movement in hindered media are discussed by Comper in two volumes (1996a, 1996b). Porous
transport, a specific type of hindering, will be covered in Ch. 6. Combined convection and
diffusion will be brought up in several later chapters.
5-11. References
Agre P, Preston GM, Smith BL, Jung JS, Raina S, Moon C, Guggino WB, and Nielsen S.
Aquaporin CHIP: the archetypal molecular water channel. Am J Physiol Renal Physiol 265:
F463-F476, 1993.
Barrer RM. A new approach to gas flow in capillary systems. J Phys Chem 57: 35-40, 1953.
Barta E, Sideman S, and Bassingthwaighte JB. Facilitated diffusion and membrane permeation of
fatty acid in albumin solutions. Ann Biomed Eng 28: 331-345, 2000.
Bean CP. The physics of porous membranes: neutral pores. Membranes: Macroscopic Systems
and Models , edited by Einsenman G. New York: Dekker, 1972, p. 1-54.
Comper WD (Editors). Extracellular Matrix, Volume 1 Amsterdam: Harwood Academic
Publishers, 1996, 476 pp.
Comper WD (Editors). Extracellular Matrix, Volume 2 Amsterdam: Harwood Academic
Publishers, 1996, 386 pp.
Costantin LL and Taylor SR. Graded activation in frog muscle fibers. J Gen Physiol 61: 424-443,
1973.
Crank J. The Mathematics of Diffusion, 2nd edition Oxford: Clarendon Press, 1975.
Curry FE. Mechanics and thermodynamics of transcapillary exchange.. Handbook of Physiology,
Sec. 2 The Cardiovascular System. Vol. 4, Microcirculation , edited by Renkin EM and
Michel CC. American Physiological Society: Bethesda, Maryland, 1984, p. 309-374.
Cussler, Edward L. Diffusion. 2nd Edition.Cambridge University Press. 1997 (not in biblio)
Deen WM. Hindered transport of large molecules in liquid-filled pores. AIChE J 33: 1409-1425,
1987.
Cussler, Edward L. Diffusion. 2nd Edition.Cambridge University Press. 1997 (not in biblio)
Maxwell, J. C. A treatise on electricity and magnetism. Volume 1 Oxford UK Clarendon Press,
1873
Nugent LJ and Jain RK. Pore and fiber-matrix models for diffusive transport in normal and
neoplastic tissues. Microvasc Res 28: 270-274, 1984.

88 Diffusion
Patlak CS and Fenstermacher JD. Measurement of dog blood-brain transfer constants by

ventriculocisternal perfusion. Am J Physiol 229: 877-884, 1975.
Redwood WR, Rall E, and Perl W. Red cell membrane permeability deduced from bulk diffusion
coefficients. J Gen Physiol 64: 706-729, 1974.
Safford RE and Bassingthwaighte JB. Calcium diffusion in transient and steady states in muscle.
Biophys J 20: 113-136, 1977.
Safford RE, Bassingthwaighte EA, and Bassingthwaighte JB. Diffusion of water in cat ventricular
myocardium. J Gen Physiol 72: 513-538, 1978.
Saxby, 1923?? Need to find this
Schnitzer, J. E. Transport functions of the glycocalyx, specific proteins, and caveolae in
endothelium. In Whole Organ Approaches to Cellular Metabolism. Ed. by Bassingthwaighte,
Goresky, and Linehan. Springer 1998, pp31-69.
Suenson, M., D.R. Richmond, and J. B. Bassingthwaighte. Diffusion of sucrose, sodium and
water in ventricular myocardium. Am.J. Physiol. 227: 1116-1123, 1974
Taylor SR and Rudel R. Striated muscle fibers: inactivation of contraction induced by shortening.
Science 167: 882-884, 1970.
Taylor SR. Decreased activation in skeletal muscle fibres at short lengths. The Physiological Basis
of Starling’s Law of the Heart. Foundation Symposium 24 . Amsterdam: Elsevier, p. 93-116.,
1974.
Tong J and Anderson JL. Partitioning and diffusion of proteins and linear polymers in
polyacrylamide gels. Biophys J 70: 1505-1513, 1996.
5. Diffusion and diffusion coefficients

5.1. The process of diffusion: thermal motion in viscous media
5.1.1. Measuring diffusion coefficients; values of D in water and gas
5.1.2. Diffusional relaxation times. L2/D
5.1.3. Reflecting boundaries and other geometric considerations
5.1.4. Effects of consumption or clearance on spatial profiles
5.1.5. Diffusion in hindering media (fibre matrix, porous media)
5.1.5.1. Extracellular diffusion (sucrose, sodium, dextrans)
5.1.5.2. Intracellular diffusion
5.1.5.3. Diffusion through cylindrical pores and rectangular clefts
5.1.5.4. Diffusion via cellular and paracellular paths (Safford water vs. Redwood et al.)
5.2. Diffusion in presence of binding sites
5.2.1. Facilitated diffusion
5.2.1.1. Facilitated diffusion with equilibrium binding to mobile proteins (Hb/O2, etc.)
5.2.1.2. Facilitated diffusion with slow binding to mobile proteins (fatty acid-albumin)
5.2.2. Hindered diffusion due to fixed binding sites (calcium, dead-end pores, immobile proteins)
5.3. Diffusive Permeation
5.3.1. Solute flux across membranes
5.3.2. Transmembrane concentration profiles
5.3.3. Effect of solubility in membrane on intramembrane concentration profiles
1. Diffusion of solute and uniform medium containing binding sites

A. Immobile binding site (DB = DSB = 0).
Situation E1. Transient for solute S.
Situation E2. Tracer *S diffusion with s and sb constant.
B. Mobile binding site (DB = DSB > 0).
Situation E1. Transients for S with binding site B mobile.

Diffusion 89
Situation E2. Tracer *S transient with s and sb constant.

C. Effect of slow binding kinetics on diffusional facilitation.
2. Path length heterogeneity.
3. Dead-end pores.
4. Parallel paths:
A. Non-interacting paths.
Interexchanging pathways: Cells uniformly dispersed in ISF.
ADD:
diff coeff in water and air and solvent viscosities
boundary layers, as e.g. in BSL p788
Weisiger intrahepatocyte diffusion of FA by flash photobleaching

90 Diffusion

05diffsep07 PDF

Uploaded by

Document Informationclick to expand document information

Copyright:

Available Formats

05diffsep07 PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

05diffsep07 PDF

Uploaded by

Copyright:

Available Formats

Chapter 5

5-1. Diffusion: thermal motion in viscous media

Viscosity: Kinematic Viscosity Viscosity: Kinematic Viscosity

07 September 2007, 5:28 pm /userA/jbb/writing/903/2005/05diff7sep07.fm

5-1.1. Brownian movement and diffusion

5-1.2. Diffusion coefficients

/userA/jbb/writing/903/2005/05diff7sep07.fm 07 September 2007, 5:28 pm

07 September 2007, 5:28 pm /userA/jbb/writing/903/2005/05diff7sep07.fm

Table 5-2: Diffusion Coefficients in Water, Dw (in aqueous solution at 20°C)

5-1.3. Influences of molecular shape on diffusion coefficients

/userA/jbb/writing/903/2005/05diff7sep07.fm 07 September 2007, 5:28 pm

Table 5-3: Frictional coefficientsa

Shape Frictional Coefficient Explanation

Sphere f = 6πηR R = sphere radius

07 September 2007, 5:28 pm /userA/jbb/writing/903/2005/05diff7sep07.fm

5-2. Fick’s “Laws”

/userA/jbb/writing/903/2005/05diff7sep07.fm 07 September 2007, 5:28 pm

This defines D, the diffusion coefficient, as

07 September 2007, 5:28 pm /userA/jbb/writing/903/2005/05diff7sep07.fm

5-2.2. Fick’s Second Law of diffusion

Over a small finite distance x2 − x1 the rate of concentration change is approximated by

5-3. Measuring diffusion coefficients

/userA/jbb/writing/903/2005/05diff7sep07.fm 07 September 2007, 5:28 pm

07 September 2007, 5:28 pm /userA/jbb/writing/903/2005/05diff7sep07.fm

5-3.2. Diffusion across an interface between unstirred regions

The standard deviation, SD, of ∂C/∂x is 2 Dt and its height, ∂C at x = 0, is Co/2(πDt)1/2 or

This provides a direct estimate of D = (slope, (area/height)2/dt)/4π.

/userA/jbb/writing/903/2005/05diff7sep07.fm 07 September 2007, 5:28 pm

Figure 5-8: Determination of diffusion coefficient from plot of (area/height)2 of ∂C/∂x

5-4. Applications to biological situations

λ = C membrane ⁄ C medium . (5-19)

07 September 2007, 5:28 pm /userA/jbb/writing/903/2005/05diff7sep07.fm

J s = – P∆C where P = Dm λ ⁄ l . (5-21)

Figure 5-9: Diffusion across a nonporous membrane. C1 and C2 are concentrations in

5-4.2. Diffusion through pores

/userA/jbb/writing/903/2005/05diff7sep07.fm 07 September 2007, 5:28 pm

5-4.3. Diffusion across walls and pores in parallel

07 September 2007, 5:28 pm /userA/jbb/writing/903/2005/05diff7sep07.fm

J s = J pores + J membrane (5-23a)

= – ( P pore + P membrane ) ∆C , (5-23b)

/userA/jbb/writing/903/2005/05diff7sep07.fm 07 September 2007, 5:28 pm

07 September 2007, 5:28 pm /userA/jbb/writing/903/2005/05diff7sep07.fm

5-5. Diffusion in Heterogeneous Media

/userA/jbb/writing/903/2005/05diff7sep07.fm 07 September 2007, 5:28 pm

Figure 5-13: Steady-state profiles of concentration with zero-order consumption in a

07 September 2007, 5:28 pm /userA/jbb/writing/903/2005/05diff7sep07.fm

/userA/jbb/writing/903/2005/05diff7sep07.fm 07 September 2007, 5:28 pm

07 September 2007, 5:28 pm /userA/jbb/writing/903/2005/05diff7sep07.fm

/userA/jbb/writing/903/2005/05diff7sep07.fm 07 September 2007, 5:28 pm

5-5.2. Diffusion across an uneven slab of tissue:

5-5.3. Diffusion through tissues with dead-end pores:

5-5.4. Diffusion through a hindering matrix

5-6. Diffusion of solutes which can be bound to absorbing sites

07 September 2007, 5:28 pm /userA/jbb/writing/903/2005/05diff7sep07.fm

5-6.1. General aspects of diffusion of solute in the presence of binding sites

/userA/jbb/writing/903/2005/05diff7sep07.fm 07 September 2007, 5:28 pm

5-6.2. Diffusion in the presence of immobile binding sites (DB = DSB = 0)

Substituting Eq. 5-40 into Eq. 5-38 gives: