Introduction - Wave Equation

Age-Structured Model
Model for Erythropoiesis

Math 531 - Partial Differential Equations

Method of Characteristics for PDEs

Joseph M. Mahaffy,
[email protected]

Department of Mathematics and Statistics

Dynamical Systems Group
Computational Sciences Research Center
San Diego State University
San Diego, CA 92182-7720
http://jmahaffy.sdsu.edu

Spring 2020

Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (1/54)

 Introduction - Wave Equation
Age-Structured Model
Model for Erythropoiesis

Outline

1 Introduction - Wave Equation

Characteristics - Wave Equation
Example
D’Alembert’s Solution

2 Age-Structured Model
Method of Characteristics
Birth Boundary Condition
Example

3 Model for Erythropoiesis

Model for Erythropoiesis
Method of Characteristics
Model for Erythropoiesis with Delays

Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (2/54)

 Introduction - Wave Equation Characteristics - Wave Equation
Age-Structured Model Example
Model for Erythropoiesis D’Alembert’s Solution

Introduction - Wave Equation

The one-dimensional Wave Equation satisfies:

∂2u ∂2u
= c2 2 ,
∂t2 ∂x

with ICs
u(x, 0) = f (x) and ut (x, 0) = g(x).

We saw that when u(0, t) = 0 = u(L, t) (fixed ends), the Fourier series solution was

∞
X
nπx nπct nπct




u(x, t) = sin L
an cos L
+ bn sin L
.
n=1

This can be shown to be the sum of forward and backward moving waves:

f (x − ct) + f (x + ct) 1
Z x+ct
u(x, t) = + g(s) ds,
2 2c x−ct

where f (x) and g(x) are odd periodic extensions of the ICs.
Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (3/54)
 Introduction - Wave Equation Characteristics - Wave Equation
Age-Structured Model Example
Model for Erythropoiesis D’Alembert’s Solution

Introduction - Wave Equation

We also demonstrated with Fourier transforms through the use of

Euler’s formula that the Wave equation on an infinite domain
starting at rest satisfies:
1
u(x, t) = 2 [f (x − ct) + f (x + ct)] ,

which again leads to waves traveling to the right and left with speed c.

The method of characteristics is introduced to solve the

one-dimensional Wave equation in greater generality.
By moving along a “characteristic” with speed c, the PDE is
reduced to an ODE and gives the solution

u(x, t) = F (x − ct) + G(x + ct).

Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (4/54)

 Introduction - Wave Equation Characteristics - Wave Equation
Age-Structured Model Example
Model for Erythropoiesis D’Alembert’s Solution

Characteristics - Wave Equation

The one-dimensional Wave Equation is given by:
∂2u ∂2u
2
− c2 2 = 0,
∂t ∂x
which can be written in “factored” form:
  
∂ ∂ ∂u ∂u
+c −c =0
∂t ∂x ∂t ∂x
or   
∂ ∂ ∂u ∂u
−c +c = 0.
∂t ∂x ∂t ∂x

Let w = ∂u ∂u
∂t − c ∂x and v =
∂u
∂t + c ∂u
∂x , then the 1
st
order wave
equations are given by:
∂w ∂w ∂v ∂v
+c =0 and −c = 0.
∂t ∂x ∂t ∂x
Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (5/54)
 Introduction - Wave Equation Characteristics - Wave Equation
Age-Structured Model Example
Model for Erythropoiesis D’Alembert’s Solution

Characteristics - Wave Equation

Consider the 1st order wave equation:
∂w ∂w
+c = 0.
∂t ∂x
We examine the behavior of this equation from the perspective of a
moving observer, x(t), so
d ∂w dx ∂w
w(x(t), t) = + .
dt ∂t dt ∂x
dx
If dt = c, then the observer sees
dw
= 0,
dt
so the solution w(x(t), t) is constant.
Thus, the observer sees no change in the solution if the observer is
moving with constant speed c, so
x(t) = ct + x0 .
Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (6/54)
 Introduction - Wave Equation Characteristics - Wave Equation
Age-Structured Model Example
Model for Erythropoiesis D’Alembert’s Solution

Characteristics - Wave Equation

Along the characteristic x(t),
w(x, t) is constant.
w propagates as a wave with
wave speed c.

General Solution:
If w(x, t) is given at t = 0

w(x, 0) = P (x),
w(x, t) = w(x0 , 0) = P (x0 ),

but x0 = x − ct, so
w(x, t) = P (x − ct).

This is a wave traveling to the right with speed c and maintaining its
shape.
Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (7/54)
 Introduction - Wave Equation Characteristics - Wave Equation
Age-Structured Model Example
Model for Erythropoiesis D’Alembert’s Solution

Characteristics - Wave Equation

The observer moves along the

red line with constant
speed c.
The observer sees no change
in the shape of the wave
as time progresses.
The solution is

w(x, t) = P (x − ct).

The full wave equation

utt = c2 uxx
starts with an initial shape, then half moves to the right with speed c
and half moves to the left with speed c.

Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (8/54)

 Introduction - Wave Equation Characteristics - Wave Equation
Age-Structured Model Example
Model for Erythropoiesis D’Alembert’s Solution

Characteristics - Example
Example: Consider the first order PDE given by:
∂w ∂w
+ 2t = cos(t)w,
∂t ∂x
with the initial condition
2
w(x, 0) = P (x) = e−0.05x .

From the method of characteristics, we can reduce the PDE to an

ODE provided
dx
= 2t, or x(t) = t2 + x0 .
dt
The ODE is
dw
= cos(t)w,
dt
which has a solution along the characteristic, x(t) = t2 + x0 ,
w(x(t), t) = kesin(t) .
Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (9/54)
 Introduction - Wave Equation Characteristics - Wave Equation
Age-Structured Model Example
Model for Erythropoiesis D’Alembert’s Solution

Characteristics - Example
2
From the initial condition, w(x0 , 0) = k = P (x0 ) = e−0.05x0 .
However, P (x0 ) = P (x − t2 ), so it follows that the solution satisfies:

2 2
w(x, t) = P (x0 )esin(t) = P (x − t2 )esin(t) = e−0.05(x−t )
esin(t) ,

which is shown below.

Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (10/54)

 Introduction - Wave Equation Characteristics - Wave Equation
Age-Structured Model Example
Model for Erythropoiesis D’Alembert’s Solution

D’Alembert’s Solution - Wave Equation

The solutions above suggest that the natural variables are not t and
x, but a change moving along the characteristics would be better.
Let ξ = x + ct and η = x − ct, which gives
ξ+η ξ−η
x= 2 and t= 2c .

By the chain rule, we have

∂ ∂ ∂x ∂ ∂t 1 ∂ ∂


∂ξ = ∂x ∂ξ + ∂t ∂ξ = 2c ∂t + c ∂x
and
∂ ∂ ∂x ∂ ∂t 1 ∂ ∂


∂η = ∂x ∂η + ∂t ∂η = − 2c ∂t − c ∂x .

From the partials above, we have

2
∂ u ∂ ∂ ∂ ∂ ∂2u 2
− c2 ∂∂xu2 = 0.



−4c2 ∂ξ∂η = ∂t + c ∂x ∂t − c ∂x u= ∂t2

Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (11/54)

 Introduction - Wave Equation Characteristics - Wave Equation
Age-Structured Model Example
Model for Erythropoiesis D’Alembert’s Solution

D’Alembert’s Solution - Wave Equation

It follows that
∂2u
∂ξ∂η (ξ, η) = 0.

Integrating with respect to ξ implies that

∂u
∂η (ξ, η) = φ(η),

which can be integrated with respect to η to give

Z
u(ξ, η) = φ(η) dη + ψ(ξ) = φ(η) + ψ(ξ).

This is D’Alembert’s solution for the wave equation.

Changing back the variables gives the form we have seen before

u(x, t) = F (x − ct) + G(x + ct).

Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (12/54)

 Introduction - Wave Equation Characteristics - Wave Equation
Age-Structured Model Example
Model for Erythropoiesis D’Alembert’s Solution

D’Alembert’s Solution - Wave Equation

Recall the ICs were

u(x, 0) = f (x) and ut (x, 0) = g(x),

so our solution u(x, t) = F (x − ct) + G(x + ct) satisfies:

u(x, 0) = F (x) + G(x) = f (x)

and
ut (x, 0) = −cF 0 (x) + cG 0 (x) = g(x).

This latter condition can be integrated to give

Z x
−cF (x) + cG(x) = g(s) ds + A.
0

Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (13/54)

 Introduction - Wave Equation Characteristics - Wave Equation
Age-Structured Model Example
Model for Erythropoiesis D’Alembert’s Solution

D’Alembert’s Solution - Wave Equation

Solving the system of equations above gives:

Z x 
1 1
F (x) = 2 f (x) − 2c g(s) ds + A
0

and Z x 
1 1
G(x) = 2 f (x) + 2c g(s) ds + A .
0

It follows that
Z x+ct
1 1
u(x, t) = 2 [f (x − ct) + f (x + ct)] + 2c g(s) ds.
x−ct

This gives the complete D’Alembert’s solution for the wave

equation for x ∈ (−∞, ∞).

Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (14/54)

 Introduction - Wave Equation Method of Characteristics
Age-Structured Model Birth Boundary Condition
Model for Erythropoiesis Example

Age-Structured Model
Age-Structured Model: Modeling with a hyperbolic PDE.

Mathematical modeling of populations often needs information

about the ages of the individuals in the population.
This modeling approach was developed primarily by McKendrick
(1926) and Von Foerster (1959).
Key Elements in Model
Let n(t, a) denote the population at time t and age a.
The birth rate of individuals b(a) depends on the age of
the adult population.
Similarly, the death rate of individuals µ(a) depends on
the age of the individuals.
Must specify the initial age distribution of the
population, f (a).

Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (15/54)

 Introduction - Wave Equation Method of Characteristics
Age-Structured Model Birth Boundary Condition
Model for Erythropoiesis Example

Age-Structured Model
Age-Structured Model: The McKendrick-Von Foerster equation is:
∂n ∂n
+ + µ(a)n(t, a) = 0,
∂t ∂a
with the birth boundary condition (Malthusian):
Z ∞
n(t, 0) = b(a)n(t, a) da,
0

and the initial condition:

n(0, a) = f (a).
Birth Function Death Function

b(a) µ(a)

a a

Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (16/54)

 Introduction - Wave Equation Method of Characteristics
Age-Structured Model Birth Boundary Condition
Model for Erythropoiesis Example

Age-Structured Model
Discussion for the Age-Structured Model
∂n ∂n
+ = −µ(a)n(t, a).
∂t ∂a
The PDE shows that age advances with time.
The right side shows that there is only a loss of population
through death with death increasingly likely with age.
The birth function:
Young individuals are incapable of giving birth
The birth function increases to peak fertility.
Births are Malthusian - proportional to the population.
After peak fertility, reproductive ability decreases, and it
could again decrease to zero.
The initial population distribution could be anything
However, in general the population distribution should decrease
with increasing time.
Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (17/54)
 Introduction - Wave Equation Method of Characteristics
Age-Structured Model Birth Boundary Condition
Model for Erythropoiesis Example

Age-Structured Model - Method of Characteristics

The Age-Structured Model:

∂n ∂n
+ = −µ(a)n(t, a).
∂t ∂a
can be written as an ODE:
d
n(t, a) = −µ(a)n(t, a),
dt
along the characteristic,

a(t) = t + c.

This has the solution:

Rt
N (t) = N0 e− 0
µ(s) ds
,

which follows the population of a particular age cohort.

Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (18/54)
 Introduction - Wave Equation Method of Characteristics
Age-Structured Model Birth Boundary Condition
Model for Erythropoiesis Example

Age-Structured Model - Survival

We can define a survival function

Ra
L(a) = e− 0
µ(s) ds
,

which gives the fraction of individuals surviving from birth to age a.

The survival from a to b is given by
Rb
L(a, b) = e− a
µ(s) ds
.

From the diagram above, we follow the characteristics to obtain the

solution of the age-structured model:

a<t: n(t, a) = n(t − a, 0)L(0, a),

a>t: n(t, a) = n(0, a − t)L(a − t, a).

Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (19/54)

 Introduction - Wave Equation Method of Characteristics
Age-Structured Model Birth Boundary Condition
Model for Erythropoiesis Example

Age-Structured Model

The age-structured model gives the dynamics of a particular age

cohort following a characteristic.
The long term behavior depends significantly on the birth process
on the boundary.
Since this is a type of Malthusian growth (with no limiting
nonlinearities), we expect a type of exponential growth (or
decline) with some rate r and having the form:

n(t, a) = Cn∗ (a)ert ,

where n∗ (a) is the stable age distribution and C depends on the

initial conditions.
For convenience, assume n∗ (0) = 1, so that n∗ (a) is the fraction of age
a individuals surviving to age a relative to age 0.

Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (20/54)

 Introduction - Wave Equation Method of Characteristics
Age-Structured Model Birth Boundary Condition
Model for Erythropoiesis Example

Age-Structure Model - Birth Function

The boundary condition of births is

Z ∞
n(t, 0) = b(a)n(t − a, 0)L(a) da.
0

Inserting the assumed stable form, n(t, a) = Cn∗ (a)ert , gives

Z ∞
rt
Ce = b(a)Cer(t−a) L(a) da,
0
Z ∞
1 = e−ra L(a)b(a) da.
0

Whether r is positive or negative determines if the overall population

grows or decays.
If r > 0, then the total population grows like Cert

Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (21/54)

 Introduction - Wave Equation Method of Characteristics
Age-Structured Model Birth Boundary Condition
Model for Erythropoiesis Example

Age-Structure Model - R0

Ecologists and epidemiologists define an important constant R0 ,

which is used to determine if a population (or disease) expands or
contracts.
For this population, define
Z ∞
R0 = L(a)b(a) da,
0

where R0 represents the average number of (female) offspring from an

individual (female) over her lifetime (integral of births times lifespan).
Note that if R0 < 1, then r < 0 and if R0 > 1, then r > 0. The latter
condition indicates that each female during her lifetime must produce
more than one female offspring for the population to grow.
Since n(t, a) = n(t − a, 0)L(a), the stable age distribution satisfies
Cert n∗ (a) = Cer(t−a) n∗ (0)L(a) = Cer(t−a) L(a),
n∗ (a) = e−ra L(a).
Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (22/54)
 Introduction - Wave Equation Method of Characteristics
Age-Structured Model Birth Boundary Condition
Model for Erythropoiesis Example

Age-Structured Model - Example

We can define the average generation time, T , to satisfy:
erT = R0 ,
so on average a mother replaces herself with R0 offspring.
The value Z ∞
1
T = aL(a)b(a) da,
R0 0
gives the average age of reproduction.
Example: Let us examine the age-structured model
∂n ∂n
+ + µ(a)n(t, a) = 0,
∂t ∂a
with the birth boundary condition:
Z ∞
n(t, 0) = b(a)n(t, a) da.
0

Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (23/54)

 Introduction - Wave Equation Method of Characteristics
Age-Structured Model Birth Boundary Condition
Model for Erythropoiesis Example

Age-Structured Model
In order to perform calculations (with the help of Maple), we take birth and
death functions

0.3, 3 < a < 8,
b(a) = and µ(a) = 0.02 e0.25a .
0, otherwise,

The birth function assumes a constant fecundity of 0.3 between the ages of 3 and
8, while the death function assumes an ever increasing function with age.

Birth Function Death Function

1 1

0.9 0.9

0.8 0.8

0.7 0.7

0.6 0.6
µ(a)
0.5 0.5

0.4 b(a) 0.4

0.3 0.3

0.2 0.2

0.1 0.1

0 0
0 5 10 15 0 5 10 15
a a

Note: These functions are very crude approximations to the forms displayed
earlier.

Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (24/54)

 Introduction - Wave Equation Method of Characteristics
Age-Structured Model Birth Boundary Condition
Model for Erythropoiesis Example

Age-Structured Model
The age-structured model had a survival function
Ra 0.25a
L(a) = e− 0
µ(s) ds
= e−0.08(e −1)
,

which gives the fraction of individuals surviving from birth to age a.

Cohort Survival
1

0.9

0.8

0.7

0.6

0.5 L(a)
0.4

0.3

0.2

0.1

0
0 5 10 15 20
a

Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (25/54)

 Introduction - Wave Equation Method of Characteristics
Age-Structured Model Birth Boundary Condition
Model for Erythropoiesis Example

Age-Structured Model
The basic reproduction number, R0 , was given by
Z ∞ Z 8
0.25a
R0 = L(a)b(a) da = 0.3e−0.08(e −1)
da = 1.1678,
0 3

which is the average number of (female) offspring from an individual

(female) over her lifetime.
With the help of Maple, we can determine the average overall
growth rate, r, for this example.
Maple solves the equation for r:
Z ∞ Z 8
0.25a
1= e−ra L(a)b(a) da = 0.3e−ra e−0.08(e −1)
da,
0 3

and obtains
r = 0.02925985.

This shows the overall population is growing about 3% per unit time.
Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (26/54)
 Introduction - Wave Equation Method of Characteristics
Age-Structured Model Birth Boundary Condition
Model for Erythropoiesis Example

Age-Structured Model
The Malthusian growth would not be sustainable over long periods of time, so
nonlinear terms for crowding and other factors would need to be included in the
model, e.g., logistic growth.
With the overall population growth rate, we can obtain the steady-state age
distribution of this population:
0.25a
n∗ (a) = e−ra L(a) = e−0.02926a e−0.08(e −1)
.

Stable Age Distribution

1

0.9

0.8

0.7

0.6

0.5 e−ra L(a)

0.4

0.3

0.2

0.1

0
0 5 10 15 20
a
Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (27/54)
 Introduction - Wave Equation Method of Characteristics
Age-Structured Model Birth Boundary Condition
Model for Erythropoiesis Example

Age-Structured Model
The average generation time, T , satisfies:
erT = R0 or e0.02926T = 1.1678.
so on average a mother replaces herself with R0 offspring in
T = 5.3024 time units.
The value,
Z ∞ Z 8
1 1 0.25a
T = aL(a)b(a) da = 0.3a e−0.08(e −1)
da = 5.33205,
R0 0 1.1678 3
gives the average age of reproduction.
In summary, the method of characteristics allows solutions for the
age-structured model, which can provide interesting information
about the behavior of a population.
Needless to say, these models must be significantly expanded to
manage more realistic populations, which in turn significantly
complicates the mathematical analysis.
Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (28/54)
 Introduction - Wave Equation Model for Erythropoiesis
Age-Structured Model Method of Characteristics
Model for Erythropoiesis Model for Erythropoiesis with Delays

Erythropoiesis

Erythropoiesis is the process for producing Erythrocytes or Red

Blood Cells (RBCs).

RBCs are the most numerous cells that we produce in our

bodies, accounting for almost 85% by numbers.
Critical for carrying O2 to our other cells, using the protein
hemoglobin.
By volume, RBCs are about 40% of blood.

Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (29/54)

 Introduction - Wave Equation Model for Erythropoiesis
Age-Structured Model Method of Characteristics
Model for Erythropoiesis Model for Erythropoiesis with Delays

Erythropoiesis
Erythrocytes or red blood cells
RBCs are one of the most actively produced cells in our bodies.
RBCs begin from a group of undifferentiated stem cells
(multipotent progenitors).
The body senses O2 levels in the body and releases
erythropoietin (EPO) inversely to the O2 in the blood.
EPO stimulates commitment of stem cells to become RBCs
and proliferate.
Progenitor cells specialize though a series of cell divisions and
intracellular changes (taking about 6 days).
Erythropoietic cells shrink, even losing their nucleus, to become
reticulocytes then RBCs, which serve as vessels for hemoglobin.
RBCs circulate in the bloodstream for about 120 days, then are
actively degraded.
Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (30/54)
 Introduction - Wave Equation Model for Erythropoiesis
Age-Structured Model Method of Characteristics
Model for Erythropoiesis Model for Erythropoiesis with Delays

Diagram for Erythropoiesis

Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (31/54)

 Introduction - Wave Equation Model for Erythropoiesis
Age-Structured Model Method of Characteristics
Model for Erythropoiesis Model for Erythropoiesis with Delays

Model for Erythropoiesis

Important Elements in Model for Erythropoiesis

BFU-E and CFU-E differentiate and proliferate in response to

EPO
Maturation requires about 6 days
EPO accelerates maturation
Lack of EPO causes apoptosis
Cell divisions every 8 hours for about 4 days
Reticulocytes do not divide - increase hemoglobin
Erythrocytes lose nucleus - live 120 days
Macrophages actively degrade RBCs
EPO released near kidneys with half-life of 6 hours

Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (32/54)

 Introduction - Wave Equation Model for Erythropoiesis
Age-Structured Model Method of Characteristics
Model for Erythropoiesis Model for Erythropoiesis with Delays

Model for Erythropoiesis

Age-Structured Model for Erythropoiesis

Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (33/54)

 Introduction - Wave Equation Model for Erythropoiesis
Age-Structured Model Method of Characteristics
Model for Erythropoiesis Model for Erythropoiesis with Delays

Model for Erythropoiesis

Age-Structured Model viewed as a conveyor system

Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (34/54)

 Introduction - Wave Equation Model for Erythropoiesis
Age-Structured Model Method of Characteristics
Model for Erythropoiesis Model for Erythropoiesis with Delays

Active Degradation of RBCs

Active Degradation of RBCs

RBCs are lost from normal leakage (breaking capillaries), which

is simply proportional to the circulating numbers
RBCs age - Cell membrane breaks down (no nucleus to repair)
from squeezing through capillaries
Aged membrane is marked with antibodies
Macrophages destroy least pliable cells based on the antibody
markers
Model assumes constant supply macrophages
Saturated consumption of Erythrocytes
- Satiated predator eating a constant amount per unit time
Constant flux of RBCs being destroyed

Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (35/54)

 Introduction - Wave Equation Model for Erythropoiesis
Age-Structured Model Method of Characteristics
Model for Erythropoiesis Model for Erythropoiesis with Delays

Constant Flux Boundary Condition

Constant Flux Boundary Condition
Let Q be rate of removal of erythrocytes
Erythrocytes lost are Q∆t
Mean Value Theorem - average number RBCs
m(ξ, νF (ξ)) for ξ ∈ (t, t + ∆t)

Balance law

Q∆t = W ∆t m(ξ, νF (ξ))

−[νF (t + ∆t) − νF (t)]m(ξ, νF (ξ))

As ∆t → 0,
Q = [W − ν̇F (t)]m(t, νF (t))

Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (36/54)

 Introduction - Wave Equation Model for Erythropoiesis
Age-Structured Model Method of Characteristics
Model for Erythropoiesis Model for Erythropoiesis with Delays

Constant Flux Boundary Condition

If macrophages consume a constant amount of RBCs at the end

of their, we obtain the natural BC

Q = [W − ν̇F (t)]m(t, νF (t))

This results in the lifespan of the RBCs either lengthening or

shortening from the normal 120 days
This implies that the lifespan of the RBCs depends on the state
of the system

Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (37/54)

 Introduction - Wave Equation Model for Erythropoiesis
Age-Structured Model Method of Characteristics
Model for Erythropoiesis Model for Erythropoiesis with Delays

Model Reduction
Model Reduction: Several simplifying assumptions are made:
Assume that both velocities of aging go with time, t,
V (E) = W = 1.

Assume the birth rate β satisfies:


β, µ < µ1 ,
β(µ, E) =
0, µ ≥ µ1 ,
Assume that γ is constant.

The model satisfies the age-structured partial differential

equations:
∂p ∂p
+ = β(µ)p,
∂t ∂µ
∂m ∂m
+ = −γm.
∂t ∂ν
Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (38/54)
 Introduction - Wave Equation Model for Erythropoiesis
Age-Structured Model Method of Characteristics
Model for Erythropoiesis Model for Erythropoiesis with Delays

Model for Erythropoiesis

The boundary conditions for the age-structured PDEs are:
Recruitment of the precursors based on EPO concentration circulating in
the blood:
p(t, 0) = S0 (E).
Continuity of precursors maturing and entering the bloodstream as mature
RBCs:
p(t, µF ) = m(t, 0).
Active destruction of mature RBCs:

(1 − ν̇F (t))m(t, νF (t)) = Q.

The negative feedback by EPO satisfies the ODE:

a
Ė = − kE,
1 + KM r
where the total mature erythrocyte population is
Z νF (t)
M (t) = m(t, ν)dν.
0

Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (39/54)

 Introduction - Wave Equation Model for Erythropoiesis
Age-Structured Model Method of Characteristics
Model for Erythropoiesis Model for Erythropoiesis with Delays

Method of Characteristics
The precursor equation generally has maturing depending on EPO, E(t), but
we assume that V (E) = 1, so time and age are in lockstep.
If we define P (s) = p(t(s), µ(s)), then

dP ∂p dt ∂p dµ
= + = β(µ(s))P (s).
ds ∂t ds ∂µ ds

The method of characteristics

suggests we want

dt
=1
ds
or
t(s) = s + t0 ,
and
dµ
=1
ds
or
µ(s) = s + µ0 .

Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (40/54)

 Introduction - Wave Equation Model for Erythropoiesis
Age-Structured Model Method of Characteristics
Model for Erythropoiesis Model for Erythropoiesis with Delays

Method of Characteristics

With the method of characteristics, the precursor equation,

dP
= β(µ(s))P (s),
ds
is a birth only population model.
The model assumes that the body uses apoptosis at the early
recruitment stage (CFU-E) to decide how many precursor cells are
allowed to mature.
The solution to the ODE above is
Rs
β(µ(r))dr
P (s) = p(t, µ) = P (0)e 0 ,

which is valid for 0 < µ < µF , focusing on the larger time solution.

Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (41/54)

 Introduction - Wave Equation Model for Erythropoiesis
Age-Structured Model Method of Characteristics
Model for Erythropoiesis Model for Erythropoiesis with Delays

Method of Characteristics
This aging process of the precursor cells is primarily a time of
amplification in numbers before the final stages of simply add
hemoglobin.
The model shows how recruited cells amplify, then enter the mature
compartment (bloodstream) to circulate and carry O2 :
Rs
β(µ(r))dr
p(t, µF ) = p(t0 , 0)e 0

= p(t − µF , 0)eβµ1 = eβµ1 S0 (E(t − µF )).

From the method of characteristics on the mature RBCs, a

similar result gives:
m(t, ν) = m(t − ν, 0)e−γν .

The continuity between the precursors and the mature RBCs

gives:
m(t − ν, 0) = p(t − ν, µF ) = eβµ1 S0 (E(t − µF − ν)).
Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (42/54)
 Introduction - Wave Equation Model for Erythropoiesis
Age-Structured Model Method of Characteristics
Model for Erythropoiesis Model for Erythropoiesis with Delays

Total RBCs
The O2 carrying capacity of the body depends on the total number of RBCs,
which is the integral over all m(t, ν) in ν:
Z νF (t)
M (t) = m(t − ν, 0)e−γν dν
0
Z νF (t)
= eβµ1 S0 (E(t − µF − ν))e−γν dν,
0
Z t−µF
= e−γ(t−µF ) eβµ1 S0 (E(w))eγw dw.
t−µF −νF (t)

We apply Leibnitz’s rule for differentiating an integral:

Z t−µF
Ṁ (t) = −γe−γ(t−µF ) eβµ1 S0 (E(w))eγw dw,
t−µF −νF (t)
h i
+eβµ1 S0 (E(t − µF )) − S0 (E(t − µF − νF (t)))e−γνF (t) (1 − ν̇F (t))

= −γM (t) + eβµ1 S0 (E(t − µF )) − Q,

Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (43/54)

 Introduction - Wave Equation Model for Erythropoiesis
Age-Structured Model Method of Characteristics
Model for Erythropoiesis Model for Erythropoiesis with Delays

Model for Erythropoiesis with Delays

After reduction of PDEs, the state variables become total mature

erythrocytes, M , EPO, E, and age of RBCs, νF .

dM (t)
= eβµ1 S0 (E(t − µF )) − γM (t) − Q
dt
dE(t)
= f (M (t)) − kE(t)
dt
dνF (t) Qe−βµ1 eγνF (t)
= 1−
dt S0 (E(t − µF − νF (t)))

This is a state-dependent delay differential equation.

Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (44/54)

 Introduction - Wave Equation Model for Erythropoiesis
Age-Structured Model Method of Characteristics
Model for Erythropoiesis Model for Erythropoiesis with Delays

Model for Erythropoiesis with Delays

Properties of the Model: Integrating along the characteristics

shows that the maturation process acts like a delay, changing the
age-structured model into a delay differential equation.

The state-dependent delay model has a unique positive

equilibrium.
The delay µF accounts for maturing time.
The state-dependent delay in equation for νF (t) comes from
the varying age of maturation.
The νF (t) differential equation is uncoupled from the differential
equations for M and E.
Stability is determined by equations for M and E

Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (45/54)

 Introduction - Wave Equation Model for Erythropoiesis
Age-Structured Model Method of Characteristics
Model for Erythropoiesis Model for Erythropoiesis with Delays

Linear Analysis of the Model

Due to the negative control by EPO, it can be shown that this

model has a unique equilibrium:

(M̄ , Ē, ν̄F ).

With the change of variables, x1 (t) = M (t) − M̄ , x2 (t) = E(t) − Ē,

and x3 (t) = νF (t) − ν̄F and keeping only the linear terms, we obtain
the linear system:

ẋ1 (t) = eβµ1 S00 (Ē)x2 (t − µF ) − γx1 (t),

ẋ2 (t) = f 0 (M̄ )x1 (t) − kx2 (t),

1
ẋ3 (t) = x2 (t − µF − ν̄F ) − γx3 (t).
Ē

Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (46/54)

 Introduction - Wave Equation Model for Erythropoiesis
Age-Structured Model Method of Characteristics
Model for Erythropoiesis Model for Erythropoiesis with Delays

Linear Analysis of the Model

Let X(t) = [x1 (t), x2 (t), x3 (t)]T , then the linear system can be
written:

Ẋ(t) = A1 X(t) + A2 X(t − µF ) + A3 X(t − µF − ν̄F ),

where
eβµ1 S00 (Ē)
   
−γ 0 0 0 0
A1 =  f 0 (M̄ ) −k 0 , A2 =  0 0 0 ,
0 0 −γ 0 0 0

and  
0 0 0
A3 =  0 0 0 .
1
0 Ē
0

We try solutions of the form X(t) = ξeλt giving:

h i
λIξeλt = A1 + A2 e−λµF + A3 e−λ(µF +ν̄F ) ξeλt .

Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (47/54)

 Introduction - Wave Equation Model for Erythropoiesis
Age-Structured Model Method of Characteristics
Model for Erythropoiesis Model for Erythropoiesis with Delays

Characteristic Equation
Dividing by eλt results in the eigenvalue equation:
 
A1 + A2 e−λµF + A3 e−λ(µF +ν̄F ) − λI ξ = 0.

So we must solve
−γ − λ eβµ1 S00 (Ē)e−λµF 0
det f 0 (M̄ ) −k − λ 0 = 0,
1 −λ(µF +ν̄F )
0 Ē
e −γ − λ

which gives the characteristic equation

(λ + γ) (λ + γ)(λ + k) + Āe−λµF = 0,
 

where Ā ≡ −eβµ1 S00 (Ē)f 0 (M̄ ) > 0.

Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (48/54)

 Introduction - Wave Equation Model for Erythropoiesis
Age-Structured Model Method of Characteristics
Model for Erythropoiesis Model for Erythropoiesis with Delays

Stability Analysis of Delay Model

Stability Analysis of the Delay Model
The characteristic equation is an exponential polynomial given
by  
(λ + γ) (λ + γ)(λ + k) + Āe−λµF = 0,

which has one solution λ = −γ.

This shows the stability of the νF equation, which was the
state-dependent portion of the delay model.
Remains to analyze

(λ + γ)(λ + k) = −Āe−λµF .

The boundary of the stability region occurs at a Hopf bifurcation,

where the eigenvalues are λ = iω, purely imaginary.

Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (49/54)

 Introduction - Wave Equation Model for Erythropoiesis
Age-Structured Model Method of Characteristics
Model for Erythropoiesis Model for Erythropoiesis with Delays

Stability Analysis of Delay Model

Properties of the Exponential Polynomial (Characteristic Equation)

(λ + γ)(λ + k) + Āe−λµF = 0.

The solution of the characteristic equation has infinitely many roots.

Discrete delay model is infinite dimensional as the initial data must be a
function of the history over the longest delay.
The exponential polynomial has a leading pair of eigenvalues and many of
trailing having negative real part (Stable Manifold Theorem).
Analysis of the delay model is easier than the generalized age-structured
model.
The models are equivalent under the assumption that V (E) = W = 1.
Stability changes to oscillatory when the leading pair of eigenvalues cross
the imaginary axis, a Hopf bifurcation.

Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (50/54)

 Introduction - Wave Equation Model for Erythropoiesis
Age-Structured Model Method of Characteristics
Model for Erythropoiesis Model for Erythropoiesis with Delays

Stability Analysis of Delay Model

Hopf Bifurcation Analysis
A Hopf bifurcation occurs when λ = iω solves the characteristic
equation,
(iω + γ)(iω + k) = −Āe−iωµF .

From complex variables, we match the magnitudes:

|(iω + γ)(iω + k)| = Ā,

where the left side is monotonically increasing in ω, and the

arguments
 
ω ω
Θ(ω) ≡ arctan + arctan = π − ωµF ,
γ k

which has infinitely many solutions.

Solve for ω by varying parameters such as γ or µF .
Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (51/54)
 Introduction - Wave Equation Model for Erythropoiesis
Age-Structured Model Method of Characteristics
Model for Erythropoiesis Model for Erythropoiesis with Delays

Constant Flux Boundary Condition

Hopf Bifurcation: One significant method for finding the roots of
the characteristic equation at a Hopf bifurcation is the Argument
Principle from complex variables.

Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (52/54)

 Introduction - Wave Equation Model for Erythropoiesis
Age-Structured Model Method of Characteristics
Model for Erythropoiesis Model for Erythropoiesis with Delays

Constant Flux Boundary Condition

Experiment:
Give rabbits regular
antibodies to RBCs.
This increases destruction
rate γ.
Observe oscillations
in RBCs.
Model undergoes
Hopf bifurcation
with increasing γ.

Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (53/54)

 Introduction - Wave Equation Model for Erythropoiesis
Age-Structured Model Method of Characteristics
Model for Erythropoiesis Model for Erythropoiesis with Delays

Model for Erythropoiesis

Model can reasonably match the rabbit data by fitting parameters
that are reasonable.

The model stabilizes with variable velocity, V (E), but a more

complicated model.
Joseph M. Mahaffy, [email protected] PDEs - Method of Characteristics — (54/54)