Jaroslav Vlček
Dept. of Mathematics and Descriptive Geometry
VŠB-TU Ostrava
Contents
4 Non-stationary models 43
4.1 One-dimensional non-stationary models . . . . . . . . . . . . . . . . . . . . . . . 43
4.2 Examples of 1D non-stationary models . . . . . . . . . . . . . . . . . . . . . . . . 45
4.3 Partial differential equations of the first order . . . . . . . . . . . . . . . . . . . . 54
4.4 Non-stationary multi-dimensional models . . . . . . . . . . . . . . . . . . . . . . 61
5 Selected problems 66
5.1 Population and logistic models . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
5.2 Motion problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
2
Chapter 1
1.1.2 Example – electrical oscilatory circuit. Derive the time dependence of the current
i(t) in serial circuit with C and L by the initial conditions i(0) = i0 , (di(t)/dt)t=0 = 0.
d2 i(t)
1 di(t)
2
+ i(t) = 0 , i(t = 0) = i0 , =0.
dt LC dt t=0
3
4 CHAPTER 1. PRINCIPLES OF MATHEMATICAL MODELING
We will solve this initial (Cauchy) problem for second-order ordinary linear differential equation
with constant coeffitients using classical method. Characteristic equation
1
λ2 + =0
LC
t t
i(t) = C1 cos √ + C2 sin √ .
LC LC
t
i(t) = i0 cos √ .
LC
1.1.3 Modeling process scheme. Presented examples illustrate simple kinds of model
problems in the two totally different application fields. However, we can discuss in detail their
common features. In the both cases, we needed to choose appropriate balance relation in rele-
vant application branch; and, to supply this one by geometrical and/or physical assumptions.
In this way, a mathematical problem was formulated on certain abstraction level. We speci-
fied it (extremal problem, initial problem for differential equation) and selected corresponding
algorithm to solve it. Finally, the results were interpreted regarding the original context.
This strategy should be enforced by the most problems in practice, which we need to mathe-
matize and to solve using effective tools. In other words, we speak about mathematical modeling,
the scheme of which is demonstrated in the following figure.
formulation ofproblem,
real problem input data
physical description,
exact formulation simplication
data specification
choose of method,
mathematical model corectness
interpretation
1.1.4 Problem types. It is profitable to be aware of the most frequented types of mathe-
matical problems that we meet by a realization of mathematical model.
Classification by the aim:
• quantitative −→ values, data, functions,
• qualitative −→ existence, uniqueness, stability,
• procedures −→ algorithms, simulations,
..
• .
By the structure:
• equations,
• extremal problems,
• proofs, verification,
• simulations,
..
• .
1.2.2 The quantities and relations. Obviously, a number of quantities appears in the
mathematical model process with different role, the estimate of which is of key importance. We
distinguish
• local quantity (temperature, velocity etc.) specified in given point and/or time,
• global quantity (mass, electric charge, ...) if considered in a volume or time interval.
Following classification is quite fundamental:
• flux quantity (velocity, heath flux, ...),
• state quantity (potential, temperature, ...).
By a state or process describing the balance relation are principial that are supplied by
especial relations between mentioned ones. We call them constitutive relations. Specific
terminology and notation is known in particular applications; the mostly used examples are
summarized in the table.
6 CHAPTER 1. PRINCIPLES OF MATHEMATICAL MODELING
1.2.3 Density of a quantity. For a 1D case we denote f (x) the density of quantity F on
the interval hx1 , x2 i, d = x2 − x1 . Thus,
Zx2
F (d) = f (x) dx
x1
If f (x1 , x2 , x3 ) is the density of sources distributed in a domain, then the quantity F gives
total production of sources on this one, which becomes oftly a constituent of balance relations.
Atributes of several processes or states
elastic balance of forces displacements stress Hooke’s rule Young modulus vnějšı́ch
deformation (moments) u σ σ = E.ε, ε = g(u) E forces
1.2. BASIC NOTIONS AND RELATIONS
Notation:
−v(x1 ) v(x2 )
-
0 x1 x2 L x
-
A balance of the flux quantity v on the element hx1 , x2 i can be expressed as follows: the
sum of in- and out- flow through the end-points equals to the total production of inner sources
with the density f (x). The symbolic form of the balance relation can be written as
Zx2
∀ x1 , x2 ∈ (0, L) : v(x2 ) − v(x1 ) = f (x) dx . (2.1)
x1
The formula (2.1) represents the global conservation law of the flux quantity called also the
balance relation. In order to be in force, the function f (x) must be integrable. To more
detailled model analysis, a relation between flux and state quantity need be at disposal that will
be discussed bellow.
8
2.1. GLOBAL BALANCE 9
2.1.2 Constitutive relations can be of various form. We present here the most usual variant
that takes place among other by transfer phenomena modeling, relaxation processes etc.
Again, we consider element hx1 , x2 i, for which we denote ∆x = x2 − x1 , and, ∆u(x) =
u(x2 ) − u(x1 ) the change of state. The mean value
∆u(x)
∆x
If the two above quantities are globally proportional, we write this property as connstitutive
relation in the form
∆u(x)
∆v(x) = − p , (2.2)
∆x
where the coefficient of proportionality p > 0 is a constant or function. The sign ”minus” express
the fact that the flux in direction of decreased state function is positive. In the most cases this
coefficient represents material properties of a medium; in particular
Combining the flux balance (2.1) with constitutive relation (2.2) we obtain global balance
of state quantity
Zx2
∆u(x)
−p = f (x) dx . (2.3)
∆x
x1
h(r2 ) − h(r1 ) ∆h
ṽ(r) = −k = −k . (2.4)
r2 − r1 ∆r
Here the filtration coefficient k depends mainly on material properties of aquifer (porosity). The
negative sign express the circumstance that the hydraulic height fall toward increasing distance
from borehole. Note that obtained formula has identical feature as the general constitutive
relation (2.2).
10 CHAPTER 2. ONE-DIMENSIONAL STATIONARY MODELS
Since the fiber cross-section can be omitted, we solve the one-dimensional problem, where
we have
• f (x) external forces density producing on the given element total force
Zx2
F = |F | = f (x) dx ;
x1
Now, we derive constitutive relation with the help of vertical components of the vector T (x) in
the end-points using well-known approximation for very small deformations:
∆u
σ(x1 ) = T (x1 ) sin α ≈ T (x1 ) tg α ≈ T (x1 ) ,
∆x x=x1
∆u
σ(x2 ) = T (x2 ) sin β ≈ T (x2 ) tg β ≈ T (x2 ) .
∆x x=x2
The mean change in the fiber section is expressed as the difference
∆u ∆u
∆σ(x) = σ(x2 ) − σ(x1 ) = T (x2 ) − T (x1 ) , (2.6)
∆x x=x2
∆x x=x1
or formally for arbitratry difference ∆x,
∆u
∆σ(x) = T (x) . (2.7)
∆x
The last two equations represents constitutive relations between the load and displacement.
Note that in more general elasticity problems the stress and deformation are expressed in tensor
form. Mutual relation is called Hooke’s law, in which the other material characteristic take place
– elastic module. In our case, the obtained constitutive relation is quite sufficient. Putting it
into (2.5) leads to global balance equation for displacement
Zx2
∆u ∆u
T (x2 ) − T (x1 ) = f (x) dx . (2.8)
∆x x=x2
∆x x=x1
x1
We will suppose that the element hx1 , x2 i ⊂ (0, L) ⊂ R can be arbitrary small in the sense of
continuum hypothesis (obr. 2.4).
0 x1 x x2 L x
u -
If the source density f (x) is only integrable, then one holds for arbitrary x ∈ (x1 , x2 )
Zx2
lim f (x) dx = 0.
x1 →x−
x2 →x+ x1
12 CHAPTER 2. ONE-DIMENSIONAL STATIONARY MODELS
This result is local conservation law of flux quantity that can be simply formulate as the
flux continuity principle: ”the all what flow in certain point must also flow out”.
2.2.2 Local form of constitutive relations. In the introduced global constitutive relation,
∆u(x)
v(x2 ) − v(x1 ) = − p(x) , (2.11)
∆x
we realize previous limit process. Therefore, the right-hand difference ratio is substituted by the
derivation:
du(x)
v(x) = − p(x) . (2.12)
dx
The flux function v(x) at a point x is proportional to the derivative (local change) of state
function u(x). As the consequence we obtain from (2.10) the transient condition for the
state quantity
du(x−) du(x+)
p(x−) = p(x+) . (2.13)
dx dx
2.2.3 Differential form of conservation law. Again we start with global balance of flux
quantity, but by the assumption of continuity of source density function f (x). We divide
the both sides of (2.9) by the difference x2 − x1 ,
Zx2
v(x2 ) − v(x1 ) 1
= f (x) dx ,
x2 − x1 x2 − x1
x1
and, we apply the integral mean-value theorem1 . It results in the quasi-local balance
v(x2 ) − v(x1 )
= f (ξ) .
x2 − x1
The limit transition by ∆x = x2 −x1 → 0 leads to the local differential form of conservation
law for the flux function
dv
= f (x) , (2.14)
dx
beacause lim f (ξ) = f (x) for any continuous function f (x).
ξ→x
Applying the local constitutive relation (2.12) in above equation we can write differential
form of the state-function conservation law
d du(x)
− p(x) = f (x) ∀ x ∈ (0, L) . (2.15)
dx dx
1 1
Rx2
∃ ξ ∈ (x1 , x2 ) : x2 −x1
f (x) dx = f (ξ) .
x1
2.2. LOCAL BALANCE 13
2.2.5 Consequences of source function properties. Both the relations (2.10) a (2.14)
follow from conservation law of the flux quantity, but by different assumptions. If the function
f (x) is continuous, these relations are equivalent; together, the smoothness of state function is
induced, so that the derivative dudx exists for any x ∈ (0, L).
On the other hand, the density f (x) need not even be integrable, nevertheless, the problem
is of the good physical nature – let consider an elastic fibre by one-point loading (Fig. 2.5).
f (x0 )
0 x0 ? L x
u -
In this case, the source function must be understand in the generalized sense, when we
interpret this one using the Dirac distribution δ(x), for instance. For an illustration, the global
stress balance (2.5) can be written as
Zx2
∆σ(x) = f (x) δ(x − x0 ) dx ( = f (x0 ) ) .
x1
It is obvious that also state function u(x) will belong to a more general function class. Conse-
quently, we speak about weak formulation of the equation (2.16).
2.2.6 Example – heat transfer in a wall. We will model both the global and local balance
by heat transfer through the wall of thickness s = x2 − x1 with heat conductivity λ(x) by the
surface temperatures T2 < T1 . The heat flux q(x) is the flux quantity, as the state one we take
the temperature T (x). Denoting f (x) a possible density of inner heat sources we can write the
global balance as
Zx2
q(x2 ) − q(x1 ) = f (x) dx , (2.17)
x1
or in local form
dq
= f (x) . (2.18)
dx
Corresponding local balance of state function arises by application of constitutive relations that
have the global and local form, respectively:
T2 − T1 ∆T dT
q̃(x) = −λ(x) = −λ(x) , q(x) = −λ(x) . (2.19)
x2 − x1 ∆x dx
14 CHAPTER 2. ONE-DIMENSIONAL STATIONARY MODELS
If we put the last one into local flux balance, then the resulting local conservation law for the
temperature T (x)
d dT
− λ(x) = f (x) (2.20)
dx dx
represents well-known stationary Fourier heat-transfer equation in the 1D case.
ZL
x ∈ (x1 , L) ... v(L) − v(x1 ) = f (x) dx . (2.22)
x1
vL
−v0 -
x1 = 0 x2 x1 x2 = L x
-
u0 uL
Local flux balance in the boundary points is derived by limit transient, where we denote
v(0+) − v0 = 0 ⇒ v(0+) = v0 ,
vL − v(L−) = 0 ⇒ v(L−) = vL .
Local state balance is again a consequence of constitutive relation (2.12), rewritten by the
limits toward to boundaries of interval:
du(0+) du(0+)
v(0+) = − p(0+) , i.e. − p(0+) = v0 , (2.23)
dx dx
du(L−) du(L−)
v(L−) = − p(L−) , i.e. − p(L−) = vL . (2.24)
dx dx
In the case, when the values u0 , uL of state function are prescribed at end-points, we usualy
establish corresponding flux on the base of transient conditions
v0 = α0 (u(0+) − u0 ) , (2.25)
vL = αL (u(L−) − uL ) , (2.26)
2.4. BOUNDARY PROBLEMS FORMULATION 15
where α0 , αL ∈ R+ are given coefficients. The local balance (2.23) a (2.24) with these right
sides give differential form of boundary conditions for state function:
du(0+)
− p(0+) = α0 (u(0+) − u0 ) , (2.27)
dx
du(L−)
− p(L−) = αL (u(L−) − uL ) . (2.28)
dx
Since this conditions follow directly from conservation law, we call this result as natural bound-
ary conditions.
2.3.2 Example – heat exchange at wall surfaces. The temperature profile T (x) in a
homogeneous wall of thickness L is predetermined by boundary conditions on the both surfaces
(Fig. 2.7). We will suppose that the outer temperatures T0 > TL are given as well as the
transient coefficients α0 , αL [W K −1 m−2 ].
The constant heat conductivity λ [W K −1 m−1 ] is the material characteristics, for which
p(0+) = p(L−) = λ in (2.27) and (2.28):
dT (0+) dT
λ = α0 (T (0+) − T0 ) , λ = α0 (T − T0 ) , x = 0 ,
dx or dx . (2.29)
dT (L−) dT
−λ = αL (T (L−) − TL ) −λ = αL (T − TL ) , x = L .
dx dx
We changed the sign in the first condition because of T0 ≥ T (x) ≥ TL (the temperature is
decreasing function of the variable x, therefore, it has negative derivative).
The right form is more clear, if we uderstand the terms in the one-side limit sense, therefore,
we prefer this manner in what follows.
2.4.2 Neumann boundary problem for the equation (2.31) folows from Newton one by
α0 = αL = 0 (the equation is unchanged):
du(x)
p(x) = h0 , x=0, (2.34)
dx
du(x)
p(x) = hL , x=L, (2.35)
dx
where h0 , hL are given values of the state function derivative in boundary points. In the especial
case, when h0 = hL = 0, we have the boundary problem with homogenneous Neumann
conditions
du(x)
p(x) = 0, x=0, (2.36)
dx
du(x)
p(x) = 0, x=L. (2.37)
dx
2.4.3 Dirichlet boundary problem for the equation (2.31) is formulated by directly pre-
scribed values of the state function u0 , uL :
It seems that we could obtain these conditions from Newton variant for p(x) = 0, but this idea
is at variance with balance equation. Although the Dirichlet conditions are frequently present
in real situations, we denote them as ”non-natural”. Again, we can work with homogeneous
variant, if u0 = uL = 0.
2.4. BOUNDARY PROBLEMS FORMULATION 17
2.4.4 Boundary problems with non-continuous data. We will study, what consequence
have more soft requirements to the input parameters. We admit a discontinuity, but by bounded
values, i.e. the functions have at least one-side proper limits. The variant with discontinuous
source density f (x) is discussed in the paragraph 2.4.7; here, we show two typical examples of
the non-continuous function p(x) in a point xp ∈ (0, L) – see Fig. 2.8:
• geometrical discontinuity (it appears rather in multidimensional models),
• material discontinuity.
The conservation law holds in the both cases (for integrable source density f (x)), but in the
sense of the flux continuity (2.10) as v(xp −) = v(xp +) in the form
du(xp −) du(xp +)
p(xp −) = p(xp +) , (2.40)
dx dx
that this is well-known transient condition (2.13). It is essential to appreciate that the
continuity assumption of the state function, u(xp −) = u(xp +), does not ensure the continuity
of its derivative u0 (x) at the point xp , because of
2.4.5 Correctness of model. A correctly formulated problem must fulfil three basic pro-
perties:
1. it has a solution (existence),
Generally, the problems with Neumann boundary conditions (2.34), (2.35) can be un-correct
– the solution exist only for several combination of the data f (x), g0 , gL , moreover, it need not
be unique.
• heat exchange with outside passes is the same on the both surfaces → α0 = αL = α,
• the left surface temperature is more than the on right one → T0 > TL .
Our aim is to find general solution and to apply the result to given data: L = 0, 5 m, T0 = 50◦ C,
TL = 20◦ C, λ = 1 WK−1 m−1 (brickwork), α = 2 WK−1 m−2 .
With regard to above assumptions takes the equation (2.41) for the temperature T (x) the
simple form
d2 T
= 0, x ∈ (0, L).
dx2
Boundary conditions of Newton type
dT
−λ = α(T0 − T ), x=0,
dx
dT
−λ = α(T − TL ), x=L
dx
were discussed in (2.29). Successive integration in second-order differential equation leads di-
rectly to the general solution
T (x) = C1 x + C2 ,
which we put into boundary conditions to obtain the constants C1 , C2 :
α λ
C1 = − (T0 − TL ) , C2 = T0 − (T0 − TL ) ;
αL + 2λ αL + 2λ
thus, we obtained final solution
αx + λ
T (x) = T0 − (T0 − TL ) .
αL + 2λ
We verify easy that this result satisfies the equation and boundary equations as well. The linear
function T (x) is continuous on the interval h0, Li that aposteriori implies the correctness of
solved model problem.
An application of given data leads to the straight line T (x) = 40 − 20x, x ∈ h0, 0, 5i; and,
also to the temperature on the surfaces: T (0) = 40◦ C, T (L) = 30◦ C. The graphical output in the
Fig. 2.9 presents the important fact that the Newton conditions do not preserve the continuity
of state function T (x) on the boundary.
f (x)
? ? ? ? ? ? ? x
-
0 L
u(x)
We will formulate and solve several typical problems that illustrate the notions from previous
text by following assumptions:
• homogenneous string is very long and thin as well, i.e. the length remarkable predominates,
that enables to model this problem as one-dimensional,
Presented analysis is aimed at the right-hand side properties as well as at boundary conditions.
1
u(x) = qx2 + C1 x + C2 , (2.46)
2
to which we will apply various boundary conditions.
A1a. Dirichlet problem corresponds to the fixed end-points, i.e. to the homogenneous
conditions
u(0) = u(L) = 0 .
Calculating the constants C1 = − 12 qL, C2 = 0 we obtain the parabola
1 1 1
u(x) = qx2 − qLx = qx(x − L) . (2.47)
2 2 2
Given problem and its result are presented in the Fig. 2.10; the solution is unique,
the model is correct.
A1b. Neumann boundary problem means that derivative of the function u(x) is prescribed
in the end-points as the tangent of solution (”free ends” – see Fig. 2.11. We write
the boundary conditions in the form
f (x)
aa !
aa α ? ? ? ? ? ? ? !!β
aa ! ! - x
0 aa ! !! L
a a !!
u(x)
Figure 2.11: Deflection of a string with free ends.
2.4. BOUNDARY PROBLEMS FORMULATION 21
x=0 ... k0 = C1 ,
x=L ... kL = qL + C1 .
It is obvious that the constant C2 can be arbitrary, therefore, the problem has infinite
number of solutions, but existing for k0 = kL − qL only in the form
1
u(x) = qx2 + k0 x + C2 .
2
It means, that the angles at end-points are invoked by given force, so that these can
not be apriori postulated. Therefore, the formulation with Neumann conditions is
not correct.
A1c. Mixed boundary problem can take many features; one from typical variants is the
bend of one-side fixed thin rod – Fig. 2.12. In this case the boundary conditions have
the form
u(0) = 0 ... on the fixed boundary,
u0 (L) = kL ... at the free end.
The derivative sign corresponds to the direction of deformation – see the figure. By
kL > 0
x
-
0 L
kL < 0
? ? ? ? ? x
-
0 a L
the model of this problem when consider the particular solution on each string part:
0 , x ∈ h0, a), u1 (x) , x ∈ h0, a),
q(x) = =⇒ u(x) =
q, x ∈ ha, Li, u2 (x) , x ∈ ha, Li.
Both the functions are connected at the point x = a by the contact conditions following
from the flux continuity principle: the string remains non-disturbed after a deformation.
Therefore, the flux function T is continuous (constant) by the assumption. With regard
to the conclusions in the paragraph 2.4.4 we get the transient conditions
It is easy to ascertain that this function fulfil both differential equation and boundary
conditions.
B2. If a loading f (x) acts in one point x0 ∈ (0, L) of a string only, the situation becomes
very specific. Since the right-hand side is not a function in classical sense, this physically
reasonable problem leads to the question, in what function class the solution should be
2.5. EXERCISES 23
expected. Among other, the ”function” f (x) can be written with the help of Dirac impulse
q(x) = qδ(x − x0 ). In the model, the bellowed relation can be used2 :
C1 , x < x0 , C1 x + C3 , x < x0 ,
Z
Z
δ(x − x0 ) dx = h(x) = h(x) dx =
C2 , x > x0 , C2 x + C4 , x > x0 .
For the problem solved by the same way as in B1 we can not suppose the continuity of
state function derivative at the point x0 , because of u01 (x0 ) 6= u02 (x0 ). Therefore, we apply
the moments balance as the fourth condition,
Z L Z L
T
u(x) dx = f (x) δ(x − x0 ) dx ,
L 0 0
which is formulated globally on the interval h0, Li. Note, that the right-hand side equals
to f (x0 ) at arbitrary point.
2.5 Exercises
2.5.1 Formulate mixed boundary problem for the one-dimensional heating of a wall. On the
outside surface the constant temperature is given, T0 = 32 ◦ C. At the inside one, the heat transfer
proceeds with the coefficient α into the medium of the temperature TL = 18 ◦ C. Solve by the
same assumptions as in the Example 2.4.6 with following input data: L = 0, 3 m, λ = 0, 2
WK−1 m−1 , α = 4 WK−1 m−2 . Establish the temperature at inner surface; demonstrate the
resulting temperature profile graphically.
2.5.2 Solve the Example 2.4.7A1a, if the string stretched by the force T is loaded by the
constant force density f on the section hx0 − a, x0 + ai. Determine the string deflection in the
point x0 by these values:
• L = 3 m, x0 = L/3, a = L/6,
• f = 12 Nm−1 , T = 325 N.
2.5.3 A string of the length L stretched by the force T = 60 is fixed at the both ends. It is
loaded at the point x0 dividing its length by the ratio 1:2. For the force density f = 10 Nm−1
solve this problem to establish the resulting form of string and the highest deflection (use the
conclusions of the Example 2.4.7B2).
2
these relation are not purely exact because of their rather formal character; indeed, an use of them in presented
context leads to required result
Chapter 3
Ω . . . a domain in R2 or R3 ,
∂Ω . . . its boundary – (piecewise) smooth curve (in R2 ) or surface (in R3 ),
Ω̄ = Ω ∪ ∂Ω . . . closed domain,
Ω0 ⊂ Ω . . . reference sub-domain,
n . . . unit normal vector of ∂Ω0 .
M ∈ Ω0 . . . inner point,
P ∈ ∂Ω0 . . . boundary point,
u = u(M ) . . . state function (usualy scalar quantity),
v = v(M ) . . . flux function (usualy vector field),
f = f (M ) . . . source function, source density (integrable).
The normal and tangential flux belong to basic quantities on a boundary (Fig. 3.2).
Denote τ the tangential vector that lies (in R3 ) in the plane determined by the vectors n and v.
24
3.1. BALANCE RELATIONS 25
Similarly, n0 , τ 0 are the unit items. Thus, we can express projections of the flux v into tangent
and normal direction:
3.1.2 Global and local field characteristics. The group of global characteristics is defined
for the functions u, v ∈ C 1 (Ω0 ) by integral operators:
On the other hand, we can estimate locally at arbitrary point of the domain Ω0 :
26 CHAPTER 3. MULTIDIMENSIONAL STATIONARY MODELS
1. a change of the scalar field u in given direction s with the help of direction derivative
∂u
= s0 · grad u,
∂s
where s0 is unit direction vector; the field gradient
∂u ∂u ∂u
grad u = , ,
∂x ∂y ∂z
represents the direction of the largest decrease (or increase) of the function u;
2. source character with regard to the divergence, because
• div v 6= 0 for source field,
• div v = 0 for non-source field,
3. vorticity regarding field rotation established using curl operator, because
• curl v 6= o in vortex field,
• curl v = o in non-vortex (potential) field.
3.1.3 From global to local characteristics. This step is realized by limit process, when
the volume (surface, curve, ...) is reduced into single point. Following explanation is slightly
simplified without loss of generality. At first we recall two important theorems of vector analysis.
The Gauss-Ostrogradsky theorem express the equality of volume integral of divergence on
a domain and the total flux through its boundary:
Z Z
div v dV = v dS . (3.6)
Ω0 ∂Ω0
The Stokes formula allows to express flux circulation along the curve K by surface integral,
I Z
v dl = rot v dS , (3.7)
K S
where S is arbitrary smooth surface passed through the curve K and being accordantly oriented
to this one. Similar statements hold also for volume integral of gradient or divergence:
Z Z
grad u dV = u dS , (3.8)
Ω0 ∂Ω0
Z Z
rot v dV = v × dS . (3.9)
Ω0 ∂Ω0
Let µ(Ω0 ) denotes a measure of the domain Ω0 (volume, surface, diameter etc.), for instance.
Supposing continuity of the function f on the reference domain, i.e. f ∈ C(Ω̄0 ), the process of
continuous reduction of Ω0 into the point M can be expressed in symbolic form as
Ω0 −→ M ⇐⇒ µ(Ω0 ) −→ 0 .
If we apply this principle to volume integral representing total source production in Ω0 , then we
obtain generalized average integral value of the density:
1
Z
lim f dV = f (M ) . (3.10)
µ→0 µ(Ω0 )
Ω0
3.1. BALANCE RELATIONS 27
Above relation enables defining the divergence of the field v by (3.6) as local characteristics at
the point M :
1
Z
lim v dS = div v . (3.11)
µ→0 µ(Ω0 )
∂Ω0
In the same way, the remaining differential operators can be locally defined using (3.8) and (3.9):
1
Z
lim u dS = grad u , (3.12)
µ→0 µ(Ω0 )
∂Ω0
1
Z
lim v × dS = rot v . (3.13)
µ→0 µ(Ω0 )
∂Ω0
3.1.4 Balance equations. As in 1D models, we will proceed from global to local formula-
tions.
(A) Conservation laws
Global conservation law of the flux quantity
Z Z
∀Ω0 ⊂ Ω : v dS = f (M ) dV . (3.14)
∂Ω0 Ω0
By (3.6), it means that the total flux through the domain boundary equals to accumulate pro-
duction of the sources inside.
(B) Constitutive relations are introduced in local vectorial form that is typical for transfer
phenomena as
v(M ) = − p(M ) grad u(M ) , M ∈Ω. (3.16)
More general feature is obtained, if the material quantity p depends also on state function u:
v(M ) = − p(M, u) grad u(M ) . (3.17)
3.1.5 Example – laminar flow of ideal liquid. Our aim is to create mathematical model
of laminar flow in given region (Fig. 3.3). The global balance of mass flux m [kg.m2 s−1 ] in
the volume Ω is the starting relation, where the source production is described by the density
f [kg.m3 s−1 ]. In the case of ideal (non-viscous) liquid we can write
Z Z
m dS = f dV .
∂Ω Ω
According the procedure introduced previously we obtain local balance of the flux quantity,
div m = f on Ω.
In practice, the use of the flow velocity v instead of mass flux is more advantegeous, therefore,
we put m = ρv, where ρ is mass density of the liquid. In the resulting relation
div (ρv) = f
ρ div v + v · grad ρ = f .
Here both left-hand parts are called as source and advective term, respetively.
In what follows, we will assume an uncompressible liquid, for which ρ = konst. The resulting
equation
ρ div v = f (3.21)
passes into well-known continuity equation div v = 0 in a volume free of sources.
Finally, we apply the requirement of laminar flow. It means, that velocity field circulation
along arbitrary closed curve in given region is equal to zero:
I Z
C= v dl = rot v dS = 0 ∀ K ⊂ Ω.
K S
rot v = o .
3.2. BALANCE ON A BOUNDARY 29
Moreover, the velocity can be written as the gradient of certain scalar function that we call
velocity potential ψ:
v = grad ψ .
Putting it into (3.21) we obtain
ρ div ( grad ψ) = f ; (3.22)
and, the classical Laplace equation ∆ψ = 0 for the potential ψ, if f = 0.
There exists the other way to the continuity equation directly from global balance without
sources. It suffices to rewrite the flux term as the sum
Z Z Z Z
ρv dS = ρv dS + ρv dS + ρv dS = 0 .
∂Ω S1 S2 S3
Since the surfaces S1 , S2 have reversely oriented normals thereby the flux through the mantle
S3 is zero-valued, one holds for ρ = konst.
S1 v1 = S2 v2 .
n(P )
v n
Γ3 Ω0
Γ2
P
Γ4 Γ1
' $
Γ0
Γ
Ω
v
& %
n
Γ
The limiting steps µ(Ω0 ) → 0 along the normal n from inside of Ω0 and in opposite direction
from outside lead to µ(Γ1 ) → 0, µ(Γ3 ) → 0, but the boundary Γ0 = Γ ∩ Ω0 retains unchanged.
Thus, we can write
1
Z Z
lim v dS = [v(P +) − v(P −)] · n dS ∀ P ∈ Γ0 ∀ Γ0 ⊂ Γ . (3.24)
µ→0 µ(Ω0 )
∂Ω0 Γ0
30 CHAPTER 3. MULTIDIMENSIONAL STATIONARY MODELS
As the local consequence of (3.23) for arbitrary boundary point P ∈ Γ we obtain [v(P +) −
v(P −)] · n = 0, therefore, by (3.1)
vn (P +) = vn (P −) . (3.25)
In this way we formulated local condition of normal flux continuity or interfacial con-
dition.
3.2.2 Condition on outer boundary. Now, we attend to the boundary ∂Ω, where the
outside flux v 0 (P ) is prescribed (Fig. 3.5). The limit step occurs at the same principle as in
previous case with following result:
[v(P +) − v 0 (P )] · n = 0 ⇒ vn (P ) = v0n , v0n = v 0 (P ) · n . (3.26)
n(P )
v 0
P Γ0
Ω0
∂Ω
Ω v
that constitutes boundary problem if supplied by several conditions below. The classification of
boundary problems is analogous as by 1D models.
∂u
p + αu = g na ∂Ω (3.33)
∂n
∂u
p =g na ∂Ω . (3.34)
∂n
3. Dirichlet condition. In this simplest variant, directly the state function is given on the
boundary:
u = u0 na ∂Ω . (3.35)
If various conditions are prescribed for different parts of the boundary ∂Ω, the boundary
problem is called as mixed.
1. Existence of solution (solvability); the properties of solution depend not only on the
input data properties (f, p, . . .) as for 1D models, but also on quality of the domain
Ω and its boundary.
(5) heat losses take the place along the side no. 2 expressed by the continuous heat flux q(y) ≥ 0
Wm−2 .
Equation
Let the rectangle sides have the size a, b, respectively. Thus, the plate inward is the domain
Ω = (0, a)×(0, b), where we will estimate the temeperature field T (x, y). For this state quantity,
the local conservation law has previously derived form
Regarding first two assumptions the equation takes more simple Laplace PDR form:
(1) ⇒ −λ div grad T = f
∆T (x, y) = 0 on Ω . (3.38)
(2) ⇒ f (x, y) = 0
Boundary conditions
Denote Γ1 , Γ2 , Γ3 , Γ4 the rectangle sides in agreement with enumeration in the figure; thus,
∂Ω = ∪Γi , i = 1, . . . , 4. We exploit the fact, that normal vectors of boundaries (rectangle sides)
are parallel to coordinate axes, therefore, we can write directly partial derivatives regarding the
3.4. FOURIER METHOD 33
variables x, y instead the gradients. It is obvious, that the orientation of normal vector n must
be taken into account on each boundary with regard to the identity
∂u
= n0 · grad u ,
∂n
where n0 is unit vector of outer normal.
• The third condition means the ideal temperature contact with surrounding medium on Γ3
that we write as Dirichlet condition
T = T0 for y = b . (3.39)
• The boundaries Γ1 and Γ4 are isolated by the assumption (4); zero-valued heat flux implies
homogenneous Neumann conditions
∂T
= 0 for y = 0 , (3.40)
∂y
∂T
= 0 for x = 0 . (3.41)
∂x
• The last condition on the boundary Γ2 is of Newton type; we write this one in the well-
known form
∂T
−λ = q(y) for x = a . (3.42)
∂x
Obtained results represents the mixed boundary problem for Laplace equation on a rectangle.
Remarks
(a) The equation ∆T = 0 by itself is less telling. The essential properties (domain form,
characteristics of boundaries etc.) are contained in boundary conditions.
(b) Verification of correctness requirements follows from the fact that input data are constants
or continuous functions. However, the situation at rectangle corners should be discussed
(what is the conclusion?).
(c) The solution of presented problem can be found e.g. by the Fourier separation method –
see the next section.
Zb
2
kf (x)k = | f (x) |2 dx . (3.44)
a
In what follows, we will consider the square-integrable functions, for which kf (x)k < ∞.
34 CHAPTER 3. MULTIDIMENSIONAL STATIONARY MODELS
Moreover, if kgn (x)k = 1, we have orthonormal system. Any orthogonal system can be
normed with the elements {gn (x)/kgn (x)k}.
Let {gn (x)} be an orthogonal system on ha, bi, and, function f (x) is square-integrable therein.
The series
∞
X
an gn (x) (3.46)
n=1
express expansion of the function f (x) with respect to the system {gn (x)}, if one holds
for the coefficients an ,
Zb
1
an = f (x) g(x) dx . (3.47)
kf (x)k2
a
Multiplying both sides of this relation by gm (x) and integrating over interval ha, bi, we obtain
by (3.45) only single non-zero term on the right side for m = n that equals to square of the
norm. However, it is obvious, that we need to find answer to two basic questions (related to the
given interval):
∞
X
| an |2 < ∞ (Bessel inequality). (3.49)
n=1
∞
an gn (x) converges on the interval ha, bi to the function f (x), if and only if
P
(2) The series
n=1
∞
X
2
kf (x)k = | an |2 (Parseval equality). (3.50)
n=1
Therefore,
∞
X
f (x) = an gn (x) (3.51)
n=1
on the interval h0, 2πi. We will prove its orthogonality and derive normed form.
0 , m 6= n
Z2π
sin mx sin nx dx = π , m = n 6= 0
0
0
, m = n = 0,
Z2π
cos mx sin nx dx = 0 .
0
The orthogonality is obvious; further, we calculate the norms of the functions contained in given
system:
Z2π 2 Z2π Z2π
1 π 2
dx = , cos mx dx = π, sin2 mx dx = π .
2 2
0 0 0
3.4.3 Fourier series. If we seek for expansion of the function f (x) with respect to the system
(3.52), then we obtain using (3.47) and (3.48)
∞
a0 X
f (x) ∼ + (an cos nx + bn sin nx) , (3.54)
2 n=1
where
Z2π
1
an = f (x) cos nx dx , n = 0, 1, 2, . . . , (3.55)
π
0
Z2π
1
bn = f (x) sin nx dx , n = 1, 2, . . . . (3.56)
π
0
The series (3.54) is called as Fourier expansion of the function f (x) on interval h0, 2πi,
the coefficients of which are given by (3.55) and (3.56).
36 CHAPTER 3. MULTIDIMENSIONAL STATIONARY MODELS
This result can be easily generalized for interval h0, Li by the transform x → 2πx/L:
∞
a0 X 2πnx 2πnx
f (x) ∼ + an cos + bn sin , (3.57)
2 n=1
L L
where
ZL
2 2πnx
an = f (x) cos dx , n = 0, 1, 2, . . . , (3.58)
L L
0
ZL
2 2πnx
bn = f (x) sin dx , n = 1, 2, . . . . (3.59)
L L
0
Thus there exists uniquely determined Fourier expansion on h0, Li of the form
∞
a0 X 2πnx 2πnx
s(x) = + an cos + bn sin
2 n=1
L L
• to the function f (x) at each point of interval (0, L), where this function is continuous,
• to the periodic enlargement f˜(x) outside interval (0, L) or to the mean of one-side limits,
where f˜(x) is not continuous.
3.4.4 Example. The function f (t) = (t − 2)2 is given for t ∈ h0, 2i. We will create the graph
of its periodic enlargement and find Fourier expansion.
The part for x ∈ h0, 4i of the parabola f (t) = (t − 2)2 is in Fig. 3.7 on the left. Function f (t)
is continuous, therefore, the discontinuities arrise after periodization only at bounds of periods,
where the expansion will converge to mean of one-side limits (it is equal to unity) – see the right
part of the figure. The Fourier coefficients are calculated by (3.58), (3.59) on the interval h0, 2i:
3.4. FOURIER METHOD 37
f (x) f˜(x)
4 c c c c 4 c c c
s s s s s s s
x c c c c c c c x
0 2 4 −6 −4 −2 0 2 4 6
R2
a0 = (x − 2)2 dx = 83 ,
0
R2
an = (x − 2)2 . cos nπx dx =
0
i2
(x−2)2
h
2 x2 1 4
= nπ sin nπx − nπ − nπ cos nπx + n2 π 2
sin nπx = n2 π 2
,
0
R2
bn = (x − 2)2 . sin nπx dx =
0
h 2
i2
= − (x−2)
nπ cos nπx +
2
nπ
x−2
nπ sin nπx + 1
n2 π 2
cos nπx = 4
nπ .
0
3.4.5 Fourier method for Laplace equation. We show the basic principle for simple
Dirichlet problem on a rectangle:
∂2u ∂2u
+ 2 =0 on Ω = {0 < x < a, 0 < y < b}, (3.61)
∂x2 ∂y
Boundary conditions.
We derive the constants A, B, C, D and unknown value µ from boundary conditions. It is very
advantageous to start with homogeneous conditions, therefore, we consider at first the relation
(3.62). In this case,
0 = (A cos µx + B sin µx).C
for arbitrary x ∈ h0, ai. Thus, C = 0, and
and,
u(x, y) = β sin µx sinh µy .
Next condition (3.63) holds at x = a:
0 = β sin µa sinh µy .
We must write the solution of problem as the superposition of all items as the series
∞
X ∞
X
u(x, y) = un (x, y) = βn sin µn x sinh µn y (3.70)
n=1 n=1
(if n = 0, then u0 (x, y) = 0). This expansion must satisfy the conditiion (3.64) at y = b:
∞
X
ϕ(x) = βn sin µn x sinh µn b .
n=1
3.4. FOURIER METHOD 39
It means, that this series is the Fourier expansion of the function ϕ(x) on the interval h0, ai,
because of ∞ X
ϕ(x) = β̃n sin µn x , where β̃n = βn sinh µn b. (3.71)
n=1
It can be shown that the system gn (x) = sin µn x is orthogonal on this interval,
Za
a
= k sin µn xk2 , m = n,
nπx mπx 2
sin sin dx = .
a a
0 , m 6= n
0
There are two ways to estimate of coefficients β̃n : a modification of general Fourier formulae
to obtain sine expansion or direct use of orthogonality. We apply the second one so, that we
multiply equation (3.71) by the function gm (x) = sin µm x and integrate over period L = a:
Za ∞
X Za
ϕ(x) sin µm x dx = β̃n sin µm x sin µn x dx .
0 n=1 0
Thus
Za
2 β̃n
β̃n = ϕ(x) sin µn x dx and βn = . (3.72)
a sinh µn b
0
Finally, we create resulting solution of given problem:
∞
X β̃n
u(x, y) = sin µn x sinh µn y . (3.73)
n=1
sinh µn b
Now, we turn back to the choose of negative sign of the constant µ2 . It follows from the fact,
that we have function of variable x in non-homogeneous boundary condition (3.64), therefore, we
require the goniometric eigen-functions also in this variable. In the opposite case (the condition
as a function of y), we should take the positive sign to have othogonal system in this variable.
In the case of non-homogeneous boundary conditions in the both variables, one from conditions
can be replaced in the equation as a source term by appropriate substitution.
3.4.6 Example – 3.3.3 continued. Derive the temperature field in the rectangular plate
(see Fig. 3.6) by following input data: a = 1 m, b = 2 m, T0 = 60 ◦ C, λ = 2 WK−1 m−1 ; and, by
heat flux
q0
q(y) = 2 (by − y 2 ) with q0 = 80 Wm−2 .
b
Demonstrate graphically temperature profile on the side, where the heat losses are acting.
∆ϑ(x, y) = 0 on Ω , (3.74)
40 CHAPTER 3. MULTIDIMENSIONAL STATIONARY MODELS
∂ϑ(x, y)
= 0, (3.75)
∂y y=0
∂ϑ(x, y)
= 0, (3.76)
∂x x=0
ϑ(x, b) = 0 , (3.77)
∂ϑ(x, y)
λ = q(y) . (3.78)
∂x x=a
∂ϑ(x, y)
(3.78) λ = q(y) ⇒
∂x x=a
∞
X q(y)
αk µk sinh µk a cos µk y = . (3.80)
k=1
λ
from which
Zb
2qk
αk = , qk = q(y) cos µk y dy . (3.81)
λbµk sinh µk a
0
Final solution
Using these results we can create following expansion of the temperature field that satisfies to
Laplace equation as well as given boundary conditions:
∞
2 X qk
T (x, y) = T0 + cosh µk x cos µk y . (3.82)
λb k=1 µk sinh µk a
Numerical results
Calculation of qk :
Zb
µk b − 2 sin µk b
q(k) = (by − y 2 ) cos µk y dy = . . . = .
µ3k
0
Temperature field:
∞
2q0 X µk b − 2 sin µk b
T (x, y) = T0 + cosh µk x cos µk y . (3.83)
λb3 k=1 µ4k sinh µk a
Temperature profile at x = a
The graphical output is presented in the Fig. 3.8. Computation has been realized in Matlab
with the first ten terms of series.
3.5 Exercises
3.5.1 Stationary temperature field. The half-strip by the Fig. 3.9 is considered. On
the finite border, parabolic temperature profile is given symmetrically along the x axis with
the maximum 2T0 at the centre of co-ordinate system. The horizontal sides of half-strip are
42 CHAPTER 3. MULTIDIMENSIONAL STATIONARY MODELS
r2
ϕ0 ,
a2
where the variable r denotes distance from common vertex. The two opposite sides have the
same constant potential ϕ0 .
Create mathematical model and solve obtained problem. Calculate the potential at the
rectangle center. Demonstrate graphically both the lines of force and equipotential lines by
appropriate scale for ϕ0 = 24, a = 10, b = 5.
Non-stationary models
Further, we introduce several important expressions to describe global balance of flux quantity.
Total source production on the element hx1 , x2 i ⊂ (0, L) and in time interval ht1 , t2 i ⊂
(0, ∞):
Zx2 Zt2
F = f (x, t) dt dx . (4.1)
x1 t1
Zt2
v(x, t) dt .
t1
Thus, the global balance in space-time domain hx1 , x2 i × ht1 , t2 i can be written in the form
43
44 CHAPTER 4. NON-STATIONARY MODELS
4.1.2 Local balance at internal point. The possibility of limit transition at arbitrary point
[x, t] ∈ (x1 , x2 ) × (t1 , t2 ) is supposed:
x1 → x− , x2 → x+ , t1 , t2 → t .
As the requirements to the functions in (4.2) are enhanced, the limiting leads to different balance
relations.
(A) If f (x, t), w(x, t) are only integrable, the equation (4.2) reduces in pure flux continuity:
v(x−, t) = v(x+, t) . (4.3)
(B) If f (x, t), w(x, t) are continuous in t and together w(x, t) will differentiable in this variable,
then we obtain after limiting in t
Zx2 Zx2
∂w(x, t)
v(x2 , t) − v(x1 , t) = f (x, t) dx − dx . (4.4)
∂t
x1 x1
(C) Moreover, if f (x, t), w(x, t) are continuous also in x, the analogous process results in local
balance
∂v(x, t) ∂w(x, t)
= f (x, t) − . (4.5)
∂x ∂t
It can be prove, that differentiability of the function v(x, t) by x follows from used as-
sumptions.
(D) Consider the previously introduced local constitutive relation modified for time dependence:
∂u(x, t)
v(x, t) = −p(x) . (4.6)
∂x
Its application in flux balance equation leads to local differential conservation law for state
quantity,
∂ ∂u(x, t) ∂w(x, t)
− p(x) = f (x, t) − , (4.7)
∂x ∂x ∂t
that has the form of second-order PDR specified by nature of the function w(x, t).
(B) If the transfer conditions are released with the coefficients α0 , αL and given functions
u0 (t), uL (t), we linearize the flux balance as
Local balance of state function leads by application of constitutive relations to the Newton
conditions
∂u(0+, t)
p(0+) = α0 [u(0+, t) − u0 (t)] ,
∂x (4.11)
∂u(L−, t)
p(L−) = αL [u(L−, t) − uL (t)] .
∂x
(C) In the case of prescribed values of state function we have simply the Dirichlet con-
ditions
u(0+, t) = u0 (t), u(L−, t) = uL (t) . (4.12)
Remark. When boundary condition at some point can not be prescribed, we can write (if no
variance with physical reality takes place)
Oftly it suffices to assume only the boundness of solution for x → ∞ (see the example 4.2.3).
4.1.4 Initial conditions. If time development of the function u(x, t) is modeled for t > 0,
we need to know initial state as a function of space variable:
u(x, 0) = ϕ(x) .
A feature of initial conditions can be extended, if also initial time derivative is required (e.g. as
the initial velocity in kinematic models):
∂u(x, 0)
u(x, 0) = ϕ(x) , = u0t (x, 0) = ψ(x), ... . (4.13)
∂t
These conditions can be also of the other meaning, when no time dependence is studied. If
we seek for solution of a differential equation at given point or on given line, the same problem
is called as Cauchy problem.
Summary. The full formulation of 1D non-stationary problems consist of the equa-
tion (4.7), initial conditions (4.13) and boundary conditions of the type (4.9), (4.11), (4.12),
eventually, of mixed ones.
where c [J kg−1 K−1 ] is heat capacity and ρ [kg m−1 ] the density (related to unit length). Right-
side dissipation function w(x, t) = cρ(x)T (x, t) express heat accumulation in considered medium
like as ∆Q = cρ∆T . In the constitutive relation
∂T (x, t)
q(x, t) = −λ
∂x
the symbol λ denotes heat conductivity of the rod. Putting it into local balance (4.7) we obtain
∂ ∂T (x, t) ∂T (x, t)
λ(x) = cρ(x) . (4.15)
∂x ∂x ∂t
Supposing homogeneous material properties, the above relation leads to well-known Fourier
equation
∂2T ∂T
a 2 = , (4.16)
∂x ∂t
where we write a = λ/(cρ) for the temperature conductivity. The three following examples
illustrate boundary problems with this equation.
4.2.2 Example – heat conduction in a thin finite rod I. We shall analyze temperature
field in a finite rod of the length `, which keeps constant temperature T0 at the left end-point;
and, retains isolated at the right end-point. Initial temperature distribution is described by the
function T0 (1 + sin πx/`).
Obviously, we need to solve Fourier equation (4.16) on the domain Ωx,t = (0, `) × (0, ∞) with
the initial condition
πx
T (x, 0) = T0 1 + sin . (4.17)
`
Boundary conditions are given as
T (0, t) = T0 , (4.18)
∂T
= 0, x = ` . (4.19)
∂x
Moreover, the asymptotic stability of solution can be ensured by the condition
Before problem solving we introduce an efficient notation that simplifies the calculation:
T − T0 x a
u(x, t) = , ξ= , κ= . (4.21)
T0 ` `2
After easy rearrangement we obtain following formulation of mixed non-stationary boundary
problem for second order partial differential equation (PDE):
1. u(ξ, 0) = sin πξ ,
2. u(0, t) = 0 ,
3. u0ξ (1, t) = 0 ,
4. u < ∞ .
4.2. EXAMPLES OF 1D NON-STATIONARY MODELS 47
We will modify formerly introduced Fourier method for this type of PDE. Setting u(ξ, t) =
w(ξ).v(t) into the equation leads to the identity
w00 1 v0
= = −α2
w κv
that must hold on the whole region Ωx,t . The double of ordinary differential equations results
as the consequence, where the negative sign follows from the condition 4 (solution boundness):
Now, we apply the last condition No. 1 to determine the coefficients Bn by solving the equation
∞
X
sin πξ = Bn sin αn ξ . (4.27)
n=1
Multiplicating the both sides of (4.27) by sin αm ξ, the integration from 0 to 1 gives
8 (−1)n
Bn = . (4.29)
π (2n − 3)(2n + 1)
The figure 4.1 demonstrates initial condition (full line) and the solution at three other time
points.
4.2.3 Example – heat conduction in a thin finite rod II. Unlike previous problem the
situation is described by rather modified assumptions.
1. Start of the heat exchange process: thin rod of the length L has constantaneous tempera-
ture T0 that is the same as outside one.
40
35
teplota [ oC ]
30
25 t = 0
t = 6
t = 18
t = 60
20
0 0.2 0.4 0.6 0.8 1
ξ
Figure 4.1: Time development of temperature profile (the legend data in seconds).
We aim to find local formulation to estimate the temperature T (x, t) at arbitrary point in
x ∈ (0, L) for the time t > 0. At first we emphasize several important atributes of this problem:
• as the rod is thin, the instantaneous temperature is the same in the entire cross-section →
one-dimensional model can be considered;
• the assumption 4 will be realized by introduction of source term depending on state func-
tion T (x, t):
α
f (x, t, T ) = (T − T0 ) , x ∈ (0, L), t > 0 ; (4.30)
L
• λ = konst. [WK−1 m−1 ] because of homogenity of the rod.
We use differential equation (4.15) supplied by source function (4.30):
∂T ∂2T α
− λ 2 = (T − T0 ) ,
cρ x ∈ (0, L), t > 0 . (4.31)
∂t ∂x L
Thus, the heat capacity c [Jkg−1 K−1 ] and density ρ [kgm−3 ] need be added to input data of our
model. Note, that all terms in obtained non-homogeneous PDR are of the same physical nature
– heat flux density [Wm−3 ].
Initial condition is prescribed by the assumption 1, i.e.
T (x, 0) = T0 , x ∈ h0, Li . (4.32)
Without loss of generality we place the assumption No. 2 into the point x = 0; and, we write
the No. 3 at the point x = L. The first one implies directly the Dirichlet condition
T (0, t) = T1 , t>0, (4.33)
the second one is represented by homogenneous Neumann condition
∂T (x, t)
=0, x = L, t > 0 . (4.34)
∂x
Performed model consists from the equation (4.31) with conditions (4.32)–(4.34). Again, it is
possible to solveit using Fourier method.
4.2. EXAMPLES OF 1D NON-STATIONARY MODELS 49
4.2.4 Example – heat conduction in a thin long rod. Very long thin rod has the initial
temperature T0 . At certain time (t = 0), it will at one end-point heated up to temperature
T1 > T0 that will be conserved. We shall determine the temperature at arbitrary point and
time.
Unlike previous problem no source term arises here that means simple equation
∂T ∂2T
cρ −λ 2 =0 , x ∈ (0, ∞), t ∈ (0, ∞) . (4.35)
∂t ∂x
Suppose that the Dirichlet condition is located at the point x = 0,
we see, that second end-point is not specified. Since the rod is very long, the straight semi-line
x > 0 can be considered as the interval, where we seek for solution. Therefore, in agreement
with physical principles, the boundness of solution toward infinity can be considered, especially
T (x, t) = T1 , x → ∞, t → ∞ (4.37)
with initial condition
T (x, 0) = T0 , x ∈ h0, ∞) . (4.38)
The formulated problem can be solved advantageously using Laplace transform. Here, we
apply quite specific approach based on appropriate substitution. At first, we simplify our formu-
lation to have at least one homogeneous boundary condition. If we set u(x, t) = T −T1 , a = λ/cρ,
then
1 ∂u ∂ 2 u
− 2 =0, x ∈ (0, ∞), t ∈ (0, ∞) , (4.39)
a ∂t ∂x
u(x, 0) = −(T1 − T0 ) = −ϑ ,
u(0, t) = 0 , (4.40)
u(x, t) = 0 , x → ∞, t → ∞ .
is dimensionless quantity; therefore, we use the above relation as the substitution that yields
∂u ∂u τ ∂2u ∂2u 1
=− , = .
∂t ∂τ 2t ∂x2 ∂τ 2 at
We get ordinary second-order differential equation
d2 u τ du
+ =0
dτ 2 2 dτ
with the conditions
Zx
2 2
erf(x) = √ e−z dz .
π
0
x
T (x, t) = T1 − (T1 − T0 ) erf √ . (4.41)
2 at
Following results were calculated with the temperatures T0 = 20◦ , T1 = 120◦ by a = 0.02 m2 s−1 .
Particularly, the temperature profile after 30 seconds is illustrated; and, time dependence of the
temperature during first 30 seconds at x = 0.5 m from heated end-point (Fig. 4.3).
4.2.5 Vibration of a string. Let we derive non-stationary boundary problem that describes
vibration of elastic fibre (string). Remember local form of stationary state balance of instan-
taneous deviation u(x) by stress force T (x) and external forces density f (x) in the paragraph
2.2.4:
d du
T (x) = f (x) , (4.42)
dx dx
4.2. EXAMPLES OF 1D NON-STATIONARY MODELS 51
where ρ(x) and a(x, t) are the mass density and the acceleration, respectively. Obviously, the
result corresponds to the second Newton law F = ma. Finally, the local equation of free
oscillations of string follows as
∂ ∂u(x, t) ∂ 2 u(x, t)
T (x) = ρ(x) . (4.43)
∂x ∂x ∂t2
The classical wave equation results by constant stress T ,
s
∂2u ∂2u T
c2 = , c= , (4.44)
∂x2 ∂t2 ρ
In this model, two initial conditions are required reagarding second-order time derivation of
state function. The first one characteries initial deviation,
∂u(x, 0)
= ψ(x) , x ∈ (0, L) . (4.47)
∂t
The wave equation (4.44) with boundary conditions (4.45) and initial conditions (4.46), (4.47)
constitute full formulation of non-stationary problem for free non-damped string vibrations. To
solve this one, the Fourier method can be used.
4.2.6 Example – oscillation of a string. Solve the problem for non-damped vibration of
the string of the length L = 1 m and stiffness c = 0.3 m/s by initial deviation
!
1 x2
ϕ(x) = x−
50 L
and initial velocity ψ(x) = 0. Demonstrate graphically instantaneous deviation at the time 1,
3, 5 and 7 seconds.
4.2.7 Typology of second-order PDE’s. For simplicity, we restrict the theory in this
subsection to linear problems with no more than two variables that we denote x and y without
physical interpretation. Partial differential equation of second order for the function u(x, y) can
be written as
The symmetric matrix A represents quadratic form related to a conic by value of δ. In this
sense we denote corresponding PDE in agreement with following definition.
If δ has stable sign on the domain Ω, then partial differential equation of second order is called
on this domain as
• parabolical for δ = 0,
Examples of classification:
4.2.8 Free advection. Suppose a very long stream, for which the flow can be considered
as one-dimensional with the velocity v = v(x, t). Concentration of a pollutant is described by
continuous function c = c(x, 0) = c0 (x) at the time t = 0. Due to water flow, the transfer of
contaminant is in progress by free advection. The other pollution sources of the local density
q(x, t) can also take the place. Our aim is to constitute balance equation to estimate of conta-
mination c = c(x, t) at arbitrary point.
At first we write global concentration balance of pollutant in the time interval ht1 , t2 i on a
flow section hx1 , x2 i:
Zt2 Zt2 Zx2 Zx2
[c(x2 , t)v(x2 , t) − c(x1 , t)v(x1 , t)] dt = q(x, t) dx dt − [c(x, t2 ) − c(x, t1 )] dx . (4.51)
t1 t1 x1 x1
The left side express total change of the flow function c.v, the right side can be considered as
a production term, where w(x, t) = c(x, t). The analogy to the general global balance (4.2) is
obvious. Since both the concentration and flow velocity are continuous in the time, we divide
the equation by ∆t and apply limit process ∆t → 0 that yields
Zx2 Zx2
∂c
c(x2 , t)v(x2 , t) − c(x1 , t)v(x1 , t) = q(x, t) dx − dx . (4.52)
∂t
x1 x1
Now, we reply this step for the variable x that results in the first-order PDE
∂ ∂c ∂c ∂v ∂c
(c.v) + = q(x, t) i.e. v+c + = q(x, t) . (4.53)
∂x ∂t ∂x ∂x ∂t
4.2.9 Free advection – model problems. Solve the equation (4.53) without internal
sources (q = 0) for initial state
cmax
c0 (x) = , k>0.
cosh kx
(a) By constant flow velocity v = v0 = const.
(b) By time-dependent velocity
v0
v(t) = .
1 + α2 t2
(c) By the velocity being the function of space variable x as
v(x) = v0 e−kx .
In all these cases the method of characteristics can be used – see the next subsection.
54 CHAPTER 4. NON-STATIONARY MODELS
(a) y ∂u ∂u
∂x − x ∂y = 0 ... linear equation,
(b) y ∂u ∂u
∂x − x ∂y = xyu
2 ... quasi-linear equation,
2
∂u
(c) y ∂x − x ∂u 2
∂y = x + y
2 ... non-linear equation.
The equation (a) is homogeneous because of zero-valued right side. In the opposite case (a
function g(x, y) 6= 0 as r.-h. term) it is non-homogeneous.
As the solution of the 1st-order PDE on the domain Ω we call any function u(x, y) fulfilling
the equation and having continuous derivatives. The last property can be written as u ∈ C1 (Ω).
4.3.2 Method of characteristics for homogeneous linear equation. This equation type
has the general form
∂u ∂u
P (x, y) + Q(x, y) =0, (4.55)
∂x ∂y
where the coefficients P, Q are continuous on a domain Ω, where we seek for solution, i.e.
P, Q ∈ C(Ω). Any solution u(x, y) represents an integral surface, equiscalar lines of which are
called as characteristics of the form
ϕ(x, y) = C . (4.56)
Thus, ϕ(x, y(x)) = C are their projections into the z = 0 plane, which we will differenciate:
ϕ0x dy
ϕ0x + ϕ0y .y 0 = 0 , ⇒ 0
= −y 0 = − . (4.57)
ϕy dx
where G(ϕ) is arbitrary differentiable function. This fact can be verified by putting it into
original equation:
The expression in the last brackets is equal to zero in agreement of problem formulation that
implies correctness of the statement.
It remains to show, how to derive particular function G with regard to given conditions. We
can demonstrate them by the following way: the graph of solution u(x, y) is passed through
prescribed curve of parametrization x = x(τ ), y = y(τ ), z = z(τ ), τ ∈ hτ1 , τ2 i. Substituting x
and y in (4.56) the relation ϕ(τ ) = C is obtained, from which we express τ as a function of ϕ.
Finally, u = z(τ ) is the particular solution – see the following problem as an illustration.
4.3.3 Homogenneous equation – an example. We will seek for the solution of the equa-
tion (a) in 4.3.1 that is passed through
1) the parabola x = τ, y = τ, z = τ 2 , τ ∈ h−1, 1i,
2) the circle x = 1, y = cos τ, z = sin τ, τ ∈ h0, 2πi.
In the given equation yu0x − xu0y = 0 we have P (x, y) = y, Q(x, y) = −x. Characteristic equation
follows as
dx dy
=− , i.e. ydy = −xdx .
y x
General solution x2 + y 2 = C represents the characteristics, therefore, the general solution can
be written as
u(x, y) = G(x2 + y 2 ) .
ϕ(τ ) = 2τ 2 ;
4.3.4 Existence and uniqueness of solution. For the problems of above type, following
general formulation is used, named as Cauchy problem or initial problem for a first-order
PDE:
∂u ∂u
P (x, y) + Q(x, y) =0, u(x(τ ), y(τ )) = z(τ ), τ ∈ hτ1 , τ2 i . (4.61)
∂x ∂y
The Cauchy uniqueness theorem (without a proof):
Let P (x, y), Q(x, y) ∈ C1 (Ω), Ω ⊂ R2 . The curve K parametrization (x(τ ), y(τ ), z(τ )), τ ∈
56 CHAPTER 4. NON-STATIONARY MODELS
hτ1 , τ2 i is given so, that the plane curve k = {x(τ ), y(τ )} lies in Ω. If one holds for any
τ ∈ hτ1 , τ2 i in the Cauchy problem (4.61) that
P (x(τ ), y(τ )) Q(x(τ ), y(τ ))
=6 0,
ẋ(τ ) ẏ(τ )
then a subdomain Ω∗ ⊂ Ω exists containing the curve k, where the solution is unique.
We show an application of this theorem to the example 2) in 4.3.3. The related determinant
is of the form
τ −τ
= 2τ .
1 1
It means that obtained solution is unique in arbitrary domain, which does not contain the point
τ = x = y = 0, i.e. the vertex of parabolic surface.
∂c ∂c v0
v(t) + =0, v(t) = , (4.62)
∂x ∂t 1 + α2 t2
where v0 , α are given constants. Initial condition is introduced by the function
cmax
c(x, 0) = c0 (x) =
cosh kx
with the input data cmax and k > 0.
At first the characteristic equation need be derived, where P (x, t) = v(t) , Q(x, t) = 1:
dx
= dt, i.e. dx = v(t) dt ,
v(t)
x = τ, t = 0, c = c0 (τ ) .
A short discussion can be added. The maximum of concentration at fixed time tk will located
at the point, where the denominator is minimized, it means at
v0
x= arctan αtk .
α
Since v(t) → 0 as t → ∞, this process ceases by time increase. Finally, the concentration will
stationarized having the maximum at the point
v0 v0 π
xs = lim arctan αt = .
t→∞ α α 2
4.3.6 Method of characteristics for quasilinear equation. We seek for a function that
satisfy the equation
∂u ∂u
P (x, y, u) + Q(x, y, u) = R(x, y, u) (4.65)
∂x ∂y
with P, Q, R ∈ C(Ω) ⊂ R3 , the characteristics of which is supposed in the form
ϕ(x, y, u(x, y)) = C .
We express partial derivatives by the both variables x, y:
Now, we add the first equation multiplied by the function P (x, y, u) to the second one multiplied
by the function Q(x, y, u):
P ϕ0x + Qϕ0y + ϕ0u (P u0x + Qu0y ) = 0 .
Since the bracketted term is equal to R(x, y, u) by (4.65), we can write
P ϕ0x + Qϕ0y + Rϕ0u = 0 .
Further, we differenciate equation of characteristics ϕ(x, y, u) = C:
ϕ0x dx + ϕ0y dy + ϕ0u du = 0 .
The last two relation hold together only if one holds
dx dy du
= = . (4.66)
P (x, y, u) Q(x, y, u) R(x, y, u)
We obtained three characteristic differential equations, from which it suffices to solve arbitrary
couple that yields two different characteristics
ϕ(x, y, u) = C1 , ψ(x, y, u) = C2 . (4.67)
Putting them into quasilinear equation (4.65) we can prove, that general solution is of the form
G(ϕ, ψ) = 0 , (4.68)
where G is a function of two variables differentiable on Ω.
A Cauchy problem is formulated in the same way as for homogeneous equation. The solving
process should be as follows:
58 CHAPTER 4. NON-STATIONARY MODELS
1. derive the characterics of the form (4.67) by the above presented procedure;
ϕ(τ ) = C1 , ψ(τ ) = C2 ;
G(C1 , C2 ) = 0 ;
dx dy du
= = .
x y u
2. The circle of radius 2 in the y = 1 plane is given curve (bold dashed in the Fig. 4.4) with
parametrization
x = 2 cos τ, y = 1, z = 2 sin τ, τ ∈ h0, 2πi .
The matrix
P (τ ) Q(τ ) R(τ ) 2 cos τ 1 2 sin τ
=
ẋ(τ ) ẏ(τ ) ż(τ ) −2 sin τ 0 2 cos τ
is of the range 2 on the whole interval h0, 2πi, the solution will be unique.
y 1 u sin τ q
C1 = = , C2 = = ⇒ C2 = 4C12 − 1 .
x 2 cos τ x cos τ
4.3. PARTIAL DIFFERENTIAL EQUATIONS OF THE FIRST ORDER 59
∂c ∂c
v(x) + = −v 0 (x)c ,
∂x ∂t
we find out that it is quasilinear. From the system
dx dt dc
= =−
v(x) 1 v(x)c
dx R dx
v(x) = dt → ϕ(x, t, c) = t − v(x) = K1 ,
dx dc
v(x) = − v(x)c → ψ(x, t, c) = v(x)c(x, t) = K2 .
60 CHAPTER 4. NON-STATIONARY MODELS
In partiular,
1 1 kx
Z
ϕ=t− ekx dx = t − e ,
v0 v0 k
cmax 2v0 cmax
ψ = v0 e−kx = 2kx .
cosh kx e +1
The parametrization of initial condition can be introduced as x = ξ, t = 0, c = c0 (ξ). Thus
2v0 cmax
ekξ = −v0 kK1 , = K2 .
e2kξ + 1
Substituting ekξ from the first relation into second one yields
2v0 cmax
K2 − 2 2 2 =0.
v0 k K1 + 1
Finally, we put ϕ, ψ instead K1 , K2 ; a short rearrangement leads to the solution
2cmax ekx
c(x, t) = .
(v0 kt − ekx )2 + 1
Validity of result can be easily verified.
There is of interess to review obtained solution for particular numerical data. Suppose that
emission of the concentration c(x, t) is abducted by wind of the velocity 40 km/h that ceases
with increasing distance from the point of its origin with the parameter k = 0.05 km−1 . At first
we estimate, when maximum of contamination shows itself at the distance xm = 10 km. This
extreme comes on by v0 kt − ekx = 0, therefore,
1 kxm
tm = e = · · · ≈ 49, 5 min.
kv0
Corresponding concentration of pollution results as
c(xm , tm ) = cmax ekxm = 1, 65 cmax > cmax .
Obviously, the maximum of contamination increases at a place with lower advection velocity by
comparison with initial maximum. Theoretically, this process is in progress by growing spacing
from the point of origin. However, this is only a consequence of quite simplified model.
4.3.9 Heat advection model. A liquid flows through very long narrow pipeline by following
conditions:
(a) flow velocity is invariable;
(b) the process is adiabatical;
(c) heat transfer proceeds in the surrounding of the constant temperature T0 on the surface of
pipeline;
(d) a temperature change of running media takes effect instantaneously in the entire pipeline
cross-section; therefore, the heat transfer coefficient [α] = WK−1 m−3 relates to the volume
unit;
(e) input medium temperature T1 > T0 is constant as well as its density ρ and specific heat c;
(f) initial temperature in the entire pipeline is equal to the surrounding temperature.
Find the solution using method of characteristics by following data:
α = 700 WK−1 m−3 , c = 4186 Jkg−1 K−1 , ρ = 1000 kgm−3 , v = 0, 2 ms−1 , T0 = 5◦ C, T1 = 85◦ C.
Demonstrate the graphs of
• temperature profile in the pipeline from the input point to the distance 250 m at the time
tk = 15 minutes,
• time course of temperature at the point xp = 200 within the first twenty minutes.
4.4. NON-STATIONARY MULTI-DIMENSIONAL MODELS 61
4.3.10 Excersises.
1. Find the solution and postulate the uniqueness condition:
∂u ∂u
+ y2 =0, x = −1, y = τ, z = τ + 1, τ ∈ R .
∂x ∂y
xy+2y+1
u(x, y) = xy+y+1 , y 6= 0
2. yu0x − xu0y = 2xy, u(x, x) = 2x2 − 1, u(x, y) = 32 x2 + 21 y 2 − 1
3. 2yu0x − xu0y = 4xyu2 , u(x, x) = 1/x2 , u(x, y) = 3
4y 2 −x2
4. Show that solving of the examples 4.2.9a,b can be generalized as an use of the substitution
Zt
ξ = x − vt or ξ =x− v(τ ) dτ .
0
(A) Boundary conditions (P ∈ ∂Ω, t > 0) by given functions p(P, t) and g(P, t):
∂u(P, t)
p(P, t) + α u(P, t) = g(P, t) . (4.77)
∂n
∂u(P, t)
p(P, t) = g(P, t) ; (4.78)
∂n
3. Thermal conductivity: λ1 inside the segment ABEF, λ2 in the neighboring one; cor-
responding temperature conductivity a1 , a2 given as well.
5. Heat transfer into outside medium with the coefficient α on the rest of of boundary.
F E D
A B C
Let Ω1 , Ω2 denote open squares ABEF and BCDE, respectively. Coordinate system with
the origin A is introduced in this way that the rectangle is located in the first quadrant and the
vertex F lies on the y axis. We will model given situation as a couple of boundary problems for
temperature fields T1 (x, y, t), T2 (x, y, t) related to the domains Ω1 , Ω2 . A coupling on common
interface will be expressed by transient conditions.
Using formerly obtained knowledge of local balance equations (4.16), (3.38) and (4.76) we
can write regarding assumptions 2 and 3
∂Tj (x, y, t)
aj ∆Tj (x, y, t) = on Ωj , j = 1, 2 . (4.82)
∂t
Also the initial condition is the same for the both regions (see 2):
Tj (x, y, t = 0) = Tp , j = 1, 2 . (4.83)
Since any boundary segment is an abscissa parallel to a coordinate axis, the normal derivatives
in (4.77), (4.78) simplify as derivatives by single space variables (see the paragraph 3.3.3, for
instance). Homogeneous Neumann conditions take place on each isolated boundary part; the
Newton conditions hold on the sides CD and DE. The transient conditions on the boundary BE
contains besides usual flux continuity also equality of temperatures by the assumption 6:
∂T1 ∂T2
T1 = T2 , λ1 = λ2 , x = a, y ∈ h0, ai . (4.84)
∂x ∂x
4.4.4 Diffusion with advection. This combined phenomena represents typical situation,
where variability of model approaches can be showed for non-stationary multi-dimensional prob-
lem. Our aim is to derive global and local balance by atmosphere contamination on the region
Ω – see Fig. 4.6. The contamination level is expressed by dimensionless concentration of a
pollutant, c(M, t), M ∈ Ω, t > 0. The soil transfer is realized partly by advection (air flow)
of the velocity v(M, t), partly as a diffusion with the coefficient k(M ). Thus, the flux function
q(M, t) is considered in the form
q = vc − k grad c . (4.85)
Further, let f (M, t) is the density of inner pollution sources (as local hearths etc.).
The mass conservation law will be the basic balance relation, where we take the injurant
concentration in volume unit c(M, t) as the state quantity. We express the fact that production
of internal sources is equal to the sum of pollution exchange with surrounding through the
64 CHAPTER 4. NON-STATIONARY MODELS
Having the global balance, we continue toward local one. At first, we apply the Gauss-
Ostrogradsky theorem to the left-hand side:
∂c(M, t)
Z Z Z
div (vc − k grad c) dV = f (M, t) dV − dV . (4.87)
∂t
Ω Ω Ω
Selected problems
5.1.1 Reproduction with migration. Constitute the problem describing time variability
of the number of residents in a closed region as a consequence of reproduction and/or migration
processes. Take into account move in, move away, nativity and deaths. Characterize the problem
type and suggest method to solve.
Let x0 denotes initial number of residents and x(t) their number at a time t. Futher, let
∆x(t) be the change corresponding to time interval ∆t. A population progress caused by move
in and away will be incorporated into common function f (t). In order to describe reproduction
processes depending on time as well as on instantaneous number of residents, we introduce
functions α(x, t) and β(x, t) for the number of born or dead residents in the time section ∆t.
In- or decrease of inhabitants ∆x in the time section ∆t can be balanced with the help of
difference equation
∆x(t)
= α(x, t) − β(x, t) + f (t),
∆t
where α, β are non-negative functions. Supposing |∆x| << x, |∆t| << t we can set time
variance of resident number as continuous that leads by well-known steps to the differential
equation
dx(t)
= α(x, t) − β(x, t) + f (t).
dt
This one constitutes first-order Cauchy problem with initial condition x(0) = x0 . Generally,
it is correctly formulated problem, but solvable only using numerical methods. An analytical
approach can be used only if the functions α, β are of favourable form.
As the simplest example we present linearized model, where
At first, we consider isolated system (f (t) = 0), for which we obtain separable linear equation
that can be easy resolved:
dx
= (k1 − k2 )x , x(0) = x0 ⇒ x(t) = x0 e(k1 −k2 )t .
dt
66
5.1. POPULATION AND LOGISTIC MODELS 67
5.1.2 Reproduction with migration - open linear system. Solve the previous problem
with non-zero migration term
f (t) = xm sin ωt, xm << x0 .
Find the solution by following conditions: time unit 1 day, x0 = 6000, xm = 5, ω = 0, 1 1/day.
Discuss the result regarding three above variants. Calculate and demonstrate graphically the
solution for the first three years by k1 = 1, 18.10−3 [1/day], k2 = 1, 15.10−3 [1/day]. Estimate
the number of inhabitants at the end of given time interval.
5.1.3 The model of prey–predator system. A closed natural system contains (in a re-
duced sense) the population of predators and prey population that serves as a food. Constitute
linear model of (continuous) time progression of the both populations with regard to following
activities:
• reproduction abibility of any population,
• prey decrease due to predator activity,
• increase of predator population as a consequence of food sufficiency.
Consider velocity of individuals number change as modeled quantity. Thus, resulting time
courses result as integrals of these one with respect to time variable. Solve the problem with
constant coefficients given as follows:
a = 0.05, d = −0.05 for reproduction ability of prey and predator population, respectively,
b = −0, 02 for prey decrease due to predator activity,
c = 0, 17 for increase of predator population.
Initial population amounts are given as 600 (prey) and 10 (predators); initial velocities of state
changes related to one day are prescribed as 0.2 by prey and 0.02 by predator population.
Solve the problem in the one-year time section. Illustrate graphically the course of population
progress and characterize the system from stability point of view.
5.1.4 Advertising effectivity. A product has N potential consumers, but only N0 of them
is informed about it. At this time, advertising campaigne is started thereby the propagation
velocity of trade references (i.e. the number of informed consumers in a time unit) is direct
proportional to geometric mean of actually informed and non-informed consumers. Formulate
mathematical model to estimate effectivity of advertising (take one day as the time unit), if N7
of consumers is acquainted with the product at the first week end. Solve by the following input
data: N = 5000, N0 = 50, N7 = 737.
Let n(t) denotes number of informed consumers at the time t. The basic balance relation is
obvious:
∆n(t) q
= k n(N − n) ,
∆t
where k is a constant that we determine later. As ∆n << N for small increment ∆t, we can
introduce initial problem for differential equation in the form
dn(t) q
= k n(N − n) , n(0) = N0 . (5.1)
dt
68 CHAPTER 5. SELECTED PROBLEMS
5.2.2 Movement of charged particle in magnetic field. A particle of the mass m with
electric charge Q is emitted into homogeneous magnetic field under the angle α to flux lines.
Constitute mathematical model to calculate the particle trajectory. Discuss obtained solution
with regard to the angle α.
5.2.3 Water outflow a vase. Consider a vase of the form arrising by rotation of the curve
y = S(x) around the x axis. The water flows freely away through a small outlet close to the
bottom with the velocity p
v = k 2gh
(value of the constant k is 0.6 for the water).