An Analytic Basis for Heat Conduction in a Fuel Rods

Dan Shields
December 15, 2011
Introduction

A Simple Fuel Rod

It is important in nuclear science and engineering to understand heat flow in and around
fuel in reactors. This is still an open area of research, although it is well understood for most
commonly used systems, solutions can be drastically changed for only subtle changes in parameters. This paper attempts to take a very
simple approach to the fist steps in building up
this theory of reactor fuel. I will apply the
concepts of reactor theory from E. Lewis text
on the subject[1], along with A. Fetters and J.
Waleckas text on classical mechanics [2].
First I will give the background necessary for
understanding the results. This will include a
brief look at the basic solution for a nuclear fuel
rods energy production distribution, a review of
thermal physics and extrapolation of a solution
to the temperature distribution of the fuel over
time.
As the equations governing this process for
nontrivial energy production inside our the fuel
is highly complicated and almost impossible to
solve for by hand analytically, I will only derive
the first principals that then could be used to
generate a numerical solution for the process.
This is exactly what modern models are built
on, so it is very useful to understand where they
come from.

Figure 1: Simple fuel rod w/ extrapolated H,R

and R)

(H

Simple Neutronics in a Rod

Lets start by looking at a cylindrical fuel rod
that is floating freely in space. This allows us to
say that all neutrons that leak out of the rod can
never return by reflection of nearby materials.
This is a good approximation to the behavior of
the rod in some materials as well, but will fail for
highly reflective or neutron producing materials.
We then can use the terminology described in
[1, pg. 167] to formulate a differential equation
that relates the scalar flux ((~r) the velocity averaged neutron distribution) to the macroscopic cross section of our material (f and a )
under Ficks approximation1 .
1

(see next page)

This solution has an arbitrary amplitude (C)

that is, by dimensional analysis, proportional to
the power of the rod (P ). This solution can then
be used to find the total power of the rod by
multiplying the scalar flux by the cross section
for fission (f ) and the energy per fission () and
then integrating over the volume of the fuel:

~ D (~
~ r) + f (~r)(~r) a (~r)(~r) = 0 (1)

Where is the average number of neutrons

produced per fission, and D is the diffusion coefficient (D can be a function of space, and is
dependent on factors of the macroscopic cross
section it is derived in [1]). Now if we provide
that our system is spatially uniform throughout
its volume, we can simplify this equation:

2 (~r) + B(~r) = 0 ; B =

k /k 1
L2

Z
f dV =

P =
V

H
2

f dzdr

(5)

H
2

Solving for C after integrating (shown in [1])

gets us our final function of :

(2)

Where k is the neutron multiplication factor,

P
z
2.405 r
(r, z) = (3.63
) J0 (
) cos (
)

k = f /a is the same factor for an infinite

f V
R
H
|
{z
}
dimensioned problem with the same materials,
C
(6)
and L2 = D/a is the diffusion length squared of
Where V is the total volume of the rod. Ala neutron in the fuel. Now we impose cylindrical
though, that we really want is the power denboundary for our ODE and find:
sity that is simply the non-integrated version of
power.Now that we know C, we can solve for the
2
000

r (r, z) + 2 (r, z) + B 2 (r, z) = 0 (3) power density (P ) at any point:
r r
z
This can be solved by separation of variables
so that (r, z) (r)(z) as the rod is cylindrically symmetric. The derivation is spelled out
in [1], but I will simply state the result 2 :
(r, z) = CJ0 (
1

2.405 r
z
) cos (
)

R
H

P 000 (r, z) = f (r, z)

2.405 r
z
P
= (3.63 J0 (
) cos (
)

V
R
H

(7)

This will be highly useful in our analysis of

the fuel, as it tells us where the sources of heat
(4)
are located. It is critical in development of a
solution to the conduction of heat over time.

This approximates the scattering processes that neutrons go through as a diffusion process that is directly
analogous to something like salt diffusing in water. See
Ficks Law.
2
Note: The boundary conditions are set to 0 at the
H
. This gives good approxiextrapolation lengths R,
mation as to the actual at the true boundaries R, H.
This is described in [1]. Also the number in our equation
comes from J0 (2.405...) = 0.

 Review of Thermal Physics

variable volume. For solids, this can be misleading and harder to solve, and one can restate
We aim to get to how heat is conducted in
this as just the opposite. This retains the same
this rod now, but first we must go over some of
meaning (p v = W ). So to make our equathe basic concepts from thermodynamics. The
tion more functional we introduce a new equamost important concepts being:
tion from a Legendre transformation of eq.8 to
find Enthalpy (H) :
The Laws of Thermodynamics:
(9)
dH = d(E + pv) = T
dS} + v dp
1st : The internal energy of a system is
| {z
|{z}
, dQ
, dW
equal to the amount of heat supplied
Now to further simplify we define a parameter
to the system, minus the amount of
work performed by the system on its for our system in terms of constant pressure: the
surroundings. (A statement of energy heat capacity:
conservation)
Q
S
H
2nd : This is an expression of the tendency
Cp = (
)p = T ( )p = (
)p
(10)
T
T
T
that over time, differences in temperature, pressure, and chemical potential
This implies that we can integrate H over a
equilibrate in an isolated physical sys- range of temperature of find the change in H:
tem. (Entropy (S) is always increasing
for spontaneous processes)
Z
T

dT 0 Cp (T 0 , p)

H(T, p) = H(T0 , p) +

(11)

T0

This can be stated quantitatively described

for a tiny element of volume in our system as
In most solids that are far from a phase tran(see [2]):
sition point, Cp is only weakly dependent on T .
This allows us to make the approximation:
(8)
dE = T
dS} p dv
| {z
|{z}
, dQ

, dW

H(T, p) H(T0 , p) + Cp0 (T T0 )

Where dE is an increase in internal energy

for an elementary reversible process (aka: it fits
with the 2nd law), T and p are the temperature
and pressure locally around and in our element,
dS is a change in entropy, and dv is a small
change in volume. (Also dQ is a tiny amount of
heat added, and dW is a tiny amount of external
work done)of our element.

(12)

Cpo = Cp (T0 )

Now we have a complete set of functions that

we can use to solve our tiny volume element of
our system. We only need to integrate this over
the entire volume now to obtain the full entropy
(Htot ) to study the effects of energy addition to
Although this equation implies that our sys- the entire system.
tem can be described as at a fixed pressure and
First we must replace our T a function depen3

dent on the position and time in the material:

An Example to Warm Up
T T (~r, t); and our heat capacity with a more
As we will see, the fuel rod problem will
convenient form: Cp0 cp d3 r. This then can
present a rather nasty PDE for us. So for now,
be used to construct:
lets start with and even simpler model for the
rods neutronics: only radial dependence in P 000
Z
for the rod. We will model this function as a simcp [T (~r, t)T0 ] d3 r+Htot (T0 , p) (13)
Htot =
ple point source of heat inside a sphere. This will
V
give us incite as to what our full problem might
Now we can use the analysis in [2] to relate look like [2]:
the rate of change in enthalpy to the transfer of
P 000
heat within the media. This is done primarily
(16)
P 000 (r) = 0 (r)
cp
on a graphical basis to form:
So our equation for the temperature of our
rod is:
Z
T (~r, t) 3
Htot
=
cp
dr
t
t
V
Z
P0000
T (~r, t)
2
=

T
(~
r
,
t)
+
(r)
(17)
~
~
= dA jh + q
(14)
t
cp
cp

T (~r, t)
~ j~h
=
t

We also enforce that at the start of time, the

temperature is uniformly zero, and the tempera~ is the a tiny element of area of a ture at the outer radius (a) is alway held at zero
Where dA
full shell surrounding the system (or part of the Thus we say:
system) , j~h is the heat current through through
that area, and q is the energy input into the sysT (r = a, t) = 0
(18)
tem per unit time. Now from [2], we say that
T (r, 0) = 0
empirically one mostly find that j~h is directly
proportional to the the gradient of the temperaThis can be solved by using a Laplace transture. This makes some sense: that heat will flow
form to replace the time dependence:
into cold area much more than other hot areas.
Z
Thus:

T(r, s) =
dt est T (r, s)
(19)
0

q
T (~r, t)
= 2 T (~r, t) +
(15)
t
cp
kth
Where : =
; kth is the thermal diffusivity
cp

So it we apply this to our PDE, we get:

s) = 2 T(r,
t) +
sT(r,

000

P0000
(r)
cp s

(20)

But this is exactly what we want: q P

from our neutronics solution, so we get a PDE
Now if we look at r > 0 we can simplify this
for T as a function of power produced by fission! expression to:
4

1 2 s
r
T T=0
r2 r r

(21)
s) =
T(r,

P0000

sinh s 2 (0 )

(26)
1
1
4cp s 2
sinh 0 s 2
This gives us an inclination as to go about
solving the problem by integrating over space
Now we need to convert back to time to undo
around the point source of heat to get a bound- our Legendre transformation in t, thus:
relaary condition at the origin for our new T
tion:
Z
1
P0000
ds st sinh s 2 (0 )
e
T (r, t) =
1
1
4cp s 2 C 2i
sinh 0 s 2
Z
(27)
s) =
d3 r T(r,
(22)
s
This is not a trivial solution to find. This
Z r<
000 Z
becomes much easier if one uses complex contour
t) + P0
d3 r (r)
d3 r 2 T(r,
c
s
p
r<
r<
integration about the poles that appear for our
contour C (as seen in the above integral). We
We can then use a similar integration about
have a countably infinite set of these, so we gen
the equation describing the conservation of heat
an infinite sum (that does prove to be convergent
flow descried in the previous section. This will
for all t > 0):
enable us to find a true boundary condition:

Z
000
s) + P0 ]
lim[ = dA T(r,
0 r
cp s

1 1 X 2
nr n2 22 t
sin
e a ]
T (r, t) =
1 [( )
a
4cp s 2 r a n=1 nr
(28)
This then a superposition of sinc like functions
from 0 until their first zero. They all decrease
exponentially in time, meaning that over time
our temperature moves towards 0. This makes
some sense: we will see a function that starts
as something like an erfc(r) that has a zero at
the boundary and as a maximum temperature
at r = 0.

P0000

(23)

This can be evaluated using the divergence

theorem (much as it is used in classical EM):

lim 4
0

P 000
T (r, s)
+ 0 =0
r
cp s

(24)

This gives us a solid boundary condition for

the slope at the origin that we can write as (re1
placing r 2 ) :
( 2

s)
dT(r,
P0000
)=0 =
1
d
4cp s 2

(25)

In [2] we find a solution to this equation that is

a summation of complex Bessel functions of the
first and second kind. This can be simplified to
find:
5

Now in terms of our Laplace operator, we

must expand in cylindrical coordinates for each
As it turns out, the solution for a spatially of our T functions and find:
distributed function of power becomes extremely
hard to predict with analytic methods. As our

2.405 r
PDEs become more and more difficult to solve.
r Tr (r) + k 2 Tr (r) = J0 (
) (32)

r r
R
For our solutions from the first section, we can
2
z
2
use a basic analytic method to try and look at
T
(z)
+
k
T
(z)(r,
z)
=
C
cos
(
)
z
z

z 2
H
the fundamentals of the solution.
Solution for a Fuel Rod

From the first two sections, our equation for

This set of ODEs is not quite as easy as it
the temperature of our rod with the inclusion of might seem. The path forward is not clear,
our P 000 (r, z) solution is:
most especially for the Tr solution. I will not
attempt to solve these here, but refer you to
the attached Mathematica notebook. This will
T (~r, t)
= 2 T (~r, t)
(29) guide you through my derivation of the rest of
t
P
2.405 r
z
this paper.
+3.63
J0 (
) cos (
)

V cp
R
H
Taking for the solution from my notebook, we
This seems to be a prime target for separation find a solution that seems reasonable for the Tz
of variables as our P 000 function is already sepa- term:
rated in r and z and has no t dependence. Thus
we say:


z
Tz (z) = [(C H Cos

H


2 + (Hk
)(Hk
+ )T0
+ CH
2

T (~r, t) Tr (r)Tz (z)Tt (~r, t) = Tr (r)Tz (z) et

(30)

Cos[kz]Sec[Hk])]

(33)

+ ))
(Hk + )(Hk

The above equation being arrived at by

inspection (The time component is trivially
Where T0 is the maximum temperature at the
solved). We then can say the specific solution
center of the rod.
to the r and z dependence is solved by the folNow when we try and use Mathematica to
lowing ODEs:
solve for our latter solution, we get a strange
result. This is due to the fact that there is not
2.405 r
(2 + k 2 )Tr (r) = J0 (
)
(31) real method to give an exact solution for such an

R
ODE. This is unfortunate, but has not stopped
z
2
2
( + k )Tz (z) = C cos (
)
us from building reactors. This is due to the fact

H
that very accurate solutions can be estimated by
numerical methods to solve for this exact type
P

) ; k=
C = (3.63
of problem.
V cp

Conclusions
So from our analysis we have build up the ma- References
chinery needed to solve any general heat prob[1] E.E. Lewis. Fundamentals of Nuclear Reaclem. This becomes very difficult and not in the
tor Physics. NY: Academic Press, 2008, pp.
least trivial for most realistic energy inputs like
139-166.
our neutronics solution. Although the solution
for P 000 can be easy to solve and interpret, heat [2] A.L. Fetter, J.D. Walecka. Theoretical Meconducts very differently, and is highly dynamic.
chanics of Particles and Continua. NY:
We must now use the governing equations in
Dover Publications Inc., 1980, pp. 406-433.
a numerical solution to try and find a close fit to
actual heat conduction. This has been shown to
be possible, and there are many ways in which to
approach it. For the interested reader, they only
Attached: Mathematica pertinent to solution of Tz , Tr
need to search nuclear papers heat conduction of
fuel rods.