3.2 PreIimjnsries
3.3.3 Discussion and Examples
We have shown that the SSD in the case
g(x)
=gx
andb(x)
= bxl,C 00 (-l)lIak"e-�
y(x)
=x
2!
(ak - 1 ) ... (ab' - 1 ) 'where a =
�
and{ (k)l
•-1 )
1 ...a
C =(k)l
1
• , t =�, m
E 1+ , t #�,m
E 1+. (3.62) (3.63) (3.M)Continuity of C
In order to show that C is a continuous function of le, we need to prove that
lim
K(�,a)r(l
_ =(ak - 1)(a2k
. .
(a(lII-l)k - 1 ) K(�,l) loga.
(3.65)1 __ -1
m.
Now using (3.42) and separating out the one factor which becomes zero as le _
m-I
, we write00
K(ak, a)
=(1 -ai-h)
n(1 -
ai-kit).II=I,llPt
(3.66)Similarly, we separate out the factor in
r(1 - i)
which causes difficulties as le _m-I ,
by applying the iterative relationshipr(x - 1)
=��l m
times, givingr ( 1 _ !. -
t) -m - I (1 - i)(2 - i) . . . «m - 1) - i> " r(l - i + m)
Substituting these expansions into the left hand side of(3.65) we have
lim K(ak a)r(l - !.)
-lim
{ r(l - i
ai-kit)
1 __ -1 ' le-
1 __ -1(1 - )(2 - . . . «m - 1) -
m -
(3.67)
(3.68)
Now the limit of the expression in square brackets can be found by a single application of
l'Hopital's rule,
(3.69)
m
Substituting le =
m-I
into the rest of (3.68) then givesNow
lim
K(ak, a)r(l - !.)
= aI-�) log a1 ... 111-1 le
(-1
)",-1(m - 1 )(m -
2). . . 1 m
(3.70)
(3.71)
so after expanding the left-hand product then changing the summation variable in the
right-hand product to get it into the form required by (3.42) we get
00
n
(1 -
aI-�) = (-1j-l(� - 1
)(�- 1)
. . .(a(lII-l)k - l)K(ak,l ).
(3.72)1I=I,llPt
Substitution ofthis expression into (3.70) now gives (3.65), so C is a continuous function of le as expected.
Moments of �he SSD
When k is an integer, we can obtain the kth moment of the SSD about the origin, Ilk =
fooo xky(x)dx,
directly from the integral condition (3.27) as1
Pt. = a
(a - 1)"
(3.73)An iterative expression for moments beyond the kth is obtainable by applying a Mellin Transform to (3.20): multiplying through by .%"'-1 , where
m
is an integer, and integratingover all
x,
remembering thaty(x)
=x-2Z(x)
gives1000 ,t"-lZ'(x)dx
= -ap",+k + a1000 ak,t"+A:[x-2Z(ax)]dt.
(3.74)Integration by parts on the left and the substitution z = ax on the right then leads to (3.75)
so we have the iterative equation
(m - 1)
(3.76)provided
m
> 1. This enables us to deduce all the moments of the SSD once we have found the kth moment by using (3.73) and all other moments up to the (t + 1 )th by integration of(3.63) directly.
For example, in the particular case t =
1,
the mean is1
(3.77)Jl = a
(a - 1)
and the second moment about the origin
(Jl2
=fO'ry(x)dx = fooo Z(x)dt)
can be obtained byintegrating (3.25) term by term to give
1
Jl2 =
a2(a
- 1 ) log a ·We can then apply (3.76) repeatedly to obtain a"(
a
-1 ) loga
-
(3.78)
1 .2 ... . 8 x - >- .6 .4 .2 .5 2.5 x
Figure 3.1 : SSD probability density functions y(x) for g(x) = gx and hex)
=
hxk for a range of values of le. The variables a anda
have been fixed at1
and2
respectively.An Example
Figure 3.1 shows the shape of the SSD function y(x) with a = 1 and
a = 2
for a range of values of /c. Consideration of the differential equation formed in the extreme cases shows that as le -- 0, the SSD y(x) approaches the Dirac delta function 6(x), and as /c __ 00 itapproaches (Koch and Schaechter [30)) the function
y(x)
= {
11 a < 11 x <.Qo , OtherwISe.
3.4 A Minimum Size for Division: The Case
b(x)
=bxkH(x
-Xl)
This case is the same as that discussed in the previous section, except that there is a minimum sizeXl
> 0 below which cells do not divide. Substitutingb(x) = bxkH(x - Xl )'
k
> 0, (where H is the Heavyside or unit step function andb
andXl
are constants) into equation(3.14)
we getZ'(X) = xf-l ( -aH(x - XI )Z(X)
+a�H(ax -XI)Z(ax» ,
(3.81 )where as in Section 3.3 we have made the substitution
a =
�.
In this case we choose to couple (3.81 ) with the integral condition(3.16)
as this simplifies to a condition dependent only onZ(x)
forX
�Xl,
namely100
-21
;q xf
Z(x)dt = a(a - 1
r (3.82)It is convenient to solve the equation in three regions starting with the region
X
>Xl .
Welet
{ ZI(X),X
�Xl
(region1 )
Z(x) =
Z:a(x)
, � �X
�Xl
(region 2)Zs(x),x
� � (region 3)where continuity considerations require
ZI (XI) = Z:a(XI )
and Zs(�)= Z:a(�).
3.4.1 Region 1, x �
Xl
In this region, (3.81 ) becomes (3.20), so the solution is
00
( -1 )
" aD
!!!{t
ZI (X) =
C ,,� (a-t - l ) .
.. (aG
'lb find the constant C, we substitute
Z(x) = Zl (x)
into (3.82) above, givingc
100 [I
( -1
)"ab'
xl-2-
/l�l
dx _1
;q ,,=0
(a-t - 1) ... (aG - 1 )
e- a(a - 1) "
(3.83)
(3.84)
(3.85)
This time there is no difficulty in immediately interchanging the integral and summation operations as the integration starts at
Xl
> 0 and therefore avoids the singularity atX =
0when le � 1 . Making the change of variable z =
a�
in each integral leads to•
Hence we can write
(le)
1 1c - -- a
(a - l)S'
whereand r(c,x) =
f%oo
r-le-'ds is an incomplete gamma (unction.In the particular case le = 1 , (3.87) reduces to c =
a(a
le - 1 )Swhere
and El is the exponential integral.
8.4.2 Region 2, � :::; x < Xl (3.86) (3.87) (3.88) (3.89) (3.90)
In this region we have H(x -Xl ) = 0 and H( c:u -Xl ) = 1 80 we may immediately integrate (3.81 ) to obtain
(3.91 )
This integration can be carried out term by term, using the substitution
z = a�,
to give00 (-1 'f
[ _*
�(x) =
C,.!O
(at - 1 )... (abt
- 1 ) e - e ,where the constant C is the same as that in the expression for Zl (x) in (3.84).
3.4.3 Region 3, x < � In this region the (3.81 ) becomes
�(X) =
0
so as we have
Z3(�)
= ZJ(�) =0
from (3.92) the solution is simplyZa(X) =
0
as expected.3.4.4 Discussion and EXamples
(3.93)
(3.94)
Combining the results above and noting Z(x) = ry(x), the SSD in the case g(x) = gx and
b(x) = bxkH(x -Xl ) is given by (3.95) where (k) l
1
c- -
- a (a - l )S (3.96) and (3.97)Figure 3.2 shows this piecewise function for y(x) for a range of mjnjmum sizes for cell
division xlt with k = 2, a =
�
=1,
and a = 2. The discontinuity in the slope ofy(x) at x = Xlis a natural consequence of the step change in the probability of cell division b(x) (from
0
to
b�) at this size. The form ofy(x) for small values of Xl approaches the form given in Fig. 2.1 for k = 2.1 .2 ... .8 x ... >. .6 .4 .2 00 .5 1 .5 2 x
Figure 3.2: SSD probability density functions y(x) for g(x) = gx and b(x) = bxkH(x -Xl),
with k = 2, a = 1 ,