1 Introduction
In the next sections we shall describe the mathematical model of a set of particles which move freely as independent Brownian motions but may proliferate at random times with certain random time-dependent rates of proliferation.
Let us anticipate a few notations.
De…nition 1 We de…ne for a function f : R ! R f (t ) := lim
s%t
f (s); the left limit of f at t;
Jf (t) := f (t) f (t ); the jump size of f at t:
Further, let (X
t)
t 0be a stochastic process. We de…ne its quadratic variation, when the limit exists and is independent of the partitions,
[X]
t= P lim
k!1 n 1
X
j=0
X
tkj+1^t
X
tk j^t2
;
where the maximal distance of two consequent sites in the partition f0 = t
k0< t
k1< <
t
kn= T g converges to 0 as k ! 1. We denote the continuous part of the quadratic variation by [X]
c, i.e.
[X]
ct= [X]
tX
s t
JX
s2:
2 The model
Assume we have a …ltered probability space ( ; F; F
t; P ), stopping times T
0a;N< T
1a;N, independent Brownian motions B
at,
I
a;N:= [T
0a;N; T
1a;N) X
ta;N=
( if t = 2 I
a;NX
(a; );NT1(a; );N
+ B
taif t 2 I
a;Nand X
a;NT1a;N
= lim
t"T1a;N
X
ta;N. Or
X
ta;N=
( if t = 2 I
a;NX
a;NT0a;N
+ B
taif t 2 I
a;Nwhere X
a;NT0a;N
= X
(a; );NT1(a; );N
= lim
t"Ta;N1
X
ta;N.
It remains to specify T
1a;N, given T
0a;Nand the other objects. Assume we have, on ( ; F; F
t; P ) independent standard Poisson processs N
t0;at 0
and denote their …rst jump times by
0;a(N
t0;ais equal to zero for t <
0;aand equal to 1 at t =
0;a; recall that
0;a
> 0 with probability one). Assume we have continuous adapted non-decreasing process
a;N
t t 0
, equal to zero for t T
0a;N, with the following property: either
a;Nt<
0;afor all t, and in this case we set T
1a;N= +1, or there exists T
1a;Nsuch that
a;Nt<
0;afor t < T
1a;Nand
a;Nt=
0;afor t T
1a;N. De…ne
N
ta;N= N
0;aa;N t:
The stopping time T
1a;Nis the …rst (and unique) jump time of N
ta;N. More precisely, we have
N
ta;N= 1
[Ta;N1 ;1)
(t) : Indeed, if T
1a;N= +1, where 1
[Ta;N1 ;1)
(t) is the function identically equal to zero, we have
a;N
t
<
0;afor all t, hence N
0;aa;N tis also identically equal to zero; if T
1a;N< 1 and t < T
1a;N, where 1
[Ta;N1 ;1)
(t) = 0, we have
a;Nt<
0;aand thus N
0;aa;N t= 0; …nally, if T
1a;N< 1 and t T
1a;N, where 1
[Ta;N1 ;1)
(t) = 1, we have
a;Nt=
0;aand thus N
0;aa;N t= N
0;a0;a= 1.
3 Preliminaries on Itô formula for processes de…ned on ran- dom time intervals
3.1 SDEs on random time intervals In the sequel, let us always assume T
1> T
0.
When we write
dX
t= b
tdt +
tdB
ton t 2 [T
0; T
1] where T
0; T
1are stopping times, we mean that
X
t= X
T0+ Z
tT0
b
sds + Z
tT0
s
dB
sfor t 2 [T
0; T
1] : When we write
dX
t= b
tdt +
tdB
ton t 2 (T
0; T
1) where T
0; T
1are stopping times, we mean that
dX
t= b
tdt +
tdB
ton t 2 T
00; T
10for every pair of stopping times T
00T
102 (T
0; T
1). Similar de…nitions apply to other cases,
like [T
0; T
1).
3.2 Itô formula on random time intervals If 2 C
1;2and
X
t= X
T0+ Z
tT0
b
sds + Z
tT0
s
dB
sfor t 2 [T
0; T
1) then
(t; X
t) = (T
0; X
T0) + Z
tT0
(@
s(s; X
s) ds + r (s; X
s) b
s) ds + Z
tT0
r (s; X
s)
sdB
s+ Z
tT0
1
2 T r
s TsD
2(s; X
s) ds for t 2 [T
0; T
1):
3.3 SDEs for processes with jumps Assume, for the process Y
t2 R, we have
Y
t= Y
T0+ Z
tT0
b
sds + Z
tT0
s
dB
sfor t 2 [T
0; T
1) Y
t= 0 otherwise.
Then Y
tsatis…es, for all t 0, the following identity:
Y
t= Z
t0
1
s2[T0;T1)b
sds + Z
t0
1
s2[T0;T1) sdB
s+ Y
T01
t T0Y
T11
t T1for t 0:
The proof of this identity is simply made by checking the validity for t < T
0, then for t 2 [T
0; T
1) and …nally for t T
1.
If needed, we may formally write
d (Y
t) = (
t=T01
T0>0 t=T1) Y
t+ 1
t2[T0;T1)b
tdt + 1
t2[T0;T1) tdB
t:
4 Back to our case
Proposition 2 Let be a function from R
df g ! R, of class C
2on R
d, equal to zero at . Then
X
ta;N= X
a;NT0a;N
1
t Ta;N0
X
a;NT1a;N
1
t Ta;N1
+
Z
t 01
s2Ia;Nr X
sa;NdB
sa+ 1 2
Z
t 01
s2Ia;NX
sa;Nds:
Proof. We have
X
ta;N= 0 if t = 2 I
a;NX
ta;N= X
a;NT0a;N
+ Z
tT0a;N
r X
sa;NdB
sa+ 1 2
Z
t T0a;NX
sa;Nds if t 2 I
a;N:
Hence it is su¢ cient to apply the result of Subsection 3.3.
Set
S
tN:= 1 N
X
a2 t
Xta;N
= 1 N
X
a2
1
t2Ia;NXta;N
so that, for as above (namely ( ) = 0) S
tN; = 1
N X
a2
1
t2Ia;NX
ta;N= 1 N
X
a2
X
ta;N:
Proposition 3 Let be a function from R
df g ! R, of class C
2on R
d, equal to zero at . Then
S
tN; = S
0N; + 1 N
X
a2
X
a;NT1a;N
1
t Ta;N 1+ M
t1; ;N+ 1 2
Z
t 0S
sN; ds where
M
t1; ;N:= 1 N
X
a2
Z
t 01
s2Ia;Nr X
sa;NdB
sa:
Proof. We have S
tN; = 1
N X
a2
X
a;NT0a;N
1
t Ta;N 01 N
X
a2
X
a;NT1a;N
1
t Ta;N 1+ 1 N
X
a2
Z
t 01
s2Ia;Nr X
sa;NdB
sa+ 1 N
X
a2
1 2
Z
t 01
s2Ia;NX
sa;Nds
= 1 N
X
a2
X
a;NT0a;N
1
t Ta;N0
1 N
X
a2
X
a;NT1a;N
1
t Ta;N1
+ M
t1; ;N+ 1 2
Z
t 0S
Ns; ds:
Now,
1 N
X
a2
X
a;NT0a;N
1
t Ta;N0
1 N
X
a2
X
a;NT1a;N
1
t Ta;N1
= S
0N; + 1 N
X
a2
X
a;NT1a;N
1
t Ta;N1
: Indeed, in the …rst sum,
N1P
a2
X
a;NT0a;N
1
t Ta;N0
, there are terms with T
0a= 0, which sum-up to S
0N; , and terms with T
0a> 0, which mean that two new particles arise at time T
0aand one particle dies; the particle which dies is included in the sum
N1P
a2
X
a;NT1a;N
1
t Ta;N 1. The sum of two new particles minus the old particle is equal to one new particle, and a way
to describe the sum of only one new particle is
N1P
a2
X
a;NT1a;N
1
t Ta;N1
. Recall that, in our scheme, we always have T
1a;N> 0 (since T
1a;N> T
0a;N), hence there is no danger to add also particles already existing at time zero, in the sum
N1P
a2
X
a;NT1a;N
1
t Ta;N1
. Proposition 4 Let be bounded (in addition to the previous properties). Then the process
M
t2; ;N:= 1 N
X
a2
X
a;NT1a;N
1
t Ta;N 11 N
X
a2
Z
t 0X
sa;Nd
a;Nsis a martingale, with respect to the …ltration G
t:= F
t.
Proof. We have
X
a;NT1a;N
1
t Ta;N 1= Z
t0
X
sa;NdN
sa;N: It is known that, since N
ta;N= N
0;aa;Nt
, the process N
ta;N a;Ntis a G
t-martingale; namely,
a;N
t
is the compensator of N
ta;N(notice: these facts have an elementary proof in the case when
a;Ntis deterministic; but the result is true also in the random case; a reference available on line is [1], lemma 3.2.
By the theory of stochastic integration with respect to point processes, the compensator of R
t0
X
sa;NdN
sa;Nis the Lebesgue-Stilties integral Z
t0
X
sa;Nd
a;Ns:
The di¤erence of the two processes, M
t2; ;N, is a martingale (not only a local martingale,
due to the boundedness of ). The result remains true under summation (this require some
technical work, omitted).
Corollary 5 If
a;Nthas the form
a;N
t
=
Z
t 01
s2Ia;NX
sa;N; S
sNds then
S
tN; = S
0N; + 1 2
Z
t 0S
sN; ds + Z
t0
; S
sNS
sN; ds + M
t1; ;N+ M
t2; ;N:
Proof. We observe that 1
N X
a2
Z
t 0X
sa;Nd
a;Ns= 1 N
X
a2
Z
t 0X
sa;N1
s2Ia;NX
sa;N; S
sNds = Z
t0
S
sN; ; S
sNds:
Let us now take, as a particular case of (x), the function
N(x
0x). We write h
Nt:=
NS
tNso, for instance, S
tN;
N(x
0) = h
Nt(x
0). Renaming, at the end of the computation, x
0by x, we get:
Corollary 6
h
Nt(x) = h
N0(x) + 1 2
Z
t 0h
Ns(x) ds + Z
t0
N
; S
sNS
sN(x) ds + M
t1;N(x) + M
t2;N(x)
where
N; S
NsS
sN(x) denotes R
N
(x y) y; S
sNS
sN(dy) and M
t1;N(x) := 1
N X
a2
Z
t 01
s2Ia;Nr
Nx X
sa;NdB
sa:
M
t2;N(x) := 1 N
X
a2
N
x X
a;NT1a;N
1
t Ta;N 1Z
t 0N
; S
sNS
sN(x) ds:
5 Estimates on the martingale terms
Let us recall a few general facts. Given a semi-martingale M
t(di che regolarità? ) we associate to it the quadratic variation [M; M ]
tand the predictable quadratic variation hM; Mi
t, which is the predictable compensator of [M; M ]
t. If M is a (local?) martingale, it is also the unique predictable process such that M
t2hM; Mi
tis a martingale (local?).
It follows
E M
t2= E [hM; Mi
t] = E [[M; M ]
t] :
Example 7 If M
t= B
t, then [M; M ]
t= t, hM; Mi
t= t, E M
t2= t.
Example 8 If N
tis an increasing jump process with discrete jumps, then [N; N ]
t= P
s t
(J N
s)
2.
Example 9 If the jumps have size 1, then [N; N ]
t= N
tand hN; Ni
tis thus the compen- sator A
tof N
titself. Set M
t= N
tA
tand assume A
tcontinuous (by a theorem, it is so if the jumps are not predicatble). Then [M; M ]
t= [N; N ]
t= N
t, hM; Mi
t= A
tand
E M
t2= E h
(N
tA
t)
2i
= E [M
t] = E [A
t] :
Example 10 In particular, if N
t= N
0t, N
t0standard Poisson, the previous facts hold with A
t=
t.
Example 11 If d
s=
sds, H is predictable and bounded, and M
t:=
Z
t 0H
sd N
0s s= Z
t0
H
sdN
0sZ
t0
H
s sds then [M; M ]
t= R
0
H
sdN
0s t. But, in case N
0t= 1
[T;1)(t), Z
t0
H
sdN
0s= H
T1
[T;1)(t) = H
TN
0tand thus Z
0
H
sdN
0st
= H
TN
0 t= H
T2N
0t= Z
t0
H
s2dN
0swhose compensator (hence the compensator of [M; M ]
t) is
hM; Mi
t= Z
t0
H
s2d
s: Summarizing,
E M
t2= E H
T2N
0t= E Z
t0
H
s2dN
0s= E Z
t0
H
s2d
s:
Given two semi-martingales M
t(i)(di che regolarità? ), i = 1; 2, we associate to them the joint or mutual variation M
(1); M
(2) tand the predictable joint variation M
(1); M
(2) t, which is the predictable compensator of M
(1); M
(2) t. If M
(1); M
(2)are (local) mar- tingales, is it true that M
(1); M
(2) tis also the unique predictable process such that M
t(1)M
t(2)M
(1); M
(2) tis a martingale (local?)? It would follow
E h
M
t(1)M
t(2)i
= E hD
M
(1); M
(2)E
t
i
= E hh
M
(1); M
(2)i
t
i :
Notice that, if M
(1); M
(2)are jump processes that do not jump simultaneously, then M
(1); M
(2) t= 0, hence E h
M
t(1)M
t(2)i
= 0.
Example 12 If a = 1; 2, N
ta= N
0;aat
, N
t0;aindependent standard Poisson, d
as=
asds, H
apredictable and bounded, and
M
t(a):=
Z
t 0H
sad N
0;aa sas
= Z
t0
H
sadN
0;aa sZ
t 0H
sa asds
= H
TaaN
0;aa tZ
t 0H
sa asds
then h
M
(1); M
(2)i
t
= h
H
T11N
0;11; H
T22N
0;22i
t
= 0 since they do not jump simultaneously. Hence E h
M
t(1)M
t(2)i
= 0.
5.1 First martingale term
By taking a …nite sum approximation and then passing to the limit, one can check that E M
t1; ;N 2= 1
N
2X
a2
E
" Z
t 01
s2Ia;Nr X
sa;NdB
sa2
# :
We have used the fact that E
Z
t 01
s2Ia;Nr X
sa;NdB
saZ
t0
1
s2Ia0;Nr X
sa0;NdB
as0= 0
for a 6= a
0, a known property for Itô integrals w.r.t. indepedent Brownian motions, when the integrands are of class M
2(here it is true since r
Nis bounded). Then by Itô isometry,
E M
t1; ;N 2= 1 N
2X
a2
E Z
t0
1
s2Ia;Nr X
sa;N 2ds
= 1 N E
"Z
t 01 N
X
a2
1
s2Ia;Nr X
sa;N 2ds
#
= 1 N E
Z
t 0D
jr j
2; S
sNE ds 1
N kr k
21Z
t0
E 1; S
Nsds:
It follows
N !1
lim E M
t1; ;N 2= 0
by an estimate on E 1; S
sNdiscussed afterwards.
Similarly,
E M
t1;N(x)
2= 1 N
2X
a2
E
" Z
t 01
s2Ia;Nr
Nx X
sa;NdB
sa2
#
= 1 N
2X
a2
E Z
t0
1
s2Ia;Nr
Nx X
sa;N 2ds
= 1 N E
"Z
t 01 N
X
a2
1
s2Ia;Nr
Nx X
sa;N 2ds
#
= 1 N E
Z
t 0D
jr
N(x )j
2; S
sNE ds :
Here we have to use the structure of
N. Precisely, if we assume that
N(x) =
Nd N1x , we have D
jr
N(x )j
2; S
sNE
=
Nd 2Z
d
N
r
N1(x y)
2S
Ns(dy) : (1) To control such term it is necessary to assume
sup
N 2N d 2 N
N < 1:
When the second moments E M
t1; ;N 2and E M
t1;N 2have been estimated, one can prove similar estimates with the supremum in time inside the expectation, using the martingale property.
5.2 Second martingale term
Consider now
M
t2; ;N= 1 N
X
a2
X
a;NT1a;N
1
t Ta;N 11 N
X
a2
Z
t 0X
sa;Nd
a;Ns= 1 N
X
a2
Z
t 0X
sa;Nd N
sa;N a;Nswhich is a martingale with respect to the …ltration G
t:= F
t. Since the jumps of N
sa;Nand N
sa0;N, for a 6= a
0never occur at the same time, we have (see 12) E
Z
t 0X
sa;Nd N
sa;N a;NsZ
t 0X
sa0;Nd N
sa0;N as0;N= 0:
Hence
E M
t2; ;N 2= 1 N
2X
a2
E
" Z
t 0X
sa;Nd N
sa;N a;Ns2
# : It is known, see 11, that
E
" Z
t 0X
sa;Nd N
sa;N a;Ns2
#
= E
"
X
a;NT1a;N 2
1
t Ta;N 1#
or also E
" Z
t 0X
sa;Nd N
sa;N a;Ns2
#
= E Z
t0
X
sa;N 2d
a;Ns: Then, using the …rst expression,
E M
t2; ;N 2= 1 N
2X
a2
E
"
X
a;NT1a;N 2
1
t Ta;N 1#
= k k
211 N
X
a2
E
"
1 N
X
a2
1
t Ta;N 1#
k k
211
N E 1; S
tN:
Alternatively, using the second expression, if
a;Nthas the form
a;N
t
=
Z
t 01
s2Ia;NX
sa;N; S
sNds we get
E M
t2; ;N 2= 1 N
2X
a2
E Z
t0
X
sa;N 2d
a;Ns= 1 N
2X
a2
E Z
t0
X
sa;N 21
s2Ia;NX
sa;N; S
Nsds
= 1 N
2X
a2
E Z
t0
X
sa;N 21
s2Ia;NX
sa;N; S
sNds
= 1 N E
Z
t 0Z
j (x)j
2x; S
sNS
sN(dx) ds
which here gives rise to a similar estimate as above.
Similarly,
E M
t2;N(x)
2= 1 N
2X
a2
E
" Z
t 0N
x X
sa;Nd N
sa;N a;Ns2
#
= 1 N
2X
a2
E Z
t0
N
x X
sa;N 2d
a;Ns= 1 N
2X
a2
E
"
N
x X
a;NT1a;N 2
1
t Ta;N 1# :
To control the last term we need to integrate in dx. The previous term is more suitable:
1 N
2X
a2
E Z
t0
N
x X
sa;N 2d
a;Ns= 1 N
2X
a2
E Z
t0
N
x X
sa;N 21
s2Ia;NX
sa;N; S
sNds
= 1 N E
Z
t 0Z
j
N(x y)j
2y; S
sNS
sN(dy) ds : If we assume that
N(x) =
Nd N1x , we have
Z
j
N(x y)j
2y; S
sNS
sN(dy) =
NdZ
d N
1
N
(x y)
2y; S
sNS
sN(dy) : (2)
6 Estimates on h
NtRecall that
h
Nt(x) = h
N0(x) + 1 2
Z
t 0h
Ns(x) ds + Z
t0
N
; S
sNS
sN(x) ds + M
t1;N(x) + M
t2;N(x) : We want an identity for h
Nt(x)
2. We need Itô formula for processes with jumps. Recall the following result, also known as generalized Itô formula.
Lemma 13 Let X be a one-dimensional semimartingale such that X
t; X
ttake values in an open set U R and f : U ! R twice continuously di¤erentiable. Then f(X) is a semimartingale and
f (X
t) =f (X
0) + Z
t0
f
0(X
s)dX
s+ 1 2
Z
t 0f
00(X
s)d [X]
cs+ X
s t
Jf (X
s) f
0(X
s)JX
sLemma 14
h
Nt(x)
2= h
N0(x)
2+ Z
t0
h
Ns(x) h
Ns(x) ds + Z
t0
2h
Ns(x)
N; S
sNS
sN(x) ds +
Z
t 02h
Ns(x) dM
s1;N(x) + Z
t0
2h
Ns(x) dM
s2;N(x) + 1
N Z
t0
D
jr
N(x )j
2; S
sNE
ds + 1
N
2X
a2 N
2N
(x X
a;NT1a;N
)1
t Ta;N 1:
Notice also that 1 N
2X
a2 N 2
N
(x X
a;NT1a;N
)1
t Ta;N1
= 1
N
2X
a2 N
Z
t 02
N
(x X
sa;N)dN
sa;N:
Proof. We apply generalized Itô formula to the function f (x) = x
2and the process X
t= h
Nt(x) (N and x given). We …rst get
h
Nt(x)
2= h
N0(x)
2+ Z
t0
2h
Ns(x) dh
Ns(x) + [X]
ct+ X
s t
J h
Ns(x)
22h
Ns(x) Jh
Ns(x)
where, after substitution of dh
Ns(x), the term R
t0
2h
Ns(x) dh
Ns(x) give rise to the sum of the terms (we analyze each one of them on a separate line)
Z
t 02h
Ns(x) 1
2 h
Ns(x) ds = Z
t0
h
Ns(x) h
Ns(x) ds
(the number of jumps of h
Ns(x) is …nite on [0; t], hence we may change the value of the Riemann integral at a …nite numebr of points)
Z
t 02h
Ns(x)
N; S
sNS
sN(x) ds = Z
t0
2h
Ns(x)
N; S
sNS
sN(x) ds Z
t0
2h
Ns(x) dM
s1;N(x) = Z
t0
2h
Ns(x) dM
s1;N(x)
(the argument about the …nite number of jumps of h
Ns(x) in [0; t] applies also to Itô integrals w.r.t. Brownian motion)
Z
t 02h
Ns(x) dM
s2;N(x) :
The quadratic variation [X]
tis equal to
[X]
t= M
1;N(x)
t+ M
2;N(x)
t+ 2 M
1;N(x) ; M
2;N(x)
t:
The quadratic variation M
2;N(x)
thas only jumps; the covariation M
1;N(x) ; M
2;N(x)
tis equal to zero since, at the jumps of M
2;N(x), M
1;N(x) is continuous. Hence term [X]
ctis equal to
[X]
ct= M
1;N(x)
t= 1 N
2X
a2
Z
t 01
s2Ia;Nr
Nx X
sa;NdB
sa= 1 N
2X
a2
Z
t 01
s2Ia;Nr
Nx X
sa;N 2ds
= 1 N
Z
t 0D
jr
N(x )j
2; S
sNE
ds:
It remains to understand the term P
s t
J h
Ns(x)
22h
Ns(x) Jh
Ns(x) . At every jump time r, we have
J h
Nr(x)
22h
Nr(x)Jh
Nr(x)
= h
Nr(x)
2h
Nr(x)
22h
Nr(x) h
Nr(x) h
Nr(x)
= h
Nr(x)
2+ h
Nr(x)
22h
Nr(x) h
Nr(x)
= h
Nr(x) h
Nr(x)
2= Jh
Nr(x)
2: Moreover,
Jh
NT1a;N
(x) = 1
N
N(x X
a;NT1a;N
) hence
X
r t
Jh
Nr(x)
2= X
a2 N
Jh
NT1a;N
(x)
21
t Ta;N 1= 1 N
2X
a2 N
2N
(x X
a;NT1a;N
)1
t Ta;N 1:
Lemma 15 One has E
" Z
t 02h
Ns(x) dM
s1;N(x)
2
#
4
Nd 2N E
Z
t 0h
Ns(x)
2Z
d
N
r
N1(x y)
2S
sN(dy) ds :
Proof. Since Z
t0
2h
Ns(x) dM
s1;N(x) = 1 N
X
a2
Z
t 02h
Ns(x) 1
s2Ia;Nr
Nx X
sa;NdB
asWe have, as above,
E
" Z
t 02h
Ns(x) dM
s1;N(x)
2
#
= 1 N
2X
a2
E
" Z
t 02h
Ns(x) 1
s2Ia;Nr
Nx X
sa;NdB
sa2
#
= 1 N
2X
a2
E Z
t0
4 h
Ns(x)
21
s2Ia;Nr
Nx X
sa;N 2ds
= 4 N E
"Z
t 0h
Ns(x)
21 N
X
a2
1
s2Ia;Nr
Nx X
sa;N 2ds
#
= 4 N E
Z
t 0h
Ns(x)
2D
jr
N(x )j
2; S
sNE ds : Using (1), we get
E
" Z
t 02h
Ns(x) dM
s1;N(x)
2
#
4
Nd 2N E
Z
t 0h
Ns(x)
2Z
d
N
r
N1(x y)
2S
sN(dy) ds :
Lemma 16 One has E
" Z
t 02h
Ns(x) dM
s2;N(x)
2
#
4
NdN E
Z
t 0h
Ns(x)
2Z
d N
2 1
N
(x y) y; S
sNS
sN(dy) ds : Proof. Since
Z
t 02h
Ns(x) dM
s2;N(x) = 1 N
X
a2
Z
t 02h
Ns(x)
Nx X
sa;Nd N
sa;N a;Nsas above we have E
" Z
t 02h
Ns(x) dM
s2;N(x)
2
#
= 1 N
2X
a2
E
" Z
t 02h
Ns(x)
Nx X
sa;Nd N
sa;N a;Ns2
#
= 1 N
2X
a2
E Z
t0
4 h
Ns(x)
2 2Nx X
sa;NX
sa;N; S
sNds
= 4 N E
Z
t 0h
Ns(x)
2Z
2N