universitÀ degli studi di bologna … · 2008-03-11 · universitÀ degli studi di bologna ......
TRANSCRIPT
UNIVERSITÀ DEGLI STUDI DI BOLOGNA ——————————————————
DIPARTIMENTO DI INGEGNERIA ELETTRICA Viale Risorgimento n°2 - 40136 BOLOGNA (ITALIA)
ANALYTICAL SOLUTIONS
FOR THE CURRENT DISTRIBUTION
IN A RUTHERFORD CABLE WITH N STRANDS
M. Fabbri, M. Breschi
Abstract −−−− The geometrical properties of the auto/mutual induction coefficient matrix among the strands of a Rutherford cable are investigated. The solution for the general linear case is
given. The comparison with a known two-strand solution is carried out. The convergence
analysis is done and the regime solution is evaluated. For the non-linear case a diagonal sys-
tem is obtained and solved under the assumption of “mean” resistance. Finally, the flux ramp
case and the perfect conductor cases are fully described.
RAPPORTO INTERNO DIEUB – APRILE 2001
DIE/UNIBO – Aprile 2001 1 di 37
Index
1. Introduction .................................................................................................................... pp. 1
2. The Solution for the Linear Case .................................................................................... pp. 5
3. Comparison with the Krempasky-Schmidt solution ....................................................... pp. 9
4. Convergence Analysis .................................................................................................... pp. 13
5. Regime Solution.............................................................................................................. pp. 16
6. Non-linear Case............................................................................................................... pp. 17
7. Flux Ramp....................................................................................................................... pp. 19
8. Perfect Conductor Case (r = 0) ....................................................................................... pp. 27
References ........................................................................................................................... pp. 32
Appendix A – Forced current distribution in a 3-strand cable............................................ pp. 33
Appendix B – 3-strand cable excited as in the Krempasky-Schmidt case.......................... pp. 35
1. Introduction
Consider a square N × N circulant symmetric matrix [A], defined by the following two properties:
1- ai,1 = ai−1,N, ai,j = ai−1,j−1, with i, j = 2, … , N;
2- ai,j = aj,i, with i, j = 1, … , N.
Assume that N is even, in particular N = 2p. In this case p+1 different elements exist in the matrix [A]
(in fact, consider the firs row of [A], for which a1,j = a1,N+2−j with j = 2, … , N. There exist p−1 pairs and
the unpaired elements a11 and a1,p+1 ha to be added). Therefore the matrix [A] can be written as follows:
[ ]
=
−+
−−
−−+
−−+
+−−
−−
+−
+
11121p,1p,11p,1141312
12112p,11p,1p,1151413
1p,12p,1111213p,11p,1p,1
p,11p,11211121p,1p,11p,1
1p,1p,11312112p,11p,1p,1
1415p,11p,12p,1111213
13141p,1p,11p,1121112
1213p,11p,1p,1131211
aaaaaaaa
aaaaaaaa
aaaaaaaa
aaaaaaaa
aaaaaaaa
aaaaaaaa
aaaaaaaa
aaaaaaaa
A
LL
LL
MMOMMMOMMM
LL
LL
LL
MMOMMMOMMM
LL
LL
LL
The eigenvalues of [A] can be obtained from the N-th complex roots of the unity 1:
π==ω k2jN e1 ⇒ p/jkN/k2jk ee ππ ==ω , with k = p, p−1, …, 1, 0, −1, …, −(p−1)
The eigenvalues are given with k = p, p−1, …, 1, 0, −1, …, −(p−1), by:
DIE/UNIBO – Aprile 2001 2 di 37
( )
( ) ( )
( ) ( ) ( )
( ) ( ) ( )
( ) ( )k1p,1
p
2s
s,111
p
2
p/k1j,1
k1p,1
p
2s
p/k1sjs,111
p
2
p/k1p2j2p2,1
k1p,1
p
2s
p/k1sjs,111
p2
2ps
p/k1sjs,1
jk1p,1
p
2s
p/k1sjs,111
N
1s
p/k1sjs,1
N
1s
1sks,1
1NkN1
1sks,1k1211k
1ap
k1scosa2a
ea1aeaa
ea1aeaa
eaeaeaa
eaaaaaa
−+π−+=
=+−++=
=+−++=
=+++=
==ω=ω++ω++ω+=λ
+=
=σ
πσ−σ+
=
π−
=σ
πσ−+σ−++
=
π−
+=
π−π+
=
π−
=
π−
=
−−−
∑
∑∑
∑∑
∑∑
∑∑KK
The equation above shows that λq = λ−q, with q = 1, … , p−1. Thus, p+1 different eigenvalues of [A] ex-
ist: the algebraic multiplicity of λ0 and λp is equal to 1 (and therefore also their geometric multiplicity
and the dimension of the corresponding subspace are equal to 1); the algebraic multiplicity of λq, with q
= 1, … , p−1, is equal to 2. To determine the existence of the associated spectral basis it’s sufficient to
demonstrate that the geometric multiplicity of λq, with q = 1, … , p−1, is equal to 2, determining two dif-
ferent eigenvectors.
It is simple to show that the vector bk (∈ CN), with k = p, p−1, …, 1, 0, defined by:
1Nkk
Tk ,,,1 −ωω= Lb
is an eigenvector of [A] (the index T denotes the transpose operator). In fact, we have:
[ ] [ ][ ]
[ ][ ]
kk
k1N
k
kk
k
112k13
1k12
1Nk
2Nk1N,111
1NkN1k
1NkN,1k1211
112N
k131N
k121N
k
2Nk1N,111
1kN1k
1NkN,1k1211
1Nk11k1312
1Nk1N,1k11N1
1NkN,1k1211
k
aaa
aaa
aaa
aaa
aaa
aaa
aaa
aaa
aaa
A
b
b
λ=
λω
λωλ
=
++ω+ωω
ω+++ωωω++ω+
=
=
++ω+ωω
ω+++ωωω++ω+
=
ω++ω+
ω++ω+ω++ω+
=
−−
−−
−
−
+−+−−
−−
−
−
−
−−
−
M
K
M
K
K
K
M
K
K
K
M
K
K
Note that of all the bk (∈ CN), with k = p, p−1, …, 1, 0, only b0 and bp belong to RN
:
1,1,,1,1 T0 L=b 1,1,,1,1 T
p −−= Lb
Concerning bk, with k = p−1, …, 1, to construct a real spectral basis the following vectors are defined:
( )
π−π=ωωℜ= −
p
q1p2cos,,
p
qcos,1,,,1 1N
qqTq LLb
( )
π−π=ωωℑ= −
−p
q1p2sin,,
p
qsin,0,,,1 1N
qqT
q LLb
DIE/UNIBO – Aprile 2001 3 di 37
with q = 1, … , p−1. The eigenvectors bq and b-q, associated to λq, are orthogonal. In fact, defining
Bq = bq + j b-q, it can be found that:
( )1N2q
2qq
Tq 1 −ω++ω+= LBB ⇒ 1 N2
qqTqq
Tq
2q −ω+=ω BBBB ⇒ ( ) 0 1 q
Tq
2q =−ω BB
Thus, for p ≠ q ≠ 0, we get:
( ) ( )qT
qqTqq
Tqq
Tqq
Tq j0 bbbbbbbb −−−− ++−==BB ⇒
=
=
−
−
0qTq
2
q
2
q
bb
bb
Finally, the real spectral basis bk, with k = p, p−1, …, 1, 0, −1, …, −(p−1) is orthogonal. In fact, recall-
ing that 2:
( )
−=∑−
= y2
1sin
y2
Nsin
y2
1Ncoskycos
1N
0k
( )
−=∑−
= y2
1sin
y2
Nsin
y2
1Nsinkysin
1N
0k
It results that, with s ≠ q (and thus s ± q ≠ ±N):
( ) ( )
( )( ) ( )( )( )
( )( ) ( )( )( ) 0
N
qssin
qssin
N
qs1Ncos
2
1
N
qssin
qssin
N
qs1Ncos
2
1
N
qs2kcos
2
1
N
qs2kcos
2
1
N
q2kcos
N
s2kcos
1N
0k
1N
0k
1N
0k
qTs
=
π−π−
π−−+
π+π+
π+−=
=
π−+
π+=
π
π= ∑∑ ∑−
=
−
=
−
=
bb
( ) ( )
( )( ) ( )( )( )
( )( ) ( )( )( ) 0
N
sqsin
sqsin
N
sq1Nsin
2
1
N
qssin
qssin
N
qs1Nsin
2
1
N
sq2ksin
2
1
N
sq2ksin
2
1
N
q2ksin
N
s2kcos
1N
0k
1N
0k
1N
0k
qTs
=
π−π−
π−−+
π+π+
π+−=
=
π−+
π+=
π
π= ∑∑ ∑−
=
−
=
−
=−bb
( ) ( )
( )( ) ( )( )( )
( )( ) ( )( )( ) 0
N
qssin
qssin
N
qs1Ncos
2
1
N
qssin
qssin
N
qs1Ncos
2
1
N
qs2kcos
2
1
N
qs2kcos
2
1
N
q2ksin
N
s2ksin
1N
0k
1N
0k
1N
0k
qT
s
=
π+π+
π+−+
π−π−
π−−=
=
π+−
π−=
π
π= ∑∑ ∑−
=
−
=
−
=−− bb
To obtain an orthonormal spectral basis, note that: 1) ||b0||2 = ||bp||2 = N; 2) ||bq||2 + ||b−q||2 = N; 3) ||bq||2 =
||b−q||2; and thus ||bq||2 = ||b−q||2 = N/2 = p. therefore, it’s possible to define the orthonormal spectral basis,
as follows:
=
N
1,
N
1,,
N
1,
N
1 T
0 Lb
−−=
N
1,
N
1,,
N
1,
N
1 T
p Lb
DIE/UNIBO – Aprile 2001 4 di 37
( )
π−π=
p
q1p2cos
p
1,,
p
qcos
p
1,
p
1 T
q Lb
( )
π−π=−
p
q1p2sin
p
1,,
p
qsin
p
1,0 T
q Lb
with q = 1, … , p−1. Since any spectral basis is complete, it’s possible to decompose uniquely any given
vector u ∈ RN as follows:
( )( )∑
−−=
=p
1pk
Tkk ubbu
and thus: [ ] ( )( )∑
−−=
λ=p
1pk
Tkkk A ubbu
Finally note that:
1) given a square N × N circulant symmetric positive-definite matrix [M];
2) given a square N × N circulant symmetric [G] such that ∑=
−=N
2k
k111 gg ;
Thus we have that:
1 - the eigenvalues of [M] are positive and given by:
( ) ( )k1p,1
p
2s
s,111k 1mp
k1scosm2m −+π−+=λ +
=∑
with k = p, p−1, …, 1, 0, −1, …, −(p−1). Thus [M] bk = λk bk;
2 - the eigenvalues of [G] are negative (except for one that is null) e given by:
( ) ( )[ ]k1p,1
p
2s
s,1k 11gp
k1scos1g2 −−−
π−−−=γ +=∑
with k = p, p−1, …, 1, 0, −1, …, −(p−1) (therefore, in particular γ0 = 0). Thus [G] bk = γk bk;
3 - the orthonormal spectral basis bk, with k = p, p−1, …, 1, 0, −1, …, −(p−1), it’s the same for [M] and
[G]:
=
N
1,
N
1,,
N
1,
N
1 T
0 Lb
−−=
N
1,
N
1,,
N
1,
N
1 T
p Lb
( )
π−π=
p
q1p2cos
p
1,,
p
qcos
p
1,
p
1 T
q Lb ( )
π−π=−
p
q1p2sin
p
1,,
p
qsin
p
1,0 T
q Lb
The analysis for N odd is really equivalent, but the formulas for the eigenvalues and for the orthonormal
spectral basis are different. Assuming that N = 2p + 1, it can be found that:
1) the eigenvalues of [M] are again positive and given by:
DIE/UNIBO – Aprile 2001 5 di 37
( )∑
+
= +π−+=λ
1p
2s
s,111k1p2
k1s2cosm2m
where the index k spans the set p, p−1, …, 1, 0, −1, …, −p.
2) the eigenvalues of g are again negative and given by:
( )∑
+
=
+−−−=γ
1
2
,112
π12cos12
p
s
skp
ksg
where the index k spans the set p, p−1, …, 1, 0, −1, …, −p.
3) the orthonormal spectral basis b is defined as follows:
=
N
1,
N
1,,
N
1,
N
1 T
0 Lb
+π
++π
++=
1p2
q2p2cos
1p2
2,,
1p2
q2cos
1p2
2,
1p2
2 T
q Lb
+π
++π
+=−
1p2
q2sin
1p2
2,,
1p2
q2sin
1p2
2,0 T
q 2p Lb
with q = 1, … , p.
2. The Solution for the Linear Case
Assume that the currents flowing in the N-strand Rutherford cable are described by the following para-
bolic system:
(P)
[ ][ ] ( ) [ ][ ] ( ) ( ) [ ] ( )
( ) ( ) ( )
( ) ( )( )
==
=α====
=∂∂++
∂∂
αα
x0t,x
N,,1con,N
tIt,Lxit,0xi
t,xGt,xx
t,xRGt,xt
MG
0
2
2
ii
vi
ii
L and i, v ∈ RN
with [R] = r [I], where [I] is the identity matrix, r is constant and [M] and [G] are constant-coefficient
matrixes with the above described properties. The boundary conditions can be synthetically written as (o)
:
(o)
Note that projecting (P) on b0, it results:
(P0)
( ) ( ) ( ) ( )
( ) ( ) ( )
==
====
=∂∂
x0t,x
N
tIt,Lxt,0x
0t,xx
0T0
T0
T0
T0
T02
2
ibib
ibib
ib
⇒ ( ) ( )N
tIt,x
T0 =ib ⇒
( ) ( ) ( )N
0tIx
0T0
==ib (compatibility condition)
DIE/UNIBO – Aprile 2001 6 di 37
( ) ( ) ( )0
N
tIt,Lxt,0x bii ====
Defining the current variations as:
( ) ( ) ( )0
N
tIt,xt,x bii −=δ
the problem can be restated as (note that [G] [M] b0 = [G] λ0 b0 = γ0 λ0 b0 = 0):
(P')
[ ][ ] ( ) [ ] ( ) ( ) [ ] ( )
( ) ( )( ) ( )( ) ( )
=−==δ
==δ==δ
=∂
δ∂+δ+∂δ∂
00
2
2
N
0tIx0t,x
0t,Lxt,0x
t,xGt,xx
t,xrGt,xt
MG
bii
ii
vi
ii
Utilizing the trigonometric basis, orthogonal in [0, L], sin(nπx/L)n, with n∈N, the following series are
defined:
( ) ( )
π=∑∞
= L
xnsintt,x
1n
nFF ⇔ ( ) ( ) ξ
πξξ= ∫ dL
nsint,
L
2t
L
0
n FF .
and then the problem can be restated as (note that the boundary conditions are essential to obtain
( ) n
2L
0
2
2
L
nd
L
nsint,
xL
2i
i δ
π−=ξ
πξξ∂
δ∂∫ ):
(P'n)
[ ][ ] ( ) [ ] ( ) ( ) [ ] ( )
( ) ( ) ( ) ( )[ ]
−−π
=−==δ
=δ
π−δ+δ
n0
0nn
nn
2
nn
11n
2
N
0tI0t
tGtL
ntrGt
dt
dMG
bii
viii
, with n∈N
Utilizing the orthonormal spectral basis bk, with k = p, p−1, …, 1, 0, −1, … −(p−1), defined above, we
get:
( )( )
( )tt k,n
p
1pk
kn η=δ ∑−−=
bi ⇔ ( ) ( )tt nTkk,n ib δ=η ( )
( )( )tt k,n
p
1pk
kn ν= ∑−−=
bv ⇔ ( ) ( )tt nTkk,n vb=ν
and then the problem can be restated as (with n∈N):
(P'n,k)( ) ( ) ( ) ( )
( ) ( )
==η
νγ=η
π−ηγ+η
λγ
0n
Tkk,n
k,nkk,n
2
k,nkk,n
kk
0t
ttL
ntrt
dt
d
ib
, with k = p, p−1, …, 1, −1, … −(p−1)
Note that k ≠ 0, in fact ηn,0 = 0 as a consequence of ( ) 0t,xT0 =δib . The problem (P'n,k) is directly solv-
able:
( )
( ) ( )
==η
ν=η
πγ
−+η
λ
0n
Tkk,n
k,nk,n
2
k
k,nk
0t
tL
n1r
dt
d
ib
⇒
DIE/UNIBO – Aprile 2001 7 di 37
⇒ ( ) ( )( ) ( )( )
ττνλ
+=η
πγ
−λ
τ−−
πγ
−λ−
∫ de1
et
2
kk
2
kk L
n1r
tt
0
k,nk
L
n1r
t
0n
Tkk,n ib
Reconstructing the solution, we get:
( ) ( ) ( )( )∑ ∑
≠−−=
∞
=
πη+=p
0k1pk 1n
k,nk0L
xnsint
N
tIt,x bbi ⇒
( ) ( ) ( )( )( )
( )( )( )
( )∑ ∑∫
∑ ∑
≠−−=
∞
=
πγ
−λ
τ−−
≠−−=
∞
=
πγ
−λ−
πτλ
τ+
+
π+=
p
0k1pk 1n
L
n1r
t
nTk
k
k
t
0
p
0k1pk 1n
L
n1r
t
0n
Tkk0
L
xnsined
L
xnsine
N
tIt,x
2
kk
2
kk
vbb
ibbbi
⇒
( ) ( ) ( )( )( )( )
( )( )( )
( )∑ ∑∫∫
∑ ∑∫
≠−−=
∞
=
πγ
−λ
τ−−
≠−−=
∞
=
πγ
−λ−
πξ
πτξλ
τξ+
+
πξ
πξξ+=
p
0k1pk 1n
L
n1r
t
Tk
k
k
t
0
L
0
p
0k1pk 1n
L
n1r
t
0Tkk
L
0
0
L
nsin
L
xnsine,dd
L
2
L
nsin
L
xnsined
L
2
N
tIt,x
2
kk
2
kk
vbb
ibbbi
⇒
( ) ( ) ( )( )( ) ( )( )
( )( ) ( )( )∑∫∫
∑∫
≠−−=
≠−−=
τ−ξΓτξλ
τξ+
+ξΓξξ+=
p
0k1pk
kTk
k
k
t
0
L
0
p
0k1pk
k0T
kk
L
0
0
t,,x,ddL
2
t,,xdL
2
N
tIt,x
vbb
ibbbi
Recalling that ( ) ( ) ( ) ( )[ ]212121 coscos2
1sinsin ω+ω−ω−ω=ωω and the definition of the elliptic theta
function ϑ3 3(oo)
: ( ) ∑∞
=
+=ϑ1n
n3 nu2cosq21q,u
2
, the Green functions can be written as (ooo)
:
(oo) the fundamental propriety of the function ϑ3 (u, q) is: ( ) ( )q,u
u4
1q,u
2
32
3
∂ϑ∂
−=∂ϑ∂
(ooo) Note that: ( ) ( ) ( ) 0t,,x
xt,,xrt,,x
t 2
k2
kkk
kk =ξ∂
Γ∂+ξΓγ+ξ
∂Γ∂
λγ , with k = p, p−1, …, 1, −1, … −(p−1).
DIE/UNIBO – Aprile 2001 8 di 37
( )
( ) ( )
ξ+πϑ−
ξ−πϑ=
ξ+π−
ξ−π=
=
πξ
π=ξΓ
πγλ
πγλ
λ−
∞
=
πγλ
λ−
∞
=
πγλ
λ−
∞
=
πγ
−λ−
∑∑
∑
2
kk
2
kkk
2
kk
2
k
2
kk
2
k
2
kk
L
t
3L
t
3
rt
1n
L
tnrt
1n
L
tnrt
1n
L
n1r
t
k
e,L2
xe,
L2
x
4
e
L
xncose
2
e
L
xncose
2
e
L
nsin
L
xnsinet;,x
with k = p, p−1, …, 1, −1, … −(p−1). Therefore the Green matrixes are given by:
( )( )[ ] ( )( )∑
≠−−=
ξΓ=ξp
0k1pk
Tkkk
0 t,,xt,,xK bb ( )[ ] ( )( )∑
≠−−= λ
ξΓ=ξp
0k1pk k
Tkk
k t,,xt,,xKbb
And the solution of (P) results to be:
( ) ( ) ( )( )[ ] ( )( ) ( )[ ] ( )τξτ−ξτξ+ξξξ+= ∫∫∫ ,t,,xKddL
2t,,xKd
L
2
N
tIt,x
t
0
L
0
00
L
0
0 vibi
Note that this solution of (P) results to be invariant by addition to the sources i(0)
(x) and v(x, t) of terms
proportional to a b0. In fact:
( )( )[ ] ( )( ) ( ) ( )( )[ ] ( )( ) ( ) ( )( )[ ] ( )( )[ ] ( ) ( )ξξ=ξξ+ξξ=ξ+ξξ 000
0000
00 ,xK,xKv,xKi,xK ibibi
( )[ ] ( ) ( ) ( )[ ] ( ) ( ) ( )[ ] ( )[ ] ( )τξτξ=τξτξ+τξτξ=τξ+τξτξ ,,,xK,,xK,v,,,xK,v,,,xK 00 vbvbv
This means that the solution does not change if in the first integral i(0)
(x) is replaced by δi(0)(x) and if the
reference node for the voltages is changed.
Note that the calculation of the integration kernels [K(0)
] and [K] may cause convergence difficulties,
since the function Γk approaches the Dirac δ distribution, for t going to zero:
( ) ( )ξ−δ=
πξ
π=ξΓ ∑∞
=→x
2
L
L
nsin
L
xnsint;,xlim
1n
k0t
Thus, if i(0)
(x) and v(x, t) are differentiable with respect to x, it’s better to utilize an alternative form of
the solution, obtained through an integration by parts. Defining the new function Γk(oooo)
:
( ) ( ) ∑∫∞
=
πγ
−λ−
ξ
πξ
ππ
=ξ′ξ′Γ=ξΓ1n
2
L
n1r
t
0
kkL2
nsin
L
xnsin
n
eL2dt;,xt;,x
2
kk
(oooo) Note that Γk satisfies the same equations of Γk: ( ) ( ) ( ) 0t,,x
xt,,xrt,,x
t 2
k2
kkk
kk =ξ∂
Γ∂+ξΓγ+ξ
∂Γ∂
λγ
DIE/UNIBO – Aprile 2001 9 di 37
with k = p, p−1, …, 1, −1, … −(p−1), and the new integration kernels [K(0)
] and [K]:
( )( )[ ] ( )( )∑
≠−−=
ξΓ=ξp
0k1pk
Tkkk
0t,,xt,,xK bb ( )[ ] ( )
( )∑
≠−−= λ
ξΓ=ξp
0k1pk k
Tkk
k t,,xt,,xKbb
The solution of (P) can be written as:
( ) ( ) ( )( )[ ] ( )( ) ( )[ ] ( )
( ) ( )( )[ ] ( )( ) ( )( )[ ] ( )( )
( )[ ] ( ) ( )[ ] ( )τξξ∂
∂τ−ξτξ−
τξτ−ξτ+
+ξξ∂
∂ξξ−
ξξ+=
=τξτ−ξξ∂
∂τξ+ξξξ∂
∂ξ+=
∫∫∫
∫
∫∫∫
=ξ
=ξ
=ξ
=ξ
,t,,xKddL
2,t,,xKd
L
2
t,,xKdL
2t,,xK
L
2
N
tI
,t,,xKddL
2t,,xKd
L
2
N
tIt,x
t
0
L
0
L
0
t
0
00
L
0
L
0
000
t
0
L
0
00L
0
0
vv
iib
vibi
and finally:
( ) ( ) ( )( )[ ] ( )( ) ( )( )[ ] ( )( )
( )[ ] ( ) ( )[ ] ( )τξξ∂
∂τ−ξτξ−ττ−τ+
+ξξ∂
∂ξξ−+=
∫∫∫
∫
,t,,xKddL
2,Lt,L,xKd
L
2
t,,xKdL
2Lt,L,xK
L
2
N
tIt,x
t
0
L
0
t
0
00
L
0
000
vv
iibi
Note that the calculation of the integration kernels [K(0)
] and [K] is no more cause of convergence diffi-
culties, since the function Γk approaches the Heaviside step function, for t going to zero:
( ) ( )
ξ>ξ<
=−ξ=
πξ
ππ
=ξΓ ∑∞
=→ x,0
x,1
2
LxU
2
L
L2
nsin
L
xnsin
n
L2t;,xlim
1n
2k
0t
In the particular case in which i(0)
(x) has equal components and v(x, t) is time-independent, the solution
further simplifies, defining:
( ) ( ) ∑∫∞
=
πγ
−λ−
πξ
π
−
πγ
−
λ=ττξΓ=ξΓ
1n
L
n1r
t
2
k
k
t
0
kkL
nsin
L
xnsine1
L
n1r
d;,xt;,x*
2
kk
with k = p, p−1, …, 1, −1, … −(p−1); and the integration kernel [K*]:
( )[ ] ( )( )∑
≠−−= λ
ξΓ=ξp
0k1pk k
Tkk
k t,,xt,,x*K *bb
the solution of (P) can be written as:
DIE/UNIBO – Aprile 2001 10 di 37
( ) ( ) ( )[ ] ( )ξξξ+= ∫ vbi t,,x*KdL
2
N
tIt,x
L
0
0
3. Comparison with the Krempasky-Schmidt solution
The solution found above is now compared with a known one for a two-strand cable 4. With the con-
ditions:
1- I(t) = 0;
2- r = 0,
3- even number of periods,
4- i(0)
(x) = 0,
5- time-independent excitation spatially bounded excitation to an interval in the center of the cable.
Hypotheses 3 and 5 lead to write:
( )
−+
ξ−δ+δ−−ξδ
Φ=ξ1
1)
22
L(U)
22
L(U
2/&v
where U is the Heaviside step function. For two strand, N = 2, p = 1; thus, given
[ ]
=
1112
1211
mm
mmM [ ]
−++−
=1212
1212
gg
ggG
1 - the eigenvalues of [M] are positive and given by:
12110 mm +=λ ; 12111 mm −=λ
2 - the eigenvalues of [G] are negative (except for one that is null) e given by:
00 =γ ; 121 g2−=γ
3 - the orthonormal spectral basis bk, with k = 1, 0, it’s the same for [M] and [G]:
=
2
1,
2
1 T
0b
−=
2
1,
2
1 T
1b
To reproduce the nomenclature of 4 the following variables are defined:
L1 = 2 (m11 − m12), G1 = g12, α = π w/(2w + δ), L = 2w + δ,
2
112 2
w2GL
4
δ+π
=τ
(Note that π / L = α / w, L / w = π / α and (πδ / 2L) + α = π / 2)
Thus we get: ( ) 2)22
L(U)
22
L(U
21bv ξ−δ+δ−−ξ
δΦ=ξ&
. When the simplified solution is utilized, we
have:
DIE/UNIBO – Aprile 2001 11 di 37
( )
∑
∑
∞
=
τ−
∞
=
πγ
−λ−
πξ
π−πλ
=
=
πξ
π
−
πγ
−
λ=ξΓ
1n
/tn
22
2112
1n
L
n1t
2
1
11
L
nsin
L
xnsine1
n
1Lg2
L
nsin
L
xnsine1
L
n1t;,x*
2
2
11
(Note that
( ) τ=
δ+π=
δ+π=
π−
=
πγ
−λ
1
w2
2
GL
1
4w2GL
1
Lg2
1
mm
1
L
112
11
2
2
2
11
2
121211
2
11
)
Therefore the kernel becomes:
( )[ ] ( ) ( )
( ) ( )
+−−+
πξ
π−π
=
=
+−−+
λξΓ=
+−−+
λξΓ=
=
−
−λξΓ
=λ
ξΓ=ξ
∑∞
=
τ−
11
11
L
nsin
L
xnsine1
n
1Lg
11
11
2
t,,x*
2/12/1
2/12/1t,,x*
2
1,
2
1
2
1
2
1
t,,x*t,,x*t,,x*K
1n
/tn
22
212
1
1
1
1
1
1
1
T11
1
2
bb
and the solution can be written as:
( ) ( )[ ] ( )
( )
( )
( )
( )
( )
−+
πξξ
π−δπ
Φ
=
−+
+−−+
πξξ
π−δπ
Φ=
=ξ
+−−+
πξξ
π−π
=
=ξξξ=
∑ ∫
∑ ∫
∑ ∫
∫
∞
=
δ+
δ−
τ−
∞
=
δ+
δ−
τ−
∞
=
τ−
2
2
L
nsind
L
xnsine1
n
1Lg
1
1
11
11
L
nsind
L
xnsine1
n
1Lg
11
11
L
nsind
L
xnsine1
n
1Lg2
t,,x*KdL
2t,x
1n
2/L
2/L
/tn
22
12
1n
2/L
2/L
/tn
22
12
1n
L
0
/tn
22
12
L
0
2
2
2
&
&
v
vi
Since i2 = − i1, only i1(x, t) is considered:
( ) ( )
( )
( )
( )
( )
( )
∑ ∫
∑ ∫
∑ ∫
∞
=
δ+
δ−
τ−
∞
=
δ+
δ−
τ−
∞
=
δ+
δ−
τ−
πξξδ
α−πα
=
=
πξξ
α−δπ
=
=
πξξ
π−δπ
Φ=
1n
2/L
2/L
/tn
2m
1n
2/L
2/L
/tn
22
m
1n
2/L
2/L
/tn
22
11
L
nsind
1
w
xnsine1
n
1I
4
L
nsind
w
xnsine1
n
1
w
LI4
L
nsind
L
xnsine1
n
1LG2t,xi
2
2
2&
DIE/UNIBO – Aprile 2001 12 di 37
where 2/wGI 1m Φ= & . The last integral can be made explicit, as follows:
( )
( )
πδ
ππδ
=
=
δ+π−
δ−ππδ
=
πξξδ ∫
δ+
δ−
L2
nsin
2
nsin
n
L2
L2
Lncos
L2
Lncos
n
L
L
nsind
12/L
2/L
Note that, since sin (nπ/2) = 0 for even n, it’s possible to restrict la summation to the odd n only. More-
over, note that
( ) ( )α
πδ+α
πδ=
α+πδ=
πnsin
L2
ncosncos
L2
nsinn
L2
nsin
2
nsin
Thus, assuming δ << L :
( )
( )( ) ( ) ( )α≅
α
πδ+α
πδ
πδ
πδ
=
πξξδ ∫
δ+
δ−
nsinnsinL2
ncosncos
L2
nsin
L2
n
L2
nsin
L
nsind
12/L
2/L
Note that this corresponds to a first order approximation of the original integral:
( )
( )( )α=δ
δ+π
δ≅
δ+πξξ
δ=
πξξδ ∫∫
δ+δ+
δ−
nsinw2
wnsin
1
w2
nsind
1
L
nsind
1w
w
2/L
2/L
and finally:
( ) ( )∑∞
=
τ− α
α−πα
=
nodd1n
/tn
2m1 nsinw
xnsine1
n
1I
4t,xi
2
If the external field ramp is stopped at time t1 and the field is kept constant, the external voltage drops to
zero. In this case the induced currents start decaying. In order to prove this, we start noting that, since
the external voltage drop to zero for t > t1, the external voltage is time dependent and it can found multi-
plying v(ξ) by U( t1 − t). Thus, the solution can be written as follows:
( ) ( ) ( ) ( )τ−ξτ−ξτξ= ∫∫ 1
t
0
L
0
tUt,,xddL
2t,x vKi
Therefore for t > t1 this equation gives
( ) ( ) ( )
( ) ( )
( ) ( ) ( ) ( )ξ−ξξ−ξξξ=
=ξ
′ξ′ξ=
=ξτ−ξτξ=
∫∫
∫∫
∫∫
−
vKvK
vK
vKi
1*
L
0
*
L
0
t
tt
L
0
t
0
L
0
tt,,xdL
2t,,xd
L
2
t,,xtddL
2
t,,xddL
2t,x
1
1
DIE/UNIBO – Aprile 2001 13 di 37
These last two integrals are the same previously solved. Thus, we can evaluate directly them as follows:
( ) ( )
( ) ( )
( ) ( ) 1
nodd1n
/t/tt
2m
1
nodd1n
/tt
2m
1
nodd1n
/t
2m
nsinw
xnsinee
n
1I
4
nsinw
xnsine1
n
1I
4
nsinw
xnsine1
n
1I
4t,x
nn1
n1
n
b
b
bi
∑
∑
∑
∞
=
τ−τ−−
∞
=
τ−−
∞
=
τ−
α
α−πα
=
=α
α−πα
−
+α
α−πα
=
Therefore, factorizing out the exponential, it can be seen that each component under the sum decays with
its corresponding time constant τn. Thus, the strand currents are given by:
( ) ( ) ( ) 1/)tt(
n1n
/t
2mn1n1 ensin
w
xnsine1
n
1I
4t,x bi
α
α−πα
= τ−−∞
=
τ−∑odd
4. Convergence Analysis
The calculation of the function Γk may cause convergence difficulties, since the summation terms are
oscillating.
( )
( ) ( )
ξ+πϑ−
ξ−πϑ=
ξ+π−
ξ−π=
=
πξ
π=ξΓ
πγλ
πγλ
λ−
∞
=
πγλ
λ−
∞
=
πγλ
λ−
∞
=
πγ
−λ−
∑∑
∑
2
kk
2
kkk
2
kk
2
k
2
kk
2
k
2
kk
L
t
3L
t
3
rt
1n
L
tnrt
1n
L
tnrt
1n
L
n1r
t
k
e,L2
xe,
L2
x
4
e
L
xncose
2
e
L
xncose
2
e
L
nsin
L
xnsinet;,x
It can be demonstrated 5 that the elliptic theta function ϑ3 admit the following representation (where
the summation terms are non-oscillating):
( )( )
∑∑∞
−∞=
π−−∞
−∞=
−− π==ϑn
s
nu
n
sns3
2
2
es
nu2cosee,u
Thus the function Γk admits the following alternative representation:
DIE/UNIBO – Aprile 2001 14 di 37
( )
∑
∑
∑
∞
−∞=
−ξ+γλ+λ−
−ξ−γλ+λ−
∞
−∞=
−ξ+γλ
−ξ−γλλ−
∞
−∞=
π−ξ+ππ
γλ
π−ξ−ππ
γλλ−
πγλ
πγλ
λ−
−πγλ
−=
=
−πγλ
−=
=
−πγλ−=
=
ξ+πϑ−
ξ−πϑ=ξΓ
n
nL2
x
t
rtnL
2
x
t
rt
kk
n
nL2
x
tnL
2
x
tkk
rt
n
nL2
x
tLn
L2
x
tL
kk
rt
L
t
3L
t
3
rt
k
2kk
k
2kk
k
2kk
2kk
k
2
2kk2
2
2kk2
k
2
kk
2
kkk
eet4
L
eet4
eL
eet4
eL
e,L2
xe,
L2
x
4
et;,x
Moreover, for what concerns Γk*, we get, after defining ττ=θ⇒τ=θ d
d2 :
( ) ( )
∑ ∫∫
∑∫
∑∫∫
∞
−∞=
−ξ+θ
γλ+λ
θ−
−ξ−θ
γλ+λ
θ−
∞
−∞=
−ξ+θ
γλ+λ
θ−
−ξ−θ
γλ+λ
θ−
∞
−∞=
−ξ+τγλ+
λτ−
−ξ−τγλ+
λτ−
θπγλ
−−θπγλ
−=
=
−θπγλ
−=
=τ
−πτ
γλ−=ττξΓ=ξΓ
n
t
0
nL2
xr
kk
t
0
nL2
xr
kk
n
nL2
xrnL
2
xrt
0
kk
n
nL2
xrnL
2
xr
kk
t
0
t
0
kk
de2
Lde
2
L
eed2
L
dee4
Ld;,xt;,x*
2
2kk
k
22
2kk
k
2
2
2kk
k
22
2kk
k
2
2kk
k
2kk
k
Consider now the summation term:
θπγλ−=θ
πγλ− ∫∫
θγλ−
−λ
θ−γ−−θ
γλ+λ
θ−
dee2
Lde
2
Lt
0
Qr
rQ2kk
t
0
Qr
kk
2
kk
kk
2
2kk
k
2
Defining r
Qr4
y
r2
yQry k
kk
2k
kkk
γ−λ+
λ+
λ=θ⇒
θγλ−−
λθ= we have that:
dyrQ4y
y1
r2
1d
k2
k
γ−++
λ=θ
Therefore:
dyrQ4y
y1ee
r4
Lde
2
L
k2
Qt
rt
yrQ2kk
t
0
Qr
kk
kk
k2
k
2
2kk
k
2
γ−++
πγ
−λ
=θπγλ
− ∫∫
γλ−−λ
∞−
−γ−−θγλ+
λθ−
DIE/UNIBO – Aprile 2001 15 di 37
Defining the functions ( ) dye2
xerf
x
0
y2
∫−
π= and erfc (x) = 1 − erf (x), we have:
( )
γ−++
γ−+−
+
γλ−−λ
π+∞−π−πγ−λ=θ
πγλ−
∫∫
∫γλ−−
λ −∞− −
γ−−θγλ+
λθ−
Qt
rt
0 k2
y
0 k2
y
kk
k
rQ2kk
t
0
Qr
kk
kk
k 22
k
2
2kk
k
2
dyrQ4y
yedy
rQ4y
ye
Qt
rterf
2erf
2e
r4
Lde
2
L
Defining ydywdwrQ4yw k22 =⇒γ−+= we have:
γ−−
γ−+π=
==γ−+
∫∫γ−+
γ−
−γ−−
kk2
rQ4X
rQ4
wrQ4X
0 k2
y
rQ4erfrQ4Xerf2
dweedyrQ4y
yek
2
k
2k
2
Thus:
γλ−+
λ−
γλ−−
λ−
γ−
λ=
=
γλ−+
λ+−
+
γλ−−
λ+
γ−
λ=θ
πγλ
−
γ−γ−−γ−−
γ−γ−
γ−−θγλ+
λθ−
∫
Qt
rterfceQ
t
rterfcee2
r8
L
Qt
rterfee
Qt
rterf1e
r8
Lde
2
L
kk
k
rQ2kk
k
rQ2rQ2kk
kk
k
rQ4rQ4
kk
k
rQ2kk
t
0
Qr
kk
kkk
kk
k
2
2kk
k
2
and finally:
( )
−ξ+γλ−
+λ
+
−ξ+γλ−
−λ
+
+
−ξ−γλ−
+λ
−
−ξ−γλ−
−λ
−
+−γ−λ=ξΓ
γ−−ξ+γ−−ξ+−
γ−−ξ−γ−−ξ−−
γ−−ξ+−∞
−∞=
γ−−ξ−−∑
nL2
x
t
rterfcenL
2
x
t
rterfce
nL2
x
t
rterfcenL
2
x
t
rterfce
e2e2r8
Lt;,x*
kk
k
rnL2xkk
k
rnL2x
kk
k
rnL2xkk
k
rnL2x
rnL2x
n
rnL2xkkk
kk
kk
kk
Note that the first two terms are time-independent and, therefore, they represent the regime solution.
Thus:
DIE/UNIBO – Aprile 2001 16 di 37
( ) ( )
−ξ+γλ−
+λ
+
−ξ+γλ−
−λ
+
+
−ξ−γλ−
+λ
−
−ξ−γλ−
−λ
−
γ−
λ+∞ξΓ=ξΓ
γ−−ξ+γ−−ξ+−
γ−−ξ−γ−−ξ−−
∞
−∞=∑
nL2
x
t
rterfcenL
2
x
t
rterfce
nL2
x
t
rterfcenL
2
x
t
rterfce
r8
L;,x*t;,x*
kk
k
rnL2xkk
k
rnL2x
kk
k
rnL2xkk
k
rnL2x
n
kkkk
kk
kk
Note that the function Γk* has two (unessential) singularity points for t → 0 and for t → +∞; in fact
erfc(−∞) → 2 and erfc(+∞) → 0.
5. Regime Solution
Note that since ( ) 0t;,x*lim k0t
=ξΓ→
it’s possible to evaluate easily the regime value of Γk*. In fact:
( )
∑
∑ ∫∫
∑∫
∞
−∞=
−ξ+γλ
λ−=
−ξ−γλ
λ−=
∞
−∞=
∞
−ξ+τγλ+
λτ−
∞
−ξ−τγλ+
λτ−
∞
−∞=
−ξ+τγλ+
λτ−
−ξ−τγλ+
λτ−∞
−
πγλ
−=
=
ττ
−ττπ
γλ−=
=τ
−πτ
γλ−=∞ξΓ
n
nL2
x
t
rs
nL2
x
t
rs
kk
n 0
nL2
xr
0
nL2
xr
kk
n
nL2
xrnL
2
xr
kk
0
k
t
e
t
e
4
L
de
de
4
L
dee4
L;,x*
2kk
k
2kk
k
2kk
k
2kk
k
2kk
k
2kk
k
LL
Recalling that ast4/a
s est
e −− π=
L 6, it follows:
( )
∑
∑
∞
−∞=
−ξ+γ−−−ξ−γ−−
∞
−∞=
−ξ+γλ
λ−=
−ξ−γλ
λ−=
−
γ−
λ=
=
−
πγλ
−=∞ξΓ
n
nL2
xr2nL
2
xr2
kk
n
nL2
x
t
rs
nL2
x
t
rs
kkk
kk
2kk
k
2kk
k
eer4
L
t
e
t
e
4
L;,x* LL
Since 0 < x < L and 0 < ξ < L, it’s simple to show that, separating the summation on the positive and
negative indexes, we have
DIE/UNIBO – Aprile 2001 17 di 37
( )
=
−+−+
−
γ−
λ=∞ξΓ
∑∑
∑∑
∞
=
−ξ+γ−∞
=
−ξ−γ−
ξ+γ−−ξ−γ−−
−
−∞=
−ξ+γ−−−
−∞=
−ξ−γ−−
1n
nL2
xr2
1n
nL2
xr2
2
xr2
2
xr2
1
n
nL2
xr21
n
nL2
xr2
kkk
kkkk
kk
eeee
eer4
L;,x*
( ) ( )
( ) ( ) ( )
( ) ( ) ( )
−+
−
γ−ξ+−γ−ξ−γ−
λ=
=
−
+−+
−
γ
−λ
=
γ−ξ+−γ−ξ−−
γ−
∞
=
γ−−γ−ξ+−∞
=
γ−−γ−ξ−−γ−ξ+−γ−ξ−−
∞
=
γ−−γ−ξ+−∞
=
γ−−γ−ξ−−
∑∑
∑∑
2
ee
1e
rxcoshrxcosh
r2
L
eeeeee
eeeer4
L
kk
k
kkkkkk
kkkk
rxrx
rL2
kkkk
1n
nrL2rx
1n
nrL2rxrxrx
1n
nrL2rx
1n
nrL2rxkk
where it has been used the fact that
1q
1
1q
1n
n
−=∑
∞
=
. Finally:
( )( ) ( ) ( )
−
γ−ξγ−−−γ
−λ=∞ξΓγ−
γ−ξ+−γ−ξ−−
1e
rsinhrxsinh
4
ee
rL;,x*
k
kk
rL2
kkrxrx
kkk
6. Non-linear Case
If the resistance of the strand is a known function of the current, i.e.
( )( )t,xirr jj = , with j = 1, …, N
the problem (P) can still reduced to a diagonal form. Assigning
[ ]( )∑
−−=
λ=p
1pk
TkkkM bb [ ]
( )∑
−−=
γ=p
1pk
TkkkG bb [ ] ∑
=
=N
1j
TjjjrR ee
where ej denote the unit vectors of the Cartesian basis, we have (with k ≠ 0):
[ ][ ] ( ) [ ][ ] ( ) ( ) [ ] ( )
( )[ ] [ ][ ]( )
( )[ ] ( )[ ] ( )[ ]
γ−∂∂+γ+
∂∂λγ=
=−∂∂++
∂∂=
∑ ∑−−= =
t,xt,xx
t,xrt,xt
t,xGt,xx
t,xRGt,xt
MG0
Tkk
Tk2
2Th
p
1ph
N
1j
hTjj
Tkjk
Tkkkk
2
2
vbibibbeebibb
vi
ii
and the following system of N−1 equations (k = −(p−1), …,−1, 1, …, p) is obtained:
DIE/UNIBO – Aprile 2001 18 di 37
(Pk)
( )[ ] ( )[ ] ( )[ ] [ ][ ]( )
( )[ ]( ) ( )( ) ( )( )
==
====
−=∂∂
γ+
∂∂λ ∑ ∑
−−= =
x0t,x
0t,Lxt,0x
t,xrt,xt,xx
1t,x
t
0Tk
Tk
Tk
Tk
Th
p
1ph
N
1j
hTjj
Tkj
Tk
Tk2
2
k
Tkk
ibib
ibib
ibbeebvbibib
Nevertheless, the equations are still coupled since ( )( ) [ ] ( )[ ]( )
== ∑
−−=
p
1ph
Thh
Tjjj t,xrt,xirr ibbe , with j = 1,
…, N. Anyway, note that for k = 0, we get:
( )[ ]( ) ( ) ( )
( ) ( )( ) ( )
====
====
=∂∂
N
0tIx0t,x
N
tIt,Lxt,0x
0t,xx
0T0
T0
T0
T0
T02
2
ibib
ibib
ib
with the usual solution: ( ) ( )N
tIt,xT
0 =ib . Thus, the solution of problem (P) results to be:
( ) ( ) ( )( )( )∑
≠−−=
+=p
0k1pk
Tkk0 t,x
N
tIt,x ibbbi
Finally, in the particular case for which
( )( )tIrrj = , with j = 1, …, N
it’s still possible to solve exactly the problem (P). In fact, with the same symbols of §2:
( )( ) ( )
( ) ( )
==η
ν=η
πγ
−+η
λ
0n
Tkk,n
k,nk,n
2
k
k,nk
0t
tL
n1tIr
dt
d
ib
⇒
⇒ ( ) ( )( ) ( )( )( )
( )( ) ( )
τ∫
τνλ
+∫
=η
πγλτ−+′′
λ−
πγλ
+′′λ−
τ∫ de1
et
2
kk
t
k
2
kk
t
0k L
nttdtIr
1t
0
k,nk
L
nttdtIr
1
0n
Tkk,n ib
Thus the solution, formally, is unchanged, provided that the function Γk is defined as:
( )( )( )
( )( )
ξ+πϑ−
ξ−πϑ
∫
=
=
πξ
π∫=τξΓ
πγλ
πγλ
′′λ−
∞
=
πγλ
+′′λ−
+τ
τ
+τ
τ∑
2
kk
2
kk
t
k
2
kk
t
k
L
t
3L
t
3
tdtIr1
1n
L
nttdtIr
1
k
e,L2
xe,
L2
x
4
e
L
nsin
L
xnsine,t;,x
DIE/UNIBO – Aprile 2001 19 di 37
Therefore the Green matrixes are expresses by:
( )( )[ ] ( )( )∑
≠−−=
ξΓ=ξp
0k1pk
Tkkk
0 0,t,,x0,t,,xK bb ( )[ ] ( )( )∑
≠−−= λ
τξΓ=τξp
0k1pk k
Tkk
k ,t,,x,t,,xKbb
and the solution of (P) can be written as:
( ) ( ) ( )( )[ ] ( )( ) ( )[ ] ( )τξττ−ξτξ+ξξξ+= ∫∫∫ ,,t,,xKddL
20,t,,xKd
L
2
N
tIt,x
t
0
L
0
00
L
0
0 vibi
7. Flux Ramp
In the particular case in which i(0)
(x) has equal components and v(x, t) is given by:
( ) ( )
><
=0
0
tt if ,0
tt if ,x*t,x
vv
the solution of (P) can be written as:
( ) ( ) ( )[ ] ( )τξτ−ξτξ+= ∫∫ ,t,,xKddL
2
N
tIt,x
t
0
L
0
0 vbi
that can be specialized in two cases:
I) t < t0
( )[ ] ( ) ( )[ ] ( ) ( )[ ] ( )ξξξ=ξ
τ−ξτξ=τξτ−ξτξ ∫∫∫∫∫ *t,,x*Kd*t,,xKdd,t,,xKdd
L
0
t
0
L
0
t
0
L
0
vvv
II) t > t0
( )[ ] ( ) ( )[ ] ( ) ( )[ ] ( )
( )[ ] ( ) ( )[ ] ( )ξ−ξξ−ξξξ=
=ξ
′ξ′ξ=τξτ−ξτξ=τξτ−ξτξ
∫∫
∫∫∫∫∫∫−
*tt,,x*Kd*t,,x*Kd
*t,,xKtdd,t,,xKdd,t,,xKdd
0
L
0
L
0
t
tt
L
0
t
0
L
0
t
0
L
0 0
0
vv
vvv
thus, after same manipulation, we obtain:
I) t < t0
( ) ( ) ( )[ ] ( )ξξξ+= ∫ *t,,x*KdL
2
N
tIt,x
L
0
0 vbi
II) t > t0
( ) ( ) ( )[ ] ( ) ( )[ ] ( )ξ−ξξ−ξξξ+= ∫∫ *tt,,x*KdL
2*t,,x*Kd
L
2
N
tIt,x 0
L
0
L
0
0 vvbi
In particular, the last integral can be written as:
DIE/UNIBO – Aprile 2001 20 di 37
( )[ ] ( ) ( ) ( )000
0
L
0
tt,xN
ttI*tt,,x*Kd
L
2 −+−−=ξ−ξξ∫ ibv
and therefore:
( ) ( ) ( ) ( )[ ] ( ) ( )0
L
0
00 tt,x*t,,x*Kd
L
2
N
ttItIt,x −−ξξξ+−+= ∫ ivbi
N.B. In the particular case in which v* is a piecewise constant function, we get:
( ) ( )∑=
+=M
1m
1mm*m x,x;xUx* vv
(with x1 = 0 and xM+1 = L) and where:
( )
><<
<=
bx fi,0
bxa fi,1
ax fi,0
b,a;xU
We have:
I) t < t0
( ) ( ) ( )[ ]∑=
++=M
1m
*m1mm0 x,x;t,x**K2
N
tIt,x vbi
II) t > t0
( ) ( ) ( )[ ] ( )[ ]∑∑=
+=
+ −−+=M
1m
*m1mm0
M
1m
*m1mm0 x,x;tt,x**K2x,x;t,x**K2
N
tIt,x vvbi
where (for 0 ≤ a < b ≤ L):
( )[ ] ( )[ ] ( )( )
( )( )∑∑∫∫
≠−−=
≠−−= λ
Γ=λ
ξΓξ=ξξ=p
0k1pk k
Tkk**
k
p
0k1pk k
Tkk*
k
b
a
b
ab,a;t,xt,,xd
L
1t,,x*Kd
L
1b,a;t,x**K
bbbb
( ) ( )t,,xdL
1b,a;t,x *
k
b
a
**k ξΓξ=Γ ∫
From the definition, we get:
DIE/UNIBO – Aprile 2001 21 di 37
( ) ( )
−ξ+γλ−
+λ
+
−ξ+γλ−
−λ
+
+
−ξ−γλ−
+λ
−
−ξ−γλ−
−λ
−
γ−
λ+∞ξΓ=ξΓ
γ−−ξ+γ−−ξ+−
γ−−ξ−γ−−ξ−−
∞
−∞=∑
nL2
x
t
rterfcenL
2
x
t
rterfce
nL2
x
t
rterfcenL
2
x
t
rterfce
r8
L;,x*t;,x*
kk
k
rnL2xkk
k
rnL2x
kk
k
rnL2xkk
k
rnL2x
n
kkkk
kk
kk
( )( ) ( ) ( )
−
γ−ξγ−−−γ
−λ=∞ξΓγ−
γ−ξ+−γ−ξ−−
1e
rsinhrxsinh
4
ee
rL;,x*
k
kk
rL2
kkrxrx
kkk
Therefore:
( ) ( )( ) ( ) ( )
−
γ−ξγ−−−γ
−λξ=
=∞ξΓξ=∞Γ
γ−
γ−ξ+−γ−ξ−−
∫
∫
1e
rsinhrxsinh
4
ee
rd
;,xdL
1b,a;,x
k
kk
rL2
kkrxrx
kk
b
a
*k
b
a
**k
Assuming ( ) krxs γ−−ξ=− , ( ) krxs γ−+ξ=+ , krs γ−ξ= and thus krddsdsds γ−ξ=== +− ,
it result:
( )( )( )
( )( ) ( ) ( )
( )
( )
( )
( )
( ) ( )[ ]
( ) ( )
( ) ( )[ ] b
akrL2
kk
b
a
rxkb
a
rxk
rbs
rasrL2
kk
rxbs
rxas
sk
rxbs
rxas
sk
rb
rarL2
kkrxb
rxa
skrxb
rxa
sk**k
rcosh1e
rxsinh
r
er4
e1xsgnr4
scosh1e
rxsinh
r
er4
e1)tsgn(r4
dsssinh1e
rxsinh
rdse
r4dse
r4b,a;,x
k
kk
k
kk
k
k
k
k
k
kk
k
k
k
k
=ξ=ξγ−
=ξ
=ξ
γ−+ξ−=ξ
=ξ
γ−−ξ−
γ−=γ−=γ−
γ−+=
γ−+=
−γ−−=
γ−−=
−−
γ−
γ−+
γ−
γ−+
γ−++−γ−−
γ−−−−
γ−ξ−
γ−λ−
+
λ
+
−−ξ
λ=
−
γ−λ−
+
λ
+
−
λ=
−
γ−λ−
λ−
λ=∞Γ
+
+
+
−
−
−
+−
∫∫∫
and finally (o)
:
( )( ) ( ) ( ) ( )
b
a
rL2
kk
rxrx
k**k
1e
rcoshrxsinh
4
ee1xsgn
rb,a;,x
k
kk
=ξ
=ξ
γ−
γ−+ξ−γ−−ξ−
−
γ−ξγ−−
+
−−ξλ
=∞Γ
Regarding the transient term, it is:
(o) Note that ( )[ ] qu
0u
uq
0
ue1)usgn(due
==
−− −=∫
DIE/UNIBO – Aprile 2001 22 di 37
( ) ( )
−ξ+γλ−
+λ
+
−ξ+γλ−
−λ
+
+
−ξ−γλ−
+λ
−
−ξ−γλ−
−λ
−
ξγ
−λ
+∞Γ=Γ
γ−−ξ+γ−−ξ+−
γ−−ξ−γ−−ξ−−
∞
−∞=∫∑
nL2
x
t
rterfcenL
2
x
t
rterfce
nL2
x
t
rterfcenL
2
x
t
rterfce
dr8
b,a;,xb,a;t,x
kk
k
rnL2xkk
k
rnL2x
kk
k
rnL2xkk
k
rnL2x
b
an
kk**k
**k
kk
kk
Assuming ( ) krnL2xw γ−−ξ−=− , ( ) krnL2xw γ−−ξ+=+ and thus krddwdw γ−ξ==− +− ,
it result:
( ) ( )
( )( )
( )( )
λ+
λ+
λ−
λ+
+
λ+
λ+
λ−
λ
λ=∞Γ−Γ
++−γ−−+
γ−−++
−−−γ−−−
γ−−−−
∞
−∞=
++
−−
∫
∫
∑
wrt2
1rterfcew
rt2
1rterfcedw
wrt2
1rterfcew
rt2
1rterfcedw
r8b,a;,xb,a;t,x
k
k
wk
k
wrnL2bx
rnL2ax
k
k
wk
k
wrnL2bx
rnL2ax
n
k**k
**k
k
k
k
k
Substitution of the relations
( )qu
0u
2
1
uq
0
uuerfeu
2
1erfce
2
1erfc)usgn(duu
2
1erfce
2=
=
Ω−−−
Ω+
Ω−Ω
−
Ω=
Ω−Ω∫
( )qu
0u
2
1
uq
0
uuerfeu
2
1erfce
2
1erfc)usgn(duu
2
1erfce
2=
=
Ω−
Ω−
Ω+Ω
−
Ω−=
Ω+Ω∫
(where rt2
1 kλ=Ω ) leads to:
DIE/UNIBO – Aprile 2001 23 di 37
( ) ( )
( )
( )( )
( )
( )
( )( )
( ) k
k
k
k
k
k
k
k
rnL2bxw
rnL2axw
k
rt
k
k
w
k
k
rt
k
k
w
k
rnL2bxw
rnL2axw
k
rt
k
k
w
k
k
rt
k
k
w
k
n
k**k
**k
wrt2
1erfew
rt2
1rterfce
rterfcwsgn
wrt2
1erfew
rt2
1rterfce
rterfcwsgn
wrt2
1erfew
rt2
1rterfce
rterfcwsgn
wrt2
1erfew
rt2
1rterfce
rterfcwsgn
r8b,a;,xb,a;t,x
γ−−+=
γ−−+=
+λ−
++
+λ−
+−+
γ−−−=
γ−−−=
−λ−
−−
−λ−
−−−
∞
−∞=
+
+
+
+
−
−
−
−
λ−
λ+
λ−
λ−
+
λ+
λ−
λ−
λ
λ−
λ+
λ−
λ−
+
λ+
λ−
λ−
λ
λ=∞Γ−Γ ∑
Simplifying, it result:
( ) ( )
( )
( )
( )
( )
( )
( ) k
k
k
k
k
k
rnL2bxw
rnL2axw
k
rt
k
k
wk
k
w
rnL2bxw
rnL2axw
k
rt
k
k
wk
k
w
n
k
**k
**k
wrt2
1erfe2
wrt2
1rterfcew
rt2
1rterfcewsgn
wrt2
1erfe2
wrt2
1rterfcew
rt2
1rterfcewsgn
r8
b,a;,xb,a;t,x
γ−−−=
γ−−−=
+λ−
+−++
γ−−−=
γ−−−=
−λ−
−−−−∞
−∞=
+
+
++
−
−
−−
λ+
+
λ−
λ−
λ+
λ+
+
λ+
+
λ−
λ−
λ+
λλ
=
=∞Γ−Γ
∑
Substituting:
DIE/UNIBO – Aprile 2001 24 di 37
( ) ( )
( )
( )
b
a
kk
rt
kk
k
rnL2x
kk
k
rnL2x
kk
rt
kk
k
rnL2x
kk
k
rnL2x
n
k**k
**k
nL2xt2
1erfe2nL2x
t2
1rterfce
nL2xt2
1rterfcenL2xsgn
nL2xt2
1erfe2nL2x
t2
1rterfce
nL2xt2
1rterfcenL2xsgn
r8b,a;,xb,a;t,x
kk
k
kk
k
=ξ
=ξ
λ−
γ−−ξ+−
γ−−ξ+
λ−
γ−−ξ−−
γ−−ξ−
∞
−∞=
−ξ+
γλ−+
−ξ+
γλ−−
λ−
+
−ξ+
γλ−+
λ−ξ++
+
−ξ−
γλ−+
−ξ−
γλ−−
λ−
+
−ξ−
γλ−+
λ−ξ−
λ+∞Γ=Γ ∑
Taking into account that for convergence regions it’s convenient to substitute erf = 1 − erfc, we get:
( ) ( ) ( ) ( )[ ]
( )
( )
b
a
kk
rt
kk
k
rnL2x
kk
k
rnL2x
kk
rt
kk
k
rnL2x
kk
k
rnL2x
n
k
n
ba
rt
k**k
**k
nL2xt2
1erfce2nL2x
t2
1rterfce
nL2xt2
1rterfcenL2xsgn
nL2xt2
1erfce2nL2x
t2
1rterfce
nL2xt2
1rterfcenL2xsgn
r8
nL2xsgnnL2xsgner4
b,a;,xb,a;t,x
kk
k
kk
k
k
=ξ
=ξ
λ−
γ−−ξ+−
γ−−ξ+
λ−
γ−−ξ−−
γ−−ξ−
∞
−∞=
∞
−∞=
=ξ=ξ
λ−
−ξ+
γλ−−
−ξ+
γλ−−
λ−
+
−ξ+
γλ−+
λ−ξ++
+
−ξ−
γλ−−
−ξ−
γλ−−
λ−
+
−ξ−
γλ−+
λ−ξ−
λ+
−ξ++−ξ−λ
+∞Γ=Γ
∑
∑
Breaking the first series in positive and negative terms, we have:
( ) ( )[ ] ( ) ( )[ ]
( ) ( ) ( ) ( )[ ]
( )[ ] ( )[ ] ba
rt
kba
rt
k
1n
ba
rt
k
ba
rt
k
n
ba
rt
k
xsgner4
1xsgner4
nL2xsgnnL2xsgnnL2xsgnnL2xsgner4
xsgnxsgner4
nL2xsgnnL2xsgner4
kk
k
kk
=ξ=ξ
λ−
=ξ=ξ
λ−
∞
=
=ξ=ξ
λ−
=ξ=ξ
λ−∞
−∞=
=ξ=ξ
λ−
ξ−λ
=+ξ−λ
=
=+ξ+++ξ−+−ξ++−ξ−λ
+
+ξ++ξ−λ
=−ξ++−ξ−λ
∑
∑
Note that, being n > 1, 0 < x < L and 0 < ξ < L, it result x − ξ − 2nL < 0, x + ξ − 2nL < 0, x − ξ + 2nL >
0, x + ξ + 2nL > 0 and thus the sign series is null. Substitution leads to:
DIE/UNIBO – Aprile 2001 25 di 37
( ) ( ) ( )[ ]
( )
( )
b
a
kk
rt
kk
k
rnL2x
kk
k
rnL2x
kk
rt
kk
k
rnL2x
kk
k
rnL2x
n
kba
rt
k**k
**k
nL2xt2
1erfce2nL2x
t2
1rterfce
nL2xt2
1rterfcenL2xsgn
nL2xt2
1erfce2nL2x
t2
1rterfce
nL2xt2
1rterfcenL2xsgn
r8xsgne
r4b,a;,xb,a;t,x
kk
k
kk
k
k
=ξ
=ξ
λ−
γ−−ξ+−
γ−−ξ+
λ−
γ−−ξ−−
γ−−ξ−
∞
−∞=
=ξ=ξ
λ−
−ξ+
γλ−−
−ξ+
γλ−−
λ−
+
−ξ+
γλ−+
λ−ξ++
+
−ξ−
γλ−−
−ξ−
γλ−−
λ−
+
−ξ−
γλ−+
λ−ξ−
λ+ξ−
λ+∞Γ=Γ ∑
Note that the erfc functions in the series do not tend a 0, for increasing n, since the argument is negative
and decreasing. In order to improve the convergence the following substitution can be done: erfc(−x) = 1
− erf(−x) = 1 + erf(x) = 2 − erfc(x). The remaining series can be treated separately:
( ) ( ) ( )[ ]
( ) ( )
( )
( )
b
a
kk
rt
kk
k
rnL2x
kk
k
rnL2x
kk
rt
kk
k
rnL2x
kk
k
rnL2x
n
k
n
b
a
rnL2xrnL2xk
ba
rt
k**k
**k
nL2xt2
1erfce2nL2x
t2
1rterfce
nL2xt2
1rterfcenL2xsgn
nL2xt2
1erfce2nL2x
t2
1rterfce
nL2xt2
1rterfcenL2xsgn
r8
enL2xsgnenL2xsgnr4
xsgner4
b,a;,xb,a;t,x
kk
k
kk
k
kk
k
=ξ
=ξ
λ−
γ−−ξ+−
γ−−ξ+
λ−
γ−−ξ−−
γ−−ξ−
∞
−∞=
∞
−∞=
=ξ
=ξ
γ−−ξ+−γ−−ξ−−
=ξ=ξ
λ−
−ξ+
γλ−−
−ξ+
γλ−+
λ−+
+
−ξ+
γλ−+
λ−ξ++
+
−ξ−
γλ−−
−ξ−
γλ−+
λ−+
+
−ξ−
γλ−+
λ−ξ−
λ+
+
−ξ++−ξ−
λ−
+ξ−λ
+∞Γ=Γ
∑
∑
In fact:
DIE/UNIBO – Aprile 2001 26 di 37
( ) ( )
( ) ( )
( ) ( )
( ) ( ) =+ξ+++ξ−+
+−ξ++−ξ−
λ−
ξ++ξ−
λ−=
=
−ξ++−ξ−
λ−
γ−+ξ+−γ−+ξ−−
γ−−ξ+−γ−−ξ−−
∞
=
γ−ξ+−γ−ξ−−
∞
−∞=
γ−−ξ+−γ−−ξ−−
∑
∑
kk
kk
kk
kk
rnL2xrnL2x
rnL2xrnL2x
1n
krxrxk
n
rnL2xrnL2xk
enL2xsgnenL2xsgn
enL2xsgnenL2xsgn
r4exsgnexsgn
r4
enL2xsgnenL2xsgnr4
( ) ( )
( ) ( ) ( ) ( )
( ) ( )
( ) ( ) ( ) ( )
( ) ( )
( ) ( )
( ) ( ) ( ) ( )1e
rcoshrxsinh
reexsgn
r4
rcoshe2rcoshe2
1e
1
r4eexsgn
r4
eeee
er4
eexsgnr4
eeee
r4eexsgn
r4
k
kk
kk
k
kk
kkkk
kkk
kkkk
kk
rL2
kkkrxrxk
krx
krx
rL2
krxrxk
rxrxrxrx
1n
rnL2krxrxk
rnL2xrnL2xrnL2xrnL2x
1n
krxrxk
−
γ−ξγ−λ+
+ξ−
λ−=
=γ−ξ+
γ−ξ−
−
λ−
+ξ−
λ−=
=++
−−
λ−
+ξ−
λ−=
=++
−−
λ−
+ξ−
λ−=
γ−γ−ξ+−γ−ξ−−
γ−−γ−
γ−γ−ξ+−γ−ξ−−
γ−ξ+−γ−ξ−−γ−ξ+γ−ξ−
∞
=
γ−−γ−ξ+−γ−ξ−−
γ−+ξ+−γ−+ξ−−γ−−ξ+γ−−ξ−
∞
=
γ−ξ+−γ−ξ−−
∑
∑
Therefore, finally:
( ) ( )
( ) ( ) ( ) ( )
( )
( )
b
a
kk
rt
kk
k
rnL2x
kk
k
rnL2x
kk
rt
kk
k
rnL2x
kk
k
rnL2x
n
k
b
a
rL2
kkkrxk
rt
rxk
**k
**k
nL2xt2
1erfce2nL2x
t2
1rterfce
nL2xt2
1rterfcenL2xsgn
nL2xt2
1erfce2nL2x
t2
1rterfce
nL2xt2
1rterfcenL2xsgn
r8
1e
rcoshrxsinh
re
r4eexsgn
r4
b,a;,xb,a;t,x
kk
k
kk
k
k
kkk
=ξ
=ξ
λ−
γ−−ξ+−
γ−−ξ+
λ−
γ−−ξ−−
γ−−ξ−∞
−∞=
=ξ
=ξ
γ−γ−ξ+−λ
−γ−ξ−−
−ξ+
γλ−−
−ξ+
γλ−+
λ−+
+
−ξ+
γλ−+
λ−ξ++
+
−ξ−
γλ−−
−ξ−
γλ−+
λ−+
+
−ξ−
γλ−+
λ−ξ−
λ+
+
−
γ−ξγ−λ+
λ−
−ξ−
λ−
+∞Γ=Γ
∑
DIE/UNIBO – Aprile 2001 27 di 37
8. Perfect Conductor Case (r = 0)
In the particular case in which r → 0, it’s possible to specialize the form of Γ* and of Γ** as follow:
( )( ) ( ) ( )
( )
( )
ξ−
ξ−−ξ+γλ−=
−γ−+γ−ξγ−
−γ−ξ++−γ−ξ−−γ
−λ=
−
γ−ξγ−−−γ
−λ=∞ξΓ
→
γ−
γ−ξ+−γ−ξ−−
L2
x
4
xxL
1rL21
rrx
4
rx1rx1
rL
1e
rsinhrxsinh
4
ee
rL;,x*
kk
k
kkkkkk
0r
rL2
kkrxrx
kkk
k
kk
Thus:
( ) ( )
ξ−
ξ−−ξ+γλ−=∞ξΓ
→
L
x
2
xx
2
L;,x* kk
0r
k
Moreover:
( ) ( )
−ξ+γλ−
+λ
+
−ξ+γλ−
−λ
+
+
−ξ−γλ−
+λ
−
−ξ−γλ−
−λ
−
γ−
λ=∞ξΓ−ξΓ
γ−−ξ+γ−−ξ+−
γ−−ξ−γ−−ξ−−
∞
−∞=∑
nL2
x
t
rterfcenL
2
x
t
rterfce
nL2
x
t
rterfcenL
2
x
t
rterfce
r8
L;,x*t;,x*
kk
k
rnL2xkk
k
rnL2x
kk
k
rnL2xkk
k
rnL2x
n
kkkk
kk
kk
thus
( ) ( )
( )
( ) +
λπ−
−ξ−γλ−
−γ−−ξ−−−
+
λπ−
−ξ−γλ−
+γ−−ξ−+−
γ−
λ=∞ξΓ−ξΓ
−ξ−γλ
−ξ−γλ
∞
−∞=
→
∑
k
nL2
x
tkkk
k
nL2
x
tkkk
n
kk0r
kk
rte
2nL
2
x
terf1rnL2x1
rte
2nL
2
x
terf1rnL2x1
r8
L;,x*t;,x*
2kk
2kk
( )
( )
λπ−
−ξ+γλ−
−γ−−ξ+++
+
λπ−
−ξ+γλ−
+γ−−ξ+−+
−ξ+γλ
−ξ+γλ
k
nL2
x
tkkk
k
nL2
x
tkkk
rte
2nL
2
x
terf1rnL2x1
rte
2nL
2
x
terf1rnL2x1
2kk
2kk
DIE/UNIBO – Aprile 2001 28 di 37
−γλ−π
+
+
−ξ+γλ−
−ξ+−
−ξ−γλ−
−ξ−
γ−γ
−λ
=
−ξ+γλ
−ξ−γλ
∞
−∞=∑
2kk
2kk nL
2
x
tnL
2
x
t
kk
kkkk
n
kkk
eet2
nL2
x
terfnL2xnL
2
x
terfnL2x
r2r8
L
and finally:
( ) ( )
−γλ−π
+
+
−ξ+
γλ−−ξ+−
−ξ−
γλ−−ξ−
λγ−=∞ξΓ−ξΓ
−ξ+γλ
−ξ−γλ
∞
−∞=
→
∑
2kk
2kk nL
2
x
tnL
2
x
t
kk
kkkk
n
kk0r
kk
eet2
nL2xt2
1erfnL2xnL2x
t2
1erfnL2x
4
L;,x*t;,x*
Note that while for the last term of this series the convergence is assured for n → ±∞, the sum of the first
two terms is semi-convergent since goes to ±2ξ for n → ±∞. In order to improve the convergence for n
→ ±∞ the following the operations are performed:
1° the substitution erf(x) = 1 − erfc(x)
2° the separation of the series from − ∞ to +∞ in two series (changing sign to the negative n):
( ) ( ) [ ]
−γλ−
πλγ
−
−ξ+
γλ−−ξ++
−ξ−
γλ−−ξ−−
λγ−−ξ+−−ξ−
λγ−=∞ξΓ−ξΓ
−ξ+γλ
−ξ−γλ∞
−∞=
∞
−∞=
∞
−∞=
→
∑
∑∑
2kk
2kk nL
2
x
tnL
2
x
t
kkn
kk
kkkk
n
kk
n
kk0r
kk
eet2
4
L
nL2xt2
1erfcnL2xnL2x
t2
1erfcnL2x
4
LnL2xnL2x
4
L;,x*t;,x*
[ ]
∑
∑
∞
=
∞
=
+ξ+
γλ−+ξ++
+ξ−
γλ−+ξ−−
λγ−
+
ξ+
γλ−ξ+−
ξ−
γλ−ξ−
λγ+
+
+ξ+−+ξ−+−ξ+−−ξ−+ξ+−ξ−λγ
−=
1n
kkkkkk
kkkkkk
1n
kk
nL2xt2
1erfcnL2xnL2x
t2
1erfcnL2x
4
L
xt2
1erfcxx
t2
1erfcx
4
L
nL2xnL2xnL2xnL2xxx4
L
DIE/UNIBO – Aprile 2001 29 di 37
−γλ−
πλγ
−
−ξ+
γλ−−ξ++
−ξ−
γλ−−ξ−−
−ξ+γλ
−ξ−γλ∞
−∞=∑
2kk
2kk nL
2
x
tnL
2
x
t
kkn
kk
kkkk
eet2
4
L
nL2xt2
1erfcnL2xnL2x
t2
1erfcnL2x
Note that, being n > 1, 0 < x < L and 0 < ξ < L it result x − ξ − 2nL < 0, x + ξ − 2nL < 0, x − ξ + 2nL >
0, x + ξ + 2nL > 0 and thus the first series is null. Moreover, the separation of the last series, evidencing
the zero-th order terms e simplifying, gives:
( ) ( )
γλ−π−
−ξ+
γλ−−ξ++
+γλ−π
+
−ξ−
γλ−−ξ−−
+γλ−π
−
+ξ+
γλ−+ξ++
+γλ−π
+
+ξ−
γλ−+ξ−−
λγ−
+
γλ−π−
ξ+
γλ−ξ+−
+γλ−π
+
ξ−
γλ−ξ−
λγ−=
=∞ξΓ−ξΓ
−ξ+γλ
−ξ−γλ
+ξ+γλ
+ξ−γλ∞
=
ξ+γλ
ξ−γλ
→
∑
2kk
2kk
2kk
2kk
2kk
2kk
nL2
x
t
kk
kk
nL2
x
t
kk
kk
nL2
x
t
kk
kk
nL2
x
t
kk1n
kkkk
2
x
t
kk
kk
2
x
t
kk
kkkk
0r
kk
et2
nL2xt2
1erfcnL2x
et2
nL2xt2
1erfcnL2x
et2
nL2xt2
1erfcnL2x
et2
nL2xt2
1erfcnL2x
4
L
et2
xt2
1erfx
et2
xt2
1erfx
4
L
;,x*t;,x*
It’s now possible to eliminate all the modules thanks to the symmetry of the erf function and being n > 1,
0 < x < L and 0 < ξ < L, so that it result x − ξ − 2nL < 0, x + ξ − 2nL < 0, x − ξ + 2nL > 0, x + ξ + 2nL >
0 e x + ξ > 0:
DIE/UNIBO – Aprile 2001 30 di 37
( ) ( )
( ) ( )
( ) ( )
( ) ( )
( ) ( )
( ) ( )
( ) ( )
γλ−π−
−ξ+
γλ−−−ξ+−
+γλ−π
+
−ξ−
γλ−−−ξ−+
+γλ−π
−
+ξ+
γλ−+ξ++
+γλ−π
+
+ξ−
γλ−+ξ−−
λγ−
+
γλ−π−
ξ+
γλ−ξ+−
+γλ−π
+
ξ−
γλ−ξ−
λγ−=
=∞ξΓ−ξΓ
−ξ+γλ
−ξ−γλ
+ξ+γλ
+ξ−γλ∞
=
ξ+γλ
ξ−γλ
→
∑
2kk
2kk
2kk
2kk
2kk
2kk
nL2
x
t
kk
kk
nL2
x
t
kk
kk
nL2
x
t
kk
kk
nL2
x
t
kk1n
kkkk
2
x
t
kk
kk
2
x
t
kk
kkkk
0r
kk
et2
nL2xt2
1erfcnL2x
et2
nL2xt2
1erfcnL2x
et2
nL2xt2
1erfcnL2x
et2
nL2xt2
1erfcnL2x
4
L
et2
xt2
1erfx
et2
xt2
1erfx
4
L
;,x*t;,x*
Regarding Γ**, from the definition it result:
( ) ( )t,,xdL
1b,a;t,x *
k
b
a
**k ξΓξ=Γ ∫
thus (*)
, assigning ( )x−ξ=σ and ξ=σ dd , it result:
( ) ( )( )( )
σσ−
ξ−ξ+γλ−=∞Γ ∫
−
−
=ξ
=ξ
→ xb
xa
b
a
22kk
0r**
k d2
1
L2
x
4
x
2b,a;,x
and finally:
( ) ( ) ( )( ) b
a
222kk
0r**
kL
x
2
xxsgnx
4b,a;,x
=ξ
=ξ
→
ξ−ξ−−ξ−ξ+γλ−=∞Γ
for what concerns the transient term, assigning ( )nL2xt2
1Lx
kk θ+ξ−θγλ−
=ω− ,
( )nL2xt2
1Lx
kk θ+ξ+θγλ−
=ω+ , with θx,θL = 0, ±1 and ξγλ−
=ω=ω− +− dt2
1dd kk it result:
(*)Note that
qu
0u
2q
0 2
u)usgn(duu
=
=
=∫
DIE/UNIBO – Aprile 2001 31 di 37
( ) ( )
( ) ( )
( ) ( )
( ) ( )
( ) ( )
( ) ( )
( ) ( )
ω
λγπ−ωωω
λγ+
+ωλγπ
−ωωωλγ
+
+ωλγπ
+ωωωλγ
−
+ωλγπ
+ωωω
λγ−
λγ−
+
ω
λγπ+ωωω
λγ+
+ωλγπ
+
ωωω
λγλγ
−
=∞Γ−Γ
+ω−
+++
−ω−
−−−
+ω−
+++
−ω−
∞
=−−−
+ω−
+++
−ω−
−−−
→
∫∫
∫∫
∫∫
∫∑ ∫
∫∫
∫∫
+
+
++
+
−
−
−−
−
+
+
++
+
−
−
−−
−
+
+
++
+
−
−
−−
−
det22
derfct4
det22
derfct4
det22
derfct4
det22
derfct4
4
det22
derft4
det22
derft4
4
b,a;,xb,a;t,x
b
akk
b
akk
b
akk
b
akk
b
akk
b
akk
b
akk1n
b
akk
kk
b
akk
b
akk
b
akk
b
akk
kk
0r**
k**
k
2
2
2
2
2
2
(
(
(
(
(
(
(
(
)
)
)
)
)
)
)
)
&&&&&&
&&&&&&
where
( )axt2
1a kk −γλ−=− , ( )bx
t2
1b kk −γλ−=− ,
( )axt2
1a kk +γλ−=+ , ( )bx
t2
1b kk +γλ−=+
( )nL2axt2
1a kk +−
γλ−=−
), ( )nL2bx
t2
1b kk +−
γλ−=−
),
( )nL2axt2
1a kk ++
γλ−=+
), ( )nL2bx
t2
1b kk ++
γλ−=+
)
( )nL2axt2
1a kk −+
γλ−−=−
(, ( )nL2bx
t2
1b kk −+
γλ−−=−
(,
( )nL2axt2
1a kk −−
γλ−−=+
(, ( )nL2bx
t2
1b kk −−
γλ−−=+
(
Simplifying e taking into account the relations
( )qu
0u
u2q
0
uq
0
22
eu
)u(erf2
1udue
2duuerfu2
=
=
−−
π+
+=π
+ ∫∫
( )qu
0u
u2q
0
uq
0
22
eu
)u(erfc2
1udue
2duuerfcu2
=
=
−−
π−
+=π
− ∫∫
we get:
DIE/UNIBO – Aprile 2001 32 di 37
( ) ( )
( ) ( ) ( ) ( ) ( ) ( )
( ) ( ) ( ) ( ) ( ) ( )
( ) ( ) ( ) ( ) ( ) ( )
πω
−ω
+ω+
πω
−ω
+ω+
+
πω
−ω
+ω
−
πω
−ω
+ω−−
+
πω
+ω
+ω+
πω
+ω
+ω−
=∞Γ−Γ
−
−
−+
+
+
+
+
+−
−
−
+
+
+−
−
−
ω−−−−
ω−+++
ω−+++
∞
=
ω−−−−
ω−+++
ω−−−−
→
∑
b
a
2b
a
2
b
a
2
1n
b
a
2
b
a
2
b
a
2
0r**
k**
k
22
22
22
eerfc2
1eerfc
2
1
eerfc2
1eerfc
2
1
2
t
eerf2
1eerf
2
1
2
t
b,a;,xb,a;t,x
(
(
(
(
)
)
)
)
&&&&&&
and finally:
( ) ( )
( ) ( ) ( ) ( )
( ) ( ) ( ) ( )
( ) ( ) ( ) ( )
( ) ( ) ( ) ( )
( ) ( ) ( ) ( )
( ) ( ) ( ) ( )b
a
nL2xt4kkkk2kk
nL2xt4kkkk2kk
nL2xt4kkkk2kk
nL2xt4kkkk2kk
1n
b
a
xt4kkkk2kk
xt4kkkk2kk
0r**
k**
k
2kk
2kk
2kk
2kk
2kk
2kk
enL2xt
nL2xt2
1erfcnL2x
t21
enL2xt
nL2xt2
1erfcnL2x
t21
enL2xt
nL2xt2
1erfcnL2x
t21
enL2xt
nL2xt2
1erfcnL2x
t21
4
tex
tx
t2
1erfx
t21
ext
xt2
1erfx
t21
4
t
b,a;,xb,a;t,x
=ξ
=ξ
−ξ+γλ
−ξ−γλ
+ξ+γλ
+ξ−γλ
∞
=
=ξ
=ξ
ξ+γλ
ξ−γλ
→
−ξ+
πγλ−
+
−ξ+
γλ−−
−ξ+γλ
−+
+−ξ−π
γλ−+
−ξ−
γλ−−
−ξ−γλ
−+
++ξ+π
γλ−+
+ξ+
γλ−
+ξ+γλ
−−
+
+ξ−
πγλ−
+
+ξ−
γλ−
+ξ−γλ
−−
−
ξ+
πγλ−
+
ξ+
γλ−
ξ+γλ
−+
+
ξ−
πγλ−
+
ξ−
γλ−
ξ−γλ
−−
=∞Γ−Γ
∑
References
1 I.S. Gradshteyn, I.M. Ryzhik, Tables of Integrals, Series, and Products, Academic Press: London, 1980, p.
1112.
2 I.S. Gradshteyn, I.M. Ryzhik, Tables of Integrals, Series, and Products, Academic Press, London, 1980, p.
29.
3 I.S. Gradshteyn, I.M. Ryzhik, Tables of Integrals, Series, and Products, Academic Press, London, 1980, p.
921.
4 L. Krempasky, C. Schmidt, “Theory of Supercurrents and their influence on field quality and stability of su-
perconducting magnets”, J. Appl. Phys. 78 (9) 1995
5 R. Courant, D. Hilbert, Method of Mathematical Physics, Interscience, NY,1962, vol.2, p. 200; vol.1, p. 75.
6 I.S. Gradshteyn, I.M. Ryzhik, Tables of Integrals, Series, and Products, Academic Press, London, 1980, p.
921.
DIE/UNIBO – Aprile 2001 33 di 37
APPENDIX A – FORCED CURRENT DISTRIBUTION IN A 3-STRAND CABLE
Consider, as shown in the figure, a 3-strand cable with an insulated input strand. Moreover, assume that
the flowing current is equally distributed on the two input strand and on the three output strands.
I(t)/3
I(t)/3
I(t)/3
I(t)/2
I(t)/2
The longitudinal resistance (r) ant the external voltage (v) are assumed to be zero. Consequently, the cur-
rents flowing in the 3-strand cable are described by the following system:
(P)
[ ][ ] ( ) ( )
( )( ) ( ) ( ) ( )
( ) ( ) ( ) ( )
======
======
==
=∂∂+
∂∂
3
tIt,Lxit,Lxit,Lxi
2
tIt,0xit,0xi,0t,0xi
00t,x
0t,xx
t,xt
MG
321
321
2
2
i
ii
and i, v ∈ R3
with I(0) = 0 and where [M] and [G] are the following constant-coefficient circulant symmetric ma-
trixes:
[ ]
=
111212
121112
121211
mmm
mmm
mmm
M [ ]
−−
−=
121212
121212
121212
g2gg
gg2g
ggg2
G
Defining the orthonormal spectral basis bk, with k = 1, 0, −1, that it’s the same for [M] and [G]:
=1
1
1
3
1 0b
−−=
2/1
2/1
1
3
2 1b
−=−
1
1
0
2
1 1b
the matrixes [M] and [G] can be written as
[ ] T111
T111
T000M −−−λ+λ+λ= bbbbbb [ ] T
111T111
T000G −−−γ+γ+γ= bbbbbb
with eigenvalues given by
12110 m2m +=λ 00 =γ
121111 mm −=λ=λ − 1211 g3−=γ=γ −
Decomposing i as follows
( ) ( ) ( ) ( ) 111100 t,xt,xt,xt,x −−η+η+η= bbbi
Problem (P) is simply divided in the following three problems:
(P0)
( )
( ) ( ) ( ) ( ) ( )
==η==η==η
=∂
η∂
3
tIt,Lx,
3
tIt,0x,00t,x
0t,xx
000
2
02
DIE/UNIBO – Aprile 2001 34 di 37
(P−1) ( ) ( )
( ) ( ) ( )
==η==η==η
=∂
η∂+∂η∂λγ
−−−
−−−−
0t,Lx,0t,0x,00t,x
0t,xx
t,xt
111
2
12
111
(P1)
( ) ( )
( ) ( ) ( ) ( )
==η−==η==η
=∂
η∂+
∂η∂
λγ
0t,Lx,6
tIt,0x,00t,x
0t,xx
t,xt
111
2
12
111
Problems (P0) and (P−1) have the simple solutions
( ) ( )3
tIt,x0 =η ( ) 0t,x1 =η−
While problem (P1) requires a different approach. Defining
( ) ( ) ( )
−+η=ϕL
x1
6
tIt,xt,x 1
gives
(Pϕ) ( ) ( ) ( )
( ) ( ) ( )
==η==η==η
γ=∂
ϕ∂+∂ϕ∂λγ
0t,Lx,0t,0x,00t,x
t,xut,xx
t,xt
111
12
2
11
which has been already treated and gives (having defined ( ) ( )
−′
λ=L
x1
6
tIt,xu 1 where the prime de-
notes differentiation with respect to the argument)
( ) ( ) ( )τ−ξΓλ
τξτξ=ϕ ∫∫ t,,x,u
ddL
2t,x 1
1
t
0
L
0
Thus
( ) ( ) ( ) ( )τ−ξΓ
ξ−τ′τξ+
−−=η ∫∫ t,,xL
16
Idd
L
2
L
x1
6
tIt,x 1
t
0
L
0
1
and the currents are
( ) ( ) ( ) ( ) ( ) 11
t
0
L
0
0 t,,xL
16
Idd
L
2
L
x1
6
tI
3
tIt,x bbi
τ−ξΓ
ξ−τ′τξ+
−−+= ∫∫
In particular, for a current ramp
( ) ( )
><<
<=′⇒
><<
<=
0
000
00
000
tt fi,0
tt0 fi,t/I
0 tfi,0
tI
tt fi,I
tt0 fi,t/tI
0 tfi,0
tI
this becomes
DIE/UNIBO – Aprile 2001 35 di 37
( ) ( )
( ) ( )
>
−ξΓ−ξΓ
ξ−ξα+
−−+
<<
ξΓ
ξ−ξα+
−α−+α
<
=
∫
∫
010*1
*1
L
0
00
0
01*1
L
0
0
tt fi,tt,,xt,,xL
1dL
2
6L
x1
6
I
3
I
tt0 fi ,t,,xL
1dL
2
6L
x1
6
t
3
t
0 tfi,0
t,x
bb
bbi
where α = I0/t0 is the ramp rate.
APPENDIX B – 3-STRAND CABLE EXCITED AS IN THE KREMPASKY-SCHMIDT CASE
Consider a 3-strand cable with the conditions:
1. I(t) = 0;
2. r = 0,
3. even number of periods,
4. i(0)(x) = 0,
5. time-independent excitation spatially bounded to an interval in the center of the cable.
Hypotheses 3 and 5 lead to write:
( ) 111 )()22
L(U)
22
L(U
2/
1
1
0
)22
L(U)
22
L(U
2/−−− ξν=ξ−δ+δ−−ξ
δΦ=
−+ξ−δ+δ−−ξ
δΦ=ξ bbv
&&
Consequently, the currents flowing in the 3-strand cable are described by the following system:
(P) [ ][ ] ( ) ( ) [ ] ( )
( ) ( ) ( )
======
=∂∂+
∂∂
0t,Lxt,0x0t,x
xGt,xx
t,xt
MG2
2
iii
vii
Decomposing i as follows
( ) ( ) ( ) ( ) 111100 t,xt,xt,xt,x −−η+η+η= bbbi
Problem (P) is simply divided in the following three problems:
(P0) ( )
( ) ( ) ( )
==η==η==η
=∂
η∂
0t,Lxt,0x0t,x
0t,xx
000
2
02
(P1) ( ) ( )
( ) ( ) ( )
==η==η==η
=∂
η∂+∂η∂λγ
0t,Lxt,0x0t,x
0t,xx
t,xt
111
2
12
111
DIE/UNIBO – Aprile 2001 36 di 37
(P−1) ( ) ( ) ( )
( ) ( ) ( )
==η==η==η
νγ=∂
η∂+∂η∂λγ
−−−
−−−−
−−
0t,Lxt,0x0t,x
xt,xx
t,xt
111
112
12
111
Problems (P0) and (P1) have the simple solutions
( ) 0t,x0 =η ( ) 0t,x1 =η
While problem (P−1) can be solved as usual. Note that the currents becomes
( ) ( ) ( ) ( ) ( )t,xit,xi,0t,xit,xt,x 32111 −==⇒η= −− bi
Thus, since
( ) ( ) ( )t,,xdL
2t,x *
11
1
L
0
1 ξΓλ
ξνξ=η
−
−− ∫
The current i2 is given by
( ) ( ) ( )( )
( )( )
δ+δ−Γδλ
Φ=ξΓξδλΦ=ξΓ
λξν
ξ= −−
−
δ+
δ−−−
−
− ∫∫ 2
L,
2
L;t,xt,,xd
Lt,,x
2d
L
2t,xi **
11
*1
2/L
2/L1
*1
1
1
L
0
2
&&
By using the same notation of Krempasky-Schmidt (but note that the numerical values are different
since the eigenvalues for the 3-strand case are different from the 2-strand case) he same solution can be
obtained:
( )
∑
∑
∞
=
τ−
∞
=
πγ
−λ−
−
−−
πξ
π−πλ
=
=
πξ
π
−
πγ
−
λ=ξΓ −−
1n
/tn
22
2112
1n
L
n1t
2
1
11
L
nsin
L
xnsine1
n
1Lg3
L
nsin
L
xnsine1
L
n1t;,x*
2
2
11
and finally
( ) ( )∑∞
=
τ− α
α−πα
=
nodd1n
/tn
2m2 nsinw
xnsine1
n
1I
6t,xi
2