Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
losses in superconductors*
H. Hashizume, T. Sugiura, K. Miya, Y. Ando t, S. Akita t, S. Torii t, Y. Kubota ~ and T. Ogasawara ~
Nuclear Engineering Research Laboratory, University of Tokyo, Tokai, Naka, Ibaraki 319-11, Japan *Komae Research Laboratory, Central Research Institute of Electric Power Industry, 211-1, Komae, Tokyo 201, Japan ~Department of Physics, College of Science and Engineering, Nihon University, Chiyoda, Tokyo 101, Japan A two-dimensional computational code to evaluate a.c. losses in superconductors has been developed using the current vector potential method (T-method), where the vector potential T is defined by V x T = J. The current distributions in both superconductors and normal conductors are calculated while changing the conductivity of the superconductor so that the current density never exceeds the critical current density. The hysteresis and coupling losses evaluated by the numerical code agree with analytical solutions, as far as these are available. In order to verify the validity of the code, an experiment to measure the hysteresis and coupling losses was carried out using Nb-Ti/Cu filaments. The total loss evaluated from the numerical code agrees with that from the experiment. The numerical analysis, however, indicates that the hysteresis loss under quasi-steady magnetic field is less than the loss in the superconductor under transient field.
xH
=J
(1)
_(aBo
vxE=
X7 B = 0
\0t
+ot-I
OBe'~
(2) (3)
J = aE
(5)
where B0 and Be are the external magnetic field and the field induced by the eddy current, respectively. Since the object is to evaluate the eddy current induced in the conductors, it is very convenient to introduce the current vector potential, T, which is defined only in the conductor region with omission of the variables in space.
601
f with the following Coulomb gauge and boundary conditions, which verify the uniqueness of the solution s v T = 0
n x T = 0
osE, as
J~ IE l '
(lEt # 0)
(13)
j =
[a J,
(7)
(8)
k.~
(IEI =o)
(14)
The following vector identity for arbitrary vector A(P), which relates to the Helmholtz theorem, is very useful (for the derivation see Appendix)
where as = conductivity of the superconductor. Since it is very difficult to determine numerically whether the absolute value of E is zero or not, the above equations are not directly available in the numerical analysis. Therefore the following iterative model, which is one modification of the Bean model, is introduced: 1 initial value of o~ is assumed sufficiently large; and 2 if IJI > Jc, then
L
o~,,=,,+t) = I ~
L
a.,=,,) - IEI
(15)
J
where 0* = 1/[47rR(P, P')] and superscript ' refers to the quantity at point P'. [See Appendix for C(P). ] Replacing A with T in Equation (9) gives the following equation
T(P) = ~
l I
J x v R(P, P') dV
(10)
1 f (n T ) V - - 1 dS 4w R(P, P')
where Equations (6)-(8) are substituted. The first term on the right-hand side of Equation (10) corresponds to the induced field in accordance with Biot-Savart's law. Therefore the relationship between Be and T becomes 0 Be(P) = goT + g4 r I T , V dS (11)
This model has a different characteristic from the Bean model in that the absolute value of current density can change from zero to Jc continuously. When IE I nearly equals zero, IJI is less than J~. Therefore the step to calculate IE I is avoided. The condition given by Equation (14) is also satisfied since the initial value of as is very large. Thus by introducing the above scheme, it is possible to treat the superconductor as if it is a particular kind of normal conductor which has non-linearity in the constitutive equation. Due to this non-linearity, a transient code is necessary even when the applied field changes sinusoidally. The calculation of Equation (15) is repeated to convergence. Since the purpose of this study is to verify the validity of the model given by Equation (15), two-dimensional analysis is enough from the viewpoint of reducing the cost and complexity. The governing equation in two dimensions, where the current distribution through the thickness is assumed constant, is deformed from Equation (12) as f o l l o w s 5'7 n. V x1 0Bo V x(T,n)=-n.---g0--
R(P, P')
cgt
OTn Ot
where T, represents n T. The governing equation based on the T-method is derived from Equations (2), (5), (6) and (11) as follows
1
o
_..On. 47r
(16) where the variable is reduced from the vector T to the scalar T,.
V x-
0Bo v xT . . . .
Ot
0T go--
Ot
(12)
Numerical analysis of a.c. losses Equation (16) is solved based on the finite element method using eight-node isoparametric elements. In order to verify the validity of the numerical code the hysteresis and coupling losses evaluated by the code are compared to analytical ones using simple geometry.
Here the critical state model is introduced in which the viscous force is ignored 4 and, for simplicity, the Bean model 6 is employed. Since a formulation which correctly expresses the Bean model has never been pro-
602
Numerical analysis of a.c. losses: H. Hashizume et al. Hysteresis loss in superconductor Theoretical evaluation of the hysteresis loss using the Bean model is given by Wilson 8 based on onedimensional geometry; a slab of superconductor with the changing field parallel to its broad face. The ioss per cycle (in J m -3) is given by Q = __B~p(fl), 2~o
r(fl) =-
These facts indicate that the Bean model can be treated numerically using the scheme proposed here. Coupling loss The theoretical solution for the coupling loss for a sandwich of superconductor and normal metal is also based on the one-dimensional model where the external field changes slowly so that the induced field becomes small enough to be negligible. The coupling loss, Po which is defined as the loss in the normal metal, and the net current across the superconductor, I, et, are evaluated by the following equations
(17)
3
1
fl
r(fl)---
2
3fl 2
forfl> 1
pc__B2m 7r ( 2w ~
7" = /XOO'n
where Bm and Bp are the amplitude of the field oscillation and the amplitude when the field penetrates to the centre of the sample, respectively. As is mentioned in the previous section, the code is two-dimensional and the current distribution through the thickness is assumed constant. Therefore the current distribution is two-dimensionally evaluated on the crosssection of the superconductor normal to the changing field. The geometry considered in the calculation is 5.0 mm length, 0.1 nun width and 10 mm thickness, in order to simulate the broad superconductor assumed in the above model. One quarter of the region of 5.0 x 0.1 mm 2 is divided into 8 x 8 meshes taking symmetry into account. The value of critical current density is assumed to be 2.0 x 1 0 9 A m -2. The results are shown in Figure 1 with the analytical solution given by Equation (17). The hysteresis loss is evaluated by calculating ,1 E in the numerical analysis. The agreement between the numerical and analytical results is excellent. Although it is not shown here how the currents invade into the superconductor when the applied field increases and then decreases, the results agree with the analytical predictions based on the Bean model.
(18)
l.o,(x) = o. B
[(;)1
x 2 -
(19)
where w, a, f, a. and ! are the normal conductor width, filament half-width, frequency, conductivity of the normal conductor and length of the cross-section of the slab, respectively. The rate of change of the field when the critical current flows over the whole cross-section of the slab is called the critical field change rate, Bo which is calculated by setting x = 0 in Equation (19), with the following result Bc = 16 aJc
a n!2
(20)
x 109 A/m 2
,-- superconductor
10
~ o = n o Region divided into mesh
*- normal
metal
II
tun :mm]
(a)
.1
01
]=,
O}
.01
-I-
Analytical solution
.001 ........ ,
.01
Normalized
.1
1
magnetic
10
field
100 change
(b)
Figure 2 Current distribution in superconductor and normal metal. (a) B/B c = 0.47; (b) B/Be = 1.88
603
0.00~
n n 0
m
OO2
Samples
1 .--
0.0010
Q
0.0005.
10 N
BmaxI
E
Z
0
0.0000
!
Pickup coils
Analytical solution
Pulse coil
2.0
Time[msec]
0.0
0.5
1. '0
1. '5
Normalized
Figure
Figure
3 Comparison of coupling losses normal conductor. When the flux change rate is high (transient state), eddy currents are induced in both superconductors and normal conductors. The loss in this case is called the pulse loss. The coupling loss is then evaluated from the difference in the pulse and hysteresis losses. Since the induced field cannot be negligible in this case, the analytical solution for the coupling loss is no longer available. In the numerical analysis the critical current density, Jc, is evaluated at half the maximum magnetic field since the flux change is too large to use Jc at zero field. One quarter of the region under consideration is divided into 8 x 8 meshes and hysteresis and pulse losses are calculated by simulating the change of the applied field. In Figure 5a the experimental and numerical results are compared in terms of the pulse and hysteresis losses. Good agreement between the experimental and numerical results is seen in the figure. This agreement also verifies the validity of the code. Figure 5b shows items of the pulse loss obtained using the numerical analysis. As shown in the figure, the numerical results indicate that the loss in the superconductor at transient state is greater than the hysteresis loss at quasi-steady state, especially when the flux change becomes high. The reason for this difference is that the current distribution in the superconductor is affected by the eddy current induced in the normal conductor, as seen in Figure 2, and the flux in the superconductor increases because of the demagnetization caused by this current. Then the difference between the pulse loss at transient state and the hysteresis loss at steady state, which is denoted by open triangles in Figure 5b and has been considered to give the coupling loss, does not correspond to the loss in the normal conductor denoted by crosses (x) in the figure. Therefore in the experiment using the samples shown in Figure 4, the coupling and hysteresis losses should be evaluated simultaneously, which means it is impossible to distinguish between the hysteresis and coupling losses in the experiment, since all that is measured is the summation of both losses.
The following parameter values are employed in the calculation; a = 0.05 (or 0.025) mm, w = 2 (or 1) mm, f=0.01Hz, o , = 2 . 7 8 X 1 0 9 f l -l, l = 3 0 m m , J = 2 x l09 A m -2 and B~ = 0.64 (or 0.32) T s -] Figures 2a and b show the current distributions when the rate of change of the field normalized by B~ is less than or greater than 1.0, respectively. In Figure 2a the current flowing in the opposite direction in the superconductor is seen, as indicated by the analytical model. In Figure 2b the current in most regions of the superconductor flows in the same direction because the flux change rate is rather high. In Figure 3 the coupling loss normalized by B2m/2~o is plotted for both B~ = 0.64 (case 1) and 0.32 (case 2). Good agreement is observed between the numerical and analytical results. Although the analytical solution does not depend on the rate of change of the magnetic field, the losses obtained by the numerical analysis for both cases 1 and 2 decrease when the rate of change of field becomes larger. This difference is due to the misuse of the analytical solution; that is, the net current near x = 0 exceeds the critical current, as can easily be derived from Equation (19). On the other hand, the net current obtained by the numerical analysis is suppressed less than the critical current, as seen in Figure 2b. Therefore the numerical analysis gives reasonable results for any range of rate of change of field.
Conclusions
The results of the present study can be summarized as follows.
604
References
1 Ciazynski, D. and Turck, B. Cryogenics (1984) 24 507-514
2 Hartmann, R.A., Dijkstra, D., van Beckum, F.P.H. and van de
[] 100
Experimental
Numerical n ~r~] 0 ~ :
O 02.A
) o t0
OI fl
~a
A A
&a
6 4
&
1
Cryogenics (1989) 29 741-747 Matsushita, T. Cryogenics (1990) 30 314-323 Hashizume, H. and Miya, K. Fusion Eng Des (1989) 7 293-321 Bean, C.P. Phys Rev Lett (1962) 8 250 Hashimoto, M., Hashizunm, H., Sugiura, T., Takagi, T., Miya, K. and Seki, M. IEEE Trans Magn (1990) MAG-26 869-872 8 Wilson, M.N. Superconducting Magnets Clarendon Press, Oxford, UK (1983) 9 Kim, Y.B., Hempstead, C.F. and Strnad, A.R. Phys Rev (1963) 129 528 4 5 6 7
. . . . . . . .1
(a)
Magnetic
flux density,B
(Tesla)
1000 Item el pulse loss(numerical.transient) + Loss in superconductor x Loss in normal conductor 0 & Hysteresis loss (experimental. steady) + Pulse minus Hysteresis 0 (~) (experimental) 0 ~ ~A~ 0 U~ U~ 0 .,d 0 10 +
C(P)A(P)= -
I00
-3
+
= V V x ~ ~*AdV(P')
o
&
&
x x
A
AX
x &
= V x i V4~*xAdV(P')
i 1 . . . . . . . .
& . . . . . . . .
1
(b)
Figure 5 Comparison of a.c. losses f r o m experimental and numerical results: (a) Hysteresis and pulse losses; (b) losses in superconductors and normal c o n d u c t o r s
= V x
[t
-
V'4~* x A dV(P')
1 A constitutive equation based on the Bean model and a new numerical scheme to treat this equation are proposed. 2 The validity of this scheme is verified through the calculation of hysteresis and coupling losses in a simple geometry. 3 Through the comparison between the numerical and experimental results, it is revealed that the hysteresis loss at quasi-steady state is less than the loss in the superconductor at transient state. This means that the difference between the total loss at transient state and the hysteresis loss at the steady state does not correspond to the coupling loss when certain types of samples are used. 4 For future work, the dependence of the critical current on the magnetic field should be taken into account, as seen in the model by Kim 9.
=V x [I*V'xAdV(P')
- l'*nxAdS(P') 1
--~7[, i' ~*V-AdV(P')
605
-I
~b*n A dS(P')]
=I V xAx V~*dV(P')
+ I (V A)V4~* dV(P') - I n x A x V4~* dS(P')
606