Stability considerations for design of a high

temperature superconductor
A. Abeln, E. Klemt and H. Reiss

Asea Brown Boveri, Corporate Research Heidelberg, D-6900 Heidelberg, Germany

Received 16 July 199 I; revised 25 September 1991

Predictions based on the well known cryogenic stability criterion for the permissible ratio
J/Jc of transport to critical current density of high Tc superconductors are compared
with a transient analysis. The results demonstrate that calculation of the transient
'temperature history' of the superconductor after a thermal disturbance, as well as
knowledge of the magnitude of this disturbance, is necessary when designing a conductor
for a particular application. While in other published work on stability analysis, constant
heat transfer coefficients were applied, an analysis like the one presented here allows one
to calculate more accurately heat transfer by using the transient temperature difference
between superconductor and coolant.

Keywords: superconductor stability; high Tc superconductors; heat transfer

field) that can be tolerated if flux jumps and quenching

In terms of classical superconductor terminology, are to be avoided. The quantities Jc, To, density and
stability analysis covers the following items (arranged in heat capacity of the superconductor have to be known as
increasing order of complexity): cryogenic stability (the 'input' values to the adiabatic stability formula. In addi-
well known Stekly criterion, which is a simple heat tion, dynamic stability requires knowledge of ct(AT) and
balance), adiabatic stability (predictions based on the of the magnetic diffusivity of the superconductor. Both
critical state model, under the influence of a thermal criteria allow inclusion of the transport current density.
disturbance) and dynamic stability (same basis as for Figure 1 describes schematically the present stability
adiabatic conditions but including thermal and magnetic scenario and the 'input' values for the different stability
diffusion and heat transfer to the coolant). A thermal models (and where the analysis to be presented here is
disturbance can be initiated by, for example, a conductor located in this scheme). These criteria, rather than
movement, a fluctuation of magnetic field or heat leaks. cryogenic stability, have to be observed when a super-
Stekly's criterion assumes that heat produced in the conductor is to be developed for a particular technical
normal conducting state of a superconductor is removed application (additional considerations have to be
by heat transfer to the coolant. It allows an estimate of directed towards an estimate of a.c. losses). With regard
the required ratio P/A of perimeter P and cross-section to thin film preparation, prescription of, for example,
A of a superconductor provided the normal state conductor thickness has a direct impact on materials
resitivity On, critical current density Jc, critical development.
temperature Tc, temperature of coolant Tcool,atand heat The above three classical tools, however, have been
transfer coefficient cKAT) to the coolant are known (AT derived using some simplifying assumptions:
denotes the temperature difference between supercon-
ductor and coolant). In some derivations of this 1 A serious drawback of the classical stability criteria
criterion, we find also the ratio X of the cross-sections is that they do not, or at least do not explicitly, con-
of superconductor to a stabilizing metallic conductor sider transient behaviour (temperature fields, current
(see, for example, Reference 1). distribution) of a superconductor. As will be
Adiabatic and dynamic stability criteria are more discussed below, this may have serious consequences
explicitly directed towards an estimate of conductor for estimates of heat transfer, which always establish
geometry. The literature mostly considers the case of an an important part of a stability analysis. A later sec-
external magnetic field orientated parallel to the surface tion of this paper deals with a comparison of results
of a superconducting slab (again see Reference 1). The obtained using the conventional cryogenic stability
adiabatic stability criterion, under the simplifying criterion and a transient analysis of temperatures in
assumption of homogeneous transport properties, then a high Tc superconductor performed with computer
yields an estimate of the slab thickness (or of penetration . . . . . simulations.
0011 - 2275/92/030269- 10
© 1992 Butterworth - Heinemann Ltd

Cryogenics 1992 Vol 32, No 3 269

Design of a high temperature superconductor: A. Abeln et al.

C l a a a l c a l Stabfllty. Crlterla thermal disturbance, in the presence of a transport cur-

rent and with heat transfer to liquid nitrogen (LN2). It
was also the aim of these investigations to allow
estimates of the maximum transport current density that
can be tolerated in a thin film, band-like superconductor.
c=~roge=~s t e b ~
MInimum Input: Pn. a(AT). The predictions from the computer simulations will be
J. Tc(J) or Jc. Tc. P/A, ~. compared with predictions based on the simple
cryogenic stability criterion. An analysis that also takes

Minimum Input: de, J. Te.

heat capacity, dimension
of slab, orientation of
magnetic field. ~.
IT--
]
Ninlmum Input: like in
Adiabatic Stability plus
a(AT), magnetic diffuslvity
into account anisotropy of transport properties will be
presented in a subsequent publication.

Computer simulations
I
Overall description of the computer program
/ X A computer program has been written for calculating
/ C o m p u t e r Strnulatlorm (this work) \ transient temperature fields in a cooled, thin film high
/ I Tc superconductor (a 'band-like' conductor) in the
1 I presence of a transport current and a thermal distur-
I / bance. Thus the program is to some extent a combina-
I
]rntrinaic Stability /
Minimum Input: de, J , To, heat capacity, /
tion of the conventional cryogenic and adiabatic stability
dimension of conductor, a(Kl3, thermal criteria mentioned above. However, it explicitly yields
conductivity tensor; a complete model /
\
would require the Dyrmmlc Stability Input / transient temperature fields in the superconductor,
plus thermal conductivity tensor, heat
\ resistance to substrate J which establish a sensitive means of stability analysis.
./
\ .1 All calculations are performed with temperature depen-
dent parameters. Like the conventional models, the pro-
gram does not take into account anisotropy of current
Figure 1 Classical stability criteria (cryogenic, adiabatic and and heat transport properties, because the models
dynamic stability), i.e. the present stability scenario, and the required for this purpose do not yet exist. Also, the
relative position ( • ) of computer simulations (this work) and of calculations have been performed without explicitly
the intrinsic stability model (References 4 and 5; see text for
details). The symbols denote the 'minimum input' required for accounting for a magnetic field (the possible use of
application of the different models for stability calculations Jc(B) as an input parameter is the only response of the
present form of this program to a magnetic field B).
Under these conditions, the question the program can
answer is: if all relevant materials parameters of a high
temperature superconductor are given, if the heat
transfer coefficients to the coolant are known, if its
2 The three classical tools assume isotropic transport transport properties are assumed isotropic and if a ther-
properties of current, heat and magnetic diffusion for mal disturbance occurs, what is the maximum allowable
the superconductor. Since these conditions are not transport current for the case of a particular film
fulfilled with high temperature superconductors, new thickness?
stability models have to be established, or the Figure 2a gives an overview of the conductor con-
existing models must be modified thoroughly. In figuration. All calculations were made in one dimension
view of the expected strong anisotropies, for exam- only. The band-like conductor (width 10 mm, with
ple of J~ or of thermal conductivity (considered in variable thickness) is deposited onto a substrate and
the ab plane or in the c direction), it is not permissi- cooled by boiling nitrogen (or forced convection; see
ble simply to apply the above formulations to below). The superconductor is connected at both ends to
ceramic superconductors. Estimates based on a copper current feed-through (no optimization is made
methods that still follow the conventional models (for with respect to the feed-throughs). A finite differences
example in References 2 and 3) can give only orders scheme is applied to define heat balances for each ele-
of magnitude. A step in the right direction has been ment, 1 _<j _< K (using K = 100 or 1000), and to solve
made recently by the introduction of an intrinsic Fourier's differential equation with heat sources and
stability criterion 4'5 which considers anisotropy of sinks.
heat transport in the conductor, but not, as yet, The heat balance for an element N is illustrated in
anisotropic critical current density or magnetic field. Figure 2b. This element exchanges heat by conduction
3 The classical models, and also recently published with its immediate neighbours, by convective or boiling
work on intrinsic stability 4"5, assume a constant and heat transfer to the coolant and by contact heat transfer
uniform heat transfer coefficient or(AT). In reality, to the substrate (or to a buffer layer or a stabilizing
this quantity depends strongly on the temperature metal).
difference AT between cooled surface and coolant. The most difficult problem when establishing heat
balances in these elements is to describe realistically the
In this paper, we will focus our attention primarily source function Qa,t- This function describes heat pro-
on the third problem. We describe computer simula- duction: (2He,t= U X I, either in flux flow or for the
tions for investigating transient temperature fields purely resistive state of each conductor element, with
in a high T~ superconductor under the influence of a voltage U over, and current 1 flowing through, the

270 Cryogenics 1 9 9 2 Vol 32, No 3

 Design of a high temperature superconductor: A. Abeln e t al.

a: Finite Differences Scheme Input data

The computer simulations were performed using
LN 2
//Superconductor experimental data of p of a superconducting thin film of
YBaCuO produced by chemical vapour deposition
. KI Connections (CVD) at the ABB Heidelberg Research Center. Details
'\ of the deposition process are given elsewhere 6.
LN= 8ubstrate A complete experimental p(T,J) diagram, as indicated
in Figure 3, was not yet available for this film. Instead,
we first measured p(J) at constant temperature (77 K).
Figure 4a shows the measured voltage U versus the cur-
b: Detail rent I of a typical CVD sample of 1.1 /zm thickness,
27/zm width and 120/zm length, that was prepared on
a SrTiO3 substrate. The critical current of this sample
aCoollng corresponds to Jc = 1.4 × 105 A cm -2.
LN= The U versus I curve shown in Figure 4a has been
measured by a pulse method. A triangular voltage pro-
QCond file from a function generator is fed (after proper
~--- OHeat amplification) on to a circuit including the sample
N N+I immersed in liquid nitrogen, and an additional ohmic
N-1
resistance. The current increases, by virtue of the
voltage profile, from zero to its maximum value within
V Oco.t,ct Substrate some milliseconds. The voltage over the sample and
over the ohmic resistance is measured in parallel and
LN= amplified by two differential amplifiers and a storage
Oe..=b= oscilloscope. Since sample voltage did not show any
hysteresis, we concluded that the sample temperature
Figure 2 Finite differences scheme for computer simulations. (a) had not risen during the short current pulse.
Overall structure of the conductor; (b) details of heat flow com- For a derivation of the required temperature
ponents and source function for superconductor element N and its dependence of p(T, J), we consider first the case
surroundings
T<L.
A curve U = U(I - Ic) was fitted to the data given in
Figure 4a, using I~ measured at 77 K. From this fit, we
sample. For calculation of U we need, in principle, a calculated the function p = p ( l - I~). This function is
three-dimensional diagram showing the electrical illustrated schematically by the curve C(77 K) in Figure
resistivity p of the superconductor in all its ther- 3. The temperature dependence of p is then accounted
modynamical states, i.e. as a function of T and J, as for as follows: the point P (see Figure 3) of the p versus
shown schematically in Figure 3 (assuming B = 0). The ( J - Jc) curve is shifted along the curve Jc(7), if T
function p(T,J) exceeds zero only in the shaded region. increases, thus giving further curves C(T').
A voltage U and thus heat generation occurs not only if Note this may imply a strong increase in p for
T > Tc (ohmic resistance) but also if J _ J~ (T < To), T _ 77 K if a transport current with density J, indicated
which relies on flux flow resistance. by the dashed line in Figure 3, is constant (because
J~(7) decreases with increasing 7). In this case, J may
be large compared with J~(T). However, even if
9(T,J) Jc(7) = 0, p determined by this procedure is always
very small compared with the corresponding value Pn
taken at T _> T~ in ohmic resistive states (see below and
Figure 4b).
Since the temperature dependence of Jc needed for
T this procedure had not yet been measured using our
sample, we applied a frequently used, linear relation
instead

L(T) = Jc(77 K) × (T~ - T)/(Tc - 77 K) (1)

The source function aHeat for the conductor element N

is then calculated using U = p [ l - l c ( 7 7 K)] ×
(L/A) x [ I - lc(T)], where L and A are the conductor
length and cross-section, respectively, of this element.
Next, we consider the case T >_ Tc. Production of
J Joule heat is calculated again using experimental values
Figure 3 Schematic diagram showing specific electrical of electrical resistivity p,(T) obtained with a CVD film.
resistance p(T, J) (J = current density, T = temperature). The
shaded region denotes resistive states of the superconductor and
These measurements were made using a sample
the dashed line corresponds to the transport current density J. mounted on a cold head. The temperature of the sample
See text for definitions of curves C(T') and C(77 K) was measured using a resistance thermometer. Voltage

Cryogenics 1992 Vol 32, No 3 271

Design of a high temperature superconductor: A. Abeln et al.

u (77K) [pv] and thus specific resistance p. were detected by applying

a small current (of several microamperes) to the sample
during cooling from 300 K to temperatures below To.
The source function is then calculated using
U = p,(T) × (L/A) × I, with linear approximations in
3001 . the two sections a and b to the measured values of
p,(T), as shown schematically in Figure 4b. A value
p* = p,(98 K) = 3.57 × 10 -7 ~ m was detected at the
intercept of the two linear approximations (filled circle
200 ] * in Figure 4b).
It is understood that the data given in Figure 4a, the
value p,* and the slope of the linear fits to p, in the sec-
tions a and b are valid only with reference to a single

1ool
thin film sample, and they might not be representative
for a large collection of other samples. A plot of
0,(300 K) versus Jc using data taken from the literature
$ * for a variety of high T~ samples is shown in Figure 4c.
Compared with the (solid line) fit curve, the Pn values
applied in the computer calculations are smaller, but
Ic they are within the scatter of data observed in this figure,
Heat transfer by conduction Qco,a from element N
Ol ~, i i i J
L (Figure 2b) to its neighbours is calculated using the
a 40 60 80 I (mA) temperature dependent (isotropic) thermal conductivity
of the bulk material. Measurements of anisotropic ther-
mal conductivity of thin films are not yet available. We
plan to measure these quantities using non-stationary
9. techniques, such as laser flash (an experimental pro-
gramme has recently been started). The temperature
dependent heat capacity is also taken from the bulk
material. The data for thermal conductivity and heat
capacity used here are from Reference 7.
No experimental data for heat resistances between thin
b film superconductors and substrates (or buffer layers or
stabilizing metal) are currently available (so we will
measure this quantity also). Accordingly, the heat
transfer OCo.t.ct between a conductor element and a
substrate is accounted for by assuming a conduction con-
tact heat transfer coefficient of 1000 W m -2 K -]. This
value is a worst case assumption in that it presumably
a underestimates the thermal coupling of the superconduc-
tor to part of its surroundings, according to experiences
with heat resistances of thin metal films with A12038. In
T other words, the heat resistance assumed here is possibly
b Tc too large. However, additional phonon scattering may
arise in thin disordered layers in either the superconduc-
tor or substrate close to the interface. A high thermal
resistance between the superconductor and (low heat
conducting) substrate would establish a 'thermal mirror'
104. to the superconductor. This would reduce the cooled
surface of the conductor to effectively about half its
value and, as a consequence, would have the same
impact as an uncertainty factor of 2 with regard to heat
E 10"4 transfer from the superconductor to a coolant!

Figure 4 (a) Voltage U v e r s u s current I measured with a pulse

) 104. technique applied to a CVD thin film YBaCuO superconductor
sample prepared on SiTi03 (Reference 6; for details o f the
measurement see text). The sample was held under liquid
nitrogen. The solid curve is a second order polynomial least
O • •
squares fit: U = U ( I - Ic(77 K)) to the data. (b) Schematic graph
104` e
O
of specific resistance Pn v e r s u s temperature T with linear approxi-
mations used in regions a and b. Experimental value P n * =
3.57 x 10 -7 ~ m (taken at T = 98 K). (c) Ohmic resistance
10-7 pn(300 K) v e r s u s critical current density J c for a variety of high Tc
superconductors. Data were taken from measurements from this
10 101 i0 = 1'08 i0' i0 s 10e i07-- w o r k and from the literature. The solid curve is a least squares fit
C Jc A/cm= to the data

272 Cryogenics 1992 Vol 32, No 3

 Design of a high temperature superconductor: A. Abeln et al.

Heat transfer to the c o o l a n t OCoolingis described in the section of the conductor element was subject to this tran-
program with data for boiling nitrogen under sition to normal resistivity. Let us take Pn* = 3.57 X
atmospheric pressure 9,t°, or with the well known 10 -7 ~ m, Jc = 109 A m -2 and, for example,
Dittus-Boelter equation for forced convection, assum- Ar = 0.1 ms. If the normal (transport) current equals
ing supercritical conditions (see below). Values of the critical current, the heat generated amounts to
a(AT) for boiling heat transfer with nitrogen given in the = 104 W h m -3 or 1.6 W h kg -t. This amount is rather
older literature (e.g. Reference 1 l) should no longer be large (about 1000 times what can be expected in the case
used (the magnitude of or(AT) has only slightly changed of a simple phase transition from the normal to the
but there are significant corrections to AT: see superconducting state; it is of the same order of
References 9 and 10). magnitude as applied in Reference 4 for a thermal distur-
Measurements of the heat transfer coefficient or(AT) bance).
between a heated surface to boiling nitrogen have been The program calculates temperatures T(N, z) for all
reported in the literature only for copper or stainless r > T'. As initial values, the program assumes T(N,
steel surfaces. Values of or(AT) using ceramics, r = 0) = 77 K for all elements, N, where 1 ___N _< K and
especially ceramic superconducting surfaces, are not for the substrate in the case of pool boiling under
known. It may be expected that the or(AT) values of the atmospheric pressure (for cases of supercritical
high temperature superconductors would be about two operating conditions, all initial values of T were set to
or three times lower than values measured on copper. 67 K). The smallest time steps were 0.1 /zs, with ele-
Surface roughness as well as degassing and thus ment lengths of 10 or 100 mm (in view of the strong
wetting behaviour effects of LN 2 with ceramics have contributions from the source and cooling functions, the
not been investigated. Another difficult problem with usual criteria for numerical stability that relate
a(AT) is connected with transient heat transfer geometrical length to the length of the time steps in con-
problems. It is not known, at least for liquid nitrogen, duction heat transfer problems are too weak and cannot
how long it takes to produce a N 2 bubble or to establish be used in the cases considered here).
a vapour film on the heated surface; in other words, with
what time-scale is it permissible to use or(AT), conven-
tionally measured under stationary conditions, in a tran- Thermal problems when applying the
sient analysis? (The situation seems to be better with cryogenic stability criterion
helium.)
Taking into consideration for a moment the above Let us start with some estimates based on the cryogenic
mentioned stability criteria, e.g. cryogenic stability, it stability criterion
can be immediately recognized that the value of or(AT)
may have equal importance in the estimation of conduc- p n * J g < (9/(AT) X ( T c - Tcoolant) X P/A (2)
tor geometry as the squared critical current density.
This influence of or(AT) must be shown explicitly in the where P denotes the cooled perimeter of the conductor
computer simulations, which will be demonstrated by and A its cross-section. This equation will be solved for
the results reported in a later section. P/A or (if width b is given) for conductor thickness d,
Besides boiling heat transfer, forced convection heat for a given ratio J/Jc of transport to critical current den-
transfer has also been considered (in sample calcula- sity (see below). Using these assumptions, Equation (2)
tions). In this case, supercritical nitrogen was con- describes the maximum allowed thickness of the conduc-
sidered as a coolant, and the heat transfer coefficient was tor for providing a ratio of cooled perimeter (or cooled
calculated by application of the Dittus-Boelter equa- surface) to conductor cross-section (or volume) large
tion. It should be noted that the critical pressure of enough to deliver the generated heat completely to the
nitrogen is = 3 3 . 5 a t m t. If it is not possible, coolant.
under extreme operating conditions, to cool the conduc- Comparison of predictions based on Equation (2) with
tor sufficiently simply by boiling heat transfer at the results obtained in the computer simulations are then
atmospheric pressure, increased pressures up to made to see the limitations of the simple Equation (2).
15 atm or even supercritical conditions cannot be Note that the application of this equation implies a per-
excluded. Such situations could impose severe addi- manent thermal disturbance of power Pn* Jc2 per unit
tional requirements with respect to mechanical strength volume. When using Equation (2) with this worst case
of the cryostat and lifetime of the superconducting thin assumption, a conservative approach to stability condi-
films. tions is usually produced. As will be show below, this
In the following it is assumed that a thermal distur- approach can fail if the heat transfer coefficient o~(AT) is
bance (caused by conductor movement or by other heat not used properly, i.e. if its temperature dependence is
generating processes) initiates quenching of the super- neglected. In Equation (2) we use the input parameters
conductor in a small volume at a time r'. This process p,* = 3.57 x 10 -7 ~ m, Jc(77 K) = 10 9 A m -2 and
will be active for a time interval AT. Heat generation ~(AT) = 104 W m -2 K -I.
resulting from the quench can be calculated, for exam- To account for a reduced heat transfer from ceramic
ple, by setting the electrical resistivity of element N (e.g. (superconducting) surfaces to LN2, these calculations
N = K/2) at the time r' to the value of p,* (the filled have been made with a reduced heat transfer coefficient
circle in Figure 4b) during the interval AT. In the simula- or(AT) that was set to one-third of the corresponding
tions, as with the application of the simple cryogenic values measured with copper. If we use the full
stability criterion, it was assumed that the full cross- o~(AT) value of copper, a considerably higher
transport current density J/Jc would be expected (this
:1 a t m = 101325 N m -2 demonstrates the impact of an uncertainty in or(AT) on

Cryogenics 1992 Vol 32, No 3 273

Design of a high temperature superconductor: A. Abeln et al.

= (AT) critical temperature T~ for the transport current density

J: T' = Tc = Tc(J). In this picture, the 'usual' critical
temperature Tc is written simply as Tc = T c ( J = 0 ) ,
which means that T~ should be measured with probe
(Xmax currents as small as possible. Accordingly, T~ is a func-
tion of J. This function is already defined by Equation
(1)

T~(J) = T~(J = 0) - (T~(J = 0) - 77 K) × J/Jc(77 K)

(3)

When inserting J and Tc(J) into Equation (2), instead of

the symbols Jc and To, we have

pn J 2 = p n * J c ( T ' ) 2 < O/(AT) × (Tc(J) - - Tcoolant)X P/A

Figure 5 Schematic diagram illustrating heat transfer coefficient
c~(AT) for (natural) convective and boiling heat transfer. AT (4)
denotes the temperature difference between the heater (e.g.
superconductor) and cooling liquid (e.g. nitrogen) T~(J = 0) = 94 K and
Let us assume that
T~oolant= 77 K; accordingly, the maximum possible
stability analysis). In particular, the value of interval of AT is
O/(zaT) = 104 W m -2 K -~ applied here corresponds to
one-third of the maximum value O/max(AT)of the O/(AT) AT = (T~(J) - Tcoolant)~ 94 K - 77 K = 17 K
curve observed in the (schematic) boiling heat transfer
diagram of nitrogen (Figure 5). Thus the value of O/(AT) is by no means fixed to
For application of Equation (2), we have adopted the O/max(AT). Instead, a variation of J between 0 < J _< J~
convention O/(AT)~---O/max(AT) from the literature. (77 K) leads to a variation of about 1 _ AT < 17 K,
However, although preferentially applied in the existing which in turn induces a strong variation of O/(AT)
work on superconductors, it should be clearly stated that ( = 2 0 0 0 - 4 0 000 W m -2 K -], the interval between
this is an overestimate of real heat transfer conditions natural convection and maximum boiling heat transfer
because only in a very few instances (if at all) will the with LN2). To make the consequences more obvious,
difference (T~ - Tcoolan t) in Equation (2) equal the value we will consider two cases:
A T = AT(O/max) where O/max is observed (AT(O/max) is
= 5 K for boiling heat transfer to liquid nitrogen under 1 Let J/J~(77 K) be large. In this case, To(J) is close
atmospheric pressure9'~°). to 77 K (Figure 6), because J is close to J~(77 K)
This uncertainty with respect to the use of O/(AT), (remember: heat production occurs with this
apparent in numerous applications of the cryogenic scenario only if J _> Jc(T)). Accordingly,
stability criterion, can be made still more obvious. If we A T = T c ( J ) - 77 K is small, which implies that
consider a particular value of transport current density O/(AT) is also small, e.g. the lower limit
J we can look for the temperature T' where this (2000 W m -2 K-~) of the O/(AT) interval given
transport current density equals the critical current den- above. In other words, cooling plays a weak part in
sity (Figure 6). This temperature can be considered the the cryogenic stability criterion if the transport cur-
rent is large.
2 If J is small, T~(J) is close to 9 4 K and
Jc(T) AT > 5 K -- ATmax- In this case, the AT values are
located in the film boiling region. However, it is not
at all obvious that stable film boiling really can be
established, and it is not known how long it would
take to establish a stable vapour film on the super-
conductor.

~J=Jc(T') In summary, application of O/(AT) taken as a constant

J and at AT(o/max) in Equations (2) or (4) leads to, except
for certain specific transport currents J, an overestimate
of cooling conditions. Consequently, J/Jc values
estimated from Equation (4) (as, for example, in
Reference 12) are too large. Nevertheless, we provi-
sionally followed these conventions in order to see how
~T they compare with computer simulations of the transient
77K T'-Tc(J) T c (J=O) behaviour of superconductor temperatures.
Still bearing in mind the cryogenic stability picture,
Figure 6 Schematic diagram showing temperature dependence
of critical current density Jc of a superconductor v e r s u s
and with the conventions explained above, we then
temperature T (the diagram is used for interpretation of the have, from Equation (4), d < - 2 b / ( 2 - a b ) using
cryogenic stability criterion; see text for details) a = pn*Jc(T')E/(o/max(AT) X (Tc(J) - 77 K)), P = 2(b +

274 Cryogenics 1992 Vol 32, No 3

 Design of a high temperature superconductor: A. Abeln et al.

d IJm) Results: c o m p a r i s o n with computer

simulations
So far we have considered the application of the simple

1
cryogenic stability criterion. The stability criterion will
now be approached using computer simulations.
Figure 8 shows calculated transient temperatures T(N,
r) (where r = time) for element N = K/2 (K = 100, ele-
15 ment length 100 mm) for different cooling conditions
and for a (strong) thermal disturbance of length
AT = 1 ms. A constant thickness d = 0.8/zm has been
assumed in the calculations.
Some of the curves given in this figure (and others
calculated with a different conductor thickness) show a
10
maximum temperature Tmaxreached by element N under
given operating conditions at the end of the heating
interval Ar (see the solid and dashed curves indicated by
J/Jc(77 K) = 0.44 and 0.45, respectively, in Figure 8).
When the disturbance is switched off, this does not
5- imply automatically that the resistance of the conductor
is again zero: the temperature of the conductor may have
become so high that it exceeds To. The conductor
temperature will then behave according to either case (a)
or (b). Case (a): TN(r)< TN(r'+Ar) for all
r _ r' + Ar because the cooling power is sufficiently
0
0 6.2 o,4 O.s J/Jc('77 high to bring the temperature of the conductor down even
if TN(r' + At) >- To, in the presence of the transport
Figure 7 Maximum conductor thickness d versus ratio J/Jc of current. Case (b): TN(r) continues to increase because
transport to critical current density (at 77 K) in a high Tc thin film
superconductor ( ) calculated by application of the cryogenic
heat production by the transport current flowing through
stability criterion, e , Results from the transient analysis the resistive conductor element is too high. In most
cases, the conductor temperature will then increase to
very high values. However, it appears that stationary
states could also be possible (see dashed curve in Figure
d) and A = bd. In Equation (4) it has been assumed that 8 indicated by J/J~(77 K) = 0.5). In the following, we
the superconductor is wetted by the coolant over its full will consider only those curves that follow case (a). The
perimeter P. The results from the cryogenic stability corresponding values of Tmaxare given in Figure 9.
criterion have been plotted in Figure 7 (solid curve). The point of intersection of the Tmax(J/J~) curves in
Obviously, the required conductor thickness depends Figure 9 with the horizontal line T = Tc = 94 K may be
strongly on J/Jc, and vice versa. interpreted as the maximum permissible ratio

100 ! T(K)

~. 0.5
/
/
/
90
J / J c (77K)

80 0.44
/
I \
I \
I ,, 0.45
70 ¸

60

T(ms)
Figure 8 Dependence on time ~"of transient temperatures T(N, r) for conductor element
N (see Figures 2a and 2b) for different cooling conditions: - - , pool boiling;
forced convection under supercritical conditions of nitrogen. The curves are calculated
for different ratios J/Jc of transport to critical current density (at 77 K). Length of ther-
real disturbance Ar is 1 ms

Cryogenics 1992 Vol 32, No 3 275

Design of a high temperature superconductor: A. Abeln e t al.

Tmax(K) LTin= (K)

d (IJm)
110
10 2
d (gm) 100
10 2 0.8 8 Tc'94K
100
Tc-94 K

80

i"
90

80 _ ~ (x c . ( d - 0 . 8 wn)
6© ,~.r
o 0.= o.4 O.e o.8
J/Jc(77K)
o 0.= 0.4 o.e J/Jc(WK=)
Figure 9 Maximum temperatures Trna× of conductor element N Figure 10 Same calculations as in Figure 9 but with forced con-
( ) calculated for different conductor thickness d, as a func- vection, using Reynolds number Re = 1.2 x 105 , for super-
tion of permissible ratio J/Jc of transport to critical current den- critical nitrogen
sity (at 77 K). Cooling conditions were pool boiling with liquid
nitrogen (1 atm) and length of thermal disturbance zlr is 1 ms.
Heat transfer coefficients were taken as one-third of those
measured in Reference 9 with copper surfaces under stationary TIn=x(K)
conditions. - - - , Full value of (stationary) heat transfer coeffi-
cient, measured with copper, for conductor thickness d = 0.8 #m AT (ms)
1 1
100
T O•94K / / 0.1
J/Jc(77 K). This is a conservative approach because in
a few cases the calculations also showed that
temperatures Tmaxexceeding Tc by even 5 K could be
tolerated, depending on the instantaneous heat produc-
tion and cooling conditions. The dashed curve calculated 80
for forced convection in Figure 8 (with the label
J/Jc = 0.5) could serve as an upper limit for the
temperature excursion over Tc since it apparently
represents a stationary state. Accordingly, the calcula-
tions showed that it was also possible to remove ohmic 60
heat for cases of small excesses of superconductor o O= 0.4 oe 08 --
temperature over Tc and if highly efficient cooling is J/Jc(77K)
supplied. If allowing only Tmax< T~, on the other hand,
Figure 11 Same calculations as in Figure 70 but for different
the simulated cooling by nitrogen will always be suffi- lengths of thermal disturbance A~ ( ) and for a 'dead time
cient to bring the conductor temperature back to 77 K interval' A~' of 2 ms ( - - - ) . All curves calculated for
after the disturbance is switched off and within time d = 0.8 #m
intervals the length of which depend on heat capacity
and cooling conditions.
If in a sample calculation the full value O~c, is used ductor thickness d = 0.8 #m is assumed). It is seen
instead of one-third of O~Cu(see above discussion), it is immediately that the permissible J/Jc ratios shift to
seen from Figure 9 that no limitation would be imposed smaller values if z17- is increased, because of the
upon J/Jc, at least within the limits 0 < J/Jc <- 0.5 increased magnitude of the thermal disturbance.
(d = 0.8/xm). This finding illustrates the consequences Because of this dependence, an appropriate definition of
that may result from uncertainties in the magnitude of the thermal disturbance, e.g. by length At, by the
o~(zXT). amount of energy deposited, by setting the conductor
A similar calculation for forced convection, assuming temperature to a certain value T' _> Tc at r = r', by
cooling with supercritical nitrogen, is given in Figure local fluctuations in Jc or its temperature dependence or
10. The curves are calculated for a Reynolds number of by local fluctuations in heat transfer, is mandatory for
1.2 × 105, which implies a flow velocity of the order stability analysis.
o f l m s -;. We have also studied the effect of a possible 'dead
Figures 8 - 1 0 were calculated using a constant value time' At' during which o~(zlT) = 0, which may occur at
of z17- = 1 ms. The impact of variable length of the inter- the beginning of a thermal excursion of the superconduc-
val =at on the maximum conductor temperature and on tor because the onset of natural convection in the coolant
the maximum J/J~ is shown in Figure 11 (constant con- or the formation of bubbles (which initiate boiling heat

276 " Cryogenics t992 Vol 32, No 3

 Design of a high temperature superconductor: A. Abeln et al.

transfer) require a measurable time step. In the case of

forced convection, as studied in Figure 11, the dead time OEI
may correspond to a possible short term fluctuation of
fluid flow close to the conductor surface. It is seen from
this figure that only in cases of very small thermal distur-
bances will a dead time of the cooling function have no
influence on J/Jc.
Maximum allowable J/J~ values from Figure 9 (when
accepting only Tmax< T~) for a given conductor
thickness d have been inserted into Figure 7 (filled °~o~
circles) for comparison with the predictions from ~ e - - .
cryogenic stability. If we had also accepted those
Tmax values that exceeded T~ by only a little
( < 1 K in case of pool boiling) the J/Jc values
would be larger by 5% at the most, which can be
estimated from the slope dTmJd(J/Jc) at T = T~ of the
curves in Figure 9. The comparison is made for a ther- ©
mal disturbance length of 1 ms. It is immediately seen 0
i i

104 b / d>
that the cryogenic stability criterion, if used with too
high a heat transfer coefficient, is necessarily weaker: it Figure 12 Stability function Cmax versus aspect ratio b/d of
width b to thickness d of the conductor, for a thermal disturbance
permits higher values of J/J~ compared with those of
of length Ar = 1 ms and with temperature (i.e. time) dependent
the transient analysis, although application of Equations Blot numbers (see text for details). The curve is calculated using
(2) or (4) implies a permanent (Az -- co) thermal distur- the filled circles in Figure 7
bance. Even with this worst case assumption, the
application of a(ATm,x) in the cryogenic stability
criterion leads to overestimates of the permissible pn*J 2 For a rough comparison of our results and those of
values. A correct formulation of the heat balance thus Reference 4 (using the curves calculated for isotropic
has to take into account exactly those ~(AT) that really thermal conductivity in this reference), in Figure 12 the
occur, i.e. we need the AT = T(r) - 77 K temperature stability function tI)ma x ~- 1 - J / J c ( 7 7 K) has been plot-
history of the superconductor, which can be obtained ted versus the aspect ratio b/d using the data (filled
only in a transient analysis. circles) given in Figure 7 of this work. These ~max
Since the thickness of CVD thin film superconductors values are larger (i.e. the permissible J/Jc ratios are
applied for devices in energy technology will probably smaller) than those reported in Reference 4. This can be
be below 2/zm, the simple cryogenic stability criterion, explained by the fact that Reference 4 applies a rather
used in the manner it frequently is, could lead to unac- large (and constant) Blot number (Bi = 0.2) in the cor-
ceptable uncertainties in J/J~; the differences (horizon- responding plot of ~max versus aspect ratio. In our case,
tal distances) between the filled circles and the solid we would have a range of Blot numbers of
curve in Figure 7 increase with increasing J/Jc. If, on 0.001 ~< Bi <_ 0.01, using Bi = t~(AT)d/(2k), where k
the other hand, a J/Jc value is prescribed, e.g. from is the isotropic thermal conductivity (-~ 4 W m -l K -l)
safety considerations, application of the same criterion and 0.8 <_ d < 10/~m, if we kept or(AT) and k constant
will lead to an overestimate in permissible film and if we used o~(A/) = C~max.In reality, both quantities
thickness. are functions of temperature and thus of time, since
T = T(r). Accordingly, the actual Blot numbers in our
calculations are even smaller, because of the variation,
Comparison with other transient analyses with AT(z), of a(AT) ___ O(max. Since these Blot numbers
indicate a considerably smaller ratio of heat transfer to
Calculations in the previous section have been made the coolant and thermal conductivity than estimated in
assuming isotropic transport properties (current, heat). Reference 4, this explains the larger t~ma x calculated
It is to be expected that the results given in the previous here (compare Figure 6 of Reference 4). In addition, the
figures will need to be corrected if anisotropic transport thermal disturbances used in our case are approximately
properties of the new ceramic superconductors are taken a factor of two larger if using Ar = 1 ms, which also
into account. While a simple criterion like the cryogenic shifts the ~maxto larger values, i.e. J/Jc to smaller per-
stability model cannot account for anisotropic transport missible values.
properties, these properties can be taken into account in
computer simulations.
Conclusions
A first step in this direction has been made in
References 4 and 5. Assuming a thermal disturbance, Stability analysis for ceramic superconductors cannot be
these references calculate T(x, y, r) or T(x, y, z, T) in an performed without appropriate models for temperature
anisotropically heat conducting superconductor. Cor- history and transient heat transfer (and anisotropic
responding to this temperature distribution, there is a transport properties of heat and current). Also the source
distribution of Jc[T(x, y, r)]. Note that the ratio function that describes the generation of heat and the
Jc[T(x, y, r)]/Jc(77 K) is a function of time. However, magnitude of a thermal disturbance has to be thoroughly
both references have applied constant (and uniform) defined. As a consequence, conductor design will not be
values of offAT), which is not permissible, as has been successful if these models and the required parameter
demonstrated above. inputs are missing or incomplete. Since materials

Cryogenics 1992 Vol 32. No 3 277

Design of a high temperature superconductor: A. Abeln et al.

development has to rely on conductor design, these 2 Iwasa, Y. Design and operational issues for 77 K superconducting
models are of great importance for high Tc supercon- magnets IEEE Trans Magn (1988) MAG-24 1211 - 1214
3 Ogasawara, T. Conductor design issues for oxide superconductors.
ductor projects. Experimental work is needed to deter- Part l: criteria of magnetic stability Cryogenics (1989) 29 3 - 9
mine the heat transfer coefficients from ceramic 4 Flik, M.I. and Tien, C.L. Intrinsic thermal stability of anisotropic
superconductors to liquid nitrogen. thin-film superconductors, paper presented at ASME Winter
A transient analysis has been performed in this work Meeting, Chicago, Illinois, USA 0988)
to quantify the maximum allowable transport current in 5 Chert, R.C. and Chu, H.S. Study on the intrinsic thermal stability
of anisotropic thin film superconductors with a line heat source
a band-like conductor, still with isotropic transport pro-
Cryogenics (1991) 31 749-755
perties, and with no magnetic field included. The 6 Schmaderer, F., Huber, R., Oetzmann, H. and Wahl, G. High
analysis shows that application of the simple cryogenic T¢ YBa2Cu307-~ prepared by chemical vapour deposition Appl Sur-
stability criterion is not sufficient for conductor design. face Sci (1990) 46 5 3 - 6 0
A more extensive comparison between computer simula- 7 Gmelin, E. Thermal properties of high temperature superconductors,
in: High Temperature Superconductors (Ed Narlikar, A.V.) Nova
tions and the other classical stability criteria, also taking Science Publishers, New York, USA (1989)
into account the anisotropic transport properties of a 8 Swartz, E.T. and Pobl, R.O. Thermal resistance at interfaces Appl
superconductor, is in preparation. Phys Lett (1987) 51 2200-2202
9 Bier, K. and Lambert, M. Comparison of pool boiling heat transfer
coefficients for different pure liquids, paper presented at Advances in
Acknowledgement Pool Boiling Heat Transfer: Eurotherm 8, Paderborn, Germany, May
(1989)
H. Reiss wishes to express his gratitude to Dr M. I. Flik l0 Riithlein, H. Autbau und Erprobung einer Apparatur zur Messung
for friendly discussions. des W/irmeiibergangs von einem horizontalen Rohr an tiefsiedende
Flfissigkeiten PhD Thesis University of Karlsruhe, Germany (1984)
11 Fastowski, W.G., Petrowski, J.W. and Rowinski, A.E.
References Kryotechnik Akademie Verlag, Berlin, Germany (1970)
12 Wipf, S. Stability of possible high T¢ oxide superconductors as cur-
I Wilson, M.N. Superconducting Magnets, Oxford Science rent carriers Proc ICEC 12 Butterworths, Guildford, UK 0988)
Publishers, Oxford, UK (1983) 931 - 935

278 Cryogenics 1992 Vol 32, No 3