Applications of High-Temperature Superconductivity - ACS Publications

Chapter 27

Applications of High-Temperature Superconductivity A. P. Malozemoff1, W. J. Gallagher1, and R. E. Schwall23, 1 Thomas J. Watson Research Center, IBM, Yorktown Heights, NY 10598-0218 Francis Bitter National Magnet Laboratory, Massachusetts Institute of Technology, NW 17-060, Cambridge, MA 02139

Downloaded by UNIV OF CALIFORNIA SANTA BARBARA on June 9, 2016 | http://pubs.acs.org Publication Date: August 28, 1987 | doi: 10.1021/bk-1987-0351.ch027

2

The new high temperature superconductors open up possibil ities for applications in magnets, power transmission, computer interconnections, Josephson devices and instrumentation, among many others. The success of these applications hinges on many interlocking factors, including critical current density, critical fields, allowable processing temperatures, mechanical properties and chemical stability. An analysis of some of these factors suggests which applications may be the easiest to realize and which may have the greatest potential. In January 1986, J . G . B e d n o r z a n d K . A . Mueller of the I B M Z u r i c h R e s e a r c h L a b o r a t o r y discovered h i g h temperature superconductivity i n a c e r tain class of c o p p e r oxides £11. O n l y a year later, superconducting transition temperatures i n this class o f materials had been pushed to almost 100 Κ (2-3), far b e y o n d the previous record o f 23 Κ that h a d remained u n b r o k e n since 1973, and well above the benchmark temperature 77 Κ o f boiling liquid n i t r o gen. In addition to the scientific excitement o f this extraordinary development, the possibilities for practical application of superconductivity clearly need to be re-examined, if only because o f the vastly greater ease i n cooling to 77 Κ c o m p a r e d to the liquid helium temperature range ( ^ 4 K ) used i n all previous applications o f superconductivity (4). While the possibility o f achieving yet higher temperatures i n n e w c o m positions continues to be actively pursued, we focus i n what follows o n the most studied group of superconductors with transition temperatures

i n the

range f r o m 9 9 - 1 0 0 K , namely the Y B a C u 0 _ 5 group with the " 1 2 3 " layered 2

perovskite superstructure

3

7

£51, w h i c h we henceforth refer

to as Y B a C u O .

Certainly the possibilities for application w o u l d be increased immeasurably if stable r o o m temperature superconductors were found. Previous

applications

o f superconductivity

have

entailed

about $ 2 0

million i n yearly sales o f finished superconducting material (mostly N b T i a n d IBM

3

visiting scientist 0097-6156/87/0351-0280$07.75/0 ©

1987 A m e r i c a n C h e m i c a l Society

Nelson et al.; Chemistry of High-Temperature Superconductors ACS Symposium Series; American Chemical Society: Washington, DC, 1987.

27.

Applications

M A L O Z E M O F F ET AL.

of High-Temperature

N b S n wire a n d tape) used primarily i n magnets. 3

total has gone into shielding or S Q U E D s .

281

Superconductivity

A quite small fraction o f the

B y far the dominant ( 7 0 - 8 0 % )

magnet application has recently been in medical magnetic resonance imaging, where the roughly quarter million dollar cost of the superconducting magnet and the approximately 4 0 0 units sold i n 1986 imply a market of order $ 1 0 0 million yearly.

M o s t of the remainder involves high energy physics or labora

tory magnets. While this is a substantial market for superconductivity, the new


high-temperature superconductors have n o w raised vastly greater hopes. In this paper we review some possible applications for the new high t e m perature superconductors.

T h e scope is huge — almost as b r o a d as the entire

field of electricity and magnetism ~ so of necessity the treatment here of any single application will be brief and superficial. F u r t h e r m o r e , even though r e search has progressed at an unprecedented pace, the amount k n o w n about the technical properties of the new superconductors is still limited. T h u s there is a great unpredictability hovering around our assumptions.

O u r conclusions

should be regarded i n this light. T h e applications we discuss here, i n magnets, p o w e r transmission, c o m puter interconnections, J o s e p h s o n devices and instrumentation, have almost all

been

studied before,

superconductivity (41.

during

two

decades

of

active w o r k

i n applied

T h o u g h t of simply as a standard superconductor with

a higher transition temperature, Y B a C u O does not by itself imply new kinds of applications, even though it may improve the commercial prospects for a p plications previously burdened b y overhead costs associated with helium r e frigeration. M o r e n o v e l applications may well emerge i n the future. O u r analysis below reveals m u c h promise i n a variety of applications, but also many challenges.

Clearly a long-term view will be essential to exploit the

full potential of this new technology. Material Parameters for Applications T h e prospects for various applications depend, of course, o n the materials properties of the superconductor.

M a n y properties are important, including

mechanical properties, interfacial interactions and contacts, high frequency loss, and so forth. H e r e we emphasize as an example the critical current d e n sity, w h i c h is one of the most vital properties for most applications. First we review some typical materials parameters obtained f r o m meas urements o n r a n d o m l y oriented ceramics (6-7). (5)

Since the Y B a C u O structure

and electronic properties 181 are highly anisotropic, the orientationally-

averaged values obtained f r o m studies of ceramics are only an initial indication until more complete experimental results o n single crystals and oriented films and ceramics become available.

F o r material with a resistivity just above the

transition of 4 0 0 μ Ω α η , a H a l l carrier density of 4 x l 0

2 1

c m ~ , and d H

c 2

/dT

of 2 T / K (6-7), one deduces a B C S coherence length £ ( 0 ) of 1.4 n m , a L o n d o n penetration depth λ ( 0 )

of

200 n m , a mean

thermodynamic critical field H ( 0 ) c

free

path (

of 1 Τ (10000 O e )

of

1.2

nm, a

a n d an upper critical


282

CHEMISTRY OF HIGH-TEMPERATURE SUPERCONDUCTORS

field H

c 2

( 0 ) o f 120 T . T h e large ratio of λ to ξ implies that Y B a C u O is a T y p e

Π superconductor, i.e, magnetic flux c a n penetrate the bulk material i n a certain range o f field strength. T h e largest low-field 77 Κ critical current density (6) measured so far i n randomly-oriented ceramic material is only 1000 A / c m , a disappointingly small value w h i c h w o u l d exclude all but a very few o f the possible applications. More

recent results

o n preferentially

oriented

epitaxial films

on S r T i 0

3

substrates (9-11) show values 100 times higher at 77 K , namely more than 10 A/cm .


5

Measurements o n single crystals {81 as well as the epitaxial films

2

show currents as large as 3 x l 0

6

A/cm

2

at 4 . 2 K , a n d the crystal measurements

show that the critical current density has a large temperature a n d field d e pendent anistropy.

The 10 A / c m 5

2

current level at 77 Κ w o u l d be adequate

for many applications, although this value was only achieved i n zero field. T h e current density improvement is apparently related either to orienting the Y B a C u O material so that current flows only along favorable directions or to

the elimination

(Josephson-coupled)

of

grain

boundaries

(12),

obstacles to current flow.

which

may form

weak

Unfortunately single crystals

and single-crystal substrates are not appropriate for many applications.

Ways

must be found, perhaps using oriented ceramics, to achieve higher current density

in a

3xl0 A/cm 4

2

poly crystalline

environment.

A

recent

report

(13)

of

i n films o n poly crystalline substrates indicates rapid progress i n

this direction. Insight into the temperature dependence o f the zero-field current density c a n be obtained b y considering a theoretical upper limit to current density of a superconductor, namely, the "depairing current density", where the kinetic energy

of the superconducting

electrons

equals

the condensation

energy

Η / 8 π . A t l o w temperatures and zero applied field, one finds (14) 2

J where J

d

d

=

10H /4fiA

(1)

c

is the critical current density i n A m p s / c m , H

c

is the thermodynamic

critical field i n O e a n d λ is the L o n d o n penetration depth i n c m . parameters given above, one calculates 3 x 1 0 deed £71.

8

A/cm

2

U s i n g the

, a very large value i n

Rather similar values are obtained for the conventional high field

superconductors because their lower transition temperatures a n d hence lower thermodynamic critical fields are compensated b y their higher electron density w h i c h lowers the penetration depth. H o w e v e r , at higher temperatures J is reduced. N e a r T , a n estimate can d

c

be made with the G i n z b u r g - L a n d a u theory (14): J

d

=

10H (1 -

ί) /3/βπλ 3/2

C

(2)

where t is the reduced temperature T / T . T h e r e is a factor 1.84 reduction i n c

the prefactor a n d a n all-important temperature dependent factor (1 — t) w h i c h can cause a drastic decrease i n J

d

,

near T . F o r example, the depairing

critical current density o f Y B a C u O with T

c

c

o f 92 Κ is reduced to 1.2x10


27.

283

Applications of High-Temperature Superconductivity

MALOZEMOFF ET AL. 2

A/cm

at 77 K , more than an order of magnitude lower than at 4.2 K .

Thus

Y B a C u O at 77 Κ has a fundamental disadvantage in zero-field critical current density c o m p a r e d to, say, N b S n at 4.2 K . 3

T h e same effect impacts the hope

for practical r o o m temperature superconductors:

T o maintain reasonable crit

ical current density at 300 Κ or 27 C , the superconducting transition would have to be closer to 4 0 0 Κ or 127 C ! T h e 77 K , zero-field depairing current density estimated above w o u l d still


be adequate for most applications. B u t another effect limits the actual critical current density J to lower values: c

T h i s is the depinning of flux or vortex lines

w h i c h penetrate T y p e Π superconductors like Y B a C u O .

Currents exert forces

o n flux lines, and w h e n they depin above a threshold current, they generate a "flux flow resistance" i n their n o r m a l cores (14).

In high-field magnet materi

als, current densities are maximized by introducing structures that p i n the flux lines, such as dislocation walls and a — T i precipitates i n N b T i (15).

B u t even

i n this well-studied case, the maximum achieved current density is still over an order of magnitude less than the depairing limit. Little is k n o w n about pinning i n Y B a C u O .

B u t the experience

with

N b S n and N b T i suggests that even after extensive effort to optimize pinning, 3

critical currents i n Y B a C u O are likely to remain at least an order of magnitude lower than the depairing limit.

T h i s implies a likely limit of 1 0 A / c m 6

2

for

Y B a C u O at 77 Κ and zero field, a factor of over ten less than what has already been demonstrated i n N b S n at 4.2 Κ and zero field. T h i s is a serious disad 3

vantage for Y B a C u O , as will be seen below.

Clearly even a modest increase

in transition temperature w o u l d improve the limit substantially; T

of 120 Κ

c

w o u l d give a factor of 3 improvement. Critical

current

density

also

depends

on

monotonically to zero at the upper critical field H slope of H applied

c

2

magnetic c

2

field,

decreasing

(16-17). In Y B a C u O the

with temperature is unusually large, of order 2 T / K w h e n field is

parallel

to

the

predominant

c o n d u c t i o n planes

of

the

structure

(18-19). T h i s implies record values (up to 200 Τ has been estimated) for the upper critical field at low temperatures and opens up the possibility of very high field magnets. T h e ability to achieve ultra-high fields (over 2 0 T ) , however, will certainly be limited in practice by the strain induced f r o m stresses connected with the containment of high magnetic fields. It is customary to have conductor tensile strains of several tenths of a percent in N b T i magnets (at fields < 1 0 T ) , for example, and it is unlikely such strains c a n be supported by the brittle ceramic Y B a C u O (20).

F o r example, consider a solenoid of 20 c m inner diameter o p

erating at 20 T .

W e assume the winding to contain 5 0 % steel by area w h i c h

carries the h o o p stress.

If the overall current density is 3 x l 0 A / c m 4

2

(which

implies over 6 x 1 0 i n the superconductor), then the h o o p stress is 6 0 0 M P a (87 4

ksi) overall and 1200 M P a (174 ksi) in the steel. G P a or 3 x 1 0

7

A n elastic modulus of 207

psi for steel yields strains of almost 0 . 6 % i n the winding.

A t fields well below 2 0 T , there are also serious obstacles to overcome. F o r applications, it is the ability to carry substantial currents i n high fields that


284


is important, not merely the value of H

c

2

determined by the zero of resistance

or, more optimisticly, by the breakaway f r o m n o r m a l resistance

behavior

(21-22). H i g h field values for zero resistance have only been observed so far i n single crystals ( 1 8 - 1 9 ) . Results so far achieved with potentially more useful ceramic wire configurations have been disappointing (12,23-24). T h e resistive transitions as a function of field are b r o a d and show a " f o o t " near zero resist ance, w h i c h causes superconductivity to be suppressed very rapidly with field (22).

Presumably this effect arises either f r o m grain boundaries or from strong


anisotropy, and again techniques for orienting ceramics may be important. Progress i n this direction could help to o p e n up applications for high field magnets operating at 4.2 K , where the critical current density of Y B a C u O i n high fields could be m u c h higher than for N b S n because of the high intrinsic 3

values of H Ref. 8.)

c 2

. ( Y B a C u O current densities of 1 . 7 x l 0

at 77 K . T a k i n g d H an H

c

2

6

were measured at 4 Τ i n

E v e n more interesting for applications w o u l d be magnets operating c 2

/ d T ^ 2 T / K for a 92 Κ superconductor, one can estimate

of 30 T , w h i c h is still larger than the 4.2 Κ value of about 20 Τ for

N b S n and 10 Τ for N b T i . 3

characterizes d H

c 2

plane (19), then H

O n the other h a n d , if one uses 0.4 T / K w h i c h

/ d T for field perpendicular to the Y B a C u O conduction c

2

at 77 Κ is only 6 T .

In this case applications are only

likely i n a low field range where the refrigeration cost advantage predominates (see below). C l e a r l y , higher transition temperatures or larger d H

c 2

/ d T would

increase the leverage of Y B a C u O vis-a-vis N b S n . 3

Superconducting M a g n e t Applications T h e impact of the critical materials parameters, especially the critical current density and the allowable operating strain, is seen quite clearly i n magnet a p plications.

Superconducting magnets are the dominant present commercial

application of superconductivity a n d underlie many of the potential but pres ently unrealized applications (15).

T h e existing commercial applications i n

clude: 1. Medical

Magnetic

Resonance Imaging

(MRI):

A s noted above, M R I is c u r

rently the dominant market for superconductive materials and devices

(25),

although the first superconductive M R I magnet was only delivered i n 1980. T h i s application does not tax the superconducting properties of existing c o m mercial N b T i wire.

Peak fields are below 2.5 T e s l a and the required critical

current density is rather low. T h e primary concerns are field homogeneity, field stability and device reliability. T h i s has led to very conservative designs with the superconductor operating at only 5 0 % to 7 0 %

of its critical current.

Stresses are rather low and large amounts of c o p p e r are used for stabilization. While there are other techniques for providing fields for M R I such as p e r m a nent magnets or resistive electromagnets, superconductive magnets have c a p tured the bulk of the M R I market by providing higher fields and better field stability w h i c h translates into superior image quality.


27.


Applications

.2. High Energy Physics (HEP):

of High-Temperature

285

Superconductivity

H E P was until recently the largest consumer of

superconducting materials. M o s t development of high field materials and m a g nets was, i n fact, performed i n H E P labs or i n response to H E P needs. T h e d e m a n d is quite cyclical however, peaking w h e n a large machine such as the T e v a t r o n is built and declining to almost zero i n the years between major m a chines. T h e next major superconducting H E P project is likely to be the highly publicized Superconducting SuperCollider ( S S C ) .

Interestingly, the currently

favored magnet design uses N b T i operating at 6 Τ and 2 . 7 5 x 1 0 A / c m , rather 5

than N b S n


3

w h i c h c o u l d be

capable

2

of higher field operation.

In

HEP,

superconductivity has made possible devices such as the T e v a t r o n and large bubble chambers w h i c h w o u l d be prohibitively expensive to operate if built with conventional resistive magnets. 3. High Field Magnets for Scientific Applications:

H i g h field magnets for scien

tific applications have been a small but important industry for over 2 0 years. With

the increasing importance

of

analytical N M R

i n the chemical and

biotechnology industries and the continued w o r k o n g y r o t r o n tubes employing superconductive magnets, there has been a modest resurgence of growth over the past few years. In these applications superconductors are used to generate very high fields i n small volumes making more widely techniques w h i c h w o u l d otherwise be restricted to a few

available large

analytical

laboratories

having multi-megawatt motor generators and high field c o p p e r magnets. 4. Magnetic

Separation:

A l t h o u g h true commercial applications of magnetic

separation are rare, many pilot projects are underway. T h e devices are used to separate minerals and scrap and to purify polluted water and flue gas. T h e role of superconductivity is to increase the field and field gradients available and hence enhance the efficiency of the operations.

A n o t h e r group of magnet applications, those not yet commercialized, is m u c h larger. Primary candidates at the present time include: 1.

Magnetic Confinement Fusion

2.

Rotating M a c h i n e r y (e.g. turbine or h o m o p o l a r generators)

3.

E n e r g y Storage

4.

L e v i t a t e d Trains - M A G L E V

5.

Magnetic Launchers

6.

Magnetohydrodynamics

In most of these applications superconductive technology is not the factor limiting commercialization, but the cost advantages offered by higher temper ature operation cannot be ignored.

T h e degree to w h i c h Y B a C u O

magnets

operating at 77 Κ impact any of these applications depends o n the performance characteristics of the material, the competing technologies (in most cases c o n -



286

ventional N b T i o r N b S n superconductor operating at 4.2 K ) a n d the e c o 3

nomics o f each application. Critical current density considerations. A s mentioned above, perhaps the most crucial materials parameter is the critical current density J . T o illustrate the c

requirements o n J , we consider a simple m o d e l system, the long thin-walled c

solenoid. T h e central field (in Oersteds) is given by


H

=

47rJt/10

where J is the current density i n A m p s / c m

(3) a n d t is the thickness o f the

2

superconducting winding i n c m . M o s t of the applications listed above require fields of 20,000 to 100,000 O e (2 to 10 T ) . F o r reasonable thickness windings this implies a winding c u r rent density well i n excess of 1 0 A m p s / c m . Since the winding area must i n 4

clude

2

stabilizing material (see below) as well as structure a n d insulation

between turns, the rninimum useful current density i n the superconductor itself is likely to be about 5 χ 1 0 A / c m . T h e requirement is likely to exceed 1 0 4

A/cm

2

2

5

i n those applications requiring high fields or high current density to

achieve precise field geometries such as multipole magnets for accelerators, high field magnets for scientific applications, high gradient magnetic separators and compact rotating machinery.

A s stated above, the achievement o f such

current densities i n p r o d u c t i o n quantities o f Y B a C u O material will require considerable additional materials a n d process development. Materials cost per ampere meter. In those applications where the critical c u r rent density a n d strain characteristics make the magnet technically feasible, we must consider the cost.

W e focus here o n the superconducting materials

cost a n d the total refrigeration cost including the cryostat a n d capitalized refrigerant costs. U s i n g typical present commercial prices for 9 9 . 9 % pure components, one finds a raw materials cost of $ 0 . 3 / c m for Y B a C u O . 3

A s s u m i n g the finished

superconductor is a factor of 4 more costly than the raw materials (as is roughly the case for N b S n wire), we c a n make a first very rough cost estimate o f about 3

$ 1 . 2 / c m . B y comparison, for multifilament N b S n wire at $ 1 0 0 per p o u n d the 3

3

equivalent number is $ 2 . 0 / c m . 3

In many applications, the relevant materials cost is the per volume cost divided b y the current density because less o f a higher current density material need be used to achieve the same required current. T h e figure o f merit c a n then be expressed i n units such as dollars per ampere-meter, where meter refers to the length of the wire. If at zero field the critical current o f Y B a C u O at 7 7 Κ remains a factor of 10 lower than that of N b S n at 4.2 K , as discussed i n the previous section, 3

the cost per ampere meter o f N b S n w o u l d be about a factor of 6 below that 3

of Y B a C u O .

A t higher fields the cost per ampere-meter depends o n the field

dependence of J i n each case ( 2 6 - 2 7 ) . In F i g . 1 we compare some speculative



MALOZEMOFF ET AL.


287

12

ω Η w ι

YBaCuO;

cu

7 7 K 6T

FIELD

(TESLA)

C o s t per ampere meter for various superconducting magnet wires: circles - N b S n values for 1 0 lbs. quantity ( 2 6 - 2 7 ) ; crosses - N b T i 3

6

values for present S S C spec (26); x's - 4.2 Κ Y B a C u O value f r o m $1.2/cm

3

cost estimate a n d measured single crystal critical current

densities at 4.2 Κ £81. F o r Y B a C u O at 77 Κ (solid and dashed lines), a current density form J ( l - H / H H

c

2

c 2

) was used with J = 1 0 A / c m

either 30 Τ (single crystal value for H±c

5

2

and

£191) or 6 Τ ( H || c).


288


costs for Y B a C u O with costs for current state-of-the-art commercial super conductors. T h e N b T i data s h o w n is for the material p r o p o s e d i n the reference design study for the Superconducting S u p e r c o l l i d e r . T h i s material has a critical current density at 5 Τ of over twice that of the material used to build the T e v a t r o n . T h e N b S n data is for large quantities (26) of multifilamentary wire 3

produced via the internal tin core process

(27).

F o r Y B a C u O at 4.2 Κ we use critical currents from single crystals (81 and the materials cost estimated above.

T h i s of course assumes that the single


crystal critical currents can be realized i n long lengths of conductor w o u n d i n a magnet geometry and also that the strain limits of the material are not ex ceeded as discussed earlier.

Because of the very high upper critical field of

Y B a C u O at 4.2 K , the critical current of this material dominates at very high fields reducing its cost-per-ampere-meter below that of its conventional c o m petitors. T h i s is an exciting opportunity for the new superconductor, although obviously it is hardly an immediate threat to the conventional materials since the arduous optimization process for Y B a C u O wire fabrication has hardly b e gun and the strain limits are far f r o m clear. A t 77 K , we estimate the Y B a C u O cost per ampere-meter by taking a current density of the f o r m J ( l -

H/H

c 2

) , with J = 10 A / c m

30 Τ or 6 T , as discussed i n the previous section.

2

and H

c

either

2

F i g . 1 shows that 77 Κ

Y B a C u O crosses under N b S n above about 10 Τ if its critical field is suffi 3

ciently high (compare the 30 Τ and 6 Τ curves i n F i g . 1), although this field range is still i n jeopardy because of the mechanical yield strain limit.

A t low

fields, the unfavorable cost-per-ampere-meter must be offset by reduced r e frigeration cost if Y B a C u O is to compete with the conventional superconduc tors.

It should be emphasized that the specific numbers for Y B a C u O given

here are only illustrative of the kinds of crossovers w h i c h are likely to occur; meaningful engineering evaluation must await data o n real Y B a C u O wires.

A

further caveat is that our analysis ignores the cost of fabricating material into a magnet.

C u r r e n t l y N b T i magnets are less costly than N b S n up to about 9 3

Τ partly because ductile N b T i is m u c h easier to use than brittle N b S n . 3

Refrigeration costs. T h e cryostat of a magnet operating at helium temperature contains components operating at three different temperatures. T h e outermost shell is a radiation shield operating at or near 77 Κ w h i c h intercepts the bulk of the heat transferred from r o o m temperature by radiation.

T h i s shield is

normally c o o l e d by a liquid nitrogen reservoir or a single stage refrigerator. T h e next shield operates at about 2 0 Κ and is cooled by c o l d gas f r o m the h e lium boiloff or by a two stage refrigerator.

T h i s intercepts almost all of the r e

maining radiation. T h e final stage is the magnet container w h i c h operates at 4.2 Κ and is c o o l e d by helium boiloff or by a three stage refrigerator.

The

magnet weight and any operating forces must also be transferred to r o o m temperature via support members. These support members are heat sunk to the intermediate heat shields and are i n many cases the dominant heat loads for the 20 Κ and 4.2 Κ thermal stations.


27.

Applications


of High-Temperature

289

Superconductivity

If we substitute an Y B a C u O magnet for that operating at 4.2 K , the shields at 77 Κ and 20 Κ are eliminated and the thermal mass previously at 4.2 Κ n o w operates at 77 K . T h i s will reduce the capital cost of the cryostat by a factor o f about 2.

T h e heat leak to the 77 Κ mass should remain unchanged

so the refrigeration cost for a Y B a C u O magnet should be essentially the same as that for cooling the 77 Κ shield o n an equivalent helium cooled magnet. F o r a commercial M R I magnet the liquid nitrogen c o n s u m p t i o n is about 1 liter/hr a n d the helium consumption is about 0.3 liter/hr.

Assuming

transferred


c r y o g e n costs of $0.25 and $8.00/liter for nitrogen and helium respectively yields a factor of 10 reduction i n overall refrigeration operating costs.

This

reduction w o u l d probably be less for cryostats c o o l e d by small closed cycle refrigerators. Interplay of Materials and Refrigeration Costs. T o estimate the effects of these materials and refrigeration costs o n magnet costs we again consider the long thin solenoid.

T h e refrigeration costs (both the cryostat and the capitalized

costs of the cryogens) are divided into two parts corresponding to the two sources of heat leak into the system. T h e first term, corresponding to radiative heat transfer, is proportional to the cryostat area. solenoid this is 4 7 r R L C , where C r l

F o r a very thin w a r m bore

is the cost of refrigeration for the heat

r l

transmitted through one c m of cryostat surface, R is the m e a n radius, L is the 2

cryostat length, and the area of the inner r o o m temperature bore is assumed roughly equal to that of the outer shell. T h e second term, corresponding to the heat leak via c o n d u c t i o n through the l o a d bearing members, is assumed proportional to the c o n d u c t o r volume a n d c a n be written 2 7 r R L t C , where C r 2

r 2

is the cost of refrigeration for the heat

transmitted through the support members for each c m

3

of conductor and t is

the winding thickness. T h e materials cost is given by 2 i r R L t C , where C m

the cost per c m of superconductor. 3

m

is

C o m b i n i n g these terms, substituting for t

f r o m E q . 3, a n d dividing by the area gives the total cost per area of cryostat as: C/A

=

C

r

l

+ ( C ^ + C )5H/(477j)

(4)

m

While this simple analysis ignores costs such as coil forms and winding labor, it does illustrate some interesting scaling relationships. W e will compare the systems N b S n operating at 4.2 Κ and Y B a C u O operating at 77 Κ i n some 3

limiting cases.

First consider the case of a large light magnet where the heat

leak is primarily via radiation. O n e can then ignore C

r 2

.

In the limit of low

field, the refrigeration costs dominate, and hence Y B a C u O is the lower cost system.

A t higher fields, the materials cost per ampere-meter will be the d e

ciding factor, and, according to F i g . 1, N b S n could be the lower cost system 3

up to about 10 T . A crossover may occur at some intermediate field, depending o n the exact parameter values. T h e other limiting case is that of a compact heavy magnet where the r a diation heat leak is negligible. T h e n the cost ratio is independent of field and depends only o n the ratio of the refrigeration costs and the superconducting


290


materials costs each normalized to the volume o f superconductor. N o t e that i n b o t h cases the costing must be done for a fixed time - usually the estimated lifetime o f the device - sinâe the C - t e r m s include the capitalized cost o f r

cryogens. A l t h o u g h any meaningful engineering analysis is impossible given the lack of g o o d cost a n d critical current data, one c a n make a n order o f magnitude e s timate. W e consider a m o d e l 100 c m bore M R I magnet a n d cryostat system providing 1.5 Τ with a cryostat area of about 2 . 1 x l 0 c m 5

volume of about 1 . 4 x 1 0 c m .


5

3

2

a n d a conductor

T h e heat leak is assumed to be evenly divided

between radiation a n d conduction (a reasonable assumption for an M R I m a g net a n d intermediate between the two limiting cases) a n d cryostat costs o f $ 1 0 0 , 0 0 0 at 4.2 Κ and $50,000 at 77 Κ are assumed.

H e l i u m cost is $ 3 0 , 0 0 0

per year for 10 years i n the 4.2 Κ case a n d nitrogen is $ 3 0 0 0 per year i n the 77 Κ case. U s i n g the cost per c m

3

for N b S n a n d Y B a C u O given above a n d 3

assuming field independent critical currents o f 1 0 A / c m 5

conservative

10

4

for Y B a C u O

gives a crossover

Y B a C u O magnet is more costly above 1.7 Tesla.

2

for N b S n a n d a 3

of about 1.7 T , i.e. the While

the actual number

obtained f r o m this exercise is clearly highly sensitive to the assumed input p a rameters, it does indicate that the possibility that Y B a C u O c o u l d be useful even at a J of 1 χ 1 0 and that the advantage w o u l d increase rapidly with increasing 4

c

J

c

and decreasing materials cost.

Stabilization, a n d C o n c l u s i o n s o n Magnets.

A n o t h e r consideration i n the d e

sign o f superconductors is "stabilization". C o n v e n t i o n a l superconductors are used i n the f o r m of multifilamentary wires w h i c h consist o f 100 to 10,000 f i l aments 6 /xm to 100/xm i n diameter distributed i n a high conductivity matrix, usually copper. T h i s fine subdivision of the superconductor is necessary to prevent thermal runaway or quenching of the conductor at currents below the critical current. T h e matrix material serves to provide a n electrical shunt around small n o r m a l zones a n d to conduct away heat. B o t h the degree of subdivision and the amount o f n o r m a l metal needed for proper stabilization depend o n the specific heat o f the superconductor and the slope o f the J versus temperature c

curve. A t this point it is not clear h o w one w o u l d stabilize a n Y B a C u O c o n ductor, but the higher specific heat o f the material at 77 Κ does significantly increase the m a x i m u m adiabatically stable strand size, thus easing this problem.

Some general conclusions w h i c h c a n be d r a w n f r o m this analysis are: •

If Y B a C u O

c a n demonstrate

the promise of usable current density at

higher fields than N b S n , it c o u l d be useful for very high field magnets 3

provided the attendant structural problems c a n be solved. •

If the cost per ampere-meter

o f Y B a C u O at 77 Κ remains higher than

conventional superconductors it will not be competitive i n applications where the superconductor materials cost dominates, i.e. very large magnets


27.

MALOZEMOFF ET AL.


291

at relatively high fields, unless there are very special demands o n space and weight. •

F o r applications at relatively low fields where little superconductor is used, refrigeration

becomes

a

significant

fraction

Y B a C u O c o u l d enjoy a cost advantage.

of

the

overall cost,

and

T h e crossover points, i n field and

device size, below w h i c h Y B a C u O is attractive will depend strongly o n the


device configuration and application. •

Additional

leverage

emerges

i n those

applications

requiring

minimal

weight, particularly in space or i n ship propulsion. Space offers the special advantage of a naturally low temperature environment so that all refriger ation could conceivably be eliminated. •

In addition to the challenge of achieving high current density at appropriate fields, manufacturing problems caused b y contacts, brittleness of the m a terial and chemical stability in the non-superconducting matrix are major obstacles yet to be solved.

Superconducting P o w e r Transmission L i n e s ( S P T L )

A n o t h e r possible transmission. veral ways.

large scale

application of superconductivity is i n p o w e r

T h i s is inherently different f r o m the magnet applications in se First, high field is not a critical factor in most designs.

superconducting magnets are i n many cases an

enabling

Second,

technology, that is

they make possible applications w h i c h have n o realistic non-superconducting competitor. In p o w e r transmission, by contrast, there exist a variety of highly developed technologies for transmitting electrical energy and the s u p e r c o n ductor is valuable only w h e n it brings a distinct cost advantage relative to c u r rent technology. M o r e o v e r

this cost advantage

must outweigh

a perceived

reliability disadvantage in an industry where unscheduled downtime must be infinitesimal. T h e costs of a S P T L divide naturally into two parts w h i c h must be c o m bined for overall costing purposes. T h e capital costs are the resources required to construct the line. T h e operating costs are primarily the cost of p o w e r c o n sumed by the line with some minor additions for maintenance. F o r a S P T L the operating costs must include the efficiency of the refrigerators, i.e. the fact that removal of one watt of heat from a 4 Κ line will consume several h u n d r e d watts of p o w e r at r o o m temperature. F o r overall costing purposes one either capital izes the operating costs or depreciates the capital costs over the lifetime of the line. In either case one must make assumptions about the cost of funds (i.e. the interest rate) a n d the busbar cost of p o w e r over the life of the line. Since b o t h of these factors vary considerably with time and geography, comparisons of various cost estimates must be made with great caution. It was largely changing views about the future busbar cost of p o w e r that led, i n the late 1970's, to the abandonment of the S P T L development projects w h i c h were begun during the


292


oil crisis of the early 70's. In view of this we will restrict ourselves here to an investigation of the scaling relationships inherent in the physics of the line using rough cost estimates only where necessary. T h e early w o r k of G a r w i n a n d M a t i s o o (28) focused o n d.c. transmission because the hysteretic losses in superconductors of the late 1960's made a.c. lines uneconomical. Transmission at d.c. is most favorable, however, for long lines a n d high powers to amortize a.c.-to-d.c. converter costs.

T h e prospects

for practical introduction of S P T L ' s , however, are greatest for smaller a.c. sys


tems, o n w h i c h we focus below.

Capital Costs.

We

consider in turn the superconducting

materials costs,

cryostat-related costs and refrigerator capital costs. -Superconductor

Cost: T h e superconductor is certain to be the most costly m a

terial in a line and hence an efficient design will minimize its volume. Its cross sectional area depends o n the fault current the line is required to carry and its surge impedance.

In the simplest approximation, the amount of superconduc

tor required will be inversely proportional to its critical current density at the line operating temperature and low field. T h e total amount of superconductor can also be expected to scale directly with the power-length product of the line (usually expressed i n M V A - m i l e s ) . T h e arguments made earlier concerning relative costs of Y B a C u O and N b S n remain valid here. T h e cost of conductor 3

per M V A - m i l e for a 77 Κ Y B a C u O line w o u l d exceed that of a 4 Κ N b S n line 3

by a factor estimated roughly as 6.

T h i s increased conductor cost must be

offset by reduced cryostat and operating costs for the line to be competitive. T y p i c a l values for the surface current density, σ of a coaxial line are 5 0 0 A / c m (where the length is measured around the circumference of the coaxial c o n ductor).

A J of 1 0 A / c m c

5

implies a 50 μιη thick film.

2

E v e n with a factor of

10 for fault current, the conductor thickness is not excessive. -Cryostat,

Dielectric and Structural Costs:

T o first order the cryostat, dielectric

and structural costs c a n be scaled as the line capacity, i.e. per M V A - m i l e in the limit of large installations. T h e estimate of a factor of 2 reduction i n cryostat costs made for the magnet case should also h o l d here. T h e cryostat cost p r o b ably scales somewhere intermediate between the cryostat area and its volume because a larger diameter cryostat will require heavier walls and support m e m bers. -Refrigerator

Capital Costs:

In small and intermediate size systems the capital

cost of a refrigeration system scales more slowly than the refrigeration power. In the limit of large systems, however, the system cost probably varies approx imately as the refrigeration capacity. T h i s is not exactly true since a long, l o w loss line will require more circulation pumps and compressor stations than a short lossy line of the same total dissipation. F o r our purposes we assume the refrigerator

capital

cost

to

be

proportional

to

the

power

input


to

the

27.


MALOZEMOFF ET AL.

293

refrigerator. T h i s is the refrigeration load at the operating temperature m u l t i plied b y the actual thermodynamic efficiency o f the system, i. e. W where T

x

=

(Tj - T ) H / T E 2

(5)

2

is the heat discharge temperature, i . e. about 3 0 0 K , T

2

is the heat

input temperature - 77 Κ for a n Y B a C u O line a n d about 7 Κ for a n actual N b S n line ( 2 9 - 3 0 ) , Η is the heat input at T , Ε is the refrigerator 3

efficiency,

2


i. e. the fraction o f ideal thermodynamic efficiency (about 0.5 at 77 Κ and 0.2 at 7 Κ (28)), a n d W is the p o w e r input at r o o m temperature.

Operating C o s t s . T h e operating cost is primarily the p o w e r required to operate the refrigerators as given b y E q . 5. T h e heat input Η has three major c o m p o nents, superconductor loss, dielectric loss a n d the cryostat heat leak. Superconductor

Loss: In a d.c. line the superconductor c a n be considered to be

lossless. In a n a.c. line however, there is a hysteretic loss connected with the movement o f flux through the material. Since the magnitude o f this loss is d e termined b y surface pinning a n d the penetration o f flux into surface irregulari ties, it scales w i t h the surface current density σ (amps/cm). T h i s loss has been well characterized for N b S n but n o measurements have been made to date for 3

YBaCuO.

Since the magnitude o f the loss is very dependent o n surface c o n

dition, one might be required to use different conductor fabrication techniques for a n Y B a C u O S P T L conductor than for a magnet conductor. -Dielectric Loss: If a n organic dielectric is incorporated into a n a.c. line the heat generated b y dielectric loss i n this material must be included i n the operating costs. Unfortunately most polymers are p o o r dielectrics (have relatively large loss tangents) at low temperatures so this term c a n be comparable

to the

superconductor loss for lines operating near 4.2 K . -Cryostat

heat leak:

T h e cryostat heat leak includes radiation f r o m r o o m t e m

perature a n d heat conducted through the structural supports.

W e assume that

the line is long enough that the leak at the tenninations is negligible.

A s i n the

magnet case, we w o u l d assume the heat leak for a 77 Κ line to be similar to that of the 77 Κ shield of a 4.2 Κ line. T h e major contribution o f a 7 7 K conductor would be to reduce the capital and operating costs associated with refrigeration.

T h e refrigerator

efficiencies

of 0.5 at 77 Κ a n d 0.2 at 4 Κ (28) yield thermodynamic efficiencies (watts i n per watt o f refrigeration delivered) of approximately 2 0 0 at 7 Κ for a N b S n 3

line a n d 6 for a 77 Κ line.

A c t u a l operating experience at the B r o o k h a v e n

demonstration line (29-30) was closer to 4 0 0 for 7 to 8 Κ operation.

One

c o u l d thus see a reduction o f about 3 0 i n the operating costs associated with refrigeration.

Reductions i n refrigerator capital costs might not be as dramatic

but should still be significant.


294


Scaling a n d the Impact o f 77 Κ Superconductor. T h e r e is a close analogy here to the earlier cost scaling discussion for magnets (see E q . 4) if one considers capacity (the power-length product) instead of magnetic field as the operative variable.

A g a i n the cost scaling depends o n whether the refrigeration and o p

erating costs are relatively independent o f capacity or whether they scale with capacity. T h e former w o u l d be the case for a d.c. or low-loss a.c. line where the heat leak through the cryostat is comparable to o r exceeds the superconductor a.c.


loss. In this situation there will be a cost-advantage crossover as a function of capacity between 77 Κ Y B a C u O and 4 Κ N b S n , with the former favorable at 3

low capacity because o f lower refrigeration cost a n d the latter favorable at high capacity because of lower net materials cost (which results f r o m the higher low-field current density). T h e r e will also be a crossover with the conventional resistive line (usually insulated b y S F gas) because o f its yet lower cooling cost 6

balanced by higher p o w e r dissipation scaling with the line capacity. These crossovers w o u l d leave a " w i n d o w " for 77 Κ Y B a C u O at inter mediate capacity. O f greatest importance, the lower refrigeration cost o f 77 Κ Y B a C u O w o u l d shift the conventional-to-superconducting crossover to lower capacity than for the earlier superconducting systems. eration cost advantage,

G i v e n the large refrig

such a d o w n w a r d shift could be substantial.

Since

hurdles to introduction of superconducting technology depend strongly o n c a pacity, such a shift c o u l d be o f great importance i n enabling commercial S P T L s . If the a.c. loss o f Y B a C u O is significant, the refrigeration costs scale more directly with capacity. T h e cost-crossovers are then m u c h more sensitive to the absolute costs, a n d meaningful evaluation awaits more reliable data o n real Y B a C u O conductors, w h i c h w o u l d allow the development a n d analysis o f r e alistic designs. In summary, Y B a C u O operation at 77 Κ c o u l d possibly reduce the c a pacity (power length product) at w h i c h the S P T L becomes attractive c o m p a r e d to conventional lines. While substitution for overhead a.c. lines will likely r e m a i n problematic given their l o w capital cost, substitution for underground compressed gas ( S F ) lines may prove practical. O n e must bear in m i n d , h o w 6

ever, that the installed transmission capacity i n the U . S . is growing at a rate o f only a few percent a year a n d hence there are limited opportunities for new large capacity lines.

Replacement o f existing lines implies a m u c h more strin

gent financial test since the savings must be sufficient to provide a rapid p a y back o f the discretionary investment. Interconnects i n C o m p u t e r s A n o t h e r possible application for high temperature superconductors is as inter connects i n computer systems with semiconducting devices (31). These c o u l d be called h y b r i d systems, since they involve b o t h superconductors a n d semi conductors.

I n particular C M O S devices are w e l l - k n o w n

to have enhanced

performance at 7 7 Κ a n d are thus potentially compatible with Y B a C u O .


27.


MALOZEMOFF ET AL.

295

T h e competitor to Y B a C u O i n this application is copper, w h i c h has a resistivity at 77 Κ of 0.24 μΩοπι, compared to a value at 3 0 0 Κ of 1.7 μ Ω α η , In fact at r o o m temperature, c o p p e r alloys with even higher resisitivity are used to reduce electromigration, w h i c h is practically eliminated at the lower t e m peratures.

T h u s c o p p e r offers a reduction i n resistance of more than a factor

of 6 at 77 K . F o r relatively short lines, as might be of interest i n interconnections o n a microcircuit chip, a lumped circuit analysis is appropriate, and the performance


c a n be estimated f r o m an R C time constant. H e r e it is not sufficient to consider only the interconnect line resistance in the total resistance R .

T h e effective

output impedance of the C M O S device also contributes and is a key feature w h i c h limits the leverage of the superconductor. T o estimate a typical C M O S device output impedance (31), one c a n use, for example, experimental results of S u n et al. (32) o n a ring oscillator with a fan-in and fan-out of 3 i n a N A N D configuration. T h e devices have a 0.5 μπι channel length and a 9 /im channel width. A t a supply voltage of 2.5 volts and a l o a d capacitance of C =

0.2 pf, the observed delay per stage is 4 5 0 psec,

implying an effective output impedance of 2 2 5 0 Ω.

T h i s is of course only a

sample value since device characteristics are highly non-linear and so the value c a n change with loading. T h e superconductor has significant leverage over 77 Κ c o p p e r only if the C u line resistance is significant compared to the output impedance.

B u t even

a line with a very small cross-section of 0.5x0.5 μ ι η and a relatively long length 2

of 5 m m has a resistance of only 4 0 0 Ω. Relative to 2 2 5 0 Ω, this w o u l d give less than 2 0 % contribution to the R C time constant. L a r g e r gate widths reduce the output impedance a n d thus increase performance and the superconductor lev erage, but at the expense of density. A n o t h e r p r o b l e m concerns the current density. F o r a typical drain c u r rent (see F i g . 5 of Ref. 30) of a few m A , the current density for a 0.5x0.5 μϊΆ

line is over 1 0 A / c m 6

2

, a value approaching the 77 Κ limit for Y B a C u O

discussed in the section o n material parameters.

F o r larger cross-section lines,

say 2.5x1 μπι , the current density drops to 1 0 A / c m , but the leverage over 2

5

2

77 Κ copper is reduced f r o m 2 0 % to a mere 2 % for the numbers discussed above. In summary, if superconducting lines c o u l d be fabricated w i t h the same ease a n d reliability as copper, they w o u l d doubtless be used i n computing sys tems with 77 Κ semiconducting devices. B u t the above estimates, while only a r o u g h guide, indicate that for the standard C M O S devices, it may be difficult to achieve sufficient performance advantage to outweigh the likely processing problems. Clearly other device configurations need to be f o u n d i n w h i c h lower output impedance offers a greater leverage for superconductivity.

A n area

where superconductivity may be more useful is i n packaging, where p o w e r losses and voltage drops of very long lines c a n be eliminated a n d current d e n sities c a n be kept lower. H e r e again copper lines offer a viable alternative with m u c h less risk. It is likely therefore that simple substitution of superconducting



296

for n o r m a l metal lines will not be enough, and more creative designs will have to be devised to exploit the advantages o f high temperature superconductivity. Electronics B a s e d o n Superconducting Devices Superconducting devices offer possibilities o f higher speed, more sensitivity and greater precision i n a b r o a d range o f electronic applications, including


mm-wave

detection, high speed digital a n d analog signal processing, ultra-

low-noise low-frequency

measuring devices a n d dc voltage standards.

Com

mercial instruments operating at o r near 4 Κ a n d utilizing J o s e p h s o n devices for high sensitivity measurements are n o w available f r o m several sources, a n d the first commercial instrument e m p l o y i n g high speed J o s e p h s o n circuitry is being introduced.

Digital computer research utilizing J o s e p h s o n devices has

been pursued seriously b y a number o f laboratories.

In addition there are a

number of noncommercial applications, e.g. the maintenance o f national v o l t age standards a n d SIS mixers i n radiotélescopes, superconducting devices.

that are best performed with

T h e question naturally arises with each o f these as

to h o w well they c o u l d w o r k at liquid nitrogen temperatures. Digital J o s e p h s o n T e c h n o l o g y Digital J o s e p h s o n technology for high performance computer applications has been pursued over the past 15-20 years i n a number o f laboratories worldwide (33-37).

While successful i n the demonstration o f working logic chips a n d

partially functional m e m o r y , this technology has seen its projected ance

advantage

narrowed

b y the rapid progress

of

perform-

semiconductor-related

technologies like V L S I , M L C packaging, g r o o v e d Si cooling and G a A s H E M T . J o s e p h s o n technology

is constrained

b y the two-terminal

nature

o f the

J o s e p h s o n device a n d the resultant need to use threshold logic. F o r general purpose digital applications, the merits o f Josephson-based technology must be considered against those o f more conventional technologies that c o u l d be available at the same time as J o s e p h s o n technology.

It

quickly becomes apparent that inductive circuit families, such as the magnetically coupled interferometer-based circuits have density potentials too l o w to be competitive. F o r logic, several circuit families that d o not involve inductive loops are k n o w n a n d have been used to make complex L S I - l e v e l chips. F o r m e m o r y , there is at present n o k n o w n alternative to inductive storage, a n d the magnetic cross-coupling between adjacent cells sets limits o n achievable m e m ory densities (36.38). L e t us examine possible ways i n w h i c h the new high temperature superconductors might affect digital J o s e p h s o n electronics: (1) T h e new s u p e r c o n ductors might allow additional performance

to be achieved with Josephson

circuits operating either at liquid helium or liquid nitrogen temperatures. Convenience

o f liquid nitrogen

cooling

might

ease

(2)

the introduction o f

J o s e p h s o n technology.


27.

MALOZEMOFF ET AL.

Applications

of High-Temperature

297

Superconductivity

T h e principal means to higher performance with the new superconductors is through the increased size of the superconducting gap Δ. T h e increased gap promises higher b a n d widths for superconducting lines (A/lirh), damental limit for the switching time of a gate (also Δ/2πΗ), creased voltage capability for driving transmission lines.

a faster f u n as well as i n

In practice the

switching speeds of devices, and hence the fastest pulses put onto s u p e r c o n ducting lines, are limited not by the gap frequency of conventional s u p e r c o n ductors, but rather b y junction capacitances C j to Z C j 0


transmission line impedance.

where Z

0

is the

Therefore the larger gap of the new s u p e r c o n

ductors is not significant as far as this aspect of high speed operation.

The

ability to better drive lines is one that c o u l d be usefully e m p l o y e d i n parts of circuits, particularly J o s e p h s o n m e m o r y , but there is n o direct effect o n m e m ory density. A n o t h e r way of making use o f the larger gap w o u l d be to r u n at higher current levels, keeping the transmission-line impedance constant, so as to improve device noise margins.

T h e limited cooling capability of liquid h e

lium, however, w o u l d necessitate lower ultimate circuit densities at the higher p o w e r levels. T h u s it does not appear that the larger gap of the high temper ature superconductor c a n significantly improve the performance projections for Josephson technology at 4 K . U s i n g the new superconductors to permit higher temperature operation of digital J o s e p h s o n circuitry is another possibility. Increased thermal noise i n this case dictates higher operating current levels, but with the increased cooling capability available at liquid nitrogen temperature increased p o w e r could be tolerated. T h u s the technology c o m p a r i s o n for 77 Κ J o s e p h s o n similar to that of 4.2 Κ Josephson.

(Refrigeration is not a primary c o n c e r n i n either case.)

O n e new o p t i o n for a digital system at 77 Κ is the use of J o s e p h s o n logic c i r cuitry with dense semiconductor m e m o r y , the prospects for dense Josephson m e m o r y not being favorable.

T h e challenge with this o p t i o n is to develop the

requisite high speed interface circuitry between the two technologies. In addition to these circuit considerations, there is a considerable amount of materials a n d processing development needed to make Josesphson technol ogy with Y B a C u O a viable option. In particular, Y B a C u O ' s short coherence length and its sensitivity to oxygen diffusion are hurdles to be overcome i n the development of high quality tunnel junctions. tions

can

be

demonstrated,

meaningful

U n t i l reasonable quality junc

prospects

for

YBaCuO

digital

Josephson systems are very difficult to assess. N e w C r y o g e n i c Device Possibilities T h e r e has been active interest i n new transistor-like devices compatible with superconductivity for a number of years (39).

Part of the interest derived f r o m

difficulties i n using the two-terminal J o s e p h s o n device i n digital circuitry. W i t h the discovery of superconductors with transition temperatures above 77 K , the device needs a n d opportunities need to be reassessed.

MOSFET's, GaAs

H E M T ' s , and heterojunction bipolar transistors ( H B T ' s ) w o r k well at 7 7 K , so



298

the d e m a n d for a new three-terminal device at 77 Κ is certainly less than it was at 4 . 2 K . Nevertheless, enhanced performance or lower p o w e r w o u l d be desir able for 77 Κ transistors. Enhancements are certainly possible within c o n v e n tional semiconductor approaches:

e.g. scaling, sharper d o p i n g profiles, etc.,

a n d there is m u c h discussion of the possibilities i n semiconductor device liter ature.

H e r e we discuss new features w h i c h superconductivity might add to

k n o w n devices ( F E T s )

a n d new devices that might be possible with super


conductors.

Superconductivity and F E T s .

F E T s with channels that c a n be induced into

the superconducting state were proposed some time ago, a n d first d e m o n strated i n the last couple of years with b o t h a Si-based device (40) InAs-based

device (41).

and a

In these devices, the channel is induced into the

superconducting state b y a proximity effect f r o m superconducting source a n d drain electrodes. tance

are

T h e strength and decay length of the proximity effect dis

governed

ξ = (^ /i/6?rmek T) 3

B

channel length.

1/2

(3i7

by n)

the 1 / 3

superconducting

coherence

induced i n the channel, relative

length to

the

H e r e μ a n d η are respectively the mobility and the carrier

density of the semiconducting channel. T h e higher mobility and lower carrier density i n I n A s resulted i n devices with channels approaching 1 μιη showing some effects, whereas the Si devices had channels of 0.1 to 0.2 μιη. T h e output voltage range over w h i c h superconductivity makes some difference is restricted to voltages o n the order of the superconducting gap voltage,

Applications of High-Temperature Superconductivity - ACS Publications

Recommend Documents