Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

THERMODYNAMIC CONSIDERATIONS FOR THE DESIGN OF ACTIVE

MAGNETIC REGENERATIVE REFRIGERATORS

J. L. Hall, C.
E. Reid, I. G. Spearing,and J. A. Barclay
Cryofuel Systems Group, University of Victoria
P. O. Box 3055, Victoria, B.c., Canada V8W 3P6

ABSTRACT

Recent work has yielded new insights into the thermodynamic operation of active
magnetic regenerative refrigerators (AMRR's). Existing theory specifies a linear profile
of adiabatic magnetization temperature change (AT.d) with temperature (T) of the
material in the flow direction of the regenerator. This theory is shown here to be
incorrect because an AMRR cannot be represented as a collection of an infinite number
of independent refrigerators, as is commonly thought. Furthermore, it is argued that
there is no unique AT.iT) profile for idealized (reversible) AMRR's; however, acceptable
profiles must satisfy an integral constraint on the total magnetic work input, as well as
satisfying boundary conditions on AT.d on the hot and cold edges of the regenerator. It
is shown that convex temperature-distance (T(r» profiles result in minimized entropy
generation in an AMRR; however, the magnitude of the magnetic work term appears is
too small to produce significant convexity. Also, it is shown that efficient AMRR's
require monotonically increasing T(r) profiles in the magnetized region. This condition
requires d(AT.d)/dT ~ -1 for any magnetic material that is used in the regenerator.
Finally, some examples are given to illustrate how these new thermodynamic
considerations are being used in the design of a rotary AMRR natural gas liquefier.

INTRODUCTION

Magnetic refrigeration is based on the phenomenon that some materials undergo


a temperature change when they are adiabatically magnetized and demagnetized. This
"magnetocaloric effect" can be combined with heat flows to and from the magnetic
material to create a thermodynamic cycle. One common example of this is the magnetic
Brayton cycle shown in Fig. 1. The material is adiabatically magnetized in process 1-2,
rising to a temperature AT'd above the initial temperature T1. The material is then cooled
back down to T1 in process 2-3, with the transferred heat being rejected to the
environment. Adiabatic demagnetization occurs in process 3-4, lowering the material
temperature below the initial temperature by AT.d. The cooling load is then absorbed in
process 4-1, which completes the cycle. Note that the use of the term "magnetic Brayton
cycle" is deliberate and underscores the direct analogy between magnetic and gas
thermodynamic cycles. In this case, the magnetization processes take the place of gas
compression and expansion in the standard Brayton cycle, and the heat transfer
operations occur at constant magnetic field rather than constant gas pressure.

Advances in Cryogenic Engineering, Vol. 41


Edited by P. Kittel, Plenum Press, New York, 1996 1653
2

S
Figure 1. Magnetic Bryton Cycle.

The magnetocaloric effect in most ferromagnetic materials is strongly temperature


and field dependent, with a maximum value at the magnetic phase transition
temperature. For example, a typical material such as Gadolinium exhibits a caret-shaped
profile as illustrated in Fig. 2. These are results from our molecular field model that are
in agreement with the experimental data of Benford and Brown.1 The peak in the profile
occurs at the Curie temperature, the point at which Gadolinium undergoes a phase
transition from paramagnetism to ferromagnetism. The magnitude of the magnetocaloric
effect is small even at the peak, a fact that is generally true for other magnetic materials
as well. Therefore, given attainable fields of the order of 10 Tesla with existing
superconducting magnet technology, maximum temperature changes of only 15 to 20
Kelvin occur. This contrasts sharply with adiabatic gas compression processes that can
produce hundreds of Kelvins of temperature change at room temperature. The relatively
small <'iT. d, combined with the narrow temperature range over which it is non-nero, has
a major impact on cycle selection for magnetic refrigeration applications. In particular,
recuperative cycles are restricted to source-sink temperature spans of the order of <'iT.d

20

/
"
.....

/
/
, "-
g 15 3.0T /
/
"-
Q) / "-,,-
Ol 6.0T "-
"
/
C
<Il / /
.r. 9.OT ,/
0 ,/
./
..,-
:s
Q)

..,- ; '
T§ 10 ,.-
,.-
Q)
,-
,-
a. ,-'-
E
Q)
I-
.~
iii
.0
<Il
-----------------
U 5
<I:

OL-__ ~ __ ~ __ ~ ___ L_ _ ~ __ ~ __ ~ __ ~ __ ~ ____ ~ __ ~~

200 240 280 320


Temperature (K)
Figure 2. Magnetocaloric Effect in Gadolinium.

1654
Figure 3. Conceptual Reciprocating AMR Device.

because the magnetic material must have a significant magnetocaloric effect over the
entire temperature range it experiences.
Applications involving large temperature spans require regenerative magnetic
refrigeration cycles. This approach separates the functions of heat pumping and
temperature spanning, allowing the small magnetocaloric effect to be leveraged across
a large temperature range. Each element of the regenerator will experience only a small
temperature change during steady-state operation; therefore, once can select materials
that are optimized for the mean temperature at every location throughout the
regenerator. Specifically, the material will be optimized if the peak in the ~T.d(T) curve
(the Curie point) coincides with the mean temperature at that location.
This idea, while simple in principle, has proven to be very difficult to implement
in practice. Considerable effort has been expended by many researchers over the past
two decades in attempts to develop and test various regenerative magnetic refrigerators,
with modest success at best.2-12 Much of the difficulty stems from the fact that the
working material is a magnetic solid, not a fluid. In particular, heat transfer within or
between solid bodies is not particularly good, leading most researchers to use a fluid to
thermally connect the magnetic solid to both the load and the environment. This requires
that the magnetic material be manufactured in the form of a porous solid through which
the fluid can be pumped. The common geometries include particle beds, perforated
plates, parallel plates, and wire screens. The need for high heat transfer coefficients
typically results in very fine-scale flow passages « 1 mm), which leads to a host of
problems associated with flow pressure drops, manufacturability, and structural
integrity. Finally, the combined effects of fluid flow, heat transfer and magnetization
result in a coupled physical problem that is very difficult to analyze mathematically.
In an attempt to gain insights into the problem, most researchers have sought
analogies between regenerative gas compression cycles and regenerative magnetic cycles.
However, unlike the recuperative magnetic Brayton cycle example described in Fig. 1,
regenerative analogies are not straightforward, and in fact have often proven to be
misleading. One reason for this is the subtlety in defining the thermodynamic system
where a solid material does the work, but a circulating fluid transports the heat. Another
key reason is that many researchers have employed an active magnetic regenerative
(AMR) device that uses magnetic material for both the work input and passive
regeneration functions, a coupling that even further blurs the boundary of the
thermodynamic system in question. Nevertheless, AMR's appear to have significant
advantages over other magnetic cycles t3•14 thereby motivating their continued use despite
the attendant analytical complexities.
The present work was motivated by the need to better understand the
thermodynamic theory behind active magnetic regenerative refrigerators (AMRR's) in
order to properly design a small-scale prototype that will liquefy natural gas. It will be
argued that the AMRR is a unique thermodynamic system that cannot be adequately
described by analogies to the various recuperative and regenerative gas cycles. We will
attempt to show where these analogies break down and how this invalidates some of

1655
the theoretical conclusions that have been drawn concerning these devices. Finally, we
will present some new insights to the thermodynamics of AMRR's and show how these
results are being incorporated into the prototype design. Further details can be found
in the recent thesis by ReidY

PROBLEMS WITH EXISTING THEORY

In its simplest configuration, an active magnetic regenerative refrigerator consists


of a porous magnetic solid surrounded by a solenoidal magnet that is periodically
turned on and off (Fig. 3). The fluid that fills the void space is moved back and forth
through the system by a pump or displacer. This fluid accepts heat at the cold heat
exchanger and rejects it at the hot heat exchanger. Note that a temperature gradient
exists across the regenerator at steady-state operating conditions. The fluid flow and
magnetization processes are synchronized in a similar fashion to the magnetic Brayton
cycle example described above. First, the fluid is held stationary while magnet is turned
on. This raises the regenerator temperature by an amount equal to the magnetocaloric
effect. Second, the fluid is pushed from the cold to the hot end, picking up heat from the
solid material and rejecting it at the hot heat exchanger. Third, the fluid flow is stopped
and the magnet is turned off. The regenerator temperature drops back down. Fourth, the
fluid is pushed from the hot to the cold end, depositing heat back into the regenerator
and accepting heat from the load at the cold heat exchanger.
A serious problem arises when we try to define the thermodynamic cycle
corresponding to this AMRR process. Every part of the system experiences different
operating conditions. The solid regenerator material exists across a range of
temperatures, resulting in a different magnetocaloric effect for each piece. Likewise, the
fluid moves back and forth through the device with each infinitesimal element oscillating
across a different temperature range depending on its starting location. Many previous
researchers have attempted to finesse this non-uniqueness problem by assuming that
some kind of composite cycle exists that can adequately describe this thermodynamic
process.4,5,16 This composite cycle approach is typically justified on the basis that the
complete regenerator is equivalent to a "cascade" of thermodynamic cycles for each solid
element. More specifically, it is postulated that each solid element functions as its own
Carnot or Brayton or Ericcson refrigerator, receiving heat from the adjacent lower
element and pumping it up to the adjacent higher element. The sum of all of these sub-
cycles therefore becomes equivalent to a single composite cycle that spans the complete
temperature range of the device, Note that this approach is directly analogous to multi-
stage gas compression devices,
However, we argue here that this underlying premise is false. The AMRR does
not consist of a cascade of sub-cycles because each solid element of the regenerator does
not directly pump heat to its neighbours, All of the solid elements are either accepting
or rejecting heat at the same time depending on which way the fluid is flowing, Heat
is not pumped from one solid element to the next; instead, it is pumped to the fluid
which then carries it across the regenerator. If the fluid exits the regenerator during this
half cycle, the heat is rejected to the environment; otherwise, the heat is deposited back
into the regenerator where the fluid stops moving, In this very basic sense, there is no
direct communication between adjacent solid elements, This is reflected in their
thermodynamic cycles plotted on the T-s plane (Fig, 4), Adjacent solid elements in the
AMRR will generally have overlapping plots, in contrast to those seen in true cascaded
(multi-stage) systems, This is indicative of the fact that those solid elements do not pump
heat from one neighbour to the next in a cascaded fashion,
Another argument can be advanced against the cascade theory, Consider two
separate AMRR devices connected by an intermediate heat exchanger (Fig, 5), In this
case, each AMRR is clearly an independent refrigerator with its own separate fluid loop.
There is no need to synchronize magnetization or fluid flows between the stages, and

1656
f

s
Figure 4. Overlapping Cycles of Adjacent AMRR Solid Elements.

in fact, they may operate at different frequencies altogether. It is also clear that heat from
the cooling load is pumped from one stage to the next in the usual cascaded fashion.
This operation contrasts sharply with the postulated sub-cycle cascading within a given
AMRR device itself. They are not the same.
The preceding arguments suggest that analogies between AMR refrigerators and
regenerative gas compression devices either do not exist or simply are not useful. In fact,
we think there is a strong case for considering the AMRR to be a unique thermodynamic
system altogether. We deliberately avoid the use of the term "cycle" here because there
is no unique cycle described by the elements of the AMRR, nor is there a composite
cycle composed of cascaded sub-cycles. We will proceed on this basis and describe how
this hypothesis leads to some significant changes in the theory of designing AMRR's.

~Tad PROFILES

Perhaps the key result of earlier theories was the specification of a linear ~T.d
versus T profile for a given AMRR design.16 This profile was derived in four steps. First,
it was argued on the basis of the sub-cycle cascade model that entropy flows from one
element of solid to the next will be equal in an ideal (reversible) device. More
specifically,

S:i2:constant (1)
T

AMRR '1 AMRR '2

Intennediate
HEX

Cold Hot
HEX HEX

Pump Pump

Figure 5. Schematic of a Two-stage AMRR.

1657
where S is the entropy flow, Q is the heat flow and T is the absolute temperature where
the heat flow occurs. Second, the net heat flows into the AMRR at the cold and hot ends
(Qc and <:4.) are directly proportional to the magnetocaloric effect. In other words, the
heat transfer fluid can only be heated up by an amount AT.d,h above the environment
temperature, and cooled by an amount AT.d,c: below the load temperature. Therefore,

(2)
and
(3)

where Inf is the fluid mass flow rate and CPf is the fluid heat capacity. The third step is
to substitute equations Eqs. 2 and 3 into the entropy balance Eq. (1) to find

(4)

where a constant specific heat capacity has been assumed. Finally, the fourth step is to
draw upon the cascade analogy and argue that Eq. 4 must apply throughout the
regenerator and not just at the endpoints. Therefore, this sets the criterion for material
selection in the AMRR: the magnetocaloric effect must be a linearly increasing function
across the temperature span of the device, with a slope given by Th/Tc.
Now, since we have argued that the cascade analogy is not valid, then the
requirement that Eq. 4 be satisfied across the interior of the active regenerator becomes
invalid as well. The ATad versus T profile does not have to be linear in an ideal device.
However, Eq. 4 still applies at the boundaries of an AMRR because this simply reflects
the need to have the same entropy flow from the load to the environment in a reversible
process (Eq. 1).
Although there are boundary conditions for the ATad(T) profile across an idealized
AMRR, there does not appear to be a unique curve for satisfying a zero entropy-
producing constraint. Consider the possible entropy producing mechanisms in· the
AMRR: fluid flow pressure drop, heat transfer across a finite temperature difference
between the solid and the fluid, and heat conduction through the solid and fluid due to
spatial temperature gradients. Note that none of these mechanisms directly depend on
the ATad(T) profile. In essence, if you can provide a zero viscosity fluid, infinite surface
area and zero spatial thermal conductivities, then the AMRR will produce zero entropy
for any adiabatic temperature profile that is specified.
However, at least one other constraint on the AT.d(T) profile in addition to the
boundary conditions at the hot and cold end. Each infinitesimal element of solid will
undergo a process similar to the magnetic Brayton cycle shown in Fig. 1. Note that the
area inside the closed curve on the T-s plane is equal to the magnetic work input per
cycle. Note also that this area is roughly proportional to the magnetocaloric effect, ATad,
although the exact relationship depends on the material properties. The integral of this
work over all solid elements must equal the total work input into the AMRR. By
definition, this must equal the difference between the heat input and heat rejection,

(5)

where W is the total work per cycle divided by the frequency. This establishes a
relationship between the integral of the AT.iT) profile and the magnetic work input to
theAMRR,

1658
12

,'''',
6.0TGdDyl
I ,,
10 6.0TGdDy2
6.0TGdDy3
6.0TDy /\
I ,
I \
I \
I
\
., ,
I
I \ . ,.,
I ).'
, '\ '
,',
/
~,"
/
/
~
:;",'" 4 ,./
//
""
~ "
./' "", ,
"- ....
2 _-----,," ....
...... _---
Temperature (K)

Figure 6. Layered Materials in an AMRA.

(6)

where the function f depends on the material properties and the details of the heat
transfer process between the solid and the fluid.
In summary, therefore, we can see that while there are an infinite number of
possible .U.iT) profiles for an idealized AMRR, they must all satisfy the boundary
conditions of Eq. 4 and the integral work constraint of Eq. 6. Linear profiles are not
required; but, they are the simplest functions that can satisfy the boundary conditions
and they remain good candidates for practical devices. Moreover, in the temperature
range that our group is currently interested in, from room temperature down to liquid
natural gas (110 K), the best available rare-earth magnetic materials tend to have
magnetocaloric effects that scale with absolute temperature anyway.16 This has motivated
us to design around a nominally linear L1T.iT) profile, even though it is not strictly
required by minimum entropy AMRR requirements. IS

T(r) PROFILES

Real AMRR's have to be concerned with the spatial distribution of the


magnetocaloric effect, L1T.ir), in order to locate the solid material at the correct absolute
temperature. This is particulary important for devices that require more than one
material to span a large temperature range between the load and the environment.9,16
Layered materials with different Curie temperatures help to maximize the
magnetocaloric effect across the entire regenerator, thereby maximizing the magnetic
work input per unit mass. Figure 6 presents one example of this, showing layered
materials spanning the range from 100 to 250 K.. However, the placement of this
material presupposes knowledge of the T(r) profile since the magnetocaloric effect is a
function of temperature, not spatial position.
It is not too difficult to argue that T(r) will be close to a linear function in an
AMRR. Certainly, passive regenerators have been shown to possess nearly linear T(r)
profiles at steady-state operating conditions, with only minor deviations due to end
effects. 17 This linearity reflects that fact that the heat transfer between the solid and the

1659
fluid is the same at all points within a uniformly porous regenerator. Given that this will
also be true in an active magnetic regenerator, the question becomes to what extent can
the magnetic work input in an AMRR cause non-linearity in the T(r) profile?
A precise, quantitative answer to this question is not yet available. The full
magneto-thermodynamic problem can only be solved with numerical methods, and there
have been relatively few studies to date. Recent work by Spearing18 suggests that
magnetic work input does not significantly alter the linearity of the T(r) profile in a
rotary AMRR model; however, definitive conclusions await further work.
We can also offer two other plausibility arguments in favour of the conclusion
that magnetic work inputs to an AMRR will not significantly change the nominally linear
T(r) profile. First, the magnetocaloric effect is small compared to the thermal heat content
of the regenerator. This scaling goes like ~Tad/T - 15/300 = .05 for typical materials,
suggesting that the magnetic work input changes the energy of the system by only a few
percent. Second, this ratio will be approximately constant for a layered bed of materials
spanning a large temperature range,16 suggesting that the magnetic work will be evenly
distributed. Evenly distributed magnetic work should have the same effect as evenly
distributed heat transfer, thereby promoting a linear T(r) profile.
Despite the seeming inability of the magnetic work to significantly change the T(r)
profile that results from passive regeneration, it is an interesting exercise to consider
what T(r) profile will minimize the entropy production of a non-ideal AMRR. Although
we have not yet solved this problem in a completely rigorous fashion, some insights can
be obtained by considering an approximate analysis of the entropy production due to
solid-fluid heat flow. This heat transfer per unit volume, (2, occurs across a small
temperature difference ~Tsf where,

(7)

and Ts and Tf are the solid and fluid temperatures respectively. The volumetric entropy
generation due to this process is given by

(8 )

where T is the absolute temperature at which the heat transfer occurs.

Now, (2 is itself proportional to ~Tsf because of the physics of convective heat


transfer; i.e.,

(9 )

where h is the heat transfer coefficient, and A''' is the surface area per unit volume.
Moreover, standard regenerator theory17 shows that

11 Tsf '" ~~ + time dependent terms (10)

The time dependent terms can be neglected in the long blow limit (large thermal mass
of the solid regenerator). Combination of Eqs. 7 through 10 with this assumption leads
to
.
Sgen'" T21 (dT)2
dr ( 11)

1660
Integration of this term across the regenerator bed gives

(12)

If we now postulate a general temperature profile of the form T(r)=ar", where a and n
are constants, then Eq. 12 becomes

This indicates that entropy production will be minimized when the exponent n is small.

(13)

Therefore, convex T(r) profiles where n < 1 will result in minimized entropy production
under these assumptions.
Another valuable insight into the design of AMRR's can be obtained from an
analysis of the interplay between the magnetic work input and heat flows from the solid
to the fluid. It is important to remember that a magnetic material will only perform net
work if the solid is cooled by the fluid while it is magnetized. This is equivalent to a
counter-clockwise cycle in the T-s plane. Therefore, it is required that,

(14)

in the magnetized region. Physically, this means that the fluid temperature must always
lag the solid temperature as fluid flows from the cold to the hot side of the magnetized
regenerator. Note that Eq. 14 refers to the temperature profile after magnetization;
therefore, this constitutes an implicit restriction on allowable dT.ir) profiles. This
restriction can be derived as follows.
Let T(r) be the temperature profile before magnetization, and T(r) be the
temperature profile after magnetization. Therefore,

T·(r) =T(r) +.6Tad (r) (15)

and,
dT· dT d.6 Tad (16)
--=-+---
dr dr dr

Using the chain rule on the last term gives,

d.6 Tad
dr
=( d.6dTTad)( dr
dT) (17)

Substituting Eq. 17 into Eq. 16 yields,

(18)

Therefore, we see that the inequality of Eq. 14 can only be satisfied if dT/dr > 0 and

1661
dt:..Tad (19)
--)-1
dT

In other words, the t:..T.d(T) profile cannot decrease by more than one degree per degree
in the regenerator.
It is not immediately obvious that Eq. 19 will be satisfied by magnetic materials
given that the typical caret-shaped profile necessarily contains negative slope regions
(Fig. 2). However, a survey of rare-earth (lanthanide) magnetic refrigerants does reveal
that the negative slope sections of their t:..Tad(T) do satisfy the one degree per degree
constraint. We have therefore proceeded to use these materials in the design of the active
magnetic natural gas liquefier prototype, selecting materials that give us the maximum
magneto-caloric effect possible across the entire temperature span of the device. ls Note
the additional requirement that the interfaces between materials must be properly
positioned to avoid steps in the composite t:..T.iT) profile. Otherwise, these steps may
violate the one degree per degree constraint.

CONCLUSIONS

It has been argued that the active magnetic regenerative refrigerator (AMRR)
constitutes a unique thermodynamic process that cannot be equated to a cascade of
thermodynamic cycles of each infinitesimal active element. This postulate means that the
existing theory requiring a linear t:..T.d(T) profile in the magnetic regenerator is incorrect.
Instead, it is argued that there is no unique t:..Tad(T) profile for idealized AMRR's,
although there are boundary conditions that must be satisfied, as well as an integral
constraint on the total magnetic work input to the system. Non-ideal AMRR's will
minimize entropy production if the spatial t:..T.ir) is convex; however, the magnitude of
the magnetocaloric effect appears to be too small to effect significant changes from the
expected linear profile resulting from passive regeneration. Finally, a derivation has been
presented to show that acceptable magnetic materials, and the interfaces between layered
materials, in a regenerator must satisfy the constraint d(t:..Tad) / dT ~ -1.

ACKNO~EDGEMENTS

This work is sponsored by the Natural Sciences and Engineering Research Council
of Canada, and Centra Gas B.c., Inc.

REFERENCES

1. Benford, S. M. and Brown, G. V. (1981). "T-s diagram for gadolinium near the Curie
temperature", J. Appl. Phys, Vol. 52, No.3, pp 2110-2112.
2. Hashimoto, T., Numasawa, T., Shino, M. and Okada, T. (1981). "Magnetic refrigeration
in the temperature range from 10K to room temperature: the ferromagnetic refrigerants",
Cryogenics, Vol. 21, No. 11, pp 647-653.
3. Barclay, J. A (1982). "The Theory Of An Active Magnetic Regenerative Refrigerator",
Los Alamos Report LA-UR-82-1792. Also NASA-CP-2287 (1983).
4. TaUSSig, C. P., Gallagher, G. R, Smith, J. L. and Iwasa, Y. (1986). "Magnetic
Refrigeration Based On Magnetically Active Regeneration", 4th International Cryocooler
Conference, Easton, Maryland, September 1986.
5. Cross, C. R, Barclay, J. A, De Gregoria, A J., Jaeger, S. R and Johnson, J. W. (1987).
"Optimal Temperature-Entropy Curves For Magnetic Refrigeration", Advances in
Cryogenic Engineering, Vol. 33, pp 767-776.

1662
6. Smith, J. L. and Iwasa, Y. (1990). "Material And Cycle Considerations For Regenerative
Magnetic Refrigeration", Advances in Cryogenic Engineering, Vol. 35, pp 1157-1164.
7. Green, G., Chafe, J., Stevens, J. and Humphrey, J. (1990). "A Gadolinium-Terbium
Active Regenerator", Advances in Cryogenic Engineering, Vol. 35, pp 1165-1174.
8. Hashimoto, T. (1991). "Recent Progress In Magnetic Regenerator Materials And Their
Applications", Proceedings of the 18th International Conference of Refrigeration,
Montreal, Quebec, August 1991.
9. Schuricht, S. R., De Gregoria, A J., and Zimm, C B. (1992). "The Effects of a Layered
Bed on Active Magnetic Regenerator Performance", 7th International Cryocooler
Conference, Santa Fe, New Mexico, November 1992.
10. Zimm, C B., Ludeman, E. M., Severson, M. C and Henning, T. A (1992). "Materials
For Regenerative Magnetic Cooling Spanning 20K to 80K", Adv. Cryo. Eng., Vol. 37, pp
893-890.
11. De Gregoria, A J. (1992). "Modeling the Active Magnetic Regenerator", Advances in
Cryogenic Engineering, Vol. 37, pp 867-873.
12. Wang, A A, Johnson, J. W., Niemi, R. W., Sternberg, A A and Zimm, C B. (1994).
"Experimental Results of an Efficient Active Magnetic Regenerator Refrigerator", 8th
International Cryocooler Conference, Vail, Colorado, June 1994.
13. Barclay, J. A (1984). "A Comparison Of The Efficiency Of Gas And Magnetic
Refrigerators", 22nd National Heat Transfer Conference, Niagara Falls, New York,
August 1984.
14. Waynert, J. A, DeGregoria, A J., Foster, R. W. and Barclay, J. A (1989). "Evaluation
of Industrial Magnetic Heat Pump/Refrigerator Concepts That Utilize Superconducting
Magnets", ANL-89/23.
15. Reid, C E. (1995). "Development of Magnetic Refrigerants For an Active Magnetic
Regenerative Refrigerator", Master's thesis, University of Victoria, Victoria, B.C, Canada.
16. Reid, C E., Barclay, J. A, Hall, J. L. and Sarangi, S. (1994). "Selection of magnetic
materials for an active magnetic regenerative refrigerator", Journal of Alloys and
Compounds, Vol. 207/208, pp 366-371.
17. Hausen, H. (1983). Heat Transfer in Counterflow, Parallel Flow and Cross Flow,
(translated by M. S. Sayer), McGraw-Hill, Inc. Chapters 11 and 13.
18. Spearing, I. G. (1995). "A Numerical Model For a Rotary Active Magnetic
Regenerative Refrigerator", Master's thesis, University of Victoria, Victoria, B.C, Canada.

1663

You might also like