Background of the Invention Thermal regenerators are used to transfer heat to and from fluids flowing through the regenerator. Typically a thermal regenerator absorbs heat from a fluid flowing through the regenerator and then releases heat to the fluid as it flows back through the regenerator in an opposite direction. Such a regenerator is a passive regenerator in that it only provides for the transfer of heat to and from a fluid while refrigeration is provided by, for example, the expansion of a gaseous refrigerant in a regenerative gas cycle.

In addition to passive regenerators, active regenerators have been developed that provide refrigeration. For example, an active regenerator can consist of a material exhibiting the magneto-caloric effect; application of a magnetic field can then be used provide heating or cooling of a fluid flowing through the regenerator.

Both active and passive regenerators preferably have large thermal masses, large heat transfer coefficients, and little resistance to fluid flow. In addition, regenerators should not conduct heat in directions parallel to the direction of fluid flow through the regenerator. The particles of the regenerator material should be solid particles, without voids and preferably have low overall porosity in a regenerator. These conditions are generally not simultaneously satisfied.

Various types of regenerators are known, including wire-screen, foil, perforated-plate, and particle regenerators. Wire-screen regenerators have layers of a fine gauge wire screen that are stacked inside a container. The wire screens provide openings for fluid flow and provide large surface areas for heat transfer. Unfortunately, it is difficult to accurately align the layers of wire screens so that fluid flow resistance tends to be variable. In addition, preparation and cutting of the wire screens is expensive. Because the wire screens are thin, a very large number of wire screens must be assembled in the container. This assembly is also expensive and difficult to automate.

Foil regenerators use thin foils of a regenerator material. In one foil regenerator, a metal foil is dimpled and then rolled, forming a series of flow channels. Metal-foil regenerators have also been made by photolithographically defining and then etching flow channels in a thin foil as described in U.S. patent 5,429,177 to Yaron et al. Foil layers with these etched flow channels are then stacked to form a regenerator. The foil layers can also be etched to provide flow passages between the flow channels.

Particle regenerators generally are porous, finely divided aggregates of particles that permit fluid flow therethrough. The aggregates have large surface areas for heat transfer between the particles and the fluid. Conventional particle regenerators use spherical metallic powders formed by, e.g., centrifugal or gas atomization. The yield of appropriately sized particles from these methods is low and such particles may have internal voids, reducing the thermal mass per unit volume of the powder.

Perforated-plate regenerators as described, for example, in U.S. patent 5,101,894 to Hendricks use a stack of perforated plates. The plates are perforated with small holes whose length-to-diameter ratios are such that the holes function as tubes in heating or cooling a fluid flowing through the holes.

Because the extent to which a regenerator can heat or cool a fluid depends on the heat capacity of the materials of the regenerator, regenerators are preferably made of materials having large heat capacities. However, if cooling to very low temperatures is intended, material selection is limited because material heat capacities diminish with decreasing temperature. In addition, many materials with relatively high heat capacities at low temperatures are brittle, so that forming the material into a shape suitable for incorporation into a regenerator is difficult and expensive. Therefore, regenerators requiring wires, wire screens, plates, spherical particles, or foils are difficult and expensive to manufacture for many advantageous materials. Moreover, regenerators are usually exposed to mechanical stresses such as differential pressures and thermal cycling that cause the particles of a particle regenerator to rub against each other. Such rubbing tends to fracture the particles of brittle materials. The fine particles resulting from the fracture contaminate the cooling system, clogging the flow passages or migrating out of the regenerator to the rest of the cooling system.

Summarv of the Invention The regenerators of the present invention take advantage of the relatively large low- temperature heat capacities of the compounds of rare earth metals and transition metals (RE/TM compounds), such as the materials described in U.S. patent 5,186,756 to Arai et al. Other compounds with advantageous properties include compounds such as Gds(Si2Ge2), Gds(SixGel x)4 (x10.5), and alloys of these materials, described in "Effect of alloying on the giant magnetocaloric effect of Gd^(Si2Ge2) and other Gd compounds," J. Magnetism and Magnetic Materials 176 (1997), L179-L184. Other materials include but are not limited to the materials described in U.S. patent 4,082,138 to Miedma et al. and U.S. patent 5,462,610 to Gschneidner et al.

Furthermore, the present invention provides regenerators and fabrication methods for regenerators that use irregularly shaped particles of these or other compounds, including brittle RE/TM compounds. The regenerators of the present invention are preferably monolithic regenerators, i.e. regenerators in which the regenerator particles are bound together to form a

self-supporting structure. Because irregularly shaped particles can be used, the compounds need not be formed into wires, screens, sheets, perforated plates, or spheres. Particle fabrication is inexpensive and can be performed with a grinding apparatus such as a mortar and pestle. Particle sizes are readily selected by sieving the particles produced by grinding. Preferably, the particles are sieved during grinding to obtain a high yield of appropriately sized particles.

To form a monolithic regenerator, particles of a regenerator material such as a RE compound are placed in a mold. An end of the mold is covered by a screen separator that contains the particles in the mold. A dilute solution of a binding agent is then applied to the particles and flows through the particles and the screen separator. Excess binding agent is then removed by forcing a gas flow through the particles in the mold. The gas flow for removing excess binding agent can be provided by a source of pressurized gas. Alternatively, a vacuum can draw a gas flow through the particles. The application of binding agent and the removal of excess binding agent can be repeated until sufficient binding agent remains among the particles while still permitting fluid flow therethrough. The number of repetitions is adjusted according to the dilution and viscosity of the binding agent.

The binding agent is then cured or hardened and the bound particles form a monolithic regenerator that is extracted as a whole from the mold. The screen separator can remain attached to the monolithic regenerator or can be removed. The finished monolithic regenerator can be cut into smaller regenerators if desired.

Because excess binding agent is removed from the particles, fluid flow through the monolithic regenerator is not appreciably impaired. In addition, rates of heat transfer to and from the particles are not appreciably changed. The binding agent holds the particles in fixed relationship, reducing particle rubbing. Thus, even brittle particles do not fracture. Low values of thermal conductivity along the fluid flow direction are maintained because the contact areas between adjacent particles are small and provide a poor path for heat conduction.

To facilitate removal of the monolithic regenerator from the mold, the mold may be a tube of a non-stick material such as TEFLON and may be coated with a mold release agent.

Molds of other shapes and materials are also suitable. A regenerator container can be substituted for the mold so that the particles are bound together and to the container. After the binding agent hardens, the monolithic regenerator (i. e., the bound particles) and the regenerator container form a regenerator assembly that can be inserted in a cooling system. Applying the binding agent with the particles molded by the regenerator container instead of a mold reduces the likelihood of fluid flow around the monolithic regenerator.

Regenerator assemblies made by these methods comprise one or more layers of regenerator particles of a RE/TM compound or other compound. The particles may be brittle and irregularly shaped. The layers of particles can be separated by screen separators. In addition, screen separators can be provided on exterior regenerator surfaces through which fluid enters or exits the regenerator during operation. The particles of the layers are bound together by a

binding agent to form a self-supporting regenerator. The binding agent is applied so as to control increases in fluid flow resistance. A 0.5 ym to 20 m coating of a binding agent on particles having effective diameters of 100 ym to 400 zm is generally satisfactory.

Monolithic regenerators formed in a mold are subsequently attached to a regenerator container. This can be done using a mixture of an epoxy and cotton fibers. Such a mixture does not penetrate into the monolithic regenerator, maintaining the fluid flow properties of the regenerator and effectively sealing the interface between the container and the regenerator.

During operation, fluid flows through interstices between the particles of the regenerator and not between the regenerator and the container. Alternatively, the particles and the container can be bound together as the binding agent is applied to the particles, forming a regenerator assembly.

Alternatively, the regenerator particles can be coated with a metal or polymer coating without binding the regenerator particles together. For example, an epoxy mixture can be applied that is too dilute to bind the particles but is sufficient to coat or partially coat the particles. After such coating, the particles can be confined in a regenerator container by screen separators or perforated plates. The screen separators or perforated plates permit fluid flow through the particles; the coating on the regenerator particles inhibits particle fracturing and tends to hold fractured particles together.

The foregoing and other features and advantages of the invention will become more apparent from the following detailed description which proceeds with reference to the accompanying drawings.

Brief Description of the Drawings FIG. 1 is a chart which contains graphs of magnetic entropy and de Gennes factors for trivalent lanthanide ions.

FIG. 2 is a schematic diagram of an arc-melting apparatus for preparing regenerator materials.

FIG. 3 is a schematic diagram of a grinding apparatus for preparing particles of regenerator materials.

FIG. 4 is a graph of volumetric heat capacity as a function of temperature for exemplary RE/TM materials, lead, and helium.

FIGS. 5 are schematic diagrams which illustrate a method for fabricating a single-layer monolithic regenerator. FIG 5(a) illustrates the application of a binding agent to a layer of regenerator particles and the removal of excess binding agent using a gas flow. FIG. 5(b) shows extraction of a single-layer monolithic regenerator from a tube.

FIGS. 6 are schematic diagrams which illustrate a method for fabricating a two-layer monolithic regenerator. FIG 6(a) illustrates the application of a binding agent to first and second layers of regenerator particles and the removal of excess binding agent using a gas flow. FIG.

6(b) shows extraction of a two-layer monolithic regenerator from a mold. FIG. 6(c) illustrates

the insertion of the monolithic regenerator of FIG. 6(b) into a regenerator container. FIG. 6(d) illustrates the sealing of the monolithic regenerator of FIG. 6(b) into the container.

FIGS. 7 are schematic diagrams which illustrate a method for fabricating a two-layer regenerator directly in a regenerator container. FIG. 7(a) shows formation of two layers of regenerator particles in a container. FIG. 7(b) shows the insertion of a plug to capture the layers of particles in the container. FIG. 7(c) illustrates application of a binding agent to the layers and the removal of excess binding agent using a gas flow.

Detailed Description Thermal regenerators using irregularly-shaped particles and methods for fabricating such thermal regenerators are provided. The regenerators are particularly suitable for cooling systems in which brittle particles of a regenerator material are used. The regenerators may be used in conventional regenerative gas-cycle devices such as Stirling, Gifford-McMahon, and Orifice Pulse Tube refrigerators. The regenerators may also be used in magnetic regenerative refrigerators where the application and removal of a magnetic induction causes the heating and cooling required for a refrigeration cycle. Aggregates of regenerator particles in which the particles are bound together to form a self-supporting structure are referred to herein as monolithic regenerators. A regenerator assembly typically comprises an aggregate of regenerator particles (such as a monolithic regenerator) and a regenerator container. The regenerator container facilitates mounting the aggregate of regenerator particles in a cooling system.

A process for fabricating a monolithic regenerator comprises the steps of selecting a regenerator material or materials, fabricating the materials, forming particles of the materials, and binding the particles into a monolithic regenerator. In addition, before the monolithic regenerator can be used in a cooling system, the monolithic regenerator must be mounted in a container for insertion into the cooling system.

Selection of a regenerator material begins by establishing selection criteria for regenerator materials. Materials with large thermal masses within an intended operational temperature range are preferred. At low temperatures, materials with large thermal masses are selected with reference to the following selection criteria. First, such materials have a low Debye temperature 0D Second, the materials have a magnetic ordering temperature in an appropriate range. Third, the materials preferably have a large magnetic entropy. Fourth, the materials should have a high number density of magnetic atoms. Fifth, the materials should contain lanthanide elements from either the beginning or the end of the lanthanide series. Sixth, the materials should have appropriately broad or narrow heat capacity peaks. These selection criteria are explained in more detail below.

Advantageous regenerator materials have low Debye temperatures 0D so that the lattice contribution to heat capacity is significant at low temperatures. In materials having high Debye temperatures eD, the lattice contribution to heat capacity is small and the heat capacity tends to be

low. Heavy lanthanides, i.e., those having atomic numbers greater than 57, such as Er and Yb, have Debye temperatures AD that are generally lower than the Debye temperatures OD of other elements, including the 3d transition metals. With reference to FIG. 2, the volumetric heat capacities C of representative RE/TM compounds Nd3Ni, Er3Ni, ErNb09Co0.l, and Er09Yb0lNi as well as Pb and He are graphed for temperatures below 30 K. Below about 10 K, the heat capacities of these representative RE/TM compounds are larger than the heat capacity of lead.

Thus, the heavy lanthanides are preferred for thermal regenerators for such low temperatures. In addition, the magnetic lanthanides have lower magnetic ordering temperatures than the 3d transition metals.

A material with a large magnetic entropy Sus it exhibits a relatively large magneto- caloric effect and thus provides superior magnetic cooling. Because Smgnenc is proportional to the total angular momentum quantum number J, elements with large values of J (i.e., with large magnetic moments) have large magnetic entropy. The heavy lanthanides generally have relatively large values of the total angular momentum quantum number J and large magnetic moments.

Therefore, the heavy lanthanides have relatively high values of magnetic entropy. While some 3d transition metals have theoretical magnetic moments comparable to the magnetic moments of the lanthanides, the measured magnetic moments are consistently lower than the theoretical values.

The fourth criterion, that the material preferably has a high number density of magnetic atoms, directs that the fraction of the magnetically active component in the material be large. If the lanthanide elements are selected, the materials should consist mostly of lanthanides. A large fraction of a non-magnetic element or an element with a small magnetic moment produces a magnetically dilute material. For non-magnetic regenerators, this consideration is not important.

In addition, the material should not be porous. It will be apparent that a porous material has a reduced number density of the magnetic component as well as a reduced specific heat per unit volume.

The fifth criterion arises because the magnetic ordering temperature tends to follow the De Gennes factor G given by Equation 1: G = J(J+l)(g-1)2 , (1) wherein g is the gyromagnetic ratio given by Equation 2: g = 1 + J(J+1)+S(S+1)-L(L- 1) (2) J(J+1)

With reference to FIG. 1, elements at either end of the lanthanide series have low magnetic ordering temperatures (i.e., low G values) and are thus preferred for magnetic regenerators for low temperatures.

The selection of materials based on the shape of the heat capacity peaks as a function of temperature generally requires consideration of the specific application of the thermal regenerator.

For example, layered hybrid regenerators having more than one regenerator material perform satisfactorily with materials having relatively narrow heat capacity peaks. With such layered regenerators, the heat capacity peaks of the different materials are matched with the anticipated temperature profile of the regenerator. In other applications, a single material is used over an extended temperature range. In this case, a broad heat capacity peak is preferred.

The width and distribution of heat capacity peaks appear to be intrinsic properties of a compound with impurities having small effect on the shape of the peaks. Therefore, regenerator material selection depends primarily on the heat capacities of the regenerator compounds with little consideration of impurities.

For the example embodiments of monolithic magnetic regenerators and methods for making such regenerators described herein, the materials Er3Ni, ErNiO gCo0 l, and ErO gYbo lNi are selected. Because the properties of ErO gYbo lNi and ErNk.9Co0. are similar, magnetic regenerators were built with ErO gYbo lNi and Er3Ni; magnetic regenerators using ErNiO gCo0 are expected to be similar. It will be readily apparent that other regenerator materials can be selected.

The size of particles for a regenerator is chosen to provide large surface areas for heating or cooling while maintaining low resistance to fluid flow through the regenerator. The rate at which particles transfer heat to and from a fluid moving through a regenerator depends on the total particle surface area; smaller particles transfer heat more effectively than larger particles.

Unfortunately, regenerators of small particles have relatively higher fluid flow resistances than larger particles. For the regenerators described herein, particles of effective diameter of 100 ,am to 400 ym are chosen. (An effective diameter is a dimension proportional to the ratio of particle volume to particle surface area.) The selection of particle size can be varied in consideration of required regenerator thermal and fluid flow properties.

After regenerator materials are selected, an arc-melter is used to prepare the regenerator materials. Because most rare-earth materials are commonly available as ingots of fixed masses, the rare-earth ingots must be cut into smaller pieces before alloying with other materials. Pure rare earth metals are difficult to cut and small chips are not readily obtained. Because RE/TM compounds for thermal regenerators contain a large percentage of a rare earth metal, it is convenient to use a mass of a rare earth metal as a starting point and then prepare appropriate masses of other constituents. Generally, RE/TM ingots of mass of 25-100 g are conveniently prepared. Alternatively, a total mass of regenerator material can be specified and corresponding

constituent masses prepared to yield the total mass of regenerator material required. A total regenerator volume can be computed based on porosity and yield values.

With reference to FIG. 2, an arc-melter 200 for preparing regenerator materials comprises a stainless steel vessel 202 which defines a chamber 201 with systems for pumping, purging, and filling the chamber 201 with an inert gas, typically 99.98% pure argon. A water- cooled copper hearth 215 and a non-consumable tungsten electrode 217 are connected to a high- current, low-voltage power supply 219. Regenerator components 221 (e.g., Er, Ni, and Co) are placed in machined pockets 223 in the water-cooled copper hearth 215. The chamber 201 is then purged by repetitively evacuating the chamber 201 with a rough vacuum pump 203 through a valve 209 and then filling the chamber 201 to pressures of up to 10 psi with argon gas from a gas cylinder 205 through a valve 213. Typically five repetitions suffice to minimize oxygen in the chamber 201. After purging, the chamber 201 is evacuated and a small piece of titanium 225 is melted a few seconds before arc-melting the components 221 to remove any remaining oxygen.

An arc is then activated, melting the components 221. The components 221 are alternately melted and flipped ten times to ensure a homogeneous regenerator material.

After the regenerator material is prepared by arc-melting, particles of the material are produced by mechanical crushing. With reference to Fig. 3, a mortar and pestle apparatus 300 for crushing the regenerator material comprises a mortar 301 and a pestle 302. The mortar 301 has a floor 311 and rotates on a shaft 303 in a direction 310 about an axis 305; the shaft 303 is held in place by a bearing plate 309. The pestle 302 rotates on a shaft 304 in a direction 306 about an axis 308. The pestle 302 is provided with a flat 313 to facilitate crushing. The shafts 303, 304 rotate independently and generally have different axes of rotation and rotation rates.

The axes of rotation 305, 308 of the shafts 303, 304 are generally parallel and offset by a distance 320. A grinding end 312 of the pestle 302 is a set a distance 322 from the floor 311 of the mortar 301. The distances 320, 322 are preferably 2.5 mm and 1.0 mm, respectively; the rotation rate of the mortar 301 and pestle 302 are preferably 60 rpm and 150 rpm, respectively.

Pressure 325 is applied to force the pestle 302 toward the mortar 301; the pressure 325 can be controlled manually.

A piece of a regenerator material 324 is placed in the mortar 301. The mortar 301 and the pestle 302 crush the piece of regenerator material 324, producing irregularly shaped particles 307 of varying sizes. The particles 307 are sieved to separate the particles 307 according to particle size. The particles 307 are continuously sieved during crushing. Without continuous sieving, the particles 307 produced tend to be too small, with effective diameters of 20-40 s4m instead of the 100400 ym diameters desired.

The regenerator material 324 is processed in the mortar 301 continuously; adding small amounts in regular intervals allows a higher yield of particles of the intended size. Without continuous sieving, the yield of appropriately sized particles from grinding brittle materials is typically 5-10%. With continuous sieving, the yield increases to 25% or larger. Particle sizes are

characterized by effective particle diameters Dp. The particle size distribution for ErNiO gCo01 is similar to that of Er3Ni and other brittle compounds.

Although particle yields in a given size range are low, mechanical crushing with the mortar and pestle apparatus 300 is both reliable and repeatable. In addition, larger particles can simply be crushed again to increase the yield of smaller sizes. Smaller particles can be re-melted and re-crushed but this is more difficult because fine particles are easily dispersed by the arc of the arc-melter 200. Providing deeper machined pockets in the copper hearth 215 better contains small particles of regenerator material during re-melting. In deeper machined pockets, turning samples or extracting finished ingots would be more difficult. However, with brittle materials such as the RE/TM compounds, the samples can be easily fragmented and extracted from the machined pockets when arc-melting is complete.

After preparing regenerator particles 307 of ErNi09Co0 l with the selected range of effective diameters, the particles 307 are formed into a monolithic regenerator. With reference to FIG. 5(a), the particles 307 are formed into a cylindrical section by a Teflon tube 108 having an outside diameter of 25.4 mm. To ensure ease of extraction, the tube 108 is internally coated with an anti-adhesive substance or a mold release agent. An example of such a mold release agent is the silicon polymer spray 122-S available from 3M Canada, Inc. It will be evident that any other anti-adhesive or mold release agent can be used. It will also be apparent that molds having cross- sections and dimensions different from the circular cross-section of the tube 108 can be used. A screen separator 105 is placed near an end of the tube 108 to seal off the tube 108 and contain the particles 307. The screen separator 105 is preferably one or more 400 mesh stainless steel screens of wire diameter of about 2.5 m. The screen separator 105 is supported by a plug 106 that is also inserted into the tube 108. Particles 307 of the regenerator material with sizes ranging from about 250 m to 300 ym are then placed inside the tube 108.

A binding agent mixture 111 is then applied to the particles 307 in the tube 108. In a preferred embodiment, the binding agent mixture 111 is a mixture of a binding agent (6.426 g of STYCAST 1266 epoxy made by mixing 5.010 + 0.002 g of part A and 1.416 + 0.002 g of part B) diluted with 40 mL of toluene. The viscosity of the STYCAST 1266 is about 650 cps before dilution with the 40 mL of toluene. A valve 113 controls the flow of the binding agent mixture 111 from a reservoir 114 through a manifold 121 to the particles 307. The manifold 121 is retained on the tube 108 by a plug 122.

The binding agent is diluted to a suitable viscosity. If the binding agent mixture 111 is too thin, the binding agent mixture 111 flows through the particles 307 without leaving enough binding agent among the particles to bind the particles 307. If the binding agent mixture 111 is too viscous, the binding agent mixture 111 does not flow between the particles 307 and remains among the particles 307, clogging the spaces between the particles 307. This causes unacceptably large fluid flow resistance in the finished regenerator.

The proper viscosity for any particular binding agent can be determined by applying the binding agent in different viscosities to corresponding test layers of particles. After the binding agent is hardened, the structural integrity and flow resistance of the test layers are assessed, establishing a suitable range of viscosities.

As shown in FIG. 5(a), the binding agent mixture 111 flows through the particles 307; excess binding agent mixture 111 passes through the particles 307 into a container 115. Because the binding agent tends to obstruct fluid flow through a regenerator, excess binding agent mixture 111 is removed. In general, a drop in porosity of 10% from approximately 50% to approximately 40% is detrimental to particle-type regenerator operation. In addition, if a thick coating of the binding agent forms on the regenerator particles 307, heat transfer is impaired. However, adequate binding agent must remain so that the finished regenerator is structurally sound. Excess binding agent also increases longitudinal conduction, i.e. thermal conduction parallel to fluid flow through the regenerator.

Excess binding agent mixture 111 is additionally removed by a pressurized gas flow to the particles 307 that is provided via a gas line 117. The gas pressure of gas line 117 is typically between 20 and 30 psi. A pressure near 20 psi of nitrogen gas is effective with the diluted STYCAST 1266 epoxy mixture. A valve 119 controls the flow of the pressurized gas through the manifold 121. When the valve 119 is open, the gas line 117 applies a pressure to the particles 307 that forces excess binding agent mixture 111 through the screen separator 105 toward the container 115. Alternatively, excess binding agent mixture 111 can be removed by pulling a gas flow through the particles 307 with a suitably-connected vacuum line.

While a single application of the binding agent mixture 111 can be sufficient to bind the particles 307 together, the diluted binding agent mixture 111 is preferably applied and the excess removed five times. After five applications of the binding agent mixture 111, the binding agent mixture is cured for 24 hours at room temperature and then post-cured for 2 hours at 93"C, following the manufacturer's recommended epoxy curing process. After the post-cure, the thickness of resulting coating on the particles 307 is estimated to be between 1 llm and 10 ym on particles having effective diameters of about 100 4m.

The number of applications of the binding agent mixture 111 can be altered by changing the dilution of the binding agent mixture 111. For the STYCAST epoxy, dilution ratios of 3:1 to 9:1 by volume of solvent (toluene) to mixed epoxy resin (part A mixed with part B) are effective.

Mixtures of approximately 5.0 g of STYCAST part A and 1.4 g of STYCAST part B diluted with 20.0, 30.0, and 50.0 mL of toluene successfully bind 200 ym copper test particles substituted for the regenerator particles 307 when excess is removed with pressurized nitrogen gas at 30, 20, and 20 psi, respectively, measured at the valve 119.

The binding agent is not limited to epoxies or other polymers. For example, a metal alloy can be deposited using a vapor which contains a mixture of metals, such as nickel or more preferably, a ductile metal such as lead or tin. After deposition, the deposited metal and the

particles 307 can be sintered together in a low temperature furnace. Alternatively, a metal coating can be applied by immersion in a bath of a low-melting point alloy. Metals used as binding agents should have chemical and magnetic properties that do not interfere with the operation of the regenerator, and have sufficiently low melting temperatures to permit sintering without thermally damaging the particles 307.

It will be understood that solvents other than toluene can be used to dilute the binding agent mixture 111 so long as the solvent is compatible with the binding agent. In addition, the number of applications and the gas flow through the regenerator particles can be adjusted for binding agents other than the STYCAST epoxy. It will be apparent that binding agents that form thin coatings or otherwise do not clog the pores between regenerator particles permit the use of smaller particles having superior heat transfer properties in a regenerator.

After the binding agent is cured, the particles 307 form a monolithic regenerator 129 as shown in FIG. 5(b). The monolithic regenerator 129 is removed from the tube 108 by applying a pressure 133 to the monolithic regenerator 129 with an extracting rod 131. Because the tube 108 is treated with a mold release agent before the particles 307 are introduced into the tube 108, the extracting rod 131 applies only a slight pressure to force the monolithic regenerator 129 from the tube 108. In FIG. 5(b), the screen separator is shown as a part of the monolithic regenerator 129 but the screen separator 105 can be removed. It is useful for the screen separator 105 to be retained to contain any regenerator particles 307 that become unbound.

After removal from the tube 108, the monolithic regenerator 129 can be shaped or cut to a desired size. For example, a diamond-charged wheel can cut the monolithic regenerator 129.

Some particles on the surfaces of the monolithic regenerator 129 can be removed by scraping, but generally the particles remain bound. The monolithic regenerator 129 can be readily cut into a number of smaller monolithic regenerators.

The monolithic regenerator 129 is advantageously glued into place in a cooling system.

Because the monolithic regenerator 129 is porous, it may be necessary to take steps to prevent the glue from invading the pores. An epoxy such as undiluted STYCAST 1266 epoxy could be used as the glue, but the viscosity of that epoxy is insufficient to prevent absorption by the monolithic regenerator 129. Instead of remaining on the surface of the regenerator 129, the epoxy would penetrate into the regenerator 129. Accordingly, before gluing the regenerator 129, 0.5 g of a cotton-fiber filler is added to an undiluted STYCAST epoxy mixture to make the epoxy more viscous and prevent penetration into the regenerator 129. With reference to FIG. 5(b), an approximately 0.5 mm layer of the cotton-fiber epoxy mixture is applied to a perimeter surface 137 of the regenerator 129, without obstructing the ends 135, 136. A second screen separator can be provided at the end 136.

With reference to FIGS. 6, a two-layer monolithic regenerator is made by a method similar to the method of FIGS. 5. In FIG. 6(a), a second layer of regenerator particles 349 of Er3Ni is provided in addition to the first layer of particles 307. After capturing the particles 307

in the tube 108, a second screen separator 145 is placed to confine the particles 307 between the screen separators 105, 145. The second layer of particles 349 is then put into the tube 108. It will be apparent that the particles 349 can be either the same or a different regenerator material than the particles 307. The total masses of the particles 307, 349 are approximately 60 g and 66 g, respectively; these masses correspond to approximate layer thicknesses of about 22 mm and 24 mm, respectively.

Fabrication then proceeds as in the first process. The binding agent mixture 111 is applied and excess removed five times. The binding agent mixture 111 is cured and a finished monolithic regenerator 149 is extracted from the tube 108 as shown in FIG. 6(b).

With reference to FIG. 6(c), a regenerator container 161 is provided that has a seal groove 163 for an application such a the second stage of a Gifford-McMahon refrigerator. A seal assembly (not shown) is generally inserted into the groove 163 and seals the regenerator container 161 into a cooling system. The regenerator container 161 defines a central regenerator chamber 162 and fluid flow channels 165, 167, 169. A layer of lead particles 171 of diameter 300 4m and 1 mm thick is placed at an end of the regenerator chamber 162 and is supported by a screen separator 173 that is in turn held in place with a retaining ring 177. The monolithic regenerator 149 is inserted into the regenerator container 161 such that the particles 349 are nearer the fluid flow channel 169 than the particles 307.

An end plug 179 is provided for the other end of the regenerator container 161. The end plug 179 comprises a plug 181, a retaining ring 182, a screen separator 183, and a perforated plate 184. The end plug 179 also has fluid flow channels 185. With reference to FIG. 6(d), the end plug 179 is inserted into the regenerator container 161 after the monolithic regenerator 149 is inserted and a second layer 187 of 300 I»m lead particles is placed in the regenerator container 161. The end plug is glued into the regenerator container with a cryogenic epoxy such as STYCAST 1266. The monolithic regenerator 149 is attached to the regenerator container 161 with the cotton-fiber/STYCAST 1266 mixture described previously.

The fluid channels 165, 167, 169 permit fluid flow into and out of the regenerator chamber 162; the fluid flow channels 185 on the end plug 179 line up with the fluid flow channels 165, 167. The perimeter surface 137 of the monolithic regenerator 149 must be sealed to the inner surface of the regenerator container 161 to prevent fluid flow between the regenerator container 161 and the monolithic regenerator 149. Such a bypass fluid flow would avoid the monolithic regenerator 149 and would not be effectively cooled by the monolithic regenerator 149.

In another method, a regenerator assembly is fabricated without the tube 108 by molding the regenerator particles directly in a regenerator container such as the regenerator container 161.

With reference to FIG. 7(a), a layer of lead particles 551 of diameter 300 ,am is placed at an end of a regenerator container 561 and is supported by a screen separator 573 and a perforated plate 574 that are in turn held in place with a retaining ring 577. The regenerator container 561 defines a central chamber 562 and fluid channels 565, 567, 569 and a seal groove 563 for sealing the

regenerator container 561 into a cooling system. Layers 507, 549, 551 of ErNi,,Co,,, " Er3Ni, and lead particles, respectively, are placed in the regenerator chamber 562, separated by screen separators 545, 546. The layers 507, 549, 551 have particle masses of about 58 g, 61 g, and 58 g and extend axially along the regenerator container 561 about 25.0 mm, 24.5 mm, and 15.5 mm, respectively. When mounted in a cooling system, the layer 551 of lead particles generally is on the warmer end.

An end plug 579 is inserted to seal an end of the regenerator container 561. The end plug 579 is similar in construction to the end plug 179. With reference to FIG. 7(b), after insertion of the end plug 579, the regenerator container 561 is positioned with the end plug 579 downward. With reference to FIG. 7(c), a binding agent such as the binding agent mixture 111 is then applied as in previous examples. The binding agent mixture 111 is applied through the fluid channel 567 and excess binding agent mixture 111 exits the regenerator container 561 through the fluid channels 565, 566. Excess binding agent mixture 111 is removed by directing a gas flow through a gas line 117. The binding agent mixture 111 is generally applied and excess removed five times before curing as discussed previously.

The layer 551 of lead particles provides an additional advantage during regenerator fabrication. A mechanical load of approximately 100 N can be applied to the end plug 579 while the binding agent is cured. Because lead is ductile, the layer 551 absorbs and distributes the load, preventing fracture of the particles in the layers 507, 549. Moreover, if the particles inadvertently fragment, then the layer 551 contains the fragmented particles, and the fragments do not contaminate other portions of a cooling system. It will be apparent that an additional layer of lead particles can be provided between the end plug 579 and the layer 507 for distributing the load during manufacture and containing particle fragments.

By forming a monolithic regenerator directly in the regenerator container 561, no further cutting or mounting is required and extraction from a mold such as the tube 108 of FIGS. 5(a)- 5(b) is not required. In addition, the binding agent seals the layers 507, 549, 551 to the regenerator container 561 so that fluid flow around the layers 507, 549, 551 is prevented.

It will be apparent that the methods described are suitable for the preparation of other regenerators. The number of layers can be readily varied and the layers can comprise particles of various materials. Screen separators can be provided within the regenerator assembly as needed to separate the layers. Alternatively, the particles can be layered without screen separators.

Regenerators can be formed in a regenerator container or transferred to a regenerator container after the particles of the regenerator are bound. While the regenerators and fabrication methods are particularly useful for irregularly shaped particles of brittle materials such as Gds(Si2Ge2) and many RE/TM compounds, other regenerator materials or materials having regular shapes can be used.

Successful use of monolithic regenerators 129, 149 requires satisfactory mounting in a cooling system. If fluid to be cooled flows around the regenerator and not through the

regenerator, the efficiency of the regenerator is diminished. Forming a monolithic regenerator within a sleeve that is part of a cooling system is therefore especially advantageous.

Superior results are obtained when the binding agent mixture 111 is an epoxy mixture diluted with a solvent such as toluene, but other polymers can also be used. In general, a polymer should be soluble in organic solvents, have a high thermal conductivity, and have a thermal expansion coefficient near that of the particles. The polymer should also have sufficient tensile strength to bind the particles and should adhere to both the particles and containers. The tensile strength of STYCAST epoxy is 20,000 psi. These factors can be traded-off, particularly if the binding agent forms thin layers as applied. With thin layers, thermal expansion and thermal conductivity are of lesser importance. However, the binding agent should withstand thermal cycling.

It will be apparent that the regenerators need not be monolithic and may be made structurally sound by similar coating and confinement methods. By applying a polymer or metal coating to the regenerator particles and then confining the particles in a regenerator container, a regenerator using brittle particles is realized. The coating reduces particle fracturing and holds fractured particles together. Because the particle coating does not bind the particles into monolithic regenerator, the coating can be thin so that fluid flow through the aggregate of particles is not impeded. It will be apparent that the coating need not completely encase the particles.

Coated regenerator particles can be used to form regenerator assemblies. Such regenerator assemblies comprise unbound but coated particles held in a regenerator container. For such regenerator assemblies, a binding agent is used to coat the particles but need not bind the particles into a monolithic regenerator. Screen separators or perforated plates confine the particles while permitting fluid flow. It will be apparent that multi-layer regenerator assemblies can be formed of these unbound, coated particles by layering coated regenerator particles in a regenerator container.

Having illustrated and demonstrated the principles of the invention, it should be apparent to those skilled in the art that the embodiments can be modified in arrangement and detail without departing from such principles. For example, regenerator materials other than RE/TM compounds, such as Gd5(Si2Ge2) and Gd5(SixGelx)4 (x<0.5) or other compounds can be used.

The particles can have various sizes or shapes. The number of layers of regenerator materials can be varied and the layers can be in direct contact or separated with screen separators or perforated plates. In addition, while the regenerators and methods described use irregularly-shaped particles, spherical particles or other regular shapes can be used. Similarly, the regenerator materials need not be brittle.