---------------------------------------------------------------------------------------------------------------

---------------------------------------------------------------------------------------------------------------

AN EXAMINATION OF METAL FELT WICKS FOR

HEAT-PIPE APPLICATIONS*

Douglas R. Adkins, Timothy A. Moss, Charles E. Andraka

Sandia National Laboratories, Albuquerque, NM

Nicos H. Andreas
Bekaert Fibre Technologies, Marietta, GA

Howard M. Cole
Porous Metal Products, Jacksboro, TX

concentrators are very nonuniform with peaks on the order of 100

ABSTRACT W/cm2. The nonuniformity in the flux profiles makes it difficult
Precision metal felts are made of filaments of micron-diameter to deliver the energy by direct illumination of tubes that carry the
wires that are chopped and layered onto a flat surface to form a engine’s working fluid.
wool-like material. Metal felts are commonly used as filters for Heat pipe receivers are currently being developed to transfer
micron-sized particles. The small diameters of the wires and the power from a solar concentrator to the engine (or engines). FIG 1
relatively open structure of these non-woven materials provide illustrates the operation of the receiver. A wick structure
porosities on the order of 90% and greater. The high porosities of distributes sodium across a solar heated dome and heat is
metal felts make them interesting candidate materials for heat removed from the dome as the sodium evaporates. Sodium vapor
pipe applications. Property measurements on selected samples of condenses on the heater tubes of the engine where the energy is
metal felts have demonstrated that typical effective pore radii transferred to the engine’s working fluid (typically helium).
range from 40 to 120 µm and the respective Darcy permeabilities Condensed liquid then flows back to the wick-covered evaporator
range from 30 to 300 µm2. Through careful compaction of these surface under the influence of gravity. The geometry of a receiver
materials, it is possible to tailor the flow characteristics of the is dependent upon the concentrator’s power distribution and a
materials to specific applications. Recently, a series of tests have receiver’s operating characteristics. In the current systems, input
been conducted at Sandia to measure the flow characteristics and powers are as high as 75 kW and receiver domes are nominally
pore structures of these materials as a function of compaction. 16-inch (41-cm) diameter hemispheres.
Results from these tests and a discussion of the current Both screens and powders have been tested as wick structures
applications of these materials in liquid-metal heat pipes for solar
power conversion systems are presented in this paper.

INTRODUCTION
The Department of Energy, in collaboration with several
industrial partners, is sponsoring programs to commercialize
solar-to-electric power generating systems that use parabolic
mirror solar concentrators coupled with a Stirling engine and
generator. Developing a suitable interface that transfers
concentrated solar energy from the focus of the parabolic dish to
the working fluid of the engine has been one of the major tasks in
this program. For optimum performance, each cycle of the engine
needs a balanced heat input, but the flux profiles from affordable
_________________________________________________
*SAND94-0657C, This work was performed at Sandia National
Laboratories, supported by the U.S. Department of Energy
under contract number DE-AC04-94AL85000. Figure 1. Operating schematic of a heat pipe solar receiver.
in previous heat-pipe receiver systems [Andraka, 1993]. Darcy the wire diameter in micrometers, and the second number is the
permeabilities of these earlier wick structures have generally been surface density of the felt in grams/meter2. The porosity, ε, of the
less than 30 µm2 and the effective capillary pore radius has been material is a function of the wick thickness, δ, and it can be
on the order of 60 µm or higher. The low permeabilities have calculated as,
made it necessary to use either thick wicks or extensive artery β δf
ε = 1− = 1− , (1)
networks to compensate for the high pressure drops as liquid ρw δ δ
passes through the wick. These options for compensating for low
permeability are undesirable because they reduce the performance
where β is the surface density, and ρw is the density of the wire
and add to the cost and complexity of the receiver.
Metal felts are now being considered as potential substitute fibers forming the felt. The ratio β /ρw (=δf) is the thickness of
wick materials for solar receivers. Felts that are made of micron- the solid sheet of fiber material (316L) with the same weight per
sized wires can have porosities greater than 90% and, in an square meter as the felt metal. For the 4/150 material, this
unsintered state, the materials are very pliable, so they can be equivalent thickness is 19.2 µm and, for the 8/300 material, the
deformed into dome shapes. It is also possible to modify the pore thickness is 38.2 µm. There are rather large uncertainties in
structure by compacting or expanding the material. measuring the thickness of felt metals; particularly in the case of
Metal felts are most commonly used in filter applications, but unsintered felts. The 4/150 and 8/300 felts were about 1-cm thick
they have been used in heat pipe systems in the past. Phillips when delivered, but the material is easily compressed as it is
[1970] experimented with nickel felts to make low temperature handled. A change of thickness from 1-cm to 1-mm, however,
heat pipes that used either methanol or water as a working fluid. only reduces the porosity of the 8/300 felt from 99.6% to 96.2%.
With the water system, Phillips reported operating flux levels of Felts can be formed by crimping long strands of wire into a
6.5 W/cm2 at a vapor temperature of 90°C. With a potassium heat steel wool-like structure or pressing steel flakes into a mat-like
pipe system, Davis and Ferrell [1974] observed operating flux structure. The Bekipor WB material is formed by laying short
levels of up to 31 W/cm2 using a stainless steel felt wick. In lengths of the micron-sized wires onto a flat sheet. In the plane of
addition to the demonstrated high flux operating levels, it has also the felt surface, the fibers have the random orientation with a
been found that stainless steel felt wicks can operate in liquid nominal spacing on the order of 100 µm (see FIG 2a). A cross-
metal environments for long periods at elevated temperatures. sectional view of the material shows that the fibers are generally
Ewell, Basiulis, and Lamp [1978] noted that there was only a parallel to the surface (see FIG 2b and 2c). Even though the
small amount of grain boundary relief in a felt-metal wick after structure is somewhat organized along the cross-section, there are
2300 hours of operation in sodium at 700°C. After 10,000 hours random voids in the media as FIG 2c illustrates.
of operation, Ewell et al reported increased grain boundary relief The fine fibers of the metal felts can easily break away from
and fracturing where some of the fibers were bent, however, they the material, but the structural integrity improves through
predicted that the wick should easily be able to withstand 50,000 sintering. As felts sinter, the material contracts and felts with finer
hours of operation. wires contract more. The level of shrinkage also increases with
One of the major obstacles to using metal felts as wicks is that increasing temperatures and heating time. Samples of 4/150 that
the properties of the materials are poorly characterized. Property were sintered in a hydrogen oven for 2 hours at 1050°C
data is presented by Phillips and Hinderman [1969] and Freggens contracted roughly 10% in area and the thickness decreased to
[1968], however, both of these earlier studies ignored the impact
of deformation on the properties of the metal felt materials. As
metal felts are applied to a surface or undergo sintering, the
materials will deform. The deformation can result in large
variations in the permeability and the pore structure of the
material. The issue of how properties of felt metals change with
deformation is explored in this paper. It is impossible to give a
complete description of wick properties for every conceivable
situation, however, the concepts presented here will give some (a) (b)
design guidance for metal felt wick materials.

MECHANICAL STRUCTURE OF FELT METALS

Sandia National Laboratories and Porous Metal Products
(c)
focused this study on Bekipor materials that are made by ⇔
Bekaert Fiber Technologies in Belgium. Other manufacturers, 200 µm
such as Technetics Corporation and Memtec America
Corporation which are both located in Deland FL, also produce
felt metal materials. Figure 2. Micrographs of 4/150 metal felt (a) surface view, (b)
Bekipor felts are made of loose metal fibers of 316L cross-section at substrate, (c) cross-section in middle
stainless steel in sheets of a non-woven labyrinth structure. of wick. Sample was sintered in hydrogen oven for 2
Materials that were tested are designated by Bekaert as WB hours at 1050°C.
04/150 and WB 08/300. The first number in this designation is
roughly half of the original height. Samples of 8/300 that were
sintered under the same conditions shrank less than 4% in area
and the thickness was reduced by 30 to 40%. All of the samples
were laying flat during the sintering operation so fibers at the
bottom the felt were subjected to the full weight of the material Mariotte bottle
while fibers toward the top were subjected to less weight. As
might be expected, the wick structure collapses more near the
bottom of the felt (see FIG 2b and 2c), and, therefore, the
porosity will be lower near the bottom after sintering. The overall
porosity, however, changed little during these limited sintering micrometer head
tests. For instance, the volume of the 4/150 material collapsed
30% through sintering at 1150°C, yet the porosity only went from H
99.8% to 99.3%.
pool
PERMEABILITY OF FELT METALS d
For the low Reynolds number flows encountered in heat-pipe
wicks, the pressure drop can be determined using Darcy’s sample
equation
r r
µv r
∇P = − o − ρ a , (2) ri ro
κ
r
where µ is the fluid’s dynamic viscosity (kg/ms), vo is the Figure 3. Wick permeability measurement system.
superficial velocity (m/s), κ is the wick permeability (m2), ρ is the
r r 4/150 felt is roughly equivalent to the permeability of the 8/300
liquid density (kg/m3), and a is the acceleration (-g z for a
felt. The similarity is most likely a result of the similar flow
stationary heat pipe). In this study, the Darcy permeability was
structures for the two felts and the fact that the porosity for both
measured by forcing liquid to flow radially through a wick
materials are roughly equal for thicknesses greater than 0.5 mm.
sample that is sandwiched between two circular plates (see FIG
As a point of comparison, the permeability of wicks in earlier
3). The distance between the two plates is set by three micrometer
heads. A Mariotte bottle was used to maintain a constant pressure heat pipe solar receivers have been on the order of 20 to 30 µm2.
drop between the inlet and discharge of the test fixture. If it is The porosities of these earlier wicks were on the order of 60%
assumed that the flow through the wick is one-dimensional in the compared to the metal felt porosities of 96% and higher.
radial direction, Darcy’s equation can be solved to give the FIG 5 shows that the permeability is extremely sensitive to the
permeability as, level of compaction in the wick. A half-millimeter reduction in
the thickness can cause a 150-µm2 drop in the permeability. This
FI is important to recognize because the material compacts easily

κ=
H K
ν V& ln ro r
i
, (3)
during handling. There are also concerns that the material
thickness could be altered by repeated freeze-thaw cycles in a
2π g Hδ heat pipe. (Tests are now underway at Sandia to determine the
where ν is the kinematic viscosity (m2/s), V& is the volumetric impact of freeze-thaw cycling on felt metals in a sodium heat
flow rate (m3/s), and H is the pressure head drop (m) between the pipe.)
inlet, ri , and outlet, ro , of the wick test area. For thicker wicks, a
correction factor is needed to account for the 2-dimensional
nature of the flow near the entrance of the wick. This factor, 1.4
which is provided in FIG 4, was determined by comparing the
2-D / 1-D PERMEABILITY

results of EQ 3 with a full 2-dimensional solution for the given 1.3

inner and outer radius of the test fixture [Adkins and Dykhuizen,
1993].
Measured permeabilities for felt metal wick samples are 1.2

presented in FIG 5. The results show that the permeability of the

1.1
r = 10 mm
i
r = 30 mm
o
1.0

0 1 2 3 4 5 6
WICK THICKNESS (mm)

Figure 4. Correction for 2-D effects in 1-D permeability

model.
 1000
For felts thicker than 0.5 mm, the tube-pack analogy predicts
900 that permeability of the 4/150 and 8/300 felts are roughly the
FELT METAL WICKS same. Based on this analogy, the permeability of a 1-mm thick
800
4/150 DATA section of 4/150 felt would be 1280 µm2 and the permeability of
700 8/300 DATA 1-mm thick section of 8/300 felt would be 1219 µm2. The
PERMEABILITY ( µm )
2

packed-tube model typically over-predicts the permeability by a

600
factor of 8. This discrepancy between the packed-tube model and
500 measured results is probably an artifact of the model’s inability to
properly account for the tortuous path through the wick. The high
400
porosity means that the metal felt structure is a very open
300 structure but the fine wires interfere with the flow more than an
array of tubes with a similar void fraction.
200

100
PORE DISTRIBUTIONS
0 In the heat-pipe solar receiver, sodium is drawn through the
0 1 2 3 4 wick by capillary forces and the magnitude of these forces is a
THICKNESS ( mm )
function of the pore structure. Metal felts have a somewhat
Figure 5. Permeability of 4/150 and 8/300 felts as a
random pore structure, so it is not straightforward to offer a single
function of compaction.
number as the effective pore radius for capillary pumping
capabilities without a fuller explanation of the origins of the
To correlate permeability with porosity, it is common to treat number.
porous materials as a series of closely packed tubes [Bird, Stewart FIG 6 shows a somewhat idealized representation of the pore
and Lightfoot, 1960]. In this analogy, the pressure drop through a structure through a wick sample. All of the pores in a felt metal
tube of length L is, wick will be interconnected. The meniscus at the liquid-vapor
interface provides the tension force that pumps liquid through the
∆P − 8 µ vo wick. When the pressure drop exceeds the available pumping
= , (4) capability at one point in the wick, the meniscus recedes to a
L Rh2 ε point where the effective pore radius is smaller. In FIG 6, the
receding meniscus would appear as the meniscus moving from
where Rh is the hydraulic radius which is defined as, Point A at the surface to Point B in the wick.
To correlate the effective pore radius with the permeability,
2 x cross − sec tional flow area porosity, and the wire diameter, Brennan and Kroliczek [1979]
Rh = . (5)
offer the expression
wetted perimeter of flow area
e
re = d w + 32 κ ε j
2. (9)
The total cross-section for the flow in a unit volume of the wick The authors did not describe the origin of this expression,
material is simply the porosity. The wetted perimeter of the flow
however, the term 32κε is the square of the diameter of a
area is approximately the total surface area of the wires in a unit
cylindrical tube with equivalent pressure drop characteristics as
volume of the material. For a wire diameter of dw, the hydraulic
the wick, so EQ 9 is a relationship between the effective pore
radius of the wick is therefore,
radius and the effective flow diameter. A section of 4/150 wick
with a thickness of 2 mm and the permeability shown in FIG 5
ε d w δ ρw ε d w δ
Rh = = . (6) would have an effective pore radius of 61 µm according to EQ 9.
2β 2 δf

By comparing Darcy’s equation with the tube-pack analogy

and applying EQ 1, the permeability becomes

ε Rh2 ε3
2
dw
κ= =
b
1 − εg
. (7)
8 2 32

Using EQ1 , the permeability can also be expressed in terms of

the thickness of the wick as

δfFδ − 1I
3 2
G J
dw
κ=
δ Hδ f K32
. (8)
Figure 6. Labyrinth structure of pores.
 One of the most common methods of measuring the effective 100
pore size is to submerge the wick in a wetting liquid and then 90

% OF PORES OF RADIUS R AND LESS

force air to pass through the wick. The effective pore radius is
80
based on the pressure required for bubbles to form on the surface. 4/150
The pore in the wick is idealized as a circular hole and the 70 4/150
99.52%
effective radius of this pore is calculated as 4 mm
98.72%
2σ
60 1.5 mm
re = , (10)
∆Pc 50

where σ is the surface tension and ∆Pc is the pressure drop across 40
the meniscus. The maximum pressure difference the wick can 30 8/300
8/300
support, which is referred to as the air entry pressure, is a 97.74%
99.45% POROSITY
20 1.7 mm
function of the largest interconnected passage through the wick. 7 mm THICK
For the idealized system in FIG 6, the air entry pressure would 10
give the effective pore radius at Point B. The largest pore along
0
the surface of the wick can be estimated by measuring the
pressure when bubbles cease to emerge from the wick. 0 20 40 60 80 100 120 140 160 180
PORE RADIUS (µ m)
FIG 7 shows the maximum effective pore radius for the 4/150
and 8/300 wick materials as a function of the wick thickness.
Figure 8. Pore distribution measured by mercury porosimitry.
These measurements were made by taking four, 25-mm diameter
specimens of each wick material and compressing them to the
radius is determined from the measured pressure using EQ 10.
desired thickness between each pore measurement. The dashed
FIG 8 shows mercury porosimetry data for 04/150 and 08/300
line in FIG 7 illustrates the average of the bubble tests, and the
samples at two different thicknesses. As the samples are
solid lines represent the predicted effective pore radius of the
compressed, the fraction of pores with smaller radii increases as
wicks based on EQ 9 and the measured permeability in FIG 5. As
might be anticipated. At a thickness of 4 mm, 75% of the pores in
the data shows, the maximum effective pore radius generally
the 04/150 felt have a pore radius smaller than 75 µm. By
decreases as the sample is compressed, but there is considerable
compressing the 04/150 felt to 1.5-mm thick, 75 % of the pores
scatter in the data from one sample to the next. This scatter is an
are less than 50 µm in radius.
artifact of the influence that a single passage can have on the
Internal cavities in the porous material can skew the results.
bubble point tests.
Voids in the material, such as the one shown in FIG 2c will not
Mercury porosimetry measures the distribution of pores in a
be filled until a sufficient pressure is reached that forces liquid
wick structure so the influence of a single passage through the
through the passages connected to the void. A void in the wick
wick is reduced. The distribution is determined by measuring the
will make it appear that a larger fraction of pores have a small
volume of mercury that is forced into the material as the pressure
radius. Bubble-point tests will provide a more conservative
is increased. Samples are first evacuated so the trapped gas will
estimate of the effective pore radius, but the results should be
not interfere with the infusion of mercury. To relate the volume
viewed in conjunction with the porosimetry tests to better assess
and pressure measurements to an effective pore size, it is
the degree of conservativeness in the design.
generally assumed that the porous matrix is a series of cylindrical
tubes. As mercury is forced into the matrix under increasing
pressure, successively smaller tubes are filled. The effective pore FELT METALS APPLIED TO SOLAR RECEIVERS
For the thin-walled domes that are used in heat-pipe receivers,
220 the evaporation from the surface of the wick corresponds directly
200
with the flux distribution on the receiver’s absorber surface. The
evaporation rate determines the mass flow rate from every point
180 8/300 FELT METAL
on the wick’s surface, and, with this information and the wick
160 properties, Darcy’s equation can be solved for the pressure
PORE RADIUS (µ m)

140
distribution throughout the wick structure [Adkins, 1988].
Capillary pumping must overcome the maximum pressure drop in
120
the wick to provide liquid sodium to all points.
100 The solid line in FIG 9 shows the relationship between the
4/150 FELT
80 PREDICTED
permeability of a wick and the effective pore radius that is
FOR 8/300 required to distribute sodium across the evaporator surface of a
60 16 inch (41 cm) hemispherical receiver dome. It is assumed in
40 this calculation that the flux profile is uniform across the surface
and the wick is 5-mm thick. The solid line in FIG 9 represents the
20 PREDICTED RADIUS FOR 4/150
pressure drop when the sun is 20 degrees above the horizon and
0 all of the sodium must be drawn from a 13-mm (1/2") deep pool
0.0 1.0 2.0 3.0 4.0 5.0 at the lower edge of the dome. In this orientation, the pore radius
THICKNESS (mm)
Figure 7. Effective pore radius measured by bubble point tests.
must be 93 µm or smaller simply to overcome the hydrostatic 160
pressure drop in the wick. 150 5-mm THICK WICK
13-MM DEEP POOL
The 4/150 felt has a minimum effective pore radius of about 140
20 DEGREE ORIENTATION
50 µm and a permeability of about 150 µm2 when it is 130
compressed to 1-mm thick. Five layers of this material would not 120 DISTRIBUTED PORE STRUCTURE

PORE RADIUS (µm)

be able to cool the absorber surface based on the results shown as 110 WITH REFLUXING
a solid line in FIG 9. There are, however, measures that can be 100
taken to improve the performance of the system. First the depth in 90
the pool can be increased to reduce the hydrostatic pressure drop 80
in the wick. Second, the wick can be thickened near the edges of 70 DISTRIBUTED PORE
the evaporator dome to improve the distribution of sodium. Third, 60 STRUCTURE
condensed liquid can be retured to locations high on the 50
evaporator surface. This lowers both hydrostatic and frictional 40
losses, but the benefits can be lost when refluxing is interrupted 30 BASELINE 75 kW RECEIVER
by cloud transients or changes in the engine’s operating power 20
level. 10
Finally, this receiver analysis may be somewhat conservative 40 60 80 100 120 140 160 180 200
for the felt wick materials. When pressure limits are exceeded, for PERMEABILITY ( µm )
2

instance, the liquid-vapor interface will recede into the wick. For Figure 9. Wick properties required for a 41-cm (16 inch)
the sintered felt materials, the pore structure is naturally more hemispherical solar receiver.
compacted near the substrate so the pores are smaller and
capillary pumping capabilities will be higher .A less conservative Mark Emerson of Vac Met, Inc. in Warren, MI for his help in
approach, and somewhat more realistic representation, would sintering.
recognize that a exceeding the maximum pore radius at a single
point will not cause the receiver to fail. REFERENCES
The dashed lines in FIG 9 indicate the improved performance Adkins, D. R., 1988, "Analysis of Heat Pipe Receivers for Point-
that will be realized when the distributed nature of the pore Focus Solar Concentrators," Sandia National Laboratories,
structure and refluxing are taken into account. For the lower Albuquerque, NM, SAND88-0093.
dashed line, it was assumed that 80% of pore volume near the Andraka, C. E., R. B. Diver, D. R. Adkins, K. S. Rawlinson, P. G.
highest stressed area could dryout before the receiver fails. Since Cordeiro, V. Dudley, and T. A. Moss, 1993, "Testing of
the wick has a distributed pore structure (see FIG 8), 20% of the Stirling Engine Solar Reflux Heat-Pipe Receivers," Proceed-
wick structure in the highest stressed area will still be available to ings of the 28th IECEC, Paper No. 93295, Atlanta, GA.
transport liquid. The upper dashed line in FIG 9 shows the impact Adkins, D.R., and R. C. Dykhuizen, 1993, "Procedures for
that refluxing to the center of the evaporator has on the receiver Measuring Properties of Heat-Pipe Wick Materials," Proceed-
system. With refluxing, two layers of 4/150 metal felt can provide ings of the 28th IECEC , Paper No. 93434, Atlanta, GA.
an adequate supply of liquid to cool the absorber surface at a 75- Bird, R.B., W.E. Stewart, and E.N. Lightfoot, 1960, Transport
kW power level. Phenomena, pp 197, John, Wiley & Sons, Inc., NY.
Brennan, P.J.,and E.J. Kroliczek, 1979, "Heat Pipe Design
SUMMARY AND CONCLUSIONS Handbook: Vol. 1," NTIS Pub. No. N81-70112.
The possibility of using metal felts in a heat-pipe solar receiver Davis, W.R., and J. K. Ferrel, 1974, "Evaporative Heat Transfer
was investigated in this report. Property measurements on the of Liquid Potassium in Porous Media," Proceedings of
AIAA/ASME Thermophysics and Heat Transfer Conference,
material showed that for an effective pore radius of 50 µm,
paper no. 74-719, Boston, MA.
the permeability was on the order of 150 µm2. This is roughly a
Ewell, G. J., A. Basiulis and T. R.Lamp, 1978, "Reliability of
factor of four improvement over wick materials that have been
Low-Cost Liquid Metal Heat Pipes,"Proceedings of the Third
used in earlier receivers. The use of felt materials could greatly
International Heat Pipe Conference, Palo Alto, CA.
improve the performance of heat-pipe solar receivers, but a
Freggens, R.A., 1968, "Experimental Determination of Wick
system to return the refluxed sodium to the evaporator system
Properties for Heat Pipe Applications," Proceedings of the 4th
will still be required to operate at a full design power of 75 kW.
IECEC, Paper 699108, pp 888-897.
This work also showed that the conventional approaches to
Phillips E.C., and J.D. Hinderman, 1969, "Determination of
theoretically determine the permeabilities of these felt materials
Properties of Capillary Media Useful in Heat Pipe Design,"
will overestimate the permeability by about a factor of 8. The
Proceedings of the ASME-AICHe Heat Transfer Conf., Paper
permeability is extremely sensitive to the level of compaction in
No. 69-HT-18, Minneapolis, MN.
the felt material, so care must be exercised in processing the
material to achieve the desired wick properties.

ACKNOWLEDGMENTS
The authors would like to express their thanks to Sharlene
Flachs for her assitance in conducting the permeability tests, and