165

Review Paper

Mechanical seal face materials

B S Nau
BHR Group Limited, Cranfield, Bedford

Abstract: Mechanical seal face materials are described with an explanation of the properties affecting
performance and the significance of surface texture including bi-Gaussian surface statistics. Aspects of
seal behaviour attributable to face materials are discussed in detail, including random fluctuations of
friction and thermal excursions. Boundary lubrication mechanisms of carbongraphites and other ceramics
are described. The role of tribolayers and transfer layers is highlighted. Failure modes are discussed
including structural fracture, surface crazing, pitting and scoring, blistering, solids deposition and `squeal'
(`ringing'). Many references are given.

Keywords: mechanical seals, ceramics, carbon, tribochemistry, failure modes

NOTATION

ac critical flaw size

b radial width of the sealing interface
c specific heat
E Young's modulus
f friction coefficient
G duty parameter Vb=W
H heat flux
Hv Vicker's hardness
k thermal conductivity
Kc fracture toughness
Kw wear coefficient linear face wear rate=PV
m Weibull modulus, an inverse measure of variability
P net specific load on interface Fig. 1 Example of a mechanical seal
RT thermal stress resistance t (1 )=(E)
S shape factor are the most critical part of a seal; the main requirements
V interface sliding speed are tribological. Faces are also exposed to chemical attack
W net load on the sealing interface by the wide range of fluids met in service. Therefore
corrosion is an important consideration in selecting
thermal expansion coefficient
materials, although outside the scope of this review. The
k thermal diffusivity k=(rc)
structural behaviour of seal components also plays a crucial
Poisson's ratio
role in determining performance. This is because the
r density
interfacial lubricating film is extremely thin, the thickness
stress
being typically in the range 0.11.0 m. Therefore any
t tensile strength
deflections at this size of scale directly affect the perform-
ance of the seal. Table 1 lists examples of seal face
1 INTRODUCTION materials with typical values of physical properties.
The sealing interface is between a pair of annular
This review is concerned with materials used for the
sliding faces, now usually based on some form of ceramic
sealing faces of rotary mechanical seals (Fig. 1). The faces
(i.e. non-metallic microcrystalline material). In particular,
The MS was received on 8 January 1997 and was accepted for publication one face is almost always a carbongraphite composite,
on 4 July 1997. often including a metallic or polymeric impregnant. For
J00397 # IMechE 1997 Proc Instn Mech Engrs Vol 211 Part J
166 B S NAU

Table 1 Examples of face material physical properties

Strength
k E Tensile Compressive Hv r k RT
Material (W=m K) (106 K 1
) (GPa) (MPa) (MPa) (kgf =m2 ) (kg=m3 ) (mm2 =s) (K)
Carbonantimony 20 3.5 33 48 280 2300 13 420
Carbonphenolic resin 9 3 23 41 230 900 1800 5.7 590

Aluminium bronze 42 18 130 345 960 7600 150

`Meehanite' 42 10 110 210 720 7200 12 190
`Ni-Resist' 40 19 96 200 150 7300 110

316 steel 16 16 190 620 185 8100 4 200

`Stellite 1' 15 11 250 620 600 8700 220

Alumina, 95% 30 6.9 365 240 3200 1500 3900 10 95

SiC, reaction-bonded 150 4.3 410 249 3000 3100 62 135
SiC, sintered 100 4.8 390 240 10 000 2800 3100 128

WC, cobalt-bonded 105 4.5 650 880 6900 1300 15 000 300

brevity these non-typical carbon-based ceramics will be sive strength much higher than tensile strength, while the
referred to as `carbon', unless the context requires a more seal design usually, but not always, results in radial
specific definition. The seal faces are nominally plane, compression. However, when the higher fluid pressure acts
operating with mixed-film or fluid-film lubrication, and on the inside the ring is in tension and care is needed to
frictional heat is dispersed by conduction through the body avoid tensile failure. A metal-reinforcing band may be
of the seal. The residual waviness is about 0.11 m fitted on the periphery.
peak-to-peak; this is significant as it is comparable to the The brittle nature of ceramics presents quite different
lubricating-film thickness. Working conditions are com- design considerations to those for ductile metals (5). A
monly in the following ranges, but these are by no means flaw larger than a critical size ac grows rapidly, and one
absolute limits: net specific load 0.15 MPa, sliding speed such flaw can result in structural failure. The critical size
120 m=s and sealed-fluid temperature 20300 8C. The increases with the material's fracture toughness K c . Table
chemical nature of the fluid `lubricant' is diverse and can 2 gives indicative values with mild steel included for
be liquid or gas, and may change phase in the interface. It comparison. The value of ac in a region of stress is
is whatever fluid is to be sealedan infinite range of given by
inorganics, hydrocarbons, etc. Such diversity has implica-  
tions for the tribochemistry of boundary lubrication in S Kc 2
a p (1)
mechanical seals.
Unfortunately, much tribo-testing reported in the litera-
ture is carried out under conditions quite different to the where S is a shape factor, approximately unity for a
above and therefore of doubtful relevance to mechanical seal ring. As an example, for a silicon carbide material
seals. Thus in pin-on-disc tribometers the disc sees a the critical flaw size in a stress field of 100 MPa is
periodic transit of the pin but most of the time is 0.3 m.
unloaded and exposed. Both mechanical and tribochem- Notice that tensile strength of ceramics varies more
ical environments are very different to those in a seal and widely than that of metals, being governed by the prob-
care is clearly necessary when interpreting such data. A ability distribution of the size of microscopic (or even
compendium of general studies of ceramic tribology is macroscopic) flaws. Weibull statistics defining the scatter
given in reference (1) and a recent review in reference of tensile strength are used to quantify the batch consis-
(2). The former includes several significant mechanical tency of ceramics (6). The variability of a batch is
seal papers, including those of Paxton and Hulbert (3), indicated by the Weibull parameter m, being greater
which gives insight into seal carbon tribology, and Labus
(4), which contains useful friction data.
Table 2 Examples of fracture toughness K c in
MPa m0:5 at 20 8C
2 SIGNIFICANCE OF PHYSICAL PROPERTIES Material Fracture toughness
Mild steel 140
2.1 Strength Cast iron 620
Tungsten carbide, cobalt-bonded 15
Strength is not usually a major concern since most seal Alumina 4
Silicon carbide 3
rings are ceramic based and most of these have compres-
Proc Instn Mech Engrs Vol 211 Part J J00397 # IMechE 1997
 MECHANICAL SEAL FACE MATERIALS 167

when m is smaller; typically m is about 10 for engineer- tance RT (8C) is the maximum temperature difference
ing ceramics. that can be tolerated without tensile failure. The product
For a specific material quality the probability of occur- kRT is thus a measure of resistance to thermal shock and
rence of a critical flaw increases with the volume of surface crazing and is sometimes approximated by k RT
material, so is more likely in a large ceramic seal than in (7, 8). To complete the picture requires a measure of the
a small one. Also, tensile strength of ceramics is often transient heat flux H (for a transient face rub this is
measured by a beam-bending test where stress varies fPV), while for transient cooling it depends on the heat
linearly through the thickness, from compression through transfer rate to the cold fluid. In general the risk of
zero to tension. Only in the tensile surface is a flaw thermal stress damage increases with j Hj=(kRT ).
exposed to the maximum tensile stress; hence bending
strength is greater than tensile strength (about 31:7).
2.4 Surface energy
Therefore in using tensile strength data for ceramics care
must be exercised to take account of the method of From time to time, surface energy, or wettability, of faces
measurement. has been considered as a factor that might control the
sealing function, by virture of meniscus curvature. The
2.2 Elastic properties surface energy of engineering surfaces is not normally that
of a chemically clean surface as contamination from the
Young's modulus E affects the stiffness of a sealing ring atmosphere commonly includes grease, which reduces
and, in turn, ring-mode face deflections (coning) and wettability (9). However, there is little detailed documented
induced surface waviness. It is generally easier to control evidence for the role of wettability, especially at the
these if the ring stiffness is high, values ranging between elevated temperatures and very high shear rates that occur
20 GPa for resin-impregnated carbon and 650 GPa for between seal faces in practice. Often mechanical seals
tungsten carbide. Hardness H v is important when sealing appear to `seal'. There is no visible liquid leakage, but
abrasive media; values are given in Table 1. vapour phase leakage is measurable with a sensitive
detector. This is explained by a boiling interface some-
where between the inner and outer edges of the seal.
2.3 Thermal properties
Surface energy also plays a role in solids deposition on
Thermal conductivity k has a crucial role in the dispersal of seal faces (10, 11). Figure 2 shows that deposits form on
heat, from the sealing interface through the seal body to the carbon `C' when its wetting energy is high but not when
surrounding fluid. Other things being equal, the higher the low, and similarly with the alumina counterface `A'. If
value of k the lower is the interface temperature and the both faces have high wetting energy then deposits form
less the risk of the interface film boiling and lubrication on the carbon. Such deposits lead to abrasive wear and
failing. Imagine a mechanical seal of 50 mm diameter and premature failure, discussed later.
suppose that 100 W of heat is conducted away through one
of the mating sealing rings. The temperature differential
2.5 Tribological properties
needed to conduct this heat is proportional to k. For resin-
impregnated carbon the temperature differential is calcu- For face material combinations, tabulated values of opera-
lated to be 70 8C; at the other extreme, for reaction-bonded tional properties such as friction coefficient, the factor
silicon carbide, only 5 8C is required. The advantage of the (PV )max and wear coefficient K w are less useful than might
high conductivity of silicon carbide is clear (but its value be expected. This is because mechanical seals operate most
falls significantly with increasing temperature).
The thermal expansion coefficient affects several
aspects of seal behaviour.
1. Coning of the faces and, if thermal expansion varies
circumferentially, face waviness are affected.
2. Relative expansion between a seal ring and its shrink-
fit housing can cause the ring either to loosen or be
overstressed and changing stresses also modify face
alignment and hence seal performance.
3. Thermal stress increases with E, affecting susceptibil-
ity to failure by surface crazing (discussed later) and
structural fracture. Clearly, lower values of are
generally advantageous.
Fig. 2 Solids deposition on the faces of a carbon (C)
Thermal diffusivity k measures the ability to flatten versus alumina (A) seal in relation to face wetting
transient temperature gradients; the higher its value the energy of each face (0 indicates no deposits)
smaller are any transient gradients. Thermal stress resis- [based on references (10) and (11)]
J00397 # IMechE 1997 Proc Instn Mech Engrs Vol 211 Part J
168 B S NAU

of the time with full-film or mixed-film lubrication of the instance, 316 steel has a thermal conductivity of only 16
sealing faces. The parameters therefore vary with the many W=mK (at 20 8C)about half that of alumina.
complex factors that determine hydrodynamic pressures in
the interface. Values also vary between seals because 3.2 Typical ceramics
interactions between design and materials change face
deformations. A value of (PV )max is sometimes quoted for Ceramics are composites (Figs 3 and 4) and most are based
a specific seal design but, of course, this can only refer to a on oxides, carbides or nitridescarbongraphite is a
specified fluid at a specified temperature. It is not a general notable exception. Seal face grades of ceramics will be
constant, even for this seal. Table 3 lists some values of this classified here, somewhat arbitrarily, as `typical' and
parameter to give a general indication of the variation `untypical'. In typical ceramics there is normally only a
between materials. trace of binder and the porosity is inherently low. Reaction-
When seal faces do make contact then tribological bonded silicon carbides contain a relatively large amount
characteristics of the face material combination determine of silicon binder, about 10 per cent, but it is convenient
survival or failure of the seal. In this situation the here to include this with the typical ceramics. `Untypical'
operational parameters f, (PV )max and K w may be sig- ceramics, carbongraphites and cermets, normally contain
nificant but, as will be seen later, even then instability of a significant amount of metal or hydrocarbonresin. These
the boundary lubrication mechanism can be overriding. act as a binder and reduce porosity; the proportion is
typically about 10 per cent but sometimes it is much more.
Corrosion resistance depends on the nature and quantity of
the binder or impregnant, the ceramic phase being very
3 FACE MATERIAL CHARACTERISTICS resistant. Table 4 illustrates corrosion rates.
The `typical' ceramics used in mechanical seals are
For convenience the materials commonly used as mechani- mainly grades of alumina or silicon carbide, both notable
cal seal running faces will be grouped under the following for low density and silicon carbide for its high thermal
headings. conductivity (Table 1). Silicon nitride has been used in
roller bearings in recent years, but the ability to with-
stand localized high stresses is not a major consideration
3.1 Metal alloys
in mechanical seals so it does not find a place in the
The following metal alloys are often used as low-cost seal present list. Chromium oxide has been used as a hard
faces, almost invariably running against a carbon: coating on seal faces, as have various other ceramics and
chromium plating, but these are not widely used. Further
`Meehanite' cast iron
background information on ceramics is given in refer-
`Ni-Resist' cast iron
ences (12) and (13).
Aluminiumbronze
Alumina (Al2 O3 ) seal face materials are sintered poly-
Stainless steel
crystalline materials ranging in purity between 95 and
`Stellite' (CoCrW alloys)
99.5 per cent. The grains are 15 m in size, variably
The principal failure mechanism for these is tribological, separated by a glassy phase (Fig. 3a). Properties are very
normally associated with loss of fluid-film lubrication dependent on microstructure; this varies with powder
due to operational or design factors. The cast irons and purity, firing temperature and duration and therefore
bronze tend to fail progressively rather than suddenlya varies between grades and batches, and sometimes within
useful feature. The low thermal conductivity of stainless batches (14, 15).
steels accounts for their limited use as face materials. For Silicon carbide (SiC) forms a particularly important
group of materials for mechanical seals. They are poly-
crystalline materials with SiC crystallites 280 m in
size, i.e. large compared with the interfacial fluid film in
Table 3 Approximate PV limit values most mechanical seals. The size distribution may be
(MPa m=s) for water at 40 8C unimodal or bimodal and the crystallite aspect ratio can
(8) vary, e.g. 1:15:1, the SiC may be an or crystal
Face Counterface (PV )max
SiC Carbon 20 Table 4 Examples of ceramic corrosion rates (g=m2 yr)
SiC SiC 20
WC Resincarbon 8 Sintered SiC WC 6% Co Alumina, 99%
WC WC 4
`Stellite' Metalcarbon 3 Sulphuric acid, 98% 20 10 000 650
`Ni-Resist' Carbon 3 Hydrochloric acid, 25% ,2 900 700
Alumina Carbon 3 Sodium hydroxide, 50% 3 50 750
Leadbronze Metalcarbon 2 
Based on data in: Corrosion=erosion resistant components for the
Stainless steel Metalcarbon 1 chemical processing industry, Carborundum Company, 1983.

Proc Instn Mech Engrs Vol 211 Part J J00397 # IMechE 1997
 MECHANICAL SEAL FACE MATERIALS 169

(a)
(b) 10 m

(c) 20 m

Fig. 3 Schematics of the structure of `typical' ceramics: (a) alumina, (b) sintered SiC, (c) reaction-
bonded SiC

phase and free carbon is present in some grades. Physical graphite ring with silicon vapour to form a surface layer
properties and performance vary significantly between of SiC, ca. 2 mm thick. This is a lower cost alternative
grades and, as with alumina, between batches and even to the preceding, but less robust, so is not recommended
within batches (14, 16, 17). There are two important for abrasive duties.
groups of SiC materials used in mechanical seals: GraphiteSiC is made from carbongraphite infiltrated
with liquid silicon to give homogeneous SiC with 3050
1. Sintered SiC is made from -SiC powder with a trace
per cent free graphite, to improve boundary lubrication.
of binder (, 2% boron or aluminium), sintered at ca.
This material is not widely used.
2000 8C without pressurization. It is widely used in
mechanical seals, especially for corrosive duties. Fig-
ure 3b illustrates the structure. 3.3 Untypical ceramics
2. Reaction-bonded SiC is made from -SiC powder and
`Untypical' ceramics comprise carbongraphites on the
graphite, infiltrated with silicon liquid or vapour to
one hand and cermets on the other, discussed separately
form a bonded composite of -SiC, -SiC and silicon.
below.
It is widely used in mechanical seals, having particu-
larly good tribological properties, including a high
thermal conductivity. The silicon binder is vulnerable 3.3.1 Carbongraphites
to chemical attack, limiting suitability for use with
The many commercial grades of carbongraphites (`car-
aggressive fluids. Figure 3c illustrates the structure.
bon') are the most widely used materials for mechanical
Siliconized carbon is made by infiltrating a carbon seal faces. Traditionally they are not regarded as ceramics
J00397 # IMechE 1997 Proc Instn Mech Engrs Vol 211 Part J
170 B S NAU

although this is what they are. They are not `typical' material and binder, e.g. synthetic or natural graphite,
ceramics since they nearly always contain a significant detailed structure, nature of trace contaminants, etc. They
proportion of resinous or metallic impregnant, to reduce also depend on details of the manufacturing process.
porosity and enhance other properties (Fig. 4a). The These factors make it difficult to get truly identical
proportion of impregnant is typically 515 per cent, but in materials from different suppliers. There are many stan-
grades designed for moulding in mass production, resin is a dard specifications for measurement of carbongraphite
much higher proportion. Normally it is the impregnant that properties [e.g. see reference (19)] but experience sug-
is degraded when exposed to aggressive fluids or high gests that control of such properties is not necessarily
temperatures. sufficient to control tribological performance.
The basic constituent is grains composed of graphite Phenolic resin impregnation is widely used for light
crystallites, the latter being 0.0050.100 m in size, i.e. duties but antimony impregnation is preferred for its
small compared with typical interfacial film thicknesses higher conductivity in more demanding applications, at
in mechanical seals. The binder is carbonized material higher speeds and pressures for instance. Grades impreg-
derived from such precursors as coal-tar pitch, petroleum nated with other metals (e.g. silver, copper, babbitt) are
pitch, resin, etc., and may be amorphous or graphitic in sometimes used but are susceptible to chemical attack.
varying degrees. There can also be additives including Non-toxic grades for food processing may have polyester
processing aids, antioxidants and boundary lubrication resin impregnation. For high-speed service, e.g. gas tur-
promoters. Paxton (18) describes the complexities of the bine engines, a graphitized fine-grained grade with anti-
manufacturing process. oxidant incorporated may be used.
The properties of carbons are very dependent on the Porosity, hardness and thermal conductivity vary
nature and sources of the raw materials used for base widely between grades and, as with corrosion resistance,
temperature limits are determined by the properties of the
filler and less often by oxidation of carbon. Connected-
cell porosity is monitored during manufacture; it deter-
Crystallite grains mines fluid permeation through the seal ring, though this
is rarely a problem in service. Closed-cell porosity in
excess can lead to serious delamination failure (20).
Binder
Principal failure mechanisms of carbons are tribologi-
cal failure and blistering (discussed later).
Impregnant

3.3.2 Cermets
Porosity
Cermets came to the fore in the 1960s and 1970s but are
now supplanted by silicon carbide. Cermets contain
ceramic crystallites bound by a metallic binder, and
crystallite size is typically about 1 m (Fig. 4b). The
ceramic is usually tungsten carbide (WC) but titanium
(a)
carbide has also been used, being more corrosion resistant,
although expensive. The usual binder is cobalt (e.g. 715
per cent) or nickel (e.g. 6 per cent). Cermets are
manufactured by pressing and presintering; then conven-
tional machining is possible. After final sintering any
finishing is carried out by diamond, SiC or BC grinding or
lapping.
As usual, physical properties vary significantly be-
tween grades, which can differ in grain size and binder
type and amount. The principal failure mechanisms are
surface crazing, tribological, or matrix corrosionagain,
any corrosion normally attacks the metal binder. Abrasion
resistance increases with fracture toughness, microhard-
ness and smaller grain size (21).

(b) 10 m 4 FACE COMBINATIONS

Fig. 4 Schematics of the structure of `untypical' cera- So far attention has focused on face materials individually.
mics: (a) carbongraphite, (b) cermet However, in service faces are used in mating pairs. This
Proc Instn Mech Engrs Vol 211 Part J J00397 # IMechE 1997
 MECHANICAL SEAL FACE MATERIALS 171

can be illustrated by some representative examples in hydrodynamic lubrication under full-film and mixed-film
common use. conditions. The residual surface waviness of lapped car-
For light duties (about 5 m=s, 5 bar pressure, say) the bons is commonly modified by exposure to liquid or
first choice, carbon, is usually a resin-impregnated grade changing temperature (8), and further changes occur dur-
running against one of several light-duty counterfaces. ing running (Table 5).
For example, a `Meehanite' cast iron counterface could Roughness refers to a similar height variation but with
be used for oil, aluminiumbronze for water or `Ni- a horizontal scale of only 10100 m and a ratio of
Resist' cast iron for mildly corrosive fluids. An alterna- 10 103 (Fig. 6). Roughness is not normally an issue
tive to the latter is 304 stainless steel, although its unless the seal operates in a mixed-film mode. The
thermal conductivity is too low for really satisfactory roughness of a lapped unused ceramic seal face does not
performance. `Stellite 1' is an option if the fluid is also have the simple Normal height distribution commonly
abrasive, but alumina (95 per cent pure) is hard and assumed when applying the average-flow form of Rey-
reasonably resistant to corrosion so is a more economic nolds equation to rough surfaces [e.g. see references (8)
choice. Where reliability is important a combination from and (23)]. In reality, the height distribution is extremely
a higher duty category would be considered. skewed due to removal of peaks by lapping (Fig. 6).
For higher dutieshigher speeds, pressures, tempera- Leefe and Williams (24) show that the profile is domi-
turesthe choice of carbon is commonly antimony-im-
pregnated and the counterface would usually be silicon
Table 5 Development of carbon waviness in service (22)
carbide. For the latter a reaction-bonded grade with high
(resincarbon versus Ni-Resist, 2900 r=min, 70
thermal conductivity is advantageous or sintered grades
8C, distilled water, 6 seals)
with corrosive attack are a possibility.
For abrasive duties both faces need to be hard materi- Time (h) 0 0.1 0.25 1 24
Wave height (m p.t.p.) , 0.2 1.644 35 1.56 528
als. Two practical options are silicon carbide against
tungsten carbide or silicon carbide against itself. If the
duty is also corrosive, only sintered SiC against either
itself or perhaps WC Ni is likely to give a reasonable
life. For the most corrosive duties, without abrasives,
an option is glass-filled PTFE (polytetrafluoroethylene)
against 99.5 per cent pure alumina.

5 SURFACE CHARACTERISTICS

5.1 Surface topography

Circumferential surface waviness and surface roughness
are characteristic of both the material and the finishing
process and will be considered in turn.
Waviness in this context refers to surface height varia-
tions extending over 10100 mm circumferentially, with
amplitudes of 0.11 m peak-to-peak (p.t.p.) and a scale
ratio of 104 106 (horizontalvertical) (Fig. 5). Since the
1960s waviness has been known to play a key role in

Fig. 6 Examples of seal face roughness: (a) carbon

Fig. 5 Example of residual waviness of the seal face graphite, (b) WC, (c) alumina (22)
J00397 # IMechE 1997 Proc Instn Mech Engrs Vol 211 Part J
172 B S NAU

nated by two roughness scales: a large-amplitude peak- Secondly, lapping is a tribological process in its own right
lopped component and a small-amplitude `sinusoidal' and may therefore generate a tribolayer on the lapped
component. It appears that the first characterizes the surface. Clearly the nature of the lapping fluid used could
method of finishing and the second the grain structure of be significant in this respect.
the material. Undoubtedly there are finer-scale compo-
nents as yet unstudied. The dominant components have
interesting properties revealed by plotting the percentage 6 CERAMICS TRIBOLOGY
of surface below a given height versus height. This shows
a bilinear relationship resembling a Weibull plot with two
6.1 Carbongraphites
failure modes, i.e. the two components of height distribu-
tion have separate Gaussian distributions, so the combina- Often it is not appreciated that the wear and friction of
tion is `bi-Gaussian' (Fig. 7). Both unused and used carbons can be very high in an inert environment, such as
ceramic faces exhibit this characteristic (25). The ability an inert fluid or a vacuum. This became apparent in World
to characterize surface profiles in this way provides a War II when carbon brushes on electrical generators in
valuable tool for gaining insight into the relationship high-flying aircraft suffered high wear rates. Savage (31
between surface finishing, material structure and seal 33) showed that this is the normal behaviour of chemically
performance. Although little used as yet, Leefe has clean graphite. Only in the presence of `contaminant'
applied it in a computer model of the elastohydrody- molecules (e.g. H2 O, oxygen and various hydrocarbon and
namic lubrication of real seal faces in the mixed-film inorganic species) does `carbon' exhibit low-friction, low-
mode (25). wear behaviour. Boundary lubrication of carbongraphite
Other surface topography parameters have been investi- involves a rather complex mechanism. Initially, hyper-fine
gated experimentally (2629) from which the radius of particles of comminuted carbon crystallites are generated
curvature of asperities appears to play a key role in seal by wear of the unprotected carbon. Due to its large
tribology. Friction and wear decrease with increasing surfacevolume ratio the particulate behaves like activated
radius and the leak rate increases; lapping scores, charcoal, although orders of magnitude more reactive. In
although extremely fine, also affect the leak rate. the presence of suitable `contaminant' molecules it links up
to form a surface coating of submicrometre thickness (Fig.
8). So long as this `tribolayer' remains intact, friction and
5.2 Effects of lapping
wear remain low and stable. The existence of a tribolayer
Apart from the effects of lapping on the surface topography, was subsequently noted in mechanical seals (34, 35).
attention should be drawn to two other effects of lapping Figure 9 shows the relative dimensions of carbongraphite
which deserve further study. Firstly, lapping can result in constituents, from atomic spacing to grain size in manufac-
faces becoming impregnated with particles of the lapping tured seal carbons.
compound (30), which could affect seal performance. Savage's findings were also confirmed by Lancaster

5
Scaled bearing fraction

4
5 0 3
Height above datum (m)

Fig. 7 Bi-Gaussian surface-profile height distribution of a lapped carbongraphite seal face (24). (The
bearing fraction is related to the proportion of `solid' along a transect at a given height above
the datum; a normal probability scale is used here)
Proc Instn Mech Engrs Vol 211 Part J J00397 # IMechE 1997
 MECHANICAL SEAL FACE MATERIALS 173

Tribolayer
(a)

Fig. 8 Schematic of a carbongraphite surface with a

tribolayer
(b)
Mechanical
seal
interface Manufactured
film
carbon
Grains (crystallite clusters)
Carbongraphite crystallites

Plane-to-plane
C-hexagon Graphite
crystal
CC
Fig. 10 Transition to high-friction, high-wear-rate for
0.1 nm 1 10 100 1 m 10 100 carbon on carbon [based on reference (37)]
Size

Fig. 9 Size scales of constituents of manufactured seal

carbongraphites

(3642) using electron microscopy and other techniques

during pin-on-ring studies of carbon tribology for aircraft
brakes. This work also showed that intermolecular bonds
holding the carbon film together may thermally desorb at
150200 8C. The protective tribolayer then disintegrates,
triggered by a high sliding speed (Fig. 10) or high load,
as well as high ambient temperatures, resulting in in-
stability or total failure. A similar behaviour could be
expected in mechanical seals.
In fact, studies of carbon tribology in mechanical seals Fig. 11 Relict tribolayer on a carbongraphite seal face
in the author's laboratory show that the situation is even
more complex than described above (4345). It appears
that `good' seal carbons build up the protective tribolayer appears to be of similar origin and nature to the tribo-
rather slowly. Eventually, however, the tribolayer reaches layer on the carbon face. In mechanical seals the transfer
a thickness at which it is mechanically unstable and it layer usually builds up more slowly than the tribolayer on
then breaks up (Fig. 11). A `good' seal carbon then the carbon and therefore tends to be more stable and so
recovers, even if the transient temperature during tribo- have less impact on seal performance.
layer break-up temporarily exceeds the desorption temp- Recent publications on carbon tribology are listed in
erature. It repeats the entire sequence in a `cyclic reference (46).
equilibrium' and the seal operates satisfactorily. On the
other hand, when the tribolayer of a poor carbon breaks
6.2 Tribology of typical ceramics
up, the interface temperature rapidly rises well beyond
the desorption temperature and total seal failure with The tribology of `typical' ceramics has been studied
severe abrasive wear ensues. Midgely and Teer (35) extensively in recent years, particularly as bearing materials
reported similar cyclic behaviour in a mechanical seal for [e.g. see references (1), (4) and (47) to (51)], showing that
a nuclear gas circulator. these too develop tribolayers and transfer layers. There are
Turning to the counterface, a `transfer layer' of car- significant differences in detail. For example, an alumina
bongraphite material forms on this surface (Fig. 12). It running face becomes hydrated in the presence of water
J00397 # IMechE 1997 Proc Instn Mech Engrs Vol 211 Part J
174 B S NAU

Figure 14 summarizes the effects of various `contaminant'

molecules on the friction and wear of SiC and alumina
sliding against themselves. The labelled data points
indicate the ambient atmosphere and show how friction and
wear rate vary with this. For example, both friction and
wear are high in dry nitrogen (a and A) and much lower in
benzene vapour (d and D), the effect varying somewhat
depending on the material combination.
When there are competing boundary lubrication me-
chanisms, as when carbon runs against SiC for example,
it appears that the softer (carbon) material dominates. It
is a common experience to find the carbon face condition
much modified while the hard counterface is still in an
`as-lapped' condition apart from modest carbon transfer.
From the above discussion it is evident that mechanical
seal performance depends on the sealed fluid, contami-
nants and ambient atmosphere. In real world applications
there may be quite diverse, and sometimes unexpected,
substances present in trace amounts, but in amounts
sufficient to influence seal face tribochemistry.
The work of Fischer and co-workers shows that water-
lubricated silicon carbide in a pin-on-disc tribo-
meter can exhibit dramatically low friction for a rather
different reasonbehaviour it shares with silicon nitride
Fig. 12 Carbon transfer layers on hard counterfaces: (a)
(4749). It appears that this is due to the tribolayer (and=or
dense layer on WC, where low-angle illumination
transfer layer) developing a near-flat surface profile and so
highlights lapping scores; (b) sparse layer on
providing a powerful hydrodynamic effect (4749). Similar
alumina
effects have been observed in a mechanical seal geometry
(Fig. 15). Examples of the high load capacities of which
molecules and forms an hydroxide tribolayer (Fig. 13). As various seal material combinations are capable are listed in
with carbon, both alumina and silicon carbide exhibit high Table 6. With the exception of alumina these increase with
friction and wear in inert environments, and the presence of Young's modulus of the higher modulus material of the
atmospheric oxygen and=or water molecules enhances combination.
performance. Also, vapour of such hydrocarbons as ethanol
and benzene reduce friction and wear very effectively.
6.3 Tribology of cermets
The tribology of cermets has been studied much less than
Lapped surface
typical ceramics or carbons. In principle, either the tri-
bolayer or transfer layer mechanism described above or
(a) normal metal boundary lubrication considerations could
apply to these materials, but more work is needed to clarify
the boundary lubrication mechanisms of cermets in mech-
anical seals.
Debris in pores
6.4 Seal behaviour
(b) Mechanical seal friction might be expected to follow a
Stribeck-type curve. Figure 16b shows an example where
this is indeed the case. The face materials are SiC versus
Debrishydroxide
antimony-impregnated carbon at 6 m=s, in distilled water at
1 bar. The transition between fluid-film and mixed-film
(c) regimes is well defined and data scatter is slight. Figure
16a shows the corresponding transition in seal face temp-
erature while Fig. 17 shows the wear-rate transition at a low
Fig. 13 Development of a tribolayer on alumina in the duty parameter for a commercial seal.
presence of H2 O by tribochemical formation of The well-behaved friction data in Fig. 16 is the ex-
hydroxide [based on reference (51)] ception, as the following examples illustrate. Figure 18a
Proc Instn Mech Engrs Vol 211 Part J J00397 # IMechE 1997
 MECHANICAL SEAL FACE MATERIALS 175

1.0
Pin-on-disc:
A...I alumina versus alumina
a...i SiC versus SiC (sintered)

Friction coefficient

0
1012 Specific wear rate (mm/s mm) 105

a, Anitrogen e, Eair
b, B water vapour f, F 50 % relative humidity
c, C ethanol vapour g, G 100 % relative humidity
d, D benzene vapour h, Hwater, 0.1 m/s
i, Iwater, 0.8 m/s

Fig. 14 Measured friction and wear of SiC and alumina sliding against themselves in the presence of
various `contaminants' [data from reference (50)]

200
Face temperature (C)

Fail

100

20
0 5 10
Specific load (MPa)

Fig. 15 High load capacity of seal faces (44), apparently due to fluid film hydrodynamics in
conjunction with the ultra-flat tribolayer

shows data for a commercial seal with waterglycerol output measuring seal housing reaction; it was supported in
solutions at 120 bar pressure, 105000 r=min of Ni- bearings. Figure 18b shows random thermal excursions of
Resist versus resincarbon (52). The wide random scatter another commercial seal in a different test rig; the faces
of the friction is characteristic of many similar tests on this were Ni-Resist versus phenolic-impregnated carbon and the
and other commercial seals, with a variety of face material fluid was distilled water (22). The excursions are relatively
combinations. The scatter is upwards from an `envelope' infrequent, but random. Figure 18c is yet another seal
resembling a Stribeck curve and approaches an upper limit configuration in another test rig. The faces are reaction-
corresponding to boundary lubrication. The seal was well bonded SiC versus phenolic-resin-impregnated carbon and
away from the thermal instability due to vaporization of the data were logged at a relatively high rate by computer (45).
fluid film. Data were read periodically from the transducer The 4:1 fluctuations in friction are too rapid to affect
J00397 # IMechE 1997 Proc Instn Mech Engrs Vol 211 Part J
176 B S NAU

Table 6 Specific net transition load (P, MPa) and Young's supply frequency and the 42 Hz peak may be a resonance
modulus (E, GPa) of the higher modulus materi- of the transducer installation. It is not obvious why the
al, for plane annular rings in distilled water at 6 friction should fluctuate synchronously with the shaft.
m=s (44) Fluctuations of friction in mechanical seals have been
Material combination P E observed by numerous workers [e.g. see references (53)
to (56)] and a similar effect has been seen with dry-
WC versus SiC 14 580
WC versus antimonycarbon 11.513.5 580
sliding alumina in a four-ball tester. This was suppressed
SiC versus antimonycarbon 5.78.7 410 by contaminants (57).
SiC versus resincarbon 4.27.3 410 From the available evidence there can be little doubt
Ni-Resist versus antimonycarbon 2.2 110
Ni-Resist versus resincarbon 2.2 110
that the random fluctuation of friction in mechanical
Alumina versus resincarbon 1.2 350 seals is a real effect. The similarity of fluctuations during
wet and dry operation eliminates fluid-film instability as
an explanation. It seems likely that the effect is asso-
ciated with the tribolayer (or transfer layer), and this is
carbon temperature, due to thermal inertia; the interface supported by the condition of faces observed in tests
continued in cyclic equilibrium for 90 hours until stopped. halted in different operating regimes. Seals stopped and
Figure 18d shows behaviour during a continuation of the dismantled when operating at above-normal face tempera-
preceding test after draining the water, with the seal ture (e.g. 140 8C) have been found to have flake-like
running dry. The friction still fluctuates and the lower remnants of a thick tribolayer on the carbon face (Fig.
envelope drifts up; the face temperature built up until the 11). The very high load capacities sometimes observed
run was abandoned at 250 8C. Some carbons do not suffer may also be attributable to an intact tribolayer (and=or
this thermal run-away and are able to run dry in cyclic transfer layer). However, plausible explanations have yet
equilibrium for 24 hours or more. Figure 19 is a fast to be offered for two aspects of carbon behaviour in
Fourier transform spectral analysis of the fluctuating seals. How can a tribolayer cause friction fluctuations
friction during dry running in a similar test, (a) during when the seal is operating in the hydrodynamic region of
stable operation and (b) during a friction excursion. The the Stribeck curve? Secondly, in tests under carefully
main peak is shaft rotation (70 Hz), 50 Hz is the electrical controlled conditions, carbon performance exhibits sto-

120
Seal face temperature (C)

(a)

20
1.0
(b)
Friction coefficient

.1

Slope 0.5
.01

.001

.0001
1010 109 108 107 106 105
Duty parameter

Fig. 16 Performance of plane annular test samples of SiC versus antimony-impregnated carbon in
distilled water: (a) face temperature versus duty parameter; (b) friction versus duty parameter;
[redrawn from reference (44)]
Proc Instn Mech Engrs Vol 211 Part J J00397 # IMechE 1997
 MECHANICAL SEAL FACE MATERIALS 177

20

Specific wear rate (m/104 cm)

0
109 Duty parameter G 104

Fig. 17 Dependence of specific wear rate on duty parameter for a commercial mechanical seal
[redrawn from reference (52)]

chastic behaviour wherein even carbons from one batch and propagate into the interior and the pads curl up at the
can vary widely in time-to-failure under nominally iden- edges, like drying mud in a puddle. This forms an
tical conditions. What is the underlying mechanism? extremely abrasive surface which abrades the counterface
and severe leakage soon develops.
Surface crazing occurs with all types of ceramic. The
7 FAILURE MODES cause is generally believed to be transient thermal stres-
ses in the surface, induced either by sudden cooling or
by sudden heating of the face. Sudden cooling can result
7.1 Fracture
from an influx of cold liquid to the sealing interface and
After tribolayer (or transfer layer) instability the most heating can result from a brief face rub. The resulting
common material failures of typical ceramics are brittle differential thermal expansion of the surface relative to
fractures. At its simplest this can be structural, originating the substrate generates subsurface shear stresses. These
at a stress raiser such as a drive slot. A less obvious cause stresses are sufficient to cause the material to fail in
has been reported in carbon rings where the crack compression or tension, depending on whether it is
originates on the sealing face (5860). This is thought to heated or cooled. Several factors affect crazing:
be due to mismatch of thermal expansion of fillers.
1. Sufficiently high temperature gradients are only likely
Alumina fractures have been attributed to straightforward
to arise under transient conditions. Therefore a high
thermal shock (14).
thermal diffusivity of the material is beneficial in
If both seal faces are hard ceramic and the seal
suppressing surface crazing.
operates continuously with its interfacial film temperature
2. The ability of the material to resist such stresses is
close to boiling, damage arises in a different fashion.
indicated by the thermal stress resistance (see Section
Fluid-phase instability causes the faces to separate peri-
2.3 and Table 1).
odically (at one side), as the fluid film vaporizes and
3. A higher dry-sliding friction coefficient is more likely
expands. The faces then crash together as hydrodynamic
to cause crazing in the event of a rub as the heat input
and hydrostatic load support is lost. The resulting impacts
will be greater.
can cause extensive structural damage to brittle ceramics
(Fig. 20).
7.3 Face pitting and scoring
7.2 Surface crazing
A quite different and very localized form of surface
In less extreme circumstances fracture is commonly fracture occurs in all types of ceramics but is particularly
localized in the surface layer of the running face, forming a common in carbons. This is characterized by formation of
network of shallow hair-line cracks, `surface crazing' (or small pits in the surface. Material detached from the pit is
`heat checking'). The cracks are either more-or-less dragged away downstream and an arc-shaped score mark
uniformly spaced radial cracks or form a `crazy-paving' forms, fading as the material is ground up (Fig. 22). Such
network of `pads' of rather uniform size (Fig. 21a, b). pits are commonly about 0.1 mm in depth and diameter.
During the early stages cracks are sparse and very fine, but Sometimes there are large numbers and damage is
visible at 310 magnification. As they develop they broaden sufficiently serious to cause seal failure. The initial particle
J00397 # IMechE 1997 Proc Instn Mech Engrs Vol 211 Part J
178 B S NAU

1.0

Slope 0.5

Friction coefficient
0.1

0.01
1011 1010 109 108 107 106 105 104
(a) Duty parameter G

40
T/(C)

0
0 100 200 300
(b) Run time (h)
Face temperature (C)

Face temperature (C)

100 100
Friction coefficient ( 100)

Friction coefficient ( 100)

0 0
(c) Run time (h) 100 (d) Run time (h) 5

Fig. 18 Fluctuating friction of mechanical seals. (a) Commercial seal in waterglycerol over a range
of speeds, pressures and viscosities; resincarbon versus Ni-Resist [redrawn from reference
(22)]. (b) Temperature transients of a commercial seal face in distilled water at constant
operating conditions; resincarbon versus Ni-Resist [redrawn from reference (22)]. (c)
Friction and face temperature of seal rings in a material test-rig at constant operating
conditions in distilled water; reaction-bonded SiC versus phenoliccarbon; data logged by
computer (45). (d) Continuation of run shown in (c) but running dry for tribolayer breakdown
at 4 h (45)
Proc Instn Mech Engrs Vol 211 Part J J00397 # IMechE 1997
 MECHANICAL SEAL FACE MATERIALS 179

(a)
Amplitude (arbitrary units)

(b)

Fig. 20 Photograph of remains of an SiC seal ring

operated with boiling interfacial film; WC coun-
terface (43)

(a)

0 50 100
Frequency (Hz)

Fig. 19 Spectral analysis of a test as in Figs 18c and d: (a) (b)

during stable dry running; (b) during a dry
running friction excursion (45)

detachment could result from fatigue due to a rotating

hydrodynamic pressure field, differential thermal expan-
sion or fluid attack at a grain boundary. Further study is
needed to establish the mechanism. Material structure also
Fig. 21 Two forms of surface crazing caused by thermal
affects the occurrence of pitting. Strong direct bonds
stress: (a) `parallel', (b) network
between ceramic grains reduce the risk of grain detachment
(14) as they are normally stronger than those between
ceramic grains and the binder. Consequently, in material
containing isolated ceramic grains formed during sintering,
these isolated grains are particularly vulnerable. Porosity
near the surface may also promote grain detachment as
adjacent grains are less firmly held. From these observa-
tions it may be concluded that a close-packed structure is
advantageous and therefore ceramic grains should be
Fig. 22 Seal face pit with associated score marks
uniform in size and shape.

connect with the blister (Fig. 23). Blistering is associated

7.4 Face blistering
with higher viscosity fluids, notably oils, and also with a
Blistering is an important failure mode of mechanical seal very flat counterface. Formation appears to be a two-stage
carbons which has been extensively studied (6167). One process. Firstly, fluid diffuses into the surface of the carbon
or more low volcano-like blisters form on the running face during normal operation. Then, any transient increase of
and cause leakage due to the resulting face separation. Fine interface temperature causes thermal expansion of the fluid
fractures, either essentially horizontal or steeply angled, trapped in the surface layers of the carbon. As the trapped
J00397 # IMechE 1997 Proc Instn Mech Engrs Vol 211 Part J
180 B S NAU

a ridge of abrasive solid causes severe grooving of the

counterface and premature failure (Fig. 24). A change of
face materials may relieve the problem by enhancing heat
transfer or reducing interface heat generation, but it is
better to eliminate the cause at source.

7.6 Squeal
A seal otherwise running normally may emit unacceptable
high-frequency squeal at about 0.3 or 1.2 kHz (also called
`ringing'). This is difficult to eliminate and can be an
Fig. 23 Schematic section through a seal face blister annoying problem in domestic washing machines and
[based on reference (65)] automotive coolant pumps. This behaviour reflects the high
resonant frequency of small lightweight seals under
torsional excitation. Squeal has been the subject of study
fluid is unable to flow out fast enough to relieve the over many years [e.g. see references (71) to (76)]. It occurs
resulting pressurization of the internal fluid, the surface of more with tap water than ethylene glycol solution and it
the carbon is deformed upwards, fracturing locally to form also occurs over a wider speed range as the fluid tempera-
a blister. This mechanism depends on porosity of the ture increases. The origin lies in stick-slip at the sealing
surface of the carbon and therefore on the grade and quality interface; this, of course, depends on the slope of the
of carbon; more resistant grades are available. It has also frictionspeed relationship. Torque fluctuations up to
been found that a high value of k= reduces the occurrence 1.5 kHz have been recorded in such seals (72).
of blistering (62). Avoiding an exceptionally flat counter-
face also helps, as the hydrodynamic lift associated with a
certain amount of face waviness reduces the risk of large
thermal transients. 8 CONCLUSIONS

The significance of individual material properties and face

7.5 Solids deposition
seal materials in common use has been summarized, and
This is of particular importance when handling fluids boundary lubrication mechanisms have been discussed at
containing dissolved solids (see Section 2.4). Even though some length, with the emphasis on carbongraphites.
the concentration is only `ppm', solute can precipitate on Modern high-duty seal face materials are normally ceramic
the sealing face and cause leakage. Solids may originate based and share common features in their boundary
directly from the sealed fluid, a common experience with lubrication mechanisms, although these differ in detail.
inorganic salt solutions or with silicate corrosion inhibitor, Important shared characteristics are a protective tribolayer
which can be deposited from automotive engine coolant on one or both faces and, on the harder face, a transfer
(10, 11, 6870). However, interface solids can also origi- layer. These layers appear to exist in a state of cyclic
nate indirectly, from material dissolved from components equilibrium, gradually building up and then disintegrating
contacting the fluid. Thus copper from a brass fitting in an and reforming, at least in `good' samples of face material.
otherwise all-steel system has been known to `plate' seal The behaviour seems to have a stochastic character as there
faces, causing a seal failure (22). is considerable variation in performance within and be-
Surface deposition is a problem in automotive coolant tween batches. These protective layers are either formed
systems. The coolant is ethylene glycol diluted with from comminuted wear particles bound by `contaminant'
water plus additives (silicate corrosion inhibitors, etc.), molecules or by tribochemical reaction between ceramic
and incidentals such as leachate from polymeric compo- (wear particles?) and fluid. Carbongraphites depend on
nents, zinc from galvanized storage tanks in the assembly the former, alumina and silicon carbide on the latter.
plant, non-ferrous metals from coolant system compo-
nents or dissolved salts from mains water supplies used
in topping-up fluid. A particular problem was noted with
mercaptan benzo triazole, a rubber vulcanizing agent
leached from hoses (11). Such effects could also be
responsible for failures in other applications, but go
unnoticed since the evidence is usually destroyed.
Solid deposition can be distributed or localized. A
uniform coating may block surface porosity and blister-
like delamination may result. If restricted to a narrow Fig. 24 Solids deposited in a narrow band near the mid-
circumferential band where the interface fluid vaporizes, face [based on references (10) and (11)]
Proc Instn Mech Engrs Vol 211 Part J J00397 # IMechE 1997
 MECHANICAL SEAL FACE MATERIALS 181

There are also physical interactions between the fluid 8 Lebeck, A. O. Principles and Design of Mechanical Face
and face material. Diffusion of fluid into the pore struc- Seals, 1991 (John Wiley, New York).
ture of the substrate, and subsequent thermal transients, 9 Jagger, E. T. Rotary shaft seals: the sealing mechanism of
can cause blistering. This can be either in the form of synthetic seals running at atmospheric pressure. Proc. Instn
localized `volcanoes' or involve delamination over a Mech. Engrs, Part 1, 1966, 181.
10 Komiya, M., Kiryu, K., Masuya, H., Okada, K. and
wider area. In another class of problems, solids from the
Hirabayashi, H. Effect of silicate in coolant on sealing
sealed fluid, including leachate from system components,
performance of mechanical seals for water pumps of auto-
even at ppm levels, are deposited on the faces. This may motive engines. In Proceedings of Japan International Tribol-
occur locally, causing abrasion, or extensively, causing ogy Conference, Nagoya, 1990.
delamination=blistering. Deposition depends on surface 11 Kiryu, K., Hirata, O., Yoshino, A., Masuya, H., Okada, K.
energy and fluid phase transition between the seal faces. and Hirabayashi, H. The relationship between sealing per-
The surface texture of lapped ceramic seal faces has formance of mechanical seals and composition of coolants for
two main components: a peak-lopped component charac- automotive engines. In ASTM STP 1192 (Ed. R. E. Beal),
teristic of the machining and lapping processes, and a 1993 (American Society for Testing and Materials, Philadel-
finer component characteristic of the material structure. phia, Pennsylvania).
The combination gives a bi-Gaussian distribution of 12 Richerson, D. W. Modern Ceramic Engineering: Properties,
Processing and Use in Design, 1982 (Marcel Dekker, New
cumulative height. Decomposition of the profile into
York).
these components facilitates computer modelling of
13 Kato, K. Tribology of ceramics. Wear, 1990, 136, 117133.
mixed-film lubrication and provides a basis for under- 14 Wallis, R. Influence of the microstructure of ceramic mate-
standing the separate roles of material and the manufac- rials on their wear behavior in mechanical seals. Lubric.
turing processes. Engng, 1994, 50(10), 789799.
Other phenomena impairing seal performance include 15 Chen, Y. M., Rigaut, B. and Armanet, F. Friction and wear
structural fracture, thermal stress surface fractures (craz- of alumina ceramics at high sliding speeds. Lubric. Engng,
ing) and face pitting; `squeal' is a cosmetic problem of 1991, 47(7), 531537.
lightweight seals subject to fluctuating torque. 16 Klimek, E. J. Selection, properties and quality assurance of
face materials for rotating mechanical seals. Lubric. Engng,
1987, 44, 719725.
17 Klimek, E. J. Microstructure of silicon carbide materials.
ACKNOWLEDGEMENTS Microstructural Sci., 1987, 16, 295304.
18 Paxton, R. R. Manufactured Carbon: A Self-lubricating
Material for Mechanical Devices, 1979 (CRC Press, Boca
The author wishes to acknowledge the support of BHR
Raton, Florida).
Group Limited and organizations that have supported the
19 Anon. ASTM Standards on Manufactured Carbon and Gra-
mechanical seal work on which this account has drawn, phite Products. PCN 06-305081-09, 1981 (American Society
including: Amoco, BP, European Space Agency, Britoil, for Testing and Materials, Philadelphia, Pennsylvania).
Caltex, ICI, Shell Expro, BW=IP, John Crane, EG&G 20 Nau, B. S. Research in mechanical seals. Proc. Instn Mech.
Sealol, Lucas Aerospace, T&N, Morganite Special Carbons Engrs, Part C, 1990, 204(C6), 349376.
and the Department of Trade and Industry. 21 Wayne, S. F., Baldoni, J. G. and Buyljan, S.-T. Abrasion and
erosion of WCCo with controlled microstructures. Tribology
Trans., 1990, 33(4), 611617.
22 Flitney, R. K. and Nau, B. S. A study of factors affecting
REFERENCES mechanical seal performance. Proc. Instn Mech. Engrs, Part
A, 1987, 201(A1), 1728.
1 Jahanmir, S. (Ed.) Tribology of Ceramics, 1987 (STLE, Park 23 Patir, N. and Cheng, H. An average flow model for
Ridge, Illinois). determining effects of three dimensional roughness on partial
2 Gee, M. G. Wear testing and ceramics. Proc. Instn Mech. hydrodynamic lubrication. Trans. ASME, J. Lubric. Technol.,
Engrs, Part J, 1994, 208(J3), 153166. 1978, 100, 1217.
3 Paxton, R. R. and Hulbert, H. T. Rubbing friction in radial 24 Leefe, S. E. and Williams, M. `Bi-Gaussian' pre-test surface
face seals. Lubric. Engng, 1980, 36, 8995. topography measurements on turbopump seal materials. BHR
4 Labus, T. J. The influence of rubbing materials and operating Group Limited report CR 6276, 1994.
conditions on the power dissipated by mechanical seals. 25 Leefe, S. E. Face seal technology: contact friction and wear
Lubric. Engng, 1981, 37, 387394. final report. BHR Group Limited report CR 6277, 1994.
5 Ashby, M. F. and Jones, D. R. H. Engineering Materials: An 26 Shimomura, T., Kiryu, K., Hirabayashi, H. and Nakajima,
Introduction to Their Properties and Application, 1980 T. A study of the relationship between sealing performance
(Pergamon Press, Oxford). and surface characteristics of end face seals. Lubric. Engng,
6 Ashby, M. F. and Jones, D. R. H. Engineering Materials 2: 1989, 45(12), 785791.
An Introduction to Microstructures, Processing and Design, 27 Shimomura, T., Hirabayashi, H. and Nakajima, T. A study
1986 (Pergamon Press, Oxford). of the relationship between frictional characteristics and
7 Mayer, E. Mechanical Seals, 1977 (Newnes-Butterworth, surface condition of mechanical face seals. Tribology Trans.,
London). 1991, 34(4), 513520.

J00397 # IMechE 1997 Proc Instn Mech Engrs Vol 211 Part J
182 B S NAU

28 Shimomura, T., Hirabayashi, H. and Nakajima, T. A study friction and wear of alumina, zirconia, silicon carbide and
of the relationship between sealing performance and surface silicon nitride. In Wear of Materials, 1989, Vol. 1, pp. 409
condition of mechanical face seals. Tribology Trans., 1992, 417 (American Society of Mechanical Engineers, New York).
35(1), 659666. 51 Gee, M. G. The formation of aluminium hydroxide in the
29 Shimomura, T. and Hirabayashi, H. The relationship be- sliding wear of alumina. Wear, 1992, 153, 201227.
tween the surface condition and the tribology of mechanical 52 Flitney, R. K. and Nau, B. S. Performance testing of
face seals. Jap. J. Tribology, 1992, 37(7), 827838. mechanical seals. In Proceedings of Thirteenth International
30 Barylski, A. W. Surface contamination of cast iron after Conference on Fluid Sealing, 1992, pp. 441466 (BHR Group
lapping. Lubric. Engng, 1996, 52(1), 6367. Limited, Cranfield).
31 Savage, R. H. Graphite lubrication. J. Appl. Physics, 1948, 53 Paxton, R. R. and Shobert, W. R. Testing carbon for seals
19(1), 110. and bearings. Lubric. Engng, 1971, 27, 2733.
32 Savage, R. H. Physically and chemically adsorbed films in the 54 Paxton, R. R. and Hulbert, H. T. Rubbing friction in radial
lubrication of graphite sliding surfaces. Ann. New York Acad. face seals. Lubric. Engng, 1979, 36, 8995.
Sci., 1951, 53, 862869. 55 Massaro, A. J. The `mating pairs' concept for mechanical
33 Savage, R. H. and Schaefer, D. L. Vapor lubrication of face seals. Lubric. Engng, 1988, 44(5), 436446.
graphite sliding contacts. J. Appl. Physics, 1956, 27(2), 136 56 Ryde-Weller, A. J., Thew, M. T. and Wallis, R. The
138. performance of small commercial mechanical seals using
34 Paxton, R. R. and Shobert, W. R. Testing high speed seal various face materials when subjected to intermittent and
carbons. ASLE Trans., 1962, 5, 308314. variable speed running. In Proceedings of Twelfth Interna-
35 Midgely, J. W. and Teer, D. G. An investigation of the tional Conference on Fluid Sealing, 1989, paper G3 (BHRA,
mechanism of friction and wear of carbon. Trans. ASME, J. Cranfield).
Basic Engng, 1963, 85, 488494. 57 Gates, R. S., Yellets, J. P., Deckman, D. E. and Hsu, S. M.
36 Lancaster, J. K. The friction and wear of non-graphite Considerations in ceramic friction and wear measurements. In
carbons. ASLE Trans., 1977, 20(1), 4354. ASTM STP 1010 (Eds C. S. Yust and R. G. Bayer), 1988
37 Lancaster, J. K. Transitions in the friction and wear of (American Society for Testing and Materials, Philadelphia,
carbons and graphites sliding against themselves. ASLE Pennsylvania).
Trans., 1975, 18(3), 187201. 58 Shimomura, T., Nishihira, E., Chiba, K., Yoshino, A.,
38 Lancaster, J. K. Instabilities in the frictional behaviour of Tanoue, H. and Nakajima, T. A study of cracking phenomena
carbons and graphites. Wear, 1975, 34, 275290. on carbon rings of mechanical seals for an automotive air
39 Lancaster, J. K. Carbons and graphite in tribology. In conditioning compressor. SAE paper 900338, 1990.
Proceedings of Fourth International Carbon and Graphite 59 Shimomura, T., Yanagisawa, T., Chiba, K., Hirabayashi, H.
Conference, 1976, 663. and Nakajima, T. Analysis of cracking phenomena on
40 Lancaster, J. K. Additive effects on the friction and wear of carbon-rings of mechanical seals for an automotive compres-
graphite carbons. In Proceedings of Third LeedsLyon sor. SAE paper 910531, 1991.
Symposium, London, 1978 paper 8(i) (Mechanical Engineer- 60 Komiya, M., Matsuda, K. and Kaneta, M. Experimental and
ing Publications, London). analytical study on thermocracking of alumina ceramic ring in
41 Lancaster, J. K. Stabilization of the friction and wear of non- a mechanical seal. Tribology Trans., 1994, 37(2), 245252.
graphite carbons by additives. In Proceedings of Second 61 Strugala, E. W. The nature and causes of seal carbon
International Conference on Solid Wear, 1978, paper ASLE blistering. Lubric. Engng, 1972, 28, 333339.
SP-6, pp. 176188. 62 Shimomura, T., Yoshino, A., Matsumoto, S. and Hirabaya-
42 Lancaster, J. K. Material-specific wear mechanisms: rele- shi, H. Sealing performance of end-face type seals for an
vance to wear modelling. Wear, 1990, 141, 159183. automotive air conditioning compressor. SAE paper 870513,
43 Nau, B. S. The operation of unbalanced mechanical seals at 1987.
high pressure. BHR Group Limited report RR 1819, 1982. 63 Wakely, K. Mechanical seals: some developments in face
44 Nau, B. S. Mechanical seal material performance. BHR Group materials. Tribology Int., 1986, 19(4), 198203.
Limited report CR 3009, 1989. 64 Miyazawa, M., Izumi, J., Uchibori, Z., Tanoue, H. and
45 Williams, M., Flitney, R. K. and Nau, B. S. Factors affecting Fujita, T. Observations on generation of the carbon blister
mechanical seal face-material performance. BHR Group phenomena on mechanical seals. Lubric. Engng, 1988, 44(6),
Limited report CR 6126, 1993. 520526.
46 Yen, B. K. Influence of water vapor and oxygen on the 65 Uchibori, Z., Miyazawa, M., Matsui, S., Kiryu, K., Tanoue,
tribology of carbon materials with sp2 valence configuration. H. and Kaneta, M. Recent results of observations concerning
Wear, 1996, 192, 208215. blister generation on carbon mechanical seal surfaces. Lubric.
47 Tomizawa, H. and Fischer, T. E. Friction and wear of silicon Engng, 1990, 46(12), 794803.
nitride and silicon carbide in water: hydrodynamic lubrication 66 Uchibori, Z., Miyazawa, M., Lomiya, M., Tanoue, H.,
at low sliding speed obtained by tribochemical wear. ASLE Hirabayashi, H. and Kaneta, M. Fundamental study of the
Trans., 1987, 30, 4146. carbon blister generation at the starting period on mechanical
48 Fischer, T. E. Tribochemistry. Ann. Rev. Mat. Sci., 1988, 18, seals. Lubric. Engng, 1991, 47(10), 847858.
303323. 67 Uchibori, Z. and Kaneta, M. Conditions of carbon blister
49 Jahanmir, S. and Fischer, T. E. Friction and wear of silicon generation on mechanical seals. Lubric. Engng, 1992, 48(8),
nitride lubricated by humid air, hexadecane and hexadecane 657664.
0:5% stearic acid. Tribology Trans., 1988, 31, 3243. 68 Kiryu, K., Tsuchiya, K., Shimomura, T., Yanai, T., Okada,
50 Sasaki, S. The effects of surrounding atmosphere on the K. and Hirabayashi, H. The effect of coolant additives and

Proc Instn Mech Engrs Vol 211 Part J J00397 # IMechE 1997
 MECHANICAL SEAL FACE MATERIALS 183

seal composition on performance of water pump seals of Sealing, 1969, paper 13 (BHRA, Harlow).
automotive engines. SAE paper 890609, 1989. 72 Kiryu, K., Yanai, T., Matsumoto, S. and Koga, T. An
69 Kiryu, K., Tsuchiya, K., Yonehara, Y., Shimomura, T. and analysis of `ringing' phenomena on a water pump mechanical
Koga, T. An investigation of deposits formation on sealing seal. ASLE Trans., 1985, 28(2), 261267.
surfaces of water pump end face seals. Lubric. Engng, 1989, 73 Kiryu, K., Yanai, T., Matsumoto, S. and Koga, T. An
45(1), 4955. analysis of `ringing' phenomena on a water pump mechanical
70 Kiryu, K., Hirata, O., Yoshino, K. and Hirabayashi, H. The seal (Part II). ASLE Trans., 1988, 29(1), 2534.
relationship between sealing performance of mechanical seals 74 Kiryu, K., Yonehara, Y., Matsumoto, S. and Koga, T. An
and composition of coolants for automotive engines. In ASTM analysis of `ringing' phenomena on a water pump mechanical
STP 1192 (Ed. R. E. Beal), 1993 (American Society for seal (Part III). ASLE Trans., 1988, 31(2), 269275.
Testing and Materials, Philadelphia, Pennsylvania). 75 Nau, B. S. Research in mechanical seals. Proc. Instn Mech.
71 Hirabayashi, H., Oka, K. and Ishiwata, H. The relationship Engrs, Part C, 1987, 204(C1), 1728.
between ringing, heat transfer and sealing condition. In 76 Tanoue, H. Sound and vibration of contact seals. Jap. J.
Proceedings of Fifth International Conference on Fluid Tribology, 1990, 35(5), 541549.

J00397 # IMechE 1997 Proc Instn Mech Engrs Vol 211 Part J