The relationship between hydrolytic

stability of adhesive joints and

equ 'brium water content
M.R. Bowditch, D. Hiscock and D.A. Moth

(Admiralty Research Establishment, UK)

The effects of a range of equilibrium water contents (EWCS) on the adhesion between
model adhesives and substrates were investigated using finely divided "adherends"
dispersed within an "adhesive" matrix. No evidence was found for the existence of a
limiting equilibrium water content (LEWC) below which joints were unaffected by water.
It was concluded that attack of adhesion interfaces by water was reversible and that at
any given EWC the full potential of water for degradatiotl of adhesion interfaces was
modified by the partitioning of water between the adherend surface and adhesive
matrix.

Key words: adhesion; water absorption tests; modulus reduction; equilibrium water
content; water partitioning

A number of workers have noted that some adhesive
joints appear to be stable over long periods in the
presence of limited water concentrations. Maloney
et aLI found no significant weakening of epoxide to
aluminium alloy joints after exposure to 45% relative
humidity (RH) at about 20C for 10 000 hours, whilst
De Lollis2 has spoken of epoxide to aluminium joints
which survived ambient laboratory humidity for I I
years with no ill effect. In addition, Brockmann 3
suggested that bonds to steel have, in his experience,
successfully withstood humidities of up to about 60%
for as long as 10 years. Whilst studying epoxide-
bonded, grit-blasted, mild steel butt joints, Gledhill
et al. 4 noted that exposure to 55% RH at 20C for 2500
hours failed to result in any loss of joint strength,
although exposure to liquid water at progressively
higher temperatures led to increasing loss of joint
strength. Information of this kind led these authors to
introduce the concept of a critical water concentration
which must be present before degradation may occur.
The objective of the present work was to study
model adhesive joint systems with a view to
determining quantitatively such limiting equilibrium
water contents (LEWCS)below which the joints appear
to be hydrolytically stable and to determine the way in
which such LEWCSchange with adhesive chemistry
and/or with the nature of the adherend. An important
0143-7496/91/030163-07 1991 Butterworth-Heinemann Ltd
INT.J.ADHESION AND ADHESIVES VOL. 11 NO. 3 JULY 1991
feature of these studies would be that observations and
measurements would be made only on systems which
had equilibrated with their environment.
Background
A major problem associated with durability
investigations is that equilibrium conditions are slow to
achieve with conventional adhesive joints where
generally impermeable adherends sandwich adhesive
between them. For test-pieces with geometries of this
type water can gain access only from around the
perimeter of the joint and the pathway for the water is,
necessarily, relatively long. Extended periods of
exposure are required, therefore, if tests are to be
conducted on equilibrated specimens.
A secondary difficulty is that such joints have then
to be tested to destruction in order to assess the extent
to which they have succumbed to the ravages of water.
The number of observations that may be made are,
therefore, limited by the number of joints prepared at
one time if specimens with a common history are to be
used throughout the investigation. The consequence is
that the frequency of testing is inhibited by the limited
number of specimens available.
At the Admiralty Research Establishment we have
largely overcome these problems by use of a technique
163
in which the normal structure of a joint is reversed and
impermeable adherend is surrounded by permeable
adhesive: in short, the technique depends upon the use
of finely divided "adherend' (filler) surrounded by
'adhesive" (matrix). The presence of finely divided
"adherend" adds substantially to the modulus of the
unfilled 'adhesive" and a drop in modulus signifies the
LEWC below which an adhesive joint is apparently
unaffected by the presence of water. Once mixed, the
adherend-filled adhesive is then cast into a thin sheet,
about 1 mm thick, which is cured and machined into
dumb-bell shapes. After equilibration with the
environment, testing involves making modulus
measurements, the magnitude of each of which is a
function of the adhesive matrix, the presence of the
particulate adherend and, most importantly, the
adhesion between matrix and filler. Strain levels are
restricted to values within the elastic limit of the system
under test and the method is therefore non-destructive.
Confidence in the use of this test method derived from
earlier work 5 in which the effects of water on a filled
polyurethane elastomer were investigated and showed
that a number of tests of this kind could be conducted
on the same specimen without any loss of performance.
This same work showed that attainment of equilibrium
was the rate-determining stage and that there were no
further changes in properties associated with longer
exposure times. Test-pieces of this kind can equilibrate
with the environment much more rapidly because
water has only to travel about 0.5 m m to reach the
centre of the specimens.
Experimental
Work was undertaken with the hope that LEWCs could
be identified and plotted against some function of
adhesive chemistry. To this end, model 'adhesives' were
produced from stoichiometric mixes of standard
epoxide resin (the diglycidyl ether of bisphenol A with
an epoxide equivalent of about 200) with blends of two
polyoxypropylene diamines of different molecular
weights. By using such blends, cross-link density was
arranged to vary between a m a x i m u m achieved by the
use of the lower molecular weight amine alone to a
minimum resulting from the sole use of the higher
molecular weight reactant.
A number of finely divided 'adherends" were
investigated (see Table 1), including pure barium
sulphate, an impure form of which had been used as a
filler in the poly}a,rethane rubber which had received
attention earlier ~. Rather more practical substrates,
such as ferric oxide, titanium dioxide and zinc oxide,
were also included in the work. The concentration of
filler used depended, to some extent, on particle size
but in general the content was kept up in order to
Table 1. Specific surface areas of model adherends
Adherend Specific surface area (m 2 g-l)
Calcium carbonate 0.29
Barium sulphate 2.80
Zinc oxide 3.49
Ferric oxide 6.46
Titanium dioxide 8.06
164 INT.J.ADHESION AND ADHESIVES JULY 1991
develop as high a modulus as was possible compatible
with the maintenance of a suitably elastic product. In
this way it was hoped that even relatively low levels of
debonding by water would be indicated by a significant
fall in modulus.
After thoroughly mixing the two or three liquid
components of the model adhesives, the finely divided
adherend was added and also thoroughly blended in.
After degassing under vacuum, the resulting filled
resins were cast in about 1 mm thick section between
Cellophane-clad glass plates. After 18 hours at room
temperature, the cure was completed by exposure to
80C for 24 hours in an oven. On removal from the
oven, castings were allowed to cool to room
temperature before demoulding and the preparation of
dumb-bells to BS 2782, Part 3. If the nature of the
product allowed, dumb-bells were cut from the sheet by
use of a cutter in a hand press, but with higher
modulus materials specimens were routed to shape
with tungsten carbide cutters. Once prepared,
specimens were spotted in preparation for
extensometry, numbered and then placed in a
desiccator to thoroughly dry for a minimum of 14 days.
Various equilibrium water contents were introduced
into the specimens by exposure to environments of
known relative humidity obtained by the use of a range
of saturated salt solutions in thermostatically
controlled enclosures. Required exposure times were
obtained from experiments on small coupons of
material of the same thickness as the dumb-bells which
were exposed to the chosen environment maintained
within a small desiccator in a cabinet situated under a
recording microbalance. The specimens were
suspended in the desiccator by means of a fine wire
passing down the centre of a long glass tube fitted into
the top of the desiccator. The output from the balance
was fed to a microcomputer programmed to plot
percentage mass uptake against root time in hours. The
time to equilibrium was indicated by the appearance of
a plateau on the plot. Modulus measurements were
made soon after the dumb-bells had reached
equilibrium with the environment.
Modulus measurements were made on an Instron
1185 tensile test machine fitted with a 1 kg load cell
and a Lloyd optical extensometer at a jaw separation
rate of 500 mm m i n - k Six replicates of each
composition were tested after thorough drying in a
desiccator over magnesium perchlorate (Anhydrone)
with the mean values used to plot points on the
modulus/water content curve. Normally, the same six
dumb-bells were used after equilibration with
environments of increasing water activity levels.
However, a larger number, usually 12, were always
processed in order that reserves were available to take
the place of specimens which did break. The generally
repeated use of the same test-pieces meant that less
scatter was found with the results than would otherwise
have been the case. Because the accuracy of the
experimental method was less good at low strain levels,
modulus values were calculated at 20% strain wherever
possible, but otherwise the values at 10% strain were
recorded. The experimental method, therefore, limited
us to a range of lower modulus materials but, on the
other hand, because of their relatively low glass
transition temperatures (Tgs) the presence of quite
small EWCs had little effect on the moduli of the
Table 2. Composition of the four model adhesives
employed with cross-link density increasing from A IOG
to D with increasing D 2 3 0 content

Component Mix (phr)

A B C D

Epoxide resin 100 1O0 100 1 O0 , \\.

(Araldite GY250)
o~ 8( _ \ \h,\\ ',, \ \
Hardener 1 166.2 140.2 110.8 83.1
(Jeffamine D2000)
Hardener 2 12.7 16.2 19.0 22.2
(Jeffamine D230) E

o
unfilled materials which were also evaluated side-by- 6o
-- ~ N\\
side with filled systems.

R e s u l t s and d i s c u s s i o n
The formulations of the four adhesives used are given \
in Table 2. Water absorption isotherms for these \

compositions appear in Fig. 1 where it can be seen that \

the difference between them is so small that EWC 4C I I I I ^\
20 40 60 80 I00
values appear to follow the same curve.
Relative humidity (%)
At an early stage these model adhesives were used
with finely divided AR grade barium sulphate as a Fig. 2 Plot showing the percentageof dry modulus retained by three
model adherend. Modulus measurements were made on barium sulphate-filled systems when equilibrated with a range
ofrelative
humidities
dumb-bells prepared from these materials after drying
and then after equilibration with a range of
progressively increasing relative humidities. Plots of as Fig. 2. The solid lines show discontinuities which
results obtained from three of the four systems appear would be expected to indicate the presence of LEWCS.
The values for these LEWCS may be obtained by
reference to the appropriate absorption isotherm.
3.0
However, these data are not presented because, despite
repeated attempts, it has not proved possible to

Formulotion I
Ax /
B
C o
a
o
reproduce such curves, and neither have we observed
similar behaviour with any other adhesive/adherend
combination. It is suggested that these data, with the
possible exception of the results obtained with
composition C, could equally be plotted as curves
D z~ without the discontinuity, as shown by the hatched

/
2.0 lines which were obtained by linear regression
analyses.
Insofar as the horizontal portions of the solid curves
may reflect reality, it is suggested that they result from
the simultaneous operation of two mechanisms
resulting from ingress of water, one of which has a

o/
degrading influence on adhesion, and hence modulus,
whilst the second may enhance such properties. There
is much experience of degradation resulting from
~. I.O
LU J' hydrolytic attack, but there is also evidence for the
presence of small quantities of water having a
beneficial effect on adhesive joint strength. We have,
for example, observed increases in joint strength
associated with the prolonged immersion in seawater6
of steel/steel butt joints. Such increases may have been
the result of water plasticizing the adhesive and
reducing the adverse effects of built-in stresses
0 I I following shrinkage during the curing process. In a
0 20 40 60 80 I00
review, Minford 7 has attributed to Pleuddemann the
Relotive humidity (%)
claim that shrinkage associated with the curing process
Fig. 1 Waterabsorptionisothermsforunfilledadhesives is a principal factor in bond failure. Brewis et al. s have

INT.J.ADHESION AND ADHESIVES JULY 1991 165

shown that aluminium alloy lap joints bonded with a 3.0
DGEBA-DAPEE system increased in joint strength with
increasing water content and they too suggested relief Barium sulphate o "]
of stresses within the joints as the explanation. Zinc oxide x I Filled
On this basis it may well be that the tolerance that
Calcium carbonate A
adhesive joints have shown to various ambient
humidities results from the simultaneous operation of
these two mechanisms rather than to an inherent
resistance of adhesion interfaces to aqueous attack 2.0 n
below certain critical water content levels. g
Our attempts to identify the existence of LEWCS g
included the investigation of many combinations of
adhesive and adherend. Apart from the use of four
formulations, already referred to, five different E=
particulate adherends were involved in the study
though not in every, combination with each of the resin
g
systems. uJ
Fig. 3 shows results obtained with adhesive A, the 1.0--
least cross-linked system used. The lower graph shows Unfilled
that the modulus of the unfilled system is apparently
unchanged as a result of progressive equilibration with
a range of relative humidities up to about 95%. The
modulus of this material is very low (about 4 MPa) and
at laboratory temperatures it was rubbery and well e
e x
above its Tg. In such circumstances the presence of
about 2.5% water, which is the EWC found with this o I I I
material when subjected to an environment of 95% RH, o 20 40 60 80 100
would not be expected to exert a noticeable effect on Relative humidity (%)
modulus. However, it is very clear that water present in Fig. 4 Equilibrium water content of adhesive A derived from data
the adhesive/adherend composites does have an effect obtained from filled materiels (plotted points) compared with the
on modulus. It can be seen that modulus reduces absorption isotherm for the unfilled adhesive (solid curve)
with increasing relative humidity and also, therefore,
with increasing EWe.
In Fig. 4 the continuous line represents the The plotted points represent EWCS of the unfilled
absorption characteristics of the unfilled adhesive A. adhesive calculated from water absorption data
obtained from the filled systems on the basis that water
5C take-up figures were a function of the organic content
Barium sulphate ~ "J only and were independent of the inorganic
component. It will be noted that the calculated values
Zinc oxide o Filled
follow the solid line quite closely.
Calcium carbonate x Fig. 5 features results obtained from the use of
adhesive B which had a marginally higher cross-link
density than did adhesive A. Here, a slight but
noticeable reduction in modulus with relative humidity
was observed with the unfilled adhesive. All three
2(
systems, containing 250 phr filler, show a progressive
decrease in modulus with increasing EWC.
Because of some doubt over modulus values
o obtained with the barium sulphate composite at the 0%
and 10% RH points, at the end of the experimental
sequence - - when measurements after equilibration
"o with 95% RH had been made - - the material was
redried in a desiccator and the modulus was
remeasured. The value obtained, together with that
obtained after re-equilibration with the 10% RH
environment, is shown in addition to the original
values.
The fact that the modulus of the unfilled adhesive is
relatively unchanged with increasing EWC indicates

0I
0
I
20 40
I

Relative humidity (%)

I
60 80
I I
100
that this recovery, is not merely the result of removing a
plasticizer, but is largely attributable to the re-
establishment of adhesively bonded interfaces. To
demonstrate the reversibility of this process more
clearly the modulus of model adhesive A, using 400 phr
Fig, 3 The relationship between the modulus of model adhesive A, both of zinc oxide as the model adherend material, was
filled and unfilled, with various substrates and relative humidity measured in the d o ' condition and then again after

166 INT.J.ADHESION A N D ADHESIVES JULY 1 991

 Barium sulphate x l
I Zinc oxide o Filled
Ferr ~c oxide ,.',
Recovery points

O_
2O

-Z
u~

2~

Fig. 6 Scanning electron micrograph showing the fracture surface of the

d~ dumb-bell

Unfilled

I I I I
20 40 60 80 I00
Relative humidity (%)
Fig. 5 The relationship between the modulus of model adhesive B when
unfilled and filled with various substrates and relative humidity

Table 3. Illustrating the reversibility on redrying of

the degrading effect of w a t e r on a zinc oxide-filled
formulation A

Operation 'Dry" modulus 'Wet" modulus Fig. 7 Scanning electron micrograph showing the fracture surface of a
dumb-bell after equilibration with 66% RH
no. (M Pa) (M Pa)

1 23.8 17.8
2 23.8 18.0
3 23.9 15.9
equilibration with 86% relative humidity. The
composite dumb-bells were redried and the operation
was repeated on two further occasions. The results
obtained are given in Table 3 and show that the
process is reversible. These results, obtained with an
epoxide matrix, parallel the behaviour observed earlier ~
where we found similar reversibility, when using a
polyurethane matrix.
Figs 6 and 7 show scanning electron micrographs of
the fractured surfaces of a "dry" and "wet" dumb-bell
respectively. By comparison with Fig. 6, Fig. 7 reveals
the presence of a much higher incidence of what is
apparently interfacial failure and provides further
visual evidence that the loss of properties resulting
from aqueous attack is largely attributable to the
disruption of interfacial adhesion forces. Because the
specimens had exceeded their elastic limit, interfaces in
the region of the fracture were separated beyond the
ability of the system to repair itself and the process was
very obviously irreversible.
The results obtained with adhesive C were entirely
compatible with those obtained with the other model
adhesives but have not been included because of the
similarity of its performance with adhesive D.
The results obtained from the use of adhesive D
show much higher initial modulus figures, but these
reflect not only the greater cross-link density of mix D
but also the fact that modulus values at 10% strain
were measured rather than the usual 20% values. Much
lower elongation at break values for this formulation
imposed the selection of a lower strain for modulus
calculations upon us.
This formulation was also examined when used with
180 phr of three different particulate adherends and
here, as can be seen in Fig. 8, the properties changed
in a very similar way with the extent of debonding
being a function of EWC.
To show that the water diffusion process was the
rate-determining stage for the epoxide system, modulus
measurements were made on formulation D filled with
180 phr of titanium dioxide when dry, and after
equilibration with an environment at 81% RH. The
results are presented in Fig. 9 and from these it
appears that the full extent of the disruption of the
adhesion interlace is rapidly achieved and is not
subsequently progressive with the passage of time.
Each value is the mean of the results obtained from
the three specimens which were tested at each
withdrawal. These results also parallel those found

INT.J.ADHESION AND ADHESIVES JULY 1991 167

 iSOr
I phr of titanium dioxide (see Fi~. 8). Even here. ~vherc
I Barium suiohate o
I I the joint area was about 3.76 m~- g-l of bonded
' ~ T~taniurn d~ox~de A
composite, the EWE7(about 0.27%) resulting from
..)
I I Or-- Ferric oxiCe x exposure to _0,, Rtt is sufficient to occupy the entire
!
adherend surface with about 20 molecular layers which
should be sufficient to cause the loss of all effective
bond strength and an appropriate drop in modulus.
This calculation was based on the figure of
0
1.56 x 10-h~ m e for the area occupied bv one molecule
<- of water '~. Nevertheless, the facts are their degradation
is still lar from complete at this stage and filler/matrix
adhesion is still contributing significantly to the
o
modulus of the composite.
It is suggested that this may be explained by'
invoking a partitioning of the water between the
adhesive matrix and the adherend surface. In an
adhesive joint the energetic surface of the particulate
20--
adherend vdll be competing for water molecules with
the adhesive matrix which has an abundant potential
10- for hydrogen bonding within the polyoxypropylene-
rich structure.
t I 1 I
40 60 80 I00 The operation of such a mechanism would explain
Relative humidity (%) commonly observed behaviour where loss of joint
strength after prolonged exposure to a given moist
Fig. 8 The relationship between the modulus of model adhesive D w h e n
filled with various substrates vs. relative humidity
environment levels out at some value depending on the
nature of the bonded system. If, as a result of
equilibration with a humid environment and the
90 operation of a partitioning mechanism, only a limited
proportion of the absorbed water is available for
interfacial attack and this is less than that necessary
80 B

for total disruption of adhesive bonding, then there will

obviously be a residual .joint strength.
70 If water does partition itself in this way between
\
\ adhesive matrix and adherend surface, the partition
\ coefficient should reflect the changing hydrophilicity
\
60 N associated with the varying chemistry of the adhesive
\ Approximate time
13. \ to equilibrium
in contact with a given adherend and the water-
x
attracting nature of different adherends bonded with a
50 _ N
N common adhesive. An examination of the results
.o \ __x x
x suggests that, for the adhesives investigated, the
x
40 differences in chemistry were unfortunately too small
for any particular trend to be apparent.
o
50
Conclusions
With the materials and methods used, and despite
20 repeated attempts, we were quite unable to identify the
presence of LEWCS in the sense that the overwhelming
evidence was that water - - even at veD' low
IO
concentrations - - caused reductions in modulus which
were mainly attributed to disruption of adhesion
o I I I between 'adhesive' and "adherend'. It seems probable
10 I0 2 I0 3 I0 4
that there may well be apparent LEWCs for certain
Exposure time ( h ) bonded systems where invasion of the joint by water in
Fig. 9 Plot reflecting the extent of degradation of adhesion associated limited quantities results not only in a degree of
with a given EWC and its independence of time of exposure debonding but also in some reduction of internal
stresses resulting from shrinkage on cure for example.
The net effect may be for the strength of such joints to
earlier when working with a pol}a,trethane matrix 5. be retained and for the joints to appear as though they
Against this background it is of interest to note that had been unaffected by the water content at these low
for all adhesive/adherend combinations described, concentrations.
degradation proceeds with further additions of water It was apparent even after equilibration with low
long alter one would have expected maximum RHS, and with adhesive forces still abundantly present,
degradation. By way of illustration, the maximum that the Ewcs were tar in excess of what one might
bonded area associated with minimum resin content expect to be required to effect total disruption of
was tbund with adhesive lbrmulation D containing 180 adhesion interfaces.

1 68 INT.J.ADHESION AND ADHESIVES JULY 1 991

 It is concluded that these facts support the concept
of water partitioning itself between the relatively
energetic adherend surfaces and hydrophilic regions
within the adhesive matrix such as those hydrogen-
bondable sites offered by the substantial polyether
element of the epoxide structure.
Acknowledgements
Helpful discussions with Dr B.E. Foulger and Dr R.F.
Long are gratefully acknowledged.
Copyright Controller, HMSO, London, 1990.
References
i Moloney, A.C., Brewis, D.M., Comyn, J. and Cope, B.C. in
"Adhesion 5' edited by K.W. Allen (Applied Science Publishers,
London, 1981) p 133
2 De Lollis, N.J. Nat SAMPE Syrnp Exhibition 22 (1977) p 673.
3 Brockmann, W. in 'Durability of Structural Adhesives" edited by
A.J. Kinloch (Applied Science Publishers, 1983) p 300
4 Gledhill, R.A., Kinloch, A.J. and Shaw, S.J. J Adhesion 11 No 1
(1980) p 3
5 Bowditch, M.R. and Stannard, D.J. J Adhesion 21 Nos 3-4
(1987) p 329
6 Bowditch, M.R., Clark. J.D. and Stannard, K.J. in "Adhesion 1 I"
edited by K.W. Allen (Elsevier Applied Science, London, (1987) p 1
7 Minford, J.D. in "Durability of Structural Adhesives" op. cit. p 151
8 Brewis, D.M., Comyn, J. end Tegg, J.L. Int J Adhesion and
Adhesives 1 (1980) p 35
9 McClellan, A.L. and Hernsberger, H.F. J Colloid Int Sci 23 {1967)
p 577
Authors
The authors are with the Admiralty Research
Establishment, Holten Heath, Poole, Dorset, UK.
Enquiries should be directed to Mr Bowditch.
INT.J.ADHESION AND ADHESIVES JULY 1991 169