Polymer 42 (2001) 5285±5291

www.elsevier.nl/locate/polymer

Effective adhesive modulus approach for evaluation of curing stresses

D.J. Macon*
Thiokol Propulsion, P.O. Box 707, M/S 243, Brigham City, UT 84302-0707, USA
Received 28 June 2000; received in revised form 21 September 2000; accepted 26 September 2000

Abstract
A study was performed to measure the curing stresses of several epoxy systems using the bimaterial beam approach. In addition, the stress
relaxation pro®le of the adhesive bonded to the metallic strip (bimaterial strip) was compared to the relaxation pro®le of the bulk adhesive. It
was found that the shape of the relaxation curves was consistent for both samples, but that the magnitude of the two curves differed. It is
argued that the difference between the curves is a result of curing stress and that the long-term effective modulus as measured from the
bonded adhesive should be used to establish the curing stress. Use of an improper modulus is one possible cause of confusion regarding the
determination of curing stresses as reported in the literature. q 2001 Thiokol Propulsion. Published by Elsevier Science Ltd. All rights
reserved.
Keywords: Residual stress; Viscoelastic modeling; Thermoset cure

1. Introduction will ®rst gelate and then vitrify. The gelation process results
in an equilibrium elastic modulus, whereas vitri®cation is
The use of adhesives as engineering materials requires an associated with the transition from a rubbery modulus to a
understanding of their mechanical properties. This includes glassy modulus [7]. The change in elastic modulus asso-
knowledge of the nature and magnitude of curing stresses ciated with gelation and vitri®cation is the dominating
formed when the material is used in an adhesive joint. These factor in the occurrence of the contractive stress rather
stresses can superimpose themselves on externally applied than just the volume contraction itself [8]. When the adhe-
stresses and thus cause premature failure of the adhesive sive is cured above the ultimate glass transition temperature,
bonds [1]. Cure stresses can also negatively impact the frac- curing stresses resulting from volume changes are mini-
ture properties of the adhesive [2]. Factors that will in¯u- mized because of the low magnitude of the rubbery modulus
ence the magnitude of the stresses include volume changes [9,10].
during curing, geometry of the constraining hardware, Curing stresses also result as the adhesive joint is cooled
curing time and temperature, material properties of the from the cure temperature to ambient temperature. Thermal
system, mechanical deformation, and dimensional changes stresses occurring from cooling arise from differential ther-
resulting from thermal ¯uctuations [3±5]. mal expansions in the joint. The magnitude of these stresses
Curing stresses can result from volume changes that will depend on the material constants of the adhesive includ-
occur during the polymerization process and will depend ing the thermal expansion coef®cient, Poisson's ratio, and
on the chemical nature of the material and the curing proce- elastic modulus [4]. These constants depend on the extent of
dure [6]. The adhesive can go through two main transitions cure, test temperature, and time scale at which they are
during the cure, gelation and vitri®cation. Gelation corre- evaluated. The magnitude of the curing stresses is also
sponds to the formation of an in®nite network. Vitri®cation affected by how constrained the adhesive is in the joint. If
occurs when the glass transition temperature, Tg, of the the material is subjected to a one-dimensional constraint, the
adhesive reaches the cure temperature. If the material is resulting stress will scale linearly with the dimensional
isothermally cured above the ultimate glass transition change resulting from thermal expansion and chemical
temperature, only gelation will occur. When the material shrinkage. This trend is true if the material behaves accord-
is cured below the ultimate glass transition temperature, it ing to the following assumptions: material isotropy, incre-
mental linear elastic behavior, history-dependent elastic
* Tel.: 11-435-863-3035; fax: 11-435-862-2202. coef®cients, and no viscoelastic behavior. The curing stres-
E-mail address: macondj@thiokol.com (D.J. Macon). ses under biaxial constraints will be larger by a factor of
0032-3861/01/$ - see front matter q 2001 Thiokol Propulsion. Published by Elsevier Science Ltd. All rights reserved.
PII: S 0032-386 1(00)00725-4
5286 D.J. Macon / Polymer 42 (2001) 5285±5291

1=…1 2 y† where y is Poisson's ratio. The curing stresses for strip having the dimensions of 5.79 £ 0.635 £ 0.013 cm.
the three-dimensionally constrained system will increase The adhesive thickness was controlled to 0.005 cm. After
over the uniaxial case by a factor of 1=…1 2 2y†: The effect coating with the adhesive, the bimaterial strip was placed at
of the constraining geometry can be examined by consider- 418C for 48 h and then rapidly cooled to 228C for the TIGA
ing an adhesive that is constrained in three-dimensions and 321 and EA 946 adhesives. This cure cycle was chosen
is cured above its ultimate glass transition temperature. For because it duplicates the cure schedule of these adhesives
this case, the material exhibits elastomeric-like behavior and as they are used on the nozzle of the Space Shuttle booster
Poisson's ratio approaches a value of 1/2. The denominator motor produced by Thiokol Propulsion Group. The Epon w
in 1=…1 2 2y† approaches zero with the result of a large 828 1 Versamid w 140 epoxy was cured at 1218C for 15 h
induced stress for any process that results in shrinkage [3]. and then rapidly cooled to 228C.
From an engineering approach, measuring the stress asso-
ciated with the curing process is desirable. A number of 2.3. Curing stress
experimental approaches have been used to make these
After the bimaterial strip was cured and cooled to ambient
measurements and include wafer de¯ection [11], photo-
temperature, it began to curve. The de¯ection of the center
elasticity [12], strain gage [5,13,14], bimaterial strip
of the strip, d , was optically measured and related to the
[1,14±17,19], impulse viscoelasticity [18], and cantilever
radius of curvature, r , by the following relation
methods [10,20]. In addition, mathematical methods
have been used to estimate curing stresses in adhesive l2
joints. On the practical front, most of these analyses are dˆ …1†
8r
elastic and make no allowance for stress relaxation of the
adhesive [20]. where l is the length of the bimaterial strip and it is assumed
Of particular interest to this work is the bimaterial strip that d ! r:
approach. This method is similar to the analysis of the bime-
tallic strip developed by Timoshenko [21]. If a layer of 2.4. Stress relaxation
adhesive bonded to a metallic strip contracts or expands, Recently, a DMA was used to measure the stress-free
the strip will curve in response. The extent of curvature is temperature of a bimaterial strip [23]. In this work, a Rheo-
a measure of the stress state in the adhesive if linear limits metrics RSA II DMA was used to measure the stress relaxa-
are maintained. Inoue and Kobatake also developed a rela- tion pro®le of both the bimaterial strip and bulk adhesive
tion to describe the residual stress for a composite strip as a using a 3-point bending setup. The bimaterial strip is placed
function of the radius of curvature [22]. Their derivation with the concave surface facing up and the loading arm is
differed from that of Timoshenko in that each ®lament in placed at the center of the strip. During testing, a step strain
the bimaterial strip is assumed to be cemented tightly of 0.1% was applied to the sample. The choice of 0.1%
instead of only at the ends. strain was well within the linear viscoelastic regime for
In this work, we utilize a bimaterial strip along with the these samples. After application of the step strain, the load-
relation of Inoue and Kobatake to measure the curing stress ing arm was able to monitor the applied force as a function
as a function of an equilibrium modulus. Afterwards, a of time. This stress relaxation test was repeated with the
dynamic mechanical analyzer (DMA) is used to establish concave surface of the bimaterial strip facing down and
the stress relaxation pro®le of the bimaterial strip. The no difference in the relaxation pro®le was observed. A simi-
results are compared to the relaxation pro®le of the bulk lar approach allowed the stress relaxation pro®le to be eval-
adhesive. Finally, emphasis is placed on the role of the uated for the bulk adhesive molded into thin strips. The
measured modulus and how it relates to the curing stress. DMA was equipped with an oven that allowed these
measurements to be carried out over a relatively wide
2. Experimental temperature range.

2.1. Materials 2.5. Coef®cient of thermal expansion

The coef®cient of thermal expansion was evaluated on the

The epoxy resins used in the study were TIGA 321 w
bulk adhesive using a Perkin±Elmer Thermal Mechanical
manufactured by Resin Technology Group (RTG) and EA
Analyzer (TMA) 7. The instrument was out®tted with a ¯at
946 w manufactured by Dexter-Hysol. In addition, a simple
tipped expansion probe. Samples were tested from 275 to
epoxy system of 34 parts by weight Epon w 828, 55 parts
1508C at 108C/min.
by weight Versamid w 140, and 4 parts by weight TS-720
Cab-O-Sil w was considered.
2.5.1. Glass transition temperature
2.2. Preparation of bimaterial strips The onset Tg of the adhesive was measured using a
Mettler-Toledo differential scanning calorimeter (DSC)
The epoxies were vacuum mixed and then coated on a brass 821. Samples were tested from 275 to 1508C at 108C/min.
 D.J. Macon / Polymer 42 (2001) 5285±5291 5287

w adhesive. Fig. 1 shows the cross-section of the bimaterial

beam. The x-coordinate direction is perpendicular to the
plane of the paper and the y-coordinate is de®ned as positive
towards the bottom of the page. For a beam made of two
Adhesive
materials, the bending moment, M, is
Z Z
c Mˆ s xA y dA 1 s xB y dA …4†
A B
tA
where s xA is the stress of the adhesive acting in the x-direc-
tion, s xB is the stress of the brass shim acting in the x-
direction, y is the distance from the neutral axis plane, and
z Neutral Axis dA is the differential change in area. The subscripts A and B
will denote properties relating to the adhesive and brass,
y respectively. Making use of Hookes law, s ˆ Ee; and the
strain in the x-direction, e x ˆ y=r; Eq. (4) becomes
Brass  Z Z 
tB 1 2 2
Mˆ E y dA 1 EB y dA …5†
r A A B

Fig. 1. Coordinate systems of adhesive bonded to a brass shim as viewed If the observed moment of the bimaterial beam, M ˆ EI=r;
axially through the cross-sectional area. is used, Eq. (5) becomes

3. Results and discussion EI ˆ EA IA 1 EB IB …6†

R 2
where I ˆ y dA:Using the parallel axis theorem gives
3.1. Stress relaxation  2
1 3 1
IA ˆ wt 1 tA w c 2 tA
In a typical 3-point bending test, the relation between the 12 A 2
de¯ection at the center of the beam, d , and the applied load,
P, is  2
1 3 1
IB ˆ wtB 1 tB w tA 1 tB 2 c
PL3 12 2
dˆ …2†
48EI where c is the distance from the top of the section to the
where L is the length of the support span, E is Young's neutral axis, w is the width of the beam, tA is the thickness of
modulus of the beam, and I is the moment of inertia for the adhesive, and tB is the thickness of the brass.
the beam [24]. In the case of the bimaterial strip, the It is also noted that there are no external forces acting on
measured modulus as given by Eq. (2) will depend on the the strip from the x-direction, therefore,
moduli of both the adhesive and the metallic strip. If X Z Z
the effective modulus of the bimaterial strip is viscoelastic, Fx ˆ 0 ) s xA dA 1 s xB dA ˆ 0
A B
Eq. (2) changes. For a stress relaxation test, the displace-
ment term will remain constant, but the load becomes a and remembering that s xA ˆ EA y=r and s xB ˆ EB y=r; then
function of time, as does the ¯exural rigidity term, EI. the summation of forces becomes
The ¯exural rigidity of the bimaterial strip depends on the Z Z
viscoelastic behavior of both the brass shim and adhesive. EA y dA 1 EB y dA ˆ 0 …7†
A B
For the experimental times considered, the modulus of the
brass shim can be assumed constant with the time-depen- To perform this integral, Fig. 1 is divided into three
dent contribution to the measured modulus coming solely centroids. The ®rst is the area from the top of the section
from the adhesive. The moment of inertia will in turn to the neutral axis. The second is the area from the neutral
depend upon the neutral axis plane of the composite strip, axis to the adhesive and brass interface. The third is the area
which is also changing with time. For a stress relaxation test, of the brass shim. Eq. (7) then becomes
Eq. (2) becomes    
1 2 1 2 1
EA w 2 c 1 …tA 2 c† 1 EB tB w tA 2 c 1 tB ˆ 0
P…t†L3 2 2 2
EI…t† ˆ …3† …8†
48d
where P…t† and EI…t† are the applied load and ¯exural Eq. (8) can be solved for EA and then substituted into Eq. (6).
rigidity as a function of time, respectively. This along with the use of the ¯exural rigidity of the compo-
The change in ¯exural rigidity of the composite is related site beam as given by Eq. (3) allows the relaxation modulus
to the individual contributions from both the brass and of the bonded adhesive to be measured.
5288 D.J. Macon / Polymer 42 (2001) 5285±5291

4.0

295 K
3.5 266 K
277 K

305 K

3.0 315K
log E(t) (MPa)

2.5

2.0
Data Master Curve: Reference Temperature = 295 K
319 K

1.5
-2.0 -1.0 0.0 1.0 2.0 3.0 4.0 -10 -5 0 5 10

log time (sec)

Fig. 2. Master stress relaxation curve for TIGA 321 bonded to a brass shim as measured by 3-point bending test on TMA.

3.2. Master stress relaxation curve the shift factor [25]. This relation is
2C1 …T 2 Tg †
If a relaxation spectrum of the adhesive is measured log aT ˆ …9†
at different temperatures, a master curve can be C2 1 T 2 Tg
constructed by shifting the experimental logarithmic
where T is the temperature of interest, C1 and C2 are
curves along the time axis to obtain superposition.
experimentally determined constants. In addition, a
The distance shifted represents the logarithm of the
small correction is required in the form of a vertical
shift factor, aT : Williams, Landel, and Ferry proposed
shift due to temperature variations and changes in
an empirical equation for the temperature dependence of
volume resulting from temperature changes [26]. The
correction factor is
r…Tg †Tg
E…Tg ; t† ˆ E…T; t=aT † …10†
r…T†T
EA946 with brass shim
EA946 without brass shim
3.0
where E…Tg ; t† is the shifted modulus, r…Tg † is the
density measured at Tg ; r…T† is the density at tempera-
2.5 ture, T, E…T; t=aT † is the modulus measured at tempera-
log E(t) (MPa)

ture, T, and shifted along the time axis by aT : Since the

2.0 contribution from the mass to the density does not
change with temperature, Eq. (10) may be rewritten as
1.5
V…T†Tg
E…Tg ; t† ˆ E…T; t=aT † …11†
V…Tg †T
1.0

where V…T† is the volume of the body at temperature, T,

and V…Tg † is the volume of the body at the glass transi-
-6 -4 -2 0 2 4
log time (sec)
tion temperature.
Fig. 3. Master stress relaxation curve for EA 946 for bulk adhesive and It has been observed that the length of an isotropic body
adhesive bonded to brass shim. increases with temperature over a fairly broad temperature
 D.J. Macon / Polymer 42 (2001) 5285±5291 5289

3.5
3.5

TIGA321 with brass shim

3.0
log E(t) (MPa)

3.0

log E(t) (MPa)

TIGA321 without brass shim

2.5 2.5

Epon 828 + Versamid 140 with brass shim

Epon 828 + Versamid 140 without brass shim

2.0 2.0

-10 -5 0 5 10 -5 0 5 10
log time (sec) log time (sec)

Fig. 5. Master stress relaxation curve for Epon 828 1 Versamid 140 for
Fig. 4. Master stress relaxation curve for TIGA 321 for bulk adhesive and
bulk adhesive and adhesive bonded to brass shim.
adhesive bonded to brass shim.

tures but this value will not differ signi®cantly from the
range as [27]
measured long-term modulus. The adhesive used to gener-
ln li …T† ˆ ln li …T0 † 1 a…T 2 T0 † …12† ate Fig. 5 was cured above its ultimate glass transition
where li …T† is the length of the body at temperature, T, li …T0 † temperature (878C). This allowed higher test temperatures
is the length of the body at some reference temperature, T0 ; to be used and a more proper equilibrium modulus to be
and a is the linear coef®cient of thermal expansion. Also, measured.
indicial notation is used. Since V ˆ l1 l2 l3 ; then Eq. (11) For the curves shown in Figs. 3±5, the effect of the curing
when used in conjunction with Eq. (12) becomes stress on the relaxation spectra is of particular interest.
Kubat and Rigdahl noted, at conditions of low stress and/
Tg or temperature, the following relation for both metals and
E…Tg ; t† ˆ e3a…T2Tg † E…T; t=aT † …13†
T polymers in a stress relaxation test [29±33]
Eq. (13) can be used to shift the relaxation curves to make a s…t† 2 s i ˆ K…t 1 a†2n …14†
master curve. Fig. 2 illustrates a master relaxation curve for
TIGA 321 bonded to a brass shim. where s…t† is the time-dependent measured stress, s i is the
internal stress and is assumed independent of time, t is
3.3. Relaxation curve for bulk adhesive versus bonded the time, K, a, and n are constants. The term on the right
adhesive of the equal sign is a decaying power law that yields a
relaxation spectrum. The terms on the left indicate that the
Figs. 3±5 show the relaxation curves for EA 946, TIGA resulting relaxation curve is shifted vertically by a constant,
321, and Epon 828 1 Versamid 140 for both the bulk and
which in this case they ascribe to the internal stress. In
bonded resin. The most noticeable features are that the
polymers, relaxation phenomena are often interpreted in
relaxation moduli for the bonded adhesives are of greater
magnitude than for the bulk adhesive and are generally
separated by a constant. This phenomenon has been 0.0 EA946 Without Brass Shim

observed before when an adhesive modulus has been

EA946 With Brass Shim

measured in butt joints [28]. The measured or apparent -0.5

modulus of the adhesive is of greater magnitude than that

of the unbonded material. This is because the adhesive -1.0
log E(t)/E(t=0)

is constrained and the substrates are more rigid than the

adhesive [1]. -1.5

The relaxation curves shown in Figs. 3 and 4 do not show

a proper long-term equilibrium modulus. This is because -2.0

both adhesives are cured at 418C for 48 h which does not

allow either adhesive to completely cure. To prevent addi-
-2.5

tional chemical aging during the stress relaxation test, the -6 -4 -2

log time (sec)
0 2 4

test temperatures were restricted to slightly above the cure

temperature. A true equilibrium modulus is not reached Fig. 6. Normalized master stress relaxation curve for EA 946 for neat resin
during the stress relaxation test at the chosen test tempera- and adhesive bonded to brass shim.
5290 D.J. Macon / Polymer 42 (2001) 5285±5291

terms of relaxation time spectra, changes in free volume, 3.3

etc. That is to say in terms of approaches where internal Instantaneous modulus with shim
Instantaneous modulus without shim

stress does not appear explicitly. However if the internal Long term modulus with shim
Long term modulus without shim

stress is included in these interpretations, special care

must be taken on how this internal stress is described. 2.3

White indicates that the internal stress as written in

log F(m,n)
Eq. (14) can refer to two quite different physical phenomena
[34]. One type is a result of normal curing stresses found in
thermosets. The other type is often used in the ®eld of 1.3

physical metallurgy to indicate a resistance to deformation.

Polymers can behave in a phenomenologically similar
manner to that of metals though the responsible mechanisms
are almost certainly different. The two types of internal
0.3

stress are interactive and depend upon the processing 0 2 4 6 8 10 12 14

n = h A/h B
conditions.
Normally, stress relaxation can be written as Fig. 7. Structural function, F…m; n†; for EA 946 adhesive.
s…t† ˆ E…t†e o …15†
derived by Inoue and Kobatake [22]
where E…t† is the time-dependent relaxation modulus and eo
is the applied strain. Instead an effective relaxation modulus EB h3B 1
as a function of time, Eeff …t†; can be de®ned for a bimaterial si ˆ F…m; n† …18†
12hA r…hA 1 hB †
strip as
where EB is Young's modulus of the brass shim and is equal
Eeff …t† ˆ f …s i †Ebulk …t† …16† to 110 GPa; hA the thickness of the adhesive; hB the thick-
where f …s i † is a time-independent constant that depends at a ness of the brass shim; m ˆ EA =EB ; n ˆ hA =hB ; and F…m; n†
minimum on the internal stress of the material and Ebulk is a structural factor given by
the relaxation modulus of the bulk adhesive. Eq. (15) may
F…m; n†
be rewritten as
log s eff …t† ˆ log f …s i † eo 1 log Ebulk …t†eo …17† …1 2 mn2 †3 …1 2 m† 1 ‰mn…n 1 2† 1 1Š3 1 m…mn2 1 2n 1 1†3
ˆ
…1 1 mn†3
where s eff …t† is the effective measured stress for the
…19†
composite beam as a function of time. The log f …s i †eo
term is time-independent and will shift the relaxation In Eq. (18), the dependence of the internal stress on EA is
spectrum by a constant. This behavior is seen in Fig. 3 given by the function F…m; n†: As derived, EA is assigned to
where the lower curve was generated from the bulk the instantaneous modulus and does not allow for any stress
adhesive and the upper curve was measured for the adhesive relaxation. Under experimental conditions, the measured
in a stressed state. Similar behavior is seen in Figs. 4 and 5. equilibrium stress will depend on an equilibrium modulus
If both curves in Fig. 3 are normalized by the initial not an instantaneous one. The choice of the adhesive modu-
measured modulus, they can be superimposed over each lus value to be used strongly impacts the resulting measured
other as seen in Fig. 6. This indicates that a constant is all stress. Examination of Figs. 3 and 4 show that four possible
that differentiates the two different curves. moduli for a given adhesive can be substituted into Eq. (19).
When evaluating the engineering properties of the adhe- Two of the possible choices are the instantaneous moduli for
sive, it is important to consider whether the adhesive is the the bulk and bonded adhesive. The other two choices are the
same in bulk form as when bonded between substrates. long-term moduli for the bulk and bonded adhesive. Fig. 7
Curing stresses are present in greater magnitude when the shows log F…m; n† for the EA 946 adhesive versus the varia-
adhesive is cured in a constrained geometry. This in turn can tion in adhesive to substrate thickness …hA =hB † for the four
in¯uence the mechanical properties of the bulk adhesive different moduli. When the substrate is much thicker than
compared to the bonded material [2]. This behavior was the adhesive (i.e. n ! 0†; the differences in the four types of
observed by Knollman, who measured a change in shear adhesive moduli do little to affect F…m; n†: However as n
modulus as the distance from the adhesive/adherend inter- approaches the ratio of adhesive to substrate thickness used
face increased [35]. in this work …n ˆ 14†; the impact of the modulus choice
becomes signi®cant. The best choice for modulus is one
3.4. Measurement of curing stress of the long-term variety. More speci®cally, the effective
long-term modulus should be used because this is an accu-
The stress associated with the curing process, s i ; can be rate measure of the bimaterial beam's resistance to defor-
measured for a bimaterial strip using the following relation mation. The other measures for the moduli shown in Fig. 7
 D.J. Macon / Polymer 42 (2001) 5285±5291 5291

give an estimate of the possible error if the wrong Acknowledgements

modulus is chosen. Using an equilibrium effective
modulus for EA 946 of 8.9 MPa in Eq. (18) gives a The author is greatly indebted to D.L. Lamont who
curing stress measure of 0.75 MPa for EA 946. If the substantially contributed to the present work.
long-term modulus for the bulk adhesive is chosen, the
calculated curing stress would be underestimated by
25.5%. Either of the short-term moduli would over-
References
estimate the curing stress by several orders of magni-
tude. Using the long term effective moduli, curing [1] Dannenberg H. SPE J 1965:669.
stresses of 1.1 and 8.8 MPa were established for [2] Kim YG, Lee DG. J Adhes 1998;65:163.
TIGA 321 and Epon 828 1 Versamid 140, respectively. [3] Plepys AR, Farris RJ. Polymer 1990;31:1932.
It has been pointed out that for a similar chemical [4] Zihlif AM, Feldman L, Farris RJ. J Mater Sci 1989;24:3267.
system, different results for the curing stress have been [5] Ochi M, Yamashita K, Yoshizumi M, Shimo M. J Appl Polym Sci
1989;38:789.
obtained by different researchers [13]. The choice of [6] Paipetis SA. Colloid Polym Sci 1979;257:934.
modulus is one possible source of confusion. In some [7] Lange J, Toll S, Manson JE, Hult A. Polymer 1997;38:809.
instances, the instantaneous modulus for the bulk mate- [8] Igarashi T, Koudo S, Kurokawa M. Polymer 1979;20:163.
rial has been used [7,14]. If no other corrections are [9] Shimbo M, Ochi M, Arai K. J Coatings Technol 1984;56:45.
used, this will lead to an overestimation of the residual [10] Ochi M, Yamashita K, Shimbo M. J Appl Polym Sci
1991;43:2013.
stress. In other circumstances, an effective modulus [11] Ree M, Swanson S, Volksen W. Polymer 1993;34:1423.
measurement of the adhesive in the test geometry was [12] Kim YG, Lee DG. J Adhes 1998;65:163.
performed. But again an instantaneous modulus was [13] Shimbo M, Ochi M, Inamura T, Inoue M. J Mater Sci
measured, leading to a possible overestimation of the 1985;20:2965.
curing stress [1,23]. Finally, the contribution of the [14] King D, Bell JP. J Adhes 1988;26:37.
[15] Brahatheeswaran C, Gupta VB. Polymer 1993;34:289.
adhesive to the stiffness has often been neglected
[16] Wang HB, Yang YG, Yu HH, Sun WM, Zhang YH, Zhou HW. Polym
under the assumption that the adhesive layer thickness Engng Sci 1995;35:1895.
and modulus are much less than that of the metallic [17] Nakamura Y, Tabate H, Suzuki H, Iko K, Okubo M, Matsumoto T.
substrate [7±9,15,17,36]. These assumptions are reason- J Appl Polym Sci 1986;32:4865.
able as long as the conditions are met. But in checking [18] Farris RJ, Vratsanos MS. Int J Fract 1989;39:93.
[19] King D, Bell JP. In: Dickie RS, Labana SS, Bauer PS, editors. Cross-
the assumptions, the effective long-term modulus should
linked polymers: chemistry, properties, and applications. Washing-
be used, especially as the adhesive layer thickness ton, DC: American Chemical Society, 1988. p. 221.
increases. [20] Li JCM. Can J Phys 1997;45:493.
[21] Timoshenko S. J Opt Soc Am 1925;11:233.
[22] Inoue Y, Kobatake Y. Appl Sci Res 1958;7:314.
[23] Dillard DA, Park TG, Zhang H, Chen B. Paper Presented at Proceed-
4. Conclusions ings of the 22nd Annual Meeting of the Adhesion Society, 21±24 Feb
1999, Panama City Beach, FL, p. 336.
From the presented data, the following conclusions were [24] ASTM D 790.
[25] Williams ML, Landel RF, Ferry JD. J Am Chem Soc 1955;77:3701.
drawn:
[26] Tobolsky AV, McLaughlin JR. J Polym Sci 1952;8:543.
[27] Blatz PJ. Polymer networks: structural and mechanical properties.
1. The effective modulus of an adhesive in a bimaterial strip New York: Plenum Press, 1971.
will be greater than for the bulk adhesive and may depend [28] Rutherford JL, Bossler FC, Hughes EJ. Rev Sci Inst 1968;39:666.
on the extent of curing stress. [29] Kubat J, Rigdahl M, Int. J Polym Mater 1975;3:27.
2. The long-term effective modulus is the most accurate [30] Kubat J, Rigdahl M. Mater Sci Engng 1975;21:63.
[31] Kubat J, Petermann J, Rigdahl M. Mater Sci Engng 1975;19:185.
measure for determining the curing stress for a bimaterial
[32] Kubat J, Rigdahl M. Polymer 1975;16:925.
strip. [33] Kubat J, Petermann J, Rigdahl M. J Mater Sci 1975;10:2071.
3. Confusion in the literature regarding the curing stress [34] White JR. Polym Testing 1984;4:165.
may have arisen from incorrect measures of the adhesive [35] Knollman GC. Int J Adhes Adhesives 1985;5:137.
modulus. [36] Lange J, Hult A, Manson JE. Polym Mater Sci Engng 1993;68:222.