Details of the new test method are described elsewhere [9]; briefly, it involves two steps: a preconditioning step and the property measurement step. In the

test, the two metal beams containing the sealant specimen are pulled in tension in the direction perpendicular to the long axis of the specimen. The strain his­tory is schematically shown in Fig. 2. In the first step, the specimens were sub­jected to two loading-unloading-recovery cycles. The motivation for this step is to quantify the Mullins effect and eliminate its influence in the subsequent char­acterization measurement. In order to do this, the maximum strain in the two cycles must be larger than any strain seen in the specimen’s previous history, and the same maximum strain must be used whenever the procedure is con­ducted, so results can be compared. For the experiments here, a maximum ten­sile strain of 26% was employed. This strain level was chosen because it exceeded the typical test movements of ± 12.5% and ±25% used in ASTM C719. In addition, 26% strain should not be large enough to introduce any damage into the specimen. The loading-unloading tests utilized a crosshead speed of 2.64 mm/min so that the total time under load (to in Fig. 2) was 150 s. In order to allow for viscoelastic recovery, the specimen was held at 0% strain for 1500 s (10to) between cycles and before the next step. The test procedure developed in Ref 9 assumes that there is complete or nearly complete recovery in the time pe­riod. The criterion was that the compressive stress required in order to maintain zero strain at the end of recovery be less than 1.5% of the maximum stress achieved during the tensile cycles. To evaluate the Mullins effect [9], the loading curves on the two cycles were compared, and the magnitude of the effect was defined as the fractional drop in stress between the first and second loading curves at a particular strain level kx

Magnitude of Mullins = ff1(kx) r2(kx) (1)


where r1 and r2 represent the stresses during the first and second loadings. At very low strains, the magnitude of the Mullins effect is difficult to determine because the experimental uncertainty is very high. In this range, small differen­ces in the position of zero strain generate large differences in the Mullins effect. In contrast, as the strain approaches the maximum value achieved during the

FIG. 2—Strain history used for Mullins cycles and stress relaxation tests.

Copyright by ASTM Int’l (all rights reserved); Tue May 6 12:07:08 EDT 2014 Downloaded/printed by

loading cycle, the Mullins effect approaches zero. Between these extremes, how­ever, there is a range of strains at which the magnitude of the Mullins effect remains relatively constant, changing by 10% or less. In the experiments here in which the maximum strain in the loading curves is 26%, the magnitude of the Mullins effect was determined at strains of 10%, 15%, and 20%.

Once the Mullins effect was quantified and eliminated, the viscoelastic properties of the sealant were measured in the second step of the procedure using a stress relaxation experiment (see Fig. 2). The specimens were loaded rapidly (70 mm/min) in tension to a maximum tensile strain of 18%, which was chosen arbitrarily between two limits, i. e., it must be significantly less than the strain level used in the Mullins cycles while not being so low that accurate meas­urements of strain and load become difficult. Once this strain level was attained, the specimen was held at that strain while the load was monitored as a function of time. The specimen reached the hold strain in just under 1 s. Data points dur­ing the first 5 s after loading commenced were ignored in order to eliminate confounding effects due to our inability to instantaneously load a specimen to the predetermined strain.

From the stress relaxation data, Ea was calculated using a relationship based on the statistical theory of rubberlike elasticity [9,11-13]


W and B = width and breadth of the sealant (Fig. 1),

L = load, t = time, and

k = extension ratio, which is given by

k = 1 + H (3)


A = crosshead displacement, and H = undeformed height of the sealant.

Results and Discussion

The results from the visual observations made in the ASTM round robin after exposures are as follows: Sealant B exhibited complete adhesive failure during the test, with separation primarily between the sealant and the metal beam (Fig. 3). Moreover, these specimens’ dimensions exhibited permanent deforma­tion as seen in Fig. 3. Sealant A did not fail, but it also exhibited permanent de­formation. In addition, Fig. 4 shows that the metal beams were no longer parallel, suggesting that specimen loading might not have been symmetric. Other than the permanent deformation, however, sealant A showed no signs of

Copyright by ASTM Int’l (all rights reserved); Tue May 6 12:07:08 EDT 2014 Downloaded/printed by

Rochester Institute Of Technology pursuant to License Agreement. No further reproductions authorized.

cracking, debonding, or color change. Sealants C and D displayed no visual changes over the stress relaxation period.

During the Mullins cycles, all four sealants met the criterion for complete or nearly complete recovery. This was true despite the observation above that sug­gests sealants A and B might exhibit permanent deformation if to is sufficiently large. Figure 5 shows the average magnitude of the Mullins effect for two sam­ples of each sealant determined at a strain of 15%. Average values from tests on the exposed samples are shown as the cross-hatched area in each bar (no exposed data are available for sealant B specimens because they failed during the exposure tests). Similar results were obtained at strains of 10% and 20%. Note first that sealants A and D show a larger Mullins effect than sealants B and C. This indicates that the network structures in A and D have more junction points that can be disrupted by strains of 26% than do those in sealants B and C. Tests on specimens after exposure show a Mullins effect that is much less than that observed for fresh specimens but well above zero. Two hypotheses could explain this observation, either when considered on their own or in com­bination. First, if in the exposure tests the maximum strain never reached 26%,

Copyright by ASTM Int’l (all rights reserved); Tue May 6 12:07:08 EDT 2014 Downloaded/printed by


Sealant Material

some Mullins effect would be expected in the characterization experiments. Sec­ond, a number of weeks passed between the exposure tests and the characteriza­tion experiments, and the specimens were under no load during this period. It has been shown [9] that many sealants recover (at least partially) some of the Magnitude of Mullins effect during such a period with no strain applied.

Figure 6 shows stress relaxation curves for all four fresh specimens. A wide variation in the stress relaxation modulus curves for the four sealants is appa­rent. Sealants A and D are virtually indistinguishable over the range of times

FIG. 6—Stress relaxation curves of fresh specimens of sealants A, B, C, and D.

Copyright by ASTM Int’l (all rights reserved); Tue May 6 12:07:08 EDT 2014 Downloaded/printed by

tested. Relative to the range of modulus seen in previous sealant testing [9], sample B is near the soft end of the range, whereas sealant C is near the firm end. At very short times, sealants A and D display an upturn toward a glassy modulus, suggesting that they might have higher glass transition temperatures than sealants B and C. The curve for sealant C shows a slight downturn at long times, suggesting that flow-like behavior or a secondary relaxation mechanism might occur in these specimens over extended periods of time. Sealants A, C, and D show a clear rubbery plateau, whereas sealant B exhibits a continuous decrease in modulus through the rubbery zone. This decrease, combined with the relatively low modulus in sealant B, suggests its network structure might have fewer junction points than the networks in the other sealants.

All three sealant B specimens failed during the stress relaxation period, so it was not possible to characterize the behavior of this sealant over the specified stress relaxation period. It is worth noting, however, that the lack of a plateau in the curve shown in Fig. 7 for this sealant is consistent with the generation of a permanent deformation in the specimen that is held under load for some time. It would be interesting to perform diffusion studies to see whether the relatively low density of junction points in the network might facilitate migration of envi­ronmental species, such as water, into the specimen, because this could weaken the interface between the sealant and metal.

As noted above, the visual examination of the exposed C and D specimens revealed no visible changes in the physical appearance of these sealants. The characterization curves, however, show that molecular changes have occurred in these sealants. The stress relaxation behaviors after exposure are dramati­cally different, as shown in Figs. 7 and 8. Also included in these plots are the data from the tests on fresh specimens for comparison. The relaxation curves for sealant D shift down to approximately half of the values of the unexposed counterparts, though the shape of the relaxation curve remains unchanged. This means that the time-dependence of the apparent modulus is unaltered by exposure, at least in the range examined here. Although the precise mechanism

FIG. 7—Stress relaxation curves of fresh and degraded specimens of sealant D.

Copyright by ASTM Int’l (all rights reserved); Tue May 6 12:07:08 EDT 2014 Downloaded/printed by

FIG. 8—Stress relaxation curves of fresh and degraded specimens of sealant C.

governing the decrease in apparent modulus remains ambiguous, the results support a view that structural changes after exposure are brought about by a reduction in the density of effective junction points in the network structure. Likewise, the magnitude of apparent modulus for sealant C decreases after ex­posure by about 30%. Although the shapes of the curves are similar, there is a less distinct downturn at long times than seen with the fresh sealants.

As stated above, exposure of sealant A produced no cracks, debonds, or color change, but some permanent deformation was present (Fig. 4). Two of the three exposed specimens were deformed to the point that further exposure was not possible. Results of experiments on the third specimen are shown in Fig. 9. As with sealants C and D, the relaxation behavior for this specimen shows a dra­matic change after exposure. Specifically, a noticeable increase in apparent modulus and substantial change in curve shape are clearly evident. The plateau

FIG. 9—Stress relaxation curves of fresh and degraded specimens of sealant A.

Copyright by ASTM Int’l (all rights reserved); Tue May 6 12:07:08 EDT 2014 Downloaded/printed by

region at long times is completely absent. This change in the time dependence of the apparent modulus indicates that drastic structural modification has occurred in the exposed specimens. To ensure the reproducibility of this result, the characterization was performed for a second time, and the relaxation curves for the two runs are virtually identical (Fig. 9).

In order to test this result, attempts were made to forcefully return the two highly deformed specimens of sealant A to their original dimensions. Wedges and clamps were inserted in order to achieve the original shape, and the speci­mens were held in this way for several weeks. After the wedges and clamps were removed, the samples were allowed to set for several days. Some of the deforma­tion returned, but not enough to make the resumption of testing possible. Stress relaxation measurements were then performed on the specimens, and the results are shown in Fig. 10. The new data show trends similar to what was found with the initial exposed sample: the apparent moduli showed an increase, and the curve shape changed significantly. Although the differences between the new curves and the curve for the first exposed specimen could be the result of sample-to-sample variation, it is far more likely that the differences are arti­facts introduced by testing deformed specimens. Consequently, although tests on deformed specimens might be useful for showing general trends, quantita­tive comparisons should be avoided.

The relaxation curves for fresh sealants A and D are identical (see Fig. 6). Moreover, neither shows any cracking, debonding, or color change after expo­sure. The stress relaxation tests, however, show that the effect of exposure on the two sealants is completely different. Whereas sealant D exhibits a decrease in modulus with no change in curve shape, sealant A shows an increase in mod­ulus and a dramatic change in curve shape (see Figs. 7 and 9). This result clearly demonstrates the different viscoelastic response of the sealants, which is not surprising given that their chemistries and formulations are different. This is

Baseline 1 Baseline 2 Exposed 2 Exposure 2 run 2 Exposure 1 ad| Exposure 3 adi


Time (s)

FIG. 10—Different stress relaxation curves obtained from forcing distorted degraded specimens of sealant A to return to their original dimensions. The data for fresh speci­mens are also included for comparison.

Copyright by ASTM Int’l (all rights reserved); Tue May 6 12:07:08 EDT 2014 Downloaded/printed by

Rochester Institute Of Technology pursuant to License Agreement. No further reproductions authorized.

potentially important information that could not be obtained from conventional visual inspection.


A test method for assessing the durability of building joint sealants using a stress relaxation approach has been examined. Specimens of four commercial sealant materials that underwent exposure conditions according to ASTM C1519-10 as part of a round robin were obtained and utilized as model systems in order to compare this new test method with the current, descriptive method­ology involving visual inspection for defects. The results here show that the new test method not only allows meaningful quantitative evaluation of sealant char­acteristics but also provides qualitative information about the molecular struc­ture of the sealants. It has been shown that important additional information that is not provided by visual inspection can be obtained by using this test method. This is particularly evident in the results for sealants C and D, for which visual inspection fails to reveal any changes after exposure even though significant changes are detected by the stress relaxation measurements. Conse­quently, the results here indicate that the viscoelastic characterization provides a robust methodology for quantitatively evaluating the durability of building joint sealants.