THE 2-YEAR CONCRETE PRISM TEST FOR ASR – IS IT WORTH THE WAIT?

David Stokes, FMC Corporation, david_stokes@fmc.com

Dan Johnston, South Dakota Department of Transportation, Dan.Johnston@state.sd.us

Roger Surdahl, Federal Highway Administration, Roger.Surdahl@fhwa.dot.gov

Sometimes concrete cracks...

Concrete cracks when the tensile stress within the concrete exceeds the tensile strength. The presence of such

discontinuities represents an impairment to the concrete‟s ability to withstand future stresses from the environment.

Tensile stresses within a concrete matrix arise from varied sources; both from a consequence of its service

in situ, and from chemical reactions that the concrete undergoes during its existence. Also, these sources of stress

are seldom, if ever, totally independent of one another. Wetting and drying of concrete exerts stress. Freezing and

thawing (when the concrete is sufficiently moist) exerts stress. Thermal gradients within the concrete exert stress.

Static and dynamic loadings exert stress. Chemical reactions within the cementitious matrix can exert stress

(delayed ettringite formation, for example). Chemical reactions within the concrete involving the aggregates used to

make the concrete can exert stress (alkali-silica reaction (ASR), for example). And so on.

No stress is an island unto itself…

Tensile stresses from different proximate causes are additive (at least, the components acting in the same direction

are). It makes no difference what the sources of the individual stresses are that combine to make up the overall

stress component that acts on a part of a concrete structure; the effects are simply additive. Figure 1 illustrates this

concept. An element of a concrete structure in service is depicted as a beam, supported on its ends, and with various

sources of stress depicted as weights on the center of the beam. The relative magnitudes of the different stress

sources shown in this abstract diagram would, of course, vary for different concrete compositions existing as

structural elements in different service environments, and would also be variable at any given moment in time.

The important point for the discussion in this paper is that even though a single source of stress in a

concrete element may not be sufficient to crack the concrete on its own, if the sum of the stresses is larger than the

ability of the concrete to withstand the overall stress component, then the concrete will crack. And it is logically

inconsistent to try to assign the root cause of the resulting crack to one particular source of stress, except perhaps in

rare cases where the magnitude of one source of stress far outweighs the magnitude of others that are invariably

acting (at least to some degree) in a given element of a concrete structure. Generally speaking, there is usually no

single source of stress that breaks the concrete‟s back, when actual concrete in service is being considered.

Meanwhile, back at the lab…

However, with laboratory testing, one can study systems that are very different from real world conditions. In fact,

the converse is usually never achieved – that‟s why there are field studies in addition to laboratory studies. In the

lab, messy, real-world complications can often be eliminated from systems under study, allowing certain aspects of

various phenomena to be considered in isolation from otherwise large realms of complexity.

This is the case with ASR-related specification testing. Most sources of stress that concrete structures will

have to endure in service are removed to fairly negligible levels, in order to isolate the effects of ASR from all other

effects. The two main types of tests in use in specifications today are the accelerated mortar bar test (AMBT) and

the concrete prism test (CPT).

These tests show up in slightly different varieties in specifications around the world, but two significant

examples of the AMBT are ASTM C1260

(for evaluating aggregates) and ASTM C1567

(for evaluating

supplementary cementitious materials combined with aggregates). For the CPT, the main example is ASTM

C1293

. ASTM C1293 is for testing aggregates, but it is also recommended by many for testing SCMs as well.

When testing SCMs, the duration is generally recommended to be extended to two years

, instead of the one year

required for aggregate testing only.

In both the AMBT and the CPT, high levels of alkali are used to drive the reactions, which, in turn, produce

tensile stresses, and, often, net expansions of the mortar or concrete samples being evaluated. In addition, the

temperature of the AMBT storage condition is higher than that which is experienced by concrete in most civil

engineering applications. In general, the only significant tensile stresses that occur in these two tests are from ASR

that may (or may not) develop in the specimens.

As ASR progresses in a given concrete in field applications, alkali is consumed by the reaction and thus the

pH of the pore solution decreases in time. In the AMBT, the mortar bars are submerged in a 1N NaOH solution, and

while alkali is consumed during the test, the decrease in the solution concentration across the usual duration of the

test would be less than it would be for concrete after decades in the field. This is then a stronger driving force

(combined with the higher temperature) for the reaction than would be the case in the field. (Even though some

concretes in actual service may have higher initial pore solution concentrations than the soak solution in the AMBT,

because the overall mass of the alkali in the AMBT system as a percentage of the cementitious binder is much

higher than any concrete, the relatively high concentration falls off more slowly, and to a lesser degree than it would

in concrete after decades of field service.) But, like the CPT, the driving force for the reaction in the AMBT is

primarily a chemical driving force. In the CPT, the alkali to drive the reaction is included within the mix, and so is

closer to what would be the case for concrete in service. However, not only does the pH of the pore solution

decrease because alkali is consumed in the reaction, but significant leaching occurs as well, which results in the pH

of the pore solution decreasing more quickly in the CPT than it does in actual field concrete. And, since the

temperature of the CPT test is only 38

C and no wetting or drying is occurring, this test tends to underestimate the

extent of the reaction that would take place in a field concrete made with the same mix as the test.

Thus, for equivalent mortars or concretes made the same as the specimens for the AMBT and the CPT, and

even in the absence of the actual additional stresses that the field concrete would be exposed to in service, the

AMBT is harsher than field service, while the CPT is milder than field service. The only way for the CPT to be

harsher than the exposure it is supposed to represent is if the concrete it characterizes has a significantly lower alkali

loading, and does not experience major stresses in the field. One example where this might be the case is for

concrete inside of large structural elements, where thermal stresses are minimal, wetting and drying is minimal,

freezing and thawing is minimal, and no significant fatigue loadings are experienced in service, and/or for

significantly lower alkali loadings. Other than for the case of reduced alkali loadings, which may or may not be the

case, this minimal stress condition generally excludes pavements.

A grim fairy tale…

Non-mandatory information in ASTM C1293 proposes using 0.04% expansion in the CPT as a limit defining

deleterious from non-deleterious expansions in concrete, and it has also been proposed to use 0.1% at 14 days in the

soak solution in the AMBT as a limit which corresponds well to the CPT result. In a 2005 PCA publication

, a case

was made defending these limits, and is one of the most thorough presentations of the evidence on this subject

known to the authors of this paper. (This 2005 PCA publication also appeared in a slightly edited form in an ACI

journal in 2007

.) Particular attention was given (in both versions of the paper) to field cases in support of the limits

proposed by the authors.

Four field cases were discussed in the PCA paper in support of the test limits. Only one case described a

structure with a significant age (30 years). The other three were 7 to 10 years of age. In the bulk of the work, actual

samples matching those used in the field were not tested –either similar materials were obtained, and/or materials

from the same sources were acquired years later and tested. One of the sites discussed used low alkali cement, and

the CPT and AMBT indicated higher levels of SCMs than would be necessary, given the condition at 10 years of

age. (This last supporting case was not mentioned in the 2007 ACI journal version of the paper.) Assuming these

are the best examples that exist in support of the 2-year limit, the evidence from the field is rather limited.

The other significant discussions in the paper revolved around comparisons of the AMBT, the CPT, and

blocks and slabs stored outdoors with mixes that were the same as the CPT, and for blocks and slabs with lower

levels of alkali than the comparable CPT (these specimens had no static or dynamic loading). One of the studies

mentioned was with a single aggregate from the U.K; many aggregates were tested and discussed from a Canadian

study. The case was made here, and it has been made elsewhere, that a significant correlation exists between the

0.1% expansion limit at 14 days in the AMBT and the 0.04% expansion limit in the CPT at 2 years

However, a closer look at the evidence from the outdoor storage specimens and the 2-year CPT result can

be made by looking at a list of results from this Canadian study that was published in 2004

. Table 4 (in that

reference

) discusses 57 mixes in the study (which is a subset of the actual numbers of mixes in the program). Some

of these mixes have more or less alkali than the standard CPT. Of the mixes that were standard, approximately 25%

passed the 2-year limit, but are exceeding that limit in the comparable slabs (with the same alkali content).

Approximately 20% of the standard mixes passed the AMBT 14-day limit, but are exceeding 0.04% in the slabs.

Thus, there is correlation between the 14-day AMBT limit and the 2-year CPT limit, but both show significant

failure to predict the excessive expansion occurring in the companion slabs stored outdoors. The age of the outdoor

specimens at the time of this comparison was 8 to 11 years. It is clear that the 2-year limit in the CPT significantly

underestimates the potential for expansion with an equivalent concrete stored outdoors. Also, remembering the

earlier discussion above, while the outdoor specimens at least experience wetting and drying cycles, freezing and

thawing cycles, and thermal gradients, these would be considerably less challenging than the thermal gradient-

related stresses that develop in actual pavements, and with none of the mechanical fatigue loadings that pavements

experience. (In addition, pavements in colder climates may also be exposed to deicing chemicals which can

exacerbate ASR-related damage.)

To be fair, the authors of the papers discussed above only recommend the 2-year CPT limit for concretes

made with a maximum 1% total alkali portland cement. However, it is clear that the CPT underestimates what

happens in their outdoor specimens significantly for the same alkali loading as is used in the CPT test method at

only 10 years of age, and this is with specimens that see no significant loading compared to a highway or airfield

pavement. However, in a paper discussing an alternative use of the data from these tests given in 2006

, some other

data from the same Canadian program clearly showed an example where the 2-year limit failed to predict failure in

outdoor specimens both for the same alkali loading as the standard CPT and for the lower 0.9% total alkali

specimens.

Figure 2 shows a typical, but stark, example of the difference that traffic loadings can make in the

development of ASR-related distress in pavements. Pictured are a driving lane and a shoulder of a 12-year old

concrete interstate, as it is drying out from a recent rain (which better highlights the cracking in the pavement). Both

driving lanes and the shoulder were paved at the same time, and so contain exactly the same concrete. The

macroscopic ASR-related distress is considerably worse in the driving lane, and practically non-existent in the

shoulder. The outdoor storage specimens used as proof of concept in the Canadian study, are like the shoulder, and

underestimate the challenge to the actual concrete in service in the driving lanes posed by the additional stress to the

pavement from the ongoing ASR.

The expansion limit of 0.04% in the CPT is considered to be an expansion level that correlates to the point

where macroscopic cracking becomes evident. Promoters of the 0.04% expansion limit in the 2-year CPT are using

the result to suggest that if ASR does not develop sufficiently in the 2-year test to expand to 0.04% and thus crack

the concrete, then no significant ASR will develop in a concrete with a comparable mix design. What they fail to

realize is the point illustrated in Fig 1 – that ASR does not have to crack the concrete on its own to be a problem for

the service life of the concrete. If the concrete expands to 0.03% in the 2-year concrete prism test, then the ASR on

its own has not developed enough to crack the concrete on its own, but is still producing significant tensile stress

within the concrete. That stress, combined with other stresses that may be present in the concrete in service

(particularly where pavements are concerned) can be sufficient to begin cracking the concrete – even if the ASR-

produced stresses could not do it on its own. This severely compromises the utility of the 0.04% limit in the 2-year

CPT for specification use – particularly for pavements.

Since the use of 0.1% expansion at 14 days in the AMBT correlates more or less with the 2-year CPT limit,

it too has significant potential for being too risky for general use in preventing significant levels of ASR, particularly

in pavements.

Since the AMBT can be more severe than the CPT, particularly for longer storage in the soak solution than

14 days, it allows for a more rigorous specification than does the 2-year CPT. Table 1 shows a comparison of some

more results from the Canadian program discussed above, including many that were in the 2004 paper referenced

above. Using the higher alkali outdoor slab specimens as the standard, Table 1 gives a list of results comparing the

0.1% 14-day limit in the AMBT, the 0.04% 2-year limit in the CPT, and a limit of 0.08% at 28 days in the AMBT.

Here, a false negative means that the limit was passed in the tests, but the slab has exceeded 0.04% expansion. Out

of 28 mixes (8 different aggregates, with and without various levels of fly ash or silica fume) using 0.1% at 14 days

as a limit in the AMBT gave 42% false negatives, using 0.04% at 2 years in the CPT gave 32% false negatives, and

using 0.08% at 28 days in the AMBT gave 7% false negatives. Thus 0.08% at 28 days in the AMBT was in much

better agreement with the outdoor slab results than the 14 day limit of 0.1% in the AMBT and the 2-year 0.04% limit

in the CPT. (At the time of this writing, the specification limit for military airfield pavement was „not to exceed

0.08% at 14 days‟. Use of this criteria in place of the more traditional 14 day limit of 0.10%, yields 29% false

negatives with this data set.)

Also with this data set, using a limit of less than 0.05% at 14 days in the AMBT would give a pass/fail

correlation equivalent to the use of 0.08% at 28 days, but unfortunately the coefficient of variation with the AMBT

for values below 0.1% at 14 days rises exponentially

, and thus a specification this low at 14 days would have

unacceptably high multi-lab variability, in the opinion of the authors of this paper.

Because of actual failures in the real world using 0.1% at 14 days in the AMBT, the NW Region of the

FAA has moved to a 28-day limit, as well as some state DOTs.

One size fits all…

The two most significant factors affecting whether or not ASR will develop to any significant level in a given

concrete are the reactivity of the aggregate and the alkali loading of the concrete (assuming sufficient moisture will

be present to supply the ongoing reaction). Because ASR takes so long to develop in the field, accelerated methods

must be used to facilitate its occurrence in laboratory specimens. Both the CPT and the AMBT test at one alkali

level only, and that is quite high compared to most concrete as it exists in actual practice. Also, only a single mix

design is proscribed by the tests; actual mix designs are not evaluated. Similarly, there are no provisions to test

differently, or to use different limits for different exposure or service conditions. Thus, one size fits all. And thus

for many situations, the level of prevention that will satisfy the test may be overly conservative, but until either

different testing becomes available or more useful methods become available to interpret the test data, this will be

the situation.

Horseshoes and hand grenades…

Both because the test result has to apply to many different mix designs, alkali loadings, and service conditions, and

the fact that there is an error level associated with any test, the results of these tests may be considered to, at best,

yield only an approximate value for the necessary level of mitigation that will need to be employed. Figure 3 shows

the results from a recent paper of 2-year CPT testing with an ASR-susceptible aggregate with different levels of a

lithium admixture

. The data from the paper show that a 150% dose of lithium admixture was necessary to control

this aggregate in the 2-year CPT. The precision statement for ASTM C1293 states that the “results of two properly

conducted tests in different laboratories on the same aggregate should not differ from each other by more than 65 %

of their average, nineteen times out of twenty.” These lines are plotted also in Figure 3. Looking at this range in

acceptable results from two labs show that, in this case, another lab could have obtained a passing result at 100%

dose of lithium admixture, and this would be acceptable variation within the precision statement, or another lab

could have obtained the result that a 150% dose of lithium admixture did not pass, and this would be acceptable

within the multi-lab precision statement. So without even considering the variability of the situations the test

prediction is supposed to apply to, the values obtained from the CPT test itself are not terribly precise.

Practically speaking…

Given that the supporting data from the field is rather sparse, and that the 2-year CPT fails to predict expansions of

companion specimens in outdoor storage to a considerable degree, and that the answer that is obtained is not very

precise, and adding to this the significant inconvenience of waiting two years for a test result, the 2-year concrete

prism test is simply not worth the wait. The authors support the use of 28 days in the AMBT, with a limit of less

than 0.08% expansion.

Acknowledgement

Thanks to CANMET for the use of data from their excellent research program in some examples discussed in the

paper.

References

1) ASTM Standard C 1260, 2007, “Standard Test Method for Potential Alkali Reactivity of Aggregates (Mortar-Bar

Method),” ASTM International, West Conshohocken, PA, www.astm.org.

2) ASTM Standard C 1567, 2007, “Standard Test Method for Determining the Potential Alkali-Silica Reactivity of

Combinations of Cementitious Materials and Aggregate (Accelerated Mortar-Bar Method),” ASTM International,

West Conshohocken, PA, www.astm.org.

3) ASTM Standard C 1293, 2006, “Standard Test Method for Determination of Length Change of Concrete Due to

Alkali-Silica Reaction,” ASTM International, West Conshohocken, PA, www.astm.org.

4) CSA A23.2-00, “Methods of Test for Concrete,” Canadian Standards Association, Mississauga, Ontario, Canada,

2000.

5) Thomas, M.D.A., Fournier, B., Folliard, K., Shehata, M., Ideker, J., and Rogers, C., Performance Limits for

Evaluating Supplementary Cementing Materials Using the Accelerated Mortar Bar Test, PCA R&D Serial No.

2892, Portland Cement Association, Skokie, Illinois, USA, , 2005, 22 pages.

6) Thomas, M.D.A., Fournier, B., Folliard, K., Shehata, M., Ideker, J., and Rogers, C., “Performance Limits for

Evaluating Supplementary Cementing Materials Using Accelerated Mortar Bar Test”, ACI Materials Journal, V.

104, No. 2, March-April 2007, pages 115 to 122.

7) Thomas, M.D.A., and Innis, F.A., “Use of the accelerated mortar bar test for evaluating the

efficacy of mineral admixtures for controlling expansion due to alkali-silica reaction,”

Cement, Concrete, and Aggregates, Vol. 21, No. 2, 1999, pages 157 to 164.

8) Fournier, B.; Nkinamubanzi, P.-C.; and Chevrier, R., “Comparative Field and Laboratory Investigations on the

Use of Supplementary Cementing Materials to Control Alkali-Silica Reaction in Concrete,” Proceedings of the

Twelfth International Conference Alkali-Aggregate Reaction in Concrete, V. 1, T. Mingshu and D. Min, eds.,

International Academic Publishers/World Publishing Corp., Beijing, China, 2004, pp. 528-537.

9) Stokes, D., “Concerning the Use of Expansion Data from ASR Testing”, Proceedings of the 8

CANMET/ACI

International Conference on Recent Advances in Concrete Technology, Marc-Andre Berube Symposium, Montreal

Canada, May-June 2006, pages 93 to 109.

10) Tremblay, C., Berube, M., Fournier, B., Thomas, M.D.A., and Folliard, K., “Effectiveness of Lithium-Based

Products in Concrete Made with Canadian Natural Aggregates Susceptible to Alkali- Silica Reactivity”, ACI

Materials Journal, V. 104, No. 2, March-April 2007, pages 195 to 205.

11) Peters, V., 1999. Draft Report „Precision Testing of AASHTO Test Method T 303 for Alkali Silica Reactivity.‟

Alkali-Silica Reactivity Interlaboratory Testing: SHRP Concrete and Structures ASR Showcase Test and Evaluation

Project 34. Federal Highway Administration, Washington, D.C.

List of Tables and Figures:

Tables:

1) This table shows comparisons of AMBT, CPT, and outdoor slabs from an ASR research program in

Canada (from Benoit Fournier, from a personal communication in 2005).

Figures:

1) An abstract diagram illustrating multiple sources of tensile stresses experienced by concrete in service.

2) A photograph of an interstate pavement illustrating the different levels of ASR-related distress between the

shoulder and the driving lane.

3) This diagram illustrates the effect that the multi-lab precision has on the approximate nature of determining

a safe level of ASR mitigation in ASTM C1293.

Table 1: Comparison of ASR Tests and Outdoor Specimens – Passing (P) or Failing (F) is marked in each of the

test columns. A passing result is one that is < 0.10% for the 14-day AMBT, < 0.08% for the 28-day AMBT, <

0.04% for the 2-Year CPT, and < 0.04% in the slab. Using the slab results as the „true‟ value, using 0.1% at 14 days

in the AMBT gave a false negative (giving a passing result when the slab is giving a failing result) 42% of the time

in this data set, using 0.04% at 2 years in the CPT gave a false negative 32% of the time in this data set, and using

0.08% at 28 days in the AMBT gave a false negative only 7% of the time with this data set. (At the time of this

writing, the specification limit for military airfield pavement was „not to exceed 0.08% at 14 days‟. Use of this

criteria in place of the more traditional 14 day limit of 0.10%, yields 29% false negatives with this data set.)

Mixture

14-day AMBT

28-day AMBT

2-year CPT

Outdoor Slab

(P/F)

%exp

(P/F)

%exp

(P/F)

%exp

(P/F)

%exp

Aggregate A

0.391

0.617

0.171

0.255

Aggregate A w/ 10% silica fume

0.142

0.402

0.038

0.086

Aggregate B

0.278

0.464

0.139

0.110

Aggregate B w/ 20% class F fly ash

0.048

0.125

0.008

0.044

Aggregate B w/ 30% class F fly ash

0.021

0.040

-0.007

0.030

Aggregate B w/ 7.5% silica fume

0.112

0.282

0.030

0.042

Aggregate B w/ 10% silica fume

0.078

0.225

0.023

0.018

Aggregate C

0.093

0.193

0.226

0.176

Aggregate D

0.173

0.212

0.087

0.171

Aggregate D w/ 7.5% silica fume

0.089

0.169

0.036

0.103

Aggregate D w/ 10% silica fume

0.083

0.178

0.030

0.075

Aggregate D w/ 12.5% silica fume

0.066

0.141

0.024

0.085

Aggregate E

0.463

0.700

0.270

0.395

Aggregate E w/ 20% class F fly ash

0.065

0.137

0.050

0.145

Aggregate E w/ 30% class F fly ash

0.034

0.036

0.021

0.087

Aggregate E w/ 7.5% silica fume

0.130

0.276

0.079

0.152

Aggregate E w/ 10% silica fume

0.120

0.284

0.039

0.081

Aggregate F

0.360

0.587

0.092

0.158

Aggregate F w/ 20% class F fly ash

0.037

0.067

0.016

0.048

Aggregate F w/ 30% class F fly ash

0.026

0.041

0.011

0.019

Aggregate F w/ 7.5% silica fume

0.090

0.222

0.028

0.062

Aggregate F w/ 10% silica fume

0.068

0.177

0.028

0.037

Aggregate G

0.419

0.603

0.221

0.254

Aggregate G w/ 20% class F fly ash

0.175

0.271

0.040

0.101

Aggregate G w/ 20% class F fly ash

0.062

0.091

0.020

0.038

Aggregate H

0.854

1.040

0.231

0.400

Aggregate H w/ 20% class F fly ash

0.395

0.461

0.085

0.247

Aggregate H w/ 30% class F fly ash

0.088

0.141

0.050

0.128

Figure 1. Abstract depiction of many sources of stress combining to form an overall level of stress that exceeds the

concrete‟s tensile strength. The individual sources of stress are insufficient to crack the concrete on their own.

Figure 2. Photograph of an interstate pavement drying out after a recent rain (which highlights the ASR-related

cracking in the pavement). The driving lane and the shoulder (and the passing lane) were all paved monolithically –

they are precisely the same concrete, placed at precisely the same time. The difference in distress is due primarily to

greater mechanical loadings on the driving lane. The shoulder is more similar to cast test specimens simply stored

outdoors. This highlights the fact that concrete simply stored outdoors underestimates the demanding service

conditions of a pavement.

CPT Multilab Precision (95% conf.)

(from ASTM C1293)

0.1

0.2

0.3

0.4

0.5

0.6

0 20 40 60 80 100 120 140 160

% Li dose

2 yr CPT

Limit @ 2yrs is 0.04%

Given this set of test results ...

… duplicates from another lab

should be expected to fall within

these limits 19 times out of 20

100% passes on the low end

150% does not pass

on the high end

Figure 3. The effect of multi-lab precision on the approximate nature of the result. In this case, various lithium

admixture dosages were tested with a reactive aggregate. The test showed that about 150% dosage of lithium

admixture was required to pass the test. However, based on the multi-lab precision statement, another lab could

have tested 100% and obtained a lower, passing result, and neither value would be considered more correct than the

other based on the precision statement in the test. Similarly, another could have tested the 150% dose and obtained

a higher, failing result, and it would also be considered as „correct‟ as the original passing result. (Precision lines are

based on Reference 3; the plotted data are from Reference 10.)