ALKALI AGGREGATE REACTION IN NZ CONCRETE: MINIMISING DAMAGE IN NEW STRUCTURES

SUE FREITAG1 JAMES MACKECHNIE2

1 WSP Opus; 2 Concrete NZ SUMMARY This paper describes the background to current NZ specifications for minimising the risk of AAR damage in new concrete structures, and a test programme currently underway to inform an upcoming review of these specifications. Using Waikato River and Rangitikei River concrete sands, this programme shows how a combination of standard test methods including petrography, the quick chemical test, and accelerated mortar bar and concrete prism expansion tests, may be used to assess the potential alkali reactivity of concrete aggregates. It also suggests outcomes from the test programme to be considered in future revisions of NZ industry recommended practice. BACKGROUND TO CURRENT NZ SPECIFICATIONS FOR MINIMISING AAR DAMAGE Alkali aggregate reactions (AAR) are chemical reactions between alkalis in the pore solution of hardened concrete and some types of minerals within concrete aggregates. The product is a gel that expands in the presence of moisture. Its expansion can cause the concrete itself to expand. The tensile stresses generated may crack the concrete and may affect mechanical properties such as strength, elastic modulus, creep, and bond to reinforcement. Expansion of the concrete may close movement joints. Cracks may facilitate the ingress of moisture and aggressive agents and thus reduce the concrete’s potential durability. AAR generally occurs over many years. It cannot easily be stopped once underway, therefore the risk of AAR damage to concrete structures needs to be managed during construction by selecting appropriate concrete materials and mix designs. The most common type of AAR is alkali silica reaction (ASR), which occurs with some types of silica minerals found in many of the rock types used as concrete aggregate. ASR is the only type of AAR known to damage concrete in structures in New Zealand. This paper uses the generic term ‘AAR’ rather ‘ASR’, although the two may be used interchangeably in this context. AAR is relatively common in parts of New Zealand where reactive aggregates have been used, particularly in the Auckland-Waikato region, where highly alkali reactive Waikato River sands have been used widely. The damage is usually minor, thanks to precautions taken since the mid-1940s to minimise the risk of damage, including a Public Works Department (PWD) requirement that low alkali cement be used where potentially reactive aggregates may be used. The PWD was such a large purchaser of cement that for many years most cement manufactured in New Zealand was low alkali to meet its requirements, leading to the widespread assumption that AAR was not a problem in New Zealand. The chance discovery in the 1980s of AAR in concrete structures indicated further controls were required.

In 1991 CCANZ published TR3, the first New Zealand guidelines for minimising the risk of AAR damage in New Zealand. Developed by a pan-industry working party, TR3 was based on UK practice, which limited concrete alkali content, rather than cement alkali content, when using reactive aggregate. TR3 also recorded the findings of government funded AAR research in a form that could be utilised by the wider concrete industry. Recognising that New Zealand’s volcanic aggregates were highly reactive, TR3 adopted a maximum concrete alkali limit of 2.5 kg/m3 compared to the UK limit of 3.0 kg/m3. The limit of 2.5 kg/m3 was chosen because evidence suggested that significant AAR damage was unlikely at lower alkali contents, not because alkali contents any higher than this would automatically increase the risk of damage. TR3 was revised in 2003 to account for increased understanding about the incidence of AAR in New Zealand structures, changes in international developments in test methods and specifications related to AAR, and changes in local industry specification practices introduced by NZS 3104:2003. The 2003 edition maintained the 2.5 kg/m3 limit as a default for Normal Concrete. For Special Concrete it introduced a range of preventive measures depending on the assessed risk of damage associated with the individual concrete structure or element, based on Canadian and RILEM guidelines at the time (since adopted by other organisations). These included a wider range of maximum concrete alkali contents (1.8 to 3.0 kg/m3), the use of supplementary cementitious materials (SCM), or rejecting the proposed aggregate. TR3 was updated again in 2012, recognising observations of AAR in the South Island, and interest in the use of recycled materials as concrete aggregate. The 2.5 kg/m3 default limit on concrete alkali content was considered a robust and practical solution, therefore was neither reviewed nor altered in 2003 or 2012. For further information about AAR in New Zealand, readers are referred to CCANZ TR3 and the references therein. UPCOMING REVIEW OF CURRENT NEW ZEALAND SPECIFICATIONS Demand to use Waikato River sands has been increasing to overcome shortages of other suitable concrete sand in the Auckland-Waikato region. In 2016, concrete producers reported that increasing demands for structural concrete with high early strengths made it difficult to stay within TR3’s concrete alkali limit for Normal Concrete. Consequently, CCANZ was asked to consider increasing TR3’s concrete alkali limit from 2.5 to 2.7 kg/m3 to facilitate the use of Waikato River sands in Normal Concrete. 2.7 k/m3 was proposed because it would overcome the immediate problem for many concrete applications, while being more conservative than the limit of 2.8 kg/m3 recommended by current Australian guidelines, SA HB 79:2015 (‘HB 79’), for specific combinations of risk and aggregate reactivity. HB 79 does not, however, recommend limiting concrete alkali content as the only control for critical structures, high risk situations, or for reactive aggregates in moderate risk situations. Other changes in the market and in industry practice compound the problems faced by concrete suppliers. Significant volumes of imported cement entering the New Zealand market mean cement can no longer be assumed to be low alkali by default. Specifiers, even on major projects, are often unwilling to consider risk-based management of AAR as set out in TR3 and in international protocols, instead requiring assurance that aggregates selected for their projects be non-reactive and/or the 2.5 kg/m3 alkali limit be applied. Commercial laboratory testing services with appropriate experience in determining the potential alkali reactivity of aggregates in concrete are not readily available in New Zealand. And in 2016, ASTM C289, long considered as a quick and reliable test for NZ aggregates, was withdrawn. In 2016, CCANZ (now Concrete NZ (CNZ)) convened a working party to address these issues and update TR3. To inform its decisions, a test programme was developed in 2016 to examine

several aspects, in particular to ascertain how the risk of AAR damage is affected by concrete alkali contents higher than 2.5 kg/m3, and whether the risk is measurably greater at alkali content between 2.5 and 3.0 kg/m3. Centred on a blend of Waikato River sand with Hunua greywacke mixed with type GP cement, and extending previous investigations into the reactivity of Rangitikei River sand, the programme follows internationally accepted testing protocols and thus also demonstrates how they may be used to answer questions that may arise about the potential reactivity of other aggregates or aggregate blends. This testing began in 2017 and will be largely completed at the end of 2018. Once these tests are completed, the working party will consider how TR3 can be modified to meet industry needs, potentially increasing the maximum concrete alkali limit for Normal Concrete, recommending testing protocols that reflect current international practice and availability of testing services, and updating information about the use of SCMs for reducing AAR expansion. The tests complement a wider Australian research programme initiated in late 2016 and supported by CCANZ to identify key features of the composition of aggregates, cementitious binders, and concrete that influence the occurrence of AAR in individual concretes. The Australian work is being undertaken by the University of Technology Sydney (UTS) under the Australian Research Council’s Research Hub for Nanoscience-Based Construction Material Manufacturing (ARC Nanocomm). The materials and results from the test programme described in this paper provided a systematic data series that is being utilised by the UTS researchers. In turn, the wider UTS research programme will give New Zealand industry better understanding of the inherent properties of the aggregate and cementitious binders used in the 2017-18 test programme, and thus explain some of the behaviours exhibited therein. The wider programme and initial test results are described by Nsiah-Baafi et al (2018). This paper describes the current New Zealand test programme, focussing on the test methodology and the testing around concrete alkali limits. Results and outcomes will be presented in future papers and reports when all testing has been completed. AIM OF TEST PROGRAMME The 2017-18 test programme set out to answer several questions, including:

1. Can the default concrete alkali limit for Normal Concrete be increased to 2.7 kg/m3 without significantly increasing the risk of AAR damage?

2. Do the accelerated mortar bar and concrete prism expansion tests now used widely overseas give results consistent with the quick chemical test previously used widely in New Zealand, and with in-situ performance, when used to assess the potential alkali reactivity of New Zealand aggregates and aggregate blends?

3. Can these tests be used to detect pessimum proportion behaviour? 4. Does initial curing at elevated temperatures increase the subsequent likelihood of

deleterious AAR expansion? 5. Are standard test methods now available to New Zealand industry on a commercial

basis suitable for assessing the potential reactivity of New Zealand aggregates and aggregate-cement combinations?

6. Are differences between results from the rapid mortar bar and concrete prism expansion tests related purely to test temperature, or are the specimen size and mix proportions also influential?

Investigations around questions 1 to 5 were centred on a commercial blend of Waikato River sand and Hunua greywacke PAP7, with Hunua coarse aggregate. A previous investigation of the potential reactivity of Rangitikei River sand carried out for Rangitikei Aggregates Ltd was extended in this programme to investigate points 5 and 6.

As well as answering these specific questions in the context of TR3’s upcoming review, the results will also indicate whether the same methodologies can be used by industry to answer future questions that may arise. By familiarising the New Zealand industry with current international practices for minimising AAR damage, the programme aims to encourage their use, and thereby enable informed decisions that will facilitate the use of concrete materials that are sustainable and economic for individual local markets, and will produce concrete that meets the needs of contractors and structure owners. MATERIALS USED Materials used in the test programme are as follows. All samples were supplied in early 2017. Table 1 Materials

General description C= crushed N = natural

Samples supplied by

How used in test programme

Hunua greywacke 19 mm & 13 mm (C)

GBC Winstone Coarse aggregate for concrete prism tests.

Hunua greywacke PAP7 (C)

GBC Winstone Blended with Waikato River sand to provide a fine aggregate of suitable grading, also tested alone.

Waikato River sand (N)

Waikato Aggregates Ltd

Blended with PAP7 or non-reactive control sand to form a fine aggregate of suitable grading.

Rangitikei River sand (N)

Rangitikei Aggregates Ltd

Concrete prism tests (a) in the same proportions of sand and cement as used in the rapid mortar bar test, and (b) in standard test proportions.

Peats Ridge basalt (C)

Test laboratory Non-reactive control coarse aggregate for concrete prism tests, and non-reactive control fine aggregate for rapid mortar bar test.

Maroota sand (N)

Test laboratory Non-reactive control fine aggregate for concrete prism tests.

Type GP Portland cement

GBC Winstone All mortar bar and concrete tests.

The Hunua greywackes were from a source that had not yet been used at the time it was sampled. Evidence from previous testing on New Zealand greywackes and from general performance of New Zealand greywackes in concrete structures suggest it is unlikely to react significantly in most concrete but may react in tests designed to accelerate reactivity in such materials. If results from this test programme show it does not react in the accelerated mortar bar and concrete expansion tests used then it could be used as a non-reactive control in future tests carried out locally. Waikato River sand is known from in-situ behaviour and laboratory testing to be highly reactive at concrete alkali levels exceeding the current limit. It is responsible for most cases of AAR in the Auckland-Waikato region. It is blended with other sands to obtain a fine aggregate of suitable grading for use in concrete. A 60/40 blend of Waikato River sand and either Hunua PAP7 or a non-reactive control fine aggregate was used in this test programme, based on recommendations by Firth Industries. Rangitikei River sand contains quantities of potentially reactive minerals close to the pessimum proportion of similar materials observed in early DSIR testing, therefore is classed as potentially reactive. It has exhibited limited reactivity in previous laboratory test programmes, but very few cases of AAR damage associated with this material have been identified on concrete structures.

Peats Ridge basalt and Maroota sand were supplied by the test laboratory as non-reactive reference materials meeting the performance requirements specified by the test methods. Maroota sand exhibits slight expansion in the accelerated mortar bar expansion test, therefore Peats Ridge manufactured fine aggregate was used in this test. Maroota sand is a natural sand, and was used in the concrete prism tests because its particle shape and size distribution produce better workability. All concrete and mortar test specimens were made with Golden Bay Cement’s Type GP cement supplied in bulk to the test laboratory. TEST METHODS Current knowledge about the causes and extent of AAR in New Zealand is based on laboratory investigations involving a combination of petrographic examinations, quick chemical testing, accelerated mortar bar expansion testing to methods like the NZS 3111:1986 method, and limited non-standard concrete testing. In the 1980s to 2000s these were augmented by investigations of AAR in concrete structures, mostly bridges, representing a wide range of ages and concrete mix designs. Most cases were attributed to concrete alkali contents exceeding 2.5 kg/m3. The ideal way to investigate the risk of AAR in concrete with alkali contents in the range of interest in this programme (2.5 – 3.5 kg/m3) would have been to inspect in situ concrete with original alkali content in this range, with similar mix design, reinforcement, and environmental exposure, and old enough to exhibit AAR damage. A suitable population of accessible structures could not be identified for such a study within the required timeframe. Consequently, the current test programme is based solely on laboratory testing. To ensure tests were repeatable and reproducible, standard test methods were used. To demonstrate the efficacy of test methods now readily available on a commercial basis to New Zealand industry, Australian standard test methods were used where possible, rather than AASHTO, ASTM, CSA, or RILEM methods. AS methods were used rather than similar VicRoads and RMS (NSW) methods because they are more generally applicable and may be more actively maintained in future. Test methods used are shown in table 2. Readers are referred to HB 79 and the test methods themselves for details of the test methods. This section and the following section describe aspects relevant to this test programme. Table 2 Test methods

Principle Standard method used in this programme

Examples of similar standard methods

Petrographic analysis to assess potential reactivity of aggregate

AS 1141.65 (aggregate) ASTM C856 (mortar/concrete)

ASTM C295 RILEM AAR-1

Quick chemical test ASTM C289 -

Rapid / accelerated mortar bar expansion test (‘AMBT-80’)

Mortar bars stored in sodium hydroxide solution at 80°C for 14-28 days and expansion measured.

AS 1141.60.1 RMS T363 VicRoads RC 376.03 ASTM C1260 RILEM AAR-2 CSA A23.2-25A

Accelerated concrete prism expansion test (‘CPT-38’)

Concrete prisms stored over water at 38°C for at least 1 year and expansion measured.

AS 1141.60.2 (RMS T364 was used for tests on Rangitikei R sand, to be consistent with previous testing)

RMS T364 VicRoads RC 376.04 ASTM C1293 RILEM AAR-3 CSA A23.2-14A

Accelerated concrete prism expansion test (‘CPT-60’)

Concrete prisms stored over water at 60°C for at least 4 months and expansion measured.

AS 1141.60.2 run at 60°C (RMS T364 was used for tests on Rangitikei R sand, to be consistent with previous testing)

RILEM AAR-4

The Australian test methods were developed primarily to assess the potential alkali reactivity of individual aggregates and aggregate combinations. They are based on test methods widely used overseas. Figure 1 at the end of this paper shows the RILEM (2016) recommended process for assessing aggregate reactivity using such tests, which may be used to evaluate proposed new aggregate sources or the consistency of supply from existing sources. This programme sought answers to questions currently facing the New Zealand industry, therefore did not attempt to evaluate new test methods currently being developed overseas. All testing was carried out by Boral Material Technical Services in NSW. AMBT-80 and CPT- 38 testing was covered by Boral’s NATA accreditation. Petrographic analysis was carried out by Geochempet Services under subcontract to Boral. DSIR reported that ASTM C289 test results were highly sensitive to aggregate preparation, therefore the test should be carried out by operators with appropriate experience. In countries where AAR is associated with aggregates that react slower than New Zealand’s fresh volcanic reactive aggregates, ASTM C289 has been superseded by AMBT-80 tests, which are considered more reliable for slowly reactive aggregates. ASTM C289 is no longer supported by ASTM, therefore may become more difficult to procure in future. Although all based on the same principles and very similar mix proportions, each individual AMBT-80, CPT-38 and CPT-60 standard test was developed to best represent experience with local materials. Thus workability, water to cement ratio, and specified mix designs and initial curing may vary slightly between jurisdictions. Consequently, criteria developed to assess the test results from one standard test method may not be directly transferrable to results from a similar test. If, as in New Zealand, test methods have not been directly calibrated against the known in situ reactivity of specific aggregates to develop accurate assessment criteria, then overall trends in the results need to be considered in addition to the expansion measured at a specific age. An extended test period and/or testing of more than one combination of materials may be necessary, particularly for unfamiliar materials. Differences between various test methods are discussed in HB 79. For example, higher water to cement ratios may produce higher expansions, though the overall outcome of the test should not be affected. Of the test methods based on the same principles, the authors of this paper consider none to be inherently better or worse than similar methods in other jurisdictions. Rocker et al (2015) also found the outcome of CPT-38 tests was not significantly affected by the specific test method used, including the assessment criteria. Instead, the reliability of the result from any individual method depends on the skill of the tester, and the quality of the outcome depends on the client and tester understanding the principles and the limitations of the test and the assessment criteria so that the test results are interpreted appropriately. Petrographic analysis AS 1141.65 was developed specifically to identify the presence of potentially reactive components in concrete aggregate, therefore was used as the basis for the petrographic analyses carried out in this programme. Unlike TR3, this method does not allow an aggregate source to be classified as non-reactive based on the absence of potentially reactive components. ASTM C856 includes specific clauses relating to the detection of deleterious reactions in concrete, including AAR. AMBT-80 In keeping with the test requirements, the Waikato River sand was tested ‘as supplied’, while the greywacke and Peats Ridge aggregates were tested in the standard specified grading. The test specimens made from Waikato River sand were less workable than the others.

CPT-38 In keeping with the test requirements, the natural fine aggregates were tested ‘as supplied’, while the coarse aggregates and greywacke PAP7 were tested in the standard specified grading. AS 1141.60.2 requires that concretes be mixed with a water to cement ratio 0.42 to 0.45 to give a slump as close as possible to 80 mm. Trial mixing was required to obtain appropriate workability. In the end, a fixed water to cement ratio of 0.45 was used for all concrete. Slumps were between 55 and 100 mm, except for the concretes made from greywacke only, which recorded slumps between 40 and 55 mm. The greywacke concretes could be readily placed and consolidated despite their lower slumps, therefore were used without further modification. CPT-38 (and CPT-60) accelerate AAR by elevating not only the exposure temperature but also the concrete alkali content. Alkali from the concrete materials is boosted to the required level by adding sodium hydroxide to the concrete. The alkali equivalent of the cement measured by the test lab in this programme differed from the value measured by the manufacturer. Subsequent analyses of the same cement sample by three different labs produced alkali equivalent values from 0.51 to 0.65%. A similar comparison run for a different programme produced a similar range of results, as did an international interlab (ATILH, 2016). TR3 reports interlaboratory differences of ±0.05% sodium oxide equivalent. This test programme used a range of alkali contents (see next section), therefore change in alkali content was considered more significant than the absolute value, and concrete alkali contents were calculated from the cement alkali equivalent of 0.58% measured by the manufacturer. Previous tests on Rangitikei River sand carried out for the supplier were planned before AS 1141.60.2 was published, therefore used RMS test methods. Tests on Rangitikei River sand in the current test programme also used the RMS method so that the results from the two test programmes could be compared readily. CPT-60 This test is essentially the same as the CPT-38 but run at higher temperature to produce results sooner. Test periods of four to six months have been suggested by RILEM and HB 79. In this test programme the tests were run for at least 6 months. PROGRAMME METHODOLOGY The combination of tests carried out is based on methodologies described by RILEM (2016), which shows how to use the generic test procedures to answer questions relating to potential reactivity of individual aggregates, aggregate combinations, and concrete mix designs, including determination of concrete alkali limits. Petrographic analyses The petrographic analyses were performed to identify potentially reactive components, and to characterise the aggregate samples used in this test programme so that these samples can be compared with similar materials tested previously and in the future. On completion of the expansion tests, selected mortar bars and concrete prisms were petrographically analysed to ascertain whether AAR had in fact occurred, and if so, to identify the reactive components of the aggregate. This will ensure that dimensional changes caused by other factors are accounted for and not mistakenly attributed to AAR expansion.

ASTM C289 This test was carried out on Waikato R sand, the greywacke PAP7, and a 60/40 blend of Waikato R sand and greywacke PAP7. AMBT-80 AMBT-80 tests were used to characterise the reactivity of Waikato R sand alone, when blended 60/40 with greywacke PAP7, and when blended 60/40 with a non-reactive control sand. Greywacke PAP7 was also tested by itself to ascertain the reliability of the test for New Zealand greywackes, and whether it is sufficiently inert to be used as a non-reactive control. CPT-38 CPT-38 tests are generally considered the most reliable indicators of potential aggregate reactivity and concrete expansion. In their simplest form they are used to identify whether a rock type is potentially alkali reactive. This involves testing the aggregate in a standardised mix design at 5-6 kg/m3 alkali (e.g.AS 1141.60.2 and ASTM C1293 use 5.25 kg/m3; RILEM AAR-3 uses 5.5 kg/m3, and RMS T364 uses 5.8 kg/m3). The test may also be used to assess the potential reactivity of aggregate combinations, the potential for a particular concrete mix design to cause deleterious AAR expansion, and to determine maximum concrete alkali contents for particular combinations of materials. The different methodologies and protocols for interpreting results are described by RILEM (2016). Following RILEM (2016) principles, this project tested each combination of aggregates at five nominal concrete alkali contents: 2.5, 3.0, 3.5, 4.0, 5.0 and 5.25 (or 5.8) kg/m3 to establish the effect of concrete alkali content on potential reactivity. 2.5 and 3.0 kg/m3 were included to represent the actual alkali contents of interest. RILEM (2016) advises that maximum alkali limits should be based on the highest alkali increment that does not give a significant expansion in the test. A safety margin is applied to this value, to allow for experimental effects (such as leaching of alkali), site batching variability, and the difference between laboratory and field specimens. RILEM (2016) suggests a safety factor of 1.0 kg/m3 alkali content, such that test results indicating significant expansion above 3.5 kg/m3 would translate to an imposed alkali limit of 2.5 kg/m3. Thus the alkali contents of interest in this test for informing the limit for Normal Concrete are 3.5 and 4.0 kg/m3. An intermediate value of 3.8 kg/m3 was included in one test series to obtain more detail within the range 3.5 to 4.0 kg/m3. A larger safety factor may be needed to account for uncertainty associated with the measurement of cement alkalis and thus concrete alkalis, and inherent variations between batches of the same cement. Three aggregate blends were tested to assess the effect of concrete alkali content on the reactivity of Waikato River sand and the Hunua greywacke:

• Waikato River / non-reactive fine aggregate blend and non-reactive coarse aggregate, to determine the effect independent of other aggregates used with Waikato R sand.

• Waikato River / greywacke PAP7 fine aggregate blend and greywacke coarse aggregate, to determine the potential reactivity of a likely commercial aggregate combination. An additional test was carried out with this combination at 5.25 kg/m3 alkali to ascertain whether the risk of expansion is affected by accelerated curing. This concrete was cured at up to 70°C following a typical curing regime used by precasters, rather than being cured in accordance with the test method.

• Greywacke coarse and fine aggregate, to determine the potential reactivity of this greywacke as evidenced by this test method, and therefore whether the test method reflects observed in situ reactivity, whether it indicates a potential reactivity not evident under normal mix designs and ambient curing conditions, and whether this material may be used as a non-reactive control in such tests in future.

Additional tests were carried out with the Waikato River / non-reactive aggregate blend to examine how the potential for expansion was affected by the quantity of reactive material in the aggregate blend and thus whether the test was sensitive enough to evaluate the risk associated with pessimum proportion effects. A further test series investigated how the proportions of cement and reactive material in test specimens affect the relationship between AMBT-80 and CPT-38 results. This test was carried out using Rangitikei River sand. AMBT-80 and CPT-38 test results for this sand were obtained from a previous investigation for the sand supplier. The results may also explain the disparity between AMBT-80, CPT-38, and in-situ behaviour of this sand and thus remove some of the doubt about its reactivity. CPT-60 CPT-60 tests were carried out on three aggregate combinations to find out whether the higher test temperature could produce equivalent results to CPT-38 in a shorter and therefore more practical test period:

• Waikato River / non-reactive fine aggregate blend and non-reactive coarse aggregate

• Waikato River / greywacke PAP7 fine aggregate blend and greywacke coarse aggregate,

• Rangitikei River sand and non-reactive coarse aggregate. These tests were carried out over the same five alkali contents as the CPT-38 tests. In the absence of Australian protocols for interpreting the results, RILEM (2016) guidelines were followed. INITIAL FINDINGS Results will be interpreted primarily in accordance with the individual test methods and associated protocols. ASTM C1778 will be used for additional guidance, e.g. if results are unexpected or conclusions unclear. Petrographic analyses of Waikato River and Rangitikei River sands gave results consistent with existing knowledge. The Hunua greywacke material was found to contain material that may cause mild or slow AAR, including small amounts of intermediate and devitrified acid volcanic rock fragments (see comments below about AMBT-80 results). ASTM C289 tests on Waikato River sand, greywacke PAP7, and a 60/40 blend of the two indicated the Waikato River sand and the sand blend were potentially deleterious, while the greywacke itself was innocuous. This is consistent with results from similar materials reported in TR3 Appendix D, and inspections of New Zealand concrete structures in the 1980s to 2000s. AMBT-80 tests showed the Waikato R sand to be reactive when tested alone, blended with the greywacke PAP7, and blended with the non-reactive control sand, which is consistent with in-situ experience. Test results were not significantly affected by reducing the alkali concentration of the storage solution. Test results from the greywacke PAP7 showed it to be ‘slowly reactive’, in keeping with expectations based on behaviours of similar rock types internationally, which this test was designed to detect. (When interpreted according to ASTM C1778, the results indicate the greywacke is non-reactive; possibly ‘moderately reactive’ if the test precision is also considered.) This means it should not be used as a non-reactive control in AMBT-80 tests. It also suggests that although no cases of AAR damage have been reported in concrete made from greywacke in New Zealand, New Zealand greywackes may undergo AAR in some circumstances, particularly when alkali contents are relatively high. Petrographic analyses of selected test specimens, including those made from the PAP7 alone, showed that the expansions recorded were probably caused by AAR.

AMBT-38 tests may not show significant expansions until after nine months or more, sometimes more than 12 months. At the time of writing, 7-month results show the concretes made from Waikato River and Rangitikei River sands and inert coarse aggregate at the highest alkali contents are starting to expand, and the heat-cured specimens have expanded more than their counterparts cured at ambient temperature. No clear trends are evident yet from the concretes made with the greywacke PAP7 or with lower alkali contents. AMBT-60 test results at 4 months indicate the Waikato River sand is potentially reactive when used with the inert aggregate, but non-reactive when used with the greywacke (although expansion had started). The reason for this inconsistency may become clearer when the CPT- 38 results, or results from further testing within the ARC programme, are available. The results for the Rangitikei River sand indicate the sand is potentially reactive, in keeping with previous AMBT-80 results. The 4-month results do not indicate a meaningful alkali limit for any of the combinations of materials tested. LIKELY OUTCOMES A potential range of 0.1% sodium equivalent between measurements of cement alkali content by different laboratories suggests that increasing the permissible maximum alkali content of Normal Concrete by 0.2 kg/m3 is within the margin of error in determining concrete alkali contents, and is therefore unlikely to significantly increase the risk of AAR damage. TR3 provisions may, however, be revised to allow for varying levels of uncertainty in the determinations of cement and concrete alkali contents. The results from standard tests indicate the likelihood of AAR damage in concrete at ambient curing and in-service conditions. Reactivity induced by exposure to elevated temperatures will be indicated by comparing the results from tests at different temperatures. TR3 provisions may be revised to include this additional level of assessment, to inform decisions related to the use of accelerated curing and for larger pours. The results will also enable TR3 to provide guidance on the use of standard test methods for assessing aggregate reactivity. For example, results to date suggest:

1. Petrographic analysis remains the easiest method of assessing the potential reactivity of an aggregate. AS 1141.65 provides suitable guidelines for geologists undertaking petrographic analyses related to AAR. When used in New Zealand, the findings should be interpreted in the context of TR3.

2. ASTM C289 remains a reliable test for assessing the reactivity of New Zealand aggregates for which in-situ reactivity is well understood. Should only be undertaken by experienced laboratories.

3. AMBT-80 tests may over-estimate reactivity therefore should be used with caution, but may be useful for initial screening of a new aggregate supply, indicating the consistency of an aggregate supply, or the effects of elevated temperature and other unusual service conditions. More informative if run for 28 days and results assessed from 14 day and 21 day expansions, as in AS 1141.60.1, rather than from 14 day results only as in ASTM C1260/C1778.

4. Overall trends should be considered, rather than simply comparing results with published criteria, especially if inherent measurement errors are not accounted for.

REFERENCES

ATILH (2016). Interlaboratory Testing Programme - Final Report. (Available to participants). CCANZ (2012). Alkali Silica Reaction. Minimising the risk of damage to concrete. Guidance notes and recommended practice. Technical Report 3, 2nd ed, incl Amendment 1. December 2012.

Nsiah-Baafi, E., Vessalas, K., Thomas, P. and Sirivivatnanon, V. (2018). “Mitigating alkali silica reactions in the absence of using SCMs: A review of empirical studies”, International Federation for Structural Concrete 5th International fib Congress 2018, Melbourne, 6-11 October 2018. RILEM (2016). RILEM Recommendations for the Prevention of Damage by Alkali Aggregate Reactions in New Concrete Structures. State-of-the-Art Report of RILEM Technical Committee 219-ACS. Eds Nixon, P.J. and Sims, I. RILEM State-of-the Art reports v.17. Rocker. P., Mohammadi, J., Sirivivatnanon, V., South, W. (2015). Linking New Australian Alkali Silica Reactivity Tests to World-Wide Performance Data. Proc. 27th biennial national conference of the Concrete Institute of Australia: pp. 502–513 Standards Australia (2016). Alkali Aggregate Reaction – Guidelines on Minimising the Risk of Damage to Concrete Structures in Australia. Handbook SA HB 79:2015. Test methods: AS 1141.60.1 Methods for sampling and testing aggregates – potential alkali silica reactivity - Accelerated mortar bar method AS 1141.60.2 Methods for sampling and testing aggregates – potential alkali silica reactivity – Concrete prism method AS 1141.65 Methods for sampling and testing aggregates – alkali aggregate reactivity – qualitative petrological screening for potential alkali silica reaction. ASTM C289 Test Method for Potential Alkali Silica Reactivity of Aggregates (Chemical Method) ASTM C295 Guide for Petrographic Examination of Aggregates for Concrete. ASTM C856 Standard Practice for Petrographic Examination of Hardened concrete. ASTM C1260 Test Method for Potential Alkali Reactivity of Aggregate (mortar bar method) ASTM C1293 Test Method for Determination of Length Change of Concrete due to Alkali Silica Reaction. ASTM C1778 Guide for Reducing the Risk of Deleterious Alkali Aggregate Reaction in Concrete. CSA 23.2-14A Potential Expansivity of Aggregates; Procedure for Length Change Due to Alkali-Aggregate Reaction in Concrete Prisms CSA 23.2-25A Test Method for Detection of Alkali-Silica Reactive Aggregate by Accelerated Expansion of Mortar Bars RILEM AAR-1.1 Detection of Potential Alkali Reactivity – Part 1: Petrographic Examination Method RILEM AAR-2 Detection of Potential Alkali Reactivity – Accelerated Mortar Bar Test Method for Aggregates RILEM AAR-3 Detection of Potential Alkali Reactivity – 38°C Test Method for Aggregate Combinations Using Concrete Prisms RILEM AAR-4 Detection of Potential Alkali Reactivity – 60°C Test Method for Aggregate Combinations Using Concrete Prisms RMS T363 Accelerated Mortar Bar Test for the Assessment of Alkali Reactivity of Aggregate. RMS T364 Concrete Prism Test for AAR Assessment. VicRoads RC 376.03 Potential Alkali Silica Reactivity (Accelerated Mortar Bar Method) VicRoads RC 376.04 Alkali Aggregate Reactivity Using the Concrete Prism Test.

ACKNOWLEDGEMENTS The authors acknowledge the contributions made to this test programme by the following organisations and individuals:

• ConcreteNZ, NZRMCA, BRANZ, and MBIE funded the laboratory testing within the research programme ‘Ensuring affordable concrete supply post 2010’ (ref.LR0526).

• WSP Opus Research provided technical overview on behalf of ConcreteNZ.

• The testing methodology was based on test programmes developed by Ahmad Shayan of ARRB for WSP Opus Research for this and previous investigations.

• Firth Industries advised on suitable aggregate blends and sources of materials.

• Rangitikei Aggregates Ltd permitted findings from previous testing to be incorporated into this test programme.

• GBC Winstone arranged delivery of all materials to the testing laboratory.

• Boral Materials Technical Services, NSW, provided testing services.

• Technical support, including development and documentation of concrete and mortar mix designs and test results, investigation of cement alkali contents, and liaison between industry and research partners, was provided by Vute Sirivivatnanon, Elsie Nsiah-Baafi, and Kirk Vessalas from University of Technology of Sydney.

The authors also wish to acknowledge CCANZ and Concrete NZ for developing and maintaining TR3 as not only as a code of practice but also as a record of NZ’s experience of AAR; and the input of Working Party members past and present, particularly Don St John and David Barnard for instigating the original edition.

Figure 1 RILEM assessment scheme (after RILEM (2016), Figure 1). Shading indicates tests included in the programme described in this paper.

AAR-1.1 Petrographic Examination carried out?

Yes No

Class II: potentially reactive or reactivity uncertain Class III: very likely to be reactive

Silica & Carbonate (classes IISC /IIISC)

Carbonate (classes IIC & IIIC)

Silica (classes IIS / IIIS)

Rapid Screening Test? Rapid Screening Test?

Yes Yes No No

AAR-2 80ºC Ultra Accelerated

Mortar Bar Test

AAR-2 & AAR-5 Concrete Microbar Test

Reactive or Potentially Reactive

Non-reactive

Petrograp

hy

Reactive or Potentially Reactive

Non-reactive

Co

ncrete Exp

ansio

n Te

sts R

apid

Screenin

g Tests

And/or Either

Class I: very

unlikely to be

reactiveactivele

AAR-4.1 60 ºC Concrete Prism

Test

AAR-3 38 ºC Concrete Prism

Test

No further action

required