Materials and Structures

, 42:485

Expansion behaviour of glass aggregates in different testing for alkali-silica reactivity

Authors

    • GuangXi Institute of Building Research & Design
  • Wen Chen
    • School of Materials Science and EngineeringWuhan University of Technology
  • Wei Zhou
    • GuangXi Institute of Building Research & Design
  • Ewan A. Byars
    • Centre for Cement & Concrete, Department of Civil & Structural EngineeringThe University of Sheffield
Original Article

DOI: 10.1617/s11527-008-9396-4

Cite this article as:
Zhu, H., Chen, W., Zhou, W. et al. Mater Struct (2009) 42: 485. doi:10.1617/s11527-008-9396-4

Abstract

The potential for reaction between amorphous silica in recycled glass used as aggregate in concrete and alkalis in cement is the subject of debate in current concrete literature. Whilst the ASTM C1260 accelerated mortar bar method is conventionally used for rapid ASR assessment, there is doubt about its suitability for glass aggregates. This paper reports upon a comparison of the relative ASR reactivity of various colours of recycled glass aggregates using the ASTM C1260 and C227 test methodologies. The results show that with limited exception the ASTM C1260 method does not cause glass aggregates to react by the end of the prescribed test period. In contrast, the ASTM C227 method causes all glass aggregates to react within 2 weeks, despite the test being designed for 12 months or even longer if necessary. This paper compares and contrasts the results of the two methods over a wide range of glass aggregate and cementitious systems made with two sizes of mortar bar, draws conclusions about the reasons for the differences observed and makes remarks on the expansion behaviour of glass aggregates in cementitious systems.

Keywords

ASRGlass aggregateASTM C1260ASTM C227Reaction mechanisms

1 Introduction

Crushed waste glass was first used as a concrete aggregate in 1974 [1] with disappointing results due to expansive alkali-silica reaction (ASR) between the amorphous silica in the glass and the alkaline cement paste. Over the past decade however, a number of major research projects in the USA and the UK [214] have been conducted in this area. The studies agree that all glass aggregates exhibit rapid ASR in normal OPC concrete and that pozzolanic cement replacement materials may significantly reduce, but not always mitigate, the expansive ASR reaction.

The three test approaches used by industry to assess ASR reactivity of aggregates [15] are (i) petrographic analysis of aggregate, (ii) direct expansion measurement of mortar and concrete made with test aggregates and (iii) chemical tests. Direct expansion measurement has been widely adopted for concrete aggregates but a variety of mortar bar and concrete prism methods including ASTM C1260, ASTM C227, BS 812-123, ASTM C1293 and RILEM TC106-3 [1620] are used. Brief descriptions of these methods are given in Table 1 [23].
Table 1

International dimensional change ASR methods

Test method

Comments

ASTM C 227: Standard test method for potential alkali reactivity of cement-aggregate combinations

High alkali cement (Na2Oeq > 0.6%) mortar bar test to determine cement-aggregate reactivity

Specimens stored in high-humidity containers at 38°C

Requires on year or even longer if necessary for completion

Excessive leaching of alkalis from specimens reported [21]

ASTM C 1260: Standard test method for potential alkali reactivity of aggregates (mortar-bar method)

Mortar bar test for aggregate reactivity

Bars immersed in 1N NaOH solution for 14 days at 80°C

Suitable for screening, but because of exposure severity, potentially unsuitable as absolute test. Results of concrete prism test should prevail [22]

ASTM C 1293: Standard test method for concrete aggregates by determination of length change of concrete due to alkali-silica reaction

Concrete prism test, regarded as best indicator of field performance, conducted at 38°C/100% RH

High-alkali cement (1.25% Na2Oeq), with a cement content of 420 kg/m3

Coarse aggregate test (non-reactive fine aggregate) or vice-versa. Requires one year for completion

Can be used to test effectiveness of suppressants over 2 years [22]

Widely accepted, but lengthy test method

BS 812-123: Testing aggregate—method for determination of alkali-silica reactivity—concrete prism method

Concrete prism test, generally regarded as best indicator of field performance, conducted at 38°C/100% RH

High-alkali cement (>1.0% Na2Oeq), with a cement content of 690 kg/m3

Coarse aggregate test (non-reactive fine aggregate) or vice-versa. Requires one year for completion

Length, but widely accepted in the UK

RILEM TC106-3: Detection of potential alkali-reactivity of aggregates—method for aggregate combinations using concrete prisms

Concrete prism test, conducted at 60°C/100%RH

High-alkali cement (0.9–1.2% Na2Oeq, raised to 1.25% by adding NaOH), with a cement content of 440 kg/m3

Test requires 20 weeks for completion, mainly used in mainland EU

In several major published studies [24, 613, 2427] the ASTM C1260 test method has been used to assess the ASR reactivity of glass aggregate, mainly because it gives rapid, repeatable results on relatively small samples. The ASTM C227 test method has been used less widely [21, 27] on glass in concrete because of its disadvantages such as long curing period, less reliable for slow or late expansive aggregates.

The ASTM requirement of the mortar bar size for ASR length changing measurement is 25 × 25 × 285mm (long-thin bars) [16, 17]. However, bars of this size are long, slender and fragile. As the numbers of bars cast for testing in this study was very large with a planned test duration of up to 5 years, the main body of research was carried out using a bar size of 40 × 40 × 160 mm (short-fat bars). A subsidiary study examined the effect of bar sizes on ASR expansion using the ASTM C1260 and C227 prescribed bar size and the size adopted in the main research programme. It needs to clarify that the test criteria with the short-fat bars may be different to that of the long-thin bars. However, the recommended test criteria are adopted in this study for simplicity and comparison purpose.

2 Experimental details

In this investigation, the C1260 and C227 test methods were used to measure the ASR reactivity of clean green (Gr), amber (Am), flint (Fl) and blue (Bl) recycled glass aggregates in 40 × 40 × 160 mm bars. The chemical compositions of the test glasses were given in Table 2. This shows that the chemical composition of these four colours are similar and that the alkali content is high (above 13%). Thus, when using as aggregates, glass may behaviour differently compared with other commonly used aggregates such as quarried rocks or gravels, and may show pozzolanic reaction when used as very fine particles [11, 14].
Table 2

Chemical composition of waste glass

Oxide (%)

Flint (Fl.)

Amber (Am.)

Green (Gr.)

Blue (Bl.)

Na2O

13.13

13.71

13.12

13.02

K2O

0.61

0.16

0.63

0.51

SiO2

72.17

72.08

71.24

72.05

CaO

10.92

10.45

10.79

10.96

MgO

1.10

0.72

1.57

1.32

Al2O3

1.35

2.19

1.63

1.37

Fe2O3

0.07

0.22

0.32

0.16

Cr2O3

<0.05

<0.05

0.22

0.03

SO3

0.09

0.05

<0.05

0.07

TiO2

0.06

0.10

0.07

0.07

BaO

<0.05

<0.05

<0.05

<0.05

Total

99.69

99.83

99.77

99.69

LOI (%)

0.31

0.17

0.23

0.31

2.1 Materials

2.1.1 Cement

Ordinary Portland cement (OPC) of strength class 42.5N to BS EN 197–1 [28] with sodium oxide equivalent (Na2Oeq) of 0.6–0.65% was used to make the C1260 test specimens. High alkali Portland cement (HAPC) with Na2Oeq of 1.08% was used to prepare the C227 test specimens.

2.1.2 Aggregate

A crushed, siliceous sand from Trent Valley (UK) was used as a control fine aggregate. The glass aggregates used consisted of recycled container and window glass, crushed, washed and tumbled, then graded to match the exact ASTM C1260 and C227 guidelines as shown in Table 3.
Table 3

ASTM grading requirement for aggregate

Sieve size

Mass (%)

Passing

Retained on

4.75 mm

2.36 mm

10

2.36 mm

1.18 mm

25

1.18 mm

600 μm

25

600 μm

300 μm

25

300 μm

150 μm

15

2.2 Mix proportions and test procedures

2.2.1 ASTM C1260 test

Mortar proportions of cement: aggregate: water of 1:2.25:0.47 were used. For each mix, three 40 × 40 × 160 mm mortar bars were cast at room temperature and cured for (24 ± 2) hours at 20°C in plastic bags to sustain high ambient RH. After demoulding, these were stored at 80°C in water for another 24 h, then transferred and immersed in 1N NaOH at 80°C until test. The length change of the prisms was measured using a length comparator specified by BS 812-123. An initial reading was taken immediately after demoulding, a zero reading after storing in distilled water at 80°C for 24 h. Subsequent readings were taken following the C1260 test procedure within 1 min of removal from the 80°C alkali storage conditions.

2.2.2 ASTM C227 test

Following C227 [16], mortar mix proportions of cement: aggregate: water of 1:2.25:0.47 were used. For each mix, three 40 × 40 × 160 mm mortar bars were cast at room temperature and cured for (24 ± 2) h at 20°C in plastic bags around 100% RH. After demolding, the length and initial readings of the bars were measured at 20°C. The bars were then wrapped in damp cloth, sealed with plastic wrapping and placed in sealed plastic bags (one bag per mix). The specimens were then transferred to a controlled environmental room maintained at (38 ± 2)°C and ≥98% RH. Subsequent readings were taken every fortnight for the first 26 weeks and every 13 weeks thereafter. The samples were then stored at (23 ± 2)°C for at least 16 h prior to measurement to remove the possibility of thermal expansions or contractions affecting the results.

2.3 Effect of mortar bar size on ASR expansion

In addition to the 40 × 40 × 160 mm bars, an additional series of bars of 25 × 25 × 285 mm (specified by both C1260 and C227 methods) using both methods to determine the effect of bar size on apparent ASR reactivity of flint and green mortars.

3 Results and discussion

3.1 ASTM C1260 test

Short term (up to 28 days) and long term (up to 133 days) ASTM C1260 test results are detailed below.

3.1.1 Short term results

ASR expansion results of the control and test glass aggregates up to an age of 14 days are shown in Fig. 1. Interestingly, with the exception of the very reactive blue glass aggregate, green, amber and flint glass aggregates show less than 0.1% expansion in the C1260 test regime. In strict accordance with C1260, this result implies that they should be classified as non-reactive and other sources [12, 13] have reported green glass as such after a C1260 test. It can also be seen from Fig. 1 that although a high early rate of reaction in the control sand (OPC/Ctrl) implies potential reactivity (14-day expansion of 0.1–0.2%), the green, amber and flint glass aggregates show insignificant or even negative expansions over the same period.
https://static-content.springer.com/image/art%3A10.1617%2Fs11527-008-9396-4/MediaObjects/11527_2008_9396_Fig1_HTML.gif
Fig. 1

ASR expansion of glass aggregates tested to ASTM C1260 to 28 days

3.1.2 Long term results

ASR expansion results for the control and test glass aggregates, up to an age of 133 days, are shown in Fig. 2. This shows that in all cases large expansions are observed if the aggressive conditions of C1260 are maintained beyond the test end point of 14 days. Initiation of an ASR reaction in flint, amber and green glass took between 2 and 4 times the recommended C1260 test duration but once initiated the rate of reaction was rapid but reduced with increasing time to initiation.
https://static-content.springer.com/image/art%3A10.1617%2Fs11527-008-9396-4/MediaObjects/11527_2008_9396_Fig2_HTML.gif
Fig. 2

ASR expansion of glass aggregates tested to ASTM C1260 up to 133 days

The other significant observation from these results is that the time to initiate ASR reaction and the rate of reaction thereafter varies with glass colour. Since the major chemistry of the green, blue, amber and flint glass is very similar, Table 2, this implies that the metals used to impart colour to glass may have an effect on the ASR reactivity [12, 13].

3.2 ASTM C227 test

ASR expansion results of control and glass aggregate mortars made with HAPC and tested to C227 method are shown in Fig. 3. These show that all glass aggregate mortars tested very quickly exceeded the 26-week reactive aggregate expansion limit (0.1%). Indeed, the limit was exceeded at only 2 weeks for blue and flint glass aggregates and 4 and 10 weeks respectively for amber and green glasses. These results further confirm that all glass aggregates tested are ASR reactive (see also Fig. 2).
https://static-content.springer.com/image/art%3A10.1617%2Fs11527-008-9396-4/MediaObjects/11527_2008_9396_Fig3_HTML.gif
Fig. 3

ASR expansion of glass cullet tested to ASTM C227 up to 52 weeks

A large discrepancy between the C227 and C1260 results, however, was observed with the control sand, which is well within C227 test limits up to a test age of 52 weeks (Fig. 3), yet appears to be “potentially reactive” when tested by C1260 (Fig. 2).

To further confirm the highly ASR reactive of glass aggregate observed in the moist-curing C227 test, a parallel study [14] has been carried out using BS 812-123 [18] concrete prism method. Test concretes were made with HAPC, 3–6 and 6–12 mm green, amber, flint and blue glasses as 100% coarse aggregate replacement and non-reaction normal sand. Detailed expansion results up to 52 weeks were given elsewhere [14] but summarised results are shown in Fig. 4. As expected, all coarse glass aggregates were highly ASR reactive and failed BS 812-123 test at as early as 10 week. Since the test conditions for the BS 812-123 and C227 are the same (38°C/100% RH, which is regarded as the best for field concrete performance [29]), these results further confirm that all glass aggregates tested are ASR reactive when combined with HAPC and tested in moist conditions (see also Fig. 3).
https://static-content.springer.com/image/art%3A10.1617%2Fs11527-008-9396-4/MediaObjects/11527_2008_9396_Fig4_HTML.gif
Fig. 4

ASR expansion of coarse glass aggregate tested to BS 812-123 up to 52 weeks

3.3 Comparison of ASTM C1260 and C227

A comparison of the ASTM C1260 and C227 test results is shown in Fig. 5a–d. With the exception of blue glass, C1260 conditions cause lower expansion at early ages for flint, amber and green glasses. This raises questions about the effects of cement alkali levels and temperature on the time to initiation and rate of ASR reactions of glass aggregates in concrete and the suitability of the two methods for estimating how the aggregate might perform in a real concrete system. Additionally, on the practical side, the elevated temperature (80°C) used in the C1260 test is expensive to maintain, requires additional operator protection and health and safety training and may lead to errors in measuring the expansion due to thermal contraction effects if a rigid time-dependent measurement methodology is not maintained. For example, the assumption that elevated temperature promotes more rapid ASR appears to be untrue (Fig. 5) at least with the glass aggregates and cements used in this study. Additional work is required in this area to isolate and identify the effects of temperature and cement alkali level on the rate of ASR and to propose changes to C1260 such that reasonable correlation between the tests is achieved.
https://static-content.springer.com/image/art%3A10.1617%2Fs11527-008-9396-4/MediaObjects/11527_2008_9396_Fig5_HTML.gif
Fig. 5

Combination of ASTM C1260 and C227 results

3.4 Effect of bar size on apparent ASR expansion using the C1260 and C227 test methods

ASR expansion results of flint and green glass aggregates combined with HAPC and tested with the C1260 and C227 methods using different bar sizes, namely 25 × 25 × 280 mm (D25) and 40 × 40 × 160 mm (D40) are shown in Figs. 6 and 7 respectively.
https://static-content.springer.com/image/art%3A10.1617%2Fs11527-008-9396-4/MediaObjects/11527_2008_9396_Fig6_HTML.gif
Fig. 6

Effect of bar size on ASR expansion using the ASTM C1260 method and 25 × 25 × 280 mm (D25) and 40 × 40 × 160 mm (D40) mortar bars

https://static-content.springer.com/image/art%3A10.1617%2Fs11527-008-9396-4/MediaObjects/11527_2008_9396_Fig7_HTML.gif
Fig. 7

Effect of bar size on ASR expansion using the ASTM C227 method

Figure 6 shows that with the C1260 method, the longer bars (280 mm) with smaller cross-section (25 × 25 mm) start expanding earlier than shorter bars (160 mm) with larger cross-section (40 × 40 mm). With the C227 method, however, this expansion trend reverses: the expansion of shorter bars with larger cross-section is higher than that of longer bars with smaller cross-section, Fig. 7. Grattan-Bellew [30] has shown similar results. Lu et al. [31] examined the effect of bar length with equal cross sectional areas (40 × 40 × 160 & 40 × 40 × 285 mm) on alkali-carbonate reaction (ACR) using a test condition similar to C1260 test and found that the expansion results due to ACR of these bars are similar. Bakker [32] investigated the effect of bars with different cross sectional areas and same length (20 × 20 × 160, 40 × 40 × 160 and 100 × 100 × 160 mm) on ASR expansion tested to C227 and concluded that a significant increase in expansion was found as the bar cross-sectional area increased. These previous studies consistently suggest that the cross sectional area of the test bars plays an important role in the perceived ASR expansion. No reference known to the authors has offered an explanation of why this may be, nor addressed the issues that bar size may have on the C1260 test results.

3.5 Discussion

3.5.1 Effect of test type (C1260 vs. C227)

The test conditions of the C1260 and C227 mortar bar test methods are very different: the C1260 immerses specimens made with any cement in 1N NaOH solution at 80°C, whilst C227 stores specimens made with HAPC in damp cloth at 38°C/100% RH. With the C1260 test, because the main reactant (1N NaOH) ingresses from the sample perimeter, the reaction proceeds initially at the bar edges and moves inwards towards the centre at a rate controlled by the permeability of the test specimens. At early stages of reaction, the expansion induces cracks on the concrete surfaces, Fig. 8a, which may increase the rate of OH diffusion at later ages. A cross-section of glass aggregate concrete tested by the C1260 method then failed in bending is shown in Fig. 8b. This shows that the external perimeter of the bar is badly deteriorated but the interior appears relatively intact. This confirms the intrusive nature of the C1260 deterioration mechanism, which can be affected by the permeation properties of the cement paste.
https://static-content.springer.com/image/art%3A10.1617%2Fs11527-008-9396-4/MediaObjects/11527_2008_9396_Fig8_HTML.jpg
Fig. 8

(a) Typical ASR cracks and (b) cross-sectional optical image of glass concrete tested to ASTM C1260

In the C227 test, the alkali level of the cement paste is high and the reaction takes place throughout the specimen from the test initiation. However, some leaching of alkali into the wet cloth may occur and can be confirmed by simple pH testing. If the surface alkali levels reduce significantly due to the combination of consumption by reaction and leaching, it could be expected that the rate of ASR expansion might be affected. In this study, the expansion rates of the relatively faster-reacting blue and flint glasses started to reduce after 39 weeks, whilst for the relatively slower-reacting amber and green glasses, this did not occur until after 52 weeks, Fig. 3.

3.5.2 Effect of bar size

The above discussion can also explain the opposite effects of bar size on the C1260 and C227 tests. With C1260, the smaller the bar cross-sectional area is, the more quickly OH- ions can completely penetrate the pore system, initiate reaction and thus yield the higher expansions shown in Fig. 6. The opposite is true for C227, because of the consumption of alkalis due to ASR reaction and the leaching of alkalis out of the bars. Moreover, when the cement alkali level reduces beyond a certain point, the reaction rate slows and finally stops. Figure 7 shows that the rate of expansion of flint and green glass aggregate with the smaller cross sections decreases to zero after 8 and 26 weeks respectively, whilst the expansions are still increasing up to 52 weeks for larger cross section bars made with the same materials.

3.5.3 Time-lag of ASR expansion in the ASTM C1260 test

The notable time-lag in the C1260 expansion curves (Figs. 2 and 4) are consistent with the fact that the C1260 test yields results in direct relation to the rate of penetration of alkali from the 1N NaOH immersion solution. An additional concern that may apply is high degree of variability in the diffusion coefficients of cement pastes, particularly when pozzolanic materials are used as the pore refinement of these may be significantly accelerated by high temperatures [33].

To further demonstrate the effect of pozzolanic materials on ASR expansion, run-of-the station PFA (CPFA) has been used to replace 20 and 30% cement in flint glass aggregate mortars and concretes and their measured expansion results using the ASTM C1260 (25 × 25 × 285 mm bars) and the BS 812-123 [18] methods are shown in Fig. 9a and b respectively.
https://static-content.springer.com/image/art%3A10.1617%2Fs11527-008-9396-4/MediaObjects/11527_2008_9396_Fig9_HTML.gif
Fig. 9

Effect of CPFA on ASR expansion of flint glass aggregate tested using the ASTM C1260 and BS 812-123 methods

According to C1260 (Fig. 9a), the apparent effect of CPFA is to totally mitigate the ASR expansion of highly reactive flint glass aggregate up to a test age of 56 days (4 times longer than the recommended 14-day test duration in C1260). However the expansion results of the BS 812-123 test on the same flint glass aggregate, Fig. 9b, shows that the CPFA improves ASR resistance but nevertheless fails the BS 812-123 test. It is outside the scope of this paper to discuss the relative effectiveness of pozzolanic materials for mitigating ASR reactivity of glass aggregate but the different apparent reactivity with the ASTM C1260 and BS 812-123 tests suggests that the pozzolanic reaction of this material is stimulated by the high temperature (80°C) and 1 N NaOH solution used in the C1260 test [14]. Thus there may be a propensity for the C1260 test to yield false negatives and caution is recommended with this test, particularly when highly reactive aggregates, such as glass, are used in concrete.

The above discussion can also explain the time-lag behaviour of the ASR expansion in the C1260 test. It has been found [7, 14] that glass in a finely ground form has pozzolanic reactivity and that [9, 14] a reduction of ASR expansion was observed when the size of glass particles was less than 1.18 mm when C1260 test conditions are used up to a test age of 189 days (see Fig. 10 as an example). It can be seen from Fig. 10 that concrete made with flint glass particles <1.18 mm exhibits similar or even less expansion than the control mix, which implies a degree of ASR mitigation for these particle size ranges, whilst ASR expansion rates increase with particle size above 1.18 mm. According to the ASTM grading requirement (Table 3), a total of 65% of the glass aggregate used to make test mortars was less than 1.18 mm and 40% less than 0.6 mm. Therefore, it is expected that this large proportion of fine glass particles may react pozzolanically to some extent at 80°C 1N NaOH solution and partially mitigate ASR reaction of coarse glass particles, at least within the 14 days of the C1260 test duration, Figs. 12. Thus it is further demonstrated that the C1260 test is unsuitable for assessing the ASR reactivity of glass aggregate.
https://static-content.springer.com/image/art%3A10.1617%2Fs11527-008-9396-4/MediaObjects/11527_2008_9396_Fig10_HTML.gif
Fig. 10

Effect of flint glass particle size on ASR reactivity (ASTM C1260 condition)

4 Conclusions

A number of points have been raised in this paper about the relative reactivities of glass aggregates and the variable results that may be observed with C1260 and C227 test methods.
  1. (i)
    Relative reactivity of glass aggregates
    1. (1)

      All four colours of glass aggregates assessed in the research were very alkali-silica reactive when tested using ASTM C1260 (with extended test duration) and ASTM C227 (within normal test duration). However the reaction rate varies and appears to be ranked as follows: blue > flint > amber > green.

       
     
  2. (ii)
    Comparison of C1260 and C227 methods
    1. (1)

      The C1260 test method did not indicate that the glass aggregates under test were ASR-reactive (except for the blue glass), despite them appearing to be reactive when tested using the C227 method.

       
    2. (2)

      The bar size used in the reported tests has a significant and opposite effect on the perceived ASR potential of glass aggregates tested by the two methods. With the C1260 method, where the aggressive alkali permeates into the sample, the smaller the cross-sectional area used, the faster the reaction. With the C227 method where alkali may leach out of the specimen, the opposite is true.

       
     
  3. (iii)
    Appropriate AST testing of glass aggregate
    1. (1)

      The C1260 test conditions (80°C and 1N NaOH solution) are likely to promote highly accelerated reaction of any pozzolanic material used to make the test specimens. In the case of glass aggregate, the grading is such that some 40% of the total aggregate under test may be pozzolanic. Thus it is inevitable that the C1260 test method will yield unreliable false negative results when used to test glass aggregates. It is therefore suggested that the particle size of glass aggregates should be 1.18 mm when C1260 test method is used.

       
    2. (2)

      The C227 test method, due to the aggregate grading, may also be partially susceptible to the failings of the C1260 test method, albeit to a less degree because of the lower temperature (38°C).

       
    3. (3)

      Where possible, concrete prism test methods such as BS 812-123 and ASTM C1293 conducted at 38°C/100% RH should be used in the first instance [14].

       
     

Acknowledgments

The authors are grateful to The Waste & Resources Action Programme for partially funding this research work and to Mr. Kieran Nash for his extensive technical support on this wide research study.

Copyright information

© RILEM 2008