- Email: [email protected]
Alkali aggregate reaction in a 60-year-old dam in Australia
The International Journal of Cement Composites and Lightweight Concrete, Volume 10, Number4
November 1988
A l k a l i a g g r e g a t e reaction in a 60-yearold d a m in A u s t r a l i a A. Shayan *
Synopsis Examination of drilled cores taken from a 60-year-old dam in Australia has shown that the dam has suffered from alkali aggregate reaction. Cracking has been observed on the crest of the dam and its concrete railing. The aggregate used in the concrete is a local dacite excavated from the dam site, and shows strong reaction rims particularly in the upper 10-12 m of the dam wall. Deeper portions of the dam wall appear to be free from reaction rims but the reasons for this have not been investigated. The reaction product was characterised by X-ray diffraction, infrared spectroscopy, scanning electron microscopy, and electron micro-probe analysis and was shown to be similar to previously reported alkali aggregate reaction products. New aggregate from the same locality as that used in the dam was tested for alkali reactivity and was found to be reactive with alkali in concrete. Keywords Alkali aggregate reactions, dams, expansion, dacite, core testing, concrete durability, cracking (fracturing), accelerated tests, electron microscopy, infrared spectroscopy, concrete construction.
INTRODUCTION Reported cases of alkali aggregate reaction (AAR) in Australia are very few, although there are a number of structures including a major jetty, bridges, dams and possibly buildings affected by it. Cole et al. [1] reported the occurrence of AAR in an old Australian dam, which was recently re-examined [2] and confirmed as an AAR case. A bridge in Perth, Western Australia, was also reported [3] to have suffered from AAR. Aggregates from the same sources as those used in the dam and the bridge were experimentally shown to be reactive [2, 4]. Another Australian dam containing a local dacite aggregate has recently been inspected and found to have developed cracks in various parts of the dam. Figure 1 shows pattern-cracking observed on the dam crest, and Figure 2 illustrates the cracking of pillars supporting the concrete hand railing along the sides of the crest. A few old cracks, now filled with dirt and vegetation, were also observed in the dam apron (downstream). This paper reports on the investigation of concrete cores drilled from the dam, and on laboratory testing of dacite aggregate from the same source as that used in the dam.
* CSIRO Division of Building, Construction and Engineering, P.O. Box 56, Highett, Victoria 3190, Australia. Received 29 July 1988 Accepted 31 August 1988 ~) CSIRO Australia 1988
0262-5075/81910408259/$02.00
EXPERIMENTAL WORK Concrete extracted from the dam Many cores (63 mm diameter) totalling over 300m long had previously been drilled from various parts of the dam by the controlling authority in order to assess the overall condition of the dam, with particular reference to cracking and permeability. However, investigation of AAR had not been part of the assessment. These cores had been taken at various angles and directions from the top of the crest and from the drainage gallery so that the dam wall had been sampled over its full height and length. After the author noted strong symptoms of AAR in one core, all the cores were broken at 300 mm interval, or less if necessary, in order to determine the extent of the reaction in the dam wall. Selected samples containing reacted and unreacted aggregate pieces were used for determining the petrographic and mineralogical properties (X-ray diffraction (XRD)) of the aggregate, and for observing the extent of microcracking and the presence of gel in the concrete matrix and aggregate particles. Samples of the reaction product were removed carefully with a blade and, after being powdered, were subjected to XRD and infrared spectroscopic analyses in order to determine their mineralogical composition. Powder samples of about 2 mg were pressed under high pressure into smooth, flat flakes for elemental analysis using quantitative electron microprobe analysis (EMPA) by the energy dispersive method. Pieces of concrete containing reacted
259
Alkali aggregate reaction in a 60-year-old dam in Australia
~.'nakm ~
or 1.74% Na20 equivalent. These mortar bars were stored under the conditions of the standard method (38 4- 2oc, 100% RH). One set of mortar bars containing no added alkali (i.e. containing cement of 0.84% alkali) was subjected to the accelerated test as described by Shayan et al. [6]. Concrete prisms (75 x 75 x 285mm) were made using a 20mm nominal single-sized aggregate and a known sand. The cement mentioned above was used (407kg/m 3) with the mix proportions of coarse aggregate: sand: cement :water of 2.62: 1.55: 1 : 0.46, The three levels of alkali and the test procedures used for the mortar bars were also used for the concrete prisms.
Figure 1 Pattern cracking on the crest of the dam
aggregate were employed for scanning electron microscopy (SEM) and qualitative energy dispersive X-ray analysis (EDS) to determine the morphology of the reaction products and their qualitative compositions.
RESULTS A N D DISCUSSION Concrete from the dam Inspection of the drilled cores showed that the reaction rims were limited to the top 12m of the dam wall, as cores from deeper portions and those drilled from the drainage gallery at different directions did not show any sign of reaction. In the 12m of reacted concrete, the reaction varied from mild to strong as judged by the
Laboratory testing of the dacite aggregate Many spalls of the dacite aggregate were provided by the authority controlling the dam to correspond as closely as possible with the rock that was excavated from the dam site and used in the construction of the dam. Petrographic and XRD analyses were done on representative samples of spalls, after which they were crushed and bulked for testing. A powdered (<751~m) sub-sample was used to determine the elemental composition of the dacite. Mortar bars were made according to the Australian Standard AS 1141-38 [5] using a portland cement containing 0.84% available alkali (Na20 equivalent). In addition, mortar bars were made in which the level of available alkali was raised by the addition of NaOH to 1.38
Figure 3 Broken surface of a core showing strong black reaction rims and white rims of reaction product within the aggregate boundaries
Figure 2 Cracking in pillars of the concrete railings
260
severity of rimming and the amount of reaction product present, Figure 3 shows the broken surface of a core in which strong reaction rims and a considerable amount of reaction product are visible in all of the dacite aggregate particles. Cores from deeper parts of the dam that did not show reaction rims also contained the same aggregate type, possibly indicating that differences in the cement type and content between the two parts of the dam could have been responsible for the observed difference in the reactivity. The unreacted portion was not investigated further.
Alkafi aggregate reaction in a 60-year-old dam in Australia
Shayan
Figure 4 (a) A quartz phenocryst with rounded edges, embayed with cryptocrystalline quartz, and (b) groundmass of the dacite aggregate showing a cryptocrystalline to granular texture. Cross-polarised light
Characterisation of the aggregate The only coarse aggregate used in the dam wall, both in the reacted and unreacted parts (and also the spalls that were provided for testing), is a medium to dark grey porphyritic dacite. Petrographic examination of thin sections showed abundant quartz phenocrysts ranging in size from 1-4mm, and often being rounded and embayed with cryptocrystalline quartz as shown in Figure 4a. These features of the quartz phenocrysts may have arisen due to enhanced resorption as a result of interaction of gas bubbles with the quartz in the superheated melt [7] and may have rendered the quartz reactive with alkali [8]. Other phenocrysts include plagioclase and perthitic alkali feldspar and pleochroic hypersthene and smaller crystals of biotite, which all show some degree of alteration. The groundmass of the dacite varied in texture from cryptocrystalline to granular (Figure 4b), containing quartz and feldspar. XRD analysis also indicated quartz and feldspar as the major phases and n,ica and chlorite as the minor phases; mica originated from the sericitisation of the feldspar and chlorite from the alteration of the hypersthene. Cryptocrystalline quartz is considered to be a
Table 1 Chemical composition of the dacite aggregate*
SiO2 TiO2 AI203 Fe203 MnO MgO CaO Na20 K20 P205 Ignition loss Total
67.4 0.62 14.2 3.66 0.05 1.22 2.42 2.90 2.98 0.17 3.52 99.14
* Analysis by the Australian Mineral Development Laboratories
reactive component in aggregates [8, 9]. The siliceous nature of the dacite aggregate is also apparent from its chemical composition (Table 1). Examination of the concrete and the reaction products Whereas no cracking was evident in the aggregates in the unreacted concrete, macrocracking was common in reacted aggregates (Figure 5a), a feature also observed in reacted aggregates from other dams [1, 10, 11] and from structures reported to have suffered AAR [3, 12]. Thin sections of the reacted concrete also showed gel-filled microcracks in the reacted aggregates, extending into the mortar (Figure 5b), a characteristic of AAR. Figure 6a shows a scanning electron micrograph of the boundary between the reacted aggregate and the mortar. The cracked gel (top) corresponds to the surfaces of the black rims and their contiguous mortar in Figure 3, and the band labelled with a triangle is the reaction product shown as white rims within the aggregate boundary in Figure 3. The surface of the aggregate away from the band of the reaction product towards the centre (bottom of photograph) is also covered with islands of the reaction product. The white spots seen on the gel surface in Figure 6a represent outgrowths of fibrous crystals (Figure 6b) but the gel itself shows a massive non-crystalline structure (Figure 6c). The band of reaction product (Figure 6a) consists of platy crystalline materials (Figure 6d) which, under higher magnification, appear to be growing out of a massive non-crystalline reaction product as seen in Figure 6e. The qualitative compositions of the phases shown in Figure 6, obtained by EDS analysis, are shown in Figure 7. The gel and its fibrous outgrowths are calcium silicates containing small amounts of sodium and potassium and both yielded similar energy dispersive spectra (Figure 7a), although the fibrous crystals contained less Na. The reaction band shown in Figure 6a is similar but contained much less sodium than potassium; the massive material in Figures 6d and 6e yielded the spectrum
261
Alkali aggregate reaction in a 60-year-old dam in Australia
5hayan
Figure 5 (a) View of a cracked portion of a core from the upper 12m of the dam. The cracks run through the aggregate particles and in the mortar. (b) View of thin section of a reacted core showing a gel-filled crack through the aggregate, that extends into the mortar (far left). Plane polarised light
shown in Figure 7b, and the platy materials (Figure 6e) produced the spectrum in Figure 7c. In view of differences in the surface roughness of the two materials and the effect of roughness on the intensity of emitted X-rays, spectra b and c should be considered similar; both indicating the alkali-rich nature of the product. Spectra similar to those in Figure 7 were reported by Davies and Oberholster [13] for the various phases of AAR products in several concretes. Results of quantitative EMPA done on pressed flakes of the white reaction product in the core (Figure 3) are shown in Table 2. Two phases, one alkali-rich and the other calcium-rich, could be recognised; the former is that shown in Figure 6d, which forms the bulk of the flakes, but the exact nature of the other phase is unclear. It must be emphasised that out of 18 analyses only four were of the calcium-rich phase, so that it is a minor constituent of the analysed flakes. Apart from minor differences, the composition of the alkali-rich phase is similar to those of the AAR products reported in the literature [3, 13, 14, 15].
Table 2 Results of EMPA of the reaction product*
SiO2 AI203 Fe203 CaO Na20 K20
Alkali-rich phase
Calcium-rich phase
67.00 0.40 0.55 16.26 5.25 10.54
26.44 3.77 0.80 64.54 1.54 3.01
* On the basis of unhydrous material. The average total for the analyses for the alkali-rich phase was about 82 %, leaving 18 % that could be attributed to H20 and CO2
262
The XRD pattern of the reaction product showed a strong peak at about 12 A, characteristic of the crystalline AAR product, and weaker peaks as shown in Figure 8. The peaks at 4.24, 3.33, 2.47 and 2.26/&, are due to quartz, and that at 3.03A due to calcite. The quartz produced diagnostic absorption bands at 780 and 795 cm -1 and the calcite at 875 and 1430 cm -1 in the infrared spectrum of the reaction product (Figure 9). Considering that the AAR product is unstable and the lattice spacings vary with heat and changes in relative humidity, the XRD pattern shown in Figure 8 is very similar to those reported for other AAR products [1,3, 13, 14] and is probably composed of peaks due to more than one phase, or phases with different hydration states. The hump between 18 and 32 ° 28 in the XRD pattern is due to amorphous silica-rich material, consistent with the SEM observations of the reaction product. The infrared spectrum of this AAR product (Figure 9) is very similar to that of the AAR product from a bridge [3] and both appear to show more pronounced absorption bands than that of the AAR product from an aged concrete [14], although this difference could have arisen because a better instrument was used for the former ones. The foregoing observations all support the existence of AAR in the uppermost 12 m of the dam.
Laboratory testing of t h e decite aggregate
Mortar bars and concrete prisms were subjected to the accelerated test [6] for quickly evaluating the potential reactivity of the dacite. Results shown in Figure 10 clearly indicate the reactivity of the aggregate because the mortar bar expansion has far exceeded the proposed criterion of 0.1% expansion at 10 days [6]. The results also confirm that concrete prisms are unsuitable for the accelerated test as suggested earlier [6]. Expansions of mortar bars and concrete prisms made at the different
Alkafi aggregate reaction in a 60-year-old dam in Australia
Shayan
Figure 6 Scanning electron micrographs of a reaction zone showing (a) gel covering the black rim of an aggregate (Figure 3) and its contiguous mortar (top), the white reaction product (triangle), and islands of reaction product covering the inner part of the aggregate (bottom), (b) surface of the gel, shown in (a), with fibrous outgrowths of a calcium silicate, (c) detail of the gel surface showing its massive nature, (d) detail of the reaction product (triangle in (a)) showing crystalline material, and (e) detail of platy crystals which appear to be growing out of a massive gel-like material
263
Alkali aggregate reaction in a 60-year-old dam in Australia
~.nay,ar~
Si £a
o
fn rrt
J _13
I
,I
I
40
35
30
I
I
25 20 DEGREE5 29(£u Ko<}
I
I
15
10
i
5
I
2
Figure 8 X-ray powder diffraction pattern of the white reaction product (Figures 3 and 6d). Q - quartz and C = calcite. Spacings are given in A units
,b
J
K
~ I
I
0
Si
lu ,I
2
I
I
I
/, ENERGY (KeV)
I
6
C I
I
8
Figure 7 Energy dispersive X-ray spectra showing the qualitative composition of (a) gel phase shown in Figure 6a. The fibrous crystals in Figure 6b also had this composition; (b) and (c) the massive material and the platy crystals shown in Figure 6e respectively, showing their alkali-rich nature
alkali levels and stored at 38 _+ 2°C and 100% RH are shown in Figure 11. Although the mortar bars containing even the highest level of alkali and concrete prisms without added alkali (i.e. 0.84% Na20 equivalent) showed very little expansion at one year, concrete prisms containing cement with 1.38 and 1.74% alkali showed considerable expansion and cracked at 20 and 24 weeks respectively (Figure 11 ). The different expansion behaviours of mortar bars and concrete prisms are not unusual and have been observed by other research-
264
ers [16, 17]. The lack of significant expansions in concrete prisms in the accelerated test may be due to the denser and less pervious nature of the cement paste in the concrete relative to that in mortar bar (i,e, lesser access of the NaOH solution to the aggregates in concrete), and this is under investigation. The laboratory tests clearly show the potential reactivity of the dacite aggregate in high-alkali concrete. The lack of reaction in the lower parts and the extensive reaction in the upper 12m of the dam indicates that either the original cements used were of different alkali contents, or probably alkali was lost due to leaching from the lower parts, but became concentred due to evaporation in the upper part and caused AAR in this restricted zone. Such differential reactivities have been observed in other dams [11] which allegedly arose due to the use of cements with different alkali contents, although concen-
I
~
I
3OOO
I
2O0O
1500
1000 800 (~0 /~0
Figure 9 Infrared absorption spectrum of the white reaction product. Q = quartz, C -- calcite and W = water
Alkali aggregate reaction in a 60-year-old dam in Australia
0"6
Shayan
tration of alkali in certain parts of a structure could occur due to moisture migration [18]. However, concretes from the various parts of the dam need to be fully investigated for a better understanding of the reaction and its absence in different parts of the dam.
-
~5
0.1+ o~ ACKNOWLEDGEMENT
z0.3
The author thanks the authority responsible for the dam for access to the drilled cores. He also thanks I. Ivanusec and R. Diggins of this Division for making and measuring the length changes of the laboratory concrete and mortar specimens.
x 0"2
0"1 CONERETE REFERENCES 0
10
20
30
50
40
60
TIME(DAY) Figure 10 Expansion of mortar bars and concrete prisms subjected to the accelerated test
0.1/,
0"12
0.10
(1.38%) z
0-08
g z
~x
o.06
ILl
0"04 ---- M(1.7/,%) M(1-25%)
0"02
I
10
20 30 TIME(WEEK)
/,0
C(0-84%) I
50
Figure 11 Expansion of mortar bars (M) and concrete prisms (C) made at different alkali levels and stored at 38 _+ 2°C and 100% RH
1. Cole, W. F., Lancucki, C. J. and Sandy, M. J. 'Products formed in an aged concrete', Cement and Concrete Research, Vol. 11, No. 3, May 1981, pp. 443-54. 2. Shayan, A. 'Re-examination of AAR in an old concrete', Cement and Concrete Research (in press). 3. Shayan, A. and Lancucki, C. J. 'Alkali-aggregate reaction in the Causeway Bridge, Perth, Western Australia', in 'Concrete alkali-aggregate reaction', edited by P. E. Grattan-Bellew, Noyes Publications, New Jersey, USA, 1987, pp. 392-7. 4. Shayan, A., Diggins, R., Ritchie, D. F. and Westgate, P. 'Evaluation of Western Australian aggregates for alkali-reactivity in concrete', in 'Concrete alkali aggregate reaction', edited by P. E. Grattan-Bellew, Noyes Publications, New Jersey, USA, 1987, pp. 247-52. 5. Standards Association of Australia, Australian Standard AS 1141 - Section 38, 'Potential alkali reactivity by mortar bar', 1974. 6. Shayan, A., Diggins, R., Ivanusec, I. and Westgate, P. 'Accelerated testing of some Australian and overseas aggregates for alkali-aggregate reactivity', Cement and Concrete Research (in press). 7. Donaldson, C. H. and Henderson, C. M. B. 'A new intepretation of round embayments in quartz crystals', Mineralogical Magazine, Vol. 52, 1988, pp. 27-33. 8. Dolar-Mantuani, L. 'Handbook of concrete aggregates, a petrographic and technological evaluation', Noyes Publications, New Jersey, USA, 1983, pp. 79-112. 9. Mielenz, R. C. 'Petrographic examination of concrete aggregate to determine potential alkali reactivity', National Research Council, Washington Transportation Research Board, Research Report No. 18-C, 1958, pp. 29-35. 10. Mather, K. 'Condition of concrete in Martin Dam after 50 years of service', Cement, Concrete and Aggregates, Vol. 3, 1981, pp. 53-62.
265
Alkali aggregate reaction in a 60-year-old dam in Australia
11.
12.
13.
14.
15.
Stark, D. 'Alkali-silica reactivity in five dams in South Western United States', United States Department of the Interior, Bureau of Reclamation, Report REC-ERC-85-10, 1985, 64pp. Idorn, G. M. 'Durability of concrete structures in Denmark', Technical University of Denmark, Copenhagen, 1967, pp. 84-136. Davies, G. and Oberholster, R. E. 'The alkali-silica reaction product, a mineralogical and an electron microscopic study', Proceedings of the 8th International Conference on Cement Microscopy, Orlando, Florida, USA, 1986, pp. 303-26. Cole, W. F. and Lancucki, C. J. 'Products formed in an aged concrete, the occurrence of okenite', Cement and Concrete Research, Vol. 13, No. 5, September 1983, pp. 611-8. Oberholster, R. E. 'Alkali reactivity of siliceous aggregates: diagnosis of the reaction, testing of
266
.Shavar,
16.
17.
18.
cement and aggregate and prescription of preventive measures', Proceedings of 6th International Conference on Alkalis in Concrete, Copenhagen, Denmark, 1983, pp. 419-33. Grattan-Bellew, P. E. and Litvan, G. G. 'Testing Canadian aggregates for alkali expansivity', Proceedings of Symposium on the Effect of Alkali on the Properties of Concrete, London, UK, 1976, pp. 227-45. Oberholster, R. E. 'Alkali-aggregate reaction in South Africa - A review', Proceedings of 5th International Conference on Alkali-Aggregate Research in Concrete, Cape Town, South Africa, 1981, Paper $252/8. Nixon, P. J., Collins, R. J. and Rayment, P. L. 'The concentration of alkalies by moisture migration in concrete - a factor influencing alkali aggregate reaction', Cement and Concrete Research, Vol. 9, No. 4, July 1979, pp. 417-23.
November 1988
A l k a l i a g g r e g a t e reaction in a 60-yearold d a m in A u s t r a l i a A. Shayan *
Synopsis Examination of drilled cores taken from a 60-year-old dam in Australia has shown that the dam has suffered from alkali aggregate reaction. Cracking has been observed on the crest of the dam and its concrete railing. The aggregate used in the concrete is a local dacite excavated from the dam site, and shows strong reaction rims particularly in the upper 10-12 m of the dam wall. Deeper portions of the dam wall appear to be free from reaction rims but the reasons for this have not been investigated. The reaction product was characterised by X-ray diffraction, infrared spectroscopy, scanning electron microscopy, and electron micro-probe analysis and was shown to be similar to previously reported alkali aggregate reaction products. New aggregate from the same locality as that used in the dam was tested for alkali reactivity and was found to be reactive with alkali in concrete. Keywords Alkali aggregate reactions, dams, expansion, dacite, core testing, concrete durability, cracking (fracturing), accelerated tests, electron microscopy, infrared spectroscopy, concrete construction.
INTRODUCTION Reported cases of alkali aggregate reaction (AAR) in Australia are very few, although there are a number of structures including a major jetty, bridges, dams and possibly buildings affected by it. Cole et al. [1] reported the occurrence of AAR in an old Australian dam, which was recently re-examined [2] and confirmed as an AAR case. A bridge in Perth, Western Australia, was also reported [3] to have suffered from AAR. Aggregates from the same sources as those used in the dam and the bridge were experimentally shown to be reactive [2, 4]. Another Australian dam containing a local dacite aggregate has recently been inspected and found to have developed cracks in various parts of the dam. Figure 1 shows pattern-cracking observed on the dam crest, and Figure 2 illustrates the cracking of pillars supporting the concrete hand railing along the sides of the crest. A few old cracks, now filled with dirt and vegetation, were also observed in the dam apron (downstream). This paper reports on the investigation of concrete cores drilled from the dam, and on laboratory testing of dacite aggregate from the same source as that used in the dam.
* CSIRO Division of Building, Construction and Engineering, P.O. Box 56, Highett, Victoria 3190, Australia. Received 29 July 1988 Accepted 31 August 1988 ~) CSIRO Australia 1988
0262-5075/81910408259/$02.00
EXPERIMENTAL WORK Concrete extracted from the dam Many cores (63 mm diameter) totalling over 300m long had previously been drilled from various parts of the dam by the controlling authority in order to assess the overall condition of the dam, with particular reference to cracking and permeability. However, investigation of AAR had not been part of the assessment. These cores had been taken at various angles and directions from the top of the crest and from the drainage gallery so that the dam wall had been sampled over its full height and length. After the author noted strong symptoms of AAR in one core, all the cores were broken at 300 mm interval, or less if necessary, in order to determine the extent of the reaction in the dam wall. Selected samples containing reacted and unreacted aggregate pieces were used for determining the petrographic and mineralogical properties (X-ray diffraction (XRD)) of the aggregate, and for observing the extent of microcracking and the presence of gel in the concrete matrix and aggregate particles. Samples of the reaction product were removed carefully with a blade and, after being powdered, were subjected to XRD and infrared spectroscopic analyses in order to determine their mineralogical composition. Powder samples of about 2 mg were pressed under high pressure into smooth, flat flakes for elemental analysis using quantitative electron microprobe analysis (EMPA) by the energy dispersive method. Pieces of concrete containing reacted
259
Alkali aggregate reaction in a 60-year-old dam in Australia
~.'nakm ~
or 1.74% Na20 equivalent. These mortar bars were stored under the conditions of the standard method (38 4- 2oc, 100% RH). One set of mortar bars containing no added alkali (i.e. containing cement of 0.84% alkali) was subjected to the accelerated test as described by Shayan et al. [6]. Concrete prisms (75 x 75 x 285mm) were made using a 20mm nominal single-sized aggregate and a known sand. The cement mentioned above was used (407kg/m 3) with the mix proportions of coarse aggregate: sand: cement :water of 2.62: 1.55: 1 : 0.46, The three levels of alkali and the test procedures used for the mortar bars were also used for the concrete prisms.
Figure 1 Pattern cracking on the crest of the dam
aggregate were employed for scanning electron microscopy (SEM) and qualitative energy dispersive X-ray analysis (EDS) to determine the morphology of the reaction products and their qualitative compositions.
RESULTS A N D DISCUSSION Concrete from the dam Inspection of the drilled cores showed that the reaction rims were limited to the top 12m of the dam wall, as cores from deeper portions and those drilled from the drainage gallery at different directions did not show any sign of reaction. In the 12m of reacted concrete, the reaction varied from mild to strong as judged by the
Laboratory testing of the dacite aggregate Many spalls of the dacite aggregate were provided by the authority controlling the dam to correspond as closely as possible with the rock that was excavated from the dam site and used in the construction of the dam. Petrographic and XRD analyses were done on representative samples of spalls, after which they were crushed and bulked for testing. A powdered (<751~m) sub-sample was used to determine the elemental composition of the dacite. Mortar bars were made according to the Australian Standard AS 1141-38 [5] using a portland cement containing 0.84% available alkali (Na20 equivalent). In addition, mortar bars were made in which the level of available alkali was raised by the addition of NaOH to 1.38
Figure 3 Broken surface of a core showing strong black reaction rims and white rims of reaction product within the aggregate boundaries
Figure 2 Cracking in pillars of the concrete railings
260
severity of rimming and the amount of reaction product present, Figure 3 shows the broken surface of a core in which strong reaction rims and a considerable amount of reaction product are visible in all of the dacite aggregate particles. Cores from deeper parts of the dam that did not show reaction rims also contained the same aggregate type, possibly indicating that differences in the cement type and content between the two parts of the dam could have been responsible for the observed difference in the reactivity. The unreacted portion was not investigated further.
Alkafi aggregate reaction in a 60-year-old dam in Australia
Shayan
Figure 4 (a) A quartz phenocryst with rounded edges, embayed with cryptocrystalline quartz, and (b) groundmass of the dacite aggregate showing a cryptocrystalline to granular texture. Cross-polarised light
Characterisation of the aggregate The only coarse aggregate used in the dam wall, both in the reacted and unreacted parts (and also the spalls that were provided for testing), is a medium to dark grey porphyritic dacite. Petrographic examination of thin sections showed abundant quartz phenocrysts ranging in size from 1-4mm, and often being rounded and embayed with cryptocrystalline quartz as shown in Figure 4a. These features of the quartz phenocrysts may have arisen due to enhanced resorption as a result of interaction of gas bubbles with the quartz in the superheated melt [7] and may have rendered the quartz reactive with alkali [8]. Other phenocrysts include plagioclase and perthitic alkali feldspar and pleochroic hypersthene and smaller crystals of biotite, which all show some degree of alteration. The groundmass of the dacite varied in texture from cryptocrystalline to granular (Figure 4b), containing quartz and feldspar. XRD analysis also indicated quartz and feldspar as the major phases and n,ica and chlorite as the minor phases; mica originated from the sericitisation of the feldspar and chlorite from the alteration of the hypersthene. Cryptocrystalline quartz is considered to be a
Table 1 Chemical composition of the dacite aggregate*
SiO2 TiO2 AI203 Fe203 MnO MgO CaO Na20 K20 P205 Ignition loss Total
67.4 0.62 14.2 3.66 0.05 1.22 2.42 2.90 2.98 0.17 3.52 99.14
* Analysis by the Australian Mineral Development Laboratories
reactive component in aggregates [8, 9]. The siliceous nature of the dacite aggregate is also apparent from its chemical composition (Table 1). Examination of the concrete and the reaction products Whereas no cracking was evident in the aggregates in the unreacted concrete, macrocracking was common in reacted aggregates (Figure 5a), a feature also observed in reacted aggregates from other dams [1, 10, 11] and from structures reported to have suffered AAR [3, 12]. Thin sections of the reacted concrete also showed gel-filled microcracks in the reacted aggregates, extending into the mortar (Figure 5b), a characteristic of AAR. Figure 6a shows a scanning electron micrograph of the boundary between the reacted aggregate and the mortar. The cracked gel (top) corresponds to the surfaces of the black rims and their contiguous mortar in Figure 3, and the band labelled with a triangle is the reaction product shown as white rims within the aggregate boundary in Figure 3. The surface of the aggregate away from the band of the reaction product towards the centre (bottom of photograph) is also covered with islands of the reaction product. The white spots seen on the gel surface in Figure 6a represent outgrowths of fibrous crystals (Figure 6b) but the gel itself shows a massive non-crystalline structure (Figure 6c). The band of reaction product (Figure 6a) consists of platy crystalline materials (Figure 6d) which, under higher magnification, appear to be growing out of a massive non-crystalline reaction product as seen in Figure 6e. The qualitative compositions of the phases shown in Figure 6, obtained by EDS analysis, are shown in Figure 7. The gel and its fibrous outgrowths are calcium silicates containing small amounts of sodium and potassium and both yielded similar energy dispersive spectra (Figure 7a), although the fibrous crystals contained less Na. The reaction band shown in Figure 6a is similar but contained much less sodium than potassium; the massive material in Figures 6d and 6e yielded the spectrum
261
Alkali aggregate reaction in a 60-year-old dam in Australia
5hayan
Figure 5 (a) View of a cracked portion of a core from the upper 12m of the dam. The cracks run through the aggregate particles and in the mortar. (b) View of thin section of a reacted core showing a gel-filled crack through the aggregate, that extends into the mortar (far left). Plane polarised light
shown in Figure 7b, and the platy materials (Figure 6e) produced the spectrum in Figure 7c. In view of differences in the surface roughness of the two materials and the effect of roughness on the intensity of emitted X-rays, spectra b and c should be considered similar; both indicating the alkali-rich nature of the product. Spectra similar to those in Figure 7 were reported by Davies and Oberholster [13] for the various phases of AAR products in several concretes. Results of quantitative EMPA done on pressed flakes of the white reaction product in the core (Figure 3) are shown in Table 2. Two phases, one alkali-rich and the other calcium-rich, could be recognised; the former is that shown in Figure 6d, which forms the bulk of the flakes, but the exact nature of the other phase is unclear. It must be emphasised that out of 18 analyses only four were of the calcium-rich phase, so that it is a minor constituent of the analysed flakes. Apart from minor differences, the composition of the alkali-rich phase is similar to those of the AAR products reported in the literature [3, 13, 14, 15].
Table 2 Results of EMPA of the reaction product*
SiO2 AI203 Fe203 CaO Na20 K20
Alkali-rich phase
Calcium-rich phase
67.00 0.40 0.55 16.26 5.25 10.54
26.44 3.77 0.80 64.54 1.54 3.01
* On the basis of unhydrous material. The average total for the analyses for the alkali-rich phase was about 82 %, leaving 18 % that could be attributed to H20 and CO2
262
The XRD pattern of the reaction product showed a strong peak at about 12 A, characteristic of the crystalline AAR product, and weaker peaks as shown in Figure 8. The peaks at 4.24, 3.33, 2.47 and 2.26/&, are due to quartz, and that at 3.03A due to calcite. The quartz produced diagnostic absorption bands at 780 and 795 cm -1 and the calcite at 875 and 1430 cm -1 in the infrared spectrum of the reaction product (Figure 9). Considering that the AAR product is unstable and the lattice spacings vary with heat and changes in relative humidity, the XRD pattern shown in Figure 8 is very similar to those reported for other AAR products [1,3, 13, 14] and is probably composed of peaks due to more than one phase, or phases with different hydration states. The hump between 18 and 32 ° 28 in the XRD pattern is due to amorphous silica-rich material, consistent with the SEM observations of the reaction product. The infrared spectrum of this AAR product (Figure 9) is very similar to that of the AAR product from a bridge [3] and both appear to show more pronounced absorption bands than that of the AAR product from an aged concrete [14], although this difference could have arisen because a better instrument was used for the former ones. The foregoing observations all support the existence of AAR in the uppermost 12 m of the dam.
Laboratory testing of t h e decite aggregate
Mortar bars and concrete prisms were subjected to the accelerated test [6] for quickly evaluating the potential reactivity of the dacite. Results shown in Figure 10 clearly indicate the reactivity of the aggregate because the mortar bar expansion has far exceeded the proposed criterion of 0.1% expansion at 10 days [6]. The results also confirm that concrete prisms are unsuitable for the accelerated test as suggested earlier [6]. Expansions of mortar bars and concrete prisms made at the different
Alkafi aggregate reaction in a 60-year-old dam in Australia
Shayan
Figure 6 Scanning electron micrographs of a reaction zone showing (a) gel covering the black rim of an aggregate (Figure 3) and its contiguous mortar (top), the white reaction product (triangle), and islands of reaction product covering the inner part of the aggregate (bottom), (b) surface of the gel, shown in (a), with fibrous outgrowths of a calcium silicate, (c) detail of the gel surface showing its massive nature, (d) detail of the reaction product (triangle in (a)) showing crystalline material, and (e) detail of platy crystals which appear to be growing out of a massive gel-like material
263
Alkali aggregate reaction in a 60-year-old dam in Australia
~.nay,ar~
Si £a
o
fn rrt
J _13
I
,I
I
40
35
30
I
I
25 20 DEGREE5 29(£u Ko<}
I
I
15
10
i
5
I
2
Figure 8 X-ray powder diffraction pattern of the white reaction product (Figures 3 and 6d). Q - quartz and C = calcite. Spacings are given in A units
,b
J
K
~ I
I
0
Si
lu ,I
2
I
I
I
/, ENERGY (KeV)
I
6
C I
I
8
Figure 7 Energy dispersive X-ray spectra showing the qualitative composition of (a) gel phase shown in Figure 6a. The fibrous crystals in Figure 6b also had this composition; (b) and (c) the massive material and the platy crystals shown in Figure 6e respectively, showing their alkali-rich nature
alkali levels and stored at 38 _+ 2°C and 100% RH are shown in Figure 11. Although the mortar bars containing even the highest level of alkali and concrete prisms without added alkali (i.e. 0.84% Na20 equivalent) showed very little expansion at one year, concrete prisms containing cement with 1.38 and 1.74% alkali showed considerable expansion and cracked at 20 and 24 weeks respectively (Figure 11 ). The different expansion behaviours of mortar bars and concrete prisms are not unusual and have been observed by other research-
264
ers [16, 17]. The lack of significant expansions in concrete prisms in the accelerated test may be due to the denser and less pervious nature of the cement paste in the concrete relative to that in mortar bar (i,e, lesser access of the NaOH solution to the aggregates in concrete), and this is under investigation. The laboratory tests clearly show the potential reactivity of the dacite aggregate in high-alkali concrete. The lack of reaction in the lower parts and the extensive reaction in the upper 12m of the dam indicates that either the original cements used were of different alkali contents, or probably alkali was lost due to leaching from the lower parts, but became concentred due to evaporation in the upper part and caused AAR in this restricted zone. Such differential reactivities have been observed in other dams [11] which allegedly arose due to the use of cements with different alkali contents, although concen-
I
~
I
3OOO
I
2O0O
1500
1000 800 (~0 /~0
Figure 9 Infrared absorption spectrum of the white reaction product. Q = quartz, C -- calcite and W = water
Alkali aggregate reaction in a 60-year-old dam in Australia
0"6
Shayan
tration of alkali in certain parts of a structure could occur due to moisture migration [18]. However, concretes from the various parts of the dam need to be fully investigated for a better understanding of the reaction and its absence in different parts of the dam.
-
~5
0.1+ o~ ACKNOWLEDGEMENT
z0.3
The author thanks the authority responsible for the dam for access to the drilled cores. He also thanks I. Ivanusec and R. Diggins of this Division for making and measuring the length changes of the laboratory concrete and mortar specimens.
x 0"2
0"1 CONERETE REFERENCES 0
10
20
30
50
40
60
TIME(DAY) Figure 10 Expansion of mortar bars and concrete prisms subjected to the accelerated test
0.1/,
0"12
0.10
(1.38%) z
0-08
g z
~x
o.06
ILl
0"04 ---- M(1.7/,%) M(1-25%)
0"02
I
10
20 30 TIME(WEEK)
/,0
C(0-84%) I
50
Figure 11 Expansion of mortar bars (M) and concrete prisms (C) made at different alkali levels and stored at 38 _+ 2°C and 100% RH
1. Cole, W. F., Lancucki, C. J. and Sandy, M. J. 'Products formed in an aged concrete', Cement and Concrete Research, Vol. 11, No. 3, May 1981, pp. 443-54. 2. Shayan, A. 'Re-examination of AAR in an old concrete', Cement and Concrete Research (in press). 3. Shayan, A. and Lancucki, C. J. 'Alkali-aggregate reaction in the Causeway Bridge, Perth, Western Australia', in 'Concrete alkali-aggregate reaction', edited by P. E. Grattan-Bellew, Noyes Publications, New Jersey, USA, 1987, pp. 392-7. 4. Shayan, A., Diggins, R., Ritchie, D. F. and Westgate, P. 'Evaluation of Western Australian aggregates for alkali-reactivity in concrete', in 'Concrete alkali aggregate reaction', edited by P. E. Grattan-Bellew, Noyes Publications, New Jersey, USA, 1987, pp. 247-52. 5. Standards Association of Australia, Australian Standard AS 1141 - Section 38, 'Potential alkali reactivity by mortar bar', 1974. 6. Shayan, A., Diggins, R., Ivanusec, I. and Westgate, P. 'Accelerated testing of some Australian and overseas aggregates for alkali-aggregate reactivity', Cement and Concrete Research (in press). 7. Donaldson, C. H. and Henderson, C. M. B. 'A new intepretation of round embayments in quartz crystals', Mineralogical Magazine, Vol. 52, 1988, pp. 27-33. 8. Dolar-Mantuani, L. 'Handbook of concrete aggregates, a petrographic and technological evaluation', Noyes Publications, New Jersey, USA, 1983, pp. 79-112. 9. Mielenz, R. C. 'Petrographic examination of concrete aggregate to determine potential alkali reactivity', National Research Council, Washington Transportation Research Board, Research Report No. 18-C, 1958, pp. 29-35. 10. Mather, K. 'Condition of concrete in Martin Dam after 50 years of service', Cement, Concrete and Aggregates, Vol. 3, 1981, pp. 53-62.
265
Alkali aggregate reaction in a 60-year-old dam in Australia
11.
12.
13.
14.
15.
Stark, D. 'Alkali-silica reactivity in five dams in South Western United States', United States Department of the Interior, Bureau of Reclamation, Report REC-ERC-85-10, 1985, 64pp. Idorn, G. M. 'Durability of concrete structures in Denmark', Technical University of Denmark, Copenhagen, 1967, pp. 84-136. Davies, G. and Oberholster, R. E. 'The alkali-silica reaction product, a mineralogical and an electron microscopic study', Proceedings of the 8th International Conference on Cement Microscopy, Orlando, Florida, USA, 1986, pp. 303-26. Cole, W. F. and Lancucki, C. J. 'Products formed in an aged concrete, the occurrence of okenite', Cement and Concrete Research, Vol. 13, No. 5, September 1983, pp. 611-8. Oberholster, R. E. 'Alkali reactivity of siliceous aggregates: diagnosis of the reaction, testing of
266
.Shavar,
16.
17.
18.
cement and aggregate and prescription of preventive measures', Proceedings of 6th International Conference on Alkalis in Concrete, Copenhagen, Denmark, 1983, pp. 419-33. Grattan-Bellew, P. E. and Litvan, G. G. 'Testing Canadian aggregates for alkali expansivity', Proceedings of Symposium on the Effect of Alkali on the Properties of Concrete, London, UK, 1976, pp. 227-45. Oberholster, R. E. 'Alkali-aggregate reaction in South Africa - A review', Proceedings of 5th International Conference on Alkali-Aggregate Research in Concrete, Cape Town, South Africa, 1981, Paper $252/8. Nixon, P. J., Collins, R. J. and Rayment, P. L. 'The concentration of alkalies by moisture migration in concrete - a factor influencing alkali aggregate reaction', Cement and Concrete Research, Vol. 9, No. 4, July 1979, pp. 417-23.