Laser scanning confocal microscopy for in situ monitoring of alkali–silica reaction

Authors


Professor K. E. Kurtis. Fax: +1 404 894 2278; e-mail: kkurtis@ce.gatech.edu

Summary

Alkali–silica reaction (ASR) occurs in concrete between reactive siliceous components in the aggregate and the strongly alkaline pore solution, resulting in the formation of a potentially expansive gel product. Lithium additives have been shown to reduce expansion associated with ASR, but the mechanism(s) by which lithium reduces expansion have not been understood. Therefore, development of an in situ method to observe reactions associated with ASR is highly desirable, as it will allow for non-destructive observation of the reaction product formation and damage evolution over time, as the reaction progresses. A technique to image into mortar through glass aggregate by laser scanning confocal microscopy (LSCM), producing three-dimensional representations of the sample was developed. This LSCM technique was utilized to monitor the progress of alkali–silica reaction in mortar samples prepared with alkali-reactive glass aggregate both in the presence and in the absence of lithium additives: LiNO3, LiCl or LiOH. The method proved to be effective in qualitatively monitoring crack formation and growth and product formation, within cracks and at the paste/aggregate interface. In particular, dendritic products were observed at the paste/aggregate interface only in those samples containing lithium, suggesting that these products may play a role in ASR mitigation.

Introduction

Expansion in concrete caused by swelling of alkali–silica reaction (ASR) gel is a concern in areas of the world where the materials available for concrete construction include aggregate containing potentially reactive silica and high-alkali cementitious and pozzolanic materials. ASR occurs when alkalis (from mixture components or an external source) present in concrete pore solution indirectly increase the hydroxyl ion content, catalysing the dissolution of silica (Powers & Steinour, 1955; Iler, 1979). Amorphous or poorly crystalline silica or silicates present in some aggregates are more susceptible to ASR than more ordered structures. After the pore solution becomes saturated with dissolved silica species, silica will precipitate, at the same time binding alkalis, forming alkali–silica gel (Hobbs, 1988). The alkali–silica gel can then take in water causing it to swell (Hobbs, 1988) and exert an internal pressure on the concrete. If the internal pressure exceeds the tensile strength of concrete, the concrete may crack (Swamy, 1994). Over time, ASR damage causes a loss in concrete mechanical performance, durability and service life.

Common methods for avoiding ASR in new construction include using non-reactive aggregate, low-alkali cement, lower water-to-cement ratio, reducing the cement content and/or incorporating supplementary cementitious materials (Stark, 1992; Ramachandran, 1998). In some locations, however, appropriate aggregate, cement or supplementary cementitious materials may not be available. In addition, tests for potential reactivity of aggregate may be unreliable. The use of chemical additives is an alternative means to avoid damage by ASR in new construction. McCoy & Caldwell (1951) first demonstrated that some chemical salts, including lithium salts, may decrease expansion in mortars undergoing ASR. However, the mechanism by which lithium acts is still unknown. To date, research on ASR gel expansion and the effects of lithium additives on ASR has not provided a comprehensive understanding of the reaction mechanism(s). A more complete understanding of the mechanism(s) of control will allow the determination of additives and dosages effective in controlling ASR-induced damage, and predict their duration of control. Then, lithium additives may be used as a practical application for managing ASR damage in concrete.

In situ, time-resolved observation of alkali–silica reaction in the absence and presence of lithium, coupled with more traditional measurements of expansion in companion samples, should provide valuable insights into reaction mechanisms and the role of lithium in suppressing expansion. The focus of the work described herein includes the development of an appropriate characterization method to allow visual manifestations of ASR damage to be monitored in situ over time.

High-resolution microscopic methods such as scanning electron microscopy (SEM), tunnelling electron microscopy (TEM), transmission soft X-ray microscopy and laser scanning confocal microscopy (LSCM) have been utilized to study the surfaces of cement-based materials. More specifically, alkali–silica reaction products have been observed by SEM and transmission soft X-ray microscopy (Kurtis et al., 1998, 1999; Aquino et al., 2001; Marfil & Maiza, 2001). However, the use of SEM and TEM for monitoring in situ microstructural changes over time as a result of ASR is limited by the requirement that the sample be dried or exposed to a vacuum during examination. Exposing cement-based materials to drying or vacuum may alter the composition and morphology of hydration and other water-containing reaction products in the sample, and such exposure will also halt or alter many ongoing reactions. Transmission soft X-ray microscopy allows hydrated samples to be observed over time, but the sample thickness must be less than ∼40 µm, making this technique poorly suited for studying larger samples, including mortars (Kurtis et al., 1999). LSCM also allows imaging of hydrated samples, but with examination of relatively larger samples than transmission soft X-ray microscopy.

Based upon the confocal principle in which out-of-focus data are rejected at each focal plane, and a three-dimensional (3D) representation may be assembled from the in-focus data, LSCM produces a higher resolution image than ordinary light microscopy (Sheppard & Shotton, 1997). In addition, in-focus images of rough or cracked surfaces may be obtained by LSCM. Lange et al. (1993) first utilized LSCM on cement-based materials to characterize surface roughness. Other significant benefits to the LSCM are high magnification, light intensity map and volumetric quantification capabilities. Intensity maps can be made across the paste–aggregate interface at successive depths through the aggregate with the expectation that different materials within the mortar will yield different relative intensities. Also, unlike other microscopy methods that only give surface information, the volume of optically transparent materials within the mortar can be determined by the generation of 3D images from the LSCM data.

A new technique has been developed to allow imaging through glass aggregate in mortar samples by LSCM. The research presented here demonstrates the applicability of through-aggregate imaging by LSCM to monitor and quantify ASR gel formation and damage in situ. Polished mortar samples containing glass aggregate and various lithium additives at various dosages were observed over time by LSCM.

Experimental method

Through-aggregate imaging of polished mortar samples containing glass aggregate and various amounts of lithium additives was performed with an LSCM. Microscopy samples were prepared to monitor and quantify visual manifestations of ASR-induced damage.

Sample preparation

Mortar samples were prepared in accordance with ASTM C 227 except that the specified graded aggregate was replaced with borosilicate glass beads (similar to ASTM C 441), and sodium hydroxide was added to the mix water to increase alkali content (based on ASTM C 1293). The mortars were prepared with Type I cement, a water-to-cement ratio (w/c) of 0.37, borosilicate glass beads and various quantities of lithium additives. Results from an oxide analysis of the cement and accompanying Bogue potential compositions are given in Table 1. NaOH was added to the mix water to increase the Na2Oequivalent to 1.0 wt%. Graded round borosilicate glass beads (obtained through Fisher Scientific), with diameters of 1, 2 and 3 mm, were used in place of traditional aggregate to allow for through-aggregate imaging. In addition to a control prepared without lithium additive, samples were prepared with a specific lithium additive (LiOH, LiNO3 or LiCl) added by molar concentration to the mix water, so the effect of each respective additive could be compared (Prezzi et al., 1998). The lithium additive dosages tested were (Li2O)/(Na2Oeq) = 0, 0.5 and 1.5 (LiNO3 only). The mortars were cast in 1 × 1 × 11.25-in (25 × 25 × 285-mm) mortar bar prism moulds, which were then cut to approximately 1.25–1.5 in (31.75–38.1 mm) in length prior to polishing.

Table 1.  Cement composition.
 Wt%
Oxide analysis
 SiO219.85
 Al2O3 5.67
 Fe2O3 2.69
 CaO64.81
 MgO 0.78
 Na2O 0.19
 K2O 0.49
 Na2O equivalent 0.51
 TiO2 0.51
 Mn2O3 0.05
 P2O5 0.19
 SrO 0.22
 BaO 0.00
 SO3 3.15
 Loss on ignition 1.44
 Insoluble residue 0.17
Potential cement composition
 C3S62.07
 C3A10.46
 C2S10.10
 C4AF 8.20

Polishing method

After curing for 24 ± 2 h in a humid container at 23.0 ± 2.0 °C, the samples were saw-cut axially using a water-cooled diamond saw. The cut surfaces were polished using sandpaper grit sizes 120, 240, 320 and 600 and water as the lubricant, on a Buehler ECOMET 4 metallography grinder/polisher. It should be noted that the sample was not impregnated with epoxy prior to polishing, as is traditionally done, in order to allow the alkali–silica reaction to continue. Owing to imperfections in the glass aggregate surface generated during the initial polishing step, a method to use CeO2 powder (2-µm particles, obtained from His Glassworks, Inc.) mixed with water to a paste consistency (volume ratio of about 0.75 powder/0.25 water), with a felt bob drill attachment to further polish the glass surface was developed. The cylindrical bob, 1.5 in length by 1.5 in diameter, was made of compact felt and was obtained from Duro-Felt Products, Inc. The flat end of the bob was used for polishing. After about 5 min of polishing, the sample was rinsed thoroughly, patted dry and the surface of the glass was inspected for imperfections under the microscope. If necessary, the process of polishing with cerium oxide was repeated until scratches in the glass surface were largely eliminated. This allowed light from the microscope to penetrate the surface of the glass beads, and not refract due to imperfections. After this final glass polishing, images using the LSCM could be obtained through the entire depth of the aggregate to the paste–aggregate interface below the aggregate. In order to prevent interference with the time-dependent reaction products, no further polishing was done after the initial polishing treatment. Polished mortar samples were mounted, with heat- and moisture-resistant epoxy (Sikadur 32 High-Mod) to aluminium templates drilled with three holes. The three holes were spaced so that the template could fit on a custom-made sample stage with corresponding raised pegs in a unique position. This allowed positions to be revisited over time using an x–y indexible sample stage and coordinate reader from Boeckeler Instruments. Mortars were stored in metal containers, at high relative humidity at 38.0 ± 2.0 °C and were monitored over time (as per ASTM C 227).

LSCM method

LSCM images were obtained with a Leica TCS NT microscope powered by a single Ar laser (λ = 488 nm) in reflected light mode. A schematic representation of the LSCM is shown in Fig. 1. First, the eyepiece was focused on the surface of the sample. Then the microscope was turned to scanning mode and the photomultiplier tube (PMT1) detector gain value, which controls the detector voltage, was adjusted such that there was no, or very little, image saturation. Image saturation was determined using the ‘glow over’ colour look up table (LUT), in which over-saturated pixels appear bright blue. The PMT1 value was only determined in this manner for the images obtained at 1 day at each x–y location. Consecutive images were obtained over time using the same PMT1 value that was used initially at each location. Images were taken with 2.5×/0.07 N PLAN, 5×/0.15 HC PL FLUOTAR and 20×/0.40 H PLAN objective lenses. The imaging locations were mapped in the x and y planes using a cross hair, indexible stage and coordinate reader. Images of each sample at each distinct x–y location were obtained at approximately 1, 7, 14, 28 and 56 days. Images at each x–y location were taken through successive z-depths of the glass aggregate. Three-dimensional (3D) images and relative reflected light intensity maps were generated using Leica Confocal Software Version 2.0.

Figure 1.

Schematic of the LSCM optics.

Results and discussion

LSCM proved to be a useful method for monitoring reactions in mortars prepared with and without lithium additive. LSCM images were taken through successive z-depths of the glass aggregate in an initially undamaged control mortar containing no lithium additive, using a 2.5×/0.07 N PLAN objective, and were compiled using Leica Confocal Software Version 2.0 to obtain the 3D rotational image in Fig. 2(a–e). Figure 2(a) shows the surface of the sample, the x–y plane, Fig. 2(b–d) show the gradual rotation about the y-axis, and Fig. 2(e) shows the y–z plane through the glass aggregate. Each image was generated by the LSCM detector collecting laser light reflected by the optically opaque surface in the focal plane. In the mortar samples, the light was reflected by the paste–aggregate interface when imaging through aggregates. Analysis of 2D LSCM images, at various depths, showed that the reflected light intensity changed with the material being imaged. (LSCM images shown with corresponding intensity maps, including those in Figs 3, 4, 7, 11 and 13, have not been ‘enhanced’ to improve brightness or contrast; these adjustments were not made such that the images presented match the intensity maps.) Figure 3(a,b) show images taken at the sample surface, z = 0 mm, and into the aggregate at z = −0.21 mm, together with their corresponding intensity maps. The intensity map corresponding to Fig. 3(a) shows that light was reflected back at a higher relative intensity at the surface of the glass than at the paste surface or when imaging through the surface of the glass to the paste/aggregate interface, as expected. Imaging through the aggregate to the paste/aggregate interface is believed to be a novel application of LSCM.

Figure 2.

3D LSCM rotational image of glass bead aggregate in microscopy mortar sample: (a) x–y plane, (b) rotation, (c) rotation, (d) rotation, (e) y–z plane.

Figure 3.

LSCM images and corresponding image maps of the reference sample at depths of (a) 0 and (b) −0.21 mm.

Figure 4.

LSCM image and intensity map showing gel rings and debonding in the reference sample at 49 days.

Figure 7.

LSCM images of cracking in the reference sample shown at different z depths over time: (a) z = 0 mm at 2 days, (b) z = −0.49 mm at 2 days, (c) z = 0 mm at 26 days, (d) z = −0.49 mm at 26 days.

Figure 11.

LSCM images of the sample prepared with LiCl, [Li2O]/[Na2Oeq] = 0.5, at (a) z = 0 mm at 14 days, (b) z = −0.15 mm at 14 days, (c) z = 0 mm at 21 days, (d) z = −0.16 mm at 21 days.

Figure 13.

LSCM images of the LiOH sample, [Li2O]/[Na2Oeq] = 0.5, at (a) z = 0 mm at 1 day, (b) z = −0.24 at 1 day, (c) z = 0 mm at 7 days, (d) z = −0.23 mm at 7 days.

The control sample was prepared without the addition of a lithium additive to determine the base case of damage resulting from ASR gel expansion so the effects of lithium additives could be assessed. Damage commonly associated with ASR, such as gel rings around the aggregate, aggregate debonding, and map cracking through the paste and aggregate, was observed only in the control mortar sample. Figure 4 shows an LSCM image of gel rings and aggregate debonding, obtained with a 20×/0.40 H PLAN objective, and corresponding intensity map at 49 days. The gel ring has a greater and more varied relative intensity than the aggregate in this image. The locations of separation of the aggregate from the paste are indicated as voids on the intensity map, and result in a relative intensity of 0. When the PMT1 gain value was held constant throughout testing dates and x–y position, images taken throughout the depth of an aggregate that debonded over time showed a large decrease in the intensity of light reflected from the bottom of the aggregate at the paste/aggregate interface as debonding occurred. Often, in debonded aggregates, the paste could not be observed through the aggregate at all.

Over time, cracks grew in number and in width as a result of continuing ASR damage in the control sample. Figure 5 shows LSCM images, using a 20×/0.40 H PLAN objective, of cracking in the aggregate and paste progressing over time. Figure 5(a) shows the aggregate, undamaged, at 2 days. Figure 5(b–d) show the gradual cracking and crack widening in the aggregate and the paste. Crack widening in the paste alone was also observed by LSCM. Figure 6 shows a crack in the paste, using a 20×/0.40 H PLAN objective that progressively widens over 67 days. Images, obtained with a 5×/0.15 HC PL FLUOTAR objective, and corresponding intensity maps of cracking in an aggregate, at different depths, are shown in Fig. 7. Figure 7(a,b), taken at 2 days, show the aggregate before there is any sign of cracking, as is seen in Fig. 7(c,d), taken at 26 days. Cracking at the surface of the aggregate is evident in Fig. 7(c). Figure 7(d) shows reaction product in the crack, which extends through the aggregate to a depth of at least −0.49 mm.

Figure 5.

LSCM images of crack formation and progression in the reference sample at (a) 2, (b) 7, (c) 14 and (d) 26 days.

Figure 6.

LSCM images of crack widening in the reference sample paste at (a) 7, (b) 14, (c) 26 and (d) 67 days.

With ordinary optical microscopy, this type of image is of poorer quality because the reaction products cannot be observed distinctly from the paste and aggregate. For comparison, Fig. 8 shows a stereo optical microscope image, magnified ×80, of reaction product within an aggregate crack. In addition, the location and possibly the quantity of reaction product within the aggregate may be determined by LSCM, but not by stereomicroscopy. Cracking within the aggregate and paste is also better observed by LSCM because of its high-resolution capabilities. For example, Figs 5 and 6 show distinguishable crack widths of about 0.01–0.02 mm, whereas the resolution of the aggregate crack in Fig. 8 renders the exact width unquantifiable.

Figure 8.

Stereoimage of cracking in the reference sample.

Reactions in mortar samples prepared with lithium additives were also monitored over time by LSCM. Over the period of examination, samples prepared with lithium showed no cracking, but a change in the paste/aggregate interface was observed, as is seen in Fig. 9, obtained with a 20×/0.40 H PLAN objective at 3, 8 and 14 days. It is believed that the change observed at the paste/aggregate interface demonstrates the formation of a chemical reaction product, rather than being an indication of aggregate debonding. Debonding of the aggregate from the paste in the control sample resulted in a mean LSCM relative intensity of less than 1, and here the mean intensity varies between approximately 12 and 50. The products at the interface appear to have ordered structures and possess lath-like or dendritic morphologies. Based on these observations, it is proposed that they may be crystalline in nature. Each individual dendrite appears to be less than 80 µm in length. These products were first noticed at 1 day of age after demoulding and polishing and were present in all samples prepared with Li, independent of salt type. That these products were observed only in samples containing Li suggests that they may play a role in ASR mitigation.

Figure 9.

LSCM images of the paste/aggregate interface in samples prepared with LiNO3, [Li2O]/[Na2Oeq] = 0.5 at (a) 3, (b) 8 and (c) 14 days.

However, the quantity of the product and how it changed with time varied with location and lithium additive type. LSCM images of the paste/aggregate interface of samples prepared with LiNO3 at [Li2O]/[Na2Oeq] = 0.5, taken over 14 days, show the reaction product growing in size. A size increase in the reaction products at the paste/aggregate interface was also observed in the sample prepared with LiCl at [Li2O]/[Na2Oeq] = 0.5. Figure 10 shows the paste/aggregate interface of the LiCl sample at 14 days, obtained by LSCM outfitted with a 20×/0.40 H PLAN objective. The reaction product at the interface in the LiCl sample (Fig. 10) does not appear to have the same morphology as the reaction product in the LiNO3 sample (Fig. 9c) at the same age and [Li2O]/[Na2Oeq]. Figure 11 presents LSCM images, obtained with a 20×/0.40 H PLAN objective, and intensity maps to show how the LiCl sample changed with depth and time. The relative intensity of the paste at the mortar surface at 14 days (Fig. 11a) is not consistent with the relative intensity of the paste at the paste/aggregate interface (Fig. 11b). This difference in relative intensity could be due to the formation of reaction products at the interface. After 21 days, the surface of the glass aggregate (Fig. 11c) showed a decrease in relative intensity at its edges possibly due to reaction of the glass surface. In addition, at 24 days, the paste/aggregate interface (Fig. 11d) showed a large decrease in relative intensity, which could be due to debonding of the aggregate. Aggregate debonding can have a significant impact on the mortar strength because strength is typically limited by the properties of the interfacial zone that exists between aggregate and the hydrated cement paste.

Figure 10.

LSCM image of the past/aggregate interface in the LiCl, [Li2O]/[Na2Oeq] = 0.5, sample at 14 days.

In samples prepared with LiOH, a reduction in the amount of reaction products at the interface occurred over time. However, the relative amount of change in the LiOH samples was not as great as was observed in the LiCl and LiNO3 samples. Figure 12, obtained with a 20×/0.40 H PLAN objective, shows the LiOH samples prepared at [Li2O]/[Na2Oeq] = 0.5 at 1 and 7 days of age. Figure 13, obtained with a 5×/0.15 HC PL FLUOTAR objective, shows the LiOH sample at different depths changing over 1 week. The surface of the aggregate and the aggregate/paste interface appear to have changed similarly to the LiCl mortar sample.

Figure 12.

LSCM images of the paste/aggregate interface in the sample prepared with LiOH, [Li2O]/[Na2Oeq]= 0.5 at (a) 1 and (b) 7 days.

Conclusion

The use of ‘through-aggregate’ imaging by LSCM, a technique for imaging reactions in concrete through glass aggregate, was developed and was shown to be effective for examining alkali–silica reaction in situ. Three-dimensional representations of the aggregate, images of reaction product both within cracks and at the paste/aggregate interface, and quantitative measurement of gel ring thickness at the surface are all examples of types of information gained by LSCM that have not, at this point in time, been possible with other microscopy methods.

LSCM and stereomicroscopy images showed definitive evidence that ASR occurred in the sample prepared without lithium additive; characteristic crack patterns, gel rings, common reaction products and debonding were observed. Samples prepared with lithium showed no evidence of ASR. However, through-aggregate imaging showed reaction product formation occurring at the paste/aggregate interface in which apparently crystalline reaction products were formed. Further characterization of these reaction products is ongoing.

Acknowledgements

We are grateful for insight provided by Nikhila Naik and Dan Dyer. This research was supported by NSF POWRE Award CMS-0074874.

Ancillary