7th CANMET/ACI International Conference on Superplasticizers and Other Chemical Admixtures, Berlin, October 2003

Test Methods for Evaluating the Efficacy of Chemical Admixtures for Controlling Expansion Due to ASR 7th CANMET/ACI International Conference on Superplasticizers and Other Chemical Admixtures, Berlin, October 2003 By David Stokes, Stephen Baxter, Michael Thomas and Russell Hill Abstract The use of lithium-based chemical admixtures for controlling expansion due to alkalisilica reaction (ASR) is now well proven. The usual lithium dose recommendations given suggest a simple proportionality based on the alkali loading of the system under consideration. However, the data available indicate that the level of lithium required depends upon a number of parameters including the type of lithium compound, the nature of the reactive aggregate and certain characteristics of the cementing material (e.g. alkalinity, and the type and level of any supplementary cementing material present). Consequently, prescriptive specifications covering the use of lithium compounds are, by necessity, complicated, and this highlights the need to develop specifications based on performance testing. In this work, four ASTM test methods, C1293, C441, C227, and C1260, were modified to include lithium-based admixture, alone and in combination with fly ash. In the case of C1260, lithium admixture was also incorporated into the caustic soak solution. The results from these tests using a wide range of aggregates with and without fly ash are compared. Of these four method modifications, the ones with C1293 and C1260 show the greatest promise to be developed into performance tests, although the long test period necessary with C1293 will likely hinder its use in specification testing. With C441 no interaction with actual aggregate is possible, and C227 often fails to yield significant expansion with many aggregates known to be reactive in the field. Some of the results from C227 strongly indicate a pronounced effect of alkalinity of the system on the necessary lithium admixture dose, with higher alkalinity requiring proportionally higher lithium doses. Thus, a potential weakness in both test regimes is the inability to evaluate lower alkali situations. Keywords: ASR, Li admixture, ASTM C1293, ASTM C1260, ASTM C441, ASTM C227

ACI member David Stokes is Manager of Concrete Technology at FMC Corporation s Lithium Division in Gastonia, North Carolina. Previously he was Chief Chemist for the Department of Transportation in Delaware. He has given numerous presentations on ASR and Lithium in North America and abroad. Stephen Baxter is the principal at Stephen Baxter Research, Inc., and has been working in ASR research for most of the past decade. ACI member Dr. Michael Thomas is a Professor of Civil Engineering at the University of New Brunswick. He is a prolific author, editor, and researcher, and has received many awards and honours from ACI and other institutions. ACI member Russell Hill is Chairman of Committee 201 Durability of Concrete and is VP of Technology at Boral Materials Technologies. INTRODUCTION Lithium based admixtures are being increasingly used to control deleterious expansions in concrete caused by alkali silica reaction (ASR). Nonetheless, there is no standard ASTM classification for these admixtures, and no standard ASTM test to measure their effectiveness. There are increasing numbers of national and local specifications used and being developed to allow their use, but they generally vary from being either totally prescriptive, or varying in the specification of performance testing. That is, they either suggest the levels at which the admixture will be dosed (generally based on the alkali supplied by the cement in the concrete mixture) and/or other tests are detailed which are generally modifications of otherwise standard ASTM or AASHTO methods, or the like. A survey of the literature shows general agreement that so-called standard dosages of the admixture go a long way towards suppressing expansions from ASR (1-6). While in most cases all the test methods are in agreement that the standard dose will yield control, at least in the test methods employed, there have been reputable findings that somewhat higher dosages would be necessary with some aggregates. And not all methods result in precisely the same result regarding the amount necessary for control (6). And in the more complicated case of combining the lithium admixture with supplementary cementitious materials (SCMs), there is perhaps even greater divergence among the different methods employed. (Of course, this is also the situation for SCMs and these test methods in general.) Thus the current state of affairs requires of the prudent specifier to use conservatively high amounts of lithium admixture for prevention of deleterious expansions from ASR (6). It would therefore be useful, and more economical, to have standard tests that would allow more precise dosages to be ascertained. This would also enable a more widespread use of lithium admixtures than is currently the case.

As a step towards the development of a useful standard test, and in an attempt to gain better insight into the response of different mortar and concrete systems to varying lithium admixture dosages, four ASTM tests were modified to include lithium admixtures C1293, C1260, C227 and C441. In this work we will discuss the modifications of these tests that were employed and the results obtained. METHODS AND MATERIALS Two portland cements having a high and low alkali content (1.1 and % Na 2 Oe, respectively), and meeting the ASTM C 150 specification for Type I cements were used. These two cements were blended in various proportions to produce four cements with different alkali contents as follows: VA Very high alkali 1.1% Na 2 Oe HA High alkali 0.9% Na 2 Oe MA Moderate alkali 0.7% Na 2 Oe LA Low alkali % Na 2 Oe Four fly ashes were selected to represent the range of calcium contents currently encountered with commercially available sources in North America; chemical analyses of the fly ashes are presented in Table 1. The calcium contents ranged from 1.2 to 30.0% CaO. The efficacy of fly ash in controlling damaging expansion due to ASR is strongly influenced by the calcium content of the fly ash (7). Lithium was added where required in the form of a solution containing 30% LiNO 3 ; this product is commercially available as chemical admixture for concrete in North America. The reactive aggregates used were Pyrex glass for the C441 mixes, with the majority of the rest utilizing four sources of natural aggregate selected to represent a range of aggregate reactivity. All are known to be reactive in the field. The natural aggregates were as follows: VR Very high reactivity Rhyolitic gravel from Albuquerque, New Mexico. Reactive phase mixed volcanics. HR High reactivity Crushed siliceous limestone from the Spratt Quarry in Ontario, Canada. Reactive phase opaline silica MR1 Moderate reactivity Crushed greywacke aggregate from Pennsylvania. Reactive phase microcrystalline and strained quartz. MR2 Moderate reactivity Crushed gravel aggregate from Texas. Reactive phase chert.

Mortars and concretes were prepared with different reactive aggregates, cement alkali levels, different types and levels of fly ash, and various doses of lithium nitrate solution. Lithium nitrate solution was added at doses ranging from of 0 to 120% of the recommended dose of 4.6 litres of LiNO 3 for each 1 kg of alkali equivalent (Na 2 Oe) in the mortar or concrete. In all cases, only the alkali from the Portland cement component of the mix was considered in the calculation of the lithium dose. The 100% dose represents a lithium-to-alkali molar ratio of [Li]/[Na+K] = 0.74, which has generally been found to be sufficient to suppress deleterious ASR with most reactive aggregates (1-6). The procedures followed were generally those outlined in ASTM standard test methods C227, C441 and C1260 for mortar bar tests and C1293 for concrete prism tests with the following deviations. When mortar bars containing lithium were tested in ASTM C1260, lithium nitrate was added to the standard soak solution (1 M NaOH), usually to achieve the same lithium to alkali molar ratio, [Li]/[Na+K], as that used in the production of the mortar (in the majority of the mixes studied). For example, mortars that contained a 100% LiNO 3 dose had a molar ratio of [Li]/[Na+K] = 0.74 and soak solution was modified to contain 1 M NaOH and 0.74 M LiNO 3. In the majority of the mixes, the % dose in the mortar bars and the soak solutions matched. However, a subset of mixes had lower doses in the soak solution compared to the mortar bar dose to study this variable as well. The standard C1260 procedure is not suitable for testing lithium admixtures as the influx of NaOH during storage in the 1 M NaOH solution results in a reduction of the lithium-alkali ratio (8). Using a modified soak solution helps to maintain the lithiumalkali ratio throughout the test period. Concrete prism tests were conducted using either a cementitious material content of 420 kg/m 3 with the alkali content of the portland cement component raised to 1.25% Na 2 Oe by adding NaOH to the mix water as per the standard C1293 mix (mixes designated VA+) or by using the same cementitious material content with no additional NaOH using either the high-alkali (HA = 0.90% Na 2 Oe) or moderatealkali cement (MA = 0.70% Na 2 Oe). Concrete prism tests were extended to two years. Mixes with the MR2 aggregate were not used in the C1293 work, and the C ashes were not evaluated in the C1260 work. Altogether 100 C1293s, 200 C227s, 83 C441s, and 266 C1260 mixes were cast in this study. Only certain subsets highlighting the most significant findings are discussed herein. RESULTS AND DISCUSSION The modified C1293 mixes performed similarly with these materials to other published studies, and the inclusion here is mainly as a benchmark to compare to the other test procedures. While not a perfect predictor of field performance (no single test is), this test is generally considered the best estimator of field performance by many (6), at least for supplementary cementitious materials. For example, Figure 1 shows results for mixes with the very highly reactive (VR) aggregate with all four fly ashes and varying lithium nitrate dosages. As the lithium dose increases, the expansion decreases, and the lower the CaO content of the ashes, the lower the expansion. The exception for these mixes is with the very high lime ash C2, dosed at 30%, which increased in expansion before starting to drop again. Also, the alkali content for this particular set of mixes was 0.9%, and was not

raised to 1.25% with NaOH. Figure 2 shows the 2- year expansion at 100% dose and control for 3 of the aggregates described above. The VR and HR mixes are shown at 0.9% alkali (HA), and the MR1 aggregate is shown at the standard 1.25% level (VA). It is noteworthy that the overall reactivity level assigned to the aggregates (based on their 1293 expansions) does not correlate well with their response to the lithium nitrate admixture. MR1 shows the least effect with the admixture, followed by HR, while VR and MR2 are easily controlled. The C441 results were also in line with other published work (9), in that the lithium admixture suppressed expansions significantly, alone as well as in combination with the ashes. What is more interesting is the lack of significant differentiation between the 2 F ashes and the 2 C ashes in this test at 20% cement replacement with high alkali (HA) cement (see Figure 3), although there is generally a significant difference in the response with actual aggregates, as shown in the C1293, C227, and C1260 results (F ashes only in the C1260 samples). However, 30% cement replacement levels showed much better differentiation with the ashes. Also important is that the differences were greater at all replacement levels for medium alkali (MA) cement, and in some cases the expansions were actually higher with MA cement. These types of effects cannot possibly be seen with C1293 and C1260 without modifying these methods to allow for different alkali levels. In the case of C1260 this would be more complicated because the soak solution strengths would also have to be modified. In general, behaviour at lower alkali levels is extrapolated from these tests, and it is usually assumed that the high alkali level investigated is the more conservative view. That is, lower alkali levels are always assumed to give lower expansions. As in the C441, C227 tests were made with various alkali levels. Unfortunately with C227, most aggregates do not give a significant expansion. Only VR and MR2 gave expansions above the % threshold at 1 year, and with MR2 this was with VA cement. VR was above the threshold regardless of alkali content, and is known to be reactive in the field with low alkali cement. Figure 4 shows 1-year C227 results with VR and the fly ashes. Note that 20% C1 and 20% and 30% C2 are higher than the controls, and are also generally higher at lower alkali levels. Figure 5 shows 1-year C227 results with VR and Li. Note the great sensitivity of the Li dose with cement alkali level at 50%, and to a lesser extent at the higher dosages. This strongly suggests that the Li dose is not simply proportional to the alkali level in the cement, but is also a function of the alkali level itself. This implies that tests with Li admixture run at high alkali levels will generally indicate higher levels of Li admixture than may actually be necessary for a lower alkali situation. The problem, of course, is to have a test to evaluate lower alkali levels that give large enough expansions in the laboratory to be useful, and in reasonable periods of time. Figure 6 shows 3 dimensional plots of Li dose versus age (in days) versus % expansion for Li modified soak solutions in C1260 with MR1 and VR. Besides the very large expansion values for VR, 2 main things can be noted on the graphs. One is the steeper curve for VR versus MR1 with Li dose, showing a greater responsiveness of VR to Li

compared with MR1. This coincides with results from C1293. The other is how very low the response of VR is at high Li doses compared to MR1. MR1 seems to maintain a residual resistant expansion value, compared to VR, even at high Li doses. And similarly, C1293 results with the standard Li dose show reduction in expansion, but not control. Figure 7 shows 14 and 28-day expansions in modified C1260 tests with VR and Li. The labels on the X-axis show the Li dose in the bar, followed by the Li dose in the soak solution, as follows Bar dose : Soak dose. For example, the 4th data pair shown is 100 : 70. This means that the bar was dosed at 100% dose (molar ratio of [Li]/[Na+K] = 0.74) and the soak solution contained a molar ratio of 70% of this ([Li]/[Na+K] = 0.74 * 0.7 ~ 2). (Actually, in the case of the soak solution, the ratio [Li]/[Na+K] = [Li]/[Na], since there is no K added to the soak solution.) For the 0 : 0 case, this is simply the standard C1260 test. For the 100 : 0 case, this means that Li was added to the bar only, and not to the soak solution. This is equivalent to work performed elsewhere (8) and clearly does not correlate at all with either concrete prism test results or outdoor exposure site results with large blocks and slabs. Again, this demonstrates the need to balance, in some way, the Li : Na ratio in the soak solution as mentioned earlier. The 0 : 100 case also does not correlate with more realistic concrete case, but is much closer than 100 : 0. This shows the greater importance of the Li in the soak solution compared to the dose in the bar. The 100 : 70 case is getting closer to correlating with actual long term test data, but again is giving somewhat higher results, particularly when the test is run for longer periods of time than the standard, than are seen in actual concrete specimens. The 100 : 100 case, while correlating in this particular instance in a pass/fail way with C1293, nonetheless is quite low, and for some of the other aggregates discussed shortly, seems to be overestimating the Li dose effectiveness compared with the concrete prism test. Some mixes looked at very low versus very high alkali contents, shown in Figures 8 through 11. While the effect for the aggregate alone is not pronounced (the 0 : 0 case in each graph) the response for some of the other conditions is quite mixed. For the MR2 aggregate (Figure 8), there are no significant differences, but for VR (Figure 9), there is a strong decrease in expansion when higher alkali cement is used. While this seems counterintuitive, it needs to be remembered that when the alkali content is higher, the amount of Li in the bar at a given dose is higher. And VR reacts very quickly compared to many other aggregates, and also appears to respond strongly to the Li in any test environment. However HR (Figure 10) begins to show a little of the opposite behaviour, and this aggregate reacts more slowly, and is more resistant to ASR control with Li in the concrete prism test. And the effect is most opposite with MR1 (Figure 11) that is also slower to react and is the least responsive to Li by the concrete prism test. While this is a limited data set, it is worth noting, and worth further investigation with a broader data set. Figure 12 shows correlations between 1 year C227 data and 14-day and 28-day 1260 results. Basically there is strong correlation in these across mixes and aggregate types. And for the Li only mixes, the correlation is fairly strong with 2-year 1293 results and

14-day and 28-day 1260 results (Figure 13). However, when all of the 1293 mixes are included (combinations with fly ash and Li) the correlations are very low for HR and MR1, and show that, for this data set at least, the C1260 modification does not correlate enough to be a useful predictor of C1293 across the board based on expansions. The fact that the C1260 values correlate much better with the C227 data than the C1293 indicates that the aggregate size is an important factor in this type of testing, because the storage conditions for C227 are quite similar to C1293, and quite unlike the conditions for C1260. One hypothesis that fits these conditions is that for some reactive aggregates, if the reaction occurs mainly on the outer surfaces as opposed to deep within the aggregates, then it follows that Li would be more effective for the outer surface aggregates, as Li in general has a lower ionic mobility than the other ions of interest in ASR. So for aggregates where the reaction tends to occur more deeply, or simply is more permeable to other ions of importance to ASR than to Li, then the larger the aggregate size, the larger would be the concentration gradients that develop within an aggregate particle, and the greater the extent of reaction would become before reactive silica surfaces could be stabilized by the Li ion. This fits also the C441 data, where Li seems to work very well, as the reaction occurs mainly on the outer surfaces of the glass particles. This would then have potentially significant consequences for any ASR test for concrete materials, as mortars would then tend to underestimate the potential development of the reaction, as the smaller aggregate particles would develop smaller concentration gradients. This also fits with the different levels of correlation between the individual aggregate subsets and the lower correlation for the set as a whole. Since the effect of aggregate size will be different for each aggregate, each aggregate would correlate differently, and thus combinations of different aggregates with large differences in sensitivity to aggregate size would correlate less well than groups with similar sensitivities to aggregate size effects. But there may be more information in the data that could prove useful for correlation than simply comparing expansion results at certain periods of time, as is the usual treatment of these test results. For example, Figure 14 shows expansion curves for 4 pairs of duplicates, with their averages drawn. There is a certain range in expansion values that exists. However, the shape of the curves themselves is quite reproducible, which can be seen by observing these pairs after being normalized to the same expansion values at 28 days (Figure 15). This information may prove useful in correlating the results of the C1260 modified tests in a more robust way with other tests, because it has a greater information content than a single expansion value at some arbitrary time period. An example of this approach has been reported elsewhere (10), and is not explored further in this work. However this type of analysis is ongoing with this data set and will be reported later.

CONCLUSIONS While C441 responds well to Li admixture, it is not possible to evaluate individual aggregate responses and so is unsuitable for evaluating Li admixtures with any specificity for a particular mix. Standard dose of Li admixture in a C1260 bar along with a matching dose in the soak solution may overestimate the effectiveness of the Li admixture when using % expansions as the sole criteria. Testing smaller aggregate sizes may underestimate reactivity. Work should be done at lower alkali contents to investigate possible underestimation of Li effectiveness in C1293. Results with lower alkali cements suggest that [Li]/[Na+K] ratios may be a function of alkali level, and increase with increasing alkali content. Work should be done to try and utilize all the information contained in the expansion data besides using only the expansion value at one point in time. REFERENCES 1. McCoy, W.J. and Caldwell, A.G. (1951) A New Approach to Inhibiting Alkali- Aggregate Expansion, Journal of the American Concrete Institute, Vol. 47, 1951, 693-706. 2. Stark, D., Morgan, B., Okamoto, P. and Diamond, S. Eliminating or Minimizing Alkali-Silica Reactivity. SHRP-C-343, Strategic Highway Research Program, National Research Council, Washington, D.C., 1993, 226p. 3. Blackwell, BQ, Thomas, MDA and Sutherland, A. 1997. Use of lithium to control expansion due to alkali-silica reaction in concrete containing U.K. aggregates. Proceedings of the Fourth CANMET/ACI International Conference on Durability of Concrete, (Ed. V.M. Malhotra), ACI SP-170, Vol. 1, American Concrete Institute, Detroit, pp. 649-663. 4. Lumley, J.S. ASR suppression by lithium compounds. Cement and Concrete Research, Vol. 27 (2), 1997, pp. 235-244. 5. Thomas, M.D.A., Hooper, R. and Stokes, D. Use of lithium-bearing compounds to control expansion in concrete due to alkali-silica reaction. Proceedings of the 11 th International Conference on Alkali-Aggregate Reaction in Concrete, (Ed. M.A. Berube et al.), CRIB, Quebec City, 2000, pp. 783-792. 6. Folliard, K.J., Thomas, M.D.A., and Kurtis, K., Guidelines for the Use of Lithium to Mitigate or Prevent ASR Task A Draft Report, FHWA Project DTFH61-02-C- 00051, 2002.

7. Shehata, M.H. and Thomas, M.D.A. The effect of fly ash composition on the expansion of concrete due to alkali silica reaction. Cement and Concrete Research, Vol. 30, 2000, pp. 1063-1072. 8. Barringer, W.L. Using accelerated test methods to specify admixtures to mitigate alkali-silica reactivity. Cement, Concrete, and Aggregates, Vol. 21 (2), 1999, pp. 165-172. 9. Lane, D.S. Preventive Measures for Alkali-Silica Reactions Used in Virginia, USA. Proceedings of the 11 th International Conference on Alkali-Aggregate Reaction in Concrete, (Ed. M.A. Berube et al.), CRIB, Quebec City, 2000, pp. 693-702. 10. Johnston, D. and Fournier, B. A Kinetic-Based Method for Interpreting Accelerated Mortar Bar Test (ASTM C1260) Data. Proceedings of the 11 th International Conference on Alkali-Aggregate Reaction in Concrete, (Ed. M.A. Berube et al.), CRIB, Quebec City, 2000, pp. 355-364.

% Expansion @ 2 years Table 1 Chemical Analysis of Fly Ashes Oxide F1 F2 C1 C2 SiO 2 50 50.90 34.40 30 Al 2 O 3 30.90 21.90 19.80 18.10 Fe 2 O 3 5.10 5.70 7.00 5.20 Sum 1 86.50 78.50 61.20 53.90 CaO 1.20 12.10 25.50 30.00 MgO 0.80 2.30 4.70 5.30 SO 3 0.00 0.90 2.00 2.60 Na 2 O 6 2.92 2.06 1.92 K 2 O 1.90 8 5 1 Na 2 Oe 1.81 3.37 2.36 2.12 AA 2 1.7 1.8 1.5 LOI 1.90 0 0 0 <45 21.6 23.5 14.7 16.9 1 SiO 2 + Al 2 O 3 + Fe 2 O 3 2 Available alkalis by ASTM C 311 5 5 No ash 15%F1 20%F2 25%C1 30%C2 5 0.05 Limit is 0.04% 0 0 10 20 30 40 50 60 70 80 90 100 % Li Dose Fig. 1 ASTM C1293 results with VR aggregate and HA cement.

% Expansion % Expansion 0 5 0 Limit @ 2 years control 100% Dose 5 0 0.05 0.00 VR (HA) HR (HA) MR1 (VA) Fig 2. Control and standard dose results @ 2 years w/ 1293. MA HA 0.0 0%Li 75%Li 100%Li 20%F1 20%F2 20%C1 20%C2 30%F1 30%F2 30%C1 30%C2 Fig. 3. ASTM C441 results with Li and the fly ashes.

% Expansion % Expansion 0.9 0.8 0.7 Control 20% F1 20% F2 20% C1 20% C2 30% C1 30% C2 0 0.7 0.8 0.9 1 1.1 1.2 1.3 Cement Alkalinity Fig. 4. ASTM C227 results with VR and the fly ashes. 5 5 5 5 No Li 50% Dose 75% Dose 100% Dose 0.05 0 0.7 0.8 0.9 1 1.1 1.2 Cement Alkalinity Fig. 5. ASTM C227 results with VR and Li.

ASTM C1260 Modified Soak Solution (Greywacke) 00 00 00 00 50 00 75 50 25 00 0.090 0.080 0.070 0.060 0.050 0.040 0.030 0.020 0.010 0.000 ASTM C1260 Modified Soak Solution (NM) 1.000 0.900 0.800 0.700 00 00 00 00 50 00 75 50 25 00 0.090 0.080 0.070 0.060 0.050 0.040 0.030 0.020 0.010 0.000 Fig. 6. 3D plots of modified ASTM C1260 results for MR1 (Greywacke, top) and VR (NM, bottom). Note MR1 has residual expansion at high Li dose compared to VR. Also, response to Li dose is steeper with VR than MR1.

% Expansion % Expansion % Expansion % Expansion 1.4 1.2 1.0 0.8 0.0 1.4 1.2 1.0 0.8 0.0 0:0:LA (% Bar Dose : % Soak Dose) 14 day 28 day 0:0 100:0 0:100 100:70 100:100 Fig. 7. Modified C1260 results with VR. Variations shown in the bar dose and the soak dose. 0:0:HA 100/70:LA (% Bar Dose : % Soak Dose : Cement Alkali) 100/70:HA 100:100:LA Series1 Series2 100:100:HA 50:50:LA 50:50:HA 50:35:LA Fig. 8. As in Fig. 7, but showing variations with cement alkali as well. 50:35:HA 0.7 0.0 0:0:LA 0:0:HA 100/70:LA (% Bar Dose : % Soak Dose : Cement Alkali) 100/70:HA 100:100:LA Series1 Series2 100:100:HA Fig. 9. As in Fig. 8, but with aggregate MR2. 0.8 0.7 0.0 0:0:LA 0:0:HA 50:50:LA 50:50:HA (% Bar Dose : % Soak Dose : Cement Alkali) 100/70:LA 100/70:HA Series1 Series2 100:100:LA Fig. 10. As in Fig. 9, but with aggregate HR. 100:100:HA 50:35:LA 50:35:LA 50:35:HA 50:35:HA

Correlation Coefficient Correlation Coeeficient % Expansion 0.7 0.0 1.0 0.9 0.8 0.7 0.0 0:0:LA 0:0:HA 100/70:LA (% Bar Dose : % Soak Dose : Cement Alkali) 100/70:HA 100:100:LA Series1 Series2 100:100:HA 50:50:LA 50:50:HA 50:35:LA 50:35:HA Fig. 11. As in Fig. 10, but with aggregate MR1. Note the opposite response to cement alkali as compared to VR in Figure 8. PL(VA) PL(HA) PL(MA) SP(VA) SP(HA) GR(VA) 14 day 28 day Fig. 12. Correlations between Li mixes with 1 year C227 data and 14 and 28-day C1260 data. 1 0.9 0.8 0.7 0 VR (Li) HR(Li) MR1(Li) VR (all) HR(all) MR1(all) All 14 28 Fig. 13. Correlations between Li mixes with 2 year C1293 data and 14 and 28-day C1260 data. Cement alkali subsets shown are VR-HA, HR-HA, MR1-VA. Sets marked Li are Li only mixes, sets marked all are combinations of Li and fly ashes.

% Expansion (normalized) % Expansion 0.8 0.7 0 0 5 10 15 20 25 30 Age (days) Fig. 14. Four pairs of duplicates showing variability of individual tests. The only piece of information normally used in a test such as this would be a single point in time. (Bold lines indicate averages of the pairs each pair uses the same symbol sets to identify them as pairs.) 1 0.9 0.8 0.7 0 0 5 10 15 20 25 30 Age (days) Fig. 15. The same four pairs as in Figure 14, with the magnitudes of the expansions at 28 days normalized to 1. Here it is clear that the shapes of the curves are quite reproducible. And since the curves are made up of sets of points, there is more information available by using the entire curve instead of simply one data point in time, as is normally the case with these types of tests. Use of this information may allow these rapid mortar bar tests to overcome limitations caused by sensitivity of the reaction to aggregate size effects.