Reactivity of FA and PGP
To evaluate the reactivity of coarse and fine PGP and to provide a comparison against low-calcium FA, their strength activity indices (SAIs) were determined as illustrated in Fig. 4, together with respective strength gains in 7 days and 28 days. As expected, cement substitution with 20% PGP and FA resulted in comparatively lower strengths than the control in both seven and 28 days. The degree of strength reduction was, however, dependent on the SCM used; coarse PGP reached the lowest level of strength, followed by FA and fine PGP, respectively. Although derived from the same source, the mortar with the fine PGP reached 52% higher compressive strength than the mortar with coarse PGP in 28 days, highlighting the influence of the fineness of PGP on strength gain. In addition, the fine PGP (that had similar PSD to that of FA, as shown in Fig. 1) achieved more than 15% higher compressive strength than FA mortars. It is hypothesised that this variation in strength is related to the difference in their chemical compositions. The fine PGP achieved the highest SAI followed by FA and coarse PGP, respectively. More importantly, the SAI of the coarse PGP and FA failed to reach 75% of the control while only the fine PGP reached the mark in both seven and 28 days. Therefore, the fine PGP conforms to the requirements set by ASTM C618-19  to be regarded as a pozzolan (particularly Class N). It should be noted that the FA used in this study closely failed to reach the Class F criteria in ASTM C618-19 due to the SAI index being less than 75%. However, this FA conforms to AS 3582.1  and is a widely used commercial product in Australia. The reactivity of the fine PGP can derive from its highly amorphous structure, relatively high contents of Na2O and CaO (more than 10%, as seen in Table 1), and the significantly low LOI in contrast to cement and FA. The relatively high alkali and CaO content may contribute to enhanced cement hydration , leading to better strength development, particularly in the first 28 days of hydration . An inverse relationship between LOI of SCMs and compressive strength was also reported in the previous studies [68, 69] and thus may corroborate the higher strength gain of fine PGP. However, the nature of this reactivity of PGP was assessed to be latent hydraulic in nature, as discussed further in Sects. 3.2 and 3.3 of this paper.
The current classification methods for pozzolans, particularly ASTM C618-19  is somewhat subject to interpretation, as the factors outlined in the standard to classify pozzolans such as, LOI, chemical oxide composition, and fineness, cover a broad range and materials, and within the range different performances may be observed. In addition, factors such as the quantity of amorphous phase and ability to consume CH are not reflected. These parameters can change even for certain well-recognised pozzolans (e.g., FA), depending on their source, and this variability is unavoidable. The pozzolanic property of the fine PGP, although conforms to the standardised definition and is in line with findings in the literature [34, 70, 71], cannot be taken as absolute without undertaking further analytical assessments. The SAI on its own may not provide a decisive outcome to characterise PGP as pozzolan because some anomaly may exist between the index and its true performance as a pozzolan . As such, further technical evidence is gathered in this research to assess fine PGP’s performance as an SCM. Nevertheless, the SAI test provided a comparative outcome, especially in demonstrating the higher reactivity of PGP when ground to a finer median size of 19.3 µm.
Heat evolution of pastes
The effects of cement substitution with FA and PGP on the rate of heat evolution (normalised to unit mass of binder and unit mass of cement) were investigated using an isothermal calorimeter and data is presented in Fig. 5. The typical stages of cement hydration  are observed in both heat flow curves in Fig. 5a, c. An initial peak of high magnitude occurs within approximately 30 min of hydration, which corresponds to the binder wetting and dissolution, followed by a short induction period. The end of the induction period marks the onset of the first acceleration period that reaches the peak at around 9–10 h. At this stage, two distinct peaks can be identified for all paste mixes. The first peak corresponded to the accelerating growth of cement hydration products (C–S–H), mainly through the hydration of alite (C3S). The second peak (also known as the sulphate depletion peak ) featured further formation of ettringite from hydration of the aluminate phases (C3A and C4AF) . The gypsum identified in Fig. 2 controls the rate of the aluminate phase hydration to avoid flash set and loss of workability. Following the sulphate depletion peak, the heat flow was decelerated, indicating the decreasing hydration reaction rate . The heat flow curve flattened beyond the deceleration and hydration continued in a steady-state phase as marked by the growing cumulative heat curves in Fig. 5b, d.
Normalising the heat of hydration to the total mass of binder and cement provides important information. The reduced heat flow and cumulative heat of Fig. 5a, b could mainly be the result of the dilution effect of both FA and PGP; the dilution is more prominent at a higher replacement level of 20%. However, at the same replacement level of PGP and FA, greater heat was generated in the PGP pastes than that of FA pastes, indicating the relatively superior reactivity of PGP. This higher cumulative heat of PGP pastes may corroborate the higher strength achieved in the SAI tests illustrated in Fig. 4. When heat flow is normalised per gram of cement (Fig. 5c), an increase in the heat flow is evident with the inclusion of SCMs. Both FA and PGP acted as fillers between cement grains, providing extra space and nucleation sites for the hydration products at the same water to binder (total solids) ratio [28, 77]. This filler effect was also reflected in the cumulative heat curves presented in Fig. 5d. With the addition of FA and PGP, the cumulative heat released per unit mass of cement increased; the rate is higher for a higher degree of substitution. Interestingly, PGP offers a higher heat flow than FA at the same replacement levels. Considering Class F fly ash are inert in the first 24 h of hydration, this increased heat release using PGP may indicate an early-age reaction potential of PGP. The previous studies [43, 44] also showed some extent of hydraulic property of powdered glass due to its CaO content.
It is noteworthy that the nature of the heat flow curves in Fig. 5a, c did not alter with the addition of FA and PGP—the slope of the acceleration curve remains almost identical. However, critical features were still noticeable with the time of occurrence of the acceleration and the sulphate depletion peaks. With the addition of SMCs, the acceleration peak was slightly delayed by approximately 20 min and 30 min for 10% and 20% FA, and 30 min and 50 min for 10% and 20% PGP addition. This delay suggests that both FA and PGP retards cement hydration to some extent with a relatively longer induction period. The literature reports that fly ash causes the delay from two competing phenomena—increase in the nucleation sites and retardation due to the adsorption of Ca2+ that would otherwise take part in CH formation [78, 79]. A similar mechanism may or may not explain the delay for PGP inclusion. However, by providing extra nucleation sites for cement hydration (as reflected in Fig. 5c), both SCMs counterbalance the retardation and thus, the delay in the formation of acceleration peak is marginal.
Thermogravimetric analysis (TGA)
The CH contents obtained from control, 20PGP, and 20FA using TGA are compared in Fig. 6 for up to 56 days of hydration. A gradual increase in the CH content for the control cement paste was observed up to 28 days, as expected, indicating continuing hydration of the cement. The CH content in the first four weeks of hydration could be linked to the hydration of alite (C3S). However, following 28 days, a small drop in CH (0.0032 g/g of anhydrous cement) was observed in the cement paste. Such a small change may be the experimental error caused by the detection limit of the data. The 5.9% calcite in the GP cement used in this study may form carboaluminates at a later age , triggering a slight drop in CH content in control and 20PGP between 28 and 56 days. Weerdt et al.  observed a similar drop in CH content in cement paste with 5% limestone and attributed the drop to secondary effect of carboaluminate formation or simply dilution effect.
On day seven, the CH content per gram of anhydrous cement in the pastes with 20% of FA and PGP was slightly higher than that obtained from the control. This seems to be due to the filler effect of FA and PGP as also identified in heat flow curves which increase the hydration degree of the cement particles.
As the pastes mature to 56 days, the consumption of CH in the FA paste was evident with a steady decline in the CH content over time. The declining trend suggested the FA’s continuing pozzolanic reaction in consuming CH to produce secondary C–S–H in the system . Attention can be drawn to the SAI of FA in Fig. 4—in 28 days, the FA specimens reached 71% strength of the control as opposed to 66% in 7 days. The TGA suggested that the pozzolanic reactivity of FA caused this improved strength at a later age. However, a contrasting observation was made for the paste with PGP. There is no clear evidence of CH consumption in Fig. 6 for 20% PGP replacement. On day 56, the relatively lower CH in the PGP paste may be attributed to the reaction of limestone of the GP cement to produce hemicarbonates , and may not originate from CH consumption by PGP. The CH content of the PGP sample was still greater than the control at 56 days, indicating enhanced hydration of the cement from the filler effect, a mechanism also corroborated by the calorimetric analyses.
Since the thermal analysis could not confirm the pozzolanic reactivity of PGP, the question remains as to the reasons for the significantly enhanced compressive strength achieved with PGP in the SAI test. The PGP used in this study may possess a latent hydraulic property that would explain the SAI achieved (more than 75%) in both 7 and 28 days of hydration. Suraneni et al.  observed a similar increasing trend of CH over time when using solid waste incineration FA with relatively higher CaO content. Another hypothesis is that glass constituents, instead of Ca from cement, get incorporated in forming of the CSH gels during hydration, allowing more CH to be formed . The PGP of this study, with about 11% CaO, may have prompted a similar mechanism; i.e., latent hydraulic reactivity while undergoing possible hydration of PGP.
Pore size distribution (below 60 nm)
The cumulative pore size distributions (below 60 nm) of all samples are presented in Fig. 7 for 7, 14, 28, and 56 days of maturity, which provide the basis for several critical observations. It should be noted that nitrogen adsorption/desorption fails to quantitatively capture pore sizes around and greater than 60 nm [60,61,62], and discussions hereafter are as such, reflective of pore sizes smaller than the range. These pores are micropores and mesopores in hydrated cement paste as defined in . The water to binder ratio of the pastes was 0.47 and at such level, the residual free water from cement hydration is expected to form coarser capillary pores [60, 83] that may not be captured in NAD isotherms . However, the cumulative pore size distribution within the micro and mesoporous region can still be evaluated qualitatively.
Figure 7a shows the gradual pore size transformation in hydrated cement paste over time. Between days 7 and 14, the volume of pores under 60 nm increased, due to the formation of small pores from the progressive filling of additional C–S–H gels in the void spaces from the ongoing hydration. The larger pores (> 60 nm) went through pore refinement and the volume of pores smaller than 60 nm increased. Between days 14 and 56, a volume reduction is observed of these pores, as identified by the curve progressively shifting downward. This indicates that the pore refinement of the control (100% cement) specimen was mostly achieved due to the fast reaction and continued hydration of cement in a steady-state condition, which may densify the cement matrix by reducing overall pore volumes of 10–60 nm. In essence, the larger pores (> 60 nm) transform to smaller ones and, with time, are further refined to result in lower pore volumes.
For FA incorporated paste in Fig. 7b, the temporal pore volume change is different from that observed for the control paste. Due to the dilution effect, the rate of hydration product formation and, consequently, the rate of pore refinement per unit volume of FA paste is expected to be slower than the control. Therefore, up to 28 days a slight, but gradual, refinement was marked by increasing volume of pores under 60 nm, likely from a higher degree of hydration for GP cement particles . However, between 28 and 56 days, when cement hydration is slower, an increasing volume of micro and mesopores are still observed in Fig. 7b, and this appears to be associated with the pozzolanic reaction of the FA as identified in the TGA analysis. As such, the secondary hydration products due to the pozzolanic reaction continuously fill the larger capillary pores and, consequently, create the small or median capillary pores (< 60 nm).
The pore size transformation of PGP incorporated paste in Fig. 7c is different to that of control but somewhat like that of FA. The volume of pores below 60 nm steadily rises over time, preferably from the continuing refinement of pores over 60 nm size. As per the case of FA paste (20 FA), the inclusion of the PGP led to a dilution effect, leading to the slower formation of hydration products (more porous matrix). With time, the large pores were continually being filled with hydration products from the hydration of cement. This may explain the observation that the volume of pores within 10–60 nm range did not drop, unlike in the control. Any contribution from the pozzolanic reaction cannot be confirmed for this pore refinement, as TGA analysis did not confirm PGP to consume CH; however, the latent hydraulic potential of PGP identified from the SAI test may contribute towards the gradual increase in the volume of pores below 60 nm size, due to the accumulation of additional reaction products from continued hydration. This is also evident in the heat release, where PGP offered relatively greater heat release than FA, likely from the combination of a higher number of nucleation sites and superior reactivity and thus a greater volume of larger than 60 nm pores.
Influence of recycled glass as fine aggregate
Accelerated mortar bar tests (AMBT)
The mortar bar expansions for inclusion of glass aggregate at 20% and 40% were recorded up to 21 days (AS 1141.60.1 ) and 14 days (ASTM C1260 ) and the aggregates were classified following the AS and ASTM limits (Fig. 8a, b). The incorporation of glass as a replacement for sand facilitated ASR and causes measurable expansions in the mortar bar specimens. All aggregate proportions over 10% are ‘reactive’ as per AS and ‘potentially deleterious’ as per ASTM specifications. An aggregate mix containing 10% GFA and 90% of river sand is categorised as ‘slowly reactive’ as per AS or ‘uncertain’ as per ASTM. The expansion steadily increases with higher proportions of GFA as illustrated in Fig. 9.
There is a significant decline in expansion when cement is partially replaced with FA and PGP (mitigation series in Table 3) at 10 and 20%, as observed in Fig. 8c, d. The ASTM C1567  was followed for the mitigation series, which allows the determination of the potential alkali-silica reactivity of combinations of cementitious materials and aggregate. A high potential for FA and PGP to limit ASR expansions was observed, with expansions of less than 0.10% after 14 days of conditioning were recorded for all mixes and, as such, the aggregates in such systems were innocuous. At a higher SCM replacement of 20%, the expansions were lower, indicating the greater mitigation potential for higher SCM dosages. Moreover, mixes with FA demonstrated a relatively better potential for ASR mitigation than PGP. The role of FA and PGP in the mitigation of ASR expansion can be further discussed in the context of microstructural characterisations undertaken in this study.
Several mechanisms can contribute to the ASR mitigative potential of SCMs, including alkali dilution, alkali binding, reduced mass transport, consumption of portlandite in GP cement (thus, reducing pore solution pH), changes in pore solution chemistry, etc. [19, 39]. Although both FA and PGP were successful to keep ASR expansions within the limit (0.10%), their mechanisms may be different. It is recognised that AMBT conditions specimens at 80˚C accelerates reactions of SCMs . On the one hand, the data seems to be still relevant to specimens exposed to ambient conditions, despite the accelerated reactions at higher temperature, as the mechanism behind reduced expansions is similar [84,85,86]. On the other hand, the alkali contribution of SCMs on the alkalinity of the pore solution remains unclear, and is worthy of further investigation.
When replacing 20% cement with FA or PGP, there was an alkali and calcium dilution in the pore solution that, somewhat, restricts the process leading to damaging expansions. However, given that the specimens had continuous alkali supply from immersion in 1 mol/L NaOH solution throughout the test duration, this may not be the critical mechanism behind the expansion mitigation. FA demonstrated its ability to consume CH, making soluble calcium less available for the formation of expansive ASR gels; the presence of CH is critical for expansions due to ASR [19, 27]. Moreover, the pozzolanic reactivity of the FA means that the FA binds alkali and hydroxyl ions in the pore solution limiting their availability in ASR gel formation . As a result, the efficiency of FA in expansion mitigation was marginally better than PGP. However, with a constant supply of alkalis in the AMBT test, the most critical mechanism is likely the limiting of mass transport from the inclusion of FA or PGP; both materials have taken part in pore refinement of large pores as suggested by the NAD analysis. This appears to limit the mass transport of alkalis from the 1 M NaOH solution. Although the contribution of the pozzolanic reactivity of FA was evident, the mechanism of PGP to some extent was unclear. However, its potential for ASR inhibition remains noteworthy. This may indicate that the amorphous silica of PGP (being several times finer than GFA) and its relatively high content of CaO may take part in both pozzolanic reactions and hydraulic reactions that can contribute to the formation of secondary C–S–H gels. The reactions may have been promoted at the temperature range of 80 °C adapted in the AMBT test , limiting the expansions due to ASR within the acceptable limit of 0.10% as per ASTM 1567 . Mirzahosseini and Riding  reported that glass powder does not demonstrate pozzolanic reactivity at 10 °C and 23 °C; however, at 50 °C, it consumes 17% CH in 28 days at a replacement level of 25%. At the 80 °C conditioning adopted in this study for the AMBT tests, both the FA and PGP demonstrated enhanced reactivity. As such, they take an active part in pore refinement and reduced the mobility of the alkalis, thus limiting ASR gel formation. For this reason, the ABMT samples with SCMs show a rare occurrence of ASR gels in their microstructure, which is discussed in the next section.
SEM imaging of the samples with and without SCMs were undertaken to identify changes in morphological features, enabling qualitative identification of ASR gels. EDS analysis was also undertaken to observe if the presence of FA and PGP alter the chemical composition of the ASR gels. The SEM images present a clear distinction between the control specimens (Fig. 10a, b) to those with 10% FA (Fig. 10c) and 10% PGP (Fig. 10d) after 14 days of exposure to 1 mol/L NaOH solution at 80 °C as per ASTM C1567 . The control specimen was severely damaged with a large volume of ASR gel formation within the glass aggregates, often to an extent causing total disruption of the aggregate particle (Fig. 10a). The swelling of the ASR gels in the control specimen also led to the formation of microcracks within the matrix, severely damaging the integrity of the specimens. Extensive cracking of the mortar prisms was also visually observed in some specimens. To the contrary, with the substitution of 10% cement with FA or PGP, the occurrence of ASR gel was significantly reduced. Only a trace amount of ASR gel is be observed in Fig. 10c, d, with several glass aggregates showing no indication of gel formation. The ASR gel thickness formed in the control specimens range from 50 to 500 µm; whereas, in the SCM specimens, the gel thickness reduced to about 2–5 µm, meaning the gel did not swell sufficiently to cause microcracking. The microstructure corroborates the expansion readings of Fig. 8c, d, where a significant reduction is achieved with SCM substitution.
Approximate Ca/Si and (Na + K)/Si ratios were measured based on EDS analysis on ASR gels formed within the glass particles, and in the microcracks, in the vicinity of glass particles. Figure 10e–g show representative locations where EDS analysis was undertaken; not all images have been included in this paper. The results of elemental analysis are summarised in Table 5. Mirzahosseini and Riding [88, 89] and Afshinnia and Rangaraju [88, 89] report that ASR gels initially form in the cracks within glass particles (cracks originating from the mechanical crushing of recycled glass) before expanding through the microcracks within the matrix around the glass particle and progressively inducing damage. Assessment of the elemental compositions demonstrates that the composition of ASR gel was location-dependent but not dependent on the binder composition . The results also suggest that the reactive silica from the amorphous FA, and the PGP, likely contribute to reaction product formations, rather than in ASR gel formations, unlike species from the GFA.
The Ca/Si ratio measured on ASR gels within glass particles was of the range of 0.23–0.26, characterising typical gel compositions in alkali-reactive aggregates [10, 90, 91]. However, the Ca/Si ratio of ASR gel measured within microcracks along the interfacial transition zone (ITZ) was 0.56. This change in the elemental ratio confirms that the composition of the gel is location dependent. Also, the Ca/Si ratio measured within the cracks was much lower than the typical Ca/Si of a neat Portland cement paste (~ 1.75) . This obvious difference suggests that the locations chosen for the elemental analysis are indeed ASR gels rich in alkali and Si, and low in Ca, that have propagated in the microcracks around the glass aggregates . The increased Ca/Si ratio in the matrix, compared to within GFA, is linked to alkali recycling due to the calcium replacement of ASR gels . Similar findings with higher Ca/Si ratio are reported in [90, 91, 95, 96].
The Ca/Si ratio of ASR gels in the FA and PGP incorporated mixes were similar to that of the control, with no appreciable difference in composition. This comparable composition of the ASR gel in the control and SCM specimens indicates that neither FA or PGP necessarily alter ASR gel composition . Thus, it is not the nature of the ASR gel that dictates the reduced expansion, rather the substantial reduction of ASR gel formation, in contrast to the control specimens, that leads to the reduced expansion recorded in the AMBT test. SCMs limit the unfavourable conditions for a damaging level of ASR gel formation, such as high alkali content in the pore solution, the availability of reactive silica, and easy access for ion transport, etc. They do not necessarily alter the chemistry of the ASR gel towards reduced expansion.