A Step by Step Methodology for Building Sustainable Cementitious Matrices

Featured Application: A formulation and testing protocol for the development of low Portland cement content cementitious binders. Abstract: In an e ﬀ ort to produce cost-e ﬀ ective and environmentally friendly cementitious binders. mainly ternary (Portland cement + limestone + pozzolanas) formulations have been investigated so far. Various proportions of constituents have been suggested, all, however, employing typical Portland cement (PC) substitution rates, as prescribed by the current codes. With the current paper a step by step methodology on developing low carbon footprint binary, ternary and quaternary cementitious binders is presented (PC replacement up to 57%). Best performing binary (60% PC and 40% LS (limestone)) and ternary formulations (60% PC, 20% LS, 20% FA (ﬂy ash) or 43% PC, 20% LS 37% FA) were selected on the grounds of sustainability and strength development and were further optimized with the addition of silica fume. For the ﬁrst time a protocol for successfully selecting and testing binders was discussed and the combined e ﬀ ect of highly pozzolanic constituents in low PC content formulations was assessed and a number of successful matrices were recommended. The present paper enriched the current state of the art in composite low carbon footprint cementitious binders and can serve as a basis for further enhancements by other researchers in the ﬁeld.


Introduction
Concrete, and subsequently Portland cement, is one of the commodities whose demand is ever increasing. According to the 2050 projections of CEMBUREAU, developing countries and emerging markets such as India and China are expected to need almost six billion tonnes of cement by the year 2050 [1]. Taking into consideration that production of cement is a major cause for carbon dioxide emissions, accountable for about 5-8% of anthropogenic CO 2 emissions worldwide, concrete and cement research has been focusing on the production of low Portland cement formulations using supplementary cementitious materials, such as limestone, fly ash, ground granulated blast furnace slag and silica fume.
Although the present research is focused and based on European Standards it should be noted that a global shift towards performance based rather than prescriptive specifications (composition based formulations) can be observed. In fact, particularly with respect to fly ash addition, the American Concrete Institute suggests a maximum limit of 35% of fly ash addition in binary compositions while

•
Low carbon footprint binary and ternary cementitious matrices ( Figure 1) suitable for further modifications with the addition of silica fume were identified • Best performing cementitious matrices were selected in terms of embodied energy and compressive strength • Divergent results on the effect of the addition of limestone in composite cement binders were clarified • The pozzolanic reactivity in ternary and quaternary cement paste formulations was investigated • The combined effect of highly pozzolanic constituents in low PC content formulations was assessed • The current knowledge on cement binder optimization was broadened.
In greater detail, apart from the industrial cement CEM II/A-L42.5 conforming to BS EN 197-1 [20] which was studied on its own, given the name B1 in the binary combinations ( Figure 1) three more binary limestone-cement combinations were created with reduced Portland cement (PC) content in a gradual manner (B2 contained 80%PC, B3 contained 70%PC and B10 contained 60%PC. At the same time, ternary blended cements were studied at various fly ash contents starting from 20%. Then a 35% FA content formulation was studied in two different pastes (B5 and B8). The specific amount is the upper limit according to ACI [2] for normal other than mass concretes in binary combinations and with respect to the European Standards used in this research, according to BS EN 197-1 [20], the Portland cement (PC) clinker content should not be less than 65% in pozzolanic blends containing limestone. Therefore, the ternary ones pose a significant challenge due to the further reduction of PC content.
Next, the performance-based concept was adopted, using as key indicators the 28 day compressive strength of the ten formulations, the hydration reactivity at 7 and 28 days of curing and the embodied CO 2 , as a result of the reduction of PC content and increase in the limestone and fly ash content. Of the ten formulations, three were selected for scanning electron images in order to detect durability issues in the forms of microcracks, which were taken into consideration for the additional quaternary formulations. This gap is bridged by the current research. In fact a step by step methodology for the development of cementitious matrices is presented, based on strength and sustainability criteria. In greater detail, the following items were pursued in this study: • Low carbon footprint binary and ternary cementitious matrices ( Figure 1) suitable for further modifications with the addition of silica fume were identified • Best performing cementitious matrices were selected in terms of embodied energy and compressive strength • Divergent results on the effect of the addition of limestone in composite cement binders were clarified • The pozzolanic reactivity in ternary and quaternary cement paste formulations was investigated • The combined effect of highly pozzolanic constituents in low PC content formulations was assessed • The current knowledge on cement binder optimization was broadened.
In greater detail, apart from the industrial cement CEM II/A-L42.5 conforming to BS EN 197-1 [20] which was studied on its own, given the name B1 in the binary combinations ( Figure 1) three more binary limestone-cement combinations were created with reduced Portland cement (PC) content in a gradual manner (B2 contained 80%PC, B3 contained 70%PC and B10 contained 60%PC. At the same time, ternary blended cements were studied at various fly ash contents starting from 20%. Then a 35% FA content formulation was studied in two different pastes (B5 and B8). The specific amount is the upper limit according to ACI [2] for normal other than mass concretes in binary combinations and with respect to the European Standards used in this research, according to BS EN 197-1 [20], the Portland cement (PC) clinker content should not be less than 65% in pozzolanic blends containing limestone. Therefore, the ternary ones pose a significant challenge due to the further reduction of PC content.
Next, the performance-based concept was adopted, using as key indicators the 28 day compressive strength of the ten formulations, the hydration reactivity at 7 and 28 days of curing and the embodied CO2, as a result of the reduction of PC content and increase in the limestone and fly ash content. Of the ten formulations, three were selected for scanning electron images in order to detect durability issues in the forms of microcracks, which were taken into consideration for the additional quaternary formulations. The last step included fine tuning of the selected formulations by increasing the FA content to 37% for one of the ternary formulations, yielding; PC60LS40, PC60LS20FA20 and PC43LS20FA37. The latter two formulations were further investigated after the addition of two different proportions of silica fume (formulations typed in red in Figure 1). The resulting matrices were analyzed again against the key indicators determined above; compressive strength 7, 28 and 56 days of curing, the hydration reactivity at 7, 28 and 56 days of curing, the embodied CO 2 and scanning electron imaging for microstructural defects, which could undermine durability. All the above, have been studied and analyzed in the subsequent sections, which paved the way for further modifications with the use of various nanoparticles (Figure 1).

Materials
The following materials and any commercially available data are listed below. These materials were characterized with a selection of the techniques, before implemented in the various cement paste combinations: The following methodology was adopted for the mixing procedure: • Dry mixing of all powders with a spatula by hand.

•
With the addition of water, the paste was mixed by a with an automatic dual shaft mixer at 1150 rpm for 3 min for pastes up to 300 g, time increasing with amounts of paste.

•
Cylindrical specimens of 60 mm height and 30 mm diameter were prepared according to BS EN 196-1 [21].

•
The cylindrical specimens of the initial cementitious matrices (B1-B10) were air-cured. However, when the final selection of binary and ternary matrices was made, the three selected cementitious matrices were reproduced, air-cured for 1 day, demoulded and then water cured until the day of testing.

Step 1: Design of Cementitous Matrices
The design of the cementitious matrices was carried out in a step-by-step manner and was aimed at designing and testing a low carbon dioxide footprint cement, which, at the same time, would have the highest possible compressive strength. As shown in Figure 2, according to BS EN 197-1 [20], for binary cements, containing clinker and limestone, only two types are allowed, CEM II/A-L and CEM II/B-L. In the former, the limestone content should not exceed 20%, whereas in the latter, the limestone content should not exceed 35%. Starting with B1, which was the as-received-cement by the industrial partner of the project FIBCEM, the amount of clinker was gradually reduced in the subsequent formulations (B2, B3 and B10).
Similarly, for ternary cementitious matrices containing clinker, fly ash and limestone, according to BS EN 197-1, the Portland cement (PC) clinker content should not be less than 65%. Therefore, ternary combinations (B4, B5, B6, B7, B8 and B9) where designed with a PC clinker upper bound content of 65%.

of 25
Best performing pastes were identified and selected ternary formulations were subsequently optimized with the addition of silica fume, as the last step of the current research. The exact procedure is described in Sections 2.3.1 and 2.3.2, which is completed in Section 3, with the analysis of the results.
Appl. Sci. 2020, 10, x FOR PEER REVIEW 5 of 26 Best performing pastes were identified and selected ternary formulations were subsequently optimized with the addition of silica fume, as the last step of the current research. The exact procedure is described in Sections 2.3.1 and 2.3.2, which is completed in Section 3, with the analysis of the results.

Binary & Ternary Cement Formulations
The primary constituent is Portland limestone cement, CEMII/A-L42.5, which is a binary combination on its own, since it contains 86% of PC and 14% of LS, as verified by TGA/dTG, shown in Section 3.1.2. The first specimen produced, then, was named B1 and contained only CEMII. Eight more samples were produced next. B2 and B3 contained LS at 20% and 30%, respectively. B4 to B9 represented ternary formulations of PC, LS and FA. All 9 initial bases were mixed at a water to binder (w/b) ratio of 0.3. The exact proportions are shown in Table 1 that follows:

Quaternary Cement Formulations
B7 and B8 were found to perform best as will be discussed they were further enhanced by the addition of μS. Initially, two quaternary formulations, having 60% PC by mass, in which the μS content was subtracted by the PC content, were designed as shown in Table 2

Binary & Ternary Cement Formulations
The primary constituent is Portland limestone cement, CEMII/A-L42.5, which is a binary combination on its own, since it contains 86% of PC and 14% of LS, as verified by TGA/dTG, shown in Section 3.1.2. The first specimen produced, then, was named B1 and contained only CEMII. Eight more samples were produced next. B2 and B3 contained LS at 20% and 30%, respectively. B4 to B9 represented ternary formulations of PC, LS and FA. All 9 initial bases were mixed at a water to binder (w/b) ratio of 0.3. The exact proportions are shown in Table 1 that follows:

Quaternary Cement Formulations
B7 and B8 were found to perform best as will be discussed they were further enhanced by the addition of µS. Initially, two quaternary formulations, having 60% PC by mass, in which the µS content was subtracted by the PC content, were designed as shown in Table 2. Then, a lower PC content paste was designed, containing 43% PC, 20%LS and 37% FA by mass as a reference paste, which was further enhanced by the addition of µS, subtracted by the LS content, this time, in order to maintain a w/b ratio of 0.3, as shown in Table 3.

Preparation of Pastes for Characterization
For the characterization of the pastes, arrest of hydration was performed following two different methodologies. In brief, if images were to be generated and the microstructure to be evaluated, such as SEM/FESEM and MIP, solvent exchange was the technique employed for the arrest of hydration. For experiments in which chemical reactions would occur or where the chemical composition had to be evaluated, such as TGA, XRD, the oven drying technique was adopted as explained in detail in [22]. For the SEM/FESEM and MIP studies, the samples were crashed and immersed in isopropanol for 48 h. Isopropanol was selected as the most appropriate solvent according [23,24]. The samples were, then, vacuum dried for 24 h. The cement paste samples were uncoated when examined at the SEM. For the chemical analyses studies the samples were crushed and placed in the oven at 60 • C for up to 24 h. Then, they were ground and sieved through a 125 µm sieve and placed back in the oven at 60 • C for another 24 h. All samples were placed in the desiccator in a labelled air-tight container until they were analyzed. For the compressive strength tests, the three samples prepared for each paste formulation were removed from the curing tank a few minutes before being tested. Any uneven surfaces were levelled with a diamond saw, avoiding capping. Consecutively, the specimens were weighed and measured.

Experimental Techniques
The entire experimental program carried out in this research paper is summarized in Figure 3.
2.5.1. Thermogravimetric Analyses and Differential Thermogravimetry (TGA/dTG) Thermogravimetric analyses were carried out using a TGA92 instrument (Setaram, Caluire, France). Approximately 20 mg of each sample were placed in an alumina crucible and heated at a rate of 10 • C/min from 20 • C to 1000 • C under 100 mL/min flow of inert nitrogen gas. Mass change, differential mass change and heat flow measurements were recorded and analyzed using the built-in software. The differential thermogravimetric curve (dTG) was derived by the TG curve. The first derivative curve was produced for the various samples tested and was used for comparisons instead of the mass loss curve, as it yields sharp distinctive peaks.  TGA/dTG was carried out for both constituent materials and cement pastes. With respect to hardened cement pastes, three broad areas, can be distinguished by the TGA; the first one is related to the dehydration of C-S-H ettringite and monosulfate, between 100 °C and 200 °C. It can be postulated that the greater the loss measured the greater the amount of C-S-H and ettringite was produced by the hydration of the paste.
The second area of interest is associated with the dehydroxylation of Ca(OH)2 between 400 °C and 510 °C. The amount of Ca(OH)2 present in the paste at different ages can be computed by the stoichiometric elaboration of the mass loss results within the specific temperature range [25].
The third area of interest is the decomposition of CaCO3 known as decarboxylation, occurring between 610 °C and 810 °C. The amount of CaCO3 present in the paste at different ages can be computed by the stoichiometric elaboration of the mass loss results within the specific temperature range. If it is greater than the limestone content of the reference paste, it constitutes an indication of sample carbonation. Particularly for the quantification of Ca(OH)2, TGA is considered a very accurate technique [26], especially if coupled with XRD [27][28][29][30].

Scanning Electron Microscopy (SEM)/X-Ray Energy Dispersive Spectroscopy (SEM/EDX)
The surface morphology was observed with SEM. Back scattered micrographs of uncoated samples at various magnifications and beam currents were obtained using a Jeol 6480 LV scanning electron microscope. For SEM/EDX elemental composition analyses sample was placed uncoated on a sheet of molybdenum for unbiased elemental analyses. A matrix of 5 × 5 spectra was acquired and the median of the elemental composition was presented.

Particle Size Distribution Analysis (PSD)
Particle size distribution analysis with optical system can provide the PSD of a wide range of scales of particles from 10 nm to 3 mm. Particle size distribution analysis was carried out at the MANIT laboratory, Bhopal, India, using an LA950 Laser Scattering Particle Size Distribution Analyzer (Horiba, New Delhi, India).

A) Constituent materials B) Binary and ternary cementitious matrices
Scanning electron microscopy (SEM) imaging was carried out for; PC86LS14, PC60LS20FA20 and PC60LS40 at 28 days of curing. PC60LS40 images at 1 day of curing were also taken.
Thermogravimetric analyses and differential thermogravimetry (TGA/dTG) (all formulations were tested) at 7 and 28 days of curing  TGA/dTG was carried out for both constituent materials and cement pastes. With respect to hardened cement pastes, three broad areas, can be distinguished by the TGA; the first one is related to the dehydration of C-S-H ettringite and monosulfate, between 100 • C and 200 • C. It can be postulated that the greater the loss measured the greater the amount of C-S-H and ettringite was produced by the hydration of the paste.
The second area of interest is associated with the dehydroxylation of Ca(OH) 2 between 400 • C and 510 • C. The amount of Ca(OH) 2 present in the paste at different ages can be computed by the stoichiometric elaboration of the mass loss results within the specific temperature range [25].
The third area of interest is the decomposition of CaCO 3 known as decarboxylation, occurring between 610 • C and 810 • C. The amount of CaCO 3 present in the paste at different ages can be computed by the stoichiometric elaboration of the mass loss results within the specific temperature range. If it is greater than the limestone content of the reference paste, it constitutes an indication of sample carbonation. Particularly for the quantification of Ca(OH) 2 , TGA is considered a very accurate technique [26], especially if coupled with XRD [27][28][29][30].

Scanning Electron Microscopy (SEM)/X-Ray Energy Dispersive Spectroscopy (SEM/EDX)
The surface morphology was observed with SEM. Back scattered micrographs of uncoated samples at various magnifications and beam currents were obtained using a Jeol 6480 LV scanning electron microscope. For SEM/EDX elemental composition analyses sample was placed uncoated on a sheet of molybdenum for unbiased elemental analyses. A matrix of 5 × 5 spectra was acquired and the median of the elemental composition was presented.

Particle Size Distribution Analysis (PSD)
Particle size distribution analysis with optical system can provide the PSD of a wide range of scales of particles from 10 nm to 3 mm. Particle size distribution analysis was carried out at the MANIT laboratory, Bhopal, India, using an LA950 Laser Scattering Particle Size Distribution Analyzer (Horiba, New Delhi, India). XRD measurements were performed for the FA using a D8 ADVANCE X-ray diffractometer (Bruker, Coventry, U.K.) with CuKα radiation. Spectra were obtained in the range 4 • < 2θ < 20 • at an angular step-size of 0.016 • 2θ.

Mechanical Properties Testing Techniques
The strength of the supplied cement was validated according to BS EN 196-1 [21]. All samples were tested at a loading speed of 0.5 MPa/s on a 100 kN servo hydraulic machine (Dartec, Stourbridge, U.K.). Mean strength values of the specimens tested were calculated.

Carbon Dioxide (CO 2 ) Footprint
Preliminary estimation of the CO2 footprint of the initial pastes was derived from compressive strength of cement pastes (MPa) evaluated against embodied CO 2 . (ECO 2 ). The levels of ECO 2 have been agreed in the UK and are evaluated in terms of kg CO 2 per tonne of binding material [31].
It can be seen that in terms of CO 2 footprint the use of FA as a supplementary cementitious material is beneficial. This is the reason for which it was used in the paste design despite of its shortage in the UK. Additionally: • µS = 14 kg CO 2 /tonne [32]

Characterization of Constituent Materials
All commercially available materials listed in Section 2.1 were characterized with a selection of techniques (PSD, TGA/dTG, XRD and SEM/EDX), before implemented in the cementitious matrices.

PSD Results of CEM II/LS/FA/Condensed µS
The particle size distribution results of the four constituent materials reacted with water were prepared by colleagues in MANIT, India and are presented below. Results are typical for this type of industrially available materials, as can be seen in Table 4 and in the following Figures 4-7.

X-Ray Diffraction (XRD)
XRD measurements were performed for the FA using a D8 ADVANCE X-ray diffractometer (Bruker, Coventry, U.K.) with CuKα radiation. Spectra were obtained in the range 4° < 2θ < 20° at an angular step-size of 0.016° 2θ.

Mechanical Properties Testing Techniques
The strength of the supplied cement was validated according to BS EN 196-1 [21]. All samples were tested at a loading speed of 0.5 MPa/s on a 100 kN servo hydraulic machine (Dartec, Stourbridge, U.K.). Mean strength values of the specimens tested were calculated.

Carbon Dioxide (CO2) Footprint
Preliminary estimation of the CO2 footprint of the initial pastes was derived from compressive strength of cement pastes (MPa) evaluated against embodied CO2. (ECO2). The levels of ECO2 have been agreed in the UK and are evaluated in terms of kg CO2 per tonne of binding material [31].
It can be seen that in terms of CO2 footprint the use of FA as a supplementary cementitious material is beneficial. This is the reason for which it was used in the paste design despite of its shortage in the UK. Additionally:

Characterization of Constituent Materials
All commercially available materials listed in Section 2.1 were characterized with a selection of techniques (PSD, TGA/dTG, XRD and SEM/EDX), before implemented in the cementitious matrices.

PSD Results of CEM II/LS/FA/Condensed μS
The particle size distribution results of the four constituent materials reacted with water were prepared by colleagues in MANIT, India and are presented below. Results are typical for this type of industrially available materials, as can be seen in Table 4 and in the following Figures 4-7.

TGA Results of CEM II/LS/FA/μS
The thermogravimetric analyses results of the four constituent materials are presented in Figure  8. It can be seen that practically the only mass loss upon heating is observed for, CEMII and LS, the differential mass loss of which was measured to be 42.4% and 5.95%, respectively. The initial amount of CaCO3 was calculated via stoichiometry [25] and yielded 14% of CaCO3 present in CEMII and 97% in LS. Therefore, the LS powder is considered to be almost pure CaCO3 and the composition of CEMII is considered as 86% PC and 14% LS. The minor mass loss of FA with a peak at about 850 °C can be attributed to the volatile matter in the unburnt coal [33].

TGA Results of CEM II/LS/FA/μS
The thermogravimetric analyses results of the four constituent materials are presented in Figure  8. It can be seen that practically the only mass loss upon heating is observed for, CEMII and LS, the differential mass loss of which was measured to be 42.4% and 5.95%, respectively. The initial amount of CaCO3 was calculated via stoichiometry [25] and yielded 14% of CaCO3 present in CEMII and 97% in LS. Therefore, the LS powder is considered to be almost pure CaCO3 and the composition of CEMII is considered as 86% PC and 14% LS. The minor mass loss of FA with a peak at about 850 °C can be attributed to the volatile matter in the unburnt coal [33].

TGA Results of CEM II/LS/FA/μS
The thermogravimetric analyses results of the four constituent materials are presented in Figure  8. It can be seen that practically the only mass loss upon heating is observed for, CEMII and LS, the differential mass loss of which was measured to be 42.4% and 5.95%, respectively. The initial amount of CaCO3 was calculated via stoichiometry [25] and yielded 14% of CaCO3 present in CEMII and 97% in LS. Therefore, the LS powder is considered to be almost pure CaCO3 and the composition of CEMII is considered as 86% PC and 14% LS. The minor mass loss of FA with a peak at about 850 °C can be attributed to the volatile matter in the unburnt coal [33].

TGA Results of CEM II/LS/FA/µS
The thermogravimetric analyses results of the four constituent materials are presented in Figure 8. It can be seen that practically the only mass loss upon heating is observed for, CEMII and LS, the differential mass loss of which was measured to be 42.4% and 5.95%, respectively. The initial amount of CaCO 3 was calculated via stoichiometry [25] and yielded 14% of CaCO 3 present in CEMII and 97% in LS. Therefore, the LS powder is considered to be almost pure CaCO 3 and the composition of CEMII is considered as 86% PC and 14% LS. The minor mass loss of FA with a peak at about 850 • C can be attributed to the volatile matter in the unburnt coal [33].

TGA Results of CEM II/LS/FA/μS
The thermogravimetric analyses results of the four constituent materials are presented in Figure  8. It can be seen that practically the only mass loss upon heating is observed for, CEMII and LS, the differential mass loss of which was measured to be 42.4% and 5.95%, respectively. The initial amount of CaCO3 was calculated via stoichiometry [25] and yielded 14% of CaCO3 present in CEMII and 97% in LS. Therefore, the LS powder is considered to be almost pure CaCO3 and the composition of CEMII is considered as 86% PC and 14% LS. The minor mass loss of FA with a peak at about 850 °C can be attributed to the volatile matter in the unburnt coal [33].

XRD and SEM/EDX results of FA
Moreover, the different minerals present in the FA powder were determined by XRD crystallographic analysis [34]. As shown in Figure 9, mullite, quartz, hematite, lime and calcium carbonate are present [9].

XRD and SEM/EDX results of FA
Moreover, the different minerals present in the FA powder were determined by XRD crystallographic analysis [34]. As shown in Figure 9, mullite, quartz, hematite, lime and calcium carbonate are present [9]. Lastly, 20 spectra were acquired via the SEM/EDX carried out on a FA sample, five of which are indicatively shown in Table 5. The presence of sulphur which has been reported to increase the ettringite formation during cement hydration was confirmed [35].

Step 2: Testing and Step 3: Selecting best Performing Pastes of the Binary and Ternary Cementitious Matrices
After Step 1 was presented in Section 2.3, in which the concept behind the designing of cementitious matrices was unfolded, in this section the results of sample testing and selecting best performing pastes is analyzed. Apart from the pure CEMII paste (B1), nine additional Portland limestone cement initial formulations (B2-10) were designed, containing different proportions of CEMII and limestone or CEMII, limestone and fly ash, as already discussed. Best performing pastes were judged according to the delivered compressive strength and embodied carbon dioxide content. As shown below, best relations between strength and embodied CO2 were obtained for B7: PC60LS20FA20 and B8: PC45LS20FA35, which delivering 80 MPa/t/t and 110 MPa/t/t, respectively. These two formulations were selected to be further optimized, later, forming the quaternary cementitious matrices. Thermal analyses were also performed on all initial cementitious matrices (B1-B10). All results are presented in the following sections. Lastly, 20 spectra were acquired via the SEM/EDX carried out on a FA sample, five of which are indicatively shown in Table 5. The presence of sulphur which has been reported to increase the ettringite formation during cement hydration was confirmed [35].

Step 2: Testing and Step 3: Selecting best Performing Pastes of the Binary and Ternary Cementitious Matrices
After Step 1 was presented in Section 2.3, in which the concept behind the designing of cementitious matrices was unfolded, in this section the results of sample testing and selecting best performing pastes is analyzed. Apart from the pure CEMII paste (B1), nine additional Portland limestone cement initial formulations (B2-10) were designed, containing different proportions of CEMII and limestone or CEMII, limestone and fly ash, as already discussed. Best performing pastes were judged according to the delivered compressive strength and embodied carbon dioxide content. As shown below, best relations between strength and embodied CO 2 were obtained for B7: PC60LS20FA20 and B8: PC45LS20FA35, which delivering 80 MPa/t/t and 110 MPa/t/t, respectively. These two formulations were selected to be further optimized, later, forming the quaternary cementitious matrices. Thermal analyses were also performed on all initial cementitious matrices (B1-B10). All results are presented in the following sections.

Compressive Strength of Binary and Ternary Cementitious Matrices (B1-B10)
Binary and ternary cementitious matrices (B1-B10) were tested in compression at day 7, 28 and 56, as shown in Figure 10. Binary and ternary cementitious matrices (B1-B10) were tested in compression at day 7, 28 and 56, as shown in Figure 10. The compressive strength tests allowed the rejection of B1, B2 and B3 on the grounds of high PC content and performance that can be achieved with greener combinations. For example, B5 showed better long term performance than B1. B9 was also rejected due to poor performance.

Embodied CO2 of Binary and Ternary Cementitious Matrices (B1-B10)
Then, the sustainability potential of all binary and ternary cementitious matrices was calculated with respect to the embodied CO2 (E.CO2) of PC (930 kg CO2/tonne), LS (32 kg CO2/tonne) and FA (4 kg CO2/tonne) [36,37]. The results are presented in Table 6. By observing the results provided in Table 6, it is evident that B6 to B10 had the lowest total E.CO2. Furthermore, B7, B8 and B10 had the lower environmental impact and exhibited the highest Cylinder compressive strength (MPa) The compressive strength tests allowed the rejection of B1, B2 and B3 on the grounds of high PC content and performance that can be achieved with greener combinations. For example, B5 showed better long term performance than B1. B9 was also rejected due to poor performance.

Embodied CO 2 of Binary and Ternary Cementitious Matrices (B1-B10)
Then, the sustainability potential of all binary and ternary cementitious matrices was calculated with respect to the embodied CO 2 (E.CO 2 ) of PC (930 kg CO 2 /tonne), LS (32 kg CO 2 /tonne) and FA (4 kg CO 2 /tonne) [36,37]. The results are presented in Table 6. By observing the results provided in Table 6, it is evident that B6 to B10 had the lowest total E.CO 2 . Furthermore, B7, B8 and B10 had the lower environmental impact and exhibited the highest mechanical properties in terms of compressive strength compared amongst the initial pastes. As an effect, B7, B8 and B10 were the three selected bases for further optimization with the addition of microsized elements.

Thermal Analyses of Binary and Ternary Cementitious Matrices (B1-B10)
TG analyses were carried out at 7 and 28 days, the results of which are shown in Figures 11 and 12. It should be noted that initially the first nine bases were dry cured (B10 was only water cured), however none had carbonated since the CaCO 3 content is lower than the initial LS content in the paste.
Appl. Sci. 2020, 10, x FOR PEER REVIEW 12 of 26 mechanical properties in terms of compressive strength compared amongst the initial pastes. As an effect, B7, B8 and B10 were the three selected bases for further optimization with the addition of microsized elements.

Thermal Analyses of Binary and Ternary Cementitious Matrices (B1-B10)
TG analyses were carried out at 7 and 28 days, the results of which are shown in Figures 11 and  12. It should be noted that initially the first nine bases were dry cured (B10 was only water cured), however none had carbonated since the CaCO3 content is lower than the initial LS content in the paste.  In addition to the thermal analyses carried out at day 7 and day 28, the three selected cementitious matrices (PC60LS40, PC60LS20FA20 and PC43LS20FA37), were reproduced, water cured after day 1 and TG analyses were carried out at day 1, 7, 28 and 56. For PC60LS40, TG analyses were carried out at day 90, as well. At this point, the Ca(OH)2 consumption will be discussed ( Table  7) in an effort to shed more light into the hydration processes. To begin with, it can be noticed that until day 28, that PC continues hydrating, the Ca(OH)2 content of PC60LS40 increased and from then after it remained almost constant. It can be claimed, therefore, that the presence of FA in PC60LS20FA20 increased the consumption of the Ca(OH)2 content significantly over all ages.   Appl. Sci. 2020, 10, x FOR PEER REVIEW 12 of 26 mechanical properties in terms of compressive strength compared amongst the initial pastes. As an effect, B7, B8 and B10 were the three selected bases for further optimization with the addition of microsized elements.

Thermal Analyses of Binary and Ternary Cementitious Matrices (B1-B10)
TG analyses were carried out at 7 and 28 days, the results of which are shown in Figures 11 and  12. It should be noted that initially the first nine bases were dry cured (B10 was only water cured), however none had carbonated since the CaCO3 content is lower than the initial LS content in the paste.  In addition to the thermal analyses carried out at day 7 and day 28, the three selected cementitious matrices (PC60LS40, PC60LS20FA20 and PC43LS20FA37), were reproduced, water cured after day 1 and TG analyses were carried out at day 1, 7, 28 and 56. For PC60LS40, TG analyses were carried out at day 90, as well. At this point, the Ca(OH)2 consumption will be discussed ( Table  7) in an effort to shed more light into the hydration processes. To begin with, it can be noticed that until day 28, that PC continues hydrating, the Ca(OH)2 content of PC60LS40 increased and from then after it remained almost constant. It can be claimed, therefore, that the presence of FA in PC60LS20FA20 increased the consumption of the Ca(OH)2 content significantly over all ages.  In addition to the thermal analyses carried out at day 7 and day 28, the three selected cementitious matrices (PC60LS40, PC60LS20FA20 and PC43LS20FA37), were reproduced, water cured after day 1 and TG analyses were carried out at day 1, 7, 28 and 56. For PC60LS40, TG analyses were carried out at day 90, as well. At this point, the Ca(OH) 2 consumption will be discussed ( Table 7) in an effort to shed more light into the hydration processes. To begin with, it can be noticed that until day 28, that PC continues hydrating, the Ca(OH) 2 content of PC60LS40 increased and from then after it remained almost constant. It can be claimed, therefore, that the presence of FA in PC60LS20FA20 increased the consumption of the Ca(OH) 2 content significantly over all ages. As demonstrated in Table 7, PC60LS40 at day 1 contained 7.5% Ca(OH) 2 and 9.1% at day 7. For the first two pastes, PC60LS40 and PC60LS20FA20, which have the same PC content, the rate of hydration of PC at day 1 and 7 is expected to be essentially unaltered. However, it can be observed that at early ages, the Ca(OH) 2 content of PC60LS20FA20 is significantly lower, particularly at day 7. This is a strong indication that the pozzolanic activity of the FA present in this paste had already started by day 7, although it has been claimed that at high FA contents (PC45FA55) other researchers had noticed Ca(OH) 2 consumption by day 3 [38]. Nonetheless, FA content is one matter and PC content is another. For this, the authors have adopted the PC/FA ratio, which is more representative in thermal analyses, together with the full formulation dosages, since FA is feeding from the Ca(OH) 2 (Table 7), implying that the remaining unreacted FA was engaged in reactions by day 56. Although no carbonation was detected, even if that was the case, the amount of Ca(OH) 2 at day 56 is that low that any increase due to carbonation would still be negligible.

Microstructural Characterization of Binary and Ternary Cementitious Matrices (B1, B7 and B10)
Backscattered (BSC) images were collected for PC86LS14 (B1), PC60LS20FA20 (B7) and PC60LS40 (B10) and are presented below. As illustrated in Figures 13-16, the microstructure of the PC86LS14, PC60LS20FA20 and PC60LS40 exhibited typical characteristics of such pastes at day 28. A number of CaCO 3 flakes, Ca(CO) 2 (annotated as CH) crystals and nest-like C-S-H (annotated as CSH) are shown in the PC/LS paste. In Figure 13, flake-like C-S-H is mainly detected and some crystals of Ca(CO) 2 covered by C-S-H are also noted. The presence of microcracks is visible as well. In Figure 14A, hydration products surrounding reacted FA (annotated as rFA) particles can be identified. In Figure 14B a common hydrated FA rim known as the "pull-out feature" was captured. It is reported as the "reminiscent" of the fully reacted FA particle [39]. In Figure 14A, hydration products surrounding reacted FA (annotated as rFA) particles can be identified. In Figure 14B a common hydrated FA rim known as the "pull-out feature" was captured. It is reported as the "reminiscent" of the fully reacted FA particle [39].  Figure 14. BSC micrograph of B7-PC60LS20FA20 28D at (A) 3000× magnification, (B) 1000× magnification [24] and (C) 100× magnification [24]. Reproduced with permission from [24], Copyright ICE Publishing, 1987. magnification [24] and (C) 100× magnification [24]. Reproduced with permission from [24], Copyright ICE Publishing, 1987. Unreacted fly ash particles are annotated as uFA. Furthermore, in Figure 14B, an entire plane of Ca(CO)2 crystals can be distinguished above the flat plane of the rest of the surface, creating what looks like a crack, but in essence is the difference in levels between the two planes. In fact, in Figure  14C the entire surface is smooth and furthermore, a number of FA spherical particles can be observed. Formulation PC60LS40 was imaged at day 1 and day 28 due to the high interest it attracted. In Figure 15, different magnifications are offered, all showing a dense, cohesive microstructure lacking microcracks, with a significant number of Ca(CO)2 crystals covered by C-S-H flakes. In Figure 16, a dense microstructure again filled with hydration products can be observed at different magnifications. There are no signs of microcracks, while dense areas of flake-like C-S-H are covering most parts of the micrographs. Similar images have been collected after 90 days of hydration of this specific composition [40].

Step 4: Optimizing Quaternary Cementitious Matrices
The first microstructural enhancement was completed with the addition of two different dosages of μS (5% and 10% by mass) on PC60LS20FA20, creating quaternary cementitious matrices. The results on compressive strength, Ca(OH)2 consumption and microstructure are shown in the In Figure 14A, hydration products surrounding reacted FA (annotated as rFA) particles can be identified. In Figure 14B a common hydrated FA rim known as the "pull-out feature" was captured. It is reported as the "reminiscent" of the fully reacted FA particle [39].
Unreacted fly ash particles are annotated as uFA. Furthermore, in Figure 14B, an entire plane of Ca(CO) 2 crystals can be distinguished above the flat plane of the rest of the surface, creating what looks like a crack, but in essence is the difference in levels between the two planes. In fact, in Figure 14C the entire surface is smooth and furthermore, a number of FA spherical particles can be observed.
Formulation PC60LS40 was imaged at day 1 and day 28 due to the high interest it attracted. In Figure 15, different magnifications are offered, all showing a dense, cohesive microstructure lacking microcracks, with a significant number of Ca(CO) 2 crystals covered by C-S-H flakes. In Figure 16, a dense microstructure again filled with hydration products can be observed at different magnifications. There are no signs of microcracks, while dense areas of flake-like C-S-H are covering most parts of the micrographs. Similar images have been collected after 90 days of hydration of this specific composition [40].

Step 4: Optimizing Quaternary Cementitious Matrices
The first microstructural enhancement was completed with the addition of two different dosages of µS (5% and 10% by mass) on PC60LS20FA20, creating quaternary cementitious matrices. The results on compressive strength, Ca(OH) 2 consumption and microstructure are shown in the following sections. Furthermore, although B8-PC45LS20FA35 showed optimum performance a slightly different paste was adopted for further optimization; PC43LS20FA37. This was decided in an effort to reduce the PC content even further. Hence, another microstructural enhancement was completed with the addition of two different dosages of µS (3% and 6% by mass) on PC43LS20FA37, creating another series of quaternary pastes. The results on compressive strength and Ca(OH) 2 consumption are also shown in the following sections.

Compressive Strength of Quaternary Cementitious Matrices
As depicted in Figure 17 the addition of µS in pastes containing at least 20% FA by mass of binder is causing a retardation of the strength development, delivering lower compressive strengths. More than that, 10% µS, delay the strength gain by 28 days. These results were expected since according to CEN [41] the maximum allowable µS content in PC/µS cements is 10% by mass. In other words, for every 1% µS by mass, at least 9% of PC content by mass, is required so that there will be enough Ca(OH) 2 produced by PC hydration for the µS to react with. Therefore, the 10% µS addition was excessive and as an effect, unreacted µS particles were predicted to be distinguished by SEM investigations. As in the case of μS enhanced PC60LS20FA20, the μS enhanced low PC content PC43LS20FA37, showed similar compressive strength gain characteristics, although the μS content was lower in both combinations ( Figure 18 As in the case of µS enhanced PC60LS20FA20, the µS enhanced low PC content PC43LS20FA37, showed similar compressive strength gain characteristics, although the µS content was lower in both combinations ( Figure 18): • At day 7 both series had similar starting points (about 21 MPa for PC60LS20FA20 and 25 MPa for PC43LS20FA37), indicating that pozzolanic reactions had already been initiated within the first 7 days. • After day 28 the compressive strength of the reference paste is almost the same for both reference pastes, but for the µS enhanced ones, an improvement was observed. Therefore, although high LS contents are reported to be causing deterioration of the compressive strength at later ages the presence of µS offered a limited (due to the additional presence of FA) enhancement.
A marginal improvement in strength was also observed at day 1, followed by deterioration at day 28 and 56 with respect to the reference paste. These results were in agreement with other findings, suggesting that the increasing OPC replacement is liable for the decrease in both compressive and flexural strengths [12]. The limited amount of PC and the high amount of FA present created some antagonism with the µS particles. Moreover, since for 1% µS, at least 9% of PC content by mass, is required so that the µS can react with the Ca(OH) 2 produced by PC hydration for these series of pastes, the µS content should have been limited to approximately 4.8% by mass. Furthermore, knowing that for high replacements of PC with FA and LS the resulting strengths will be low, it was of particular interest to investigate if the nanoparticles could alter the performance of the low PC content blended cements, carried out in other research [42].

Embodied CO2 of Quaternary Cementitious Matrices
In addition to the E.CO2 values of PC, LS and FA pastes, the E.CO2 of μS modified pastes was also evaluated (taking as μS E.CO2 = 14 kg CO2/tonne according to [43]. The results are presented in Table 8, where the second quaternary series based on PC43LS20FA37 had the lowest total E.CO2. Furthermore, PC43LS20FA37 exhibited the highest mechanical properties in terms of the ratio of compressive strength/E.CO2. Lastly, for such pastes the lower μS content seems to be delivering the most optimal results.

Embodied CO 2 of Quaternary Cementitious Matrices
In addition to the E.CO 2 values of PC, LS and FA pastes, the E.CO 2 of µS modified pastes was also evaluated (taking as µS E.CO 2 = 14 kg CO 2 /tonne according to [43]. The results are presented in Table 8, where the second quaternary series based on PC43LS20FA37 had the lowest total E.CO 2 . Furthermore, PC43LS20FA37 exhibited the highest mechanical properties in terms of the ratio of compressive strength/E.CO 2 . Lastly, for such pastes the lower µS content seems to be delivering the most optimal results.

Thermal Analyses of Quaternary Cementitious Matrices
As shown in Figure 19, given that the reference paste contained at the various ages approximately 8% Ca(OH) 2 , a paste containing 55%PC would be expected to generate 7.3% Ca(OH) 2 and a paste containing 50%PC would be expected to generate approximately 6.6% Ca(OH) 2 Therefore, it can be argued that in all cases the µS addition cause a further Ca(OH) 2 consumption. Lastly, as shown in Figure 20 no carbonation has taken place as the CaCO 3 content is lower than the 20% LS content present in the pastes in all combinations.  The Ca(OH)2 consumption is depicted in Figure 21. It is interesting to note that after day 28 and by day 56, the Ca(OH)2 content in the reference paste reduced to 1.5% by mass. It has been reported that for high PC replacement by FA (> 60%), depletion of Ca(OH)2 may be observed after one year [16,17]. Therefore, given more time, Ca(OH)2 in this ternary system is expected to be depleted, if the sample has not carbonated. However, the phenomena observed in composite cements can be more complex. In fact, it has been reported that combinations containing 65% OPC and 35% of a formulation of FA and LS produced similar amounts of Ca(OH)2 after day 1. However, the minimum  The Ca(OH)2 consumption is depicted in Figure 21. It is interesting to note that after day 28 and by day 56, the Ca(OH)2 content in the reference paste reduced to 1.5% by mass. It has been reported that for high PC replacement by FA (> 60%), depletion of Ca(OH)2 may be observed after one year [16,17]. Therefore, given more time, Ca(OH)2 in this ternary system is expected to be depleted, if the sample has not carbonated. However, the phenomena observed in composite cements can be more complex. In fact, it has been reported that combinations containing 65% OPC and 35% of a formulation of FA and LS produced similar amounts of Ca(OH)2 after day 1. However, the minimum amount of Ca(OH)2 was reached for the 30% FA and 5% LS formulation and not the 35% FA The Ca(OH) 2 consumption is depicted in Figure 21. It is interesting to note that after day 28 and by day 56, the Ca(OH) 2 content in the reference paste reduced to 1.5% by mass. It has been reported that for high PC replacement by FA (> 60%), depletion of Ca(OH) 2 may be observed after one year [16,17]. Therefore, given more time, Ca(OH) 2 in this ternary system is expected to be depleted, if the sample has not carbonated. However, the phenomena observed in composite cements can be more complex.
In fact, it has been reported that combinations containing 65% OPC and 35% of a formulation of FA and LS produced similar amounts of Ca(OH) 2 after day 1. However, the minimum amount of Ca(OH) 2 was reached for the 30% FA and 5% LS formulation and not the 35% FA formulation [12]. This was attributed to the synergistic interaction of FA and LS. Although according to De Weerdt et al. [12] the presence of LS leads to the formation of carboaluminate hydrates, this was not detected by the TG analyses (Figure 22), neither by the XRD analyses shown in detail in other publication [25].
that for high PC replacement by FA (> 60%), depletion of Ca(OH)2 may be observed after one year [16,17]. Therefore, given more time, Ca(OH)2 in this ternary system is expected to be depleted, if the sample has not carbonated. However, the phenomena observed in composite cements can be more complex. In fact, it has been reported that combinations containing 65% OPC and 35% of a formulation of FA and LS produced similar amounts of Ca(OH)2 after day 1. However, the minimum amount of Ca(OH)2 was reached for the 30% FA and 5% LS formulation and not the 35% FA formulation [12]. This was attributed to the synergistic interaction of FA and LS. Although according to De Weerdt et al. [12] the presence of LS leads to the formation of carboaluminate hydrates, this was not detected by the TG analyses (Figure 22), neither by the XRD analyses shown in detail in other publication [25].

Microstructural Characterization of Quaternary Cementitious Matrices
Indicatively, four BSC micrographs of PC55LS20FA20+5%μS are shown in Figure 23. As discussed earlier in Portland limestone cement pastes with high FA contents except for the needlelike hydration products (figures 23 C,D), CaCO3 flakes and Ca(OH)2 crystals a number of rims and reacted FA particles can be detected. It is interesting to note that the investigation of different areas did not yield indications of unreacted FA particles. A fact correlating well with the TG measurements showing that there was enough Ca(OH)2 still present in the hydrated paste after day 28 and 56 should the FA have needed it to react with. Reacted FA (rFA) particles are circled in Figure 23B. Furthermore, in Figure 23B the reaction of a FA particle leaving a characteristic mullite sphere [39] was identified (enclosed in the rectangle). It was suggested by Xu and Sarkar that the glass phase in FA has been reacted and the crystalline phase is left almost intact. Adding to this, the typically acicular hydration product of FA in the first 28 days can be seen in Figure 23B (enclosed in a rectangle). Lastly, the presence of microcracks is prominent, possibly signaling cracking due to elevated heat of hydration. Therefore, this specific formulation could be enhanced by reducing the silica fume content.

Microstructural Characterization of Quaternary Cementitious Matrices
Indicatively, four BSC micrographs of PC55LS20FA20+5%µS are shown in Figure 23. As discussed earlier in Portland limestone cement pastes with high FA contents except for the needle-like hydration products (Figure 23 C,D), CaCO 3 flakes and Ca(OH) 2 crystals a number of rims and reacted FA particles can be detected. It is interesting to note that the investigation of different areas did not yield indications of unreacted FA particles. A fact correlating well with the TG measurements showing that there was enough Ca(OH) 2 still present in the hydrated paste after day 28 and 56 should the FA have needed it to react with. Reacted FA (rFA) particles are circled in Figure 23B. Furthermore, in Figure 23B the reaction of a FA particle leaving a characteristic mullite sphere [39] was identified (enclosed in the rectangle). It was suggested by Xu and Sarkar that the glass phase in FA has been reacted and the crystalline phase is left almost intact. Adding to this, the typically acicular hydration product of FA in the first 28 days can be seen in Figure 23B (enclosed in a rectangle). Lastly, the presence of microcracks is prominent, possibly signaling cracking due to elevated heat of hydration. Therefore, this specific formulation could be enhanced by reducing the silica fume content.
Comparing the reference paste PC60LS20FA20 (Figure 14) with the 5% µS ( Figure 23) and the 10% µS ( Figure 24) enhanced quaternary formulations and particularly the 200 times magnification there is an indication that the µS enhanced pastes are less dense. This cannot be attributed to particle packing that is expected to be enhanced, but credited to the high amount of FA present. The FA was already feeding from the Ca(OH) 2 , therefore the more reactive pozzolanic particles of µS may not be able to access the Ca(OH) 2 crystals encapsulated within the reacted FA particles. That is to say, although Ca(OH) 2 is not depleted in these quaternary formulations, the µS particles may be deactivated by the inaccessibility of Ca(OH) 2 created by the hydration products around the much greater quantities of FA. As depicted in Figure 24, a similar microstructure to the lower μS content paste can be observed. The only difference being the plethora of needle-like hydration products detected at all magnifications and the many unreacted μS particles (small, spherical bright elements). The typically acicular hydration product of FA in the first 28 days can be seen in Figure 24B (enclosed in a rectangle). Another FA particle with hydration gel around it, is circled in Figure 23C. Reacted FA particles are circled in Figure 24C. In agreement with Lothenbach et al. [16], the μS enhanced pastes showed a more refined pore structure, but are expected to deliver higher total porosities. That is to say, capillary pores were expected to be reduced and finer pores slightly increased. This could partly explain the microcracks present in the images taken. The increased heat of hydration could be the other reason for the development of microcracks as discussed for the +5%μS addition. Knowing that the presence of the cracks is related to lower durability of the formulations, future research on quaternary pastes using this formulation should contain substantially lower μS contents. In fact, this observations led to the reduction of the μS contents in the design of the quinary combinations, based on PC43LS20FA37. As shown in published research four different formulations competed against the reference paste PC43LS20FA37; (a) PC43LS17FA37 + 2.5%μS + 0.5%nS (where nS is the abbreviation for nanoparticles of SiO2), (b) PC43LS17FA37 + 2.0%μS + 1.0%nS, (c) PC43LS14FA37 + 5.5%μS + 0.5%nS and (d) PC43LS14FA37 + 5.0%μS + 1.0%nS [42]. The lower additions of nS particles performed better, however, the presence of high quantities of pozzolanic and highly pozzolanic materials, delayed hydration reactions and strength gain. In other words, the nanoparticles were unable to leverage the already expected delayed reactions due to the presence of high volume fly ash and silica fume content. As depicted in Figure 24, a similar microstructure to the lower µS content paste can be observed. The only difference being the plethora of needle-like hydration products detected at all magnifications and the many unreacted µS particles (small, spherical bright elements). The typically acicular hydration product of FA in the first 28 days can be seen in Figure 24B (enclosed in a rectangle). Another FA particle with hydration gel around it, is circled in Figure 23C. Reacted FA particles are circled in Figure 24C. In agreement with Lothenbach et al. [16], the µS enhanced pastes showed a more refined pore structure, but are expected to deliver higher total porosities. That is to say, capillary pores were expected to be reduced and finer pores slightly increased. This could partly explain the microcracks present in the images taken. The increased heat of hydration could be the other reason for the development of microcracks as discussed for the +5%µS addition. Knowing that the presence of the cracks is related to lower durability of the formulations, future research on quaternary pastes using this formulation should contain substantially lower µS contents. In fact, this observations led to the reduction of the µS contents in the design of the quinary combinations, based on PC43LS20FA37. As shown in published research four different formulations competed against the reference paste PC43LS20FA37; (a) PC43LS17FA37 + 2.5%µS + 0.5%nS (where nS is the abbreviation for nanoparticles of SiO 2 ), (b) PC43LS17FA37 + 2.0%µS + 1.0%nS, (c) PC43LS14FA37 + 5.5%µS + 0.5%nS and (d) PC43LS14FA37 + 5.0%µS + 1.0%nS [42]. The lower additions of nS particles performed better, however, the presence of high quantities of pozzolanic and highly pozzolanic materials, delayed hydration reactions and strength gain. In other words, the nanoparticles were unable to leverage the already expected delayed reactions due to the presence of high volume fly ash and silica fume content. The BSC micrograph of PC43LS20FA37 at day 28 is presented below (Figure 25). A number of reacted and unreacted particles, can be observed as expected according to the TG analyses which lead to the conclusion that at this age almost half of the FA particles were still unreacted), CaCO3 flakes and Ca(OH)2 crystals were also present and areas of dense matter, believed to be C-S-H. Similar description is valid for curing until day 1 and day 56 [42]. No cracks can be observed, rendering the specific formulation promising for further optimization.

Conclusions
The present paper is the last part of an extensive research program carried out at the University of Bath, aimed at investigating the effect of nanoparticles in sustainable blended cements. Although a significant number of papers have been published on the effect of various nanoparticles on low carbon footprint composite cementitious formulations [ The BSC micrograph of PC43LS20FA37 at day 28 is presented below (Figure 25). A number of reacted and unreacted particles, can be observed as expected according to the TG analyses which lead to the conclusion that at this age almost half of the FA particles were still unreacted), CaCO 3 flakes and Ca(OH) 2 crystals were also present and areas of dense matter, believed to be C-S-H. Similar description is valid for curing until day 1 and day 56 [42]. No cracks can be observed, rendering the specific formulation promising for further optimization. The BSC micrograph of PC43LS20FA37 at day 28 is presented below (Figure 25). A number of reacted and unreacted particles, can be observed as expected according to the TG analyses which lead to the conclusion that at this age almost half of the FA particles were still unreacted), CaCO3 flakes and Ca(OH)2 crystals were also present and areas of dense matter, believed to be C-S-H. Similar description is valid for curing until day 1 and day 56 [42]. No cracks can be observed, rendering the specific formulation promising for further optimization.

Conclusions
The present paper is the last part of an extensive research program carried out at the University of Bath, aimed at investigating the effect of nanoparticles in sustainable blended cements. Although a significant number of papers have been published on the effect of various nanoparticles on low carbon footprint composite cementitious formulations [

Conclusions
The present paper is the last part of an extensive research program carried out at the University of Bath, aimed at investigating the effect of nanoparticles in sustainable blended cements. Although a significant number of papers have been published on the effect of various nanoparticles on low carbon footprint composite cementitious formulations [5,8,10,25,[27][28][29][30]40,42,[44][45][46][47], the starting pastes and the methodology of selecting formulation on performance based criteria, has never been discussed before. The paper has introduced, analyzed, compared and discussed, a suite of binary (PC + LS) and ternary (PC + FA + LS) cementitious matrices. Three ternary pastes were selected which were further optimized with the addition of silica fume. Competing constituents and areas of improvements to pave the way for the addition of the nanoparticles were identified. Moreover, the pozzolanic reactivity in ternary and quaternary cement paste formulations was investigated, while the lowest possible Portland cement content was employed. In addition to the above, a number of further conclusions were made: • Current research can expand on challenging formulations, which allow for significant reduction in CO 2 emissions, beyond allowable limits by current codes.

•
It was demonstrated that optimizing blended or composite Portland cement pastes at high PC replacement levels is an iterative process. This step by step methodology presented in the paper can serve as a basis for other researchers in the field.

•
In areas where pozzolanic materials are scarce, binary formulations can be developed as the control mix, further enhanced by highly reactive materials, such as nanoparticles of SiO 2 or montmorillonite [10,29,44,47].

•
In areas where fly ash is abundant and needs to be valorized, ternary blends can be developed and again can be further enhanced with nanomaterials [22,25,42]. Similarly, quinary cementitious matrices can be developed with both micro-and nanosized silica [42] for superior strength and denser microstructure.

•
It is essential that material characterization is thorough and extensive testing is necessary.

•
The complicated interaction between the constituents was investigated with respect to strength development, changes in the hydration reaction and microstructural enhancement.
Lastly, adding to the future research suggestions, apart from the addition of different types of nanoparticles for strength and durability improvements, which has been previously mentioned, matrices containing different types of supplementary cementitious materials can be generated. Some options could include ground granulated blast furnace slag or geopolymers altering water demand, strength and microstructural characteristics. This research has rendered the surpassing of the boundaries defined by the current codes of practice, possible and therefore, further optimizations of blended pastes can be expected in the near future.
Author Contributions: Conceptualization, S.P. and K.P.; methodology, S.P.; validation, S.P. and K.P.; formal analysis, S.P. investigation, S.P.; data curation, S.P.; writing-original draft preparation, S.P.; writing-review and editing, S.P. and K.P.; funding acquisition, K.P. Both authors have read and agreed to the published version of the manuscript. All authors have read and agreed to the published version of the manuscript.