PI 42.5 cement, mineral powder, and silica fume were adopted as raw materials, as shown in Table 1. The average particle size of PI 42.5 cement, mineral powder, and silica fume is 36.96, 20.30, and 0.22 μm, respectively. The molecular structure of commercial Sika high-performance polycarboxylate is shown in Figure 1.

Isophorone diisocyanate, hydroxyl-terminated silicone, polyether glycol, N-methylpyrrolidone, 1,4-butanediol, dimethylolpropionic acid, and prenyl alcohol polyoxyethylene ether (TPEG) were used for synthesis; the pure acrylic acid (AA), thioglycolic acid, p-toluenesulfonic acid, polyethylene glycol-200 (PEG-200), ascorbic acid (Vc), hydrogen peroxide, dibutyltin dilaurate, ammonium persulfate, and deionized water were analyzed.

During the synthesis, alcohol and amine compounds and multi-functional groups were introduced into the main chain of polycarboxylate molecules, so as to bring about more short branches, promote the alternating distribution of long polyether branches and short branches of chain extenders of alcohol and amine compounds, and increase the dispersibility and adaptability of the superplasticizer.

22.2 g isophorone diisocyanate was placed in a three-necked flask, polyethylene glycol 1000 containing dibutyltin dilaurate (N-methylpyrrolidone solution with 0.2 g dibutyltin dilaurate), and dimethylolpropionic acid in 25 g polyethylene glycol 1000 (3.35 g dimethylolpropionic acid dissolved in 5 ml N-methylpyrrolidone) was added dropwise in turn with a dropping funnel, and then 50 g water was added. The mixture was heated to 70°C and kept warm. Then, the mixed solution of hydroxyl-terminated polysiloxane and 1,4-butanediol (9 g of hydroxyl-terminated polysiloxane mixed with 3 g of 1,4-butanediol), 2.5 g of ethylenediamine, and 1.5 g of sodium dodecylbenzenesulfonate were added dropwise while stirring before feeding. When the free-NCO content in the mixed solution was detected to be 16%, heating was stopped to obtain the polyurethane prepolymer. The principle of side chain reaction was shown in Figure 2.

To ensure the equal molar concentration of double bonds in the system, the concentration of free radicals and molecular weight regulators was controlled, and the molar ratio of acrylic acid to TPEG was adjusted. TPEG (120 g), purified water (80 ml), and PEG-200 (10 g) were put into a 250 ml four-necked flask. After fully stirring the mixed solution, 0.1 g dibutyltin dilaurate, 1.2 g of p-toluenesulfonic acid, and 30% hydrogen peroxide were added. A and B solutions were prepared for later use.

Component A (acrylic acid + sodium bisulfite, Vc, or sodium formaldehyde sulfoxylate).

Component B (thioglycolic acid + ammonium persulfate).

A peristaltic pump was used to drip components A and B into the beaker at the rate of 2 ml/min, which lasted for 2.5 h. After that, the mixed solution was kept warm for a period of time, and the pH value was adjusted to 6–7 with NaOH to obtain mother liquor of polycarboxylate (M-PCE). The structure of synthesized M-PCE was shown in Figure 3.

In view of GB/T8077-2012 Test Method for Homogeneity of Concrete Admixtures, the water reduction rate of polycarboxylate mortar is tested using the following formula: Mortar water reduction rate = M 0 − M 1 M 0 × 100 % . (1) Here, M ₀ is the water consumption of benchmark mortar flowability (mm) and M ₁ is the water consumption when the flowability of mortar with admixture reaches 180±5 mm.

The content of polycarboxylate is 2% (mass ratio) of cementitious components, the water-to-binder ratio of ultra-high performance concrete is 0.18, and the sand-to-binder ratio is 1.0. Table 2 shows the mixture ratios.

According to the standard GB/T2419-2005 “Method for Measuring Fluidity of Cement Mortar,” fine sand below 2.36 mm was selected, the mass of cementitious material was 200 g, the sand-to-cement ratio was 1:1, the water-to-binder ratio was 0.18, and the polycarboxylate mother liquor content was 2%, then, the flowability of mortar was tested.

In the free shrinkage test, reference cement and standard sand were used, with the ratio of sand to cement of 1:1. The flowability of cement mortar was higher than 140 mm, and a set of 40 × 40 × 160 mm specimens was reported. The test molds were triple molds and were filled and maintained for 24 h. The temperature of the curing box was controlled to be 20°C ± 1°C, and the relative humidity was equal to or higher than 90%. Then, the mold was transferred to the steam curing room. The temperature in the drying room was 20 ± 3°C. Then, the strength at the 1st, the 3rd, the 7th, the 14th, and the 28th day was tested.

In view of JC/T603-2004 “Test Method for Drying Shrinkage of Cement Mortar,” a group of 25 × 25 × 280 mm specimens should be formed in the drying shrinkage test of cement mortar, and the specimens should be demolded after being cured for 24 ± 1 h from the time of adding water. Then, the length of the test at the 1st day, 3rd day, 7th day, 14th day, 21st day, 28th day, 56th day, and 90th day was measured. The result was calculated according to the following formula, accurate to 0.001%: S n = ( L 0 − L 28 ) × 100 280 , (2) where Sn is the drying shrinkage rate of the cement mortar specimen at the nth day (%); L₀ is the initial measurement reading (mm); L₂₈ is the measured reading at the 28th day (mm); and 280 is the effective length of the specimen (mm).

The auto-shrinkage of mortar in self-compacting cement mortar was measured with the combination of a bellow and a non-contact probe. The bellow has an inner diameter of 20 mm and a length of 340 ± 5 mm. The cement mortar flowability was higher than 140 mm, and the auto-shrinkage of cement mortar was continuously measured for 72 h.

The water consumption of benchmark mortar without superplasticizer was 93.7 g, which was 186 and 175 mm with an average of about 180 mm after rest. The experimental results are shown in Table 3.

The water reduction rate of mortar of M-PCE was similar to that of Sika. In terms of the molecular structure, the dispersing group of Sika was mainly a long-chain alkyl macromonomer. However, the dispersing ability of M-PCE was enhanced by the strong chemical bond of silicone groups in M-PCE and silicate phase (Jones et al., 1995). Due to its chemical structure, M-PCE has longer side chains than PCE, indicating good steric hindrance (Levi et al., 2002). Therefore, the dispersibility of M-PCE in cement is also excellent, which is consistent with the results of Kong et al. (Thompson et al., 1997) Silicone chains provide spatial repulsion between cement particles in some way, exerting dispersing effect on cement slurry. The anchoring ability of silicone groups has been confirmed by the enhanced adsorption behavior of silane-modified polycarboxylate in cement slurry (Kagi and Ren, 1995).

The flowability of UHPC under the cement mortar–silica fume–mineral powder gelatinization system is shown in Table 4.

According to Table 4, the flowability of M-PCE was better than that of Sika under the same conditions, indicating that M-PCE had better dispersion than Sika. In addition, the flowability of ultra-high performance concrete increased from 119 to 172 mm as silica fume content increased from 0 to 30%. However, when the amount of silica fume was too large, the flowability of concrete decreased. A large amount of superplasticizer will be adsorbed on the surface of silica fume with small particle size (Ganesh and Surya, 1995; Khatri et al., 1995). Under the action of superplasticizer, silica fume, as ultra-fine particles, disperses cement particles and releases more free water (Duval and Kadri, 1998). For ordinary concrete, the dosage of silica fume is lower than 10%. For high-strength concrete, the dosage of silica fume should be increased to 10% (Khayat et al., 1993). UHPC silica fume consumption is usually lower than 15% (Mazloom et al., 2004). For UHPC studied in this paper, if the amount of silica fume is over 30%, it will make concrete extremely viscous, reducing the flowability of concrete.

The specific surface area of mineral powder is larger than that of cement. Therefore, mineral powder significantly affects the flowability of UHPC. When the amount of mineral powder exceeds 40%, the flowability of UHPC is less than 140 mm.

Under the cement mortar–silica fume–mineral powder gelatinization system, M-PCE and Sika polycarboxylate were, respectively, used to prepare UHPC.

As can be seen from Table 4, UHPC prepared with the two different superplasticizers had little difference in compressive strength. Table 5 shows that, with the increase of silica fume dosage, the early compressive strength of UHPC increased. After the amount of silica fume exceeds 25%, the increment of strength slows down. When the amount of silica fume increased from 0 to 15%, the 28-day strength of concrete rose from 92.5 to 117.4 MPa. Therefore, silica fume could improve the compressive strength of ultra-high strength concrete. It is worth noting that, at the dosage of silica fume lower than 15%, the 28-day compressive strength of UHPC increased with the dosage of silica fume. Table 6 also shows the same trend. In addition, we found that the addition of mineral powder would reduce the 3-day compressive strength of UHPC. Especially when the content of mineral powder was more than 25%, the early strength of concrete decreased from 71.0 to 63.6 MPa. At the 28th day, when the dosage of mineral powder increased from 10 to 20%, the compressive strength of UHPC reached the maximum value, and the strength reached 103.6 MPa. It is not difficult to see whether adding mineral powder alone would significantly improve the compressive strength of UHPC. However, when mixed with silica fume, slag would enhance the compressive strength of UHPC. According to previous studies, to improve the strength of concrete, the best constitution of cementitious materials is mainly associated with the calcium-to-silicon ratio (Erdem and Kırca, 2008; Elahi et al., 2010). When the calcium-to-silicon ratio is 1.30, the strength of UHPC is greater than that of silica fume (Taylor, 1997).

Under the cement mortar–silica fume–mineral powder gelatinization system, M-PCE and Sika polycarboxylate, respectively, were adopted to prepare UHPC. The drying shrinkage changes of UHPC are shown in Figures 4, 5.

As shown in Figures 4, 5, UHPC prepared with M-PCE had slightly higher drying shrinkage as compared with UHPC prepared with Sika. Sika as a compound product has stronger capacity of reducing the surface tension of pore solution, so UHPC prepared with Sika will have a smaller drying shrinkage value.

Figures 4, 5 show that the drying shrinkage value of UHPC was the smallest when the dose of mineral powder was about 50%. In addition, the drying shrinkage of UHPC increased with the addition of silica fume (N2 and N5 curves) because silica fume has high activity and can react with the hydration product Ca(OH)₂ to form CSH gel (Yazıcı, 2007). Theoretically, there is a positive correlation between the degree of pozzolanic reaction and the dose of silica fume. As a consequence, the dose of silica fume will increase the shrinkage value of UHPC (Zhang et al., 2003; Hong and Glasser, 2004).

M-PCE and Sika polycarboxylates were, respectively, used to prepare UHPC. The auto-shrinkage changes of UHPC are shown in Figures 6, 7.

According to Figures 6, 7, under the same conditions, the influence of M-PCE on the auto-shrinkage of UHPC was similar to that of Sika. The auto-shrinkage of UHPC prepared with M-PCE or Sika is relatively smaller than that of UHPC prepared with pure Portland cement.

In addition, it was observed that the dosage of silica fume decreased the auto-shrinkage of UHPC (as shown in Figures 6, 7). In Figure 6, the content of silica fume increases from 15 to 30% and the auto-shrinkage ranges from 3,800 to 5,400. In contrast, in Figure 7, this value ranges from 3,700 to 5,500, which is consistent with reference (Duan et al., 2020), mainly due to the fact that the silica fume has large fineness, exerting special filling on the pore structure of cement-based materials. With the increasing dosage of silica fume, the capillary diameter will be reduced, causing greater capillary pressure and greater auto-shrinkage of UHPC (Rong et al., 2014).

In Figure 6, the content of mineral powder increases from 25 to 50% and its auto-shrinkage ranges from 4,200 to 6,800. In Figure 7, the auto-shrinkage ranges from 4,400 to 6,600. From the point of view of mineral powder, the auto-shrinkage of UHPC is also greatly affected by single addition. The pozzolanic activity and hydration degree of mineral powder were higher than those of Portland cement, leading to accelerated speed of water consumption and faster internal drying process (He et al., 2019).