Sinopharm Chemical Reagent Co., Ltd, 99.0% pure laboratory reagents with a fineness of at least 74 μm were used throughout: calcium carbonate (CaCO₃), calcium phosphate (Ca₃(PO₄)₂), alumina (Al₂O₃), silica (SiO₂), ferric oxide (Fe₂O₃) and alcohol. All the materials were precisely weighed to the proportions specified in the phosphoaluminate cement design (12) and 0, 10, 11, 12, 13 or 15% Fe₂O₃ was added to prepare specimens respectively labelled A, B, C, D, E and F. Sample G had the same composition as sample A but was sintered at a higher temperature.

The components were thoroughly blended, mixed with laboratory grade alcohol and then dried at 105 °C for 4 hours. They were subsequently pressed into round specimens measuring Φ60 mm×10 mm, sintered at 1380 °C for 2 hours (ramping the temperature at a rate of 5 °C/minute) and cooled under fast-flowing forced air. Specimen G was sintered separately at 1560 °C for 2 h, ramping at the same rate as above and fan-cooled. The seven clinkers, including specimen G, were ground to pass the No. 200 sieve. The powder was mixed with water and the resulting paste was poured into 20×20×20‑mm³ moulds for curing at 20 °C and 90% relative humidity for 1 day. After removal from the moulds the pastes were stored in water for 1, 3, 7, 28 or 90 days, at which times they were tested for compressive strength. The fragments were then immersed in alcohol to detain hydration and vacuum dried at 30 °C for further analysis.

Clinker fineness was determined by negative pressure sieving as per Chinese standard GB/T1345-1991. The unsieved residue was held within 0.5–3 wt%. Compressive strength was determined on a 50-kN MTS CMT5504 test frame (China). XRD patterns were recorded on a Bruker D8 Advance diffractometer (Germany) fitted with a Cu Kα X-ray tube and operating a 40 kV and 40 mA. Readings were taken between 2θ angles of 5 to 60° with a step size of 0.02° and a scan speed of 0.2 s. SEM analyses were conducted on a FEI QUANTA FEG (USA) scanning electron microscope operating at 20 kV and 20 mA. The EDS findings were obtained on an Oxford Instruments INCA energy X-MAX-50X analyser (UK). A TAM Air eight-channel, thermometric isothermal conduction calorimeter was used to determine heat of hydration and heat flow. DSC data were logged with a TGA/DSC1/1600HT analyser from ambient temperature to 650 °C, ramping at a rate of 10 °C/min.

The XRD patterns for anhydrous phosphoaluminate cement with varying Fe₂O₃ contents sintered at 1380 °C and for two reference specimens without the oxide sintered at 1380 and 1560 °C are reproduced in Figure 1. As the figure shows, the main phases identified in specimen A (0% Fe₂O₃) included modified CA_1-Y(P_Y), C₁₂A₇ and CP_1-z(A_z), while only traces of LHss were detected. In contrast, large quantities of LHss (lines at 23.751°, 33.858°, 41.756°) were present in specimen G (likewise with 0% Fe₂O₃), whose diffractogram showed no signals for CA_1-Y(P_Y), C₁₂A₇ or CP_1-z(A_z). In other words, LHss was observed to form at high temperatures even without Fe₂O₃.

In specimens B to E, with Fe₂O₃ ranging from 10 to 13%, the main phases were CA_1-Y(P_Y) and CP_1-z(A_z). In these patterns, the LHss signal was more intense than on the pattern for specimen A, while the line for C₁₂A₇ disappeared and a new phase, C_AAF, formed. According to these findings, Fe₂O₃ would favour the reaction between C₁₂A₇ and phosphorus oxide, yielding mineral LHss at the lower temperature, while C₄AF would be the product of the reaction between Fe₂O₃ and C₁₂A₇. The intensity of the LHss diffraction line rose while the signals for C₄AF, CA_1-Y(P_Y) and CP_1-z(A_z) weakened with rising Fe₂O₃ content. Therefore, like high temperature, moderate percentages of Fe₂O₃ favoured LHss formation from CA_1-Y(P_Y), CP_1-z(A_z) and C₄AF.

When the Fe₂O₃ dose was increased to 15%, however, as in specimen F, LHss formation declined. The possible explanation lies in Fe₂O₃'s role as intermediate network oxide in partially molten clinker. At low percentages, it would act as a network modifier oxide, lowering the viscosity of the molten phase and favouring LHss formation. At an overly high content, however, Fe₂O₃ would act as a network former, hindering ion migration and inhibiting crystal precipitation (19).

The SEM and EDS element maps for specimen E (13% Fe₂O₃), taken as an example of mineral morphology and Fe₂O₃ distribution in the cement, are shown in Figure 2. Al, P and Ca overlapped in the regularly shaped α particles, which consequently consisted of calcium phosphoaluminate or modified monocalcium aluminate. The P/Ca overlap in the β particles was evidence that they consisted of calcium phosphate. The Ca content was observed to be higher in calcium phosphate than in calcium phosphoaluminate and modified monocalcium aluminate. Si was distributed evenly across all the minerals.

As Figure 2(f) shows, the iron phase was found in interstitial positions between the crystals, where it acted as a mineraliser during clinker sintering, confirming the XRD findings discussed above. The EDS-determined ferrite composition at seven points on the iron phase (20, 21) are given in Table 1. The respective chemical formulas, calculated from the EDS data, are listed in the last column of the table. These findings revealed substantial variation in ferrite composition, which in this phosphoaluminate cement comprised a series of solid solutions comparable to the solutions in portland and calcium aluminate cements (22–24).

The compressive strength values of the pastes at different ages are listed in Table 2. Figure 1 showed that the main minerals present in A were modified monocalcium aluminate (CA_1-Y(P_Y)) and dodecacalcium hepta-aluminate (C₁₂A₇). The former would clearly afford the paste early age strength, which rose to day 7 and remained constant thereafter. In the XRD pattern for paste G, which unlike the other cements was sintered at 1560 °C, the diffraction line for LHss predominated over all the other minerals present.

The data in Table 2 show that while early age strength was much lower in paste G than in paste A, the 28- and 90‑day strength values were higher in the former. Paste C, which had higher early age strength than specimen E, according to Figure 1, contained more CA_1-Y(P_Y) and less LHss than specimen E. The higher rate of strength development in E between 3 and 90 days would indicate that LHss enhances cement strength at later ages. Early age compressive strength declined significantly with rising percentages of Fe₂O₃, while later age strength rose visibly. The conclusion that may be drawn is that Fe₂O₃ favoured the conversion of CA_1-Y(P_Y) to LHss, thereby raising later age compressive strength. This reasoning is consistent with the XRD findings. Hence, the compressive strength of phosphoaluminate cement paste may be modified by controlling the Fe₂O₃ dosage used.

The XRD patterns for the 1-, 3-, 7-, 28- and 90‑day A, C and E cement pastes are reproduced in Figure 3. The signal for CA_1-Y(P_Y) was weak after just 1 day, an indication that it was largely consumed. In contrast, relatively intense diffraction lines for this mineral on the patterns for 3‑day paste A showed that the addition of Fe₂O₃ favoured CA_1-Y(P_Y) hydration. As CA_1-Y(P_Y) hydrated, the intensity of the signals at 2θ angles of 8.460° and 16.941° for C₂(A_1-X-YP_XSi_Y)H₈ grew. These values were shifted from the 8.256° and 16.525° recorded for C₂AH₈ due to the replacement of Al by P and Si. The line for C₂(A_1-X-YP_XSi_Y)H₈ practically disappeared in the 28‑day pattern for paste C, however, with its conversion into C(A_1-X-YP_XSi_Y)H_n at later ages (25). These findings were confirmed by the DSC analysis of paste C shown in Figure 4.

The intensity of the signals generated by C(A_1-X-YP_XSi_Y)H_n (6.201°,12.299°) rose with curing age. Since CA_1-Y(P_Y) was almost entirely consumed in the first day, the hydration product detected, C(A_1-X-YP_XSi_Y)H_n, must have been the result of LHss hydration. Unlike CAH₁₀ (6.219°,12.352°), which is unstable in high alumina cement, C(A_1-XP_XSi_Y)H_n was stable due to the replacement of Al by P and Si (25). The diffraction lines associated with LHss declined in intensity after 90 days, when more C(A_1-XP_XSi_Y)H_n was found in the system. As the solid solution of P and Si in LHss prevented C₂AH₈ and CAH₁₀ from converting to C₃AH₆, compressive strength rose continuously in these specimens. Figure 5 shows the 1‑day SEM micrograph and EDS analysis for specimen C, in which flaky hydration products were observed. EDS identified points 1 and 2 as C₂(A_1-X-YP_XSi_Y)H₈, a finding consistent with the XRD results.

The heat flow curves plotted are shown in Figure 6. The figure shows that hydration peaked in paste A at 16.19 mW/g after 7.86 h and in paste C at 10.55 mW/g after 9.19 h. Both peaks were associated with swift CA_1-Y(P_Y) hydration. The paste E hydration peak was recorded at 7.03 mW/g after 17.86 h. Unlike paste C, paste E exhibited no significant heat peak in the first 10 hours. The reason was that the Fe₂O₃ added had already induced CA_1-Y(P_Y) conversion to LHss.