The effect of boron oxide on the microstructure and hydration of calcium sulfoaluminate phase

Boron-rich waste causes numerous environmental problems when discharged directly into the environment. Here, various quantities of boron oxide (B2O3) were added to calcium sulfoaluminate (C4A3$) during the sintering process to demonstrate a potential use of boron-rich waste. The microstructure and hydration performance of C4A3$ with various B2O3 contents were investigated with scanning electron microscopy, x-ray diffraction, isothermal conduction calorimetry, thermogravimetric studies and compressive strength tests. B2O3-doped C4A3$ had a larger grain size than the pure phase; and were surrounded by amorphous phases. The presence of B2O3 was shown to promote the phase transition process through which C4A3$ changes from the orthorhombic to the cubic structure; and the substitution of Al3+ for B3+ in AlO4 tetrahedra was surveyed by structural refinements. As the B2O3 content increased, the induction period of C4A3$ increased while the hydration rate decreased because of the amorphous phases around the C4A3$. However, the hydration degree of doped C4A3$ increased due to the slower reaction rate. Thus, when an appropriate amount of B2O3 was added to the C4A3$ during sintering, a significant improvement in the compressive strength of pastes was observed.


Introduction
Boron ores are used in the production of industrially-important compounds such as boric acid, boron oxide (B 2 O 3 ) and borax. Various types of boron-rich wastes such as clay wastes and radioactive wastes are continuously produced and require careful treatment [1][2][3]. The most widely used stabilization process for stabilizing boron-rich waste is cementation, where the waste is added as a retardation additive to ordinary Portland cement (OPC), or to a lesser extent, calcium sulfoaluminate cement (CSA) [4][5][6]. Boron-rich waste retards cementation by forming a protective coating of amorphous or poorly-crystallized calcium borate over individual cement grains and thus inhibits the hydration process [7][8][9]. Boron also retards or inhibits the dissolution of aluminate phases [10,11]. Borax or B 2 O 3 has been added to the belite-ye'elimite-ferrite cement (BYF) during sintering in order to stabilize α-belite, which are more reactive with water. As a result, the ye'elimite (calcium sulfoaluminate or Klein's salt; C 4 A 3 $) phase was found in only the cubic structure [12][13][14]. Álvarez-Pinazo et al [15] attributed this effect to the substitution of Al +3 by smaller framework ions such as B 3+ , Si 4+ or Fe 3+ . C 4 A 3 $, the main constituent of low-energy cement, belongs to the tectoaluminosilicate sodalite family, which have a body-centered cubic structure comprising AlO 4 tetrahedra in a cage framework, and Ca 2+ or SO 4 2-ions which fill in the skeleton structure to balance the charge [16][17][18]. C 4 A 3 $ undergoes reversible phase transitions between cubic (I43m, c-C 4 A 3 $) and orthorhombic (Pcc2, o-C 4 A 3 $) structures under different conditions [19][20][21]. This process can be controlled with ion doping [15,22]. For example, iron-containing cubic ye'elimite is achievable by substituting Al 3+ with Fe 3+ ions [23]. c-C 4 A 3 $ can also be obtained by doping Ba 2+ or Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI. Sr 2+ ions in Ca 2+ sites, where the ions with larger ionic radii cause the expansion of the frame structure to produce a non-collapsed cubic structure [24,25]. C 4 A 3 $-containing binders have received considerable attention for their rapid hardening and high strength [26][27][28]. Different crystal structures of C 4 A 3 $ have different hydration behavior, but the difference between the two phases is not obvious [29]. The main hydrates of C 4 A 3 $ are monosulfate with variable water content (C 4 A$H 12 , AFm 12 ; C 4 A$H 14 , AFm 14 ), ettringite (C 6 A$ 3 H 32 , AFt), and amorphous aluminum hydroxide (AH 3 ) [30][31][32]. Gypsum both controls the hydration process of C 4 A 3 $ and changes the ratio of the main hydrates (AFt and AFm) [33][34][35]. Calcium hemicarboaluminate hydrate (C 4 AČ 0.5 H 12 , Hc), which belongs to the AFm family, forms when the calcite content of cement pastes is low or zero [36][37][38]. The excellent binding performance of C 4 A 3 $ is attributed to the rapid formation of AFt skeleton structure, and the precipitation of other hydration products. However, the hydration process and the hardening time of C 4 A 3 $ are hard to control, and a retarder (borate, boron, citric acid etc) is often needed to obtain a suitable time between mixing and hardening. It is worth noting that although the retarder delays the start of hydration, the hydration reaction is still rapid [1,6,39].
The C 4 A 3 $ hydration must be suitably adjusted to meet the requirements of the intended application. B 2 O 3 has an important influence on pastes hardening and therefore the addition of B 2 O 3 during calcination may further optimize the C 4 A 3 $ properties. Currently, the effect of B 2 O 3 on the C 4 A 3 $ microstructure and hydration is poorly understood, but it is known that C 4 A 3 $ exhibits only a cubic structure when B 2 O 3 is added to BYF [12,13,15]. Thus, various quantities of B 2 O 3 were added to C 4 A 3 $ during sintering, and microstructure and hydration performance of C 4 A 3 $ were investigated.

Experimental section
2.1. Synthesis of C 4 A 3 $ Five C 4 A 3 $ named C0, C0.1, C0.5, C1.0, and C2.0 were synthesized via the solid state reaction according to the literature [40]. C0 denotes without any dopant, while C0.1, C0.5, C1.0 and C2.0 denote C 4 A 3 $ with 0.1, 0.5, 1.0 and 2.0 wt% B 2 O 3 , respectively. Doped and undoped C 4 A 3 $ was prepared using stoichiometric amounts of calcite, aluminium oxide and gypsum, and appropriate amounts of B 2 O 3 (analytical reagent). The raw materials were ball-milled for 2 h and pressed into discs at the pressure of 15 MPa. The discs (Φ 5.0 × 1.0 cm) were heated at 1350°C for 2 h, then cooled in air. Finally, parts of the sintered discs were broken away for scanning electron microscopy (SEM) analysis. Another part was ball-milled for 2 h and characterized by x-ray diffraction (XRD), isothermal conduction calorimetry (ICC), particle size measurements and Fourier transform infrared spectroscopy (FTIR) analysis.
The particle size distribution of ball-milled C 4 A 3 $ with various B 2 O 3 content was measured using a laser grain size analyzer. The volume-mean grain size (d 50 ) of C0, C0.1, C0.5, and C1.0 were 9.6 μm, 9.3 μm, 9.4 μm, and 10.0 μm respectively (figure 1). Additionally, C 4 A 3 $ with 1.0 wt% B 2 O 3 was slightly less easy to grind than other samples. Overall, the particle size distributions of C 4 A 3 $ with various B 2 O 3 content are similar, so the effect of different particle size distributions on the hydration of C 4 A 3 $ can be ignored. For a comparison, the C0 was homogenized with 0.1, 0.5, and 1.0 wt% B 2 O 3 for hydration analysis, and denoted B0.1, B0.5, and B1.0, respectively.

Preparation and performance of pastes
A water/solid ratio of 0.5 was used to prepare the specimens of C 4 A 3 $ paste. Pastes were cast into 20 × 20 × 20 mm 3 cubic moulds and left for 1 day. The cement paste was then removed and cured at 20°C ±1°C with a relative humidity >95% for 3, 7, or 28 days. At each curing age, six cubic specimens were removed and tested for their compressive strength with a compression measuring equipment at a loading speed of 1.0 kN s −1 . C 4 A 3 $ hydrates fast and the main hydration reaction is within 1 day [41,42]. Thus, C0 and C0.5 were cured for 5 min, 12 h or 1 day, and then filtered or broken. The hydration reaction was stopped by soaking in an excess ethanol for 1 day and then dried in a vacuum desiccator at ambient temperature. At last, a portion of each of the specimens was ground to pass a 75 μm mesh sieve for XRD and thermogravimetric (TG) analysis. Another section was broken away for SEM analysis.

Methods
XRD patterns of the clinker and pastes were obtained between 2θ = 5°-45°with a Smartlab-3kw x-ray diffractometer (Rigaku Ltd., Japan) using CuKa radiation (λ = 0.15405 nm), and operated in a reflection geometry (θ/θ) at room temperature. A scanning speed of 5°min −1 with a step size of 0.02°was used. The x-ray tube was operated at 40 kV and 30 mA. The XRD patterns were analyzed using the software GSAS for Rietveld refinement and phase quantification [43,44]. The refinement parameters include background coefficients, LP factor, cell parameters, zero-shift error, peak shape parameters, absorption and preferred orientation. In addition, 20 wt% ZnO was homogenized with C0, C0.1, C0.5, and C1.0 as an internal standard.
FTIR analysis of the clinker was performed using a Nexus 670 FTIR spectrometer (Nicolet Ltd., USA). 1 mg of sample was ground with 100 mg KBr in an agate mortar, and the mixture was then pressed into a transparent tablet. FTIR spectra were collected between 400-1400 cm −1 .
The sample microstructure of clinkers and cement was characterized by SEM using a JSM-6510 (JEOL Ltd., Japan), which was equipped with a W-filament and operated under an accelerating voltage of 15 kV. The spot size of the electron beam was set to 40. The samples were sputter-coated with Au prior to analysis.
The hydration process of some samples was monitored by ICC using a Thermotric TAM Air calorimeter (Ta Instruments Ltd., America) at 20°C for 90 h using an internal mixing procedure. The pastes were prepared using a water/solid ratio of 0.5. Once the water was injected, and mixed with the powder, the heat flow was recorded at 30 s intervals.
TG analysis was performed using a STA 449C simultaneous thermal analyzer (NETZSCH Ltd., Germany) between 40°C to 500°C at a heating rate of 10°C min −1 under an N 2 atmosphere.

Results and discussion
3.1. XRD analysis of C 4 A 3 $ XRD analysis of C0, C0.1, C0.5, C1.0, and C2.0 suggested that all samples constituted both orthorhombic and cubic C 4 A 3 $ phases (figure 2(a)). Boron-rich phases were not detected by XRD. The XRD pattern for C0 featured low intensity signals attributed to mayenite (C 12 A 7 ) and tricalcium aluminate (C 3 A), which were formed due to the volatilization of sulfur in the raw materials [45]. The peaks attributed to C 12 A 7 and C 3 A were not observed in patterns from C0.1, C0.5 and C1.0, which suggested that B 2 O 3 hindered the volatilization of sulfur and thereby contributed to C 4 A 3 $ formation. However, peaks attributed to a by-product calcium dialuminate, CA 2 appeared in the pattern for C2.0, which indicated that C 4 A 3 $ was unable to accommodate excessive B 2 O 3 .
The position of the strongest diffraction peak of the C 4 A 3 $ shifted to higher angles as the B 2 O 3 content increased ( figure 2(b)). This indicated that the lattice parameters of C 4 A 3 $ became smaller as B 3+ substituted for Al 3+ in AlO 4 tetrahedra. The relative intensities of the strongest peaks from C0, C0.1, C0.5, and C1.0 compared to the strongest peak attributed to ZnO were 3.1, 3.5, 3.1, and 3.0, respectively. This suggested that as B 2 O 3 content increased from 0.0 to 0.1 wt%, the relative content of C 4 A 3 $ first increased due to the disappearance of byproducts C 12 A 7 and C 3 A. As the B 2 O 3 content increased further to 1.0 wt%, other amorphous phases were generated which were not detectable by XRD. The intensity of the (121) diffraction peak (18.02°) of o-C 4 A 3 $ was gradually weakened, but the peak centre did not change as the B 2 O 3 content increased ( figure 2(c)). This suggested that while the content of o-C 4 A 3 $ decreased, its lattice parameters did not change. Only the lattice parameters of c-C 4 A 3 $ became smaller.
XRD patterns of C 4 A 3 $ samples prepared with various B 2 O 3 contents were then interrogated using the software package GSAS for Rietveld refinement (figure 3). The refined data showed a satisfactory fit with the raw data, which confirmed the reliability of the refinement protocol. The Rietveld quantitative phase analysis (RQPA) of clinkers is shown in table 1. Approximately 5 wt% of byproducts C 12 A 7 and C 3 A were observed in C0. The refinement data suggested that there were no byproducts in the clinker, and only a small amount of B 2 O 3 was needed to promote C 4 A 3 $ formation. Besides, as the B 2 O 3 content increased, the c-C 4 A 3 $ content increased while the o-C 4 A 3 $ content decreased. In C1.0, the content of c-C 4 A 3 $ in the clinker was 38.4 wt%. We determined that B 2 O 3 can only partially promote the conversion of o-C 4 A 3 $ to c-C 4 A 3 $, and that the results were different from the phase transition law established for C 4 A 3 $ in the B 2 O 3 -rich BYF clinker [13,15,46]. Therefore, further promotion of the conversion of o-C 4 A 3 $ to c-C 4 A 3 $ would require doping with numerous  different ions simultaneously. In C2.0, 8.6 wt% CA 2 was formed, and the C 4 A 3 $ content decreased. Due to the limits of the RQPA method, and the use of diffraction-based source data which is better suited for ordered crystalline phases, the amorphous content in the clinker samples was not accurately determined by the results of internal standard tests. It was inferred that the amorphous content was therefore low (<10 wt%).
Rietveld refinement studies also yielded C 4 A 3 $ refined lattice parameters (table 2). The lattice parameters of o-C 4 A 3 $ barely changed as the B 2 O 3 content increased, which was in agreement with the raw data analysis as presented in figure 2(c). Meanwhile, the lattice parameters of c-C 4 A 3 $ were unchanged in C0.1, which suggested that very small amounts of B 2 O 3 promoted C 4 A 3 $ formation and it may act as a melted agent. As the B 2 O 3 content increased from 0.5 to 2.0 wt%, the lattice constant, a, of c-C 4 A 3 $ continued to decrease. Therefore, B 2 O 3 not only promoted c-C 4 A 3 $ formation, but also entered the crystal lattice. The refined lattice parameters were used to determine Al/B-O bond lengths of c-C 4 A 3 $, which were shown to decrease as the B 2 O 3 content increased ( figure 4) [22,47]. The refinement data showed that B 2 O 3 promoted C 4 A 3 $ formation by acting as a molten phase which enhanced mass transport, partially promoted the conversion of o-C 4 A 3 $ to c-C 4 A 3 $; and substituted for Al 3+ in AlO 4 tetrahedra in c-C 4 A 3 $, especially at high B 2 O 3 contents.

FT-IR analysis of C 4 A 3 $
FTIR spectra of C 4 A 3 $ with various B 2 O 3 content obtained within the range 400-1400 cm −1 featured absorbance maxima which were assignable to different functional groups ( figure 5). The absorbance maxima   We note that the shifts in FTIR spectra were small and qualitative since most C 4 A 3 $ was in the undoped orthorhombic phase.

SEM analysis of C 4 A 3 $
Secondary electron micrographs were obtained from all samples to assess the effect of B 2 O 3 content on the grain size and shape of the C 4 A 3 $ phase (figure 6). C0 featured small (0.5∼2 μm) and polygonal grains (figure 6(a)), whereas a larger grain size (1∼5 μm) was observed in C0.1 ( figure 6(b)). The larger grain size observed in C0.1 was attributed to the promotion of C 4 A 3 $ formation by B 2 O 3 , which also led to the increased wt% content of C 4 A 3 $ as indicated by the increase in intensity of the strongest peak attributed to C 4 A 3 $ from the XRD analysis (see figure 2). The C 4 A 3 $ grains in C0.5 were still large, but the polygonal morphology and grain boundaries were difficult to distinguish ( figure 6(c)). This was attributed to the solidification of the amorphous phase on the grain surfaces during cooling. The amorphous phase was thought to be the byproduct in C0.5 which lowered the overall wt% content of C 4 A 3 $, but which was not detected directly by XRD. The amorphous phases may have mainly consisted of poorly crystalline C 4 A 3 $ or related compositions. Micrographs of C1.0 revealed further interconnectivity of grains to form a relatively dense, smooth structure caused by the deposition of a larger amount of amorphous byproduct ( figure 6(d)). This explained the further reduction in overall wt% content of C 4 A 3 $, and the observation that C1.0 was harder to grind.

ICC testing of C 4 A 3 $
For comparison, heat flow and heat released data from ICC was first used to investigate the hydration reaction of C0, B0.1, B0.5, and B1.0 at 20°C within the first 90 h of hydration ( figure 7). Induction times for C0 and B0.1 were short at approximately 2 h. The sharp heat flow peak suggested that the hydration reaction was rapid and lasted approximately 18 h ( figure 7(a)). The hydration heat flow peak of B0.1 was slightly smaller than that of C0. The induction time for B0.5 was longer at 15 h, and the only exothermic heat flow peak intensity was smaller than B0.1. The hydration reaction of B0.5 was still violent although it was delayed. Finally, the induction time for B1.0 was longer than the 90 h time frame examined, which suggested that the hydration reaction was severely hindered and the total heat released was close to zero ( figure 7(b)). The induction period for CSA with a 2 wt% borax content was 24 h, which is considerably shorter than that observed in B1.0 [13]. This confirmed that the retarding effect of B 2 O 3 was clearly greater than that of borax. The total heat released from B0.1 was slightly larger than that from B0.5, and both were larger than that of from C0. This suggested that the retarding effect of B 2 O 3 was uncontrollable due to the violent, rapid nature of the hydration reaction.
Heat flow and heat released data were also obtained from C0, C0.1, C0.5, and C1.0 at 20°C during the first 90 h of the hydration reaction ( figure 8). In comparison with the rapid, violent hydration reaction in C0, the induction time of C0.1 was considerably longer at 10 h, and the heat flow data featured two low-intensity maxima ( figure 8(a)). Therefore, the C 4 A 3 $ grain size had a significant impact on the hydration process. As the grain size increased from C0 to C0.1 (see figure 6(b)), the hydration rate decreased such that the timespan of the reaction increased from approximately 10 to 20 h.   increased was attributed to both the increase in C 4 A 3 $ grain size and an increasing hindering effect of the amorphous phase on the hydration reaction (see figure 6). Furthermore, B 3+ released from C 4 A 3 $ during hydration may have also contributed to the continuous inhibition of C 4 A 3 $ dissolution resulting in longer hydration reaction times [10,11]. C0.1, C0.5, and C1.0 exhibited the longer reaction time and they also yielded the higher total heat release ( figure 8(b)). In particular, C0.5 yielded the highest total heat release of all samples analyzed. A progressively higher B 2 O 3 content in the clinker caused a gradual increase in the induction time of the hydration of C 4 A 3 $, an apparent decrease in the hydration rate, and an increase in the total heat released.

XRD analysis of pastes
XRD patterns were obtained for C0 and C0.5 after curing for 5 min, 12 h, or 1 day (figure 9). After curing C0 for 5 min, tiny amounts of AFt was formed immediately, while large amounts of AFt, AFm 12 , and AFm 14 were observed after curing for 12 h. A small, broad peak was observed around 20°-21°, which was attributed to the formation of a small amount of AH 3 [49]. Due to the presence of CO 2 in the air, some of the AFm 12 and AFm 14 was converted into H c after curing for 1 day. Hydration products were not detected in C0.5 after curing for 5 min, and only a little AFt was formed by hydration after curing for 12 h. A large amount of hydration products had formed in both samples after curing for 1 day, but there was more AFm in C0.5 than in C0. Calcium borate or other boron-rich phases was not detected by XRD. This data showed that B 2 O 3 hindered the hydration of C 4 A 3 $ during sintering, which was in agreement with ICC data.  3.6. TG Analysis of pastes TG and differential thermogravimetric (DTG) analysis was carried out on C0 and C0.5 in order to analyze the hydration products (figure 10). The TG curve of C0 after curing for 5 min was almost straight, whereas significant mass losses were detected after curing C0 for 12 h and 1 day ( figure 10(a)). This indicated that there was negligible product formation in the early curing stages and extensive product formation in the later curing stages. The DTG curve of C0 after curing for 5 min (the inset in figure 10(a)) featured two peaks at 70°C-130°C and 140°C-300°C, which were attributed to water removal from AFt and AFm phases, respectively [50,51]. This suggested that a small amount of AFt and poorly-crystalline AFm, which was undetectable by XRD, was formed from C0 after curing 5 min. After curing for 12 h or 1 d, large amounts of AFt and AFm were formed and the decomposition temperatures increased slightly due to the increased crystallinity. AFm 14 , AFm 12 and Hc, which belong to the AFm family, were indistinguishable due to the overlap of the peaks. In addition, AH 3 was not observed by TG analysis.
TG curves of C0.5 after curing for 5 min and 12 h were also almost straight, while a significant mass loss was detected after curing for 1 d ( figure 10(b)). DTG curves revealed there were no hydration products in C0.5 after the curing ages of 5 min, and only a little AFt was formed after curing for 12 h (the inset in figure 10(b)). This suggested that the presence of B 2 O 3 delayed and retarded the C 4 A 3 $ hydration process during sintering, which was in agreement with the data and discussion described above. After curing C0.5 for 1 d, a large amount of AFm was formed but the quantity of AFt was very small. The results of TG analysis strongly agreed with XRD studies.

SEM analysis of pastes
Secondary electron micrographs of C0 and C0.5 cured for 5 min, 12 h or 1 d were obtained to provide further insight into the composition of the pastes (figure 11). Many needle-like AFt crystals formed on the C 4 A 3 $ grain surfaces after curing for 5 min, where undoped C 4 A 3 $ grains in C0 could undergo rapid hydration ( figure 11(a)). Many plate-like AFm and rod-like AFt were observed after curing C0 for 12 h ( figure 11(b)). The precipitation of various hydration products after curing for 1 d yielded a dense mixture of various structures ( figure 11(c)) [11,35,36]. Overall, C0 had a lower hydration degree because the reaction was too fast, and some minerals had not been able to react.
After a curing C0.5 for 5 min, no hydration products were detected, but instead the amorphous phases were shown to surround C 4 A 3 $ grains, which were known to hinder the hydration reaction ( figure 11(d)). After curing for 12 h, a few short rod-like AFt were generated on the surface of C 4 A 3 $ grains, but the remaining amorphous phase continued to prevent the hydration of C 4 A 3 $ ( figure 11(e)). The protective layer made of phases such as calcium borate which formed over the C 4 A 3 $ grains was not observed. Finally, after curing C0.5 for 1 day, well-developed AFm crystallites were observed, the high quality of which was attributed to the slower reaction rate ( figure 11(f)). These data agreed with ICC, XRD and TG data. Although the amorphous phases formed during sintering hindered the hydration reaction, their mechanism of retardation was different to that achieved by adding B 2 O 3 to pastes. The latter arises from the precipitation of an amorphous or poorly crystallized calcium borate over cement grains during the hydration process, or inhibition of the dissolution of aluminate phases [7,8,10,11].

Compressive strength
The effect of curing time on the compressive strengths of different C 4 A 3 $ pastes was evaluated ( figure 12). The addition of B 2 O 3 during the hydration process (i.e. samples B0.1 and B0.5) moderately improved the   compressive strength. After curing for 3 days, B0.1 had a slightly higher compressive strength of 13.9 MPa compared to C0 and B0.5. This was expected from ICC studies, where the total heat released during first 90 h of hydration was higher in B0.1 than in C0 and B0.5. Due to the rapid, violent nature of the hydration reaction, the increase in compressive strength of C0, B0.1 and B0.5 after 28 days was small, although the hydration is delayed.
The compressive strengths of C0.1, C0.5 and C1.0 were higher than those of C0, B0.1 and B0.5, regardless of the curing time. In particular, C0.5 had the highest compressive strength of 18.2 MPa after 3 days, which was expected as C0.5 had the highest total heat released during the first 90 h of hydration. When an appropriate amount of B 2 O 3 was added to C 4 A 3 $ during the sintering process, the compressive strength of C0.1, C0.5 and C1.0 improved significantly after extensive curing. The highest compressive strength of 20.9 MPa was obtained from C1.0 after curing for 28 days. The improvement of mechanical properties of C0.1, C0.5 and C1.0 was attributed to the hindered, controlled hydration which ensured that the ettringite skeleton structure was built gradually to match the continuous precipitation of various hydrates.

Conclusion
An appropriate amount of B 2 O 3 promoted the formation of C 4 A 3 $, and also partially promoted the conversion of o-C 4 A 3 $ to c-C 4 A 3 $ during sintering, although a large amount of o-C 4 A 3 $ remained unconverted. A greater degree of phase transformation from o-C 4 A 3 $ to c-C 4 A 3 $ therefore requires the simultaneous doping with other ions. As the B 2 O 3 content increased, C 4 A 3 $ grain size increased and became more extensively surrounded by amorphous phases. B 3+ could substitute for Al 3+ in the AlO 4 tetrahedra in c-C 4 A 3 $, causing a gradual decrease in the lattice constants of c-C 4 A 3 $. When the B 2 O 3 content is above 2 wt%, the CA 2 byproduct appeared because C 4 A 3 $ could not accommodate excessive B 2 O 3 . The addition of B 2 O 3 also caused the induction time of the hydration of C 4 A 3 $ to increase, while the reaction rate decreased. However, the final hydration degree was increased. After hydration of 12 h in the presence of 0.5 wt% B 2 O 3 , only a few short rod-like AFt were generated on the surface of C 4 A 3 $ grains because the amorphous phases which surrounded the C 4 A 3 $ grains hindered the hydration reaction. The compressive strength of doped C 4 A 3 $ was higher than that of undoped C 4 A 3 $, and that cured in the presence of 1.0 wt% B 2 O 3 for 28 days yielded the highest compressive strength.