Hydration of Gypsum and Cement — Dissolution, Precipitation, Connection, and Pores

Back to: Mochizuki Research Group Education — Inorganic Materials Science > Virtual Experiments, Inorganic Materials Science resources

This page is the explanation that pairs with hydration-simulator-en.html. Calcined gypsum (hemihydrate gypsum) and cement harden when water is added, not because the water evaporates, but because the powder dissolves in the water and a different solid that has taken up water (a hydrate) precipitates and joins up. This page explains that mechanism in the following order: the reaction equations and the mass and volume balance, solubility and supersaturation, nucleation and growth, the heat-release curve, and Powers’ model of volumes. How to read the simulator screen, and the equations and values used in the calculations (distinguishing literature values, empirical formulas, and provisional values), are collected in Appendix A; the items we would like to confirm with Prof. Mochizuki are collected in Items to Confirm.

The backbone of this page ① Why does it harden? The dissolved components precipitate as a solid that has taken up water as water of crystallization or structural water, and these solids join up (hydration). ② Where does it precipitate, and in what form? In gypsum, needle-like dihydrate crystals form in the supersaturated liquid and interlock; in cement, C–S–H and other products grow from the particle surfaces and bind the particles to one another. ③ What happens to the excess water? Water that is not used in the reaction remains as pores, so the more water is added, the more pores there are (§5, §9).

1. Introduction — What “Hardening When Water Is Added” Means

Both calcined gypsum and cement lose their fluidity within tens of minutes to a few hours of being mixed with water (setting), and keep getting stronger afterward (hardening). During this process the powder dissolves in the water, and the dissolved components precipitate as a different solid that has taken up water. In gypsum this solid is dihydrate gypsum, in which water molecules have entered the crystal (water of crystallization); in cement it is C–S–H, calcium hydroxide, and the like, in which water has entered the structure. Even if you dry the material after it has hardened, this water does not leave. Only the water that was not used in the reaction is removed by drying.

Another important point is the route by which the material dissolves first and then precipitates (dissolution–precipitation). The powder particles do not simply swell and stick together; the material first passes through the liquid and then precipitates in a different place and in a different form. This idea goes back to Le Chatelier (1887), and today it is observed in situ by synchrotron diffraction and X-ray CT (Refs. 6–10). The way matter is transported is completely different from sintering (sintering simulator), in which heating joins powder particles to one another by atomic diffusion.

2. Reaction and Mass Balance of Gypsum

2.1 Reaction equation and stoichiometry

(1) CaSO4·12H2O +32H2O →CaSO4·2H2O

The molar masses are 145.15 g/mol for hemihydrate gypsum (calcined gypsum), 172.17 g/mol for dihydrate gypsum, and 18.015 g/mol for water. The amount of water needed to convert 1 g of hemihydrate completely into dihydrate (the theoretical amount of water) is

(2) (w/p)theor= 1.5×18.015145.15=0.186

In actual mixing, water at around w/p = 0.6 is often used so that the paste can be poured (an example for the β form; to be confirmed), and the entire difference from the theoretical amount of water becomes “excess water” (§5). So that the simulator conserves mass exactly, it sets M(hemihydrate) = M(CaSO₄) + ½M(H₂O) and M(dihydrate) = M(CaSO₄) + 2M(H₂O) (M(CaSO₄) = 136.14 g/mol).

Table 1. Mass balance for 100 g of powder + 60 g of water (w/p = 0.60)Calculated values for complete reaction. This is the same calculation as the readout in simulator tab ①
StateHemihydrateDihydrateFree waterTotal
Just after mixing100.0 g0 g60.0 g160.0 g
After the reaction is finished (before drying)0 g118.6 g41.4 g160.0 g
After drying0 g118.6 g0 g118.6 g

The dried solid is 18.6 g heavier than the original 100 g of powder. This is the water taken in as water of crystallization, and it is direct evidence that the material does not harden by “the water evaporating.”

2.2 Heat release and volume

The hydration of hemihydrate to dihydrate is an exothermic reaction, and the code of CEMHYD3D (NIST’s cement hydration model) uses 132 J per 1 g of β-hemihydrate (Ref. 4). Calculating from the heats of formation gives ΔH ≈ −19.2 kJ/mol, which agrees with this value. The value from a thermodynamic database that corresponds to the α form gives −17.0 kJ/mol (118 J/g) (Ref. 3). When the temperature is actually measured in a container that is not adiabatic, it rises from 20 °C to about 30 °C for the β form (Ref. 6). The simulator assumes that the temperature is held constant in a thermostatic bath.

As for volume, hemihydrate 52.97 cm³/mol (density 2.74 g/cm³) + water 27.0 cm³ → dihydrate 74.21 cm³/mol (density 2.32 g/cm³), so the product is smaller than the reactants by 5.8 cm³/mol (7.2 %; the value with the density of water taken as 1 g/cm³) (chemical shrinkage). On the other hand, looking at the solid alone, the volume increases by a factor of 1.40, and this gradually fills the gaps between the grains.

3. Solubility and Supersaturation — Why Calcined Gypsum Hardens

Hemihydrate gypsum is more soluble in water than dihydrate gypsum. At 25 °C the solubility of the dihydrate is 15.1 mmol/kg (PHREEQC Example 2) and that of the hemihydrate is about 58 mmol/kg, so the ratio is about 3.9 (Table 2). When the hemihydrate dissolves, the liquid immediately becomes supersaturated about fourfold with respect to the dihydrate. When the dihydrate precipitates from the supersaturated liquid, the liquid becomes more dilute, and more hemihydrate dissolves to make up for it. This repetition of “hemihydrate dissolves ⇄ dihydrate precipitates” is what dissolution–precipitation really is.

Table 2. Solubility of each CaSO₄ phase (mol/kg water)Values read from the figure of Azimi et al. 2007 (reading error of about ±0.003). The 25 °C value for the dihydrate is 15.1 mmol/kg from PHREEQC Example 2
TemperatureDihydrate (gypsum), DHHemihydrate, HHRatio Smax = cHH/cDH
0 °C0.01250.0786.2
25 °C0.01500.0583.9
50 °C0.01550.0362.3
75 °C0.01450.0221.5
100 °C0.01350.0141.0

Fig. 1. Solubilities of the hemihydrate and dihydrate, and the upper limit of supersaturation Smax (the values in Table 2 joined by straight lines; the simulator also uses this interpolation)

Expressing the driving force as a Gibbs energy, from the difference between the equilibrium constants in the thermodynamic database llnl.dat (log K: hemihydrate −3.66, dihydrate −4.48), at 25 °C we obtain

(3) ΔG=−RTln⁡100.82≈−4.7 kJ/mol

This value is based on Ref. 3. Raising the temperature greatly lowers only the solubility of the hemihydrate, so Smax becomes smaller and reaches 1 at about 99 °C (at higher temperatures, conversely, the dihydrate turns into the hemihydrate). For this reason, warming does not necessarily make it harden faster. The rate coefficient of the reaction increases with temperature, but the driving force (supersaturation) becomes smaller. For dental and building gypsum, it is often explained that raising the temperature of the mixing water makes setting faster up to around 40 °C and slower above that (primary sources were not checked in this survey).

What happens when water is added to uncalcined dihydrate powder Even when dihydrate gypsum powder dissolves in water, the liquid becomes concentrated only up to the solubility of the dihydrate (S = 1). The liquid does not become supersaturated, so no new crystals form and the grains do not join up. Calcined gypsum hardens because it is dihydrate that has been heated to drive out part of its water, turning it into hemihydrate, which dissolves more readily than the dihydrate. You can check this with comparison condition G in simulator tab ①.

4. Nucleation, Growth, and Setting

4.1 Nucleation

For a crystal to form from a supersaturated liquid, the supersaturation must be large enough to outweigh the surface-energy cost of a small crystal. In classical nucleation theory, the nucleation rate J takes the form

(4) J=J0exp⁡(−B(ln⁡S)2) ,S=ccDH

(B is a constant proportional to the cube of the surface energy). When S is small, J drops sharply, so almost no nuclei form at high temperature or in dihydrate powder. The simulator uses this form, and J0 and B are provisional values for teaching. Nuclei are assumed to form in proportion to the amount of surface of the hemihydrate grains (the sum of their perimeters) (a schematic of heterogeneous nucleation). La Bella et al. (2023) report that the contribution of dihydrate forming directly and epitaxially on the surface of the hemihydrate is small (Ref. 7). The simulator therefore does not attach nuclei directly to the grain surface but places them in the concentrated liquid around a grain. One grain is chosen with a weight given by its original perimeter (a grain that has completely dissolved is not chosen, and the next grain is taken), and the nucleus is placed at a distance from the grain center between rL − 0.3δ and rL + δ (rL is the radius of the grain at w/p = 0.40, and δ = 14 px is the width of the concentrated liquid around a grain; schematic). Because the position is the same regardless of w/p, it is easy to compare the setting times when w/p is changed (when w/p is large the grains are small, so this band shifts outside the original outline r0. The shift is at most about 5 px at w/p = 1.0, and for large grains the lower edge of the band also lies outside the original outline). The fraction of nuclei that form inside the space left by a dissolved grain depends on the time (counting the nuclei whose center lies inside the original outline of some grain; 20 °C). At the moment of setting it is about 10 % or less for any w/p (4–11 %), with no trend in w/p. At 120 min it is about 20 % for w/p = 0.4 and less than 10 % for 0.7 and above, so in the later stage the larger w/p is, the smaller the fraction. When the center would fall inside the current grain, the nucleus is placed again outside the current edge (up to rL + δ), and when it would fall inside a neighboring grain, it is moved out beyond the edge of that grain. In this way the nuclei form either in the space left by a dissolved grain or in the concentrated liquid outside, that is, in the liquid. The number that selects the grain and the direction are not random but come from a sequence with golden-ratio steps, which distributes the nuclei evenly among the grains and over the directions (to reduce the scatter of the setting time caused by the random seed; §4.3). Because the grain is chosen with a weight given by its original perimeter, in the later stage, when the grains have become small, this deviates slightly from the formula for the nucleation rate (proportional to the current perimeter) (the difference up to about the time of setting is small). A grain that has completely dissolved is not chosen, so once a grain has dissolved away, no nuclei form near it. However, the grains are packed densely, most of the liquid outside the grains lies within δ of a grain, and the needles grow to many times δ in length, so the bias in the positions of the needles is hardly visible in the figure.

4.2 Growth and degree of reaction

The dihydrate that forms grows into needles and later becomes thicker (Refs. 7, 8). In the simulator, for each representative needle, the length ℓ and the thickness w are grown by

(5) dℓdt=gLk(T)(S−1)1+ℓ/ℓc, dwdt=gWk(T)(S−1)

and the volume fraction of the dihydrate is expressed by the fraction of the area occupied by the needles in the figure, Σℓw/A (the area fraction in the two-dimensional figure is regarded as the volume fraction). The rate at which the hemihydrate dissolves is d nHH/dt ∝ −(grain perimeter)·(cHH − c), and the concentration of the liquid c is determined by the difference between the amount dissolved and the amount precipitated. E = 30 kJ/mol in k(T) = exp[−(E/R)(1/T − 1/293.15 K)] is also a provisional value. The curve of the degree of reaction α obtained in this way (the fraction of the hemihydrate that has become dihydrate) is S-shaped, and it is known to be well described by the Avrami form α = 1 − exp[−k(t − t0)n] (n ≈ 3 for the β form; Ref. 6).

The coefficients were chosen to be close to “about 25–30 min to reach 50 % and about 60–70 min to exceed 90 %” for the β form of Gurgul et al. (2019) (w/p = 0.8, about 20 °C) (read from the abstract; to be confirmed), and to the initial set of 10.4 min and the final set of 16.0 min of building plaster (w/p = 0.5) in Darmawan et al. (2020). In the simulator’s representative condition (w/p = 0.8, 20 °C), it takes 27 min to reach 50 % and 52.5 min to reach 90 %, and the setting at w/p = 0.5 is at 14.6 min (Appendix A.5). The induction period (until α reaches 1 %) is about 3 min, which is shorter than the 9 min of Gurgul et al.

4.3 Setting = connection of the needle network

When the needles cross one another and form a network connected from one end of the sample to the other, the paste begins to hold its shape. Morgado et al. (2019) report that the fraction of needles belonging to the percolating (connected) cluster is a good indicator of the onset of setting and agrees with the rise of the shear modulus (Ref. 10). In the simulator, needles that cross as line segments (or whose half-thicknesses touch) are regarded as belonging to the same cluster, and the time at which a cluster connecting the left and right edges of the window forms is defined as “setting.” It is known that line segments of length ℓ scattered in two dimensions become connected over an infinite extent when the number density exceeds Nℓ²/A ≈ 5.64 (Ref. 23). From this we see that, for the same amount of dihydrate, the longer and thinner the needles, the earlier they connect. When w/p is increased, the amount of dihydrate in the window decreases, so more reaction is needed before the needles connect, and setting is delayed. Because the needles form around the grains (§4.1), how they connect also depends on the arrangement of the grains. In the simulator the grains are placed with the same random sequence regardless of w/p (they are placed for the amount of powder at w/p = 0.40, and only the radii are shrunk to match w/p), so that only the spacing between grains changes when w/p is changed. When nuclei are placed near the grains, the setting time depends on the arrangement of the grains and nuclei (the random seed). Over 80 different seeds, the setting at w/p = 0.60 and 20 °C scatters over 12.4–22.4 min (median 15.6 min; 80 % of the runs lie within 13.2–19.4 min). The property that setting becomes later as w/p increases from 0.40 to 1.00 (in steps of 0.1) is preserved in 73 of the 80 seeds, and even for the seeds in which it breaks down, the reversal is 1.2 min or less. The default seed is a sequence chosen to be close to this median, and the setting at w/p = 0.60 is 15.4 min (this is not a value matched to the 15.6 min of the previous version, but a value obtained by choosing a seed near the median). The progress of the degree of reaction does not depend on the positions of the nuclei, but because the arrangement of the grains (the size distribution and the perimeters) was changed to a sequence independent of w/p, the time to reach α = 0.9 shifted by at most 1.5 min from the previous version (Table A1).

5. Excess Water and Pores

Assuming that the apparent volume is the same as at the time of mixing, the porosity after complete reaction and drying is

(6) φ=1−VDH/MHH1/ρHH+w/p =1−0.5110.365+w/p

Here the units are cm³/g (VDH = 74.21 cm³/mol, ρHH = 2.74 g/cm³). The porosity is 7 % at w/p = 0.186, 33 % at 0.4, 47 % at 0.6, and 63 % at 1.0. It is known that real gypsum expands slightly on setting (the growth pressure of the crystals), but we could not confirm the amount in this survey, so Eq. (6) is an approximation of “constant apparent volume.” Strength decreases as porosity increases, but the coefficients could not be confirmed in the literature, so simulator tab ① does not display strength and shows only the porosity and the network fraction.

6. Composition of Portland Cement

Portland cement is made by adding a few percent of gypsum to clinker, which is limestone, clay, and the like fired at about 1450 °C, and grinding the mixture. In cement chemistry notation, C = CaO, S = SiO₂, A = Al₂O₃, F = Fe₂O₃, S̄ = SO₃, and H = H₂O, and the four main minerals are C₃S (alite), C₂S (belite), C₃A (aluminate), and C₄AF (ferrite). Paraphrasing the key points of lecture 01, slide 14: alite (C₃S) develops strength from an early age over a long period, belite (C₂S) reacts slowly and contributes to the strength after many years, and aluminate (C₃A) is a mineral with a large heat of hydration (the role of ferrite (C₄AF) is not given on the slide). Low-heat cement has more C₂S and less C₃A in order to use this property to suppress heat release. The simulator’s default values are the mineral compositions of lecture 01, slide 14 (Table 3). Comparing with the source data of Prof. Mochizuki’s bachelor’s thesis (Ref. 26), the values for moderate-heat and low-heat cement correspond to values calculated with Taylor’s equation (without the SO₃ correction), and those for ordinary cement correspond to values from Bogue’s equation (including the SO₃ correction, the same equation as in the table of the bachelor’s thesis). Aligning with Bogue’s equation gives 47.8/30.5/2.3/13.0 for moderate-heat cement and 34.8/48.0/1.5/9.2 for low-heat cement (C₃S/C₂S/C₃A/C₄AF, %). The simulator uses the values on the slide as they are.

Table 3. Mineral composition (mass%)Ordinary, moderate-heat, and low-heat are from lecture 01, slide 14 (the ordinary type corresponds to values from Bogue’s equation, and the moderate-heat and low-heat types to values from Taylor’s equation). CEM I 42.5R is the XRD quantification of Jagličić et al. 2025 (Ref. 12)
CementC₃SC₂SC₃AC₄AF
Ordinary (OPC)59.915.88.29.6
Moderate-heat (MPC)53.326.42.313.0
Low-heat (LHC)40.943.41.59.2
CEM I 42.5R51.819.57.57.5

The four minerals add up to 93.5–95 % for ordinary, moderate-heat, and low-heat cement and to 86.3 % for CEM I 42.5R, and the rest is gypsum (sulfate), calcite, and so on. The slide does not give the amount of gypsum, so a provisional value of 4.5 %, obtained by converting the SO₃ content of 2.10 % of a product example (Ref. 16) to dihydrate gypsum, is used for moderate-heat, low-heat, and CEM I 42.5R. Only for ordinary cement did we use 4.9 % (SO₃ 2.3 %). In the model, a lot of gypsum is consumed by the reaction of C₃A just after mixing (§8.1), so if 4.5 % were used for ordinary cement, the gypsum would run out (at 11.0 h) before the main peak of C₃S. In real cement, the amount of gypsum is adjusted so that the sulfate runs out after the main peak (§8, Ref. 12), so we made the ratio of gypsum to C₃A for ordinary cement almost the same as that of CEM I 42.5R, where this is the case in measurements (provisional). Note that the sulfate in CEM I 42.5R is 3.5 % anhydrite and 1.2 % hemihydrate (Ref. 12), and the amount of SO₄ is 0.034 mol/100 g, about 30 % more than that of 4.5 % dihydrate (0.026 mol/100 g). In reality part of the sulfate is also used for the hydration of C₄AF and for adsorption on C–S–H, but in the model only C₃A uses it, so 4.5 % and 4.9 % are provisional values meaning “the effective amount used by C₃A” (§8.1, Items to Confirm, item 6).

7. Hydration Reactions and Products of Cement

Portland cement is said to have begun when J. Aspdin, a bricklayer in Leeds, England, obtained a patent in 1824 (lecture 01, slide 14), and the mechanism by which it reacts with water and hardens is given by the following hydration reactions.

Table 4. Main hydration reactions and heats of hydrationCalculated from the molar volumes and heats of formation in CEMHYD3D (Ref. 4). The composition of C–S–H was taken to be C₁.₇SH₄
ReactionHeat of hydration (J/g of mineral)Volume of product / solid reactant
C₃S + 5.3H → C₁.₇SH₄ + 1.3CH517 (phase-specific value)2.12
C₂S + 4.3H → C₁.₇SH₄ + 0.3CH262 (phase-specific value)2.25
C₃A + 3CS̄H₂ + 26H → C₆AS̄₃H₃₂ (ettringite, AFt)16722.36
2C₃A + C₆AS̄₃H₃₂ + 4H → 3C₄AS̄H₁₂ (monosulfate, AFm)1762 (per C₃A)1.03
Hydration of C₄AF (when AFt forms)725—

For C₃S and C₂S, the values calculated from the heats of formation are 535 and 222 J/g, which differ slightly from the phase-specific values of 517 and 262 J/g that CEMHYD3D uses for its heat calculation (the difference comes from how the composition of C–S–H is set). The simulator uses the phase-specific values. The products have different shapes, so they are drawn differently in the inside figure.

The reason for not drawing everything as the same needles is that the difference in how the materials harden (§10) comes from the difference in the shapes and locations of the products. The shapes and arrangement are nevertheless schematic. The default of the inside figure in simulator tab ② looks the same as the “flow → setting → hardening” figure in lecture 01, slide 15. Unhydrated particles are drawn as gray polygons, water as light blue, the hydrates C–S–H, CH, AFm, and C₄AF together as an ocher layer (with a pattern of black fibers), and AFt as short dark-blue rods. Choosing “Color-coded by phase” gives a figure in which the four kinds above are drawn with different shapes and colors (Appendix A.3).

8. Heat-Release Curve and Stages of Hydration

The heat-release rate of cement measured with an isothermal calorimeter can be divided into the following stages. A sharp peak just after mixing (surface dissolution and the formation of AFt; it reaches its maximum within a few minutes and becomes small in about 1 hour) → the induction period with little heat release (a few hours; the paste is still fluid) → the acceleration period, in which C–S–H and CH precipitate through the hydration of C₃S → the main peak → the deceleration period, in which the layer of products becomes thicker and the reaction slows down. Right after the main peak, a further peak (a shoulder) overlaps, in which the gypsum (sulfate) is used up and the reaction of C₃A speeds up again. Cement manufacturers adjust the amount of gypsum so that the time at which this sulfate runs out comes after the main peak of C₃S (Ref. 12). For the CEM I 42.5R of Jagličić et al. (2025) (w/c = 0.5, 20 °C), the end of the induction period was tis = 3.7 h, the maximum of the acceleration was at tm = 7.0 h (1.3 mW/g, slope 0.27 mW/(g·h)), and the main peak was at tmax = 12.6 h with 2.2 mW/g; the cumulative heat release was 41.4 J/g at 6.2 h, 45.1 J/g at 7.0 h, and 81.8 J/g at 12.6 h (Table 1 of Ref. 12; the sample was placed in the calorimeter 5 min after mixing for the measurement).

8.1 The simulator’s rate equations (for teaching)

The degree of hydration αi of C₃S, C₂S, and C₄AF is expressed as the sum of an initial fast reaction, nucleation and growth (Avrami form), and a slow diffusion stage (three-parameter exponential form).

(7) αi(θ)= a0(1−e−θ/τ0) +A[1−e−(k(θ−ti))n] +Be−(τD/(θ−ti))βD

The second and third terms are added only for θ > ti. For C₃A, the reaction is held back by the AFt layer formed on the surface as long as gypsum remains, and it speeds up again once the gypsum is used up (the “shoulder” in §8). To keep this order (the gypsum runs out → C₃A speeds up) inside the model as well, only C₃A was given the following form.

(7a) αC₃A(θ)= a0(1−e−θ/τ0) +Bs(1−e−θ/τs) +Bxe−τx/min(θ,θsd) +A[1−e−(k(θ−θsd))n] +Be−(τD/(θ−θsd))βD
(7b) a0(1−e−θsd/τ0) +Bs(1−e−θsd/τs) +Bxe−τx/θsd =ngypsum3nC₃A

The fourth and fifth terms are added only for θ > θsd. The third term is a slow term in which C₃A reacts slowly through the AFt layer even while gypsum remains; we set Bx = 1 − a0 − Bs = 0.494 and τx = 300 h (provisional), and after the gypsum is used up it is added with the value at that moment held fixed. Its magnitude at θ = 15 h is 10⁻⁸ or less, so it has almost no effect on the results for ordinary cement and CEM I, in which the gypsum runs out at around 15 h. θsd is the clock of the reaction that uses up the gypsum, and it is calculated for each composition from Eq. (7b) as the time at which the gypsum used for the formation of AFt (C₃A + 3CS̄H₂ → AFt, Table 4) reaches the whole amount (n is in mol per 1 g of cement). If the right-hand side is at least a0 + Bs + Bx = 1, the gypsum does not run out and θsd = ∞. In this case too, C₃A reacts slowly through the third term while forming AFt, and approaches 1 with time (α of C₃A in moderate-heat and low-heat cement is 0.58 at 7 days and 0.81 at 28 days). For ordinary cement (right-hand side 0.313) the gypsum runs out at 14.8 h, and for CEM I 42.5R (0.314) at 15.0 h, both after the main peak (12.6 h); for moderate-heat (1.02) and low-heat (1.57) cement there is little C₃A, so gypsum remains even at 28 days (the amounts used by 28 days are 79 % and 52 %). This equation expresses the role of gypsum as a “setting regulator”: the less gypsum there is relative to the amount of C₃A, the earlier it runs out. In reality the sulfate is also used for the hydration of C₄AF and for adsorption on C–S–H, but the model does not count this (§6).

The coefficients are set so that a0 + A + B = 1 for C₃S, C₂S, and C₄AF and a0 + Bs + A + B = 1 for C₃A (while gypsum remains, a0 + Bs + Bx = 1), so that, as long as the water does not run out, every phase approaches αi → 1 with time (even while gypsum remains, C₃A heads toward 1 through the third term; what stops it is the water constraint of §8.2). The Avrami equation (Ref. 20) describes nucleation and growth, and the three-parameter exponential form exp[−(τ/t)β] is an empirical formula often used for the degree of hydration of concrete (Ref. 21; the original source has not been checked). The idea of stages, “slowing down by diffusion after nucleation and growth,” follows the model of Parrott and Killoh (1984) (original source not checked). The coefficients (Table A3) were fitted to the CEM I 42.5R values of Jagličić et al. (tis and tm with their heat-release rate and slope, the heat-release rate at 6.2 h, the time and height of the main peak, the cumulative heat release at 6.2, 7.0, and 12.6 h integrated from 5 min after mixing, and the time at which the gypsum runs out, which is after the main peak (about 15 h), with the C₃A peak at about 17 h (read from the position of tsd in Fig. 1 of that paper)), and to the heats of hydration of a product example, 332 J/g at 7 days and 384 J/g at 28 days (Ref. 16), and to the degrees of hydration of the individual phases that Gutteridge and Dalziel (1990) measured by quantitative XRD (w/s = 0.71; at 7 days 0.73 for C₃S, 0.19 for C₂S, 0.76 for C₃A, and 0.27 for C₄AF, and at 28 days 0.82, 0.44, 0.83, and 0.40; read from Fig. 2 of NISTIR 7232, Refs. 4, 24). They are provisional values chosen by least squares in this educational material. The results for CEM I are tis 3.6 h, tm 7.0 h, the main peak at 12.6 h with 2.2 mW/g, and cumulative heat release of 40.6, 44.0, and 82.0 J/g (at 6.2, 7.0, and 12.6 h); the gypsum runs out at 15.0 h, and the peak of the C₃A reaction rate is at 17.0 h. For ordinary cement (Table 3, gypsum 4.9 %) at w/c = 0.5 and 20 °C, the main peak is at 12.6 h with 2.55 mW/g, the gypsum runs out afterward, at 14.8 h, and the C₃A peak is at 16.8 h. The cumulative heat release is 330 J/g at 7 days and 402 J/g at 28 days, and the degrees of hydration of the phases (C₃S, C₂S, C₃A, C₄AF) are 0.65, 0.20, 0.74, and 0.28 at 7 days and 0.81, 0.41, 0.81, and 0.36 at 28 days; C₂S keeps increasing from 7 to 28 days. To match the measured cumulative heat release, the reaction of C₃A just after mixing (a0 = 0.19, time constant 0.37 h) was made somewhat large. The first hour of the measurement may also include a temperature disturbance caused when the sample was put in, so the coefficients for this part are especially uncertain (provisional). This a0 also determines how fast the gypsum decreases (for ordinary cement, 47 % is used for AFt in 30 min, 60 % in 1 h, and 79 % in 6 h, and it runs out at 14.8 h). The “remaining gypsum” in the readout of tab ② comes from this calculation and is a provisional value. Limitation of the model. This equation, in which a lot of gypsum is used just after mixing and only C₃A uses gypsum, cannot be transferred as it is to the compositions of real products. For example, if the composition of a product example (C₃A 9 %, SO₃ 2.10 %; Ref. 16) is entered, the gypsum runs out at about 8 h, and if the actual sulfate content of CEM I 42.5R (0.034 mol/100 g) is entered, at about 35 h (the measurement shows about 15 h). The amount of gypsum in the model (4.9 % for ordinary cement, 4.5 % for the others) is an effective value chosen to match the time at which it runs out through the reaction of C₃A alone (§6, Items to Confirm, item 6).

The heat release is Q = ΣfiHiαi (fi is the mass fraction and Hi is the phase-specific value in Table 4), and the overall degree of hydration is α = Σfiαi/Σfi. For simplicity, the reaction of C₃A after the gypsum has run out (the formation of AFm) is also counted with the same 1672 J/g as for AFt.

8.2 Temperature and the water constraint

θ is the “clock” of the reaction, and, including the effects of both temperature and water, it is advanced as

(8) dθdt= exp⁡[−ER(1T−1293.15 K)] ·1−e−wcap/WS1−e−(w/c)/WS

where E = 40 kJ/mol is the apparent activation energy recommended by CEMHYD3D (citing ASTM C1074, Ref. 4). While there is enough water (when the second factor is about 1, for example right after mixing or under water curing), this is the same as the age converted to 20 °C (equivalent age), and the rate is 0.41 times as fast at 5 °C and 2.2 times as fast at 35 °C. wcap is the capillary water available for the reaction (per 1 g of cement); it is w/c − 0.42α for sealed curing and w/c − 0.36α for underwater curing (§9), and for dry curing the evaporated water Eev is further subtracted. The curing (sealed, underwater, dry) and the temperature are decided before mixing and are not changed midway (if the sample were moved underwater midway, a separate calculation to replenish the evaporated water would be needed). For dry curing, dEev/dt = 0.06 h⁻¹ × wcap (provisional value) is used, so the capillary water leaves in about 1 day and hydration stops. The second factor φ is a “multiplier for the progress of the reaction due to water,” normalized to be 1 immediately after mixing (wcap = w/c). We set WS = 0.12 for sealed and dry curing. This is a schematic of the fact that when the capillary water decreases, the internal humidity drops (self-desiccation) and the reaction slows down even before the water runs out. For underwater curing, water keeps entering from outside, so we set WS = 0.02 so that the reaction does not slow down until just before the voids are filled with hydrates (both are provisional values). As a result, α (28 days) of ordinary cement at 20 °C is 0.62 for sealed curing at w/c = 0.30 (φ falls to 0.31), 0.71 for underwater curing, and 0.69 for sealed curing at w/c = 0.50. This expresses the tendency that the lower the w/c, the more the curing water matters (the amounts are provisional).

8.3 Measured example 2: Mochizuki 2016 (with and without a water-reducing agent)

Measured example 1 is the CEM I 42.5R of Jagličić et al. (2025) above. Example 2 consists of the heat-release curves that Prof. Mochizuki himself measured in his fourth-year undergraduate research (the bachelor’s thesis, and unpublished measurements of the same samples; Refs. 26, 27; used at Prof. Mochizuki’s instruction (2026-09-25)). The samples are research-grade ordinary Portland cement (“OPC” in the bachelor’s thesis) and commercial moderate-heat and low-heat cements, and we use the curves for the conditions without added limestone powder. The measurement used a sandwich-type conduction calorimeter (12 measurement points) and starts 3 h after the addition of water (the peak just after mixing was measured with a different instrument and is not used here). The conditions of the bachelor’s thesis are W/P = 0.32 and 0.2 mass% of a polycarboxylate-based water-reducing agent (dispersant), and the temperature is not recorded. The measurement without the water-reducing agent (April 2016, unpublished) has no record of W/P or temperature. In the heat-release graph of simulator tab ②, these three curves can be overlaid as dotted lines in the color of each type (the default is without the water-reducing agent; the original data were thinned out to every 15–30 min, and for the curve of the bachelor’s thesis the points of instrument disturbance at 3–4 h were removed).

Table 4a. Measured heat-release curves (Mochizuki 2016) and the modelThe measured values were read from the original data at 5-min intervals. The cumulative heat is from 3 to 72 h after the addition of water (not including 0–3 h). The model is for w/c = 0.32, 20 °C, sealed (calculated with Node.js)
Main-peak timeMain-peak heightCumulative 3–72 hCumulative ratio (ordinary = 1)
With water-reducing agent 0.2 % (bachelor’s thesis): research-grade ordinary / moderate-heat / low-heat22.2/29.3/30.0 h2.70/2.54/1.63 mW/g225/177/130 J/g1:0.78:0.58
Without water-reducing agent (unpublished): research-grade ordinary / moderate-heat / low-heat6.8/8.4/7.7 h4.02/2.78/2.04 mW/g233/198/143 J/g1:0.85:0.61
Model: ordinary / moderate-heat / low-heat12.6/12.7/12.7 h2.44/2.05/1.59 mW/g207/149/119 J/g1:0.72:0.57

How to read it. The water-reducing agent delays the peak by 15–22 h (this is the retardation caused by the polycarboxylate-based water-reducing agent). On the other hand, the cumulative heat up to 72 h does not change much (225 and 233 J/g for research-grade ordinary cement, 130 and 143 J/g for low-heat cement). For the curves with the water-reducing agent, therefore, we do not compare the times directly with the model, but look at the shape (induction period → acceleration → main peak → deceleration), the order of the types (ordinary > moderate-heat > low-heat), and the ratios of the cumulative heat. For the ratio of the cumulative heat, the model’s 1:0.72:0.57 roughly agrees with the measured 1:0.78:0.58 (with the water-reducing agent) and 1:0.85:0.61 (without) (the model is somewhat small for moderate-heat cement). The time of the main peak is nearly the same regardless of type for moderate-heat and low-heat cement, at 29.3 and 30.0 h with the water-reducing agent and 8.4 and 7.7 h without (the peaks without the water-reducing agent are flat over 7.7–9.0 h) (the commercial ordinary cement, which is not in Table 4a, also gives 28.8 h with and 9.1 h without), which does not contradict the fact that the model uses the same stage boundaries regardless of composition (§12). That only the research-grade ordinary cement is early is thought to be because its particles are finer (volume-based median diameter 11.8 µm; the three commercial types are 16.6–19.2 µm). The measurement without the water-reducing agent shows a peak that is earlier and higher than in the model. W/P and the temperature were not recorded, so the reason for this difference (for example, the possibility that the temperature was higher than 20 °C) is not known (Items to Confirm, item 11).

8.4 Fluidity — amount of water and water-reducing agent (apparent viscosity)

How easily the paste flows before it sets changes greatly with the amount of water. This is the flip side of lecture 01, slide 15, “reducing the mixing water reduces the voids in the hardened body and makes it stronger”: when water is reduced, the paste flows less easily and can no longer be packed into a mold. A water-reducing agent is therefore used to reduce the water while keeping the paste flowing as easily. In Prof. Mochizuki’s bachelor’s thesis (Ref. 26) and the measurements made in preparation for it, flow curves (shear rate versus shear stress) were measured with a stress-controlled rotational viscometer (double cylinder), changing the shear stress 0.54 → 200 → 0.54 Pa over 240 s (20 °C). The apparent viscosity is the slope (stress ÷ shear rate) of the straight line joining the origin and the 200 Pa point of the flow curve, that is, the viscosity when the paste is regarded as a Newtonian fluid. It allows even pastes with a yield value to be compared with a single number.

Table 4b. Measured water/powder ratio W/P and apparent viscosity (mPa·s, 20 °C)Mochizuki 2015–2016. “Without agent” (no water-reducing agent) is from the 2015 measurements, and the ordinary cement is a commercial ordinary cement (estimated). “With 0.2 % agent” (0.2 % water-reducing agent) gives the values of the bachelor’s thesis, and the ordinary cement is the research-grade one. “—” means not measured
W/PNo agent: ordinaryModerate-heatLow-heat0.2 % agent: ordinaryModerate-heatLow-heat
0.50(failed)(failed)58 (rough guide)———
0.40220268281———
0.35774505690———
0.32———211208198
0.3041351649does not flow291 (limit)326347
0.28————442 (limit)505
0.26—————976 (limit)

Without the water-reducing agent, reducing W/P from 0.40 to 0.30 raises the apparent viscosity several-fold to about 20-fold, and low-heat cement at 0.30 did not flow even when the stress was raised to 200 Pa (the yield value exceeds 200 Pa). With 0.2 % water-reducing agent, W/P 0.30–0.32 gives about the same viscosity (200–350 mPa·s) as W/P 0.40 without it, so the water can be reduced by about 20–30 % at the same ease of flow. Limit of the measurement. “Limit” is the smallest W/P at which the apparent viscosity could be measured in the bachelor’s thesis. With less water than that, the flow curve oscillates or the sample slips on the cylinder surface, and the values are not reliable. In addition, this viscometer reads water as 4.8 mPa·s (the true value is about 1 mPa·s), so values below 100 mPa·s are rough guides. The bottom graph of simulator tab ③ plots this table on a logarithmic scale and draws a vertical line at the current water/powder ratio. The model does not calculate fluidity (it only places the measurements side by side).

9. Powers’ Model — Voids and the Gel/Space Ratio

Powers and Brownyard divided the volume of cement paste into “unhydrated cement,” “hydration products (gel: solid and gel water),” and “capillary pores” (Refs. 14, 15). Taking the specific volume of cement as 0.32 cm³/g, the non-evaporable water wn = 0.23α (specific volume 0.72 cm³/g), and the gel water wg = 0.19α (specific volume 1 cm³/g), the hydration of 1 g of cement produces gel with a volume of

(9) Vgel=0.32α+0.72×0.23α+0.19α≈0.68α

The volume fractions relative to the total volume of the paste, w/c + 0.32, are

(10) φunhyd=0.32(1−α)w/c+0.32, φgel=0.68αw/c+0.32, φcap=w/c−0.36αw/c+0.32

The sum of the three is 1. The “unhydrated cement” in the readout of tab ② is calculated from the degree of hydration of each phase and the consumption of gypsum (the part used for AFt), so it differs slightly from the first of Eqs. (10) (for ordinary cement at w/c = 0.5: 0.356 at 1 h (0.371 by the equation), 0.345 at 6 h (0.365), and 0.110 at 28 days (0.120); the difference is at most 2 percentage points). The “hydration products” are the remainder (1 − unhydrated − capillary pores), so in the early stage they come out larger than in Eq. (10) (0.056 at 1 h versus 0.041 by the equation). X and the model strength are calculated with Eq. (11) (X at 28 days is 0.653; from the ratio of products to capillary pores it would be 0.657). In sealed curing, w/c − 0.42α of the capillary pores is filled with water, and the remaining 0.06α is empty voids created by chemical shrinkage (0.23α × (1 − 0.72) ≈ 0.064α). Hydration stops when the capillary water is gone, so complete hydration is possible in sealed curing only for w/c ≥ 0.42, and for underwater curing only for 0.36 or more (upper limit αmax ≈ (w/c)/0.42). Brouwers (2004) has re-evaluated this as about 0.39 for present-day CEM I (Ref. 14). In reality, in sealed curing the internal humidity drops and the reaction slows down before the water runs out (self-desiccation), so α at 28 days falls below this upper limit (§8.2).

Fig. 2. Volume fractions in Powers’ model (w/c = 0.50, sealed). As hydration proceeds, the water-filled capillary pores are replaced by gel

The fraction of the space made up of the gel itself and the capillary pores together that is occupied by the gel is called the gel/space ratio X.

(11) X=0.68α0.68α+(w/c−0.36α) =0.68α0.32α+w/c ,fmodel=234X3 MPa
How the model strength is treated 234X³ MPa is an empirical formula that Powers obtained for his own samples (cement paste) (original source not checked, Ref. 15); it is not the strength of mortar or concrete. In the simulator it is labeled “model strength,” and it is not displayed as if the measured strength were uniquely determined by the degree of reaction. That the model value of 67 MPa for w/c = 0.5 and α ≈ 0.7 is close to the 28-day strength of JIS mortar, 62–64 MPa (Ref. 16), should be regarded as little more than a coincidence.

Porosity and strength. The right-hand figure of lecture 01, slide 15 shows the relation between porosity (15–35 %) and compressive strength measured by Lian et al. (2011) for porous (permeable) concrete: the lower the porosity, the stronger the concrete, and the strength falls monotonically along a downward-convex curve as the porosity rises (the paper proposes a model of porosity and strength based on Griffith’s theory of fracture. We do not give the equation because we have not checked it in the original; Ref. 25). As the text of the slide says, reducing the mixing water reduces the voids in the hardened body, and the strength increases exponentially. Powers’ Eq. (11) also gives the same trend. For ordinary cement at 20 °C, sealed, at 28 days, the capillary porosity for w/c = 0.40, 0.50, 0.60, and 0.70 is 22.1, 30.6, 37.7, and 43.6 %, and the model strength is 95, 65, 46, and 33 MPa. Approximating this by f = f0 exp(−kp) (p is the porosity in %), a least-squares fit to the 7 points for w/c 0.40–0.70 gives k ≈ 0.049 /% and f0 ≈ 290 MPa (a derived value calculated in this educational material). The third graph of simulator tab ③ draws this relation. The materials (paste versus porous concrete) and the sizes of the voids differ, so the numbers cannot be compared with the figure on the slide and only the trend is the same. The points and the numerical values read from the figure on the slide are not used.

10. Setting — Connecting the Solids

Bentz (2006/2007) calculated the connectivity (percolation) of the solids with CEMHYD3D and compared it with the setting measured by the Vicat needle. He reports that at the time of setting only 4–8 % of the cement has hydrated, but the skeleton that joins the particles has already formed (Ref. 13). The wider the spacing between particles (the larger w/c), the more products are needed before they connect, so setting is delayed. Jagličić et al. (2025) also state that the initial set by the Vicat needle agrees well with the time tm of maximum acceleration and with the ASTM onset time tASTM (for CEM I 42.5R, tm = 7.0 h and tASTM = 6.2 h; Ref. 12).

Simulator tab ② regards the particles as connected when the C–S–H “fuzz” extending out from a particle reaches the neighboring particle, and defines the time at which the left and right edges of the window are connected as setting. In a two-dimensional figure the particles look farther apart than in three dimensions, so the length of the fuzz is set to κ = 33 times “the thickness when C–S–H is spread uniformly over the surfaces of all particles.” The amount of C–S–H is the volume obtained from the amounts of C₃S and C₂S that have reacted, using a molar volume of 108 cm³/mol (C₁.₇SH₄, Table 4). However, the C–S–H formed in the reaction just after mixing (the a0 term of Eq. (7)) is a thin layer on the particle surface and is not counted as reaching the neighboring particle. Only C–S–H is counted toward the amount of fuzz; the hydrates formed from C₃A (AFt and AFm) and the hydrates of C₄AF are assumed not to form part of the skeleton. Because C–S–H precipitates from the liquid and attaches to the surface of every particle, the fuzz is attached with the same length to all particles other than gypsum (including C₃A), and it is drawn in the inside figure with the same length on the same particles as in the setting criterion. κ is a provisional value chosen so that the setting of CEM I 42.5R comes close to tm = 7.0 h; the setting in the model occurs at 7.5 h (α = 7.5 %). The particles are placed with enough spacing that they do not overlap even when they are enlarged to match w/c = 0.30 (if the particles were touching from the start, they would become connected right after mixing when w/c is small). For ordinary cement at 20 °C, setting occurs at 6.3 h, 7.3 h, and 8.0 h for w/c = 0.3, 0.5, and 0.7, and the degrees of hydration at those times are 6.7 %, 7.7 %, and 8.4 %, which roughly agree with Bentz’s “4–8 %” (α includes about 3 % for the reaction of C₃A and C₄AF just after mixing; Appendix A.5). The setting time also changes with the arrangement of the particles (the random-number seed). With 20 different seeds, setting is spread over 6.1–7.0 h, 7.0–7.8 h, and 7.6–8.4 h for ordinary cement at w/c = 0.3, 0.5, and 0.7, and over 5.8–9.3 h for all four types together (standard deviation 0.14–0.41 h). The differences among the types at w/c = 0.5 (ordinary 7.3 h, moderate-heat 7.4 h, low-heat 7.8 h, CEM I 7.5 h) are about the same as this spread, so the model cannot distinguish differences in setting among the types of cement. Compared with the initial set of 2 h 10–15 min of a product example (Ref. 16; standard consistency, w/c ≈ 0.27), the model’s 6.3 h at w/c = 0.3 is about 2.9 times as late, so the setting time should be taken only as a guide (provisional).

11. Differences Between Gypsum and Cement, and Common Misconceptions

Table 5. How gypsum and cement set
Gypsum (calcined gypsum)Portland cement
Where it precipitatesIn the liquid (nucleation from the supersaturated liquid)Mainly on the particle surfaces (C–S–H covers the particles and grows outward)
ProductsDihydrate gypsum (needle-like to plate-like), one kindC–S–H (gel), CH (plates), AFt (needles), AFm (thin plates)
How it hardensThe needles interlock to form a networkThe products around each particle reach the neighboring particle and join the particles together
Time scale (model)Setting after about 10–20 min; the reaction is nearly complete in about 1 hSetting after a few hours; the reaction continues for weeks to years
Water needed for complete reactionw/p = 0.186w/c ≈ 0.42 (sealed)
Excess waterLeaves on drying and becomes pores (Eq. (6))Remains as capillary pores, which the hydrates fill little by little (Eq. (10))
Evidence that hardening is not “water evaporating” (1) The reaction releases heat (gypsum 132 J/g; cement about 330 J/g at 7 days). (2) Part of the water is taken up as water of crystallization or structural water and does not leave on drying (dried gypsum is 18.6 % heavier than the powder). (3) It hardens while plenty of water is still present (it has not evaporated). Both gypsum and cement set under water. (4) Conversely, if cement is dried, hydration stops and it remains weak (comparison condition F in simulator tab ②). However, it is known that gypsum becomes even stronger when the excess water leaves (the abstract of Lewry and Williamson 1994, Ref. 22). We treat “hardening” and “becoming strong by drying” as separate things.

12. The Simulator’s Model and Its Simplifications

Appendix A. What the Simulator (hydration-simulator-en.html) Shows

What this appendix covers What each tab of the paired simulator hydration-simulator-en.html draws, what each operation means, and by which equations and values the numbers are calculated. Literature values, empirical formulas, and provisional values (values chosen for teaching) are distinguished.

A.1 The whole page

There are three tabs: ① Gypsum (calcined gypsum + water), ② Cement (Portland), and ③ Compare how they set. In every tab, the upper left holds the apparatus figure (a container in a constant-temperature bath, a thermometer, a spatula, a needle), the upper right holds the schematic inside figure of the material, and below it come the graph, the legend, a one-line comment (what the current operation has changed), and the readout; on the right (below, on a smartphone) is the control panel. The apparatus figure, the inside figure, the graph, and the readout are all drawn from the same state (the result of the calculation). Appending #gypsum, #cement, or #compare to the address opens that tab; ?embed=1 gives a display for embedding with the heading and the introduction hidden; and writing ?only=cement shows only that tab (the row of tabs is also hidden). If you switch tabs or hide the screen during playback, the animation stops.

A.2 ① Gypsum (calcined gypsum + water)

Order of operations. “① Add powder” → “② Mix” (1 minute; while mixing, the powder disperses in the water in the apparatus figure and the inside figure) → “③ Advance time” (play/pause; there are also “+5 min” and “Jump to 120 min”) → “④ Dry” (40 °C, provisional). Each step carries over the state of the previous step. w/p (0.40–1.00), the type of powder, and the temperature (5–60 °C) can be changed only before the powder is added. After “④ Dry,” the apparatus figure changes to one showing the lump taken out of the mold (or, when dried before the network has connected and with dihydrate powder, which has not hardened, the powder spread on a dish) placed on a shelf of the drying oven (40 °C, provisional). “Reset” returns to the beginning, and the same conditions give the same result (the random-number seed is fixed). The time is compressed: 1 second of display corresponds to 2, 5, or 15 minutes of the experiment (selectable).

Inside figure. It draws hemihydrate grains (circles; when they dissolve, the smaller grains disappear first), ions in the liquid (Ca²⁺ as circles, SO₄²⁻ as triangles; the number is proportional to the CaSO₄ concentration of the liquid), and dihydrate needles (purple; needles that have joined the network connecting the left and right edges are dark purple). The scale is a rough guide (a 100 µm bar). Dissolution and reprecipitation around a grain are drawn according to the following conventions. (1) The background color of the liquid is painted according to the supersaturation with respect to the dihydrate, S = c/cDH (the color of water for S ≤ 1, and the darkest wisteria color at Smax; because S of the whole liquid approaches Smax early, the color fraction is taken as ((S − 1)/(Smax − 1))2.5, and the color bar at the lower right and the ▼ (the present S of the whole liquid) follow the same convention). (2) While a grain is dissolving, a color that corresponds to Smax (the solubility of the hemihydrate) at the surface and falls toward the concentration of the whole liquid going outward is overlaid in a band of width δ = 14 px around the grain (the two ends are quantities of the model; the distribution between them is schematic). (3) The outline of the original grain, r0, is left as a thin dashed line, and the edge of a dissolving grain (width 2.5 px) is drawn pale and made to ripple in small waves in proportion to the dissolution rate. (4) The number of ions leaving the edge of the grain (Ca²⁺ and SO₄²⁻ with white outlines and tails) is proportional to the dissolution rate; of these, the fraction “precipitation rate/dissolution rate” disappears toward the nearest needle within 30 px, and the rest scatter into the liquid (which becomes more concentrated). Positions are determined only by the time and the index number of each ion, so the same state gives the same figure, and the ions move only during playback. (5) Dihydrate nuclei are born in the dense liquid around a grain (at a distance of rL − 0.3δ to rL + δ from the center of the grain; a nucleus that would fall inside the grain is moved to just outside its edge) (§4.1). Because the grains are densely packed, the bias in the positions of the nuclei is hardly visible in the figure. In the calculation the liquid is uniform, and the distribution of concentration around the grains and the flow of ions are a schematic overlaid on that calculation. Graph. The upper panel shows the degree of reaction α and the heat release rate (132 J/g × dα/dt, in mW/g), and the lower panel shows the concentration of the liquid and the solubilities of the hemihydrate and the dihydrate (dashed lines). The range between the two dashed lines is the supersaturated range. Readout. The mass balance per 100 g of powder (hemihydrate, dihydrate, dissolved CaSO₄, free water, and the total; the total stays constant until drying), the water used for hydration (the water of crystallization in the solid minus the ½H₂O originally contained in the powder; it does not change after drying) and the water that evaporated on drying, the volume fractions, and the porosity after drying. After drying there is no liquid, so the concentration of the liquid and the supersaturation are displayed as “— (no liquid).”

Comparisons you can try in class (A–G under “Compare.” Choosing one shows the result calculated up to 120 minutes, and “Keep result” lets you overlay it with the next condition. “Clear previous” removes it): Compare the setting time and the porosity after drying for A Standard (w/p 0.60, 20 °C), B More water (1.00), and C Less water (0.40). With D Warm (40 °C), E Hot (60 °C), and F Cold (5 °C), see that warming does not always make things faster (at 60 °C the supersaturation is small and there are few nuclei). The one-line comment compares with the setting time at 20 °C for the same w/p and states from the result whether it is faster or slower (examples: at 50 °C there are fewer nuclei, but setting is 1.4 min faster than at 20 °C; at 60 °C it is 3.6 min slower). G Water on dihydrate powder: it does not become supersaturated and does not harden (a failure example). If you press “Dry” partway (for example at 10 minutes), unreacted hemihydrate remains (a failure example).

Table A1. Model values for the representative conditions of ① (20 °C)Calculated by calling the simulator’s functions directly in node (Appendix A.5)
w/pNumber of representative needlesTime at α = 0.5Time at α = 0.9Setting (α)Porosity after drying
0.4011127.0 min52.5 min13.6 min (0.18)33 %
0.608827.0 min52.5 min15.4 min (0.22)47 %
0.807327.0 min52.5 min17.4 min (0.27)56 %
1.006227.5 min52.5 min19.6 min (0.32)63 %

The progress of the reaction (the times at which α is reached) hardly depends on w/p; only the setting time and the porosity change with w/p. This is because the model makes nucleation proportional to the surface of the grains; in reality there are also reports that the reaction becomes slower the more water there is (Items to Confirm, item 3).

Table A2. Values used in the calculation of ①
QuantityValueCategory and source
Molar mass of CaSO₄, H₂O136.14, 18.015 g/molLiterature value
Density of hemihydrate, dihydrate / molar volume of dihydrate2.74, 2.32 g/cm³ / 74.21 cm³/molCEMHYD3D (Ref. 4). The density of β-hemihydrate may be slightly lower (to be confirmed)
Solubility (0–100 °C)Table 2Read from the figure of Azimi et al. 2007 (Ref. 1)
Heat of hydration132 J/g hemihydrateCEMHYD3D (Ref. 4)
Rate of dissolution Kd0.08 min⁻¹Provisional value
Nucleation J0, B40 needles/min, 1.7Provisional value (form of Eq. (4))
Growth gL, gW, ℓc12, 0.07 px/min, 40 pxProvisional value (Eq. (5))
Temperature coefficient of the rates E30 kJ/molProvisional value
Setting criterionNetwork connecting the left and right edgesSchematic (the idea of Morgado et al. 2019, Ref. 10). Checked every 0.2 min
Width δ of the dense liquid around a grain / position of the nuclei14 px / rL − 0.3δ to rL + δSchematic. The default seed is a sequence chosen to be close to the median setting time when the seed is varied in 80 ways (15.6 min at w/p = 0.60) (the default gives 15.4 min; the range is 12.4–22.4 min; §4.3)

A.3 ② Cement (Portland)

Order of operations. “① Add powder” → “② Mix” (3 minutes) → “③ Advance time” (“Next time point” steps through 1, 3, 6, and 12 hours and 1, 3, 7, and 28 days, and also stops at the initial set and the final set; there is also “Jump to 28 days”). The type of cement, w/c (0.30–0.70), the temperature (5–60 °C), and the curing (sealed/under water/drying) can be changed only before the powder is added. Time advances faster the later it gets (1 second of display ≈ 1.2 hours at first, and later 1 second of display corresponds to about 0.36 times the elapsed time so far). The three boxes above the inside figure show the stages of the stiffness of the paste, “Flow → Setting → Hardening,” the present stage, and the start time of each stage. The initial set is the time given by the setting criterion (the time at which the particles become connected, §10), and the final set is the time at which the needle penetration reaches 0.5 mm (the advance of the reaction clock θ since setting, 1.25 × (1 − √(0.5/39)) ≈ 1.11 h; interpolated within a time step). For ordinary cement at w/c = 0.50, 20 °C, sealed, Flow is 0–7.3 h, Setting is 7.3–8.4 h, and Hardening starts at 8.4 h (the Setting stage is 6.3–7.4 h at w/c = 0.30 and 8.0–9.1 h at 0.70). The stages of heat release (induction, acceleration, and deceleration periods) are a separate classification, and setting occurs partway through the acceleration period.

Inside figure (same appearance as the lecture figures; the default). It is drawn from the same state with the same conventions as the figure of lecture 01, slide 15. Unhydrated particles are gray angular polygons (5–7 vertices; the lightness varies slightly by phase; the area is πr²), the gypsum in the cement is a pale gray elongated shape, water is light blue, C–S–H, CH, AFm, and the hydrates of C₄AF are grouped into an ocher layer (with a pattern of short black fibers; the outer edge ripples by ±8 % for each particle), AFt is short dark-blue rods around the C₃A particles (the length follows from the amount of AFt), and empty voids are white (dashed). The layer is distributed to the particles in proportion to the reacted area (it fills the traces left by dissolution, and the remainder extends outward), and so that the area of light blue + white in the painted figure matches the capillary porosity (water + empty), a coefficient compensates for the overlap of layers, the part that extends outside the window, and the part where the AFt rods and the outer fibers lie over the water (counted on a 4 px grid and determined by bisection). Because thin fibers span more screen pixels the smaller the display magnification (device pixels per 1 px of the figure) is, the fraction covered by fibers (8 % at magnification 2 and 9.5 % at magnification 1; where rings overlap, it is increased according to the number of overlaps) and the number of outer fibers (0.7 times at magnification 1) are adjusted to the magnification. For the same reason, the area covered by the AFt rods is allowed for with a factor of 1.5 at magnification 1 and 1 at magnification 2 (linear interpolation in between). When the magnification is smaller than 1 (when the figure is small on a screen whose device pixel ratio is 1), the same line is extended down to magnification 0.5 (this is outside the range of the fit, so it is only a guide). The outline of a particle is drawn only on the inside of the polygon. Because the empty voids (white) are drawn beneath the layer and the particles, the circles are enlarged to allow for the hidden part, so that the ratio of the visible white to light blue matches capE:capW (when there is a lot of empty void because of drying, the same is done with light-blue circles). The white is drawn after the pale background of the outer fibers, so it takes on no color. When the background is white (empty voids) because of drying, only the fibers are drawn, without painting the pale background. The sparse C–S–H outside that (out to the same outer edge as in the setting criterion, R + the length of the fuzz) is drawn as pale ocher and thin fibers over the light-blue background, and is counted as water (capillary pores). On the outside of the cluster connected at setting, the background is made slightly darker so that the connection can be seen (the darkness stays within the range counted as light blue). The setting figure of the lecture is a schematic in which the hydrates are drawn thick, whereas at the model’s setting (for ordinary cement at w/c 0.50, the products are about 9 % of the volume) the ocher layer is hardly visible yet. This difference is stated in the one-line comment on the screen and below the three-stage figure. Because setting is judged with this outer edge, the connection in the figure and the setting time agree (before setting, the dense layer never extends beyond the outer edge used in the criterion). If you choose “Flow, setting, and hardening side by side (3 inside figures)” in the graph selector, the inside figures calculated under the same conditions for Flow (the middle of the induction period, θ = 2 h), Setting (the moment of initial set), and Hardening (the state at the displayed time if the displayed time is in the Hardening stage, otherwise 28 days) are placed side by side (vertically on a smartphone). Inside figure (color-coded by phase). Unhydrated particles are drawn with a different color and shape for each phase (C₃S hexagon, C₂S circle, C₃A square, C₄AF flattened pentagon, gypsum elongated rhombus). Every shape is sized so that its area equals that of a circle of the same radius (πr²), so the fraction of the unhydrated area agrees with the volume fraction in the readout (the setting criterion is evaluated with circles of the same area). The particles of each phase dissolve and shrink starting with the smallest; C–S–H is drawn as a shell covering the particle and fuzz extending outward (attached to all particles other than gypsum; fuzz that has joined the cluster connecting the left and right edges is dark blue), the hydrates of C₄AF as a light-brown shell, CH as hexagonal plates, AFt as red needles around C₃A, and AFm as thin plates. The background color represents the capillary pores, distinguishing water (blue) from empty voids created by chemical shrinkage or drying (white). The area fractions of the products and voids are also matched to the volume fractions in the readout (they deviate by the amount of overlap among products).

Graph. “Heat release rate and periods” shows the heat release rate over 0–72 h and the stages (initial, induction, acceleration, and deceleration periods), marks the main peak of an overseas measured example (Jagličić 2025) with ×, and shows the end of the induction period, 3.7 h, with a short red dotted line (with a label). These marks are values at 20 °C, so they do not move when the temperature is changed. With “Overlay measured data,” the heat-release curves of the research-grade ordinary, moderate-heat, and low-heat cements of Mochizuki 2016 (§8.3; from 3 h after adding water) are overlaid as dotted lines in the colors of the model for each type (ordinary orange, moderate-heat blue-green, low-heat purple; the model line for CEM I is dark blue). The choices are “None,” “no water-reducing agent” (default; unpublished, W/P and temperature not recorded), and “bachelor’s thesis” (W/P 0.32, water-reducing agent 0.2 mass%), and the line of the same type as the selected cement is drawn darker (the color of the model line also changes with the type). The conditions are written in the legend, and notes such as “The water-reducing agent delays the peak by 15–22 h, so do not compare the times directly with the model; look at the shapes, the order, and the ratios of the integrated heat” are written below the graph. At the end of the readout, the measured integrated heat release over 3–72 h (bachelor’s thesis 225/177/130 J/g, ratio 1:0.78:0.58; without water-reducing agent 233/198/143 J/g, ratio 1:0.85:0.61) is placed next to the values calculated with the model at w/c = 0.32 (20 °C, sealed), close to the measured W/P (207/149/119 J/g, ratio 1:0.72:0.57; calculated within the page). A brown dashed line appears at the time when the gypsum is used up. “Degree of hydration of each phase” shows α of the four phases and of the whole on a logarithmic time axis, together with the upper limit set by running out of water (sealed: (w/c)/0.42; under water, the upper limit at which the voids become filled: (w/c)/0.36). “Porosity and gel/space ratio X” shows the capillary porosity, X, and the model strength 234X³. “Flow, setting, and hardening side by side” is the three inside figures described above. When you press “Keep result,” the previous line is overlaid in gray on every graph and the conditions appear in the legend. The readout item “Factor for the progress of the reaction due to water, φ” is the second factor of Eq. (8).

Comparisons you can try in class (A–H): Compare the porosity and the model strength of A Standard, B Less water (w/c 0.3), and C More water (0.7) (in B, which is sealed, the capillary water decreases, “Water is running short” appears after about 6 days and the reaction slows down, and α (28 days) stays at 0.62. If you press “Keep result,” reset, set the curing to “Under water,” and calculate under the same conditions as B, it becomes 0.71. Both are strong because they have few voids). With D (5 °C) and E (35 °C), the time of the heat-release peak changes (Arrhenius). F: if it is dried, hydration stops and it is weak (a failure example). G: low-heat cement is rich in C₂S and releases little heat, and because it has little C₃A the gypsum remains until 28 days (readout), so the growth of α from 7 days to 28 days (0.41 → 0.59) is proportionally larger than for ordinary cement (0.54 → 0.69). H: compare with the measured main peak (×) using the composition of CEM I 42.5R.

Table A3. Coefficients of Eqs. (7) and (7a) (provisional values; the units of θ, ti, and τ are h, and k is in h⁻¹)For C₃S, C₂S, and C₄AF, a0 + A + B = 1; for C₃A, a0 + Bs + A + B = 1. The starting point of the 4th and 5th terms of C₃A is not ti but θsd of Eq. (7b). The 3rd term of C₃A has Bx = 0.494 and τx = 300 h. The Avrami exponent n is 2 or more
Phasea0τ0BsτstiAknBτDβD
C₃S0.0350.18——1.420.2230.07853.040.74272.50.567
C₂S0.0050.60——6.930.2150.0099420.7807391.58
C₃A0.1910.370.31530.2θsd0.05450.07092.130.4401110.329
C₄AF0.1060.50——2.250.2200.0080620.67411101.80
Table A4. Model values for ② (ordinary cement, 20 °C, sealed unless stated otherwise)Calculated in node. Setting is the time at which the left and right of the window become connected (a guide that is later than the measured initial set; §10)
w/cSetting (α)α (7 days)α (28 days)Capillary porosity (28 days)XModel strength
0.306.3 h (0.067)0.500.6213 %0.84140 MPa
0.30 (under water)6.2 h (0.067)0.550.717 %0.91178 MPa
0.406.8 h (0.072)0.530.6722 %0.7495 MPa
0.507.3 h (0.077)0.540.6931 %0.6565 MPa
0.708.0 h (0.084)0.550.7144 %0.5233 MPa
0.50 (drying)7.5 h (0.077)0.290.2948 %0.339 MPa

A.4 ③ Compare how they set

With the same water/powder ratio (0.40–0.70; w/p for gypsum and w/c for cement), the inside figures of gypsum and cement are placed side by side at three time points: “just after mixing,” “at setting,” and “after the reaction has progressed” (2 hours for gypsum, 28 days for cement) (the buttons carry the names of the stages of lecture slide 15: “Flow (just after),” “Setting,” and “Hardening (after reaction)”). The temperature is 20 °C, the cement is ordinary cement (Table 3) and sealed, and the inside figures of the cement have the same appearance as the lecture figures. The upper left of the graph shows the time to setting (minutes for gypsum, hours for cement; the scales differ) and the upper right shows the porosity after the reaction has progressed (for gypsum, after drying; for cement, the capillary porosity); both are values calculated for 0.40–0.70 in steps of 0.05. The lower graph shows “Capillary porosity at 28 days → model strength” (cement, the same 7 points) and its exponential approximation (f0 exp(−kp) in §9), and it highlights the point for the present w/c. As in the right-hand figure of lecture slide 15, you can see the trend that the fewer the voids, the stronger the material (the numbers cannot be compared; §9). You compare the difference between gypsum, in which “needles grow in the liquid and interlock,” and cement, in which “hydrates grown from the particle surfaces join the particles to one another,” and also the fact that in both, adding more water delays setting and increases the pores. The bottom graph shows the measured W/P and apparent viscosity (Mochizuki 2015–2016, Table 4b, logarithmic scale; the data without water-reducing agent were measured in 2015). The solid lines and ● are without water-reducing agent (W/P 0.50–0.30; the upward arrow at 0.30 for low-heat cement marks that it did not flow), the dashed lines and □ are with 0.2 % water-reducing agent (from W/P 0.32 down to the lowest value that could be measured; the limit point is marked with ×), and the colors indicate the type of cement. The vertical line is linked to the water/powder ratio slider (0.40–0.70), and the readout shows the values measured at that W/P without water-reducing agent and with 0.2 % water-reducing agent (if that W/P was not measured, it says so). Because most of the measurements lie between 0.26 and 0.35, the other slider, “Viscosity-graph vertical line W/P” (0.26–0.70), lets you move only the vertical line toward the low-water side (the model calculation stays at 0.40–0.70). The range with more water than W/P 0.50 has no measurements and is shaded lightly, and the region below 100 mPa·s is also shaded lightly as a guide. Look at lecture slide 15’s “reducing the mixing water makes it stronger” (the strength graph above) side by side with “reducing the water makes it flow less easily; with a water-reducing agent you can reduce the water while keeping it easy to flow” (this graph).

A.5 Checking the calculations

The simulator’s physics and chemistry functions are written within the page separately from the screen components (between MODEL-BEGIN and MODEL-END), and the following were checked by calling them directly in node.

Items to Confirm (with Prof. Mochizuki)

As of September 24, 2026, the experimental materials for gypsum and cement had not yet arrived, so the following points were made with provisional values and provisional treatments. In his reply of September 25, 2026, Prof. Mochizuki agreed that the cement measurements from his own undergraduate research (Refs. 26, 27) could be used, and they were used in the note of §6 and in §8.3 and §8.4 (resolved: item 6, the origin of the composition equations; unresolved: item 11).

  1. Is β-form (calcined gypsum) acceptable as the starting material for gypsum? If it were the α form (α-hemihydrate gypsum, also called hard gypsum or high-strength gypsum), it would need less water and give a stronger hardened body, but its dissolution and setting are generally slower than those of the β form (Abu Zeitoun et al., ZKG International 10/2020; Seiller et al. 2021, cited in the instruction sheet, is a paper on the α form).
  2. The water/powder ratio for gypsum (the default 0.60 is provisional), the time and method of mixing, the mold, the method of judging setting (Vicat needle, Gillmore needle, pressing with a finger, etc.), and the drying conditions (40 °C is provisional).
  3. Will the difference in reaction rate caused by the amount of water be observed in the lab class (in the model the progress of the degree of reaction hardly depends on w/p, and only the setting and the porosity change)?
  4. The temperature condition (room temperature or constant temperature). The change of the setting rate with temperature (fastest around 40 °C) is schematic and has not been checked against primary sources.
  5. The type of cement (is ordinary cement acceptable?), the water/cement ratio (the default 0.50 is provisional), the curing conditions and the ages to observe, and whether strength tests will be done (if so, the specimens and ages).
  6. The source of the mineral composition on lecture 01, slide 14 (the slide does not give one). From the original data of Prof. Mochizuki’s bachelor’s thesis (Ref. 26), the ordinary cement turned out to have values from Bogue’s equation (with SO₃ correction), and the moderate-heat and low-heat cements values from Taylor’s equation (without SO₃ correction) (§6). Should they be unified to one of the equations (the simulator keeps the values on the slide)? The amount of gypsum in the cement (4.9 % for ordinary and 4.5 % for the others are provisional values). In the model only C₃A uses gypsum, and it uses much of it right after mixing, so if the SO₃ 2.10 % (4.5 %) of the product example were used as it is for ordinary cement, the gypsum would be used up before the main peak (at 11.0 h). Real products are adjusted so that the gypsum is used up after the main peak, so for ordinary cement the ratio to C₃A was made almost the same as for CEM I (which, in measurements, is used up at about 15 h, after the main peak) (it is used up at 14.8 h). The actual sulfate in CEM I (3.5 % anhydrite, 1.2 % hemihydrate) is about 30 % more in moles, but since the part used for C₄AF and C–S–H is not counted, 4.5 % is used (a limitation of the model in §8.1).
  7. The citation on slide 15, “C. Liana, Y. Zhugeb, S. Beechama,” seems to be a misprint of C. Lian, Y. Zhuge, S. Beecham (Constr. Build. Mater. 25(11), 4294–4298 (2011), Ref. 25) (the affiliation marks a and b have been attached to the names; it is a study of the porosity and strength of porous concrete, not of the relation between the water/cement ratio and the strength of cement paste).
  8. What are the dark-blue needle-like crystals in the “Hardening” figure of slide 15 (ettringite or calcium hydroxide)? In this material the model’s AFt (ettringite) is drawn as dark-blue rods.
  9. If there are photographs (SEM, etc.) of the products from the lab class, we would like to match the shapes and sizes of the products in the inside figures to them.
  10. If a setting test of cement will be done, its method and the expected result (the setting in the model is matched to tm of the calorimetry of CEM I and is later than the Vicat initial set of 2 h 10–15 min of the product example). If heat release will be measured, its conditions (the time from mixing to the start of measurement).
  11. The temperature of the heat-release measurement of Mochizuki 2016 (calorimeter with 12 measurement points) and the W/P of the measurement without water-reducing agent (April 2016). Since there is no record, the screen and §8.3 say “W/P and temperature not recorded,” and the times are treated as a guide. Is it acceptable that the ordinary cement in the viscosity without water-reducing agent (2015) is a commercial ordinary cement (the research-grade ordinary cement has been used since 2016)?

The numerical values on this page were calculated from the equations and table values in the text. The rate coefficients and the setting criterion are provisional values for teaching and do not predict the properties of any particular product. All figures were drawn for this page and are not reproductions of figures from the literature.

Paired simulator: Gypsum and Cement — Why Do They Harden When Water Is Added?

  1. G. Azimi, V. G. Papangelakis, J. E. Dutrizac, “Modelling of calcium sulphate solubility in concentrated multi-component sulphate solutions”, Fluid Phase Equilib. 260, 300–315 (2007). — Solubilities in Table 2 (read from the figure).
  2. USGS, PHREEQC Version 3 documentation, Example 2. water.usgs.gov. — Solubility of the dihydrate at 25 °C, 15.1 mmol/kg.
  3. PHREEQC database llnl.dat (USGS). — log K and heats of formation of the hemihydrate and dihydrate. Y. Tang et al., ACS Omega 4, 7636 (2019). doi:10.1021/acsomega.8b03476
  4. D. P. Bentz, “CEMHYD3D: A three-dimensional cement hydration and microstructure development modeling package. Version 3.0”, NISTIR 7232 (2005). tsapps.nist.gov. — Densities, molar volumes, and heats of formation; heats of hydration of each phase; activation energy 40 kJ/mol; and heat of hydration of the hemihydrate, 132 J/g.
  5. J. W. Bullard, E. Enjolras, W. L. George, S. G. Satterfield, J. E. Terrill, “A parallel reaction-transport model applied to cement hydration and microstructure development”, Modelling Simul. Mater. Sci. Eng. 18, 025007 (2010). doi:10.1088/0965-0393/18/2/025007. — HydratiCA (the destination of the NIST link in the instruction sheet; a model separate from CEMHYD3D).
  6. S. J. Gurgul, G. Seng, G. R. Williams, “A kinetic and mechanistic study into the transformation of calcium sulfate hemihydrate to dihydrate”, J. Synchrotron Rad. 26, 774–784 (2019). doi:10.1107/S1600577519001929
  7. M. La Bella et al., J. Appl. Cryst. 56, 660–672 (2023). doi:10.1107/S1600576723002881
  8. J. Seiller, D. Bonnal, J. Adrien, S. Meille, S. Tadier, E. Maire, A. Bonnin, “4D in situ monitoring of the setting of α plaster using synchrotron X-ray tomography with high spatial and temporal resolution”, Constr. Build. Mater. 304, 124632 (2021). doi:10.1016/j.conbuildmat.2021.124632 (only the abstract was consulted)
  9. J. Adrien et al., Cem. Concr. Res. 82, 107–116 (2016). doi:10.1016/j.cemconres.2015.12.011 (only the abstract was consulted)
  10. G. Morgado et al., J. Cryst. Growth 507, 124–133 (2019). doi:10.1016/j.jcrysgro.2018.11.006 (only the abstract was consulted)
  11. I. Darmawan, O. Willy, J. A. Budiman, J. Dent. Res. Dent. Clin. Dent. Prospects 14, 167–170 (2020). doi:10.34172/joddd.2020.036
  12. A. Jagličić, T. Gädt, M. Hofmann, J. Therm. Anal. Calorim. (2025). doi:10.1007/s10973-025-14162-3. — Characteristic times of the heat-release curve and the mineral composition of CEM I 42.5R.
  13. D. P. Bentz, “Cement hydration: building bridges and dams at the microstructure level”, Mater. Struct. 40, 397–404 (2007). doi:10.1617/s11527-006-9147-3. — Degree of hydration at setting, 4–8 %.
  14. H. J. H. Brouwers, “The work of Powers and Brownyard revisited: Part 1”, Cem. Concr. Res. 34, 1697–1716 (2004). doi:10.1016/j.cemconres.2004.05.031
  15. T. C. Powers, T. L. Brownyard, “Studies of the physical properties of hardened portland cement paste”, J. Am. Concr. Inst. 18 (1946–1947). T. C. Powers, “Structure and physical properties of hardened portland cement paste”, J. Am. Ceram. Soc. 41, 1–6 (1958). (The original was not checked. The volume constants are from Ref. 14, and the strength equation 234X³ was checked in a secondary source.)
  16. Taiheiyo Cement, “Ordinary Portland Cement” product catalog (in Japanese); UBE Mitsubishi Cement, “Ordinary Portland Cement N” data sheet (April 2023) (in Japanese). — Product examples of mineral composition, SO₃, heat of hydration 332 J/g at 7 days and 384 J/g at 28 days, and JIS mortar strength.
  17. JIS R 5210 (Portland cement), ASTM C1074 (maturity method).
  18. “Inorganic Materials Science” (in Japanese), lecture 01, lecture slides 14 (history of cement and the mineral composition by Bogue’s equation) and 15 (hydration process of cement, and porosity and strength of concrete).
  19. H. F. W. Taylor, Cement Chemistry, 2nd ed., Thomas Telford (1997). — Abbreviations of cement chemistry and a general description of the products (not used for the numerical values of this material).
  20. M. Avrami, “Kinetics of phase change. I”, J. Chem. Phys. 7, 1103 (1939). doi:10.1063/1.1750380
  21. P. Freiesleben Hansen, E. J. Pedersen, “Maturity computer for controlled curing and hardening of concrete”, Nordisk Betong 1, 21–25 (1977). L. J. Parrott, D. C. Killoh, “Prediction of cement hydration”, Br. Ceram. Proc. 35, 41–53 (1984). (Neither original was checked. Only the form of the equation and the idea were borrowed.)
  22. A. J. Lewry, J. Williamson, J. Mater. Sci. 29, 5524–5528 (1994). doi:10.1007/BF00349943 (only the abstract was consulted)
  23. J. Li, S.-L. Zhang, “Finite-size scaling in stick percolation”, Phys. Rev. E 80, 040104(R) (2009). doi:10.1103/PhysRevE.80.040104. — Percolation threshold of line segments in two dimensions (not used directly in the calculations of this material).
  24. W. A. Gutteridge, J. A. Dalziel, “Filler cement: The effect of the secondary component on the hydration of Portland cement. Part I. A fine non-hydraulic filler”, Cem. Concr. Res. 20, 778–782 (1990). (The original was not checked. The degrees of hydration of each phase were read from Fig. 2 of Ref. 4.)
  25. C. Lian, Y. Zhuge, S. Beecham, “The relationship between porosity and strength for porous concrete”, Constr. Build. Mater. 25(11), 4294–4298 (2011). doi:10.1016/j.conbuildmat.2011.05.005 (bibliographic data checked against Crossref; content: only the abstract was consulted; figure not reproduced). — The source of the right-hand figure of lecture slide 15.
  26. Yasuhide Mochizuki, bachelor’s thesis, “Powder Characteristics of Cement and Fluidity of Concentrated Cement Pastes” (in Japanese), Department of Inorganic Materials Engineering, Faculty of Engineering, Tokyo Institute of Technology (February 2016). — Heat-release curves (W/P 0.32, water-reducing agent 0.2 mass%), W/P and apparent viscosity (Tables 4a, 4b), and the calculation of the mineral composition (§6). Prof. Mochizuki’s own materials (not public).
  27. Yasuhide Mochizuki, unpublished data (2015–2016; apparent viscosity without water-reducing agent (2015) and heat-release curve (April 2016) of the same samples as the bachelor’s thesis). Prof. Mochizuki’s own materials.