The water kinetics of superabsorbent polymers during cement hydration and internal curing visualized and studied by NMR

SuperAbsorbent Polymers (SAPs) can be applied as an admixture in cementitious materials. As the polymers are able to swell, they will absorb part of the mixing water and can then release that water back towards the cementitious matrix for internal curing. This is interesting in terms of autogenous shrinkage mitigation as the internal relative humidity is maintained. The mechanism is theoretically described by the Powers and Brownyard model, but the kinetics and water release still remain subject of detailed investigation. This paper uses Nuclear Magnetic Resonance (NMR) to study the release of water from the superabsorbent polymers towards the cementitious matrix during cement hydration. The release of water by the SAPs is monitored as a function of time and degree of hydration. The internal humidity is also monitored in time by means of sensitive relative-humidity sensors.

the SAP-entrained system, isolated macro pores are predominantly present compared to the finer capillary pore system present in the system without internal curing 21,22 .
For internal curing of cementitious materials by SAPs, the SAP itself -i.e. type and size amongst othersplays an important role. If the SAP is releasing the water too fast (before final setting), the total water-to-cement ratio would increase. If the SAP is releasing the water too slow, it will keep its water and will not provide it to the cementitious matrix for internal curing. The SAPs thus need to release their water at the right time during hydration of the cementitious material. Furthermore, the SAPs need to be homogenously distributed and small enough to ensure a proper surface area available for internal curing. The ideal swollen size is in the range of 100-200 µm 14,15 . A previous study 17 has investigated a SAP within this range for internal curing, together with a less ideal bigger SAP still able to partially mitigate autogenous shrinkage. These two different sized SAPs are used in this paper.
The water release and kinetics from the SAPs during hydration are an important parameter to investigate. Mönnig 23 visualised the densification of the cementitious matrix around a desorbing SAP particle when the material was hydrated. As the SAP particle releases its water, it will change the surrounding matrix and cause a shift towards finer pores 21,22 . The SAP particle shrinks and an empty macro pore remains as shown by means of neutron tomography measurements 24 . The water release seems to coincide with the first drop in relative humidity 25 . Schröfl et al. 26 showed the water release by means of neutron radiography measurements during hydration of the cementitious material.
Nuclear Magnetic Resonance (NMR) is a powerful technique to study the water amounts ( 1 H) in a material. Even other elements can be studied, such as 23 Na and 35 Cl 27 . It can be applied to study the cement hydration 28,29 and fluid transport in porous materials 30 as the hydrogen atom will have an interaction with the pore wall and as a result the signal will decay. The resulting relaxation rate is proportional to the pore structure 31 and can be quantified as well. Moisture profiles of porous materials during heating could be studied and conclusions could be drawn on how the water escaped the sample as a function of time 32,33 . The profiles can be used to study drying or saturation of a mortar sample 34,35 , the water release from a cement paste to a porous material 36 or for autogenous healing 37 . Only a few studies utilizing NMR relaxometry for water release by SAPs towards the cementitious matrix have been done. Friedemann et al. 38 showed that water is available during post-internal curing of cement pastes. Nestle et al. 39 studied the water balance in a cement paste with SAPs. The release of water from the SAPs towards the cementitious matrix started by the time of the acceleration period of cement hydration and lasted for two days. Still, the kinetics of water release of the SAPs towards the cementitious matrix need to be investigated in detail to fully understand the principle of internal curing and the effect the SAPs have on and during internal curing. By using the non-destructive NMR test, the differences in water states and pore size distribution can be investigated and compared in systems with and without SAPs.
In this paper, the water release by the SAPs and the change in water states are monitored by means of NMR tests as a function of time and degree of hydration. The results are compared to the theoretical model of Powers and Brownyard and supported by measuring the internal relative humidity.

Materials and Methods
In this section, the different studied mixtures are explained (2.2). These mixtures are chosen to study the release of entrained water by the SAPs for internal curing. For that, the swelling capacity of the SAPs and their swelling times need to be assessed first (2.1). Next, the NMR principle and apparatus together with the performed calculations to investigate the release kinetics from the SAPs to the cementitious matrix are explained (2.3). In the following part the hydration and setting properties are discussed in order to compare the obtained results with the Powers and Brownyard model (2.4). In the end, information about the internal relative humidity measurements is provided which will serve as a basis of reference of the mitigation of self-desiccation and the link to the NMR results (2.5).

SAP absorption characteristics and added SAP amounts.
To determine the amount of SAPs needed to absorb the amount of entrained water, the swelling capacities need to be known. This swelling capacity can be studied in different ways and with different methods. A recent review 40 combines all these tests and points towards two main tests which can be conducted: the tea-bag test 41 and the filtration test 13 . The latter was conducted in this research. The test fluids were demineralized water (pH = 6.5) and a cement filtrate solution. For the cement filtrate solution 100 g of cement was mixed in 1 l of demineralized water and subsequently filtered after 6 h. The filtered solution (pH = 12.8) was used for testing. The cement used was CEM I 52.5 N and the chemical composition can be found in Table 1.
For the filtration method 13 a known weight of SAP (m 1 ) is added to an excess of testing fluid (m 2 ). After 24 h the whole is filtered and the mass of filtered fluid is recorded (m 3 ). During the test, the filter paper was pre-saturated and the testing setup was covered to minimize possible carbonation and evaporation. The absorbed mass of testing fluid per gram of dry SAP equals: (m 2 -m 3 )/m 1 . The test was performed in tenfold (n = 10). The results are shown in Table 2. Two SAPs (obtained from BASF) were investigated. These were SAP A, a cross-linked copolymer of acrylamide and sodium acrylate (particle size 100.0 ± 21.5 µm (n = 100)), and SAP B, a cross-linked potassium salt polyacrylate (particle size 476.6 ± 52.9 μm (n = 100)). Both SAPs are bulk-polymerized and consist of irregular crushed particles.
To determine the swelling time, the vortex method was applied based on the one found in literature 42 . For this test, 100 g of demineralized water was added to a beaker. Then, a vortex was made using a magnetic stirrer (20 mm length at 400 rpm). From the already obtained absorption capacity, the specific amount of SAPs to absorb 100 g of demineralized water was added to the beaker. The time was recorded until the vortex disappeared (n = 10). This time served as a reference of the time of swelling.
The amount of mixing water absorbed was calculated from the comparison of the flow values of the different mixtures, following the Standard EN 12350-5. This influence on the flow values reflects the absorption of the SAPs in the mortar mixture 22,43 . These values need to be interpreted with precaution as the SAPs may absorb more water or release water prior to setting. The tests were conducted after 5 min of water addition to the cement. This corresponds to a time when the total swelling (in the order of seconds see Table 2) was already reached. Additionally, microscopic analysis was conducted to verify whether the formed macro pores had the expected size 22,44,45 . When assuming spherical SAPs, the swollen size can be calculated starting from the initial dry size and the amount of absorbed entrained water and compared to the found macro pores. In that way, the swelling capacity in the mortar could be determined.
The R0.30 mixture has a low water-to-cement ratio and is prone to autogenous shrinkage. Based on the model of Powers and Brownyard, the amount of needed entrained water in the SAPs for internal curing can be theoretically calculated and amounts to an additional water-to-cement ratio of 0.054 14 . In that way, the total water-to-cement ratio (w/c) tot equals 0.354, but with the entrained water-to-cement ratio (w/c) e it leads to an effective water-to-cement ratio (w/c) eff of 0.30.
Two different mixtures containing different SAPs were investigated. These were SAP A (cement paste Ae) and SAP B (cement paste Be). All SAPs were stored in dry and sealed conditions prior to testing or mixing in the cement paste mixture. The SAP particles were added on top to the cement and were first dry mixed to ensure a homogenous dispersion in the cement. After this dry mixing, the total water was added together with superplasticizer which was dissolved in the water prior to addition. The amount of superplasticizer in the SAP mixtures was the same as for the R0.30 mixture, to minimize the influence on the setting properties of the respective mixtures 46 . The amount of SAP to be added, after conducting the above-mentioned tests based on the mixing water absorption in section 2.1 was 0.22 m% (mass-% of binder weight) SAP A (Ae) and 0.45 m% SAP B (Be), respectively. There was no loss in workability observed.
Nuclear Magnetic Resonance (NMR) setup. The sealed sample containers (cylinder ∅ 27 mm × 100 mm high) filled with the cement pastes were put in the NMR setup as shown in Fig. 1. An external magnetic field B 0 was applied of 0.8 T corresponding to a frequency of 34 MHz. This field was provided by a water-cooled  iron-cored electromagnet with poles 50 mm apart from each other. A coil was placed around the sample for creating and receiving the radiofrequency fields during an NMR measurement. A Faraday shield was added between the coil and the sample to suppress the effect of the changes of the dielectric permittivity by variations of the moisture content. In that way, the NMR measurements are made quantitative. All NMR measurements were performed at room temperature (23 ± 1 °C). NMR is a magnetic resonance technique in which the magnetic moments of the nuclei are manipulated. The frequency of the condition is given by the Larmor frequency f 1 : Where B 0 is the externally applied static magnetic field and γ is the gyromagnetic ratio. The gyromagnetic ratio depends on the type of nucleus and for 1 H γ/2π is equal to 42.58 MHz·T −1 . Hence NMR can be made sensitive to only hydrogen. The magnetic field gradient was set at 0.3 T/m offering a one-dimensional resolution in the order of magnitude of 1.0 mm. In a so-called pulsed NMR experiment the magnetic moments of the hydrogen nuclei are manipulated by Radio Frequency (RF) pulses at the resonance frequency. The amplitude of the so-called Hahn spin-echo signal is proportional to the hydrogen density. Moreover the spin-echo signal also provides information about the rate at which the magnetic excitation of the spins decays. For the moisture signal, a Hahn-spin echo signal is obtained with signal intensity S equal to: where k is a proportional constant, ρ is the density of the nucleus, T 1 the spin-lattice relaxation and T 2 the spin-spin relaxation. The T 1 reflects the relaxation due to interactions of spins with the surrounding matrix, whereas T 2 reflects spin-spin interactions. Both the T 1 and the T 2 can give information about the distribution of the pore sizes in a porous material 34 . The repetition time T R and the echo time T E are also important parameters. The spin-echo signal S is recorded after a time T E equal to 180 µs. The relaxation time T 2 of water in a porous material can be as short as 500 µs 32 and S needs to be recorded before it has decayed. T R is the time in between two successive spin echo experiments and typically T R > 4·T 1 is chosen. The Hahn spin-echo was always determined prior of testing to decide for the parameters to be used. The following parameters were used: 29.422 MHz center frequency, 30 µs pulse time, 180 µs echo time and 64 averages. In this study T 2 was used to determine the pore-size development during hydration which was measured using Carr-Purcell-Meiboom-Gill sequences (CPMG). For CPMG following additional parameters were used: 1000 ms repetition time and 128 echoes for R0.30 and R0.354. For Ae and Be these values were 2000 ms and 2048 echoes, respectively. The resulting measured relaxation represents a complex summation of decaying signals. The data analysis of such signal involved a Laplace inversion to obtain a distribution of relaxation times or diffusion constants. In this paper, Fast Laplace Inversion (FLI) 47 was used to obtain the relaxation times T 2 as a spectrum.
All signals were normalized to the first signal to receive the percentage. The weighed percentage of each peak was then multiplied with the normalized signal to receive the relative ratio of each studied peak as a function of time. This was also studied as a function of the degree of hydration (see next section) and plotted on the For the SAP-mixtures, a comparison was made between the signal of the SAP entrained water and the free water to see when and how the water is released by the SAPs for internal curing.
Furthermore, T 2 is proportional to the pore size distribution due to the following relation 30,31,34 : where V is the volume, S is the surface and ρ 2 surface relaxivity. Assuming spherical pores with a diameter d, the volume-to-surface ratio is equal to d/6. Müller et al. 28 used planar pores of area A where S = 2·A, V = A·d and the volume to surface ratio is equal to d/2. They used a surface relaxivity of 3.73·10 −3 nm·µs −1 for a white Portland cement. The latter two are also used in this research. From the obtained T 2 spectra the peaks are located and multiplied by the volume-to-surface ratio and divided by the surface relaxivity.
Setting and hydration modelling. The Vicat needle test following the Standard ASTM C191 -08 (Method A: Reference Test Method using the manually operated Vicat apparatus) was used to determine the final setting of the mixtures. By periodically penetrating (n = 3) the thin cement paste with the 1-mm Vicat needle, the time between initial contact of cement and water and the time at which the needle did not leave a circular impression in the paste surface, was recorded and used as the time of final setting. This time of final setting is often used as the starting point of conduction autogenous shrinkage measurements 17 and is used to give referential information of the time of water release by the SAPs. Via the chemical composition (as found in Table 1) the mineral composition was calculated using the Bogue equations. The model of Parrot and Killoh 48 was used to determine the degree of hydration. The model uses the water-to-cement ratio and the mineral composition of the cement. In that way, the degree of hydration was obtained. This degree of hydration was then linked to the age of the sample during testing and this was used to link it to the intensity of the NMR signal. In that way, the model of Powers and Brownyard could be verified based on the obtained NMR data starting from the initial known amounts of water and the respective densities.
The maximum degree of hydration α max was also theoretically calculated from Powers and Brownyard's model using equations 4-5 20, 49 . When there is internal curing, equations 5-6 is used with the effective water-to-cement ratio or equations 4-5 with the total water-to-cement ratio.
where ρ c is the specific volume of cement. For R0.30 α max is 0.72 and for R0.354, Ae and Be α max is equal to 0.85.
The obtained values were used to visualize the theoretical model of Powers and Brownyard.
Internal relative humidity measurement. To see whether the SAPs were able to mitigate self-desiccation (and thus autogenous shrinkage), the internal relative humidity was monitored. Information on the effects of autogenous shrinkage mitigation of the two SAPs can be found in Snoeck et al. 17 . The internal relative humidity was monitored using sensitive relative humidity sensors. These humidity and temperature probes were obtained from Vaisala (HMP110). They have a measurement range of 0% to 100%RH and −40 °C to + 80 °C. The probe has a Vaisala Humicap 180 R sensor with an accuracy of ± 1.5%RH. The temperature sensor is a Pt1000 RTD Class F0.1 IEC 60751 with accuracy of ± 0.2 °C. Together with the probes, a protective lid system was used to protect the probes and to maintain the internal conditions of the sample near the probe.
The cement paste was poured in a small cylinder (∅ 14 mm × 85 mm high) and the probe was immediately placed hovering above the paste (Fig. 2). The protective assembled lids were used and the connection between the cylinder and the assembly, and the top was sealed with parafilm to exclude moisture escaping from the setup. In that way, a sealed environment was obtained and the relative humidity and temperature could be monitored as a function of time. The complete setup was placed on a metal grid in such a way that it was vertical for the complete measuring time.
To cancel the effect of the hydration heat -i.e. a small increase in temperature such as the heat of hydration may shift the RH measurements to false readings -the setup was submerged in a water bath placed at constant 20 ± 1 °C.
Data availability statement. The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.

Results and Discussion
In this section, the relaxation distribution and the percentage of signal intensities are discussed first (3.1) to provide information on the kinetics of SAP entrained water release for internal curing. Next, the results are linked to the Powers and Brownyard model (3.2) and the calculated pore size distributions compared to existing models (3.3). In the end, the effectiveness of an SAP for internal curing is discussed and supported by means of additional internal relative humidity measurements (3.4).

Relaxation distribution and percentage of signal intensity.
When a cement paste is hardening, the water will be consumed and products will be formed causing the material to become denser. This is seen in Fig. 3 showing the T 2 relaxation distribution profiles for all studied specimens. They are shifted downwards as a function of the degree of hydration to visually show the change in time and to compare the obtained results. The distinct peak in all systems is attributed to free water. In the reference samples, we can clearly see a shift in relaxation time from the 10 −3 s to 10 −4 s range as a function of the degree of hydration. Furthermore, a smaller peak is formed at higher degree of hydration in the 10 −3 s range. This latter peak may be attributed to some larger capillary pores which are visible after hardening of the cement paste due to the decrease in intensity of the other predominant peak.
In the SAP systems, additional peaks are allocated to SAP-entrained water in the 10 −2 s to 10 0 s range. This water is stored in the SAP particles and can be released for internal curing. In that way, as the T 2 is different, the water kinetics from the SAPs towards the cementitious matrix could be studied in detail as a function of time and as function of the degree of hydration. For the SAP A mixture (Ae), it is clear that water is still present in the SAP inclusions at later ages (after final setting); i.e. at higher values of the degree of hydration in the lower positioned curves. For SAP B (Be), this is not entirely the case and the signal for entrained water drops as a function of time and thus as a function of the degree of hydration. From this relaxation time spectrum, it is already clear that the SAP A system seems to be promising for internal curing, as for the SAP B systems, water is consumed quickly.
The signals were then transformed towards percentages of the initial signal intensity. In that way, a gradational graph was obtained showing the height of the different peaks relative to the initial signal intensity. These results are shown in Fig. 4. The largest peak from the free water has the largest portion of the total signal fraction. In time, this water is used and consumed for cement hydration causing the total signal intensity to drop in time. Initial setting occurred at approximately 8 h of age and final setting was at approximately 11 h of age as determined with the Vicat needle test. The time of initial setting corresponds to the drop in NMR intensity signal. From this point onwards, the free water was used more visibly in NMR in all studied samples.
After final setting, the SAP-entrained water was visibly consumed, i.e. for internal curing. The signal intensities for SAP A decreased more steadily compared to the signals for the SAP B mixture. Here, water was more quickly consumed and used in the overall system. As SAP A is ideal to mitigate autogenous shrinkage and SAP B Figure 2. Used relative humidity and temperature probe to monitor the internal environment of a sample container (cylinder ∅ 14 mm × 85 mm high) filled with the cement paste with protective assembly and used thermocouples to record the ambient temperature in the water bath and the surrounding climate chamber. only mitigates to a minor degree 17 , already a glimpse of the mechanisms governing internal curing can be seen by the decrease in T 2 spectrum peaks.
By comparing the relative signal intensities of the entrained water in the SAPs and the free water and dividing both by the total intensity, values of 84.3% and 85.2% free water and 15.7% and 14.8% entrained water are found for the SAP A and SAP B system, respectively. This is comparable to the theoretical values when dividing the amount of entrained water by the total amount of mixing water; 5.4/35.4 = 15.3%. The effective water-to-cement ratio is hereby 0.30, the entrained water-to-cement ratio 0.054 and the total water-to-cement ratio 0.354. This means that the amount of free water not present in the SAP is 30/35.4 = 84.7%. There is a good correlation between these theoretical values and the values obtained with NMR. The amount of SAP was determined in such a way that the absorption as determined by the flow values, autogenous shrinkage mitigation and by means of microscopic analysis was the correct one 17 .
As the cementitious matrix is formed at the time of final setting, the respective macro pores were found by image analysis on polished cross sections of hardened cement paste. The size was the one (257 µm for SAP A and 981 µm for SAP B) which could be expected when comparing the swelling capacity (mixing water in Table 2) and the initial dry size of the SAP particles. However, from this test the real behaviour of the SAP, i.e. faster release of entrained water from SAP B compared to SAP A is not detectable. This caused the SAP B polymer to realize less autogenous shrinkage mitigation, as reported in literature 17 . This shows the strength of using NMR to determine the real water kinetics from the SAPs for internal curing.
Link with Powers and Brownyard model. The found signal intensity results were plotted on the theoretical lines described in the model of Powers and Brownyard 20 as a function of the degree of hydration and can be found in Fig. 5 both as mass and volume fraction. The consumption of free water for cement hydration can again be clearly seen.
The obtained NMR results correspond nicely to the lines proposed by Powers and Brownyard. Recently, results found in literature 28,29 suggest that the free water line is not a straight line as stated in the model. A more concave downward curve is expected as found by means of NMR on cement pastes with white cement. These results correspond with the plotted curve in this research from a degree of hydration of 0.2 upwards. A more steady change in behaviour prior to 0.2 degree of hydration is found in this research. On the top horizontal curve, some noise is found due to small observed peaks in the spectrum as found in Fig. 3.
From the results of the SAP mixtures plotted on the theoretical Powers and Brownyard's lines, the entrained water stored in SAP A completely follows the theoretical line. For SAP B, this is also the case, but from a degree of hydration of 0.3 onwards, the curve slightly moves downwards. Part of the entrained water is thus prematurely released towards the cementitious matrix for internal curing. It is also possible that the detachment of the swollen SAP from the faces of the macro pore can cause a decrease in signal, but as the water spin-spin mechanism is governing in this measurement, this only has a minor influence. The release of entrained water from SAP B was also clearly visible as disappearing peaks in the T 2 spectrum found in Fig. 3d. Pore size distribution and link to existing models. The water content in the pores is found in the T 2 spectrum intensities and can be studied with time during hydration of the cement paste. T 2 is proportional to the pore size distribution 30,31,34 by multiplying the index of the found peak with the surface relaxivity and assuming planar pores. The pore size distribution results are found in Fig. 6. In time, a decrease in pore sizes is observed. This represents the densification of the cement paste due to hydration.
For the R0.30 mixture the main pore sizes found are in the range of 1.5-2 nm followed by an amount of pores with size of 8-12 nm, representing gel pores. Furthermore, some bigger capillary pores (10 to 1000 nm) are found as well. The microstructure of mortar studied by NMR in literature shows small (<10 nm) gel pores and bigger (10 to 1000 nm) capillary pores 34 . The T 2 values are also comparable to the values found by Müller et al. 28 . The latter addressed the found pores of 2.5 nm to gel pores, 8 nm to interhydrate pores. This is also the range of pores we found. The size of the pore containing the free water stabilizes at around 8-12 nm.
The model of Feldman and Sereda 50 shows mainly interlayer water and physically adsorbed water but no larger intrinsic reservoirs in C-S-H. They describe C-S-H as quasi-continuous layers with nanometer sized irregularities. The found pore size distribution also fits the Jennings Colloidal Model (CM-II) for C-S-H 51,52 . He proposes three different pore size categories. The first are the intraglobular pores (<1 nm). The second are the small gel pores that exist outside the globules (<3 nm) and the third are the larger gel pores (3-12 nm). It appears that the first and the second category are found in the spectral array of T 2 around the 1.5-2 nm scatter as obtained by NMR. The third category is seen as the pores with sizes in the range of 8 nm to 12 nm.
Clear peaks in T 2 were assigned to the entrained water in Fig. 3, even though the sizes are not corresponding with the macro pore size of the swollen SAP particle (around 257 µm in the swollen state within the cementitious material) if a calculation is made starting from the obtained T 2 (<1000 nm) and as shown in Fig. 6c,d. A reason stated in literature is the intrusion of solutes from the cement pore water in the SAP which might lead to enhanced relaxation due to paramagnetism and/or the formation of precipitates in the SAP 38 . With a Hahn spin-echo experiment of pure SAP in cement filtrate solution, we could not observe such T 2 relaxation times. Furthermore, as the signal does not decrease substantially in time, these reasons could not occur. The alkaline environment in the SAP is continuously changing 41 and no visible precipitates are found when looking at SAPs in macro pores in polished sections. Another reason for the observed SAP T 2 peaks could be that water molecules in the SAP can have diffusive exchange with water molecules at the interface with the cementitious matrix 39 . As water diffusion is almost not impeded in a SAP particle, as studied by means of a single Hahn spin-echo test, this is the main explanation for the found peaks in the T 2 spectrum. As the SAP particles are irregular in size, a scatter in this diffusive exchange is found.

Effectiveness of an SAP for internal curing.
To study the kinetics of the SAPs for internal curing, the relative signal intensities for the entrained water and the free water could be compared. The results are shown in Fig. 7. In this figure, the overall theoretical relation is also depicted. This is 30/35.4 divided by 5.4/35.4 or 30/5.4 = 5.56. This is the theoretical ratio found by using the Powers and Brownyard's model. However, as stated above in section 3.2, the line for the free water is not straight. The ratio, however, was used for a general way of comparison of the different results found in this research.
When looking at the results found for SAP A mixtures (Fig. 7a), the SAP A seems to release little amount of entrained water in the beginning (approximately at 11 h and 0.13 degree of hydration), after final set. Then the free water is consumed and at some point (approximately at 22 h and 0.30 degree of hydration), the entrained-water signal in the SAPs is decreasing again (till approximately 30 h and 0.38 degree of hydration), and the overall ratio is moving towards the theoretical one. For SAP B (Fig. 7b), this is not the case. The SAPs seem to release their stored water prematurely before initial setting (approximately at 8 h and 0.07 degree of hydration), to re-absorb part of the free water before final setting (approximately at 10 h and 0.12 degree of hydration) and again release it more quickly before the more pronounced decrease in free water (approximately from 13 h till 20 h and 0.19 till 0.30 degree of hydration). In the end, the ratio is again the same, and comparable to the SAP A mixture. However, as part of the entrained water is already released, there will be less mitigation of autogenous shrinkage or maintaining of the internal relative humidity.
When studying the internal relative humidity in Fig. 8, a different trend for each series is observed. After setting, the RH in reference samples starts to drop. This drop is more pronounced for a mixture with a lower water-to-cement ratio. The relative humidity keeps decreasing as a function of time and the rate for R0.30 and R0.354 after seven days is the same. For the SAP mixtures, the trend is different. For Be there is a partial drop in RH followed by a constant RH. This means internal curing is taking place, but not to its full extent. This could be due to the large particle size of SAP B and thus the lower surface area available for internal curing as compared to the smaller SAP A. Important is the inter-particle spacing as the water transport takes place within 2 mm from   the internal reservoir containing the internal curing water 53 . The ideal size of the SAP for internal curing is within the range of 100 µm to 200 µm 14 , which is more the case for SAP A (257 µm in swollen state within the cementitious matrix) while it is much higher for SAP B (981 µm in the swollen state within the cementitious matrix). The smaller SAPs are better distributed throughout the cementitious matrix and are effectively causing internal curing (Ae). The internal RH does not decrease too much after 28 days. The measured RH at 28 days is 96%, 90%, 80% and 76% for Ae, Be, R0.354 and R0.30, respectively. These results are comparable to the curves found by means of neutron radiography 54 . The resolution used in the latter research was, however, too low to really study the water release by the SAPs towards the cementitious matrix.
The results are also substantiating the results found with NMR. As confirmed by the results plotted on the model of Powers and Brownyard (Fig. 5), SAP A is ideally maintaining the internal RH and thus mitigating autogenous shrinkage. For SAP B, it partially maintains the internal RH and thus mitigates autogenous shrinkage partially. The quick drop in RH for the SAP B mixtures seems to correspond to the observed quick drop in signal intensity for the entrained water stored in the SAP B particles (Fig. 3).
These results correspond to autogenous shrinkage measurements on the same mixtures within corrugated tubes 17 . In the latter results, the R0.30 mixture and R0.354 mixture showed autogenous shrinkage. Most shrinkage was found in the R0.30 system as the cementitious matrix is denser and there thus is an increase in self-desiccation due to the higher hydrostatic tension forces in the capillaries. For Ae the autogenous shrinkage was completely mitigated. In the paper 17 it was opinioned that SAP B was releasing its water too slow. In this paper, it is shown that it releases its water too fast, from final setting onwards.

Conclusions
Based on the findings of the research on internal curing by SAPs as studied by means of NMR following conclusions can be drawn: • The free water and the entrained water by the SAPs could clearly be distinguished in the T 2 relaxation spectra.
The SAP signals relate to the exchange of their entrained water molecules with water molecules at the interface with the cementitious matrix. The signal for SAP B entrained water decreased more quickly compared to the SAP A entrained water signal. • The amount of mixing water absorbed by the SAP obtained from comparing the flow tests and microscopic analysis is the same as the amount determined by means of NMR. • The results fit to the model described by Powers and Brownyard 20 . The free water line, however, shows a more concave downward type of curve as a function of the degree of hydration. The results also fit the Feldman and Sereda model 50 and the Jennings Colloidal Model (CM-II) for C-S-H 52 . • The main nm pore size distribution was found in the range of 1.5-2 nm followed by an amount of pores with size of 8-12 nm. These can be attributed to the gel pores and the interhydrate pores. • SAP A is preferable considering the more effective mitigation of self-desiccation and internal curing compared to SAP B. SAP B seems to prematurely release its stored water after time of final setting, while SAP A gradually releases its entrained water in time, thus maintaining the internal relative humidity more effectively. • The study of the internal relative humidity led to the same conclusions. SAP A is ideal for internal curing and thus the mitigation of autogenous shrinkage.