The nitrogen adsorption desorption isotherms for the two types of samples (well-carbonated and non-carbonated) are presented in Figure 1. We observe that the quantity of adsorbed nitrogen on the carbonated sample is more significant than on the non-carbonated one.

Regarding the pore distribution curves (Figure 2), we observe that the carbonation results in a decrease in the micropore volume and an increase in the mesopore volume with radius around 2 nm. For the mesopores with radius larger than 2 nm, the increase in their volume is insignificant. The carbonation of portlandite is manifested by the crystallisation of numerous calcite crystals on the portlandite plates [13] . Thus, it is understandable that the carbonation results in an increase in the specific surface area (Table 1) and a modification of the pore network. However, the increase in the specific surface is very slight and can be considered insignificant. This is probably due to the insignificant increase of the mesopore volume of CEM I cement mortar. This result shows clearly that the cement materials based on CEM I and composed cements (CEM II, III, etc.) behave differently during carbonation: the increase in mesopore volume is more significant in composed cement materials than in CEM I cement materials.

While the decrease in volume of micropores is attributed to the formation of CaCO₃ which clogs the pores, causes of the increase in the volume of mesopores is still discussed. According to Eitel [14] , the increase in mesopore volume is caused by the porous structure of the silica gels that are formed during the carbonation.

Swenson and Sereda [15] reported that the increase in mesopores is caused by cracks in the CaCO₃ gangue that surrounds the portlandite crystals. Other authors have attributed the increase in mesopores to carbonation shrinkage [16] . Our Scanning Electron Microscopy (SEM) images of the cement mortars before and after carbonation show clearly apparition of micro cracks due to carbonation shrinkage (Figure 3).

The porosity accessible to water before and after carbonation determined using hydrostatic weighing [17] are presented in Table 2. We observe a clear decrease in the total porosity. This result is therefore opposite in comparison with the slight increase in the cumulative pore volume obtained by the nitrogen adsorption (Figure 4). Both techniques seem to cover different porous domains.

According to Belie et al. [6] , the size of nitrogen molecules (radius 0.21 nm) does not allow them to access the microporosity, whereas the water molecules (radius 0.1 nm) can enter these micropores. Hence, the results of nitrogen adsorption provide information mainly about the mesoporous domain.

In the domain of nano and micro pores, the calcium carbonate formed during carbonation obstructs the pores, by consequence, the water accessible porosity decreases. Some investigators believe that a major fraction of water molecules taken up by the sample enters spaces between the C-S-H rather than to be adsorbed on the existing surface. Hence, the results of porosity accessible to water provide information mainly about the nano and micropores. The decrease in the volume of micropores inferred from porosity accessible to water was also con-

firmed in Figure 2 where we presented the pore size distribution calculated from nitrogen adsorption desorption isotherms.

The carbonation Reactions (1), (2) shows that a quantity of CO₂ was captured to give CaCO₃ as a product of carbonation. Also, water released by calcium hydroxide and C-S-H on carbonation may aid the hydration of the unhydrated cement. For this reason, all the mass were measured at dry state to reflect only the gain in mass of the solid phase. An electronic scale was used to measure the mass increase of the specimens due to CO₂ uptake. CO₂ uptake was determined by the initial mass and the final carbonated mass as shown in Equation (3):

We observe in Figure 5 that the mass increases in a continuous manner. On the other hand, the apparent volume of the specimens remains constant during carbonation. Therefore, we deduce an increase in the density of the mortar after carbonation, which is confirmed by the measures presented in Figure 6. Hence, the carbonation leaded to a densification of the microstructure.

The propagation of CO₂ in the cement mortar was revealed by spraying phenolphthalein solution onto the fresh surfaces of samples. We observe in Figure 7 that the carbonation rate in an accelerated carbonation process is much more rapid than in natural carbonation where it can take several years for a penetration of just several mil- limetres [1] .

Furthermore, the calcite formed during carbonation covers the crystals of portlandite, and thus slows the car- bonation rate because it becomes more difficult for CO₂ to reach the portlandite. This explains why we observe the most important rate of propagation after the first 7 days of carbonation in comparison with the results at 14 and 32 days.

Although the samples were protected laterally in order to execute an axial carbonation, as a very active gas, the CO₂ still penetrated from the sides as shown in Figure 7. The carbonation depth was thus measured mostly in the middle of the sample in order to eliminate the effects of two-dimension carbonation. Moreover, the bot- tom of the sample is denser than the top of the sample due to the segregation of aggregates during preparation of cement mortar, which results in a more important carbonation depth at the top than at the bottom of the sample.

In Figure 8, we observe that the carbonation propagation is not always a linear function with the square root of the duration of carbonation: [1] , where A is a constant taking into account both the composition of the cement material (water/cement ratio, type of binder…) and the environmental conditions (relative humidity, temperature, pressure…). This simple linear modelling of carbonation depth is therefore not always applicable. This is because the coefficient A depends actually on the porosity which decreases during carbonation (Figure 9), therefore A decreases and the slope of the graph in Figure 8 changes.

Measures of gas permeability were performed with the experimental setup as described in Figure 10.

Gas permeability is measured using Helium gas according to recommendation standard RILEM TC 116-PCD [18] . Apparent permeability (K_a) is calculated from the Hagen-Poiseuille equation for laminar flow of a com-

pressible fluid through a porous body under steady state conditions according to formula (4).

where; K_a: apparent permeability to gas of the specimen (m²) at fixed pressure (in our case at 1, 2, 3 and 5 Bars); L: length of the sample (m); Q: measured gas flow (m³/s); A: cross-sectional area (m²); μ: coefficient of viscosity of the gas (Pa.s); P_i : applied absolute pressure = upstream pressure (Pa); P_atm: atmospheric pressure = downstream pressure (Pa). The intrinsic coefficient K is obtained by the intersection of the line connecting K_a values in function of 1/P_i with the coordinate axis.

Figure 11 presents the evolution of the gas intrinsic permeability during carbonation. Contrary to what we expected, an increase in the gas intrinsic permeability was observed. The results seem to be in conflict with the decrease of the porosity. However, the water porosity decreases after carbonation means that the totality of pores decreases, but we have no specific information about the meso and macro pores. By combination with the pore size distribution calculated from nitrogen adsorption, we have concluded that the carbonation resulted in an increase in the volume of mesopores at the expense of the volume of micropores. Therefore, it seems that the evolution of the gas permeability during carbonation is largely influenced by the changes in mesoporous domain: the increase in the volume of mesopores is the cause of the increase in the gas intrinsic permeability. The gas permeability was not influenced by the decrease in the volume of micropores.

It is also important to note that the increase in the gas permeability is not really significant. This result is in perfect correlation with the insignificant increase of the mesopores volume with radius larger than 2 nm. Indeed, we only observed a clear increase in the volume of pores with radius around 2 nm.

We present in Figure 12 the evolution of the thermal conductivity of CEM I mortar during carbonation. One measure was taken at 65% of relative humidity, another was taken when the specimens were dried. We observe an increase in the thermal conductivity as a function of the carbonation duration. Due to the low thermal conductivity of the air, the thermal conductivity varies with the density [19] . Hence, the increase in the thermal conductivity during carbonation is coherent with the decrease of the total porosity.

The results show that the thermal conductivity at dry state is smaller than that obtained at 65% relative humidity. For building materials, it is common to use the following equation to show influence of the relative humidity on the thermal conductivity:

where k is a coefficient of dimension, λ₀ is the thermal conductivity of dry material, and H is the relative humidity in percentage. The values of k were calculated and presented in Table 1. We observe that k increases and remains stable after carbonation. k is a characteristic coefficient which is unique for each material, it is therefore understandable that k remains stable once the cement mortar is carbonated (see Table 3).

The evolution of the thermal diffusivity is similar to the case of the thermal conductivity. Effectively, we observe also an increase in the thermal diffusivity (Figure 13). These results show that the carbonated cement mortar is more sensible to heat transfer than the non-carbonated one.

We present the evolutions of longitudinal (V_L) and transverse (V_T) ultrasonic velocity in Figure 14 and Figure 15, respectively. These ultrasonic velocities increase continuously when carbonation occurs. Because the ultrasonic velocity in the air is smaller than in the dense material, the observations show an increase in the density, which is coherent with the results observed earlier. The characteristics of ultrasonic wave propagation in a material can provide valuable information on material properties, microstructure and damage state. These methods have many advantages: ease of implementation, ability to work with one side of the material, can pass through large thicknesses, obtaining immediate results of measurements. Furthermore, we can calculate the Poisson’s ratio and dynamic modulus of elasticity from ultrasonic velocities as follow.

Poisson’s ratio [20] :

Dynamic modulus of elasticity [20] :

with ρ is the density of the material.

Figure 16 shows that the Poisson’s ratio remains constant before and after carbonation. In contrast, the dynamic modulus of elasticity increases after carbonation (Figure 17). These observations show a positive influence of the carbonation on the cement based material, of course without taking into account the corrosion of rebar. This positive effect of the carbonation was already applied in the concrete industry, especially for the non reinforced concrete structures.

Helium pycnometry was used to determine the solid phase volume of samples before and after carbonation. For comparison, the helium pycnometry analysis was also performed on samples which were subjected only to natural carbonation at 20˚C and 65% relative humidity without additional CO₂. The results are presented in Figure 18. We can see clearly that the accelerated carbonation resulted in a significant increase in the volume in comparison with the natural carbonated sample. The increase in the volume of the solid phase can be explained by the formation of CaCO₃ during carbonation, because the carbonation of one mole of portlandite leads to an increase in volume of 4 cm³ [21] , and the carbonation of one mole of C-S-H leads to an increase in volume of 12 cm³ [22] or 39 cm³/mol [2] . Due to differences in the volume, the carbonation product (CaCO₃) clogs the pores, thereby decreasing the porosity.