Quantitative evaluation of cement paste carbonation using Raman spectroscopy

Carbonation induced corrosion is one of the major durability issues for reinforced concrete structures. To address this issue, it is essential to understand the underlying mechanism of carbonation by detecting the ingress of the CO2 and carbonation depth quantitatively. This paper demonstrates the potential of Raman spectroscopy, as a powerful technique, to implement quantification analysis of cement paste carbonation. In present work, the correlation between the content of the CO32− and the Raman peak intensity was initially established by using several standard specimens premixed with certain amount of CO32−. The established correlation was then used to investigate the carbonation of the cement paste in a CO2 diffusion scenario, and the results show a good agreement with that obtained by thermogravimetry measurements. Therefore, the feasibility of the Raman spectroscopy to quantify the carbonation degree of the cement paste can be demonstrated.


INTRODUCTION
As one of the major degradation types, carbonation of cement paste has attracted extensive attentions with attempts to understand the underlying mechanism. Portland cement is one of the most common construction materials around the world, benefiting from its excellent mechanical property and relatively well durability. Hydration of the Portland cement mainly produces portlandite (Ca(OH) 2 ) and other calcium-bearing phases like C-S-H 1 , which contributes to the development of the mechanical properties and alkalinity of the material. These properties would, however, be affected by the carbonation, which is one of the major degradation issues for cementitious material 2 . Essentially, the carbonation begins with the ingress of the CO 2 into the pore solution to form carbonic acid which mainly reacts with Ca (OH) 2 (Reaction 1) and calcium-bearing phases (such as C-S-H (Reaction 2)) to form CaCO 3 , resulting in the neutralization of the alkaline environment 3 . The consumption of the hydroxides would certainly reduce the pH and potentially initiate corrosion of the steel reinforcement 4,5 . On the other hand, the precipitation of CaCO 3 may form a denser structure than uncarbonated cement 6 . To assess the above effects, it is of great significance to precisely follow the carbonation evolution so that the carbonation induced corrosion and alteration of the mechanical property could be followed precisely. Thereby, development of a quantitative analysis technique is attracting increasing attentions.

Ca OH
ð Þ 2 þ CO 2 ! CaCO 3 þ H 2 O: (Reaction 1) To explore the carbonation evolution, several techniques have been well documented by former researchers, such as phenolphthalein indicator 7,8 , X-ray diffraction (XRD) 9 , thermogravimetric (TG) analysis 10,11 , and Infrared spectroscopy (IR spectroscopy) 8,12 . Spraying the phenolphthalein indicator on the carbonated cement is one conventional method to assess the carbonation depth 8 and it is recognized as the most convenient method to capture the color change boundary between the carbonated and uncarbonated concrete where the pH is around 9 13 . Though phenolphthalein indicator is capable to give a continuous picture of the carbonated and uncarbonated zone, it can only identify the degree of carbonation above/below 50% (i.e., the pH is higher/lower than 9) 7 and the less carbonated location is impossible to be identified. In addition, XRD and TG are also widely employed to investigate the carbonation depth by characterizing the hydrated and carbonated products 9 . In principle, XRD is capable to investigate the crystal structure of carbonated products and thus determine the carbonation degree and carbonation front, and TG is dependent on the different decomposition temperature of Ca(OH) 2 (420-600°C) 14 and CaCO 3 (440-800°C) 15 . Both XRD and TG could achieve quantitively analysis and have provided valuable information to assess the carbonation of the concrete 10,16 . However, both XRD and TG are traditional bulk analysis techniques 17 , and powder sample should be prepared prior to each test. As it is almost impractical to obtain powder at different depth at a microscale, high resolution spatial analysis on the carbonation depth is thereby rarely achieved, which is sometimes of significant importance. Apart from the above techniques, infrared (IR) spectroscopy 8 and Raman spectroscopy have also been used to investigate the carbonation of cement paste, benefiting from their capability to follow identify the carbonated products in accordance with the characteristic peak of the C-O stretching bond. These two techniques are discussed below.
Raman spectroscopy, a scattering based technology with high chemical fingerprint sensitivity, is a complementary technique to IR spectroscopy 18 . Raman spectroscopy has attracted increasing attention since it was first employed to characterize the cementbased materials in 1976 19 , which was followed by several systematical work on the spectrum of the main compounds of the cement 20,21 . In addition, Raman spectroscopy was also extensively applied to characterize of the deterioration products of cement-based materials under carbonation, sulfate, and alkali-silicate reactions attacks 22 . Overall, Raman spectroscopy has demonstrated its potential to study both hydration and deterioration of cement-based materials in the literature. In comparison with IR spectroscopy, almost no preparation is required, and thus the negative effects induced by sample preparation could be minimized 23 . Furthermore, the spectra of the CaCO 3 , as the main carbonation products, obtained by Raman spectroscopy exhibits a much sharper peak corresponding to the C-O vibration modes in CO 3 group than that obtained by IR spectroscopy 24 . The Raman spectroscopy is thus more sensitive than IR spectroscopy to determine the carbonation evolution. Therefore, there is a few attempts to use Raman spectroscopy to evaluate the carbonation evolution and the obtained carbonation maps showed relative high resolution [25][26][27] . For instance, Torres-Carrasco et al. 27 investigated the weathering effect on the deterioration of the anhydrous cement by using Raman spectroscopy, which main focused on the carbonation effect. In addition, Raman spectroscopy has followed the carbonation depth of carbonation reaction of lime paste successfully 26 . The ability of Raman spectroscopy to evaluate the spatial distribution of the carbonation products of Portland cement was further confirmed by Corvisier et al. 25 . It can be seen that using Raman spectroscopy to investigate the distribution of carbonation products in cement has shown great potential 25,26 , the quantitative analysis is still absent due to the lack of the correlation between the Raman intensity and content of the carbonation products.
Aim of the current work is thus to achieve the quantitative analysis of the carbonation products in hydrated cement paste by using Raman spectroscopy. To achieve this aim, standard reference specimens premixed with different dosages of CO 3 2− were prepared and tested with Raman spectroscopy. Thereby, correlation between content of CO 3 2− and intensity of the Raman characterization peak was established. To verify the accuracy of the established correlation as well as the feasibility of the quantitative analysis of the carbonation evolution, an investigation into the carbonation of the Portland cement was also completed. The obtained correlation was then used to characterize progress of carbonation in hardened cement paste exposed to an accelerated carbonation scenario, so as to explore in-depth mechanism of carbonation. Meanwhile, the most commonly used techniques, i.e., XRD and TG, were used to determine distribution of CaCO3 in the tested samples to further verify accuracy and suitability of the Raman mapping technique in studying progress of carbonation.

RESULTS
Building up the database for quantitative analysis Prior to the quantification analysis, a criterion to evaluate quantify the carbonation of cement paste is essential to be established. As the main carbonation product in cement paste is CaCO 3 , the carbonation can thus be linked to the content of the CO 3 2− directly 3 . A typical Raman spectrum was obtained from analytical grade CaCO 3 powders, as shown in Fig. 1. It can be seen that the spectrum present a strongest peak at 1084 cm −1 , which should be assigned to the symmetric stretching of the CO 3 group (ʋ 1 ) in CaCO 3 as depicted in the spectrum 28 . Meanwhile, three minor peaks can also be identified at 281, 711, and 1435 cm −1 , which are associated with Ca-O lattice vibration mode, double degenerate in-plane bending mode and asymmetric stretching mode, respectively [28][29][30] . However, only this strongest peak can be observed when there is a small amount of CO 3 2− . Consequently, using the intensities of the peaks at 281, 711, and 1435 cm −1 as for quantification analysis might be not sensitive enough to identify the paste with low CO 3 2− content. To improve the sensitivity to quantify CO 3 2− with a minor content, the most intense peak at 1084 cm −1 was used as indicator to prove the existence of carbonated minerals and evaluate the carbonation degree in all the analyses 26 . Figure 2 shows the experimental results of the Raman spectroscopy measurements on the surface of the type (1) samples which were premixed with different dosages of CaCO 3 . As explained above, three radom regions were selected in each sample and the images of these regions under optical microscope are displayed in Fig. 2a. It can be seen from the optical microscope images, mixes with a higher CO 3 2− content presented a brighter and homogeneous color appearance, which is consistent with the color of CaCO 3 powder 31 . Figure 2b presents color mapping results revealing intensity of the 1084 cm −1 Raman peak (ʋ 1 CO 3 ) in the areas listed in Fig. 2a. Based on the differences of the mapping results, it can be easily derived that the intensity of the 1084 cm −1 Raman peak is associated with amount of the CO 3 2− compounds. Figure 2c presents a box chart revealing the correlation between the Raman peak intensity and the CO 3 2content of the test samples. Based on these data points, a linear relationship between CO 3 2− content and intensity of the 1084 cm −1 peak is clearly illustrated as red dash line, which is similar to correlation between intensity of Raman peak and content of other compounds, e.g., salicylic acid and pharmaceutical compound, as reported elsewhere 32,33 . Therefore, the obtained fitting line, i.e., (where x is the CO 3 2− content and y is the intensity of the Raman peak at 1084 cm −1 ) with an R 2 of 0.96, can be used to convert Raman spectra signal to CO 3 2− content in hardened cement paste samples. In such manner, potential of using the 1084 cm −1 Raman peak for quantification of CO 3 2− content in hydrated cement paste is thus proved.
Verification on feasibility of the established correlation to evaluate progress of carbonation To validate the accuracy and feasibility of the above established correlation for the carbonation process in a CO 2 diffusion scenario, an attempt was made to use Raman spectroscopy to quantify the carbonation evolution of the cement paste under accelerated carbonation. TG measurements was also carried out to validate the quantification results obtained by Raman spectroscopy.
As stated, the Raman mapping measurements was applied on slice samples sliced from type (2) samples after 0d, 1d, 3d, 7d, 14d, 28d, and 56d of accelerated carbonation. At the depth ranging from 0 to 10 mm, Raman mapping analyses were conducted, and results were collected. By using the correlation obtained from Fig. 2c, statistic results from the 1084 cm −1 Raman peak can be converted to the distribution of CO 3 2− content, as presented in Fig. 3. From the CO 3 2− content mapping results, evolution of the carbonation process and the development of carbonation at certain location can be clearly observed.
As can be seen in Fig. 3, no CO 3 2− was detected at 0 day, which confirms that good preparation and condition procedures were carried out prior to the accelerated carbonation. It can also be found that carbonation has started from the exposure surface, and the  3 2− converting from the intensity of Raman peak at 1084 cm −1 by using the correlation obtained from Fig. 2c. Fig. 4 The concentration of CO 3 2− at different depth. An area graph illustrating the increase of the CO 3 2− content with the time under accelerated carbonation against the distance to the exposure surface (the CO 3 2− content at each depth is the averaged results of 8 points and the line was smoothed by Origin TM ).
carbonation front gradually increases with the time of accelerated carbonation. Although the distribution of the CO 3 2− can be evaluated in Fig. 3, the exact quantities of the CO 3 2− at all depths can be hardly compared so that the depth of the carbonation front and fully carbonated zone could be determined. Therefore, an area graph is depicted in Fig. 4 to present the contents of the CO 3 2− at all depth for the specimens. The carbonation front could be easily found by looking at point where the CO 3 2− content reaches 0. The depth of the carbonation front has been labeled in Fig. 4 and recorded in Table 1. Based on the carbonation front, the diffusion coefficient could be calculated according to Eq. (2) (Where X c is the depth of carbonation front, k is diffusion coefficient and t is time). As the values of X c and t are both available in Table 1, the diffusion coefficient can be calculated as 2.37. For the fully carbonated zone, the threshold of CO 3 2− content should be established. In the literature, the content of the CO 2 in the fully carbonated cement paste has been reported to range from 22% to 26% 9,10 . To convert CO 2 content to CO 3 2− content, Eq. (3) was employed (Where C CO 2À 3 is the content of CO 3 2− ; C CO2 is the content of CO 2 ; M CO 2À 3 is molecular mass of CO 3 2− and M CO2 is molecular mass of CO 2 .). Based on calculation, the CO 3 2− content should be between 30 and 36% for fully carbonated cement paste. The depth of the fully carbonated zone is thus defined as the point where CO 3 2− content exceeds 30%, which has been labeled in Fig. 4 and listed in Table 1. Furthermore, the overall carbonation degree could be evaluated by the total CO 3 2− content within the specimen, which can be calculated by ratio of the area for each specimen to the whole area (100%*10 mm) as depicted in Fig. 4. The results have been listed in Table 1.
The Raman spectroscopy has demonstrated its capacity to detect the distribution of the CO 3 2− content over the sample surface, and this information can be used to evaluate the carbonation evaluation. To verify the above results, either comparing the similar results in the literature or carrying out complementary experiment is needed. However, it is almost impossible to compare the results with that reported in the literature because of the variation in experimental condition, such as the pretreatment, water/binder ratio and material, which has been proved to affect the carbonation depth significantly 34 . Complementary experiment, thermogravimetry analysis (TGA) was thus performed.
The TG measurements, as a conventional technique to quantify the content of CO 3 2− , have been performed to verify the Raman results. Figure 5 depicts the TG/DTG plots of cement paste specimens after 0, 1, 3, 7, 14, 28, and 56 days of accelerated carbonation. As explained earlier, DTG plots are employed to determine the decomposition intervals by inspecting the beginning and end of the peaks 35 . It can be seen that there are three peaks in Fig. 5, which are around 100, 410, and 700°C. According to the temperature required for the decomposition of possible compounds in the carbonated cement paste specimens, these three peaks should be attributed to the evaporation of free water (100°C), dehydration of Ca(OH) 2 (410°C) 14,36 and decomposition of CaCO 3 (700°C) 10,37 . As the purpose of TG measurements is to verify the quantification results, the peak for the decomposition of CaCO 3 is analyzed only.
In the DTG curve, the peak at around 700°C starts from the temperature 440°C and continues till 800°C, which is owing to the different polymorphs of CaCO 3 9,38,39 . The weight loss corresponding to the decomposition of CaCO 3 is thus evaluated between 440 and 800°C, and the exact ranges are determined according to the actual starting and ending points of the peak at each curve. As the weight loss results represent the weight of CO 2 , which are then converted to the CO 3 2− by Eq. (3). To compare the total CO 3 2− content obtained by TG and Raman measurements, they are   plotted in Fig. 6. As can be seen that the content of CO 3 2− obtained by TG and Raman measurements are almost the same. Therefore, the quantification results obtained by Raman measurements could be verified.

DISCUSSION
It might be worth noting that Raman results report a slightly less quantity of the CO 3 2− content, especially for the specimen after 56 days of accelerated carbonation, which might be caused by the different polymorphs of the CaCO 3 as explained above. It has been well-established that there are three polymorphs of CaCO 3 , i.e., calcite, aragonite and vaterite, and calcite is known to be the main carbonation products 40 . Both calcite and aragonite pronounce a strongest peak at around 1084 cm −125 . As explained in previously, quantitative analysis in present work is based on the intensity of the peak at 1084 cm −1 , the existence of aragonite might not affect the quantification results. However, for vaterite, the main peak splits into a doublet at 1074 and 1090 cm −130, 41 . The presence of vaterite would thus potentially induce slight underestimation of carbonation degree. To verify the existence of vaterite, XRD measurements was performed for verification.
The XRD patterns of the cement paste specimens after 0, 1, 3, 7, 14, 28, and 56 days of accelerated carbonation have been depicted in Fig. 7. The assignments of the peaks have been well documented and labeled above the patterns. Among all the composition detected in the carbonated cement paste specimens, two polymorphs of the CaCO 3 were found, i.e., calcite and vaterite. The presence of calcite could be confirmed by a strongest peak at 29.4°2θ, along with many weak peaks as marked in Fig. 7 42,43 . Vaterite has its strongest peak at 27.0°2θ as well as other 4 peaks at 24.9°, 32.8°, 43.9°, and 50.0°2θ 9,37 . Therefore, the presence of the vaterite could be confirm. Based on the peak intensities, the quantity of the vaterite is neglectable before 3 days, which gradually increases after 7 days. In the meantime, the amount of calcite, as the main carbonation products, rises much more dramatically as shown in the XRD pattern (Fig. 7). Therefore, the CO 3 2− content results obtained by Raman spectroscopy is representative to assess the carbonation degree, although the existence of vaterite may induce the minor underestimation.
Overall, in present work, the correlation between the Raman peak intensity and the content of CO 3 2− were established Initially by a linear equation: y ¼ 49500x. The established correlation was employed to study the accelerated carbonation of the cement paste in a CO 2 diffusion scenario. The results clearly show the distribution and quantity of the carbonation products. In addition, the accuracy of the obtained quantification results by Raman spectroscopy were confirmed by TG results. Overall, the feasibility of employing the Raman spectroscopy for quantitative analysis of the cement paste carbonation has been fully demonstrated.

METHOD Material and sample preparation
In this study, two types of samples were manufactured in lab, viz. (1) cylindrical hardened cement paste, with a dimension of φ20 mm × 10 mm, as depicted in Fig. 8a, and (2) a φ28 mm × 150 mm cement sample as depicted in Fig. 8b. The previous group of samples involving eight groups of hardened cement paste samples containing 0%, 5%, 10%, 20%, 30%, 40%, 50% and 60% of CO 3 2− , respectively, which was achieved by adding analytical grade calcium carbonate during mix. Two groups of type (1) specimens were prepared to improve the reproductivity of the experimental results, namely group A and group B. To facilitate a reasonable hydration of cement and a meaningful carbonation speed, a water/binder (w/b) ratio of 0.45 was designated for all the eight mixes. For the latter type of sample, the cement Fig. 7 Characterization of the chemical composition of carbonation products. XRD pattern of cement paste specimens after 0d, 1d, 3d, 14d, 28d, and 56d of accelerated carbonation. After cast, all the samples were cured in a vacuum dryer for 24 h to avoid carbonation. After curing, type (1) samples were ready for the Raman spectroscopy measurements; while type (2) samples were placed in a carbonation chamber with a condition at~74% RH, 20% CO 2 , 20°C. Prior to carbonation of the type (2) samples, two ends of the specimen were sealed to make sure that only the side of the cylinder was exposed to CO 2 . At the age of 0, 1, 3, 7, 14, 28, and 56 days of the accelerated carbonation, as depicted in Fig. 8c, a 20 mm long cylinder sample was cut off to test the development of carbonation with Raman spectroscopy, TG and XRD. It is worth pointing out that the slice obtained at 0 day was used to verify that no carbonation happened during the sample preparation.

Raman spectroscopy
Raman spectra were acquired by using a Reinshaw inVia Raman spectroscope, which is equipped with a 532 nm laser and a charged coupled device detector. Laser was focused via a Lecia ×50 (N.A. = 0.55) into a spot size of 1.40 µm in diameter on test samples. To avoid possible decomposition of carbonation products under applied energy, power of the tests was limited within 0.82 mW. All the spectra were recorded with an exposure time of 10 s and two accumulations to improve the signal-tonoise ratio.
For the type (1) sample, three regions with an area of 150 × 60 µm were selected randomly form group A and group B, named A0-A2 and B0-B2. In each region, 35 spectra forming a matrix of 7 × 5 were collected, as depicted in Fig. 9a. For the sliced disc specimens, the Raman spectroscopy was collected from a threadlike testing area ranging from surface to a depth of 10 mm and covered a width of 70 µm. In each of this area, 640 spectra forming a matrix of 8 × 80 were collected and the intervals of the analyses for the parallel and vertical directions were 10 and 125 µm, respectively, as shown in Fig. 9b.
The Raman spectra obtained from cement are always accompanied with unwanted sloping background, which may be caused by the intrinsic fluorescence as well as the incorporation of optics 45,46 . The background is thus subtracted by using WIRE 5.1 via an "intelligent fitting" process with a polynomial value of 11.

X-ray diffraction (XRD) and thermogravimetry (TG) analyses
After the Raman measurements, the disc specimens were crushed into powder with particle size of lower than 70 µm. Subsequently, the powder was immediately analyzed with TG and XRD for their CaCO 3 contents and composition of the carbonation products, so as to verify the quantification results obtained by Raman spectroscopy.
The TG measurements were carried out by using a TA SQT-Q600, and the temperature range was from 30°C to 1000°C with a heating rate of 10°C/ min. The XRD measurements were carried out using a Bruker D8 Advance diffractometer. A scanning range between 5 and 70°with a scanning rate of 0.02°/step was selected. Fig. 9 Diagrams showing the test points for data collection. a Test points for type 1 specimens. b Test points for type 2 specimens.
T. Mi et al.