Long-Term Chloride Accumulation on Concrete Surface in Marine Atmosphere Zone—Modelling the Inﬂuence of Exposure Time and Chloride Availability in Atmosphere

: Surface chloride concentration ( Cs ) is a key parameter used to feed models adopted to simulate chloride penetration into concrete and evaluate the initial period of corrosion. Although there are several models that have been proposed for the representation of Cs behaviour in the marine atmosphere zone, such models are still scarce. In this context, we analysed the behaviour of surface chloride concentration in concrete specimens exposed over 12.5 years in a marine atmosphere zone in the northeast of Brazil. The experimental work was carried out in two steps: environmental characterization, which was undertaken for temperature, relative humidity, rainfall, wind characteristics and sea-salt data; and chloride concentration measurements for the concrete surface considering three different concrete mixtures with w/b ratios of 0.65, 0.57 and 0.50. The results showed that the Cs increase over time followed three stages: a ﬁrst short stage characterised by an initial dispersion, followed by an increase period and then a ﬁnal period of stabilisation, which was not fully reached in the present study. This behaviour can be represented by a power function or a sigmoidal function, with a better ﬁt with the latter. Chloride concentration in the atmosphere plays an important role in Cs behaviour. Higher availability of chlorides means higher Cs values. The relationship between Cs and the rate of chloride deposition on a wet candle was analysed and the function Cs = C 0 + k cs · ( Dac ) n was the one that best ﬁt the experimental data.


Introduction
Surface chloride concentration (Cs) is one of the main parameters used to feed models adopted to simulate chloride penetration into concrete structures [1,2].As Cs governs the availability of the chloride ions that are transported into concrete, representing its behaviour in a more accurate way should result in more accurate forecasting of the initiation period of corrosion (the time that elapses between the exposure of the reinforced concrete structure and the reinforcement depassivation).
Regarding Cs behaviour over time, it has been observed that Cs tends to increase as time progresses [3,4].However, this increasing trend weakens over time and tends to become stabilised, which can be observed after about 10 years of exposure time in some cases [5].Taking into account that literature data on Cs behaviour over time are still scarce, Figure 1 presents available data from several authors focused on the marine atmosphere zone.Despite the short exposure time, it is possible to see signs of the weakening tendency previously mentioned for those cases where the exposure time was longer.Relationship between surface chloride content in concrete and exposure time [3,6,7].
The literature thus shows that Cs behaviour can be represented by various mathematical models (power or exponential growth functions), which are presented in Table 1.All these functions proposed in the literature and shown in Table 1 are based on exposures in the marine atmosphere zone.From this table and from Figure 1, it can be observed that not many studies have focused on the behaviour of Cs in the marine atmosphere zone and that most proposed functions are power or exponential functions, with the first being predominant.
Table 1.Models from the literature used to represent the behaviour of surface chloride concentration in concrete in the marine atmosphere zone.Following the analysis of the results presented by the studies referenced in Figure 1 and Table 1, it can be observed that, in general, Cs sharply increases in the initial years of exposure and, in the following years, this trend gives way to a more understated increase.Depending on the function adopted to represent behaviour of the data, the final part can assume a more asymptotic shape, which denotes a trend towards stabilisation over the years.Moreover, in some cases, Cs presents some fluctuation, which can be a result of environmental interaction [5,12].This can be more easily seen in the initial years of exposure, when the chloride concentrations are still low and the impact of the environmental variables, such as rainfall, on chloride content on concrete surfaces can assume a stronger magnitude.

Source of Data
Regarding the shape of the Cs curve, continuous cement-paste hydration over time is one of the aspects that may influence behaviour.As hydration of cement paste advances, the concrete surface becomes less porous and fewer chloride ions can be captured in this The literature thus shows that Cs behaviour can be represented by various mathematical models (power or exponential growth functions), which are presented in Table 1.All these functions proposed in the literature and shown in Table 1 are based on exposures in the marine atmosphere zone.From this table and from Figure 1, it can be observed that not many studies have focused on the behaviour of Cs in the marine atmosphere zone and that most proposed functions are power or exponential functions, with the first being predominant.Data from the literature 0-3 Cs = a + bt 0.5 Zhou et al. [11] 2016 Data from the literature 0-5 Cs = a(1 − e bt ) Yang et al. [4] 2017 Following the analysis of the results presented by the studies referenced in Figure 1 and Table 1, it can be observed that, in general, Cs sharply increases in the initial years of exposure and, in the following years, this trend gives way to a more understated increase.Depending on the function adopted to represent behaviour of the data, the final part can assume a more asymptotic shape, which denotes a trend towards stabilisation over the years.Moreover, in some cases, Cs presents some fluctuation, which can be a result of environmental interaction [5,12].This can be more easily seen in the initial years of exposure, when the chloride concentrations are still low and the impact of the environmental variables, such as rainfall, on chloride content on concrete surfaces can assume a stronger magnitude.
Regarding the shape of the Cs curve, continuous cement-paste hydration over time is one of the aspects that may influence behaviour.As hydration of cement paste advances, the concrete surface becomes less porous and fewer chloride ions can be captured in this region [10,13].As a consequence, the rate of the increase in Cs weakens.Furthermore, the concrete's ability to capture chlorides decreases with the increase in chloride concentration as a consequence of the increase in the bound chloride content, which also contributes, in the same way, to weakening the increase in Cs over time [12].
Another aspect to be considered in this analysis is the aggressiveness of the environment to which concrete structure is subjected.At locations with a greater availability of chlorides in the atmosphere, there is a stronger increase in Cs in the initial years, with a subsequent attenuation with time [3,14].However, it can be expected that Cs may present some fluctuation over the years, which may be related to the movement of ions towards bulk concrete or to direct environmental interaction, which can promote several removal effects, such as surface chloride removal due to rainfall [12].
Although several models have been proposed to represent Cs behaviour over time in the marine atmosphere zone, they are still scarce and there is no consensus concerning the best function to represent Cs behaviour.Moreover, the relationship between Cs and the availability of chlorides in the atmosphere has not been modelled.This work contributes to this discussion and analyses the behaviour of Cs in concrete exposed over 12.5 years in a marine atmosphere zone located in the northeast of Brazil, and analytical models to represent the relationships between Cs, time and chloride presence in the atmosphere are proposed.This is part of a long-term project studying chloride ion transportation into concrete under field exposure in the marine atmosphere zone.

Experimental Work
The experimental work carried out in this study took place in two steps: environmental characterisation and chloride concentration measurements from the concrete surface.

Environmental Characterization
The environmental characterization was undertaken with temperature, relative humidity, rainfall, wind characteristics and sea-salt data.Climatic data were collected by a Brazilian government weather station located in the region where the research took place.Sea-salt data were collected at location 10, 100, 200 and 500 m away from the sea (Figure 2) using the wet-candle method and a standardized capturing device, following the ASTM standard G140 [15].With regard to long-term exposure, chloride deposition measurements were taken monthly in the first 2 years and, afterwards, in two additional cycles of 12 month monitoring following 5 and 10 years of the first monitoring cycle.No significant differences could be observed among these cycles.As a consequence, these results are represented by an average monthly rate of chloride deposition and its standard deviation.

Surface Chloride Concentration in Concrete
Prismatic concrete specimens (0.15 × 0.15 × 1.40 m 3 ) were cast using filler-modified Portland Brazilian cement, the chemical and physical properties of which are presented in Table 2.The main oxides were determined using XRF analysis and the insoluble residue, loss on ignition, specific surface and specific density were obtained using the Brazilian standards specific for each of these properties.The coarse aggregate was granitic crushed rock with a maximum diameter of 19 mm and the fine aggregate was river sand with a maximum diameter of 4.8 mm.Their granulometric curves are presented in Figure 3.

Surface Chloride Concentration in Concrete
Prismatic concrete specimens (0.15 × 0.15 × 1.40 m 3 ) were cast using filler-modified Portland Brazilian cement, the chemical and physical properties of which are presented in Table 2.The main oxides were determined using XRF analysis and the insoluble residue, loss on ignition, specific surface and specific density were obtained using the Brazilian standards specific for each of these properties.The coarse aggregate was granitic crushed rock with a maximum diameter of 19 mm and the fine aggregate was river sand with a maximum diameter of 4.8 mm.Their granulometric curves are presented in Figure 3.The concrete mixtures, which had w/b ratios of 0.65, 0.57 and 0.50, along with their physical properties are presented in Table 3.They are denoted C65, C57 and C50.As these mixtures were produced several years ago, they have w/b ratios that are not typical nowadays but which represent a significant number of the reinforced concrete structures built in the past.Two samples for each test were used to characterise the concrete, as well as the studied cement and aggregates.The concrete mixtures, which had w/b ratios of 0.65, 0.57 and 0.50, along with their physical properties are presented in Table 3.They are denoted C65, C57 and C50.As these mixtures were produced several years ago, they have w/b ratios that are not typical nowadays but which represent a significant number of the reinforced concrete structures built in the past.Two samples for each test were used to characterise the concrete, as well as the studied cement and aggregates.The specimens were cured in a wet chamber (RH > 95 %) for 7 days.The curing period of 7 days was adopted to represent a condition more similar to the building site.Afterwards, the specimens were painted with a waterproof film at those surfaces thorough which chloride penetration should be avoided; i.e., all the surfaces except the one facing the sea.Then, the specimens were placed at the same monitoring stations used for the wet-candle devices in an unsheltered condition.
After 6, 10, 14, 18, 46, 78 and 150 months of exposure, powdered samples were extracted from the specimens to obtain chloride profiles for the concrete.These samples were obtained by progressively grinding the concrete layers (from surface to bulk) of cores previously extracted from the prismatic concrete specimens.These cores had 75 mm diameters and 150 mm lengths.Although chloride profiles were obtained at each sampling period, only the surface chloride contents are analysed here.These samples were extracted from the first millimetre of the exposure surfaces of the specimens; i.e., a concrete layer from the surface to a depth of 1 mm.This option was adopted considering the two-zone profile that is characteristic of marine atmosphere zones, composed of a convection zone (outer zone) and a diffusion zone (inner zone) [16,17].From this kind of profile, the extrapolation of the diffusion zone gives a theoretical Cs that is related to the internal peak that separates these two zones but not to the real concrete surface.Thus, directly obtaining a sample from the first concrete millimetre was considered to be closer to the real condition.
After obtaining the powdered samples, the total chloride content was determined through potentiometric titration using an automatic titrator from Metrohm and a doublejunction electrode with a silver ring.The sample preparation followed the procedures of the International Union of Laboratories and Experts in Construction Materials, Systems and Structures [18].

Environmental Parameters
The climatic results showed that the temperature ranged between roughly 18 and 33 • C over this period, with an average value of 27 • C (Figure 4a).The relative humidity presented fluctuations between 62 and 99%, with an average value of 76.6% (Figure 4b).Higher values were reached during the winter (rainy season), which mainly takes place between May and August.The monthly average wind speed data ranged between 1.8 and 4.2 m/s, with an average value in the exposure period of around 3 m/s (Figure 5a).The predominant average wind directions remained between the east (E) and southeast (SE) directions (Figure 5b).
presented fluctuations between 62 and 99%, with an average value of 76.6% (Figure 4b).Higher values were reached during the winter (rainy season), which mainly takes place between May and August.The monthly average wind speed data ranged between 1.8 and 4.2 m/s, with an average value in the exposure period of around 3 m/s (Figure 5a).The predominant average wind directions remained between the east (E) and southeast (SE) directions (Figure 5b).Average chloride deposition data are presented in Figure 6.This figure shows a strong decrease in salinity in the areas closest to the sea, which impacts chloride transportation into concrete at different levels.This drop in salinity is a consequence of the removal of marine aerosol salt particles, mainly due to the gravimetric effect.The gravimetric effect results in larger salt particles moving from the upper to the lower atmosphere layers until they are deposited on the ground, while marine aerosol is transported inland [19][20][21].Other removal mechanisms can simultaneously act alongside the gravimetric effect and increase salt particle removal, such as the presence of obstacles and rainfall removal.However, the gravimetric effect assumes a greater significance than the other mechanisms [20].
The rate of the decrease in marine aerosol salinity when moving inland changed from site to site and was strongly dependent on wind characteristics [22].However, a significant decrease in salinity when moving inland was observed in the vast majority of cases.As a result, concrete structures placed at different distances from the sea were subjected to different levels of aggressiveness.Average chloride deposition data are presented in Figure 6.This figure shows a strong decrease in salinity in the areas closest to the sea, which impacts chloride transportation into concrete at different levels.This drop in salinity is a consequence of the removal of marine aerosol salt particles, mainly due to the gravimetric effect.The gravimetric effect results in larger salt particles moving from the upper to the lower atmosphere layers until they are deposited on the ground, while marine aerosol is transported inland [19][20][21].Other removal mechanisms can simultaneously act alongside the gravimetric effect and increase salt particle removal, such as the presence of obstacles and rainfall removal.However, the gravimetric effect assumes a greater significance than the other mechanisms [20].Average chloride deposition data are presented in Figure 6.This figure shows a strong decrease in salinity in the areas closest to the sea, which impacts chloride transportation into concrete at different levels.This drop in salinity is a consequence of the removal of marine aerosol salt particles, mainly due to the gravimetric effect.The gravimetric effect results in larger salt particles moving from the upper to the lower atmosphere layers until they are deposited on the ground, while marine aerosol is transported inland [19][20][21].Other removal mechanisms can simultaneously act alongside the gravimetric effect and increase salt particle removal, such as the presence of obstacles and rainfall removal.However, the gravimetric effect assumes a greater significance than the other mechanisms [20].The rate of the decrease in marine aerosol salinity when moving inland changed from site to site and was strongly dependent on wind characteristics [22].However, a significant decrease in salinity when moving inland was observed in the vast majority of cases.As a result, concrete structures placed at different distances from the sea were subjected

Surface Chloride Concentration and Its Relationship with Exposure Time
The results for the surface chloride concentration are presented in Figure 7, considering the three different concretes, the four exposure sites and their distances from the shoreline.Regarding the general aspects of the data, they show some fluctuation in the first months of exposure, followed by a period of significant increase and, afterwards, a tendency to increase at lower rates.These results suggest a tendency to reach a maximum over time, but it not possible to fully observe this in these 12.5 years of exposure.The initial fluctuation in the chloride concentration on the concrete surface can be attributed to environmental interaction, in which rainfall can play an important role, changing the chloride concentration on concrete surfaces with a relative higher magnitude [12,23].

Surface Chloride Concentration and Its Relationship with Exposure Time
The results for the surface chloride concentration are presented in Figure 7, considering the three different concretes, the four exposure sites and their distances from the shoreline.Regarding the general aspects of the data, they show some fluctuation in the first months of exposure, followed by a period of significant increase and, afterwards, a tendency to increase at lower rates.These results suggest a tendency to reach a maximum over time, but it not possible to fully observe this in these 12.5 years of exposure.The initial fluctuation in the chloride concentration on the concrete surface can be attributed to environmental interaction, in which rainfall can play an important role, changing the chloride concentration on concrete surfaces with a relative higher magnitude [12,23].Regarding the influence of the concrete's characteristics, it is clear there was greater accumulation of chlorides on the concrete surface as concrete porosity increased (see Table 3), which was related to the greater availability of space for the accumulation of chloride ions.Furthermore, taking into account the distance from the sea, it is noticeable that the chloride concentration decayed as the distance from the sea increased, which was a consequence of the decrease in aggressiveness at sites far from the shoreline due to the lower availability of chlorides in the atmosphere, as can be observed in Figure 6.This decay did not follow a linear relationship, as can be indirectly observed in Figure 7 (and in the following figures), where it is possible to observe rates of decrease differing from one exposure site to another.This difference occurred because only some of the chlorides in atmosphere were deposited and captured on the concrete surface.
Exponential or power functions are usually used to represent the increase in Cs over time, as presented in Table 1.Nevertheless, the function that best fit the present experimental data was a sigmoidal function that considered the initial stage of fluctuation at low chloride concentrations, the sharpest growth period and the final period of stabilization.This can be better seen when comparing the determination coefficients obtained when using power functions and sigmoidal functions (Figure 8).This latter function is an alternative to the options presented in the literature that better represents the three stages of Cs growth over time.Regarding the influence of the concrete's characteristics, it is clear there was greater accumulation of chlorides on the concrete surface as concrete porosity increased (see Table 3), which was related to the greater availability of space for the accumulation of chloride ions.Furthermore, taking into account the distance from the sea, it is noticeable that the chloride concentration decayed as the distance from the sea increased, which was a consequence of the decrease in aggressiveness at sites far from the shoreline due to the lower availability of chlorides in the atmosphere, as can be observed in Figure 6.This decay did not follow a linear relationship, as can be indirectly observed in Figure 7 (and in the following figures), where it is possible to observe rates of decrease differing from one exposure site to another.This difference occurred because only some of the chlorides in atmosphere were deposited and captured on the concrete surface.
Exponential or power functions are usually used to represent the increase in Cs over time, as presented in Table 1.Nevertheless, the function that best fit the present experimental data was a sigmoidal function that considered the initial stage of fluctuation at low chloride concentrations, the sharpest growth period and the final period of stabilization.This can be better seen when comparing the determination coefficients obtained when using power functions and sigmoidal functions (Figure 8).This latter function is an alternative to the options presented in the literature that better represents the three stages of Cs growth over time.
After producing simulations using the functions adopted in Figure 8, it was possible to compare Cs values after 50 years of exposure in the marine atmosphere zone (Table 4).These results show that the sigmoidal function tends toward stabilization over a shorter time than the power function (as can also be seen in Figure 8) and that the values of Cs at the end of the 50 years may be between 3.7 and 7 times higher for the power functions.This overestimation was more pronounced for lower chloride concentrations in the atmosphere and less porous concretes.Considering the arrangement of the data in Figure 8 and previously published papers focusing on marine exposures [5], it is more reliable to expect a stabilization tendency after 10 years, which is better represented by the sigmoidal function.After producing simulations using the functions adopted in Figure 8, it was possible to compare Cs values after 50 years of exposure in the marine atmosphere zone (Table 4).These results show that the sigmoidal function tends toward stabilization over a shorter  Figure 9 gathers the literature data and the presented experimental data and shows the curves that represent each dataset.From this figure, it can be seen that the present results are from long-term field exposure and, thus, can catch part of the stabilization period.Furthermore, the literature data based on shorter exposure periods generally presented higher tendencies for increases in Cs.This can be explained not only by the fact that these studies were based on the first and second stages of the increase in Cs during the exposure time but also by the differences in the availability of chlorides in atmosphere.Nevertheless, the present data and Ghods et al. [7] data show good harmony, which may indicate similarities in environmental characteristics.
This overestimation was more pronounced for lower chloride concentrations in mosphere and less porous concretes.Considering the arrangement of the data in F and previously published papers focusing on marine exposures [5], it is more reli expect a stabilization tendency after 10 years, which is better represented by the sig function.9 gathers the literature data and the presented experimental data and the curves that represent each dataset.From this figure, it can be seen that the p results are from long-term field exposure and, thus, can catch part of the stabilizat riod.Furthermore, the literature data based on shorter exposure periods general sented higher tendencies for increases in Cs.This can be explained not only by t that these studies were based on the first and second stages of the increase in Cs the exposure time but also by the differences in the availability of chlorides in atmo Nevertheless, the present data and Ghods et al. [7] data show good harmony, whi indicate similarities in environmental characteristics.

Surface Chloride Concentration and Its Relationship with Chloride Deposition on a W Candle
Considering that the exposure time is not the main variable that influences crease in Cs but rather the availability of chlorides in the atmosphere, the relati between Cs and the accumulated deposition of chlorides on a wet candle (Da

Surface Chloride Concentration and Its Relationship with Chloride Deposition on a Wet Candle
Considering that the exposure time is not the main variable that influences the increase in Cs but rather the availability of chlorides in the atmosphere, the relationship between Cs and the accumulated deposition of chlorides on a wet candle (Dac) was analysed, which was obtained by summing the chloride deposition on the wet candle month-to-month (Figure 10).The power and sigmoidal functions were fitted to the experimental data in a similar way.
analysed, which was obtained by summing the chloride deposition on the wet candle month-to-month (Figure 10).The power and sigmoidal functions were fitted to the experimental data in a similar way.As can be seen in Figure 10, there were no significant differences between the fitting results using the power and sigmoidal functions.This can be explained by the fact that,

Figure 2 .
Figure 2. Region where wet-candle devices and concrete specimens were exposed.

Figure 2 .
Figure 2. Region where wet-candle devices and concrete specimens were exposed.

Figure 4 .
Figure 4. Maximum and minimum daily temperature (a) and daily average relative humidity (b) in the exposure region.

Figure 4 .
Figure 4. Maximum and minimum daily temperature (a) and daily average relative humidity (b) in the exposure region.

Figure 4 . 8 Figure 5 .
Figure 4. Maximum and minimum daily temperature (a) and daily average relative humidity (b) in the exposure region.

Figure 5 .
Figure 5. Monthly average wind speed (a) and monthly average wind direction (b).

8 Figure 5 .
Figure 5. Monthly average wind speed (a) and monthly average wind direction (b).

Figure 6 .
Figure 6.Average chloride deposition data for the research period.

Figure 6 .
Figure 6.Average chloride deposition data for the research period.

Figure 8 .
Figure 8. Fitting results with power and sigmoidal functions and Cs data for C65 (a), C57 (b) and C50 (c) concrete.

Figure 8 .
Figure 8. Fitting results with power and sigmoidal functions and Cs data for C65 (a), C57 (b) and C50 (c) concrete.

Figure 9 .
Figure 9. Fitting curves representing the literature data and the data from the present work

Figure 9 .
Figure 9. Fitting curves representing the literature data and the data from the present work [3,6,7].

Table 1 .
Models from the literature used to represent the behaviour of surface chloride concentration in concrete in the marine atmosphere zone.

Table 2 .
Chemical and physical properties of studied cement.

Table 2 .
Chemical and physical properties of studied cement.

Table 3 .
Concrete mixtures and properties.

Table 4 .
Simulation of Cs values using power and sigmoidal functions for 50 years of exposure.

Table 4 .
Simulation of Cs values using power and sigmoidal functions for 50 years of expos