Effect of wetting and drying cycles on the durability of bio-cemented soil of expressway slope

Cyclic wet-dry is one of the influential weathering agents which can rapidly alter the mechanical properties of soils, limiting their durability and consistent performance. This study investigates the effect of wet-dry cycles on the mechanical behaviour of bio-cemented soil. Microbial-induced carbonate precipitation-based bio-cementation is an innovative soil improvement method, which is gaining increasing attention as a potential alternative for stabilizing slope surface. As the treated surfaces are exposed to repeated rainfalls and draughts, durability analysis is essential; cyclic wet-dry tests were therefore performed as a credible indicator of durability. The soil obtained from the Hokkaido expressway slope was treated at laboratory to varying cementation levels (% CaCO3) and subjected to 50 subsequent wet-dry cycles. Physical and mechanical changes were monitored using mass loss, shear wave velocities and needle penetration tests during wet-dry cycles. The results showed that the wet-dry cycles deteriorated the physical and mechanical at two stages. The mass and S-wave velocity of specimens significantly dropped after first few cycles and then tended to reach equilibrium. The second stage of notable deterioration was observed between 30 and 50 wet-dry cycles. It is suggested that the erosion of weak and powdery deposition of CaCO3 causes the degradation at the early stage, whereas the degradation in the late stage was attributed to the microstructural deformations of intact carbonate bonds. It was also found that the increase in cementation level decreases the deterioration of bio-cemented soil under wet-dry cycles.


Introduction
Over the decades, number of cementation-based soil stabilization methods have been developed and which are being applied in geotechnical engineering projects at a global scale. MICP (microbial induced carbonate precipitation) is relatively a novel bio-cementation method, developed at the confluence of bio-geochemical processes to improve the geotechnical properties of soils. The method utilizes non-pathogenic ureolytic microorganisms to mineralize the calcium carbonate bio-cement within soil matrix. Firstly, the hydrolysis of urea is catalysed by enzyme urease of soil bacteria, resulting the formation of carbonate and ammonium ions (Eq. 1). In the subsequent step, the crystallization of CaCO 3 occurs in calcium-rich aqueous media (Eq. 2), and the precipitated carbonates coat the soil grains, bind the particles effectively at particle contacts, eventually resulting the increase in strength of the soil (DeJong et al. 2010;Achal and Kawasaki 2016;Omoregie et al. 2020). Many investigations demonstrated that the MICP can be effectively implemented to various geotechnical applications including ground improvement (van Paassen et al. 2010;Kalantary and Kahani 2019;Badiee et al. 2019), liquefaction resistance in loose sands (Montoya et al. 2013;Moosazadeh et al. 2019), erosion control (Maleki et al. 2016;Wang et al. 2018;Jiang et al. 2019), coastal protection (Nayanthara et al. 2019;Imran et al. 2019) and stabilizing dust and loess (Meyer et al. 2011;Sun et al. 2020).
Among the wide range of potentials, application of MICP for the preservation of slope surface is gaining a great deal of interest among geotechnical engineers. One major reason is that the MICP offers numerous ecological benefits over conventional measures. For an instance, the Portland cement is the mostly used binder for slope protection (Daraei et al. 2018;Tang et al. 2018); however, its manufacturing contributes to the significant emission CO 2 and SO 2 (Alshalif et al. 2020). Besides, the synthetic grouts used for surface stabilization (such as polyacrylamide and phosphor-gypsum) are reported to be highly toxic (Jiang et al. 2019). In fact, MICP is a naturally occurring biogeochemical process in the earth's crust. The required bacteria are also found to be abundantly existing in soil environments. Therefore, MICP can be an eco-friendly alternative and can be applied to the slope surface with minimal soil-environmental damage.
The slopes are typically exposed to several climatic changes such as rainfall, snowfall, snowmelt and evapotranspiration. Surface erosion is a common slope degradation process; therefore, most of the existing studies focused on the responses under simulated rainfalls and wave effects. It was repeatedly shown that the MICP can significantly improve the erosion and scouring resistance of foreshore slopes when subjected to tidal wave effects (Salifu et al. 2016;Bao et al. 2017;Liu et al. 2020). In due course, few researchers assessed the impact of short-term rainfall on the surface erosion of sandy slopes (Jiang et al. 2019;Chung et al. 2020). The outcomes indicated that the calcium carbonate bonds formed between soil particles could markedly enhance the stability of surface aggregates against erosion. Recently, to apply the MICP in cold regions, Gowthaman et al. (2020) studied the durability characteristics of treated slope surface, showing that the durability under freeze-thaw actions is attributed to the effective cementation at particle contacts.
Cyclic wet-dry (WD) is another core slope weathering agent, and that has been a serious concern in geotechnical and geological engineering. Due to the direct contact with the atmosphere, the surface undergoes periodical WD cycles during (i) diurnal changes in rainy-sunny weathers (Tang et al. 2011) and (ii) subsequent processes of rainfallevapotranspiration (Tang et al. 2016). Numerous studies demonstrated that the WD cycles cause substantial irreversible deformations in the fabric of geomaterials. For instance, presence of swelling minerals was found to cause temporal changes in soft soils during WD, resulting in desiccation cracking, shrinkage and volumetric changes (Tang et al. 2011;Zhang et al. 2013). Repeated WD process significantly alters microstructure of soil aggregates, leading to deterioration in shear strength and compressibility (Tang et al. 2016). In the case of sedimentary materials, fracture (2) Ca 2+ + CO 2− 3 → CaCO 3 (deposition) energy reduction and chemical corrosive actions are the most reported deterioration mechanisms. Several research works evidenced that WD process significantly deteriorates the mechanical properties of calcarenite rocks and sandstone materials (Ciantia et al. 2014(Ciantia et al. , 2015Khanlari and Abdilor 2015). Since the responses of MICP-treated soils (cemented by CaCO 3 ) are often comparable with natural carbonate/ sandstone sediments (Daryono et al. 2020), treated surface may more likely undergo similar deterioration during prolonged exposure to repeated WD process. A performance assessment with respect to durability is therefore necessary and recommended prior to the field applications. It is worth noting that the available information on weathering effects of MICP treatment under repeated WD actions are very limited. The flowchart (Fig. 1) illustrates the sequence of entire research works carried out to introduce the MICP technique as a new alternative for the preservation of expressway slope surface in Hokkaido, Japan. As indicated in Fig. 1, the objective of this article is to investigate the effects of repeated WD actions on the physical and mechanical deterioration of MICP treated soil at a pilot-scale. The soil obtained from the Hokkaido expressway slope was prepared in cylindrical

Representative slope
The soil used in the study was obtained from the erosion prone expressway slope (Do-O expressway line, Onuma) of Hokkaido, Japan. For the experimental purpose, disturbed soil samples were obtained at the site (the latitude and longitude are 42.388532 N and 140.284762E, respectively). Fig. S1 (refer supplementary file) presents the particle size distribution curve of the soil. Based on the Unified Soil Classification System (USCS), the soil is classified as SP, poorly graded sand, consisting particle diameter at 50% finer by mass (D 50 ) of 0.23 mm, coefficient of uniformity (C u ) of 2.5 and coefficient of curvature (C c ) of 0.8. The basic characteristics and the chemical composition obtained from energydispersive X-ray fluorescence (XRF) analysis (JSX-3100R II JOEL, Japan) are provided in Table 1.

MICP treatment procedure
The soil samples were prepared into the cylindrical syringe moulds (30 mm in diameter; 50 mm in height) to a density of 1.6 ± 0.1 g/cm 3 . (The corresponding relative density was around 60%). The experimental fabrication is shown in Fig. S2(a) (refer supplementary file). A two-phase treatment by surface percolation method was used similar to that suggested in many previous works (Cheng et al. 2013;Omoregie et al. 2019). In the first phase, the injection of bacteria culture was performed. The urease positive bacteria (Lysinibacillus xylanilyticus) which were isolated previously from the same target slope was cultivated at aerobic conditions (in shaking incubator at 25 °C and 160 rpm) for 48 h in ammonium-yeast extract media (ATCC 1376) consisting 20 g/L of yeast extract, 15.75 g/L of tris buffer and 10 g/L of ammonium sulphate. Isolation and identification processes of the bacteria can be found in our previous work (Gowthaman et al. 2019a). The optical density value (OD 600 ) and urease activity of the bacteria culture during the injection were around 4.0 and 3 U/mL, respectively (where, 1 U defines the hydrolysis of 1 micromole urea per minute at pH 7.0 at 25 °C).
In the second phase of the treatment, multiple cementation injections were performed until the required cementation level (% CaCO 3 by mass) was achieved. All the solutions were introduced to the specimen at the constant flow rate of 4 mL/min using a peristaltic pump. The chemical recipe of cementation media consists of 111 g/L of calcium chloride, 60 g/L of urea and 6 g/L of nutrient broth. As previously mentioned, the different levels of cementation were achieved by performing different number of the injections of cementation media. In this study, three cementation levels were considered: C-1, C-2 and C-3. The average cementation level of C-1 is between 12 and 13% CaCO 3 . C-2 is defined as an average cementation between 15 and 17%, and the average cementation of C-3 is 22-23% CaCO 3 . To achieve the C-1, C-2 and C-3 levels, total number of 7, 10 and 14 injections of cementation media were, respectively, performed, and the injection was on every 24 h basis. The methodology used to determine the precipitated quantity of CaCO 3 in treated specimens is elaborated in the following section.
It should be noted that as the enzymatic activity of the bacteria drops after 7 injections (which was confirmed by the drop in pH and increase in Ca 2+ concentration at effluent), the bacteria injection was performed once again on 7th day. The cementation levels considered in this study fall under the heavy cementation category (Feng and Montoya 2016). After the treatment, the moulds were cut, and the samples were carefully taken out. Finally, all samples were prudently trimmed to achieve smooth surfaces (attaining the final specimen size of 3 × 3 cm: diameter × height). The appearance of the specimens treated to different cementation levels are shown in Fig. S2(b) (refer supplementary file).

Determination of CaCO 3 content
The carbonate contents of the MICP specimens were determined by using a simplified device fabricated for the measurements of CO 2 gas pressure during reactions (Fukue et al. 1999). After obtaining the oven dried mass (105 °C, 48 h), the samples were allowed to react with 3 mol/L concentrated HCl in the closed system (of device) under constant temperature and volume conditions. Meanwhile, the internal pressure was monitored from the digital manometer connected with the system; the reading was recorded when the manometer read constant pressure. Using the calibration curve developed between quantity of calcium carbonate and pressure of CO 2 gas released during reaction, the precipitated mass of CaCO 3 was determined; hence, the cementation content (%) was determined using Eq. 3.

Cyclic wet-dry (WD) tests and test cases
The WD cycles were performed in accordance with the method suggested in ASTM (2003). In each wet-dry cycle, specimens were fully immersed into distilled water at room temperature (25 ± 1 °C) for 6 h, followed by a minimum of 42 h of oven drying at 70 ± 1 °C. Specimens were subjected to a total number of 50 cycles. By the end of each WD cycle, the dry mass of the specimen (when the mass becomes constant during drying process) was recorded to evaluate the loss in mass.
As mentioned earlier, the specimens were treated to three levels of cementation (C-1, C-2 and C-3) to evaluate the influence of cementation level on durability of MICP treated soil. The details of the specimens subjected to WD analysis are summarized in Table 2. The mechanical deterioration was continually evaluated by monitoring the shear wave velocity of the specimens and comparing the UCS of control specimens (i.e. not subjected to wetdry tests) and tested specimens (i.e. after wet-dry cycles). In order to study the initial characteristics, a set of specimens was simultaneously treated to considered levels, and their as-treated characteristics were evaluated (shown in Table 3). In each case of analysis, experiments were performed in triplicate. Wet-dry analysis C-1-1 7 1.15 C-1 12-13% C-1-2 7 1.18 C-1-3 7 1.13 C-2-1 10 1.35 C-2 15-17% C-2-2 10 1.34 C-2-3 10 1.36 C-3-1 14 1.61 C-3 22-23% C-3-2 14 1.87 C-3-3 14 1.59

Shear wave velocity measurement
Shear wave velocities (V s ) of the specimens were determined after every WD cycle using high voltage pulse receiver (Son-icViewer-SX: 5251, Japan). During measurement, MICP specimens were carefully positioned between transducers (transmitter and receiver), and the signals from the benders were captured using the two-channel digital oscilloscope at frequency bandwidth of 150 MHz. The elapsed time (digital oscilloscope, determined from the output waveforms) and the wave travel distance (sample height, measured by vernier caliper) were used to compute the shear wave velocity of the specimens, as shown in Eq. 4.
where L is the distance between the transmitter and receiver (i.e. specimen height), and t is the effective travel time of the signal from the transmitter to the receiver. t is the delay of the wave when there is no specimen placed, measured by assembling only transmitter and receiver.

UCS measurement
The UCS of the samples were estimated by needle penetration test (SH-70, Maruto Testing Machine Company, Tokyo, Japan) in accordance with JGS (2012). The tests were performed to the specimens after 12, 30 and 50 WD cycles. Following the drying process, the needle of the device was gradually penetrated into the cylindrical surface of the specimen. Based on the obtained test results (penetration resistance and penetration depth), the UCS of the specimens were accurately estimated from the regression relationship developed (Eq. 5). It should be noted that needle penetration tests are widely being used by many other researchers as well for the evaluation of MICP treated specimens (Fukue et al. 2011;Danjo and Kawasaki 2016;Amarakoon and Kawasaki 2018;Gowthaman et al. 2019b).
where y is the UCS; x is the "penetration gradient (N/mm)" which can be determined using penetration resistance and penetration depth.

SEM and XRD analysis
To study the microstructural changes in the specimens, SEM analysis was performed using Miniscope TM 3000 (Hitachi, Tokyo, Japan) before and after the WD test. The representative samples were carefully taken from the samples and observed through SEM after the oven dry at 60 °C. To better understand the chemical changes in the specimens during WD actions, XRD (X-Ray Diffraction) tests (MiniFlex™, Rigaku Co. Ltd., Tokyo, Japan) were performed before and after 12, 50 WD cycles. The powdered samples were tested at a scan rate of 6.5°/min and at angles from 5° to 70° (2θ). The results obtained from the analysis are discussed comprehensively in the subsequent sections.

Distribution of calcium carbonate
Level of cementation was found to be one of the most important factors determining the mechanical characteristics of MICP-treated soils (Feng and Montoya 2016;Gowthaman et al. 2020). As presented in Table 2, the specimens were treated herein to three levels of cementation, referred to as C-1, C-2 and C-3 (corresponding to 7, 10 and 14 numbers of cementation injections, respectively). From the CaCO 3 measurements shown in Table 3, it can be seen that the average precipitation content of CaCO 3 gradually increases with the increasing treatment numbers, which is consistent with the tendency reported by Cui et al. (2017) for the sands treated by MICP. In addition, the shear wave velocity values of C-1, C-2 and C-3 specimens are in a range between 1.1 and 1.2 km/s, 1.3 and 1.4 km/s and 1.6 and 1.9 km/s (Table 2), respectively, indicating that the velocity measurements have a close relationship with the precipitation content of CaCO 3 . As the mechanical strength due to MICP is achieved by the cementation at particle-particle contacts, the shear wave velocity measurement can be a credible measure to non-destructively assess the mechanical behaviour of treated specimens.
It is undoubtedly essential to evaluate the uniformity of CaCO 3 precipitation with the depth, because the MICP is attained by injecting solutions from the surface of the specimens. For that, treated control specimens (Table 3) were prudently sectioned into three equal segments along the height (around 1 cm each), and measured calcium carbonate contents are plotted in Fig. 2 against depth. Slightly a heterogeneity in vertical distribution of CaCO 3 is witnessed from the results of all the treatment levels. Cheng et al. (2014) also observed a similar distribution pattern when treating fine sand material. In fact, there are several factors determining the spatial distribution of CaCO 3 along the depth including distribution of bacteria cells, chemical conversion and particle size distribution (Whiffin et al. 2007;Martinez et al. 2013). As the soil studied herein is a fine-grained soil, more bacteria cells were possibly filtered at around the injection top compared to bottom zone, which led to the gradient in the distribution (Fig. 2). It should be noted that in this study, the average CaCO 3 value (%) was used to represent each cementation level (i.e. C-1: 12-13%, C-2: 15-17% and C-3: 22-23%). Figure 3 presents the physical appearance of specimens treated to varying cementation levels before and after subjecting repeated WD cycles. As observed, the surfaces of the specimens were relatively smooth before subjected to WD test. By the end of 50 WD cycles, C-1 specimens showed severely damaged surface with large open pores, indicating less durability under repeated WD actions. C-2 specimens exhibited a moderate level of surface damage, whereas the specimens cemented to C-3 level showed only a minor damage on the surface, suggesting higher resistance to the WD-induced damages compared to that of C-1 and C-3. It should be noted that in all the cemented specimens, the damages could be prominently seen on the surface after around 30 WD cycles. Figure 4a presents the cumulative mass loss of the specimens with increasing number of WD cycles. The results show that the specimens underwent an average mass loss in a range between 2 and 5% by the end of 50 cycles, which is in a good agreement with the mass loss reported to phosphate cement specimens under similar WD actions (Li et al. 2019). In addition, the level of cementation appears to influence the rate of loss in mass. By the end of the 50 WD cycles, the scaled-off mass of the C-1, C-2 and C-3 specimens were 4.4%, 2.8% and 2.4%, respectively, suggesting more or less of the cementation level would in turn increase or decrease the resistance to aggregate loss.

Visual observation and mass loss
Moreover, all the tested specimens revealed an abrupt increase in mass loss (over around 1.3% by weight) by the end of first WD cycle (Fig. 4a). However, within another few more WD cycles, the scaling-off decreased gradually and tended to reach an equilibrium state. Up to around 30 number of WD cycles, relatively flat graphs were obtained in all the test cases, suggesting that the loss of mass were negligible within this equilibrium phase. A similar tendency in mass loss was also reported by Ciantia et al. (2015) to the calcarenite specimens when subjected to repeated WD cycles. When the MICP specimens were subjected to further WD cycles (above 30), mass loss became significant once again. At this later period, increased loss rate was evidenced in C-1 specimens, whereas the loss rates experienced in C-3 specimens were lower compared to C-1 (around 1.8 times) under WD actions.

Shear wave velocity
For better understanding the changes in the mechanical behaviour of cemented specimens, shear wave velocities were measured periodically with the increase in number of WD cycles, and the results are provided in Fig. 4b. It is worth noting that the precipitated calcium carbonate at particle contact has the utmost effect in shear wave velocity of the specimens, because the shear waves can only be transmitted through solids (not through soil pores) (Al Qabany et al. 2011). Therefore, shear wave measurements were frequently used as a credible monitoring tool for MICP cemented specimens (Feng and Montoya 2016;Gowthaman et al. 2020), demonstrating that the more or less/ strong or weak connections of soil particles would result, respectively, high or low velocity values. From Fig. 4b, it can be seen that the variation in velocity measurements follows relatively a similar pattern in all the cementation cases with increasing WD cycles, i.e. a rapid drop was observed at the early stage (within around 5 WD cycles), followed by a smooth and gradual reduction. It is worth mentioning that the tendency observed in velocity measurements (Fig. 4b) shows a good agreement with pattern observed in scale-off mass (Fig. 4a). This explains that the rapid reduction in velocity at early stage (up to 25%) could be attributed to the weakening of particle connections due to the erosion induced by WD process. Furthermore, the velocity reduction taken place within 5-50 WD cycles was comparatively lower than that befell during early stage (0-5 cycles), suggesting that the impairment during early stage of WD process was noteworthy.
Unconfined compressive strength Figure 5 shows the variation in UCS of MICP treated specimens after the exposure of 0, 12, 30 and 50 WD cycles. Before subjected to the WD cycles, the average UCS of C-1, C-2 and C-3 cemented specimens were 2.3, 4.1 and 11.8 MPa, respectively. After subjected to 12 WD cycles, C-1 and C-2 specimens showed significant loss in UCS, which were, respectively, around 38% and 27% relative to those of control specimens. On the other hand, C-3 specimens showed inconsequential destruction in strength (which was only 4.5%), indicating higher durability against WD effects compared to C-1 and C-2. Nevertheless, the UCS continued to decrease gradually in all the cases with increasing cycles, and by the end of 50 cycles, the reduction in UCS were 55%, 46% and 32% for C-1, C-2 and C-3 specimens, respectively. It is clearly perceived the strength deterioration conspicuously occurs throughout the WD process, but the overall deterioration rate is more likely to be determined by the level of cementation. Fig. S3 (refer supplementary file) shows the results of XRD analysis which was performed to verify the occurrence of phase transformation of carbonate deposits, precipitation of new minerals and dissolving of existing minerals. The results suggest that the morphology of the precipitated calcium carbonate before the WD tests was completely calcite. Basically, the calcium carbonate can precipitate in various forms of polymorph such as calcite (rhombohedral shape), vaterite (spherical shape), aragonite (typically clustered needles) and amorphous calcium carbonates (ACC, nanoparticles), which is determined by reaction kinetics and thermodynamic  (Dhami et al. 2013;Tolba et al. 2016). The formation of least-stable phases (such as vaterite and ACC) at the initial stage of reaction might get transformed into stable phase (such as calcite) (Wang et al. 2019). However, the XRD results confirm that there are no observable changes in the morphology of minerals throughout the WD process. The observed peaks are identical with the increase in number of cycles, suggesting neither formation of new minerals nor phase changes were induced.

Deterioration mechanism of MICP treated soil
In this study, the MICP was investigated as a cementation material to treat the soil. The treatment is attained through the surface percolation; most of the bacteria cells are therefore tended to aggregate at particle contacts, leading to a preferential formation of calcium carbonate at or near the contact points of soil particles (DeJong et al. 2010). However, as the crystallization process had already begun during the percolation phase (Whiffin et al. 2007), not all the precipitated carbonates contribute to the contact cementation; a considerable quantity of carbonates either crystallize randomly on the soil grains or set on the already formed carbonate bridges. Micrographs depicted in Fig. 6a, b clearly illustrate different forms of calcium carbonate deposits. As shown, the secondary formations of the calcium carbonate could embed as accumulated carbonate clusters (marked as B), individual clusters (marked as C) and powdery deposits (marked as D), and the microstructure reported herein is consistent with the micrographs of many previous works (Bao et al. 2017;Wu et al. 2019).
It is clearly perceived that the physical and mechanical behaviours are significantly influenced by the WD cycles. From the results of both mass loss and V s measurements (Fig. 4a, b), it is evident that the performance deterioration of specimens occurs namely in two stages. In the early stage (within 5 WD cycles), a severe degradation was witnessed, Fig. 6 a Microstructure of MICP soil (A: CaCO 3 well deposited at particle contacts; B: CaCO 3 assemblage; C: individual crystals; D: powdery CaCO 3 deposits; E: accumulation of CaCO 3 deposits; F: deposits supporting large carbonate clusters), b powdery CaCO 3 deposits, c schematic illustration of MICP soil before subjected to wetting, d wetting process: penetration of water into pores, and e wetting process: saturated stage hereafter referred to as STD (short-term deterioration). In the subsequent stage, a gradual degradation was observed, which is hereafter referred to as LTD (long-term deterioration).

Short-term deterioration (STD)
The STD mechanism can be explained from the evolution of calcium carbonate precipitates during MICP treatment process. As demonstrated by Wang et al. (2019), during the injection of cementation solution, the calcium carbonates are initially induced as irregular precipitates (non-crystalline), and they continue to grow with the time and transform as regular calcium carbonate crystals (e.g. calcites) in the aqueous media. However, in many situations (e.g. deficiency of resources, presence of organic content in reaction system), the induced non-crystalline precipitates tend to stabilize on soil surfaces before transforming to crystalline structure. Figure 6b shows the magnified micrograph, clearly depicting such powdery forms of CaCO 3 attached ubiquitously on the grain surface. Similar powdery deposits of calcium carbonate were also observed previously during the formation of calcarenite rocks (Ciantia et al. 2015). It is therefore clear that the MICP treated soils consist of both strong and weak deposits of calcium carbonate. Figure 6c, d, and e show the serial progression of water inundation during the initial wetting process. Figure 6c schematically illustrates the microstructure of MICP treated soil matrix, representing the existence of different types of carbonate bonds. When the samples are initially subjected to the wetting, water tends to penetrate through the porous system and inundate the material, resulting the powdery deposits and their accumulations to fall into suspension (Fig. 6d,  e), followed by the erosion. This could attribute to the leap increase in mass loss observed after first WD cycle (Fig. 4a). It should be noted that at the early stage, the erosive mechanism is dominant, as the dissolution of calcium carbonate is relatively low under the standard wetting conditions (Ciantia and Hueckel 2013). By the end of first cycle, layer of powdery formations was evidenced at the bottom of the container wherein the samples were subjected for wetting, which further corroborated the mechanism suggested herein. Following the first cycle, the mass loss tended to become relatively stable (in another 3-4 cycles), suggesting that the powdery deposits have totally been eroded. Moreover, the rapid drop observed in V s measurements within first 5 cycles (Fig. 4b) could be attributed to the disintegration of powdery and weak deposits. As demonstrated in Fig. 6c, there are also carbonate clusters which are feebly bonded, unlike primary intact bonds which are well-formed. Despite weakly attached, they had a substantial contribution in strengthening particle connections through supporting primary bonds. The repeated inundation during 5 WD cycles would disintegrate the deposits as illustrated in Fig. 6d, e, resulted in weakening of particle connections. This eventually caused the rapid loss in V s and UCS of the specimens within the short-term.
Numerous WD degradation mechanisms were proposed for geomaterials such as fracture energy reduction, capillary tension decrease, chemical and corrosive deterioration, frictional reduction, etc.; however, they vary from material to material and depend on the exposure conditions (Zhao et al. 2017). Clay materials or the soils consisting certain fraction of clay are more susceptible to WD process, exhibiting rapid deterioration (Albrecht and Benson 2001;Tang et al. 2011). For instance, Tang et al. (2011) observed the formation of numerous large inter-aggregate pores in clayey soil after only 2 WD cycles due to the desiccation cracking and increase in bulk volume. The largest increase in hydraulic conductivity and shrinkage strains in clay occurred after the first WD cycle (Albrecht and Benson 2001). It is worth mentioning that the STD mechanism of MICP treated soils are more likely to be comparable with that of carbonatic rocks. Ciantia et al. (2015) found the occurrence of instantaneous debonding of depositional bonds (menisci shaped calcite ensembles) in carbonatic rocks when subjected to the saturation, and a single wetting process caused considerable loss in compressive strength.

Long-term deterioration (LTD)
When the specimens were subjected to further WD cycles, the MICP-treated specimens continuously underwent deterioration (refer to Fig. 4). However, this LTD was not as rapid as that observed in early stage (i.e. STD). Figure 7a, b presents the SEM images captured, respectively, before and after the cyclic WD process. All the micrographs presented herein are captured at same magnification and scale (indicated at bottom right of each figure). From the comparison, a substantial change in the microstructure can be observed after subjected to prolonged WD actions. Before subjected to the WD test (control specimens), the presence of large carbonate clusters was evidenced in SEM (Fig. 7a). The appearance indicates that the calcite morphology found in control specimens was cauliflower-like crystallization, evolved by the accumulation of rod shape calcite clusters ). On the other hand, by the end of 50 WD cycles, a substantial deformation was observed in carbonate deposits (Fig. 7b), and the cauliflower-like morphology could not be seen anywhere in the matrix. It should be noted that the results of XRD analysis (refer Fig. S3 in supplementary file) confirmed that the deformation is not related to the chemical changes or phase transformations; therefore, this deformation might be possibly due to the effects from the development of internal stresses during cyclic WD process as reported by Li et al. (2019). The difference in thermal expansion coefficient of calcite and soil material could result in the development of fatigue stresses during cyclic WD process (due to temperature change), led the calcite crystals to undergo deformation.
It should be noted that the destruction in MICP specimens due to the fatigue stresses was clearly observable at the later stage of WD process. For example, sharp increase in mass loss was once again prominently noticed after around 30 number of cycles (Fig. 4a). During this stage, soil particles tended to detach from the surface of the specimens, and rapid progression of macropores was experienced around surface (refer to Fig. 3). Fig. S4 (refer supplementary file) shows the microstructure of the deteriorated specimen surface (after 50 cycles). The observation reveals that the specimen surface underwent higher destruction compared to that observed around core of the specimens (Fig. 7b). As specified in Fig. S4, the calcite precipitated around the surface was observed to be impaired in addition to the deformation of calcite morphology. Microcracks were also observable at around the intact bonds. It is clear that the core zone of the sample is confined stiffly; therefore, the developed fatigue stresses caused more damages on the outer surface of the specimens, led to the weakening of soil aggregation and subsequent detachments. Similar surface degradations due Fig. 7 The SEM images of MICP treated specimens a before subjected to WD cycles and b after the exposure of 50 WD cycles to WD stresses were also observed by several researchers for cemented specimens (Zhang et al. 2012;Gao et al. 2014).
It was reported that when the calcites are flooded by water for a long time, a minor portion of carbonate would undergo chemical dissolution (Ciantia et al. 2014;Khanlari and Abdilor 2015). Ciantia et al. (2015) have confirmed that the LTD of carbonatic rocks is due to the dissolution of diagenetic calcite bonds. However, in this study, the effect of dissolution is assumed to be negligible under the standard conditions.

Effect of cementation level on the durability of MICP treatment
As explained earlier, three levels of cementation (C-1, C-2 and C-3) were investigated in this study. The WD test results clearly reveal that all the specimens experienced physical and mechanical deterioration under the exposure of continuous WD cycles. However, the cementation level is found to have a marked impact on the durability of MICP treated soil. The comparison presented in Fig. 8 reveals the effect of cementation level on the durability of MICP treated soil (UCS and mass losses after 50 WD cycles are compared). By the end of 50 WD cycles, the losses in UCS were 55%, 46% and 32% for 12-13% (C-1), 15-17% (C-2) and 22-23% (C-3) cemented specimens, respectively, which demonstrates that the durability of specimens increases with the increase in cementation level. A similar impact of cementation level on the cumulative mass loss could also be seen from the comparison (Fig. 8). The increased durability could be attributed to the formation of strong intact bonds during longer treatments. The longer treatment (e.g. C-3) resulted in higher precipitation of CaCO 3 at around particle-particle contact points, enhancing the resistance to the deterioration of calcite bridges and detachment of soil particles due to fatigue stresses developed during repeated WD cycles.
In a slope (as illustrated in Fig. S5 of supplementary file), surface zone/near-surface serves as the buffer-layer between atmosphere and inner substrates, significantly determining the slope responses such as infiltration, aggregate stability and deterioration processes (Dai et al. 2018;Zhang et al. 2018). During the MICP treatment, surface substrates would be cemented together by precipitated calcium carbonate. It is very clear that the treated surface would have a crucial role on the stability of the expressway slope by functioning as a cover material. As demonstrated in this laboratory study, the MICP-treated surface would gradually weaken under the exposure to repeated wetting and drying processes, i.e. due to rainfall, evapotranspiration and draughts. It is worth mentioning that an effective application (slope cover) requires compatibility of its permanence and durability with service life requirements. Treating the soil to 12-13% of CaCO 3 (i.e. C-1) is shown to have less durability in terms of aggregate stability and strength, yet the longevity is found to be increased by around two times when treated to 22-23% of CaCO 3 (i.e. C-3). Moreover, by deploying the MICP treatment to the slope surface, Jiang et al. (2019) found that high cementation resulted in less erosion, preserving slope surface against rainfall-induced damages. Gowthaman et al. (2020) demonstrated that the aggregate stability is attributed to the precipitation content of carbonate that resists the crystal-soil ruptures and detachments during the freeze-thaw. It is therefore clear that more precipitation of calcium carbonate corresponds to high durability of MICPtreated slope surface regardless of the environmental and/ or climatic factors.

Conclusion
MICP based bio-cementation has recently gained an increased attention as a promising alternative for the preservation of slope surface. In this study, the durability of MICP treatment upon cyclic wet-dry actions was evaluated at laboratory scale. The soil from Hokkaido expressway slope was treated by MICP and then subjected to continuous 50 number of WD cycles. During the WD cycles, changes in physical and mechanical responses were monitored using mass loss, S-wave velocities and UCS measurements. SEM was also performed to analyse the microstructural changes caused by WD actions. Based on the results, the following conclusions are drawn.
With the increase in number of WD cycles, the index properties of the specimens showed considerable deterioration in comparison with their initial values. Based on the observation, cyclic WD effects appeared to have adverse effects on the physical and mechanical characteristics of MICP-treated specimens. The study demonstrated that the deterioration is related to the degradation of calcium carbonates deposited in soil matrix, and that occurred in two stages: short-term and long-term. The short-term deterioration (STD) was significant, occurred rapidly within the first few WD cycles due to the suspension and erosion of weak carbonate deposits, resulting in significant loss of mass, shear wave velocity and UCS. The long-term deterioration (LTD) was not as significant as STD, occurred gradually, was prominently observable after around 30 number of WD cycles. SEM analysis demonstrated that intact carbonate deposits was deformed due to the fatigue stresses developed during repeated WD process. The developed stresses imposed increased degradation at surface of the specimens compared to that of core, which led to the detachment of surface particles and progression of macropores.
Moreover, it was found that the increase in cementation level would increase the durability of the MICP-treated specimens. When the soil was treated to high cementation level, the contact points were bonded strongly, which enhanced the resistance to the deterioration of calcite bridges due to WD fatigue stresses developed. The losses in mass and UCS were found to reduce by 41% and 55% when the cementation level in soil increased from 12-13 to 22-23%, indicated the enhancement in service life by around 2 times.