The addition of cementation within the pore network and on the surface of the grains results in strength enhancement and hydraulic conductivity alteration. In most hydrological applications, the hydraulic conductivity is the main property that controls any design. This section presents the two properties that change when applying an MICP injection scheme with a focus on hydraulic conductivity which is most relevant to the examined applications. The analysis focuses on both the macro- and micro-scales since both the amount of cementation and its microstructure characteristics control the response and behaviour of the resulting products.

The properties of the treated products are influenced by the characteristics of the base material (particle roughness, shape, size, and width of particle size distribution) as well as the distribution and structure of the cement within the medium (including the amount, crystal shape, size, and calcite location) [51,52,53,54,55]. Additionally, the cement properties are also governed by various other factors related to the MICP procedure and external/environmental factors, such as the number of injections, chemical concentration in each injection, retention time between injections, infiltration (or injection) rate, and bacterial density/activity, temperature, pH, and presence of oxygen [10,52,56]. Consequently, the mechanical properties may vary from one study to another, depending on the selection of biochemical parameters and the specific experimental protocol. To accurately assess the effects of each parameter, whether biological or related to the grains, it is essential to isolate other influences and follow a consistent procedure.

Both the strength enhancement and the hydraulic conductivity heavily depend on the microstructure. Three distinct and ideal types of grain/cement structures have been identified [57], as shown in Figure 3. These are:

(i)

contact-cementing, in which the carbonate crystals precipitate on and around the contacts between the grains (see schematic in Figure 3a);

(ii)

grain coating, in which cement forms a uniform film around the grains (see schematic in Figure 3b); and

(iii)

matrix-supporting, in which the precipitate is identified within the granular network (see schematic in Figure 3c).

The distribution of the cement around the grains depends on the grain characteristics but also on the choice of MICP formulation and other environmental conditions. The resulting mechanical and hydraulic properties presented in the next subsections will be linked to the microstructure, frequently referring to Figure 3.

In the existing literature, the most common method used to evaluate the strength of biocemented sands is the unconfined compressive strength (UCS), although other measurements, such as the small strain shear stiffness, G_max, tangent modulus [14,57,58], friction angle, cohesion derived from triaxial compressional tests [59,60,61], and even tensile strength [62,63], have been evaluated.

Starting from the biochemical factors, the literature suggests that the bacterial strain selected, the population and the urease activity, the type of calcium source, the concentration of chemicals, and the treatment duration and frequency of injections, but also the injecting method, affect the precipitating carbonate crystals in terms of size, type, and distribution within the pore network which, in turn, affect the resulting microstructure and, therefore, strength enhancement, as shown in Figure 3.

Previous research by Cheng et al. [53] indicates that lower urease activity can enhance particle bonding and increase unconfined compressive strength (UCS) due to slower formation of carbonate and more effective particle bridging (see Figure 3a). The findings of this study were strengthened by the study of Konstantinou et al. [46], in which a protocol was proposed on the basis of selecting bacteria with reduced urease activity as part of a low-concentration cementation solution during the MICP procedure which leads to balanced and slower reactions, thereby promoting consistent, reproducible outcomes and, eventually, leading to more efficient and better strength enhancement. However, Whiffin [9] reported significant variability in enzyme levels, suggesting that consistent urease production cannot be guaranteed even under conducive growth conditions. This variability of enzymatic activity could be either resolved by using bacteria strains of lower variability or native micro-organism isolates which could also minimize any environmental impacts or by following proposed methods to generate bacterial populations with the desired urease activity [46]. The authors [46] linked the differences in behaviour with the microscale in which it was observed that when the urease activity was higher, the carbonate crystals formed were of comparable size, exhibiting a cubic shape and uniform distribution on the grain surface. With decreasing urease activity, the crystals grew larger and tended to form clusters, appearing to be more efficient in creating bridges between the particles. These observations were particularly noticeable in higher bacterial populations and explain the better strength enhancement obtained. Various bacterial types have been also used, demonstrating the success of the procedure with almost any ureaseorganisms [19,38,53,64,65,66,67].

The bacterial population does also have effects on the size and structure of the carbonate distribution. Wang et al. [68,69,70] used a microfluidic chip to observe the various biochemical MICP parameters in response to the carbonate precipitation. An example of a microfluidic chip experiment is shown in Figure 4. The injection process of bacterial and chemical solutions, distribution of bacteria, precipitation, calcium carbonate crystal growth process, and its spatial distribution were studied. The findings indicate that a higher bacterial density results in a higher precipitation rate of CaCO₃, with more crystals but with a smaller average crystal volume [71]. This finding indicates that a medium range of optical densities leads to more efficient bonding of particles and, thus, higher strength.

The amounts of chemicals are also important parameters for these investigations. Al Qabany and Soga [51] demonstrated that reducing the chemical concentration level increases the unconfined compressive strength of biotreated specimens. In a relevant study at the grain scale level [72], the extent of the strength enhancement of the MICP-treated specimens was linked to the differences in crystal sizes and numbers and their growth dynamics. The authors identified that, regardless of the concentration of cementation, decreasing the normalized input rate (treatment frequency and duration) of the cementation solution resulted in a substantial increase in UCS. A further reduction in the normalized input rate led to a slight increase in UCS values. Poisson’s ratio in MICP-treated sands was found to decrease with increasing calcium chloride concentration, indicating less lateral deformation and volume change at the same axial strain compared to lower concentrations [57].

The second group of factors is environmental factors. These could include factors such as the pH, the presence of oxygen, the salinity level, the saturation, the environmental temperature, the presence of competitive micro-organisms, applied stresses, etc. These, of course, have effects on both the gaining of cohesion but also on the hydraulic conductivity reduction. In the study by Cheng et al. [53], the unconfined compressive strength (UCS) was measured for various degrees of saturation. The authors found that lower degrees of saturation resulted in strength enhancement during the MICP procedure, likely due to more effective cementation at particle contacts (see Figure 3a). Triaxial testing revealed that the initial tangent Young’s modulus (E_i) of MICP-treated sands was controlled by the CaCO₃ content and was less sensitive to an increase in effective confining pressure compared to untreated specimens [57]. Other studies also showed similar trends in friction angle and cohesion as a function of cementation level [59,60,61]. The durability of MICP-treated samples in freeze–thaw (FT) cycle tests was found to vary depending on the particle size distribution, with well-graded sands exhibiting improved resistance to FT cycles [53]. Additionally, conducting the MICP procedure at higher temperatures enhances the strength of the resulting samples [53,73,74]. However, at low or very high temperatures the bacterial growth was found to be lower, which has effects on crystal growth, as seen in the microfluidic experiments by Wang et al. [73] (see Figure 4). Such dynamics affect the resulting strength enhancement [75].

The third group of factors that affect the alteration of properties is granular network properties. The strength of cemented materials relies on the number of contact points and the position of the cement within the porous medium. A greater number of particles in a given volume leads to more contact points, distributing stresses evenly within the granular matrix and reducing stress on each particle. According to Konstantinou et al. [76], who systematically studied those effects, when cement is present at these contact points, it acts as a bonding agent, holding the particles together. The strength of these bonds depends on the amount of cementation around the contact points (effective cementation) which is shown in the ideal case in Figure 3a. In cases where there is no cementation, the contact points between particles become weak points. Large particles increase the likelihood of carbonate crystals forming on their surfaces, providing less effective cementation, resulting in a significant portion of the cementation not contributing much to strength improvement. While more contact points allow for more effective cementation, an excessive number of contact points requires a larger amount of cementation overall to enhance strength, as any uncemented contact points create preferential paths of failure under compression. Additionally, cementation may become random as the flow during treatment diverts from densely packed areas to those with less cementation. Very coarse sands with a larger average particle diameter had fewer particles but more surface area, making them less suitable for effective cementation, leading to consistently lower strength at any given cementation level (Figure 3b). On the other hand, very fine sands offered more contact points, making them better candidates for effective cementation, but their strength was still lower compared to fine sands [76,77]. If there are many contact points to be cemented, some might be left without cementation support, creating weak points susceptible to failure unless a substantial amount of cementation is provided. Fine-to-coarse sands strike a balance, providing the highest strength because of the combination of the absolute number of contact points and the ratio of contact points to surface grain area. An increase in the uniformity coefficient generates more contact points and reduces pore sizes, resulting in lower fluid flow during the treatment process. Therefore, mixtures with larger particles and a wide particle size distribution (PSD) achieved better strength enhancement since large particles offer fewer contact points, but the wider PSD reduces the chances of cementation depositing on the particle surfaces [76]. In materials with similar uniformity coefficients but different grain sizes, higher cementation levels were necessary to attain the same strength, primarily because smaller grains increase the likelihood of cementation occurring between particles rather than on their surfaces [76]. Particle shapes also influence mechanical properties [78]. Sub-rounded, spherical particles provided the highest strength due to their numerous contact points, while angular particles lacked sufficient contact points. Angular particles had initial strength from interlocking before cementation, and, while cementation increased strength, the effect was limited (i.e., diminishing returns with increasing cementation) [76]. On the other hand, Xiao et al. [78] demonstrated that a better base material for MICP treatments in terms of strength is provided by less round particles in mixtures of round and angular glass beads.

Figure 5 presents unconfined compressive strength values with respect to cementation level based on the results of various studies [3,51,53,72,76,77,79,80,81]. The literature contains a large number of data on UCS, yet some studies have been chosen to demonstrate its correlation with cementation level. As seen in the graph, in which the y-axis is in a log scale, an exponential fit describes the correlations, showing that cementation level is a main contributing factor despite the fact that experimental conditions are different across the studies (grain characteristics, chemical concentrations, bacterial population, and urease activity, temperature, and saturation level). These conditions are reflected in the precipitation patterns at the microscale level.

Perhaps, the hydraulic conductivity is the property that is most understudied for MICP-treated porous media. This might be due to the very variable results and the fact that until this point most of studies focused on strength enhancement which is usually the main objective in geotechnical engineering studies. However, controlling the hydraulic conductivity in applications related to hydrology and water resources is of paramount importance. It is influenced by various factors related to the pore space, including its size, shape, and tortuosity (i.e., the Kozeny–Carman equation). As stated previously, the fate of calcium carbonate crystals within the granular medium depends on the biochemical MICP parameters and the environmental factors but also on the properties of the porous medium itself.

Previous research on MICP-treated sands has consistently shown that the hydraulic conductivity decreases significantly as the cementation level increases even up to 10 orders of magnitude. This reduction in hydraulic conductivity was reported in several studies [51,76,79,82,83,84,85,86,87,88]. However, there was an exception noted in the study by Whiffin et al. [3], where the measured hydraulic conductivity showed little change after treatment. This was supported by other studies suggesting that specimens with lower amounts of precipitated calcite experience a less significant reduction in relative hydraulic conductivity [89]. In the studies by Dawoud et al. [80,90,91], the hydraulic conductivity was shown to have an initial ‘stable’ phase at the very low cementation region in which there was almost no reduction in hydraulic conductivity. Song et al. [92] presented a typical or normalized reduction profile as shown in Figure 6. In the initial stage of low cementations (stage I), the decrease in hydraulic conductivity occurs at a gradual pace and is marginal. Subsequently, there is a rapid decline in hydraulic conductivity (stage II) at moderate cementation levels. Finally, at high cementations, hydraulic conductivity reaches a nearly constant level which is very low (stage III). Considering this diagram, the findings of the previous studies could be combined and interpreted given that in the study by Whiffin et al. [3], the cementation levels were closer to the lower cementation levels compared to the rest. The general sense, though, is based on a comparison between hydraulic conductivity and UCS for Portland cemented samples and MICP-treated samples, that there is a smaller trade-off between hydraulic conductivity and strength in the case of MICP, suggesting a relative ability of retaining soil hydraulic conductivity [79].

Beyond this general profile, which only presents a correlation with the overall volume of cement added, the resulting hydraulic conductivity could be explained to a great extent by evaluating the microscale response. This involves assessing properties such as carbonate crystal type, distribution, and size, which are routed on the three groups of factors mentioned earlier.

The level of urease activity employed during treatment had only a minor effect on hydraulic conductivity reduction in the study by Cheng et al. [53] and according to Choi et al. [93]. However, in the research by Konstantinou et al. [46], the higher the urease activity was, the more clogged the specimen was at the injection point, decreasing dramatically the flow rate in subsequent injections. The difference between the two studies might be attributed to the fact that in the second work, the range of urease activities was wider. Also, a lower hydraulic conductivity was measured for higher bacterial populations [94].

Al Qabany and Soga [51] reported that the high concentration cementation solution produced a quicker and greater reduction in the coefficient of hydraulic conductivity, suggesting that higher concentrations of calcium chloride and urea lead to larger calcite crystals and a more uniform distribution of precipitation when using lower concentrations. The larger crystals can cause early clogging, as observed in their study.

Similar findings were reported by other studies [91,95]. In the study by Duo et al. [95], the hydraulic conductivity gradually decreased as the concentration of the solidification solution increased, with a maximum reduction of approximately three orders of magnitude. The hydraulic conductivity coefficient reduction was particularly prominent up to a concentration of 1.5 mol/L, beyond which it remained constant despite having higher concentrations of chemicals. The authors linked this behaviour with the microstructure stating that during biocementation, calcium carbonate accumulated on the surface of the sand particles (see Figure 3b) and filled the gaps between them, leading to a gradual reduction in the volume of sand pores and, subsequently, a decrease in hydraulic conductivity. According to the authors, these findings of absolute hydraulic conductivity values suggest that the MICP technique holds promise for seepage control in pond and landfill engineering projects in desert areas due to the low hydraulic conductivity obtained [95].

Dawoud et al. [91] classified the hydraulic conductivity reduction in three phases: during the initial stages of treatment hydraulic conductivity shows a slight decrease or remains relatively unchanged. At this point, a small amount of precipitated calcite adds stiffness to the soil without causing pore clogging. This phase can be represented by a linear relationship with a slight negative slope (in agreement with the profile in Figure 6). As the precipitation of calcite continues, a certain threshold is reached where the accumulated calcite starts to clog the pores, leading to a new steeper trend of decreasing hydraulic conductivity after each treatment. During this stage, the concentration of the chemical solutions used in the treatment significantly influences the soil’s behavior and characteristics (again in agreement with Figure 6). The study noticed that using a 1M concentration for the urea–CaCl₂ solution resulted in an earlier transition to the clogging phase. Once clogging initiates, the distribution of MICP becomes more uncertain. Blocked flow paths cause new precipitates to accumulate near the injection point, resulting in less uniform treatment across the sample. This phenomenon is consistent with the findings of Qabany and Soga [51].

The calcium source is also another biochemical factor that has effects on hydraulic conductivity. In the study by Kadhim et al. [96], the incorporation of a cementation solution containing calcium chloride derived from eggshells had a substantial impact on hydraulic conductivity. However, the effect was more pronounced in the silica sand samples rather than the river sands as in the former higher hydraulic conductivity reduction was observed. In other studies, the use of calcium acetate caused the greatest hydraulic conductivity reduction followed by calcium chloride. A very low hydraulic conductivity reduction was observed with the use of calcium nitrate [97].

MICP process under lower saturation conditions is more favorable if the goal is to improve the mechanical properties while still maintaining relatively high residual hydraulic conductivity [79]; however, it was also reported that there is a general trend of decreasing hydraulic conductivity with the increase in produced calcite content (CaCO₃) irrespective of the degree of saturation at which the soil was treated [98]. The differences observed are likely due to the interaction effects (a variable is behaving differently at various levels of another variable) added from the choice of other biochemical parameters.

The relative density (RD) and injection volume in a single injection event (VIP—void injection percentage) were also studied showing a negative correlation with hydraulic conductivity [99]. That is, as the two factors increased, there was a corresponding decrease in hydraulic conductivity. This reduction in hydraulic conductivity was attributed to several reasons. With an increase in RD, the pore volume (PV) decreased, leading to the formation of smaller and less permeable pore throats. However, at very high RD some inconsistency was observed, likely due to the pore throats becoming smaller, increasing the likelihood of clogging and facilitating the creation of preferential flow paths during bacterial suspension and cementation fluid injection. Moreover, the decrease in hydraulic conductivity with increasing VIP was attributed to localized bacterial concentration near the injection point, which promoted more significant calcite precipitation in that specific region. This localized effect led to a reduction in hydraulic conductivity in the surrounding area [99].

One of the very few environmental factors studied for MICP in relation to the reduction of hydraulic conductivity was the use of seawater. The hydraulic conductivity coefficients of the samples treated with natural seawater-based biocementation were slightly higher compared to those treated with freshwater-based biocementation. This difference was attributed to a higher precipitation of carbonates in the freshwater columns than in the seawater columns. However, despite this disparity, using seawater-based biocementation instead of freshwater cementation did not have a significant impact on hydraulic conductivity [22]. Similarly, the ureolytic bacteria utilized by Cheng et al. [67] were acclimated to high-salinity conditions by employing a growth medium containing high concentrations of ammonium sulfate ((NH₄)₂SO₄) to help prevent significant osmotic effects when exposed to seawater. The use of seawater supplemented with urea, instead of a concentrated cementation solution, had no substantial impact on the hydraulic conductivity per carbonate formed. However, the same carbonate formation resulted in higher strength in the seawater cementation trials, which, consequently, allowed for greater hydraulic conductivity for a given level of strength [67].

In terms of injection strategy, the use of continuous flow technique resulted in a significant reduction in hydraulic conductivity compared to the stopped-flow technique. This was attributed to the higher likelihood of columns treated with continuous flow to experience plugging near the injection source [84]. Also, the presence of a stationary liquid during biocementation resulted in slightly higher hydraulic conductivity reduction [100].

The work by Konstantinou et al. [76,82] on the grain characteristics effects on hydraulic conductivity revealed valuable insights on this link. Hydraulic conductivity undergoes a significant decrease when cement is present at contact points, leading to a reduction in pore throat size. Materials with numerous particle-to-particle contacts prove challenging to decrease hydraulic conductivity, as many contacts require cementing. Such materials regulate flow through paths with larger pores and pore throats, which involve uncemented particle contacts.

As the grain size increases in a packed bed of solids, the hydraulic conductivity is expected to rise as well (see example MicroCT images in Figure 7a–c). For example, the Kozeny–Carman equation incorporates the squared average particle diameter in its empirical equation’s numerator. This trend of reduced hydraulic conductivity with increased cementation was also observed in very fine to coarse sands during the study and is in agreement with other studies where a D₁₀ increase, resulted in a smaller reduction in relative hydraulic conductivity [89].

However, very coarse sands and gravel did not follow the same pattern and displayed a less controlled reduction in hydraulic conductivity [76]. In very coarse materials, carbonate precipitation mainly occurs on the surface of particles rather than at contacts among them. As a result, the reduction in hydraulic conductivity is not expected to be significant. Nevertheless, the lower number of contact points leads to fewer flow path options, explaining the rapid decline of hydraulic conductivity in very coarse particles at lower cementation levels. When comparing gravel and very coarse sand (both deemed ineffective for MICP treatment), gravel exhibited a larger reduction in hydraulic conductivity due to the lower number of available flow paths. Gravel’s erratic pore distribution and randomness, caused by its very large grains, provide for unpredictable fluid flow paths, resembling a system with small and few large pipes dominating the flow [76].

Hydraulic conductivity in relation to cementation is lower when the base material has a wider spread of particle size distribution (PSD) due to the narrower initial pore space distribution before cementation (see example MicroCT images in Figure 7d–f). The reduction in hydraulic conductivity with increasing carbonate content is also lower compared to more uniform sands, as there are too many narrow flow paths to be cemented (and, essentially, closed) when a higher number of contact points exists. This holds true for granular materials with wide PSD [76]. This was confirmed by other studies in which the hydraulic conductivity of cemented sand was found to be influenced also by its grain gradation in a similar manner. The reduction in the hydraulic conductivity coefficient during each MICP treatment cycle increased with higher values of the uniformity coefficient (C_u) and the curvature coefficient (C_c) [86].

On the other hand, the absolute value and reduction of hydraulic conductivity in two materials with similar uniformity coefficients but different grain sizes are the same, demonstrating that the dominant factor controlling flow is the spread of PSD [76]. Although the material with smaller grain sizes has more contact points, it provides more flow path options compared to the one with larger grain sizes. At the same time, the latter has a lower ratio of contact points over surface grain area, resulting in some of the cementation being consumed on the surface of the grains, leading to a lower rate of hydraulic conductivity reduction.

While particle sphericity is known to impede flow, according to the Kozeny–Carman equation, significant hydraulic conductivity differences were only observed in the case of angular sand in the study by Konstantinou et al. [76]. However, the reduction trend with respect to cementation levels was similar for angular sand, fine and coarse glass beads, and fine and very coarse subrounded sand, indicating that grain size is the dominant factor in these cases. The flow paths are affected, to some extent, by grain shapes, but the addition of cementation causes proportional hydraulic conductivity reduction in spheres, sub-rounded, and angular sands. In the study by Song et al. [101], though, the non-spherical particles (crushed Ottawa sand) experienced the highest drop in hydraulic conductivity compared to the spherical particles despite having a lower calcium carbonate content showing a specific trend: the angular grains exhibited the highest hydraulic conductivity reduction, followed by the near-spherical particles, with the spherical particles showing the least reduction [101].

To investigate how pore-scale CaCO₃ distributions affect the hydraulic conductivity of MICP-treated sands, the researchers used the Panda–Lake model [102]. This model incorporated three reduction factors: the porosity reduction factor, tortuosity reduction factor, and specific surface area reduction factor. Additionally, the Kozeny–Carman model was used to estimate hydraulic conductivity reduction with CaCO₃ content, considering only the reduction of porosity [103]. By comparing the calculated hydraulic conductivity using the Panda–Lake and Kozeny–Carman models with the measured hydraulic conductivity reported in existing literature, the researchers developed an analytical model that can reasonably predict the hydraulic conductivity of MICP-treated sands for different CaCO₃ contents and types of sands [103].

Lin et al. [103] performed an analysis on measured hydraulic conductivity values across various studies identifying that the grain coating (Figure 3b) Panda–Lake model provides reasonable fits to the data provided that the main mechanism is matrix-supporting. The Panda–Lake model takes into account, the shape, tortuosity, specific surface area (surface area of the grain/the volume of the grain), the statistical characteristics of the particle size distribution, the cement saturation of the pore space, the fraction of CaCO₃ volume to the total volume of solids, and the specific surface area of the CaCO₃ crystals. The matrix-supporting environment seen in Figure 3c shows smaller reduction of hydraulic conductivity with respect to cementation level. It follows, based on the findings of Lin et al. [103] and Konstantinou et al. [76], that the Kozeny–Carman equation would give better estimations of the reduction of hydraulic conductivity for the contact-cementing model.

Even though the findings are in good agreement, there is room for further research to examine the combinations of factors and their effects on the resulting hydraulic conductivity. Figure 8 presents the data concentrated from the available references that measured hydraulic conductivity [3,22,51,53,67,76,79,82,83,84,85,86,87,88,89,91,92,93,94,95,96,97,98,99,100,101,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128,129,130]. The y-axis is in a log scale. Despite the fact that there is a trend in reduction similar to the profile presented in Figure 5, the results are scattered showing a weak correlation with the cementation level. This is because not only the volume of cementation is required to identify the reduction in hydraulic conductivity but also the carbonate crystal size and distribution is required. Also, the initial configuration of the granular network is required, and this is the reason the Panda–Lake model performs better compared to the Kozeny–Carman equation. The results also show that UCS has a higher correlation coefficient (0.55) (Figure 5) compared to hydraulic conductivity (with a value of 0.1) showing more dependence on cementation level and less dependence on the granular and cement configuration.