Investigating Ammonium By-product Removal for Ureolytic Bio-cementation Using Meter-scale Experiments

Microbially Induced Calcite Precipitation (MICP), or bio-cementation, is a promising bio-mediated technology that can improve the engineering properties of soils through the precipitation of calcium carbonate. Despite significant advances in the technology, concerns regarding the fate of produced NH4+ by-products have remained largely unaddressed. In this study, five 3.7-meter long soil columns each containing one of three different soils were improved using ureolytic bio-cementation, and post-treatment NH4+ by-product removal was investigated during the application of 525 L of a high pH and high ionic strength rinse solution. During rinsing, reductions in aqueous NH4+ were observed in all columns from initial concentrations between ≈100 mM to 500 mM to final values between ≈0.3 mM and 20 mM with higher NH4+ concentrations observed at distances furthest from the injection well. In addition, soil Vs measurements completed during rinse injections suggested that no significant changes in cementation integrity occurred during NH4+ removal. After rinsing and a 12 hour stop flow period, all column solutions achieved cumulative NH4+ removals exceeding 97.9%. Soil samples collected following rinsing, however, contained significant sorbed NH4+ masses that appeared to have a near linear relationship with surrounding aqueous NH4+ concentrations. While these results suggest that NH4+ can be successfully removed from bio-cemented soils, acceptable limits for NH4+ aqueous concentrations and sorbed NH4+ masses will likely be governed by site-specific requirements and may require further investigation and refinement of the developed techniques.

Microbially Induced Calcite Precipitation (MICP), or bio-cementation, has shown significant promise as an environmentally-conscious alternative to geotechnical ground improvement technologies, which have traditionally relied upon hazardous grouting chemicals, high mechanical energy, and energy-intensive materials to improve the engineering properties of soils [1][2][3] . In the urea hydrolysis (ureolysis) driven process, microorganisms containing urease enzymes are used to catalyze a reaction that degrades urea, yielding total ammonium (NH 4 + ), dissolved inorganic carbon, and hydroxide ions 4 [Eqs. (1)(2)(3)]. When soluble calcium is available from either treatment solutions or groundwater, resulting alkalinity can supersaturate solutions with respect to calcium carbonate (CaCO 3 ) and initiate mineral precipitation on soil particle surfaces and contacts [Eq. (4)]. Resulting bio-cementation can improve the engineering properties of soils through large increases in shear stiffness and strength with small reductions in hydraulic conductivity and porosity [5][6][7][8] . The process has been proposed for a variety of applications including mitigation of earthquake-induced soil liquefaction, general geotechnical soil improvement, subsurface flow manipulation, and divalent contaminant immobilization among other uses 1,[9][10][11][12]  Researchers have made significant advances in the technology in recent years, including identifying alternative treatment techniques 13,14 , up-scaling the process to meter-scale [15][16][17][18][19] , and characterizing the engineering behavior of bio-cemented soils 8,20 . Despite these advances, environmental concerns regarding the fate of produced NH 4 + by-products have remained largely unaddressed 1,21 . NH 4 + is a commonly encountered source of inorganic nitrogen in soil systems, however, high aqueous NH 4 + concentrations produced following ureolytic MICP may present serious environmental and human health concerns if left untreated in soils and groundwater. For example, the presence of high NH 4 + in surface waters can encourage the growth of toxic algal blooms, which can decrease aquatic dissolved oxygen availability, produce high levels of toxins, and encourage bacterial growth, therefore threatening fish, humans, and other flora and fauna 22 . While no maximum contaminant level has been established by the U.S. EPA for total NH 4 + in drinking water, maximum concentrations for aquatic life of 17 mg/L (≈1 mM) and 1.9 mg/L (≈0.1 mM) total NH 4 + for acute and chronic exposure, respectively, have been recommended 23 . During field-scale applications, aqueous NH 4 + concentrations produced during bio-cementation will likely require removal to meet water quality standards and or maintain similar water quality as that present prior to treatments. NH 4 + concentrations near twice the concentration of applied urea are expected by reaction stoichiometry [Eq. (1)], with many past experiments resulting in the production of NH 4 + concentrations exceeding 500 mM. Limited understanding of post-treatment NH 4 + removal has been a significant barrier for bio-cementation technology that has limited field-scale applications and environmental benefits. In order for MICP to be a truly environmentally beneficial technology, methods to manage, remediate, or remove NH 4 + by-products following bio-cementation are needed.
While in-situ nitrification of produced NH 4 + to nitrite (NO 2 − ) and nitrate (NO 3 − ) (under aerobic conditions) and subsequent denitrification of NO 3 − to nitrogen gases (under anaerobic conditions) may offer a potential remediation strategy, these processes will be challenging and likely unpractical at field-scale due to the need for dramatic modifications in subsurface oxygen availability and chemical conditions. In addition, Gat et al. 21 showed that in-situ oxidation of NH 4 + can have detrimental effects on bio-cementation integrity resulting from the generation of acidity in this process, further suggesting the need to address produced NH 4 + following bio-cementation. Although few studies have examined and quantified the removal of NH 4 + by-products following MICP, "rinse" solutions have been most commonly applied to remove NH 4 + from treated areas with subsequent remediation of collected effluent completed ex-situ at water reclamation facilities 15,24. Post-treatment rinsing may require significant energy and materials when applied at field scale, however, the technique can provide an effective strategy for NH 4 + management following bio-cementation in the absence of other remediation processes and may provide an opportunity for NH 4 + recovery. Recently, centimeter-scale soil column experiments were performed to further investigate and quantify the removal of NH 4 + by-products following MICP in a clean poorly-graded Concrete Sand material using rinse solution injections 25 . Results suggested that removal of positively-charged NH 4 + ions from clean sands may be significantly more difficult due to interactions with negatively-charged soil particle surfaces. When rinse solutions of differing ionic strength and pH were applied, NH 4 + removal was improved with increases in ionic strength, while pH had no detectable influence on removal. Improved removal was attributed to the exchange of NH 4 + existing on soil surfaces with Ca 2+ cations supplied from rinse solutions. A high pH (pH = 9.0) and high ionic strength (500 mM CaCl 2 ) rinse solution was identified that improved NH 4 + removal while simultaneously limiting dissolution of bio-cementation. Although these results were promising, it remained unclear if these techniques would remain effective at meter-scale treatment distances.
In this study, five 3.7-meter long soil columns were improved using ureolytic bio-cementation, and post-treatment NH 4 + by-product removal was investigated as a function of different soil materials and treatment techniques. Columns contained three different poorly-graded sand materials including: an alluvial sand (Column 1, 2, and 3), a marine sand (Column 4), and a second alluvial sand material (Column 5). Prior to bio-cementation, all columns received different biological treatments over the first six days intended to either enrich native ureolytic microorganisms or augment Sporosarcina pasteurii (S. pasteurii) bacteria uniformly across column lengths. Three different biological treatment strategies were applied to enrich native ureolytic microorganisms to achieve high (Column 1) and low bulk ureolytic rates (Column 2, 4, 5) and augment soils with S. pasteurii at a high cell density (Column 3) to obtain a high ureolytic rate similar to Column 1. Differences in achieved ureolytic rates were intended to examine the effect of urea hydrolysis rate on the spatial uniformity and extent of bio-cementation; the results of which are more extensively described in a separate manuscript 26 . Following biological treatments, nine cementation injections containing identical calcium and urea concentrations were applied to all columns over nine (Column 1, 3) and eighteen (Column 2, 4, 5) days. All columns achieved differing degrees of bio-cementation with distance from the injection well. Following MICP treatments, 525 L of a high pH and high ionic strength rinse solution was applied to each column to remove produced NH 4 + . During rinse injections, changes in NH 4 + by-product removal and effects on bio-cementation integrity were monitored. Following rinsing, physical soil samples were collected at various locations and sorbed NH 4 + masses remaining on soil surfaces were quantified. Soil columns. Five 3.7-meter long hollow steel columns with square cross-sections (0.2 m by 0.2 m) contained 0.15 m 3 of sand and received treatment injections in one direction to simulate a single stream tube within a well-to-well half-space. Treatment wells (ID = 26.6 mm) were located on column ends and had valves, which allowed for pressurization of the injection well and removal of effluent solutions at the outlet well. Soils were placed in three ≈16 to 76 mm lifts, which allowed for bender element sensors to be placed at mid-height within columns during soil placement. Lifts were moist tamped and surfaces were scarified to limit density differences between lifts. Table 2 summarizes soil material types, soil column porosities, relative densities, and pore volumes determined from placed soil masses. In all columns, estimated porosities were between 0.30 and 0.40, relative densities were between 55% and 67%, and column pore volumes (PV) varied between 48.5 L and 63.0 L. Columns contained four bender element sensor pairs at distances of 0.31 m, 1.33 m, 2.35 m, and 3.37 meters from the injection location to monitor changes in soil shear wave velocities (V s ), indicative of bio-cementation. Four aqueous sampling ports were placed at distances of 0.15 m, 0.83 m, 1.82 m, and 2.81 m away from the injection location to monitor injection pressures and obtain solution samples at various times. Sampling ports consisted of 0.15 m long steel tubes (ID = 3.35 mm) with 25.4 mm diameter circular plastic filters  μm) open at mid-height within columns. A fifth drain port, consisting of a 60.3 mm ball valve with a porous filter material, was used to obtain additional samples and was located 3.58 m away from the injection location at the bottom of columns. Figure 1 provides a detailed schematic and images of soil columns including treatment application systems, bender element and sampling port locations, and treatment wells.
Saturation and bromide passive tracer testing. Prior to all treatments, columns were slowly saturated with an artificial ground water (AGW) solution containing 40 μM KNO 3 , 450 μM MgSO 4 , 1.75 mM CaCl 2 , 40 μM NaNO 3 , 1.1 mM NaHCO 3 , and 60 μM KHCO 3 following Ferris et al. 30 . Immediately following saturation and before all bio-cementation treatments, columns received passive tracer injections to evaluate differences in solution transport between columns. During tracer testing, 76 L of a 15 mM NaBr solution was injected, followed by 76 L of de-ionized water to examine the arrival and removal of passive Br − ions at the outlet well. A constant flow rate of 400 mL/min was used for injections and solution samples were collected at outlet wells once every 5 minutes. Solution conductivities were measured and normalized by the conductivity of the injected NaBr solution to estimate normalized Brconcentrations (C/C 0 ). One-dimensional advective-dispersive soil column transport models were developed using the geochemical software PHREEQC 31 and were used to match passive tracer experimental observations by varying porosities and longitudinal dispersivities for all columns. All models were composed of 42 cells (0.089 meters each) and received NaBr and de-ionized water injections that were identical to the physical experiment.   www.nature.com/scientificreports www.nature.com/scientificreports/ solutions were applied to either enrich soil columns for native ureolytic microorganisms (Column 1, 2, 4, 5) or augment soils with the highly active ureolytic bacterium, S. pasteurii (Column 3). Enriched columns received six enrichment treatments once daily with varying concentrations of yeast extract intended to achieve different ureolytic activities. Column 1 received higher yeast extract concentrations (0.2 g/L) following past experiments 14   www.nature.com/scientificreports www.nature.com/scientificreports/ wherein high ureolytic activities were targeted (hydrolysis of 250 mM urea within ≈8 hours). Columns 2, 4, and 5 received lower yeast extract concentrations (0.04 g/L) to obtain a low bulk ureolytic rate (hydrolysis of 250 mM urea within ≈48 hours). Column 3 was inoculated on the last day of enrichment treatments by injecting 76 L of an isotonic saline solution (9 g/L NaCl) containing S. pasteurii (ATCC strain 11859) at a cell density of 9.36 × 10 7 cells/mL, intended to match the high ureolytic rate of Column 1. Due to lower ureolytic activity than expected along Column 3, a second augmentation injection consisting of 456 L of isotonic saline with 1.4 × 10 6 cells/mL was applied after the fourth cementation injection. Following all enrichment treatments, a flush solution that was identical to cementation solutions, but did not contain urea and Ca 2+ , was applied to all enriched columns immediately before the first cementation treatment to remove high aqueous carbonate species expected after enrichment. Cementation treatments containing Ca 2+ were then applied to all columns to initiate CaCO 3 precipitation for a total of nine injections. Columns 1 and 3 received cementation treatments once every 24 hours and Columns 2, 4, and 5 received treatments once every 48 hours to allow near full hydrolysis to occur. During enrichment, augmentation, and cementation treatments, injection volumes of 76 L were applied at a constant flow rate of 400 mL/ min, during which injected solutions were continuously mixed within injection tanks. Following cementation treatments, a 200 mM CaCl 2 solution (initial pH ≈ 10.0) was applied to all columns to remove NH 4 + following previously identified strategies 25 . During NH 4 + by-product removal injections, a rinse solution volume of 525 L was injected into each column at a flow rate of 750 mL/min and lasted ≈700 minutes. Following rinse injections, columns remained saturated for 12 hours during a stop flow period until columns were disassembled. All solutions were injected using small pumps (Wayne Inc., 0.1 HP) with pressure gauges to monitor injection pressures.
Aqueous sampling. Before and immediately after treatment injections, aqueous samples of ≈10 mL were collected from all sampling ports. On select days, aqueous samples were also collected 1, 2, 4, and 8 hours after injections to monitor ureolytic activity and chemical changes in time. Prior to all sample collection events, 30 mL of solution was removed from ports and discarded to obtain representative samples. During NH 4 + rinse injections, ≈10 mL samples were collected from outlet wells once every 20 minutes and from all sampling ports and well locations once every 60 minutes. Aqueous samples were immediately frozen and stored at −20 °C until subsequent chemical analyses.
Shear wave velocity measurements. Shear wave velocity (V s ) measurements were obtained using horizontally-oriented bender element sensor pairs placed at mid-depth within columns at various distances from the injection well. V s measurements were obtained at all locations before and immediately after all biological and cementation injections. Additional measurements were performed during NH 4 + rinsing before, 120, 240, 480, and 700 minutes after the start of rinse injections and following a 12-hour stop flow period (24 hours after start of rinse injections). Bender elements were coated with epoxy, electronics wax, and an insulating coating to waterproof sensors following similar processes detailed in Gomez et al. 18 . Transmitting bender elements were excited with a 24 V 100 Hz square wave and signals from receiving bender elements were measured and recorded using an oscilloscope at a sampling frequency of 1 MHz. Shear wave arrival times were interpreted visually and V s values were calculated from measured sensor spacings.
Aqueous measurements. Solution pH measurements were completed using a semi-micro pH electrode and meter system that was calibrated daily using a three-point buffer sequence (4.01, 7.00, 10.00) and had ± 0.01 pH unit accuracy. Total NH 4 + measurements were completed using a salicylate reaction method similar to Krom (1980) 32 , wherein two reagents (Reagent A & B) were added to dilute sample volumes and absorbance values were measured at 650 nm using a microplate spectrophotometer. Reagent A consisted of 1.9 mM sodium nitroprusside, 811 mM sodium salicylate, 387 mM sodium citrate, and 515 mM sodium tartrate in water. Reagent B consisted of 1.32 mM sodium hypochlorite and 1.5 M sodium hydroxide in water. Urea measurements were completed using a colorimetric urea assay similar to Knorst et al. 33 . A colorimetric reagent consisting of 216 mM p-dimethylaminobenzaldehyde, 2.32 M hydrochloric acid, and 13.83 M ethanol in water was added to dilute samples and absorbance values were measured at 422 nm using a microplate spectrophotometer. Conductivity measurements were completed using a conductivity probe and meter. Augmented cell densities were determined through direct counting 34 . cation exchange capacity and exchangeable cation measurements. Cation exchange capacity (CEC) and exchangeable cation measurements were completed for untreated sand materials using a process similar to U.S. EPA Method 9080 35 . Soil CEC values reflect the capacity of negatively-charged soil surfaces to bind positively-charged ions, thus it was hypothesized that CEC differences between soils could influence NH 4 + transport and removal following bio-cementation. During these measurements, 10 grams of untreated dry soil and 50 mL of a 1 M NH 4 Cl solution were added to a plastic syringe. After a 12-hour residence period, soil solutions were extracted, collected, and select exchangeable cations were characterized using inductively coupled plasma mass spectrometry (ICP-MS www.nature.com/scientificreports www.nature.com/scientificreports/ from thawed moist soil samples using a centrifuge process wherein 30 grams of moist soil samples (of known water content) were added to conical centrifuge tubes with 0.45 micron nylon filter baskets and centrifuged at 4150 rpm for 20 minutes to extract solutions. A minimum volume of 2 mL was collected, frozen, and stored at −20 °C until NH 4 + concentrations were analyzed. Sorbed NH 4 + masses remaining on soil particle surfaces were quantified using a KCl extraction process following Keeney & Nelson (1982) 36 . In this process, 10 gram moist soil samples (of known water content) were mixed with 20 mL of a 2 M KCl solution and allowed to equilibrate for at least 4 hours to facilitate removal of NH 4 + ions. Soil solution mixtures were then added to another conical filter tube, centrifuged, and a 2 mL solution sample was collected, frozen, and stored at −20 °C until NH 4 + concentrations were analyzed. KCl extracted NH 4 + measurements included NH 4 + initially present in free soil solutions as well as NH 4 + masses that were initially sorbed to soil surfaces. Sorbed NH 4 + masses were therefore estimated by subtracting NH 4 + masses expected from free solution from NH 4 + measurements following KCl extraction. Sorbed NH 4 + masses were normalized per gram of dry soil.
Statement on consent to publish. The authors have notified persons shown in Fig. 1 and have received consent to publish images.

Results and Discussion
Results from soil CEC and exchangeable cation measurements are summarized below in Table 4 for all three sands. A fourth sample consisting of Concrete Sand augmented with 10% by mass reagent-grade CaCO 3 was also tested to assess the potential influence of CaCO 3 minerals on soil CEC. As shown, Delta Sand had the highest CEC (4.32 meq/100 g) with Covelo Sand exhibiting the lowest CEC of tested sands (1.64 meq/100 g). Additionally, when CaCO 3 was added to the Concrete Sand sample, no significant effects on CEC were observed suggesting that the presence of bio-cementation CaCO 3 minerals likely had little influence on the ability of soils to bind NH 4 + ions. When examining exchangeable cation concentrations, sands generally had similar values for common soil cations. Delta Sand, however, had notably higher K + , Mg 2+ , Na + , and S 2+ concentrations than other tested sands, which was consistent with the marine depositional environment from which it was obtained. Exchangeable cations in Concrete Sand and Covelo Sand were most similar, however, Covelo Sand had significantly higher Al 3+ and Ca 2+ concentrations as well as much lower Mg 2+ concentrations than other sands. While tested sands had minor CEC and exchangeable cation differences, values were generally consistent with those expected for clean quartz sands 37 . Figure 2 presents measurements of normalized solution conductivity (C/C 0 ) versus injected NaBr tracer solution volume from samples obtained at outlet wells during passive tracer testing. Differences in C/C 0 values with injected volume are reflective of porous media advective-dispersive transport properties including porosity, hydrodynamic dispersion, and diffusion 38 . Immediately after starting the 15 mM NaBr solution injection, most C/C 0 values were near zero with small values (<10% C/C 0 ) observed exiting columns due to background soil solution conductivities. In Column 4, containing Delta Sand, notably larger initial C/C 0 values were observed and were consistent with the higher exchangeable cations measured previously (Table 4). After additional tracer solution was injected, C/C 0 values increased above background levels (C/C 0 ≈ 20%) in Column 1, 2, 3, and 5 after injecting ≈20 to 30 L (0.39 to 0.62 PV). Column 4 had a much more delayed arrival, however, with C/C 0 values increasing above ≈20% only after ≈42 L (0.66 PV). The delayed breakthrough in Column 4 was consistent with the higher porosity measured in this column ( Table 2). After injecting the 76 L volume, C/C 0 values measured at the outlet well were between 94% and 96% for Column 1 and 2, 89% and 90% for Column 3 and 5, and were 98% for Column 4. While it was expected that Column 1, 2, and 3, which all contained Concrete Sand, would have similar breakthrough curves, the Column 3 breakthrough was distinctly different. Column 3 and 5, however, were found to have very similar breakthrough curve behaviors despite containing different sands. While unexpected, both Concrete Sand and Covelo Sand had similar grain size distributions and differences in transport properties may have resulted from differences in soil preparation between columns and minor preferential flow at the column boundaries. To better quantify transport differences between columns one-dimensional advective-dispersive soil column transport models were calibrated to match experimental trends by varying column longitudinal dispersivities (α) and porosities (n). Modelled results are presented in Fig. 2 and porosity and dispersivity values used for models are summarized in Table 2. Modelled porosities were lower than those calculated from soil mass measurements for all columns, likely due to incomplete saturation, which could have reduced the apparent porosity of columns. When comparing dispersivity values between columns, a large increase in longitudinal dispersivities was observed between Column 1 and 2 (α = 0.3 m) and Column 3 and 5 (α = 1.5 m), despite having similar porosities. Column 4 had a similar dispersivity (α = 0.1 m) as Column 1 and 2, but a much larger modeled porosity (n model = 0.33) as expected from soil mass measurements.  www.nature.com/scientificreports www.nature.com/scientificreports/ Figure 3 presents aqueous NH 4 + concentrations in time following the ninth cementation injection for Column 1, 2, and 3 from measurements at all port locations, immediately after injections, and 1, 2, 4, 8, 24, and 48 hours (when applicable) after injections. Aqueous NH 4 + concentrations were determined by difference and reaction stoichiometry from direct urea measurements in time. While similar data were also obtained for Column 4 and 5, only Column 2 data are shown and was representative of trends observed in all enriched low ureolysis rate columns (Column 2, 4, 5). As shown, similar NH 4 + production trends in time were observed between columns at each port location. Immediately after injections, aqueous NH 4 + concentrations were lowest at the 0.15 m port location and increased with distance from the injection well for all columns due to urea hydrolysis occurring during solution transport as well as mixing with previously reacted solutions. At a distance of 0.15 m in the high   www.nature.com/scientificreports www.nature.com/scientificreports/ ureolysis rate Column 1, NH 4 + concentrations increased from post-treatment values near ≈50 mM to ≈500 mM over 24 hours, while ≈500 mM concentrations were achieved at all other ports within 1 hour after injections. In the low ureolysis rate Column 2, NH 4 + concentrations increased from post-treatment values near ≈50 mM to ≈500 mM at a distance of 0.15 m within 48 hours, and to ≈500 mM at all other ports between 6 and 24 hours after injections. In the augmented Column 3, a slower rate of urea hydrolysis was consistently observed at all port locations relative to enriched columns. This resulted in significantly lower NH 4 + concentrations between ≈200 mM and ≈400 mM residing after 24 hour treatment periods and a significant fraction of injected urea remaining non-hydrolyzed due to low ureolytic activity. Localization of ureolytic activity near injection well locations has been previously observed in augmented experiments 39 and likely resulted from the colloidal filtration of S. pasteurii bacterial cells during augmentation of Column 3 40 . Despite much lower NH 4 + production observed in Column 3, all enriched columns (1, 2, 4, 5) achieved similar post-treatment NH 4 + concentrations near ≈500 mM after 24 hour (Column 1) and 48 hour residence periods (Columns 2, 4, 5). Figure 4 presents solution pH and aqueous NH 4 + concentration measurements for outlet well samples versus injected rinse solution volume. As shown in Fig. 4a, solutions initially exiting all enriched columns had pH values between 7.6 and 8.0. This was consistent with previously observed solution pH values for cementation solutions with urea-to-calcium ratios of 1:1 following near full urea hydrolysis. In Column 3, however, initial pH values were significantly lower (near 6.8) due to limited generation of alkalinity from limited urea hydrolysis. As rinse injections proceeded, effluent pH values approached 7.0 for all columns after injecting 100 L. In all enriched columns (Column 1, 2, 4, 5), pH values then gradually increased to near steady values between 7.4 and 8.2 for the remainder of rinsing. The temporary reduction observed in outlet solution pH values was unexpected as the initial pH of the injected rinse solution was near 10. It is hypothesized that this pH reduction may have resulted from consumption of remaining carbonate species and some limited calcite precipitation upon the initial introduction of rinse solutions with 200 mM Ca 2+ . Following the removal of sufficient solution alkalinity, increases in pH may then have occurred due to equilibration of high pH solutions with existing CaCO 3 and soil minerals. In Column 3, pH values remained lower than all enriched columns with final outlet well solution pH values near 6 to 6.5, which likely resulted from a lack of significant quantities of CaCO 3 minerals in this column.
Large reductions in effluent NH 4 + from values between 497 mM and 524 mM (Column 1, 2, 4, 5) and ≈342 mM (Column 3) to values between ≈20 and 32 mM were observed in all columns at the outlet well after injecting 200 L (Fig. 4b). Following the application of an additional 200 L, however, NH 4 + concentrations were only reduced to values between 3.5 and 9.0 mM. At the end of the 525 L injection, all effluent NH 4 + concentrations were between 0.8 and 2.5 mM. It is hypothesized that the limited improvement in NH 4 + removal after injecting significantly more rinse solution volume was influenced by soil-ion interactions and removal of sorbed NH 4 + masses, though solution mixing via hydrodynamic dispersion may have also contributed. In order to evaluate the effect of these interactions on observed NH 4 + trends, PHREEQC models were used to simulate passive tracer removal trends for all columns. As shown, concentrations were similar between measured NH 4 + and modelled passive tracers early during rinsing due to limited breakthrough at the outlet well location, however, at injection volumes greater than 100 L, modeled passive tracer concentrations were significantly lower than experimentally observed NH 4 + values. For example, reductions in passive ion concentrations to values below 10 mM required a maximum injection volume of 124 L for all models, however, physical experiments suggested that over twice that volume (≈280 L) was required to achieve these NH 4 + levels. When integrating outlet well concentrations over injected volumes, measurements suggested the removal of 27.2, 27.2, 17.6, 38.6, and 28.2 moles of NH 4 + occurred in Columns 1 through 5, respectively. This was equivalent to the removal of average pore fluid concentrations of 536 mM, 561 mM, 347 mM, 612 mM, and 560 mM NH 4 + for Columns 1 through 5, respectively. Average pore fluid concentrations removed from Columns 1, 2, 4, and 5 exceeded the maximum expected NH 4 + concentration of 500 mM and suggested that significant sorbed NH 4 + masses must have been removed during rinsing.  www.nature.com/scientificreports www.nature.com/scientificreports/ Retardation of NH 4 + transport relative to passive ions was also previously observed in centimeter-scale experiments 25 and likely contributed to delayed NH 4 + removal relative to passive tracer trends. Lastly, although NH 4 + removal was expected to be reduced in Column 4 for similar rinse injection volumes due to a larger porosity, surprisingly, when injection volumes exceeded ≈200 L, an opposite trend was observed with much lower NH 4 + concentrations exiting Column 4 compared to other columns. While unexpected, this was consistent with the greater NH 4 + removal calculated for this column. Figure 5 presents aqueous NH 4 + concentrations measured spatially along soil columns at various points in time during rinse injections and 24 hours after rinse injections following a 12 hour stop flow period. Pre-rinsing measurements reflected conditions following the ninth cementation treatment, wherein most locations in enriched columns had NH 4 + concentrations between ≈400 and 500 mM due to near full hydrolysis of the previously applied 250 mM urea injection. In Column 3, however, much lower NH 4 + concentrations between 100 mM and 324 mM were observed and were consistent with the lower ureolytic activity observed in the augmented column during cementation treatments (Fig. 3). As rinse injections proceeded in time, significantly lower NH 4 + concentrations were first observed at sampling locations closer to the injection well with large increases along column lengths. After injecting 45 L, almost no changes in NH 4 + concentrations were observed near the outlet well (distance of 3.58 m), however, NH 4 + concentrations were reduced to much lower values between 4 mM and 27 mM near the inlet well (distance of 0.15 m) in all columns. When comparing trends in time, spatial NH 4 + concentrations were similar between Columns 1, 2, and 5, however, Column 3 and 4 trends were significantly different. Despite lower pre-treatment NH 4 + concentrations in Column 3 prior to rinsing, trends in time were similar to other Concrete Sand columns when rinse volumes were less than ≈270 L. When additional rinse solution was www.nature.com/scientificreports www.nature.com/scientificreports/ applied to Column 3, however, continued reductions in NH 4 + were not observed and most locations had concentrations exceeding 1 mM. Reduced NH 4 + removal in Column 3 may have resulted from limited urea hydrolysis during the treatment period, the presence of more unoccupied cation exchange sites, and thus increased retardation of NH 4 + transport. Additionally, lower pH values observed in this column during rinsing, may have resulted in more ammonium existing in the charged form NH 4 + rather than NH 3 . In Column 4, greater NH 4 + removal was again observed for similar injection volumes, despite having a larger column pore volume. Higher exchangeable cation contents measured in Delta Sand may have prevented NH 4 + from interacting with soil surfaces during rinsing, thus improving NH 4 + removal efficiency. After injecting 525 L of rinse solutions, NH 4 + concentrations were below 15 mM, 6 mM, 19 mM, 0.4 mM, and 4 mM at all locations in Columns 1 through 5, respectively. Following the 12 hour stop flow period, however, NH 4 + concentrations generally increased and gradients in concentrations across columns became less pronounced likely due to equilibration of solutions with sorbed NH 4 + concentrations and diffusion. Figure 6 presents changes in NH 4 + concentrations during the stop flow period versus column length. Most locations had NH 4 + increases between 1 and 5 mM during the stop flow period. Again, Column 1, 2, and 5 showed similar trends, with Column 3 and 4 differing. In Column 3, much smaller increases in NH 4 + concentrations were observed with a single location achieving a 6.1 mM reduction during the retention period. Similarly, in Column 4, increases in concentrations were much lower than other columns and were generally near 0.3 mM. Smaller increases in NH 4 + concentrations in these columns during the stop flow period may reflect more limited desorption of NH 4 + resulting from less NH 4 + exposure during treatments (Column 3) and saturation of sorption sites and limited NH 4 + and soil interactions (Column 4). Figure 7 presents (a) removal of pre-rinsing NH 4 + concentrations (in percent) along column lengths and (b) cumulative NH 4 + removal (in percent) for all columns in time. As shown in Fig. 7a, all locations achieved greater than 95.7% NH 4 + removal following rinsing and the 12-hour residence period, with the exception of the most distal location in Column 3 (90.6% removal). At distances less than 2.82 m, greater removal was observed with all columns achieving greater than 98.0% NH 4 + removal. Column 4 achieved the highest NH 4 + removal of all columns with all locations achieving values above 99.2% removal. In order to better understand temporal changes in NH 4 + removal, spatial contours of NH 4 + at various times (Fig. 5) were integrated along column lengths to estimate cumulative NH 4 + removal in time. As shown in Fig. 7b, all columns started with 0% removal before rinsing, and achieved between 38% and 69% NH 4 + removal after injecting only 45 L (60 min). With increased rinse injection volumes, columns exhibited similar removal trends with the exception of Column 3, which achieved less removal in time. In all enriched columns, greater than 80% removal was achieved after injecting 90 L (120 min) and greater than 93% removal was achieved after injecting 180 L (240 min). Column 3, however, required near    Figure 8 presents NH 4 + concentrations measured in aqueous solutions obtained from sampling ports after the stop flow period as well as solutions extracted from moist soil samples obtained post-rinsing at various locations. As shown, values obtained from both methods resulted in similar magnitudes and trends in NH 4 + concentrations along columns suggesting that aqueous samples obtained from sampling ports were generally representative of internal soil conditions. NH 4 + concentrations from soil samples, however, were consistently slightly higher than sampling port values suggesting that greater concentrations of NH 4 + may have resided more locally around soil particle surfaces. Despite these differences, soil sample trends indicated post-rinsing NH 4 + concentrations between 3.7 mM and 17 mM in Column 1, 2, 3, and 5 with greater removal and values between 0.3 mM and 2.9 mM in Column 4. Figure 9a presents measurements of sorbed NH 4 + masses as determined from KCl extracted soil versus free solution NH 4 + measurements (Fig. 8). As shown, columns had measurable sorbed NH 4 + masses that varied between 4.8 × 10 −4 and 3.6 × 10 −3 millimoles of NH 4 + per gram of dry soil. In all columns, sorbed NH 4 + masses increased with distance from the injection well location, suggesting reduced NH 4 + removal from soil surfaces at larger distances. Figure 9b presents similar sorbed NH 4 + masses as a percentage of soil cation exchange capacities. As shown, all sorbed NH 4 + masses were between 1.9% and 16.1% of the total CEC of soil materials. Large differences were observed between soil types with Concrete Sand columns (Column 1, 2, 3) achieving average sorbed NH 4 + masses of 5.8% CEC, Covelo Sand (Column 5) achieving the highest average sorbed NH 4 + masses of 10.3% CEC, and Delta Sand (Column 4) achieving the lowest average sorbed NH 4 + masses of 3.3% CEC. Figure 9c presents measurements of sorbed NH 4 + masses with corresponding aqueous NH 4 + concentrations obtained from soil samples. As expected, for all soil materials, when aqueous NH 4 + concentrations were higher, sorbed NH 4 + masses also increased. The apparent linear relationship between sorbed NH 4 + masses and aqueous NH 4 + concentrations in equilibrium with soil surfaces suggested that NH 4 + sorption in columns may be reasonably described by a Freundlich adsorption isotherm over the concentrations observed 41 . Linear relationships between free NH 4 + concentrations and sorbed NH 4 + masses have been similarly observed in other studies examining ammonium sorption kinetics in soils 42,43 . Furthermore, this correlation suggests that sorbed NH 4 + concentrations were likely higher prior to rinsing and were reduced during the rinsing process. This is consistent with the higher removed NH 4 + concentrations estimated from outlet well measurements (Fig. 4). In order to better understand the amount of NH 4 + remaining on soil particle surfaces, "effective" aqueous NH 4 + concentrations were calculated assuming that all sorbed NH 4 + was instead available to surrounding aqueous solutions. Figure 9d presents Figure 10 presents contours of soil V s measured along columns lengths for all columns before cementation injections, after cementation but before rinsing, and after rinsing injections following the stop flow period. All columns had similar initial V s values, however, after cementation treatments, large differences in V s distributions were observed between columns and were reflective of differences in bio-cementation distributions. In Column 1, which had the highest ureolytic activity, high magnitudes of cementation were observed at distances less than 2.35 m with V s values ranging from 1107 m/s to 1522 m/s with a large reduction in V s to 382 m/s at a distance of 3.37 m. In Column 2, which contained the same sand but had a lower ureolytic activity, V s values ranged between 723 m/s and 1186 m/s at distances less than 2.35 m, however, a V s of 546 m/s was obtained at 3.37 m, which was significantly greater than Column 1. In Column 3, which was augmented, a V s of 1197 m/s was obtained near the injection well, however, little cementation was detected at distances greater than 0.31 m. In Column 4 and 5, which had lower ureolytic activity, similar trends were observed along column lengths as Column 2, with slightly lower V s values between 817 m/s and 871 m/s measured at a distance of 0.31 m and values between 227 m/s and 401 m/s measured at a distance of 3.37 m. For all columns, V s values before and after rinsing differed by no more than ± 50 m/s suggesting that little dissolution or precipitation had occurred. To further examine trends during rinsing, changes in V s values were plotted in time during rinse injections for all columns and bender element locations (Fig. 11). No consistent trends between columns or measurement locations were observed. Immediately after the start of injections, locations had changes in V s values between −46 m/s and +14 m/s, however, following the stop flow period changes between −33 m/s and +43 m/s were observed. These V s results suggest that rinse injections likely only resulted in minor changes in calcite contents between −0.2% and +0.3% by mass following previous relationships established by Gomez and DeJong (2017) 7 .

conclusions
The production of NH 4 + by-products during ureolytic bio-cementation presents a significant challenge for the technology, which must be addressed if MICP is to attain widespread acceptance as an environmentally-conscious ground improvement alternative. In this study, five 3.7-meter long soil columns containing three different sandy soils were treated using different bio-cementation treatment techniques to investigate NH 4 + by-product removal following ureolytic bio-cementation. During treatments, differences in enriched and augmented ureolytic activity were achieved and resulted in differences in bio-cementation distributions. While all enriched columns achieved near full hydrolysis of applied 250 mM urea injections over 24 hours (Column 1) and 48 hours (Column 2, 4, 5), the augmented Column 3 achieved limited urea hydrolysis and less NH 4 + production during cementation treatments. The localization of ureolytic activity and cementation observed near the injection well in Column 3 was attributed to the filtration of S. pasteurii bacterial cells during augmentation. Following cementation, a single 525 L volume of a high pH and high ionic strength rinse solution (200 mM CaCl 2 , pH ≈10.0) was applied to each column and NH 4 + removal and cementation integrity were monitored. NH 4 + concentrations observed at the outlet well were compared to expected trends for a passive tracer and results suggest that NH 4 + transport was retarded by soil-ion interactions with significantly greater NH 4 + removal than expected due to removal of sorbed NH 4 + from soil surfaces. When spatial changes in NH 4 + concentrations were examined during rinsing, large gradients in NH 4 + concentrations were observed across columns, however, NH 4 + concentrations below 19 mM were observed at all locations immediately after rinsing. After a 12-hour stop flow period, increases in NH 4 + concentrations between 1 and 5 mM at most locations were observed, with final cumulative NH 4 + removal between 97.9% and 99.8%, achieved for all columns.  www.nature.com/scientificreports www.nature.com/scientificreports/ Greater NH 4 + removal observed in Column 4 was attributed to higher concentrations of exchangeable cations present in the marine soil, which may have limited interactions between NH 4 + and soil minerals. In contrast, limited NH 4 + removal observed in Column 3 was believed to have resulted from both the presence of more unoccupied sorption sites from limited urea hydrolysis and lower pH values observed during rinsing. Following KCl extraction of post-rinsing soil samples, measurements suggested that significant NH 4 + remained sorbed to soil surfaces, which   www.nature.com/scientificreports www.nature.com/scientificreports/ may present challenges related to NH 4 + desorption over time. Finally, soil V s measurements suggested that NH 4 + removal had no significant effect on cementation integrity. While these results suggest that NH 4 + can be successfully removed from aqueous solutions residing in bio-cemented soils, acceptable limits for aqueous NH 4 + concentrations and sorbed NH 4 + masses will likely be governed by site-specific requirements and may require further investigation of rinsing and management techniques.

Data availability
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request. All measured data presented in the figures of this paper will be available through the NSF DesignSafe-CI Data Depot repository (https://www.designsafe-ci.org/data/browser/public/) under project number PRJ-2467.