This study was conducted using B. subtilis 98/7, a spore-forming strain isolated from dairy processing lines. Spores were produced following the method described by Faille et al. (2019) with slight modifications. Sporulation was carried out on Spo8 agar plates, incubated at 30 °C for 10 days, until the sporulation rate exceeded 95%. Spores were then collected by gentle scraping of the agar surface, washed five times by centrifugation in sterile distilled water, and finally suspended in sterile water and stored at 4 °C until use. Before each experiment, the suspensions were subjected to a 2.5 min ultrasonic treatment (Bransonic 2510E-MT, 42 kHz, 100 W, Branson Ultrasonics, United States) to minimize aggregation and ensure a homogeneous dispersion. The surface hydrophobicity of spores was evaluated using the Microbial Adhesion to Hydrocarbons (MATH) test, as previously described by Faille et al. (2019). Briefly, 3 mL of spore suspension were mixed with 500 µL of n-hexadecane (Sigma-Aldrich) and vortexed for periods ranging from 5 s to 30 min. The two phases were allowed to separate for 30 min before measuring the absorbance of the aqueous phase at 600 nm. The ratio of absorbance after mixing (A_t) to the initial absorbance (A₀) was plotted as a function of vortexing time.

Cleaning experiments were performed on AISI 316 stainless-steel coupons (45 × 15 mm, 2B finish) provided by APERAM (Isbergues, France). Before use, the coupons were cleaned using a standardized laboratory procedure. They were first rubbed with a pure alkaline detergent (RBS T105, Traitements Chimiques des Surfaces, France), then immersed for 10 min in a 5% (v/v) RBS T105 solution maintained at 60 °C. Coupons were rinsed thoroughly with tap water followed by deionized water for 5 min each, dried, and sterilized by dry heat at 180 °C for 1 h. This protocol ensured the removal of any organic or mineral contaminants and provided reproducible surface conditions. To reproduce the physicochemical properties typically found in industrial environments, an ageing treatment was applied to the stainless-steel surfaces prior to contamination. This treatment simulated repeated cycles of food contact and cleaning processes encountered in dairy production. Each ageing cycle consisted of four successive steps: immersion in a whole milk solution (150 g L^-1, INGREDIA) for 30 min at ambient temperature, rinsing under deionized water for 5 min, immersion in a 0.5% (w/w) NaOH solution at 60 °C for 30 min, and a final rinse in deionized water for 5 min. This four-step cycle was repeated fifteen times. Such ageing was found to significantly modify the surface energy and wetting behavior of the stainless steel, thus providing conditions representative of industrially used materials. Immediately before the experiments, all coupons underwent an additional cleaning step to remove any residual debris. Coupons were rubbed with RBS T105 detergent and immersed for 10 min in a 2% (v/v) solution heated to 60 °C, then rinsed with both hard and deionized water for 5 min each. They were then mounted vertically on a stainless-steel holder, separated by bolts to avoid contact, and sterilized at 180 °C for 1 h. For surface contamination, the spore suspensions were prepared in sterile ultrapure water at a final concentration of approximately 10⁶ CFU mL^-1. Sterile stainless-steel coupons were vertically immersed in 250 mL of the spore suspension contained in sterile glass beakers and maintained at room temperature for 4 h to allow adhesion of spores to the surface. After incubation, the coupons were carefully removed, air-dried under sterile conditions at ambient temperature, and directly used for the cleaning experiments.

The surface properties of the stainless-steel coupons were characterized prior to use to ensure reproducibility of the experiments. Surface topography was analyzed using a contact profilometer (Perthometer S2, Mahr, France), from which the arithmetic mean roughness (Ra) and the maximum height difference (Rz) were calculated based on the recorded surface profiles. The parameter Ra represents the average absolute deviation of the surface height from the mean line over a given sampling length, while Rz corresponds to the sum of the highest peak and the deepest valley measured within that length. In addition, the wettability of the surfaces was evaluated by measuring the static water contact angle using a DIGIDROP GBX drop-shape analyzer (France). For each material, ten measurements were performed using 1 μL droplets (five droplets on each of two independent coupons). Surfaces exhibiting contact angles greater than 90° were classified as hydrophobic, whereas those with lower values were considered hydrophilic.

Three surfactants with distinct interfacial properties were selected to evaluate how their molecular structure influences foam behavior and cleaning performance: sodium dodecyl sulfate (SDS), Lauramine oxide (Ammonyx® LO), and Capstone® FS-30. These surfactants represent different classes: anionic, amphoteric, and fluorinated nonionic, providing a comprehensive comparison of their physicochemical effects on foam formation and stability. Sodium dodecyl sulfate (SDS; ≥98.5% purity, Sigma-Aldrich) was used as the reference surfactant owing to its well-known capacity to produce stable foams and its extensive use in cleaning formulations. Lauramine oxide (Ammonyx® LO; Stepan Company, United States) is a tertiary amine oxide surfactant that exhibits nonionic behavior under neutral and alkaline conditions and cationic character under acidic environments. It is commonly used as a detergent, emulsifier, and foam stabilizer. Capstone® FS-30 (Chemours™, United States) is an ethoxylated fluorinated nonionic surfactant with high surface activity at low concentrations and strong stability in saline and hard-water media. Its chemical structure enables efficient reduction of surface tension and enhanced wetting, making it relevant for surface cleaning applications. All surfactant solutions were prepared in deionized water and mixed gently until complete dissolution. The critical micelle concentration (CMC) of each surfactant was determined experimentally using surface tension measurements carried out with a DIGIDROP GBX drop-shape analyzer (France) employing the pendant drop method. Measurements were conducted at room temperature (22 °C ± 1.5 °C) using a 1 mm stainless-steel needle. For each surfactant, solutions at various concentrations were analyzed, and the surface tension ( σ ) was recorded as a function of surfactant concentration. The CMC was defined as the concentration above which σ remained constant, indicating that additional surfactant molecules preferentially formed micelles in the bulk rather than adsorbing at the air–liquid interface.

Foams were generated and applied using a custom-built experimental rig. The open-circuit foam system consisted of a feeding tank, a foam generator, and a horizontal rectangular duct in which stainless-steel coupons were placed for cleaning tests. The setup allowed visual monitoring of the foam behavior through a transparent Plexiglas section identical in geometry to the test duct. To ensure fully developed flow conditions, the section to be cleaned was positioned more than 40 times the hydraulic diameter downstream of the foam inlet, corresponding to a distance of approximately 1 m. The main reservoir (100 L capacity) was filled with aqueous surfactant solutions prepared with deionized water (at their respective CMC values, as presented in the Results section). Three types of surfactants were tested: sodium dodecyl sulfate (SDS, anionic), Capstone® FS-30 (fluorinated nonionic), and Ammonyx® LO (amphoteric). Each solution was continuously recirculated using a positive displacement pump (VARMECA 21TL055, Leroy-Somer, France) that supplied the feed tank (50 L) positioned 3 m above the foam generator. This configuration created a constant liquid flow rate through the system by gravity, ensuring stable foam formation throughout the experiment.

Foam was produced by simultaneous injection of air and surfactant solution at a controlled flow rate of 4.5 L h^-1 each, yielding a foam quality β in the generator of approximately 50% β = Q G Q L + Q G , where Q_g and Q_l are respectively the gas and liquid flow rates). The mean foam velocity in the ducts was maintained at 1.8 cm s^-1 (Re ≈ 50), characteristic of laminar, plug-like foam flow. Although such low-Reynolds conditions limit the contribution of bulk hydrodynamic forces, local wall shear stress, generated by bubble-wall interactions and interfacial oscillations, remains a key determinant of cleaning efficiency. In closed ducts, foam performance therefore arises from the coupling between interfacial phenomena and the microscale shear events produced at the wall. Foam was allowed to stabilize for 1 m along the duct before reaching the test zone containing the coupons. Each cleaning experiment was performed at ambient temperature. The test ducts contained five stainless-steel coupons mounted horizontally, of which the three central ones were soiled with B. subtilis spores and subsequently analyzed for residual contamination. Cleaning trials were conducted under various foam exposure times (15 s–20 min). After each cleaning operation, the coupons were gently removed from the ducts and rinsed by immersion in sterile ultrapure water (1 L) to eliminate any non-adherent debris. To quantify residual spores, each coupon was transferred into a sterile container containing 10 mL of 2% (v/v) Tween 80 in peptone water (0.015 g L^-1, Biokar Diagnostics, France) and subjected to ultrasonication for 5 min in a Branson 2510 ultrasonic bath (40 Hz). This treatment, validated in a previous study (Saabi et al., 2020), effectively detaches over 99% of surface-adhered spores.

The resulting suspension was serially diluted in sterile ultrapure water, and 100 µL aliquots were spread onto nutrient agar plates prepared from 1.3% (w/v) nutrient broth (Bio-Rad, France) and 1.5% (w/v) bacteriological agar (Biokar Diagnostics, France). Plates were incubated at 30 °C for 48 h before colony counting. The number of viable spores recovered from each coupon after cleaning (N_resid) was compared to that obtained from control coupons before cleaning (N₀). Cleaning performance was expressed as the logarithmic percentage of residual spores log₁₀ ((N_resid/N₀)×100). A two-phase kinetics model was used to fit the data using GInaFIT (Geeraerd et al., 2005) through a biphasic model composed of two first-order kinetics (Dallagi et al., 2018; 2019). At least 3 repetitions were carried out for the quantitative analysis of the residual contamination after cleaning. Data were analysed using MATLAB R2024a (University of Caen, France), and Statistical analyses were performed by general linear model procedures using SAS V8.0 software (SAS Institute, Gary, NC, United States). Variance analysis and Tukey’s grouping (Alpha level = 0.05) were performed to determine how the bacteria removal, as described by the kinetic parameters (residual contamination at different cleaning times and model parameters), was affected by the tested cleaning conditions.

Foam stability tests were performed to compare the persistence and drainage behavior of foams generated from the three surfactant formulations. For each surfactant, foams were produced under identical operating conditions corresponding to a foam quality of 50% (air fraction in the foam generator). After a 1 m establishment length in the horizontal pipe, the foam underwent drainage and structural reorganization, leading to a local air fraction higher than that initially generated. This ensured that the foam reached a stable flowing regime with well-defined Plateau borders before sampling. Immediately after generation, the foam was collected in a 1000 mL graduated cylinder, and the time required for the complete disappearance of the visible foam and separation of the liquid phase was recorded at room temperature. This operational measure was used solely as a comparative indicator of foam persistence under identical conditions for SDS, Capstone® FS-30, and Ammonyx® LO. It reflects the combined effects of drainage, coalescence, and gas loss, without implying a specific stabilisation mechanism.

Table 2 summarizes the physicochemical properties of the surfactant solutions and the corresponding flow parameters of the foams generated under identical operating conditions. The capillary force acting on a bubble, Fc, corresponds to the interfacial force generated by surface tension at the bubble interface. It can be expressed as F c = σ 2 π r where σ is the surface tension and r the bubble radius (Al-Qararah et al., 2013). Here, the capillary force represents an order-of-magnitude estimate of the force transmitted at a single bubble-wall meniscus in confined foam flow, rather than the total force acting on a free bubble. These quantities characterize the resistance of a bubble to film thinning, drainage, and coalescence. The balance between viscous drag and surface tension forces acting on the foam during flow is described by the non-dimensional capillary number C a = µ l   v ¯ σ . Where μ _L is the dynamic viscosity of the liquid phase and v ¯ is the mean foam velocity (calculated as: v ¯ = Q l + Q g S . S is the cross-sectional area of the duct: 1.5 cm high × 1 cm width). Lower values of Ca indicate a regime dominated by surface tension, favoring stable lamellae and limited bubble deformation. The Reynolds number, characterizing the flow regime of the foam, is given by: R e = ρ f . v ¯ . d h μ f , where d _ℎ is the hydraulic diameter, ρ _f the foam density ( ρ f = 1 − β . ρ l + β . ρ g , and μ _f the effective foam viscosity ( μ f = μ l 1 − β 1 / 3 ) (Hatschek, 1911; Chovet, 2015). Solution viscosities μ _l were estimated from dilution calculations based on manufacturer data for concentrated products. However, at the highly dilute concentrations used (CMC levels), the surfactant contribution to viscosity is negligible, and all solutions exhibit viscosities equivalent to water ( μ l ≈ 1 ± 0.1 mPas at 22 °C). The mean wall shear stress τ ¯ was estimated from pressure drop measurements using τ ¯ = Δ P L d h 4 ; where Δ P is the pressure drop measured using a differential pressure sensor (Schlumberger type 8D, ±0.5% accuracy), d h is the hydraulic diameter, and L is the measurement distance.

The bubble size distribution within the foams was quantified using high-speed image acquisition with a Panasonic LUMIX DMC-FZ62 camera operating at 200 frames per second. Imaging was performed in a transparent Plexiglas observation section positioned perpendicular to the flow and located at the same streamwise position (1 m downstream of the foam generator, corresponding to approximately 40 hydraulic diameters) as the stainless-steel test section containing the coupons. This configuration ensured that the observed foam structure was representative of the foam in contact with the surfaces during cleaning. High-speed acquisition was required to limit motion blur associated with bubble translation and deformation at a mean foam velocity of 1.8 cm s^-1 and to enable reliable automated bubble contour detection. Images were acquired under steady-state flow conditions, and repeated acquisitions yielded consistent bubble size distributions, indicating a stationary foam microstructure at the measurement location. Bubble size analysis was performed within a fixed 15 × 10 mm region of interest located at the top wall, centred across the channel width. Only bubbles fully contained within this region were included in the analysis, while truncated or partially visible bubbles were excluded. For each experimental condition, three independent images were acquired under steady-state flow conditions. Images were calibrated using a micrometer scale to convert pixel dimensions into physical units (µm). Image processing and analysis were carried out using ImageJ and Piximètre software. Raw images were converted to grayscale, background subtraction was applied, and bubble contours were identified using manual thresholding followed by binary segmentation. When necessary, watershed segmentation was applied to separate overlapping bubbles. Equivalent bubble diameters were calculated from the detected contours. Bubble size distributions were fitted using a log-normal model by maximum likelihood estimation in MATLAB. Characteristic diameters, including the percentile diameters (D₁₀, D₅₀, D₉₀), the Sauter mean diameter (D₃₂), and the polydispersity index (D₉₀/D₁₀), were extracted. Measurement uncertainty was quantified using the standard deviation and coefficient of variation, as reported in Table 1, ensuring a robust and reproducible characterization of the bubble size distributions.

In this study, AISI 316 stainless-steel coupons were used as model surfaces to simulate those commonly encountered in dairy processing equipment. The surface characterization confirmed a smooth but microscopically heterogeneous texture, with an arithmetic mean roughness (Ra) of 0.21 ± 0.02 μm and a maximum profile height (Rz) of 1.52 ± 0.08 μm. These values reflect the presence of grain boundaries and minor polishing defects typical of industrial-grade steel. The static water contact angle measured on the aged surfaces was 49.9° ± 1.5°, indicating a moderately hydrophilic behavior consistent with stainless steel that has undergone multiple cleaning and exposure cycles.

The biological contaminant used in this work was B. subtilis 98/7, which produces hydrophilic spores surrounded by a flexible exosporium (Faille et al., 2014). This strain was selected as a model soil because Bacillus spores are highly resistant to heat and cleaning agents and are frequently encountered in dairy and other heat-treated products (Tu et al., 2021), making them a relevant worst-case scenario for assessing foam cleaning performance. The hydrophobicity of the spores was quantified using the MATH assay, revealing an aqueous affinity exceeding 90%, thus confirming their highly hydrophilic nature. This strong affinity for water suggests limited adhesion through hydrophobic interactions, making mechanical shear and capillary phenomena the dominant factors in their detachment from surfaces.

To generate the cleaning foams, three surfactant systems were employed: SDS, Ammonyx® LO, and Capstone® FS-30. Each exhibited distinct interfacial properties and critical micelle concentrations. The CMC values were experimentally determined as 2300 mg L^-1 for SDS, 11 μL L^-1 for Ammonyx® LO, and 100 μL L^-1 for Capstone® FS-30. Surface tension measurements performed at 23 °C ± 1.3 °C revealed equilibrium values of 26.2 mN m^-1 for SDS, 32 mN m^-1 for Ammonyx® LO, and 18 mN m^-1 for Capstone® FS-30 (Table 2). The SDS value is lower than typical literature values reported for highly pure SDS >99% (≈30–36 mN m^-1), likely due to the commercial-grade purity of the SDS used here (≥98.5%), which may contain impurities affecting surface tension. The measured values were retained because all experiments were performed using this formulation.

The stability of foams generated from the three surfactant systems was assessed by monitoring the time required for the complete disappearance of the visible foam under static conditions. The results (Figure 1) revealed pronounced differences in foam persistence depending on surfactant type. Foams produced with SDS remained stable for nearly 24 h before total collapse, whereas those formed with Ammonyx® LO and Capstone® FS-30 disintegrated after approximately 8 and 3 h, respectively.

Although the practical cleaning durations in our experiments were limited to 20 min, these long-term stability data provide essential insight into the intrinsic ability of each surfactant system to sustain its interfacial structure after flow conditions. Foam longevity reflects the competition between three principal destabilization mechanisms: gravitational drainage, bubble coalescence, and film rupture (Langevin, 2017; Denkov et al., 2020).

The foam microstructural characteristics revealed by representative micrographs (Figure 2) are quantitatively substantiated by the cumulative bubble size distributions fitted with log-normal models (Figure 3), providing comprehensive insight into the stability trends observed across surfactant systems. SDS foam (Figures 2a, 3) exhibited a homogeneous distribution of small, nearly spherical bubbles characteristic of a wet, plug foam regime. The quantitative analysis (Table 1) reveals exceptional size uniformity, with a log-normal distribution (R² = 0.984) displaying a median diameter (D₅₀) of 0.23 mm and remarkably narrow polydispersity (D₉₀/D₁₀ = 2.7). The low coefficient of variation (51.4%) indicates highly reproducible bubble formation during foam generation, reflecting optimal interfacial stabilization by SDS molecules. This fine, uniform structure ensures homogeneous capillary pressure distribution within the Plateau borders and minimizes localized film thinning, thereby significantly delaying film rupture through balanced disjoining pressure. In comparison, Ammonyx® LO foam (Figure 2b) presented a distinctly different bubble architecture, characterized by a bimodal distribution comprising both extremely fine bubbles and larger irregular structures. While achieving the smallest median diameter (D₅₀ = 0.1 mm), this surfactant exhibited broader size distribution (D₉₀/D₁₀ = 3.5) and higher variability (CV = 80.4%). The micrographs corroborate this statistical analysis, revealing increased bubble size heterogeneity, with a significant distribution tail extending beyond 0.4 mm. This morphological pattern suggests competing mechanisms of rapid bubble nucleation versus insufficient film stabilization, where initial bubble formation generates numerous fine bubbles, but inadequate interfacial cohesion permits subsequent coalescence events, leading to accelerated drainage rates and reduced dilatational modulus. Capstone® FS-30 foam (Figure 2c) displayed the most heterogeneous structure, characterized by large polyhedral bubbles and relatively thick parietal Plateau borders observable at the wall. The quantitative data robustly support these visual observations, demonstrating the widest size distribution (D₉₀/D₁₀ = 6.9) with a median diameter (D₅₀ = 1.64 mm) approximately seven-fold larger than Ammonyx. The substantially high Sauter mean diameter (D₃₂ = 2.3 mm) indicates low specific surface area, while the significant positive skewness (1.76) confirms the prevalence of oversized bubbles within the distribution. This morphological configuration facilitates rapid coalescence through diminished Laplace pressure gradients and enhanced gas diffusion between adjacent bubbles, consistent with the poorest foam stability observed experimentally.

The strong correlation between microscopic morphology and statistical size distribution parameters indicates that a small mean bubble size combined with low polydispersity is systematically associated with more persistent foams. SDS foams, therefore, combine a small characteristic diameter with a relatively narrow distribution, whereas Ammonyx® LO and Capstone® FS-30 show either larger bubbles or broader, positively skewed distributions. The log-normal fits are consistent with these observations and summarise the main structural differences that underpin the stability contrasts observed among the three surfactant systems. To quantify the actual foam state in the test section, the local air (void) fraction at the top of the duct was determined by image analysis of foam cross-sections using MATLAB. Binary segmentation of the images allowed discrimination between gas-filled regions and Plateau borders/lamellae, from which the areal gas fraction was computed as a proxy for local void fraction. The resulting values were 0.88 for SDS foam, 0.79 for Ammonyx foam, and 0.96 for Capstone foam, confirming that, despite the nominal quality of 0.5 imposed in the generator, the foam arriving in the measurement zone is already in a dry regime with densely packed, polyhedral bubbles. Following the framework of Forel et al. (2016), the areal gas fractions measured at the wall (0.88, 0.79, and 0.96) correspond to estimated bulk liquid fractions of approximately 0.14%, 0.45%, and 0.015% for SDS, Ammonyx LO, and Capstone FS-30, respectively. This confirms that all foams are in the very dry regime, with the surface liquid fraction being significantly larger than the bulk liquid fraction, as expected. The locally measured void fractions are fully consistent with the textures observed in Figure 2 and justify interpreting the flow as dry foam rather than a wet bubbly dispersion.

The foams generated from the three surfactants exhibited distinct physical and hydrodynamic properties under identical operating conditions (air fraction = 0.5, flow velocity v = 1.8 cm s^-1), as summarized in Table 2. The fundamental physicochemical properties of the surfactant solutions differed significantly, and these intrinsic interfacial properties governed the observed foam structures. SDS demonstrated the lowest surface tension (26.2 mN m^-1), leading to a low capillary number (Ca = 6.75 × 10⁻⁴). At such low Ca, surface tension forces dominate over viscous stresses, enabling strong Marangoni stabilization of lamellae (Breward, 2002; Vitasari et al., 2013) and promoting capillary-controlled bubble formation with narrow size distributions (Mangani et al., 2022; Ohashi et al., 2022). These combined effects favor the development of stable lamellae and a more homogeneous foam structure. In contrast, Ammonyx® LO exhibited higher surface tension (32 mN m^-1), whereas Capstone® FS-30 displayed the lowest surface tension (18 mN m^-1) but the highest Ca within the same order of magnitude. Despite these differences, the similarity of Ca across all samples indicates that viscous contributions remain negligible for the three systems.

The resulting foam properties followed consistent trends, with density increasing from SDS (499.6 kg m^-3) to Ammonyx® LO (544.1 kg m^-3) and Capstone® FS-30 (550.6 kg m^-3). The Reynolds numbers (47.5, 51.7, and 52.3 for SDS, Ammonyx® LO, and Capstone® FS-30, respectively) confirm similar flow regimes across all systems. Despite the comparable mean wall shear stress (∼2 Pa), it should be noted that, according to Dallagi et al. (2022b), conductimetry and polarography in similar foam-flow geometries showed that instantaneous local wall shear stress at the top wall can reach values up to approximately twice the mean value inferred from pressure drop. Such fluctuations play a crucial role in determining the cleaning behavior, particularly during the second phase of detachment. In the present study, this electrochemical method could not be applied systematically to all surfactants, as the ferri/ferrocyanide redox couple required for polarography destabilised SDS- and Capstone® FS-30-based foams, leading to rapid collapse.

The cleaning efficacy of foams generated from SDS, Ammonyx® LO, and Capstone® FS-30 surfactants was evaluated through detachment kinetics of B. subtilis spores from stainless-steel surfaces (Figure 4). All formulations demonstrated time-dependent log₁₀ reductions in viable spores, though with markedly different kinetic profiles. After 20 min of foam flow, SDS achieved the highest spore reduction (1.9 log₁₀), followed by Ammonyx® LO (0.83 log₁₀) and Capstone® FS-30 (0.55 log₁₀), establishing a clear cleaning efficiency hierarchy: SDS > Ammonyx® LO > Capstone® FS-30.

Kinetic analysis revealed a biphasic detachment mechanism characterized by two distinct removal rates (Table 3). The initial rapid phase (K₁) corresponds to detachment of weakly bound spores, while the subsequent slow phase (K₂) represents removal of strongly adherent spores. SDS exhibited a substantially higher K₁ value (114.75 s^-1) compared to Ammonyx® LO (12.61 s^-1) and Capstone® FS-30 (0.25 s^-1). The fraction of spores removed during this rapid phase (f) was greatest for SDS (98.2%), though statistical analysis indicated no significant differences among surfactants for this parameter. The secondary detachment rates (K₂) were consistently low across all surfactants (<0.05 s^-1), confirming that firmly anchored spores remained largely unaffected within the 20-min cleaning window. Statistical grouping revealed that while K₁ values for SDS and Ammonyx® LO were significantly different, K₂ values showed no significant variation among the three surfactant systems. These results demonstrate that surfactant selection significantly influences the initial spore detachment rate, with SDS foam achieving substantially faster removal kinetics during the critical first phase of cleaning. The consistent K₂ values across all systems suggest that the secondary removal mechanism is largely independent of surfactant chemistry within the experimental timeframe.