Environ. Sci. Technol. 2005, 39, 7853-7859
Retention and Transport of Amphiphilic Colloids under Unsaturated Flow Conditions: Effect of Particle Size and Surface Property JIE ZHUANG,† JUN QI, AND YAN JIN* Department of Plant and Soil Sciences, University of Delaware, Newark, Delaware 19716
The purpose of this study was to examine the mechanisms responsible for deposition and transport of amphiphilic colloids with a wide range of particle sizes (20-420 nm) through porous media. A series of saturated and unsaturated column experiments were conducted using amphiphilic latex microspheres and a hydrophilic silica colloid. We found that the amphiphilic latex particles were retained to a greater extent than the hydrophilic silica colloid in unsaturated media. This was attributed to colloidal attachment at the airwater interface due mainly to hydrophobic interactions. We also found that dependence of colloid retention on particle size was nonlinear. There existed a fraction of colloids with greater mobility than other fractions, which we referred to as the most mobile colloids. As particle size increased from 20 to 420 nm, colloid deposition rate first decreased to reach a minimum value at ∼100 nm then increased, indicating that different retention mechanisms were involved. We showed that conducting saturated transport experiments and analysis using filtration theory may be an effective approach for determining the most mobile colloid size(s) in porous media, perhaps even for unsaturated flow conditions. This study highlights the importance of including size effect and surface properties when predicting concentrations and fluxes of amphiphilic colloids or colloid-bound amphiphilic and hydrophobic contaminants in the subsurface environment.
Introduction Transport of colloidal particles in porous media is of primary importance for understanding and predicting behaviors of colloid-bound contaminants in subsurface environments (14). In addition, colloids, such as pathogenic microorganisms (viruses, bacteria, and protozoa), are contaminants themselves (5). This is why numerous studies have been conducted to investigate the factors and mechanisms governing colloid deposition and transport in porous media. As documented, three types of factors, including aqueous phase composition, properties of colloids and media, as well as physical and chemical conditions of the flow generally influence colloid transport (3, 5). Among them, colloid size and property, * Corresponding author phone: 302-831-6962; fax: 302-831-0605; e-mail:
[email protected]. † Current address: Center for Environmental Biotechnology, Department of Earth and Planetary Sciences, The University of Tennessee, Knoxville, TN 37996-1410. 10.1021/es050265j CCC: $30.25 Published on Web 09/09/2005
2005 American Chemical Society
polydispersity, and partial water saturation are the most relevant to the natural subsurface environments (2, 3, 5) Particle size can significantly influence colloid deposition (6-12) and the size distribution of the deposited particles on matrix surfaces changes with time and flow conditions (11, 13, 14). Size effect on colloidal transport has been demonstrated by experimental observations (7) and theoretical calculations based on filtration theory (9). Theoretical predictions indicate that, in the presence of repulsive electric double layer interaction, colloid deposition or coagulation rates depend on particle size. This is because colloid size can affect the combined electric double layer and van der Waals interactions, which consequently change the height of interfacial energy barrier. Straining, the trapping of colloid particles in pore throats that are too small to allow particle passage, also strongly depends on particle size (15-18). Harvey et al. (16) reported increased colloid retention with increasing colloid size. Bradford et al. (18) observed that colloid breakthrough concentration and the final spatial distribution of colloids retained by porous media was highly dependent on colloid size and soil grain size distribution, and a conceptual model was developed to account for colloid straining in addition to attachment and exclusion (19). The particle size effect may also depend on the hydrophilic and hydrophobic attributes of colloid surfaces. When hydrophilic colloids attach to matrix surfaces, electrostatic interactions dominate the interfacial processes. Hydrophobic interaction becomes a nonnegligible mechanism for the deposition of amphiphilic colloids (20, 21). In partially saturated media, presence of an air-phase may further increase the effect of hydrophobic interaction at the airwater interface (22-24). However, most of the previous studies on size effect were performed with hydrophilic colloids (6-8), although colloid and colloidal contaminants with amphiphilic surfaces properties are present extensively in natural environments. Those include microorganisms (e.g., bacteria and viruses), manufactured nanoparticles, proteins, and natural occurring mineral colloids with organic matter coatings. It is important to point out here that although several studies have reported the retention and transport of “hydrophobic” colloids, the colloids used in these studies were in fact amphiphilic colloids with large contact angles as well as positive or negative surface charges (20-22). This is understandable because purely hydrophobic particles tend to aggregate with each other and do not form stable suspensions, therefore, they are stabilized either by attachment of charged functional groups to the particles or addition of surfactant into the suspension. The objectives of this study were to (1) compare transport behaviors of amphiphilic latex colloids with a hydrophilic silica colloid through a quartz sand medium, and (2) examine size-dependent transport of the amphiphilic latex colloids under both saturated and unsaturated flow conditions. The results revealed that the transport of amphiphilic latex colloids of different sizes depended on their particle size, and their retention followed different mechanisms. Furthermore, hydrophobic interactions increased the effect of water content on colloid retention under unsaturated flow conditions.
Materials and Methods Porous Medium and Treatment. The porous material selected for the column breakthrough experiments was quartz sand with sizes ranging from 300 to 355 µm. The sand was sieved from Accusand 40/60 (Unimin Corporation, Le Sueur, MN) with stainless steel mesh, and chemically treated to VOL. 39, NO. 20, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
7853
remove metal oxides from the grains. To remove the metal oxides, the sand was first washed in tap water until the rinse water was free of suspended impurities and then reacted with 6 M HNO3 for 5 h at 80 °C. After that, the acid was decanted, and the sand was thoroughly rinsed with deionized water. To guarantee that the sand was free of metal oxides, the sand was further treated following a procedure modified from the method of Mehra and Jackson (25), which is described briefly here. A 500-mL portion of 0.2 M citrate buffer solution that contained 44.1 g L-1 sodium citrate (Na3C6H5O7‚2H2O) and 10.5 g L-1 citric acid (C6H8O7) was added to a 2-L polyethylene bottle containing 300 g of the water-washed sand. The bottle was put in a shaking water bath (120 rpm) maintained at a temperature of 80 °C where 15 g of sodium dithionite (Na2S2O4) was added to the bottle. The bottle was removed from the shaker and hand-shaken thoroughly for several seconds, then returned to the shaker and vigorously shaken for 1 min. Next, the speed was reduced and the sample was digested for 15 min with occasional vigorous shaking. The solution was decanted from the sand and the procedure was repeated three additional times. Finally, the sand was rinsed with deionized water until the electric conductivity of the water reached near zero. Drying curve of the water characteristic function was measured for the sand with a Tempe pressure cell (Soil Moisture Equipment Inc., Santa Barbara, CA) (26) at bulk density of 1.75 g/cm3, which was similar to that of the sand columns packed for the transport experiments. The measured water retention curve was essentially the same as reported by Saiers and Lenhart (27) for the same sand. The air-bubbling pressure was -16 cm water, and the pore length was 173 ( 15.5 µm as calculated by Saiers and Lenhart (27). Colloidal Particles and Characterization. Two types of model colloids were used in the breakthrough experiments. One was a monodispersed hydrophobic chloromethyl latex microsphere, which is stabilized by negatively charged sulfate groups (Interfacial Dynamics Corporation, Portland, OR). Containing both polar and nonpolar groups on the particle surface resulted in a contact angle of ∼90°. The latex microsphere is thus referred to as an amphiphilic colloid in this study. Latex particles of various diameters were selected: 20, 100, 350, and 420 nm, all with the same particle density of 1.06 g/cm3. The other experimental colloid was hydrophilic silica particles (Nissan Chemical Industrial, Ltd.) with a particle density of 2.19 g/cm3 and a mean particle diameter of 310 nm. Colloid concentrations were measured by UV-Vis spectrophotometry (DU Series 640, Beckman Instruments, Inc., Fullerton, CA) at appropriate wavelengths, which were determined specifically for each colloid size and type. The amphiphilic latex colloids with sizes of 20 and 100 nm were analyzed at a wavelength of 250 nm, while 350 and 420 nm latex colloids and 310 nm silica colloids were analyzed at 350 nm wavelength. Calibration curves obtained in the same background solution as used in the column experiments were used to convert UV absorbance to colloid concentrations. Column Breakthrough Experiments. Column experiments were performed in duplicate to examine the effects of particle size and hydrophobic interactions on the transport of the selected colloids under saturated and unsaturated flow conditions. A summary of the experimental conditions maintained for the columns is provided in Table 1. The saturated and unsaturated column setups used in this study were similar to those illustrated in Jin et al. (28). The acrylic column had an inside diameter of 4.5 cm and was 10.0 cm long. The column consisted of a top and a bottom plate and was sealed by an O-ring on each end. The O-ring had no contact with the inside of the column. Except for the pump tubing that was made of Tygon, tubings throughout the system were made of Teflon. For the unsaturated experi7854
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 20, 2005
TABLE 1. Parameters of the Breakthrough Experiments
exp.
colloid
1 2 3 4 5 6 7 8 9 10 11 12
latex 20 nm latex 20 nm latex 100 nm latex 100 nm latex 350 nm latex 350 nm latex 350 nm latex 350 nm latex 420 nm latex 420 nm silica 310 nm silica 310 nm
bulk saturation density velocity dispersion % Mg/m3 porosity cm/h cm2/h 100 28.2 100 28.6 100 81.8 55.0 28.9 100 28.8 100 28.8
1.68 1.68 1.65 1.71 1.68 1.68 1.68 1.68 1.66 1.74 1.71 1.69
0.378 0.377 0.390 0.367 0.378 0.378 0.378 0.378 0.385 0.355 0.367 0.372
7.40 15.80 7.29 16.39 6.53 7.37 8.51 13.80 7.28 17.17 7.40 15.31
1.70 2.21 1.53 4.59 1.05 2.80 1.36 2.48 1.97 5.32 2.51 3.98
ments, a nylon membrane with 20-µm pores and 65-cmwater air-entry value (Spectra/Mesh, Spectrum Laboratories, Inc.) was placed on the bottom plate for maintaining capillary pressure within the columns. For the saturated experiments, the same nylon membrane was also used to be consistent with the unsaturated systems. When packing a column, a deaerated background solution (0.84 mM NaCl + 0.16 mM NaHCO3, pH 7.5) was preintroduced into the column from its bottom to a certain height, and then the dry sand was slowly poured into the solution as 1-cm increments while being stirred with a plastic rod to ensure uniformity of the packing and to avoid air entrapment in the column. For each experiment, approximately 15 pore volumes of the background solution (particle-free) was first introduced to the column upward from the bottom using a peristaltic pump to (1) leach tiny impurities from the matrix; (2) allow a chemical equilibrium of the packed column with the background solution; and (3) establish a steady state flow condition. In the saturated experiments, after the column system was equilibrated, application of the input was switched to the colloidal suspension, which was composed of the experimental colloid (100 mg/L) and tracer bromide (30 mg/L NaBr), and later switched back to the background solution to flush the column. In the unsaturated experiments, after the preequilibrium with background solution was reached, the input from the bottom of the column was stopped and meanwhile injection of the same colloid-free solution was initiated from the open top of the column via a sprinkling system. The sprinkling system consisted of an injection chamber (4.5-cm i.d. and 5-cm height) with thirteen 26-gauge hypodermic needles arranged uniformly across the bottom of the chamber. During the injection of liquid from the top, effluent was induced by vacuum through a vacuum chamber connected with the bottom of the packed column. Water content in each column was controlled by adjusting the pressure in the vacuum chamber and the flow rate of the influent. Once the water content stabilized at the desired value, colloid suspension was injected through another set of the sprinkling system until the colloid concentration in the effluent stabilized at a certain relative concentration (C/ C0). Subsequently, the particle-free background solution was reapplied until the colloid concentration in the effluent returned to the initial baseline level. During the course of the experiments, samples of the effluent were collected from the top (saturated flow) or bottom (unsaturated flow) of the columns into 18-mL glass test tubes with a fraction collector at defined time intervals. Results of preliminary experiments demonstrated that removal of the latex and silica colloids by the empty column system (filled with background solution only without sand) and nylon membrane was negligible. The concentration of bromide tracer in the effluent was determined by ion chromatography (IC) using an IonPac
FIGURE 1. Bromide breakthrough curves from all column experiments.
FIGURE 2. Breakthrough curves of 350-nm amphiphilic latex colloids at different relative water saturation. Inset compares the breakthrough between bromide and the colloid under saturated flow condition for the first 4 pore volumes. AS9-HC 4-mm anion-exchange column (Dionex Corporation, Sunnyvale, CA) with 9-mM sodium carbonate as elute and at a flow rate of 1 mL min-1, and the data were used to test the hydrodynamic performance of the columns and to estimate dispersion coefficients (D). All experiments were conducted at room temperature (22(1 °C).
Results and Discussion Transport of Bromide Tracer. Breakthrough curves of bromide for all the experiments are plotted in Figure 1. Stability and consistency of the tracer breakthrough suggest that the flow conditions were controlled very well for the experiments, and the effect of velocity on ion dispersivity was minimal. The dispersion coefficients (D) calculated based on the traditional convection-dispersion equation for a nonreactive solute are presented in Table 1. Effect of Relative Water Saturation. Steady-state breakthrough concentrations of the 350-nm amphiphilic latex colloid from the quartz-sand-packed columns decreased with decreasing water saturation (Figure 2). This coincides with previous observations on the transport of silica colloid, hydrophilic latex microspheres, and viruses through waterunsaturated sand columns (22, 23, 29). In the literature, three mechanisms have been proposed to account for increased
colloid retention in unsaturated media, including attachment at solid-liquid interface, film straining, and air-water interfacial capture (22, 28-31). Colloid attachment at solidliquid interface can be examined by saturated transport experiments. As illustrated by the inset of Figure 2, breakthrough of the colloid under saturated flow conditions was slightly earlier than that of bromide, indicating that the colloid initially experienced minor preferential transport, probably due to size exclusion, but reached the same steady-state C/C0 as bromide, indicating that solid-liquid interfacial attachment was not an important mechanism governing the retention of the 350-nm amphiphilic latex colloid. The observed unsaturated transport behavior suggests that the presence of an air phase played a major role in immobilizing the colloid. As water saturation decreased, retention of the latex particles increased accordingly. In an unsaturated system, the air phase can change the geometrical pattern of the flow path and the interface dimensions (29, 32, 33). As a result, partial saturation facilitated colloid retention through limiting colloid-accessible pathways hence attachment sites and decreasing thickness of water film on sand particles, as well as providing the area and continuity of the air-water interface. Identification and distinction of these processes have been well documented through modeling and experiments in recent work by Saiers and Lenhart (27) and Lenhart and Saiers (29). Their studies indicated complex interactions between water content and pore water chemistry and the combined effect on colloid mobility in the vadose zone and showed that the dominant colloid retention mechanism transitions from straining, to air-water interface capture, and to mineral-grain attachment as the solution ionic strength increases. Effect of Colloid Surface Property. To examine the effect of surface property on colloid transport, breakthrough behavior of the 350-nm amphiphilic latex colloid is compared with that of the 310-nm hydrophilic silica colloid in Figure 3 under saturated and unsaturated flow conditions. In the water-saturated columns, both the amphiphilic latex particles and the hydrophilic silica behaved similarly to the bromide tracer (Figure 3a). The overall similarity of the complete breakthrough observed for both colloids implies that hydrophobic interactions played a very limited role in colloid retention under saturated flow conditions. The strong electrostatic repulsion between the sand surface and the particles dominated colloid-matrix interaction. Unlike saturated transport, breakthrough of both colloids decreased significantly under unsaturated flow conditions (Figure 3b). Presence of an air phase enhanced the retention of both latex and silica particles, but the effect was more pronounced on the amphiphilic latex colloids. This suggests that retention at the air-water interface was probably the more important mechanism for immobilizing the colloids as compared to film-straining because much less of the silica colloids, which were similar in size to the latex particles, were removed at the same water content. Both electrostatic and hydrophobic interactions have been reported to contribute to colloidal attachment at the air-water interface (22, 23, 34, 35). In addition to possible electrostatic interaction, hydrophobic interaction between the amphiphilic particles and the air-water interface resulted in greater colloid retention at the interface than the silica particles. Hydrophobic interaction is a solvent-induced force that consists of a part of what is commonly referred to as the potential of an average force (36). On the basis of their breakthrough experiments with two bacteriophages (MS-2 and φX174) of different hydrophobicity from Al2O3-coated sand columns, Zhuang and Jin (21) suggested that electrostatic interaction and hydrophobic interaction had a coupled effect on the attachment of the amphiphilic virus particles on Al oxide surface. Strong electrostatic attraction can enhance the VOL. 39, NO. 20, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
7855
FIGURE 3. Comparison of transport of the silica colloid (310 nm) with the amphiphilic colloid (350 nm) under (a) saturated and (b) unsaturated flow conditions. hydrophobic effect, and vice versa. Therefore, in addition to colloid hydrophobicity, the extent of hydrophobic effect is also determined by other factors, such as the availability of air-water interfacial area, collector surface properties, and solution chemistry. The importance and detailed mechanisms of hydrophobic interaction remain to be further investigated. Effect of Particle Size. Transport of the amphiphilic latex colloid was strongly particle-size-dependent under both saturated and unsaturated flow conditions (Figure 4). In the saturated experiments (Figure 4a), both the smallest (20 nm) and the largest (420 nm) colloids exhibited significant immobilization, clearly contrasting with the complete breakthrough of the intermediately sized colloids (100 and 350 nm). A more detailed analysis of size-dependent colloid retention using filtration theory is provided in the next section. In the unsaturated experiments (Figure 4b), a complete breakthrough was observed for the 100-nm colloid after ∼9 pore volumes, which was much more delayed than bromide breakthrough. Retention of the other colloids (20, 350, and 420 nm) was consistently greater under unsaturated than saturated flow conditions. These results clearly show that the size-dependent retention of colloids in porous media was sensitive to the change of water saturation. This is not surprising because change in water saturation not only impacts the number of colloid-accessible pathways (watersaturated pores as well as corner-water ducts with high degree of connectivity and dimensions excessive of colloids), but also triggers additional interaction mechanisms, such as filmstraining and air-water interfacial capture. Nonuniform flow velocity employed in the individual experiments also contributed to the observed colloid transport behavior. As shown in Table 1, the approaching velocities were different for the unsaturated and saturated experiments. 7856
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 20, 2005
FIGURE 4. Effect of particle size on transport of the amphiphilic colloids. This means that the effect of hydrodynamic interaction may be different for the various sized particles (37). However, high velocity has always been found to facilitate colloid transport (38-40). It is thus expected that more colloids would have been retained in the water-unsaturated columns if the experiments were carried out at the lower flow velocities used for the water-saturated column experiments. Moreover, worthy noting in Figure 4 is that the shape of the breakthrough curve of the 20-nm colloid is different from those of the larger colloids. The effluent concentrations of the 20-nm particles increased to a constant level at 1.4 and 2.5 pore volumes in the saturated and unsaturated experiments, respectively, and remained for the duration of the experiments. This behavior indicates that the removal of the 20-nm colloid was controlled by a first-order kinetic mechanism in both saturated and unsaturated systems. In comparison, arrival of stable breakthrough concentrations of the 100- and 350-nm latex colloids were delayed, with much larger retardation in the unsaturated than the saturated columns. However, the effluent concentrations of 420-nm colloid never reached a steady-state level in both columns but instead showed a trend continuously climbing up toward the unity C/C0 with a decreasing slope. This characteristic reflects a decline in deposition rates with increasing particle load on the solid phase, governed by a second-order removal mechanism. These results suggest that the mechanisms governing colloid removal in porous media varied with particle size and such effect could considerably complicate the deposition processes of polydispersive colloids that occur in most natural environments. Existence of the Most Mobile Colloid Size(s). To quantitatively assess the relationship between colloid attachment and colloid size, we calculated the deposition rate (Kd) of colloid using the formula Kd ) (v/L)ln(C0/Ce), where v is the pore velocity (m/h), L is the column length (m), C0 is the influent colloid concentration (g/m3), and Ce is the stable
FIGURE 5. Variation of colloid deposition rate with particle size under saturated and unsaturated flow conditions. effluent concentration of colloid (g/m3) (3, 41). Because a completely stable effluent concentration for the 420-nm colloid was not observed within duration of the experiment, outflow concentration at the end of the colloid pulse was used as an approximation of Ce in the calculation. The possible underestimation of the Kd values from this approximation is minimal because the effluent concentrations remained relatively stable for the last 3 and 5 pore volumes in the saturated and unsaturated experiments, respectively. Furthermore, a sensitivity analysis indicated that increase in the effluent concentration by 1 mg/L only decreases the value of Kd by 10 and 20% for the saturated and unsaturated experiments, respectively, which does not change the overall
trend presented in Figure 5. Results in Figure 5 show that the smallest particle size with minimum deposition rate was ∼100 nm for the amphiphilic latex microspheres in a wide range of relative water saturation, ranging from 29% to 100%. As degree of water-saturation rose, larger colloids joined this group of the most mobile colloids, and the largest size of the most transportable colloids through the medium was estimated to be between 350 and 420 nm. We refer to this size or size range corresponding to the minimum deposition rate as the most mobile colloid size(s). It must be pointed out that the equation used for calculating Kd is only applicable to saturated porous media, the parameter calculated for the unsaturated experiments is a “lumped” value, which includes the effects of colloid retention by air-water interface and film-straining. Small changes in water saturation may cause significant changes in the estimated value so it allows only qualitatively comparisons between experiments. Our results are supported by the work of Huber et al. (9). They found that the size distribution of the most mobile colloids ranged from 100 to 1000 nm in the coarse sand used in their study, and the distribution became narrower as the collision efficiency increased or when background solution changed from water to 1 mM NaCl. To further examine the size-dependent transport behavior, we used filtration theory to analyze the relationship of particle size with respect to different transport mechanisms: diffusion, interception, and gravitational sedimentation, using results from the saturated experiments. As described by Tufenkji and Elimelech (42), the overall single-collector contact efficiency (η) is positively proportional to the magnitude of the colloid deposition rate (Kd), and equals the sum of the single-collector contact efficiency by diffusion (ηD), by interception (ηI), and by gravitational sedimentation
FIGURE 6. Relationship between particle size and individual single-collector contact efficiency (η) as well as attachment efficiency (r) calculated according to refs 41-43 using the saturated column data. Parameter values used in the calculation: Hamaker constant A ) 1 × 10-20 J; Boltzmann constant k ) 1.38 × 10-23 J/K; fluid absolute temperature T ) 295 K; fluid viscosity µ ) 1.005 × 10-3 kg/m/s; fluid density Gr ) 1000 kg/m3. VOL. 39, NO. 20, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
7857
(ηG). Individual contributions of each mechanism to total colloid deposition on matrix can thus be evaluated by calculating ηD, ηI, and ηG. The calculated results according to the method of Tufenkji and Elimelech (42) are plotted in Figure 6. The particles smaller than 100 nm encounter the matrix surface much more frequently than the larger particles (Figure 6a), which are considerably more subjected to the effects of interception and gravity. This analysis seems to indicate that while filtration by diffusion dominated removal of all the colloids examined, removal by interception and gravitation was completely negligible for particles smaller than the most mobile colloids (e.g., 100 nm) (Figure 6b and c). In addition to the single-collector contact efficiency (η), attachment efficiency (R) can also impact the deposition rate (Kd). The relationships between them are Kd ) 3vRη(1 - φ)/2d, and R ) 2dln(Co/Ce)/3Lη(1 - φ), where d is the diameter of the medium grain, and φ is porosity of the medium (41, 43). Figure 6d shows that both the smaller and larger particles had higher attachment rates than the intermediately sized particles (i.e., the particles having sizes similar to the most mobile colloids). Another possible mechanism responsible for the sizedependent transport may be particle-size-induced change of energy barrier at interfaces, since particle deposition rates are exponentially related to the height of the energy barrier (44). As predicted by the DLVO theory, height of the energy barrier is positively proportional to the size of particles approaching the interfaces (6, 7). This mechanism is likely very important for the attachment of colloids smaller than the most mobile colloid size, because particle-surface collision is a prerequisite. The amphiphilic latex colloids and the sand medium used in this study were both overall negatively charged at neutral pH, and thus repulsive electric double layer interaction was present between the particle and the sand surface. According to the DLVO theory, the repulsive energy barrier is larger for the 100-nm than for the 20-nm colloids. This may be why particle deposition rate of the 100-nm colloid was small, while the 20-nm colloid exhibited a higher retention. This result provides insights on the retention and transport characteristics of nanosized particles (normally
