The first documented examples of using foam generation to remove organic compounds from water can be traced to Ostwald, and independently, Schutz in 1937 (Ostwald, 1937; Schutz, 1937). Thiel claims that water purification via skimming was introduced into the aquarium hobby in the 1960's by Huckstedt (Thiel, 1997; Huckstedt, 1972), but the practice did not gain much traction until a resurgence of interest in keeping corals brought it to the fore again in the 1990's. Another early notable advance in using foam flotation technology for saltwater purification was described by Wallace (Wallace, 1969). The early developments in water purification then led to advances in two disparate venues; wastewater remediation, and protein purification (Lemlich, 1972; Okamoto, 1979; Clark, 1983; Caballero, 1990). The application of skimming in aquarium husbandry was an outgrowth of successful implementation of foam fractionation techniques in these areas, and the development of modern skimmers owes much to these pioneering efforts. Foam fractionation in particular proved to be a valuable asset in enabling the isolation/recovery of desirable proteins from dilute solutions in many areas of food and pharmaceutical science. In this context, the goal was just the opposite of protein skimming in aquaria; recovery of valuable proteins in the foam with discharge of the depleted water phase. In contrast, of course, protein skimming in aquaria is used to remove undesirable organics from the (valuable) tank water. Nevertheless, the processes are identical, a conceptual convergence that becomes important in assessing the influence of various input parameters on skimmer performance. Specifically, the pivotal role of foam fractionation-based purification in protein recovery has prompted many research groups to conduct studies designed to optimize protein purification by tweaking input variables. It is possible that these studies can inform the aquarium skimming area as well. Much effort has been directed to measuring how changes in (a) gas flow rate, (b) liquid flow rate, and (c) bubble size influence two important figures-of-merit in the protein purification (and by inference, aquarium skimming) area; enrichment (E) and recovery (R). Enrichment (E) is defined slightly differently by different authors. Some authors describe E as the ratio of the protein concentration in the (collapsed and removed) foam head relative to the protein concentration in the skimmer feed solution (E = Crecov/Cin in Fig. 1) (Uraizee, 1996; Brown, 1990), whereas other authors define this quantity as the ratio of protein concentration in the foam head compared to the protein concentration in the output solution of the skimmer (E = Crecov/Cout in Fig. 1) (Ahmed, 1975; Schnepf, 1959). The numbers obtained by these two definitions do not differ greatly, and so this distinction is not critical. A second figure-of-merit often cited in these skimmer performance studies is recovery (R), which is defined as the amount of protein removed from the solution by the skimmer relative to the amount of protein fed into the skimmer. The recovery R can be expressed as a percent of protein removed after a specified time: i.e., 50% of the protein has been recovered after 90 minutes. These two measurable quantities typically run in opposite directions; that is, those changes that increase the enrichment typically decrease the recovery, and vice versa.
Both enrichment and recovery have counterparts in the aquarium skimming area. Dry skimming implies very little water hold-up in the foam, and this scenario is more closely aligned with enrichment. Thus, maximizing foam enrichment while dry skimming should maximize impurity removal from aquarium water. In contrast, wet skimming, with its proportionally larger liquid hold-up in the foam, falls more under the aegis of the "recovery" manifold of skimmer operation. That is, the removal of organic-rich foam and entrained aquarium water that contains organics (= wet skimming) should lead to a greater overall removal of the organic impurities in the aquarium water. In this case, maximizing recovery R should lead to maximizing water purification. To the extent that an aquarist aligns their skimming technique with one or the other extreme, then the lessons learned about optimizing either enrichment or recovery might prove insightful. Wet skimming bears the added burden of introducing possible salinity fluctuations, as the aquarium water removed in the foam phase must be replaced by water of equivalent salinity in order to maintain the overall tank's salinity. To the extent that this match is not maintained, the tank's overall salinity may vary. Thus, a compromise between wet and dry skimming often is sought. In the final analysis, any operational parameter changes that increase the amount of organic material removed from the aquarium water is equivalent to achieving better water purification. Protein purification studies typically focus on enrichment E, as the purity of the recovered protein, which may find use in food and/or biomedical applications, often is paramount. Thus, biomedical and food science researchers are willing to sacrifice removal capacity (higher R) for product purity (higher E). Nevertheless, the reciprocal relationship between the two skimmer benchmarks suggests that lessons learned in the protein purification area can provide insight into aquarium skimming as well. Typical results from the protein purification area serve to illustrate the connection, Figs. 2, 3, and 4. In these studies, a model protein, Bovine Serum Albumin (BSA) in high ionic strength water (~ 5.8 ppt of NaCl; compare saltwater ~ 35 ppt of all salts) is circulated through a model skimmer with concomitant introduction of air through a bubble-making porous frit. The collected and collapsed foam is removed and assayed for protein concentration, as are the skimmer's input and output streams. Fig. 2 illustrates how E and R run in opposition when gas flow rate is the experimental variable. Thus, higher gas flow rates decrease E but increase R, at least over the flow ranges tested. A comprehensive and quantitative explanation for these trends (E or R vs. gas flow rate) is lacking, but a qualitative rationale for these observations has been developed (Brown, 1996; Uraizee, 1996; Du, 2000; Wong, 2001; Rosa, 2007). The key is the behavior of the bubbles, both as they transverse the solution absorbing BSA, and then in the foam, where they coalesce. There are several different properties of bubbles in these different environments that impact the overall E or R, and some of these properties run in opposition. Thus, the aggregate observed behavior (increasing or decreasing E and R) reflects the competition between these opposing phenomena. Specifically,
- The surface area of the bubbles is a key parameter, as that is where protein absorption occurs. Smaller bubbles have a greater surface area per unit volume. In solution, the bubbles are nearly spherical, and the surface area, A, = 6/d (d = diameter of the bubble). In the foam, the bubbles actually adopt a dodecahedral shape (= 12 sided) as a consequence of its six nearest neighbor interactions, and A = 6.6/d (Du, 2001). So, in the liquid phase, smaller bubbles will lead to greater bubble surface area for a given volume, which in turn enhances both enrichment and recovery.
- Liquid entrained in the foam drains back down from the foam phase into the bulk solution. For a variety of complicated reasons, larger bubbles result in faster foam drainage. In general, this drainage serves to increase enrichment, E, as it removes from the foam some liquid which is not as rich in protein as are the bubbles themselves. By default, the remaining foam then is more highly enriched in protein, leading to a larger measured E. For these reasons, larger bubbles in the foam phase increase enrichment.
But, what factors influence the bubble size?
The relationship between superficial gas flow rate and bubble size is complex and appears to depend on the details of the bubble generation process. In the Brown work cited above, faster gas velocities lead to marginally smaller bubbles for the 4-10 µM frit size employed (Brown, 1990). Using frits with larger pore sizes for bubble formation than the one Brown et al. employed in generating the data of Fig. 2, Rosa and, independently, Wong, and Tanner report that the bubble size slightly increases with increasing superficial gas velocity (Wong, 2001; Du, 2002; Rosa, 2007). Enhanced bubble coalescence due to more bubble-bubble collisions at the faster gas flow rates is cited as the rationale (Wong, 2001). In no case are the changes in bubble size very large as the gas flow rate varies over the range examined.
As far as enrichment vs. gas flow rate goes, drainage from larger bubbles in the foam (point 2 above) appears to predominate over the greater protein binding surface area of the smaller bubbles in the liquid (point 1 above) under Brown's experimental conditions, and so the enrichment data in Fig. 2 result. In actuality, slower gas velocities result in complementary effects that both influence E in the same direction; (a) the aforementioned greater foam drainage as a consequence of the larger bubbles, which increases E (Gehle, 1984), and (b) greater protein absorption due to greater bubble residence time in the liquid phase (Bhattacharjee, 1997), which should increase E also (Uraizee, 1996), thus providing a physical explanation for the E vs. gas flow trend shown in Fig. 2. Of course, at the other extreme, faster gas velocities entrain more liquid into the foam. This dilution with relatively protein-poor water diminishes the relative contribution of the protein absorbed on the bubble surface to the overall protein present in the foam, thus contributing to a decrease in enrichment at these faster gas flow rates.
Recovery vs. gas velocity is a different story; in this instance, the lesser amount of foam drainage resulting from the smaller bubbles that are generated at faster gas velocities actually should increase R, the recovery. That is, since the amount of protein residing in the foam is the sum of the (concentrated) protein on the bubble surface layer and the protein dissolved in the interstitial hold-up liquid, then anything that decreases foam drainage will increase the overall amount of protein present and hence recovered from the collapsed foam, and R will increase (Fig. 2, second graph). Enhancing this effect is the fact that more liquid is entrained into the foam at higher gas velocities (Uraizee, 1996; Wong, 2001). The complementary conclusion holds as well; larger bubbles from slower gas velocities lead to more foam drainage and a decrease in recovery.
In a separate series of experiments, Brown also has shown that increasing the flow rate of the liquid through the model skimmer leads to a measurable decrease in recovery, but not much change, except at very low flow rates, in the enrichment, Fig. 3. Of course, the liquid flow rates examined in these experiments are orders-of-magnitude less that the flow rates used in aquarium skimmers, but that disconnect is balanced out somewhat by the fact that the model skimmer used in these studies is just a bit larger than a toilet paper tube. An explanation for these trends is provided below. Protein recovery as a function of liquid flow rate is determined once again by the intersection of two opposing effects. Faster liquid flow rates lead to smaller bubbles both in the liquid phase and in the foam phase (Brown, 1990; Wong, 2001; Du, 2002). The former observation is attributed to less opportunity for bubble-enlarging coalescence, whereas the latter result is explained by citing diminished coalescence of bubbles in the foam as a consequence of protein concentration/bubble surface tension effects (Wong, 2001). If this bubble size effect was dominant, we might expect that recovery would increase as liquid flow increased (= smaller bubbles), since foam drainage would be diminished. However, that expectation is not met experimentally. Therefore, another phenomenon must be in play, and Wong theorizes that faster liquid flow leads to less contact time between the rising bubbles and the protein-containing liquid phase. In this scenario, less protein will be captured on the
bubble surface, and overall recovery suffers (Wong, 2001).
Bubble size occupies a central role in aquarium skimmer performance discussions, and not surprisingly, this topic clearly retains its importance in the protein purification literature as well. Unfortunately, hard data on the direct effect of bubble size on either enrichment or recovery are scarce, and Aksay's report of enrichment and recovery as a function of bubble size represents perhaps the best information to date (Aksay, 2007; see also Uraizee, 1996). In this report, Aksay and Mazza document that the enrichment increases and the recovery decreases as bubble size increases, Fig. 4. Explanations for these trends have been discussed previously; larger bubbles lead to more foam drainage, which increases enrichment. On the other hand, these larger bubbles (a) trap less protein per-unit-volume and of course (b) enhance foam drainage, leading to a decrease in recovery. As an aside, the whole topic of measuring bubble sizes during a skimmer run is fraught with controversy; most authors use photography to characterize bubbles at the skimmer wall only (Brown, 1990; Uraizee, 1996; Wong, 2001; Aksay, 2007), but that approach has been criticized by Tanner, who has developed an indirect technique to measure bubble sizes at any position in the foam (Du, 2001; Du, 2003). He found that wall effects do indeed exist, and foam interior bubbles appear to be ~ 1.5 times larger than the wall bubbles.