I I I I I I I I I

(d) Counter-current trickle bed

Figure 3. Concentration Profiles in Different Bioreactors. (a), (b) Partial pressure in gas phase, dot-dash line (c), (d) Viable biomass concentration, dot-dash line. Dotted line, partial pressure in biophase = HS; solid line, mole fraction in gas phase.

Biophase Gas

Figure 3. Concentration Profiles in Different Bioreactors. (a), (b) Partial pressure in gas phase, dot-dash line (c), (d) Viable biomass concentration, dot-dash line. Dotted line, partial pressure in biophase = HS; solid line, mole fraction in gas phase.

unimportant). However, a goal of 95% contaminant removal now requires not G > 3, but G > 19, which gives some idea of the lower effectiveness of a completely mixed gas phase. The numbers in Table 1 show that this lower effectiveness cancels out the increased k1 a leading to conclusions similar to those for bubble columns. The extra expense of mechanical agitation is rarely justified in gas treatment bioreactors, although reactors designed for a high k1a/g with large aspect (height:diameter) ratios and multiple impellers to prevent bubble coalescence (29) have been proposed for removing hydrogen sulfide. Stirred tanks have been used successfully (39) for the microbial production of ethanol and acetic acid from carbon monoxide and hydrogen (another insoluble gas; H/RT = 51 at 20 °C), but the objective there was high volumetric productivity, rather than complete gas consumption. Several tanks were used in series, and they were pressurized to several atmospheres to improve gas transfer.

In contrast to bubble-gasified reactors, the gasbiophase interfacial area in biotrickling filters is almost constant, fixed by the choice of packing media and only slightly influenced by Ug and the superficial velocity at which the liquid is pumped over the bed U1. This makes it possible to achieve the necessary value of G for contaminants of almost any solubility by appropriate choice of the packing media and the gas flow per unit volume, g. Many types of packing are available and the best choice in a particular situation is a complex trade-off between the following considerations:

Less soluble contaminants will need the large a provided by smaller media and the small g implied by the lower permeability of such media to gas flow. For these contaminants, most of the gas-biophase masstransfer resistance is on the biophase side of the interface, so in correlations of the form of equation 7, c r 0, but A is a fairly strong function of U1.

More soluble contaminants can achieve the same G with the smaller a and larger ggiven by larger media. In this case most of the mass-transfer resistance is on the gas side of the interface, so c > 0 and A is less dependent on U1.

The conservation of mass shows that the change in concentration of a nonvolatile metabolic product in the biophase as it moves down the bed is UgAnym/ U1RT. If this number is small then the biophase approximates complete mixing. If it is large, then the concentration gradients, including that of pH, may be large enough to inhibit metabolism.

• Although a high liquid flow improves mixing and mass transfer, there is a definite upper limit of U1 at which the bed floods. This limit is predictable (28) and is much smaller for small media.

• It is difficult to generalize about the attachment of the microorganisms to the media because this depends on both the nature of the media and the type of biomass involved. However, it is clear that a bed of small media with attached biomass may quickly become plugged if the contaminant is sufficiently concentrated and the biofilm is allowed to grow. It is reasonable to suppose that the larger the media and the higher the liquid velocity, the more cells will be suspended in the liquid rather than being attached. The z parameter in Figure 1 will be correspondingly larger.

• A biophase containing suspended and attached biomass may, if too thick or too dense, have its own mass-transfer resistance (38). While such resistance is usually to be avoided, it may be beneficial. For example, in trying to separate one electron acceptor, NOx, from stack gases that contain another, O2, which is preferred by the microorganisms, it should be possible to take advantage of the much higher solubility of the NOx to create locally anaerobic environments in the biophase.

Somewhere among all of these considerations is a best choice of media for any situation. For dilute, less-soluble contaminants whose metabolism does not produce any nonvolatile metabolic products, the biotrickling filter will look more like a biofilter, with small media, little liquid flow, and attached biomass. For more soluble contaminants which do generate such products, it will tend toward the other extreme of high liquid flows and a completely mixed biophase. If there seems to be no sensible solution, the answer may be to adopt a bioscrubber. Although more expensive, it greatly simplifies the design problem by separating gas-liquid mass transfer from the microbiological considerations. It is certainly preferable to think in terms of this continuum of possibilities and to choose the one best suited to a particular application, rather than considering the different types of gas treatment bioreactors as completely separate technologies, each of which can be forced to fit any job.

Biophase Mixing and the Minimum Effluent Concentration

The single bioreactors with a completely mixed biophase analyzed here have an obvious drawback: all of the microorganisms are exposed to the same physicochemical environment, it is the worst possible environment for metabolism, with the lowest nutrient concentrations and the highest concentrations of metabolic products. One consequence appears in Figure 2. Even with no mass transfer resistance (G = <»), no product inhibition (F = <»), and no wastage of biomass (D = 0), the analysis suggests that complete contaminant removal is still impossible. With no biomass outflow there is no net cell growth, and all of the contaminant is used for maintenance metabolism (equation 2 withi = D = 0 gives q = k). Under these conditions, the dissolved contaminant concentration is not zero, but the stationary phase concentration, Ss, so the partial pressure in the effluent gas cannot be less than HSs, and the fractional removal, x, can never exceed (1 — HSs/ni), even with perfect mass transfer (equation 4 with G = <»). This maximum removal also exists for completely mixed biological wastewater treatment systems, but in that case it does not matter because Ss is much less than common effluent standards. It may matter for some insoluble gasphase contaminants because the corresponding minimum-achievable partial pressure, HSs, may be significant, even when Ss is small. Bioreactors with a completely mixed biophase are a poor choice for such contaminants.

The simplest improvement would be two bioreactors through which the gas flows in series and between which the microbial culture is continuously recirculated. Reactor 1 would have a relatively high effluent partial pressure and dissolved contaminant concentration, providing a good environment for metabolism and cell growth. If the process variables are fixed correctly (a complete analysis is beyond the scope of this chapter) reactor 2 could run with a dissolved concentration S < Ss. The biomass in reactor 2 would then be in an endogenous state, but this would no longer matter because fresh, viable cells would be continuously supplied from reactor 1, and the dying cells would either flow back there to recover or be lost in the liquid effluent. The liquid inflow would go to reactor 1, where the growth nutrients are needed, and where it would dilute the concentrations of any metabolic products.

The more bioreactors in series, the better the average physicochemical environment for metabolism. Columns containing many sieve trays or bubble-cap trays have long been used for gas-liquid contacting in the chemical process industry and should be considered for the biological treatment of gases in cases where the biomass grows best in suspension and the contaminant is relatively dilute, insoluble, and a poor microbial substrate (i.e., HSs/ni is large). These columns are inherently countercurrent contacting devices with the gas flowing upward while the microbial culture falls down from tray to tray. An individual microorganism experiences a series of environments with ever-increasing dissolved contaminant concentration, S, until it is suddenly pumped back to the top tray where conditions are endogenous (S < Ss), but where it can continue metabolizing contaminant and survive long enough to start through the cycle again. Nutrient addition and pH control can be done on each tray as needed, and the liquid effluent can be taken from near the top of the column where cell viability is lowest.

A similar effect, with the gas closer to the theoretical optimum of plug flow, can be achieved in an upflow bio-trickling filter with a small liquid recycle ratio. The only drawbacks are that pH control and nutrient addition can now only be done at the top of the bed, and the liquid can only be removed from the bottom, which is not necessarily the best arrangement. If all the microbial cells are suspended, the concentration profiles of viable biomass and contaminant will be as shown in Figure 3d. In real trickle beds some of the biomass remains attached, which is a great advantage over tray columns when a nonvolatile metabolic product is generated, because it allows a large liquid outflow to remove the product without the danger of washing out the biomass. (Mathematically, D is small because z is small, even though f is large.) Biotrickling filters can also be operated in the gas-downflow, cocurrent mode, which gives concentration profiles (again assuming suspended biomass) more like those shown in Figure 3c. Now, the microorganisms moving down the bed experience an ever-worsening environment for metabolism, with lower S and higher product concentration Sm. They leave the bottom of the bed in an endogenous, low-viability state, which is very suitable for the liquid outflow stream, but which may cause a significant lag phase when cells are recycled to the better environment at the top of the bed. The question of whether the cocurrent mode works better can only be answered definitively by experiments on a particular gas stream.

In biofilters the biophase is far from being completely mixed. It consists of an immobilized mass of up to xXeb viable cells per unit bioreactor volume (less if cell growth has been limited by some nutrient other than the contaminant), and one of the goals of media selection is to provide sufficient surface area so that these cells can be distributed as a biofilm thin enough to avoid mass transfer limitation due to diffusion through the film. However, simple models (38) suggest that biofilters should share the minimum effluent partial pressure, HSs, predicted earlier for bioreac-tors with a completely mixed biophase, because biofilms exposed to any lower partial pressures near the biofilter effluent would be in an endogenous state and eventually die. Experience suggests otherwise, probably because the adsorption of the contaminants on the media increases their effective concentration in the microorganisms' immediate environment. It follows that highly adsorptive support particles, such as compost or activated carbon, will work best, and that biofilters can treat very insoluble, hy-drophobic contaminants if they adsorb well.

Scale-Up

First-stage scale-up experiments can be done in a reactor of any size that is convenient or available, as long as it is large enough so that the flows of gas and liquid are not significantly altered by entrance effects, channeling, or wall effects. Ideally, several different bioreactor types or packing materials are studied, but this is not always possible. The objective is to vary the gas flow, g, the liquid flow, f, nutrient addition, Sji, and so on, to determine how the specified contaminant removal can be achieved most economically. The scale-up problem is then to extrapolate the results in order to calculate the necessary volume (volume = gas flow to be treated/g) and shape (specifically the height, L) of the commercial-scale bioreactor, which must satisfy the process requirements without further experimentation immediately after start up. The best combination of process variables found in the experiments must be close to the optimum identified by the point at the right end of the operating region in Figure 2. To preserve this solution at the commercial scale, the dimensionless numbers D, F, and G that define this point must be kept the same between scales. Scaling up a gas treatment bioreac-tor is thus an example of Reynold's similarity principle, which states that two physical situations are similar only if the all of the relevant dimensionless numbers are identical.

Consider first the case where there is no inhibitory metabolic product and the process needs little or no liquid outflow. F(roi) and D(r0) both drop from the list of relevant dimensionless numbers, and the scale-up problem reduces to finding combinations of gas flow, g, and bioreactor height, L, that keep G constant between scales. Given a mass-transfer correlation of the form of equation 7, G can be written ARTLf/Hg1 ~c). The problem is illustrated graphically in Figure 4, where lines of constant G are drawn on a log-log plot of g versus L. If the small-scale experiments have found different combinations of g and L that produce the specified contaminant removal, x, then this data should fall on such a line, labeled empirical in Figure 4, because the theory suggests that "constant x implies "constant G." Also shown in Figure 4 is an upper limit on the gas superficial velocity, Ug = gL, caused by physical phenomena such as slugging of bubble columns, fluidization of the media in upflow biofilters, or flooding of the impeller in stirred tanks. This limit may be slightly different at different scales. Two extreme cases ofthe scale-up problem can be identified.

For a biofilter or a biotrickling filter treating sparingly soluble contaminants, the mass-transfer factor is little affected by the gas velocity, implying that c = 0 and making the empirical line horizontal. Several important phenomena are lumped into the parameter A, but there is no reason to expect them to vary with scale because microorganisms cannot "know" if they are in a small or large bioreactor. Consequently, performance on the large scale is defined by the same horizontal line, meaning that scale-up is done at constant g, or equivalently at constant gas-residence time (= 1/g). Because g = Ug/L, constant g can be achieved either with a deep bed and a high gas velocity, the right end of the c = 0 line in Figure 4, or with a shallow bed and a low velocity, the left end of the line. Although the former are easier and cheaper to build, they have a much higher pressure drop though the bed, meaning higher costs for gas compression and possibly subsequent

Empirical/ / Projected

N^ Maximum superficial velocity

Bioreactor height, L (log scale) Figure 4. Dotted line, large scale; solid line, small scale.

cooling to return the gas to a temperature tolerable to the microorganisms. The latter are much cheaper to operate and work well as long as the beds are deep enough (at least several hundred times the particle size) to avoid problems of channeling and short-circuiting of the gas flow. Many biofilters consist of shallow beds of media on trays arranged in various configurations in a reactor designed to ensure an even flow of gas into the beds.

At the other extreme is the ideal bubble column described in the section "Gas Biophase Mass Transfer" for which c = 1. The value of G, and thus the contaminant removal, is now independent of gas flow but proportional to bioreactor height. Scale-up must therefore be done by keeping L constant but increasing the reactor width, the large-scale bioreactor operating at maximum g in order to minimize its volume (the operating point in Figure 4). Although this ideal bubble column is an abstraction, it illustrates a paradox that can arise in scaling up practical bubble columns (for which c = 0.7); if the desired removal can be achieved in a small-scale column, then a bubble column is probably not the best solution at the commercial scale. For very soluble contaminants, the L needed to obtain the desired fractional removal may be in the range of laboratory reactors, around 1 m. Scale-up would produce a very unusually shaped large-scale bioreactor, a similar height but several meters wide. There would be no point in making it any deeper, except to provide a safety factor in the design, and it could not be made any narrower and still meet the process requirements. For insoluble contaminants the necessary L may be several meters. The conclusion from laboratory experiments would then be that a bubble column is not a feasible bioreactor, even though the desired contaminant removal may be achievable (albeit with a high pressure drop) with the reactor heights used at the commercial scale.

All practical situations fall between these two extremes, so neither simple scale-up rule, constant g or constant L, is necessarily appropriate. The empirical line will have a positive slope, and a similar line for the same G at the larger scale can be projected from it if any variations of the parameters A and c with scale can be estimated from the published literature, experience, consultations with equipment manufacturers, and so on. The design of the large-scale bioreactor can be anywhere along this projected line in Figure 4, the exact design point again being constrained by the maximum Ug line, but essentially an economic trade-off between the large bioreactors demanded by a small g and the large pressure drops associated with a large L. The essential point is that the shape, and the size of the large-scale bioreactor are fixed by this procedure and should not be chosen based simply on guesswork or convenience.

Some bioreactor types allow design flexibility because they have variables that can be adjusted when drawing the projected line. Both A and c tend to decrease with scale in any bubble-gasified reactor due to poorer gas dispersion, but in mechanically agitated tanks, A is a function of the power input per unit volume, which can be varied to some extent. For trickle beds and biofilters, both A and the bed permeability are predictably related to packing size, allowing the packing to be chosen to give the large-scale reactor both a convenient shape and a reasonable pressure drop. Note that, whether or not such changes are made between scales, the estimates on which the projected line in Figure 4 is based should always be conservative in order to provide a factor of safety in the large-scale design.

When product inhibition is significant, the liquid outflow, f, becomes a critical variable, and all three dimen-sionless numbers D, F, and G must be kept constant if the physicochemical environment of the biomass (S, Sm) and the optimum bioreactor performance established in the small-scale experiments are to be extrapolated to the large scale. An exact solution is possible only when the retention of the biomass by cell immobilization or recycle (the z parameter) can be controlled, although when retention is almost complete (z r 0) only the product DF (from which z cancels) is important, implying that the liquid flow, f, must be kept proportional to the gas flow, g. Biotrickling filters have an extra dimensionless number, the liquid recycle ratio, RR, which introduces the extra difficulty that increasing the bed height, while keeping f and RR constant increases the liquid superficial velocity [Ul = Lf(1 + RR)], thus changing the mass-transfer situation and possibly flooding the bed. The procedure just described will not work in these more complex situations (constant x no longer implies constant G), but the general considerations still apply, including the physical limits on the gas superficial velocity, the desirability of tall small-scale bioreactors to reduce uncertainties, the need for compromise between capital and energy costs, and the possibility of predicting a very inconvenient aspect ratio for the large-scale bioreactor. Judgment and calculation are needed in each case because the trivial solution to the scale-up problem— "build a bigger bioreactor with the same shape, f, and gas the small one"—is unlikely to be the best. Gas treatment bioreactors are one of the few devices whose performance tends to improve with scale, but taking advantage of this tendency in practice is not straightforward.

Brew Your Own Beer

Brew Your Own Beer

Discover How To Become Your Own Brew Master, With Brew Your Own Beer. It takes more than a recipe to make a great beer. Just using the right ingredients doesn't mean your beer will taste like it was meant to. Most of the time it’s the way a beer is made and served that makes it either an exceptional beer or one that gets dumped into the nearest flower pot.

Get My Free Ebook


Post a comment