Mixing is one of the oldest unit operations in the chemical process industries (CPI) and can also be one of the most demanding. The complex challenges presented by the wide variety of mixing tasks and scaleup needs encountered in the CPI have been compounded by a rapid rise in reactor size over the last three decades (Table 1). This trend towards larger-scale manufacturing processes has occurred across a wide range of applications including bulk chemicals, specialty chemicals, formulated products (such as adhesives and coatings), foodstuffs, cosmetics and more. This article focuses on techniques for gas-liquid mixing that can improve performance, which is particularly important when considering scaleup.
(For more on mixing, not specific to gas-liquid, see CE, April 2006, pp. 46–53)
Economy of scale
The increase in reactor size has helped reduce costs and raise competitiveness in the global economy, but this trend has also increased demands on the performance of many reactor agitators. A holistic approach to increasing productivity must address all relevant issues, such as:
1. Product quality and plant productivity should be optimized. Productivity can be expressed as output per time and space as follows:
or as a specific production cost:
In an ideal world, P1 should be maximized, and Pâ‚‚ minimized
2. Raw material and energy usage should be minimized
3. Discharges to the environment and waste should be eliminated or minimized
4. Mechanical reliability should be maximized (this reduces not only maintenance costs and waste, but also extends operational availability and, therefore, productivity)
5. The plant and processes should be designed with an eye to achieving “faster times to market” (plant flexibility). Consideration of flexibility issues at the design stage can significantly shorten process development cycles, thereby representing a huge benefit to producers who need to maintain their competitiveness by regularly bringing new products to market
The competing demands of mixing tasks combined with huge increases in scale (Figure 1) mean that traditional agitator and impeller designs are unlikely to provide the most economical performance. New, recently developed designs can offer significant improvements in process performance and cost reductions. The mechanical design, seal selection, surface finishes and new materials of construction also play a part in maximizing productivity.
Significant developments in agitation and reaction technology over the last 20 years have led to a much better understanding of design criteria, and of the best technical solutions to gas-liquid mixing challenges. Gas-liquid reactor design is one of the areas that clearly shows the benefits of using modern technology to maximize P1 and minimize Pâ‚‚.
Designing for productivity
The complexities of gas-liquid mixing stem from the variety of the physical demands encountered. A low-density compressible gas has to be dispersed into a much denser liquid with a reasonably long contact time. Usually, significant turbulence must be induced into the liquid phase to aid mass transfer and reaction. In addition, rapid movement of the liquid phase is often required at heat transfer surfaces that are far away from the impellers. For good measure, the liquid phase may sometimes contain 40–60 wt.% solids, which must be maintained in suspension. Not surprisingly, the final state of the art has not yet been reached.
Reactors for gas-liquid reactions commonly consist of large pressure vessels with quite sophisticated internal components for gas feed and exhaust, liquid feed and outlet, heat transfer and baffling, as well as for agitation. Due to the hazardous nature of many processes, the mechanical seal with its peripherals is vital for plant safety. As a result, the gas-liquid reactor can represent very substantial capital and operating costs for the user.
The performance of gas-liquid reactors has been significantly improved in recent years due to advances in agitation technology for two- and three-phase systems. Simply changing to a modern agitator design can sometimes have a large positive impact on the productivity of a manufacturing plant. Even so, assessments of design improvements are not commonly performed when building new plants. Designs are often simply copied or scaled up, without implementing even straightforward improvements.
The potential for increased productivity not only helps improve the efficiency of new plant reactors, but can also justify upgrading existing gas-liquid reactors, especially in high-cost regions such as Europe and North America. Firms that specialize in mixing can often help in assessing designs for improved agitation, since they have the benefit of an in-depth understanding of mixing fundamentals and reliable performance data based on years of development work. When process-specific data are necessary to optimize a design or for scaleup, specialty mixing vendors can offer to carry out process simulation and test work in dedicated mixing test equipment (Figures 2a and 2b).
Reactions with pure gases
From a reactor-engineering point of view, there are two major types of gas-liquid reactions: those that operate with a “pure” feed gas, and those in which the gas (air, for example) contains a substantial fraction of inert gases. Typical pure gas reactants are hydrogen, oxygen, carbon dioxide, carbon monoxide, chlorine, ethylene and ethylene oxide. Pure gas applications (Table 2) are generally those in which the gaseous reactant must be completely consumed, because it cannot be released to the atmosphere (for cost, safety or ecological reasons).
For a gas-liquid reaction to take place, a molecule of gas must dissolve in the liquid phase and then meet a molecule of the reagent. Often a catalyst is also required, in which case both of the reactants must meet at an active site on the catalyst. Figure 3 is a schematic representation of the latter process.
The diffusion and reaction at the catalyst is usually very fast due to the very high, specific surface area of the catalyst. Therefore, the reaction is limited by transport of the gas through the boundary layer around the gas bubbles and into the liquid phase. The specific rate of mass transfer through the boundary layer is governed by the standard mass transfer equation.
The film mass transfer coefficient, kL, is mainly a function of the physical properties of the reactants; it is less sensitive to mixing conditions. The specific surface area, a, also depends on material properties, but can be significantly increased by changes to the plant and process design. These factors are usually expressed together as a specific mass-transfer capacity, kLa, since it is difficult to measure either one directly. The term, kLa is usually correlated using an equation of the form:
where c1, x and y are functions of the reactor design and the chemical system.
The term (c*– c) in Equation (3) must also be carefully considered to achieve efficient production. It is sensitive to changes of process design, chiefly due to influences that can be brought to bear on the value c* (the theoretical equilibrium concentration of dissolved gas). The actual concentration of dissolved gas in the liquid phase, c, is more a secondary function — it is determined by the relative rates of mass transfer and conversion of the gas.
The value of c* is, according to Henry’s law, proportional to the partial pressure of the reactive gas. Hence, the use of gas in a pure form facilitates the process efficiency, but the presence of volatile solvents can reduce process efficiency. Increasing the reaction pressure can be a simple means to boost productivity (P1), but operation and investment costs go up with pressure, increasing the specific production cost (Pâ‚‚). The target for the design engineer is, therefore, to optimize kLa and operating pressure, to achieve the required productivity at minimum cost.
We now turn to two typical examples of gas-liquid reactions using pure gases: the continuous hydrogenation of nitrobenzene to aniline; and the batch hydrogenation of a complex organic molecule. Modern technology offers substantial productivity improvements over old designs for both of these cases.
Hydrogenation processes are employed widely throughout the CPI. These reactions are generally carried out on a multiphase basis (gas-liquid-solid) with finely dispersed solid-metal catalysts (usually nickel in various porous forms; or precious metals such as platinum or palladium that are finely dispersed on a porous carbon carrier) and at elevated pressures (up to 100 bar) and temperatures (up to 350°C). The reactions are often strongly exothermic. This heat generation, together with the hazards presented by handling hydrogen and the pyrophoric metal catalysts, exacerbate the demands on the reactor system as a whole. Typical reactions include:
• reduction of nitro compounds to amines
• saturation of multiple bonds
• reduction of carbonyls and nitriles
Continuous hydrogenation. A good example of a hydrogenation reaction is the reduction of nitrobenzene to aniline (Figure 4).
This reaction is fast and very exothermic. The productivity of the reactor is usually limited either by mass transfer of hydrogen gas into the liquid phase or by the need to remove large amounts of reaction heat (a typical conversion rate of 16 m.t./h of nitrobenzene requires about 18 MW of cooling). Maximizing the mass transfer not only increases the reaction rate, but is also desirable because high hydrogen concentrations reduce the risk of producing undesired byproducts at the catalyst. To achieve the desired objectives, the reactor has to deliver high performance in terms of hydrogen mass transfer and heat removal.
Traditional hydrogenation reactors (Figure 5, left) are fitted with a gas sparge ring near the base with an impeller arranged immediately above it to disperse the gas into the liquid. Gas that is not dissolved and reacted reaches the liquid surface and enters the headspace, increasing the pressure. When the headspace pressure reaches the maximum allowable pressure, the feed must stop until hydrogen has been absorbed from the headspace back into the liquid and reacted. An impeller is often arranged at or near the liquid surface in order to generate turbulence, thereby enhancing mass transfer of hydrogen back into the liquid phase. However, “surface splashing” is very inefficient at generating mass transfer and can severely limit the reaction rate.
To improve mass transfer with gas in the headspace, some reactor designs recirculate liquid from the bottom of the vessel via an injector (Figure 5, second from left). This entrains the hydrogen and can produce significantly higher rates of mass transfer. However, it can also suffer several drawbacks:
1. The delay between reaction and removal of heat in the external heat exchanger may result in local overheating (reduced selectivity)
2. Effective mass transfer in the nozzle is followed by a relatively long period of hydrogen impoverishment during recirculation (leading to byproducts and deactivation of the catalyst)
3. The loop requires additional space and provision for thermal expansion between reactor and heat exchanger
4. The catalyst may not be homogeneously suspended throughout the reactor system (reduced effective volume)
5. The pump seal is submerged in the product (abrasive solids, increased maintenance)
Another alternative is to recirculate hydrogen from the headspace back to the bottom of the reactor using an external compressor (Figure 5, second from right). This increases the mass transfer by increasing the superficial velocity as can be seen from Equation (4), and has the advantage of requiring little modification of the reactor. However, compressors are expensive, can incur high maintenance costs and consume significant amounts of power.
A modern solution (Figure 5, right) uses a self-inducing turbine that acts like an internal compressor. This is illustrated in more detail in Figure 6. The feed gas enters the reactor through a simple feed pipe underneath a primary gas-dispersion impeller that is designed with a built-in rotating sparger. This eliminates the need for a gas sparge ring that may block or corrode. Modern primary dispersing impellers belong to the family of modified and enhanced concave turbines that retain their performance characteristics well in the presence of a gas phase. Undissolved gas that reaches the headspace is drawn down the hollow shaft and recirculated into the liquid phase by a specially designed gas-induction impeller (vane turbine). This perpetually recirculates gas from the headspace into the reaction mixture with no need for any external pump or compressor. The induction effect comes from the design of the hollow vanes. Gas is drawn through these into the low-pressure side of the impeller. When the speed is increased above a critical value, the pressure drop at the vanes exceeds the hydrostatic head and recirculation starts.
The gas recirculation rate can be calculated from a set of dimensionless numbers. It is a function of the turbine type, diameter, speed and submergence. The vane type of turbine gives higher suction efficiencies than traditional “channel turbines”, which consist of a series of radial channels between two horizontal disks, and the low pressure is generated by a two-phase flow through the channels.
The higher efficiency of the vane impeller results in a higher gas-suction rate and thus in higher kLa values according to Equation (4) (Figure 7). The overall mass-transfer performance of these impellers is generally much better than surface splashing or a vortex. Comparing self-inducing turbines alone (vane and channel types) shows the higher efficiency of the vane type. A combined gassing system can achieve even higher kLa values, similar to, or even better than, those achieved by the more complex external recirculation systems.
In order to achieve a high concentration of both reactants in the zone of high turbulence, the feed of fresh liquid reactant is directed via a dip-pipe into the liquid intake of the main gas recirculation impeller. The mixture with a high concentration of gas and liquid reactants then flows to the cooling surfaces to remove the heat of reaction and reduce the risk of local overheating.
Several other performance characteristics of the combined gassing system are also generally advantageous:
1. Formation of byproducts can be reduced, and catalyst life extended, because homogeneity is good and the risk of hydrogen starvation is low
2. The heat transfer capacity can be enhanced by installing heat transfer surfaces with specific areas up to 20 m2/m3 in the reactor chamber. Reaction heat can be dissipated where it is created, rather than in a remote heat exchanger
3. Maintenance problems due to the mechanical seal are reduced because it is located in the gas space, away from the process liquids and solids
Batch hydrogenation. Another example comes from the saturation of double bonds in a complex organic molecule. Figure 8 illustrates that improved performance can be achieved using a modern combined-gassing system instead of an old surface-gassing system. Curve (1) illustrates the rate of completion of the reaction using surface gassing, the curves (2) and (3) show the relative rate of completion using combined gassing at different specific powers (2 and 3 kW/m³).
Use of the latest technology can halve the reaction time in many cases. What is more, a well defined relationship between installed power and productivity can facilitate optimization of the plant economics.
Gases with inert components
As discussed in the introduction, many industrial processes use gases that are diluted with significant amounts of inert gas. Most commonly found are processes that use air (21% oxygen in nitrogen) or flue gases (carbon dioxide or sulfur dioxide in nitrogen, oxygen and mixed oxides), which are typically used on a once-through basis and then discharged to the atmosphere. Table 3 lists examples of applications in this category.
Processes that use gases with inerts have some unique properties compared to those that use pure gases. Different mixing and gas-feed devices may be required, and it is particularly important to pay attention to the stoichiometry of the reaction. Taking air as an example, the following list of factors must be considered:
• The mass transfer rate is very sensitive to the concentration driving force Δc = c* – c (Equation ). The equilibrium concentration, c*, is proportional to the oxygen partial pressure, pO2, which is reduced by the dilution in nitrogen to about one fifth
• Full consumption of the oxygen from air is not possible as the driving force, Δc, would tend to zero. Therefore, reactors using gases with inerts must operate with a stoichiometric excess. Industrial-oxidizer exhaust gases generally have around 4–15% residual Oâ‚‚
• The partial pressure of oxygen in the dispersed phase is changing as it is consumed, and this must be taken into account during reactor design calculations. Generally, a mean value of pO2 between feed and exhaust can be calculated by integration. In low-pressure applications, the enhancing effect of the hydrostatic pressure at the feed point on pO2 must be considered
• It is no longer possible to increase the mass transfer by recirculating gas from the headspace, since the headspace gas concentration is depleted
• The large quantity of inert nitrogen together with the requirement for a stoichiometric excess of reactant, and the saturation of the air bubbles by vapor from the liquid phase can sometimes lead to extraordinarily high gas rates in the reactor. This causes flooding of the impellers
• The loss in mass-transfer capacity can partly be compensated by higher kLa (power input, air rate and impeller type/gassing system). Practically, however, it generally leads to the requirement of larger reactors
• Due to economies of scale, many reactors used for oxidations with air are now extremely large. Bulk chemical reactors are usually between 200 and 1,000 m³, and mineral processing reactors can be up to 20,000 m³. At these scales, the problem of homogeneity in the reactor can become significant. This could be in terms of temperature, concentration of dissolved oxygen, reactants or nutrients, and pH. With insufficient mixing, zones may exist where conditions are far from ideal. Whatever the inhomogeneity, the likely consequences are reduced output, lower quality or higher raw-material consumption
Impellers for high gas rates
Traditionally, for reactors that have very high gas rates, combinations of impellers such as flat-blade disc turbines (FBDT), pitch-blade turbines (PBT), or wide-foil impellers (WFI) are used. The state of the art for many applications is now to use impellers of the concave turbines family (Figure 9), adapted to fulfill specific tasks in the reactor.
The full flow of feed gas can be dispersed into the liquid phase using a radially pumping primary disperser (PD). One or more secondary dispersers (SD) may then be installed on the same shaft, but higher in the liquid, to provide a combination of axial blending of the liquid and redispersion of the gas. Redispersion is essential for processes where the bubbles tend to coalesce rapidly. The PD can be equipped with a built-in rotating sparger so that the feed gas can be fed via a simple pipe arrangement, rather than needing a sparge ring. This can save investment costs and is operationally attractive, since a simple feed pipe is much less likely to become a source of contamination than a sparge ring with several hundred holes. Similarly, the risk of blockages or corrosion can be reduced.
Another significant advantage of modern PD and SD impellers is that their performance characteristics are more stable than traditional designs. In the presence of significant quantities of dispersed gas, traditional impellers rapidly loose power (Figure 10), whereas the concave-blade impellers of the PD and SD types are much less affected. This can simplify operations that require varying gas rates, since the agitation power is relatively stable, thereby eliminating the need to adjust the agitator speed to ensure adequate mass-transfer performance (see kLa correlation Equation ). Due to the instability of the performance of traditional impellers, a speed controller is often required, which increases investment and operational costs.
Figure 10 illustrates, using dimensionless terms, the correlation between impeller power and gas feed-rate. These curves differ with varying Froude (Fr) numbers, which are the ratios of centrifugal to gravitational forces influencing the gas-liquid flow patterns. The three fields, as defined in Figure 10, are discussed below.
Relatively low gas rates — Field 1. (Fr = 0.05–0.3, vsg = 0.005–0.02 m/s, P/V = 0.1–0.35 kW/m³) In this range, gasflow (Q) numbers are typical of leaching and bioreactors with low conversion rates. At these conditions, some wide-foil (WF) impellers exhibit an increase of power number of up to 25% due to the change of the flow pattern around the blade. This can be eliminated or minimized by using a carefully designed concave impeller system.
Moderate gas rates — Field 2. (Fr = 0.4–1.5, vsg = 0.02–0.1 m/s, P/V = 1–5 kW/m³) Q numbers are typical of the majority of gas-liquid reactors. Traditional impellers, such as PBT and FBDT, show very significant power number reductions caused by the formation of stable gas cavities on the trailing side of the blades (leading to a reduction in the blades’ drag coefficient).
Very high gas rates — Field 3. (Fr = 0.2–0.4, vsg = 0.3–0.4 m/s, P/V = 1–3 kW/m³) These Q numbers are used in some high-efficiency continuous reactors for bulk-chemicals manufacturing.
Blending in gas-liquid mixtures
Homogeneity of the reaction mixture is a key factor for productivity, product quality and the avoidance of side reactions, but it is often overlooked because of the focus on achieving adequate mass transfer. Homogeneity of not just the liquid reactants is crucial, but also of dispersed and dissolved gas, mixture temperature, pH and the concentration of all reacting components. This is true for batch, fed-batch and continuous operations.
Typical flow patterns for bioreactors are depicted in Figure 11. The reactor on the left has multiple FBDTs; the reactor on the right a combination of the radially pumping PD and radially/axially pumping SD’s. The radial flows from the FBDTs create two equal vortices above and below each impeller. These vortices “roll” on each other, generating poor material exchange in the axial direction. With the PD/SDs, the vortices have a strong axial deformation and extension, they do not roll on each other, but interact. This effect leads to a reduction of blend time by at least a factor of two (at the same power input and gas rate), hence significantly increasing the homogeneity.
There is a limit to the gas flow that an impeller of a given diameter and speed can handle. Above this limit, gas flooding occurs. Flooding can, however, occur in various ways and can look quite different from case to case (Table 4). The various flooding phenomena appear at different gas feedrates, and in a different order, depending on the agitator design and operating conditions. This can lead to considerable confusion when comparing flooding correlations from literature.
Flooding is generally less critical in small tanks, as the agitators are often over designed. But for reactors in “world-scale” plants, especially those that are handling gas streams with large amounts of inerts, a detailed knowledge of the potential flooding phenomena can be crucial.
Independent of scale, each of these phenomenon happens when a certain ratio of gas power to impeller power is exceeded. The gas power is defined as the difference of potential energy of the gas flow between entering and leaving the liquid. It is created by the hydrostatic head and is expressed as:
The ratio of gas/impeller power can be converted to the following dimensionless number and hence, to a scale-independent correlation, which is strictly valid only with geometric similarity.
If the impeller/tank diameter ratio is not constant, a corrective factor with D/T has to be added:
For a particular impeller, each of the flooding phenomenon described in Table 4 will occur at a different value of the constants for flooding correlations, câ‚‚ and c3. The values of câ‚‚, c3 and x can only be determined experimentally.
Figure 13 shows an experiment at pilot scale. As a general rule, the tank diameter for flooding experiments should be at least one meter, because at smaller scales the flow patterns with high gas content and eruptions breaking through to the liquid surface may not be fully evolved. Therefore, câ‚‚, c3 and x often appear to be scale dependent in small-scale trials. Data from small-scale trials are generally not suitable for scaling up to production scale.
“Loss of pumping capacity” is an important flooding phenomenon. Different impellers have different sensitivities to the presence of a gas phase. Figure 14 shows the relative drop of flow velocity at the same point in a tank at the impeller level caused by increasing gas feedrate (normalized as the gas feed power to agitator power ratio: Pgas/Pagit). A similar Pgas/Pagit ratio means that the impellers experience the same gas feedrate at the same impeller power. These curves clearly show how the same phenomenon appears at different gas rates for different impellers. Since flooding is defined as the gas feedrate at which the pumping capacity is reduced to 50% of that in the absence of dispersed gas, Figure 14 shows that:
• The PBT floods (loses pumping capacity) at minimal gas feedrates
• The FBDT also rapidly approaches flooding. Its rapid loss in pumping capacity seems to correspond to its power drop under gassing
• The PD retains its pumping capacity at high gas rates. It floods at gas feedrates about five times higher than the FBDT
As outlined above, mass-transfer capacity and thereby the productivity of the reactor, can be strongly affected by the design of the impellers (or combinations of impellers) and the design of the gas-feed system. The design engineer requires a tool to compare the efficiency of these systems. One such measure is the “oxygen transfer efficiency” (OTE), which is usually defined as the quantity of gas dissolved per kWh of agitator energy input.
However, OTE comparisons should be handled with care. There are many parameters that influence kLa, c* or c, and that have nothing to do with the impeller performance at all. The most common sources of error are:
• kLa depends on the gas flow via the superficial gas velocity, as described in Equation (4). Higher gas rates are paid for by higher compressor power, which does not appear in Equation (8). Hence, a comparison at different vsg will lead to different apparent OTEs
• The solubility, c*, is proportional to reactor pressure. Therefore, higher pressures lead to higher mass-transfer rates. Again, this does not appear in Equation (8). Comparisons made at different pressures will show different apparent OTEs — even with the same impellers
• The mentioned effects will also occur in tall reactors with different liquid levels, and hence hydrostatic head at the gas-feed point. In this and the above case, the increased OTE is paid by higher compressor power
• kLa is affected by many material properties and the operating conditions, which influence issues such as bubble breakup, coalescence and diffusivity. It is sensitive to viscosity, surface tension, dissolved salts, diffusion coefficient of different gas-liquid combinations and so on
• c* is also dependent on many of these parameters
• Both kLa and c* are strongly dependent on the operating temperature
• Even slightly differing concentrations in a reaction mixture can invalidate a comparison of impeller efficiencies
In summary, data from plant and pilot model tests must be handled with care. Without a detailed knowledge of the operating parameters and the measurement methods, comparisons are worthless and can easily lead to wrong conclusions. However, if the pitfalls are avoided, a reliable comparison is possible from either pilot tests with a model system, or from the “real” reaction in pilot or full scale. The OTE comparison can then be used as the basis for the pay-back calculation, and help justify revamping poorly performing reactors. Performing the comparisons in cooperation with a world-class agitator supplier will enable the supplier to optimize the design and provide guarantees regarding performance improvements in the ugraded reactors.
Recent advances in agitator performance and design technology mean that the performance of some existing gas-liquid reactors can be improved by simply replacing the agitator. Specialized vendors can contribute the latest know-how as well as extensive databases, design programs and simulation capabilities.
Modern designs and techniques can have a large impact on the productivity of a plant for a relatively modest investment. They also ensure high performance and cost-effective designs for new world-scale reactors. Other developments not covered in this article, such as improved sealing technology and maintainability, can further increase the cost benefits.
This company recently designed a new skid system (photo) capable of mixing and pumping low to moderately viscous products. The…
The newly redesigned RS-02 high-shear pilot-scale mixer (photo) is well suited for pilot plants and small-scale production, including product development,…
Improved equipment and controls, as well as continuous mixing, improve efficiency Optimization is imperative for today’s chemical processors who strive…
Five different examples are presented in which specialty mixers are used to perform tasks more efficiently than conventional approaches Specialty…
A method of quantifying mixing according to a mixing index is presented. This index can evaluate and predict mixing intensity…