1.

Separations
The ability to analyze, synthesize, and design separation processes is a "distinctive" core capability of chemical engineers. Separation is needed to achieve goals of enrichment, concentration, purification, refining, and isolation. All separations work by exploiting differences between the matter to be separated: size, shape, vapor pressure, solubility, etc. Broadly speaking, separation technology can be divided into two sets: mechanical separation diffusional separation

1.1. Mechanical separation

Mechanical separations depend primarily on differences in particle size, density, velocity, etc. Examples include screening, filtration, sedimentation, centrifugation, decanting, etc. (See Ref. [1] Chap. 29) These techniques are good for separating phases (solid from liquid), but not necessarily components within a phase. Separation of components within a single phase is more difficult energy input is required. Mixing is an "irreversibility", so it cannot be spontaneously undone without an energy input. Once the new phase has been created, mass transfer (diffusion) between the phases moves some components relative to others, so that when the phases are separated so are the components. Thus, these are called diffusional separations.

1.2. Diffusional separations

Diffusional separations exploit differences in vapor pressure, solubility, diffusivity, etc. The driving force for these separations is a difference in chemical activity or concentration which leads to migration of components across a phase boundary. The phases are then separated (typically mechanically) to produce products. In most separation processes, it is necessary to create a new phase for transport to occur. In most industrial applications, the new phase is created by direct energy addition. For instance, in evaporation energy boils off solvent to produce a new vapor phase. Energy addition is characteristic of distillation and crystallization. A new phase can also be created by adding another component (mass). This characterizes separation techniques such as extraction and absorption. An efficient separation is thus one that minimizes the energy or mass input required to reach the desired product purities.

1.3. Common separation techniques

In distillation, vaporization separates a liquid mixture into components or groups of components. In gas absorption or scrubbing, a solute transfers from a carrier gas phase to a liquid solvent phase. The reverse (transfer from liquid to vapor) is called desorption or stripping. Dehumidification is the removal of a pure liquid from a gas carrier phases by condensation. The reverse (from liquid to vapor) is drying. Adsorption is the removal of a solute from a fluid by contact with an otherwise inert solid. When a mixture is treated with a solvent that selectively dissolves certain components, it is called liquid extraction if the initial phase is liquid and leaching if the carrier phase is a solid. Crystallization and evaporation create a new phase by heat transfer. In these methods, usually only one of the product phases is valuable. Membrane technologies, including reverse osmosis, ultrafiltration, etc., allow one component to pass through a selective membrane which rejects other components. In Ref. [2] Tables 1.1-1.4 list many separation types, requirements, etc.

1.4. Analysis of separation processes

Diffusional separations are governed by transport (mass transfer limits the rate of separation) and by equilibrium (thermodynamic limits on the extent of separation). Earlier courses have provided you with tools for examining thermodynamics and momentum transport, and you are currently studying heat transport. The new material in the present course will begin with the study of component mass transfer. Essentially all separation problems require you to develop and solve: material balances (total and component) energy balances equilibrium expressions (phase and chemical) transport expressions (mostly convective)

Much of this should be somewhat familiar now we extend it and look at new applications. References: 1. McCabe, W.L., J.C. Smith, and P. Harriott, Unit Operations of Chemical Engineering (6th ed.), McGraw Hill, 2001, pp. 505-506, 986. 2. Seader, J.D. and E.J. Henley, Separation Process Principles, John Wiley, 1998, pp. 519, 23-27.

2. Equilibrium stage operations

Equilibrium stage operations are based on principles of phase equilibrium. Two phases are mixed together. Some of the components will partition between the phases as the system tries to reach equilibrium. When the phases are separated, one is enriched with the solute and the other depleted. This combination of mixing, approach to equilibrium, and separation is called an equilibrium stage. The basic calculation for an equilibrium stage is the flash calculation that you learned in your multicomponent thermodynamics class. You should review flash calculations in general, and pay special attention to bubble point and dew point calculations. A single equilibrium stage can be used to make a separation. All that is needed is a heater or cooler and a vessel with enough space that liquid and vapor can disengage. These are typically called something like "flash drums" or "knock-out pots". Both are applications of flash distillation.

2.1. Cascades
The separation achieved by a flash often fails to meet process requirements. In practice, several equilibrium stages are connected in series to form a cascade. Many separation processes (distillation, extraction, etc.) are based upon cascades of equilibrium stages. Arranging stages into cascades allows more separation or less energy input than is possible in a single stage. The effectiveness of a cascade depends on how close the stages are able to approach equilibrium. This, in turn, depends on how good the mixing and mass transfer are within the stage. We will often talk about ideal stages the theoretical ideal (or perfect) stage where the two phases are in equilibrium when they leave. In reality, the contact stages we can build are not ideal, but calculations based on ideal stages are a useful approximation. To relate ideal stages to actual ones, we can apply a stage efficiency. Key to the success of an equilibrium stage operation is the use of two separable phases. The heavier is usually assigned the symbol L, the lighter V. For example: Distillation the L phase is liquid, the V phase vapor Leaching the L phase is entrained with a solid, the V phase a liquid solvent Extraction both phases are liquid; that resulting in the desired product, or extract, is usually the V phase, regardless of density.

In this course, we will be pretty loose in how we deal with compositions your text will extend "concentration" (amount of substance per unit volume) to include mass and mole fractions. Normally, x will be used to symbolize the composition of the L phase, y the composition of the V phase. Stages can be arranged in three main ways: cocurrent, countercurrent, and crosscurrent. In a cocurrent arrangement, both streams (L and V) flow in the same general direction; stage 1 contacts the L feed with the V feed.

In countercurrent flow, the fresh streams enter at opposite ends; the first stage the L phase enters is the last stage for the V phase.

A crosscurrent arrangement typically allows one stream to flow through the cascade in series, while the other flows through the stages in parallel. The streams in an equilibrium stage process are numbered according to the stage they leave, thus L1 flows from stage 1 to stage 2; L2 from stage 2 to stage 3. I like to use L 0 to denote the fresh stream entering stage 1; however, Ref. [1] will use numbers only for the interior stages and renames the entering L phase La and the L phase stream leaving the process Lb. The V streams are numbered so that Va is on the same side of the process as L a. This doesn't really effect hand calculations, but in my opinion, renaming streams makes computer assisted calculation trickier. Most solution techniques for cascade separation systems rely on finding the intersection between two sets of equations: one describing the equilibrium, the other describing the operating conditions. The equilibrium "equation" may be expressed as a data table or a plot as well as an analytic equation. Equilibrium sets the ideal limit on the extent of separation an equilibrium stage operation can never purify a stream past its equilibrium concentration. The operating equations for a process are set by its mass and energy balances. These describe the actual amounts and compositions of material and degree of contact. They thus set the limit on the rate or amount of material that may be separated. When a solution satisfies both sets of equations equilibrium and operating it fits the process. Depending on the complexity of the unit process, the equations may be solved analytically, numerically, or graphically. In this course, we'll spend a fair amount of time working with graphical solutions primarily because these tend to help most students reach a better understanding of how the processes work; but be aware that mathematical solutions may be your best choice for the problems you try to solve. References: 1. McCabe, W.L., J.C. SMith, and P. Harriott, Unit Operations of Chemical Engineering (5th Edition), McGraw-Hill, 1993, pp. 495-505. 4

3. Determining mass transfer coefficients

It isn't reasonable to expect mass transfer coefficients to be readily available for any and all systems. The "best" solution is to experimentally measure coefficients on a bench scale (using a wetted-wall column, etc.) and then use the results to design a full scale separation column. When this isn't feasible, more approximate arrangements must be made.

3.1. Correlations
Dimensional analysis of mass transfer suggests correlations of the form:

A number of correlations matching this form are presented in Ref. [1], pp. 533-540. Treybal (1987) suggest the following correlations for use with beds packed with Raschig rings or Berl saddles:

subject to the following definitions

We will be using these in a class example.

3.2. Analogies
Since the basic mechanisms of heat, mass, and momentum transport are essentially the same, it is sometimes possible to directly relate heat transfer coefficients, mass transfer coefficients, and friction factors by means of analogies. Analogies involving momentum transfer are only valid if there is no form drag, hence they are pretty much limited to flow over flat plates and inside (but not across) conduits. Also recognize that if there is much heat or mass transfer, it may change fluid and flow characteristics enough to make analogy worthless; in some cases, a viscosity correction may be used to compensate. A simple, crude analogy recognizes that turbulent eddies transport heat and mass as well as momentum, thus one can argue that the eddie diffusivities are the same for all modes of transport, that is: ET = EH = EM. These values are seldom at hand, though. Another analogy probably the oldest is the "Reynolds analogy", which relates the Fanning friction factor for fluid flow to heat transport:

where the right hand side is the "Stanton number". The Stanton number is a dimensionless group made up of other, more familiar groups. It can be defined for heat transfer or for mass transfer.

The Reynolds analogy gives reasonable values for gases where the Prandtl number is roughly one. Note that one-half the friction factor is the ratio of the overall momentum transported to the wall to the inertial effects in the mainstream. The Stanton number represents the ratio of the overall heat transport to the wall to the convective effects in the mainstream. The Reynolds analogy says that these ratios are equal for mass and momentum transport. The Reynolds analogy postulates direct interaction between the turbulent core of the flow and the walls. If a laminar sublayer is included between these, the Prandtl-Taylor analogy applies:

This form includes the ratio of the mean velocities in the sublayer and core as well as the Prandtl number for heat transfer. Note that when the Prandtl number is equal to one, this equation reduces to the Reynolds analogy. Probably the most widely used is the Colburn (or Colburn-Chilton) analogy. It is based on correlations and data rather than on assumptions about transport mechanisms. The Colburn "jfactor" for heat transfer and the Colburn-Chilton j-factor for mass transfer are:

The heat transfer factor may be modified with the Seider-Tate viscosity correction

although this does not seem to be universally done. When the j-factors are used, the fluid properties in the Stanton number are evaluated at the mean bulk average temperature and those for the Prandtl number at the film temperature (this means two heat capacities!). The Colburn-Chilton analogy is simply

valid for turbulent flow in conduits with NRe > 10000, 0.7 < NPr < 160, and tubes where L/d > 60 (the same constraints as Seider-Tate). A wider range of data is correlated by the Friend-Metzner analogy:

which is valid when NRe > 10000, 0.5 < NPr < 600, 0.5 < NSc < 3000.

3.3. Coefficients from reference conditions

Another possibility is to estimate mass transfer coefficients by comparison with measured values for reference systems. For instance, the overall mass transfer coefficients for the oxygen-water system has been measured (see Ref. [3] Fig. 18.21, p. 581) and can be used to predict overall coefficients for other systems using

Refs. [2, 3] suggest a typical value of n = 0.3, so new values can be obtained using

For gas-film coefficients, Refs. [2, 3] provide data for ammonia-water, and recommend estimation using

References: 1. Brodkey, R.S. and H.C. Hershey, Transport Phenomena: A Unified Approach, McGraw-Hill, 1988, pp. 516-520. 2. McCabe, W.L., J.C. Smith, and P. Harriott, Unit Operations of Chemical Engineering (5th Edition), McGraw-Hill, 1993, pp. 348-552. 3. McCabe, W.L., J.C. Smith, and P. Harriott, Unit Operations of Chemical Engineering (6th Edition), McGraw-Hill, 2001, pp. 532-540, 580-588.

10

4. Algebraic solution of equilibrium stage problems

4.1. The Kremser equation
Solving problems involving equilibrium stage separations requires simultaneous solution of the equilibrium and operating (component balance) expressions. Choice of a solution technique algebraic, graphical, or numerical depends on the form of the expressions. The Kremser Equation, an "absorption factor method", provides an algebraic solution for analyzing equilibrium cascades. It cannot be used for every problem, but is convenient for several cases, notably: dilute gas absorption (when set up on "solvent free" basis) distillation (use for the extreme ends of a high purity separation where the curvature of the equilibrium curve is not significant) leaching

4.2. Modeling
The equations will be developed for a countercurrent cascade of N stages. Begin by writing the steady state component balance over n-1 stages:

The equilbrium expression will be written in terms of a "K-value" (Ref. [1] develops these equations starting with a linear equilibrium expression with slope m):

The absorption factor will then be defined. It is the ratio of the local slope of the operating curve to that of the equilibrium curve. Similar expressions can be defined to serve as "stripping factors", or "extraction factors", or "wash factors", etc.

11

The absorption factor thus varies from stage to stage. These three expressions (component balance, equilibrium, absorption factor) are then combined and rearranged

If the same steps were taken for a balance over n-2 and n-3 stages, the results would be:

These expressions are then "nested" into the first to obtain

This process is repeated, until the balance over 1 stage is incorporated

The balance will be written one more time, over n stages

Then the last two equations are set equal and rearranged:

12

If the absorbent fed is pure, x0 = 0, and the second term vanishes. It is then convenient to define the "fraction NOT absorbed", the ratio of solute leaving to solute fed

which can sometimes be used to compact the notation. This equation allows calculation of the recovery; but it is unlikely that anyone would have all the required absorption factors.

4.3. Group method approximation

The absorption factor A varies from stage to stage as the liquid and vapor flows and equilibrium shift. The "group method" approximation says that we can assume an average, "effective" value of the absorption factor that is defined to be the same for all stages. Note, though, that if both the equilibrium curve and operating curve are straight lines, no approximation is involved. This allows algebraic simplication of the recovery fraction

provided one remembers the rules for geometric series from calculus

A similar simplification can be done on the L0x0 term, noting that the order is one less. The full equation is thus

13

(Save this we'll come back and use it again later) The coefficient on L0x0 represents the consequences of both impure absorbent and the fact that vapor flow may do some stripping of the enriched absorbent. It thus makes sense to express this quantity in terms of the stripping factor:

Beginning by setting up a common denominator, the L0x0 coefficient can be rewritten to obtain

so that the overall equation is

This equation is useful in solving some problems.

4.4. Operating equation forms

Now go back to the form I said to save, and rearrange it in the operating equation form. From here on we will assume that the flow rates L and V and the equilibrium K-value are constants. This means that both the equilibrium and operating curves will be straight lines and that the absorption and stripping factors are constants.

14

Define the hypothetical equilibrium vapor composition, substitute, and rearrange.

Next we need a rearranged version of the balance over n stages

(Note that yn* = yn). This can be used to calculate A from known endpoint compositions

which is the same as equation 20.23 in Ref. [1]. We might also combine the last two equations to get

15

(same as Eq. 20.24 in Ref. [1]) which can be used to determine the number of stages needed to make a separation. References: 1. McCabe, W.L., J.C. Smith, and P. Harriott, Unit Operations of Chemical Engineering (6th Edition), McGraw-Hill, 2001, pp. 632-638. 2. Seader, J.D. and E.J. Henley, Separation Process Principles, John Wiley, 1998, pp. 242-246.

16

5. Equilibrium diagrams
In order to solve equilibrium stage problems, you must have a model or correlation for the vapor-liquid equilibrium (VLE) and physical properties of the system. When two phases are in equilibrium, VLE data enables us to relate the composition of a liquid phase to the composition of the vapor phase. Equilibrium data may be obtained by experiment, by thermodynamic calculation, or in published sources. It is typically presented either in tabular form or as an equilbrium diagram. Diagrams may take several forms: Txy, Pxy diagrams boiling point diagrams ternary diagrams solubility diagrams

The figure (also see Fig 21.3 in Ref. [4]) shows one common way of plotting equilibrium data the Txy diagram. It represents a binary mixture, and all compositions are expressed as mole fractions of the more volatile component; x in the liquid phase or y in the vapor phase. (The less volatile component is thus 1 - x and 1 - y.) The lower curve plots the bubble point of the binary mixture as a function of composition. The upper curve is the dew point. For a given temperature and composition, this diagram tells us the nature and composition of each phase of the mixture that is present. Similar drawings can be constructed at constant temperature while allowing pressure to vary. You should have seen some of these diagrams in your material balance and stoichiometry class (for instance, see Felder & Rousseau (3rd ed.), Section 6.4d).

17

5.1. yx diagrams
In this class, we will often work with plots of vapor phase composition vs. liquid phase composition, y vs. x (also see Figures 21.2 and 21.20 in Ref. [4]).

Usually we'll include a "45 degree line" (x = y) on the diagram for reference. These diagrams are typically made at constant pressure, so each point represents a different temperature. An xy diagram like this may be constructed from a Txy diagram by picking a temperature, reading the corresponding y and x and plotting them against each other.

5.2. Ideal VLE

Problems can be greatly simplified in cases where the VLE behavior of the system is ideal or can be represented in a simple form. This usually means we can work with one of three main alternatives: 5.2.1. Raoult's and Henry's laws Raoult's law (review your notes from material balances and from thermodynamics) relates the vapor and liquid equilibrium compositions of a mixture to the pure component vapor pressure and system pressure:

There are few theoretical tools for predicting the equilibria of gases dissolved in liquid solvents. One way of correlating the data is to use Henry's law, which takes the same form as Raoult's law but replaces the vapor pressure with a Henry's law coefficient (also a function of temperature). Henry's Law applies to low solubility gases dissolved in liquids (it would be a bad choice for ammonia, say ...).

18

5.2.2. Equilibrium "K-values" A "K-value" or equilibrium coefficient is the ratio of the vapor composition to the liquid composition. These are often correlated as functions of temperature and can be found in published data, as nomographs, or as equations, such as:

You likely encountered equilibrium coefficients in your multicomponent thermodynamics course. Smith, Van Ness, and Abbott (5th ed., pp. 494-495) have a set of "DePriester charts", which are plots of correlated K values for hydrocarbons. It might be smart to review using these. A related concept, the "distribution coefficient", can be used to relate the equilibrium compositions of two liquid phases. 5.2.3. Relative volatility The "relative volatility" is the ratio of the K values for two components:

which in the case of a binary system can be rearranged to:

Better yet, if the relative volatility is constant, the expression can be rearranged into the form needed to plot an equilibrium curve for a set of x-values:

(the math can be found at the end of this document). Relative volatility is generally a much less strong function of temperature than the component vapor pressures; in many systems, it is acceptable to assume that the relative volatility is constant over a range of temperatures and compositions. High relative volatilities produce xy diagrams with a great deal of separation between the equilibrium curve and the 45 degree line. A relative volatility less than one is probably "upside down" the more volatile component is in the denominator.

19

5.3. Nonideal systems

In cases where the VLE cannot be treated as ideal, more rigorous models based on multicomponent thermodynamics are required. We won't worry about these for awhile. 5.3.1. Azeotropes Most of the time, we deal with systems where the equilibrium curves have no inflection points; that is, as the concentration of the less volatile component increases, so do the dew point and bubble point. If, however, there are strong physical or chemical interactions between components, diagrams may look different. In some such systems, there is a critical composition where the liquid and vapor compositions are identical. Once this composition is reached, separation cannot continue without changing pressure. These mixtures are called azeotropes they will have minimum or maximum boiling points. Minimum boilers are more common. Composition of the vapor produced from an azeotrope is the same as the liquid, so an azeotrope can be boiled at constant pressure without changing composition.

5.4. Enthalpy concentration diagrams

For a binary mixture, cp, H, , etc. are functions of composition. All of this information can be used to construct an enthalpy concentration diagram (see Ref. [4] Fig 21.24 for another example):

These are constructed at constant pressure and each point represents a different temperature (but beware! T(x = 0.5) is not equal to T(y = 0.5)). The upper curve is a plot of saturated vapor enthalpy vs. vapor composition, the lower curve saturated liquid enthalpy vs. liquid composition; consequently, for any given composition, the difference between the two curves is the latent heat of vaporization of the mixture. Points between the two curves ("inside the

20

envelope") represent two phase mixtures. Points outside the envelope represent superheated vapor or subcooled liquid. Constant temperature tie-lines between the vapor and liquid curves are sometimes provided (although rarely as many as one might want). Developing Hxy diagrams isn't particularly difficult, but it does require attention to detail. To begin with, reference temperatures must be selected and applied with consistency. It is also important to remember that liquid solution enthalpies must include both sensible heat and heats of mixing:

It may be most convenient to set the reference temperature at that corresponding to the heat of solution data (which will be negative if heat is evolved on mixing). Remember that enthalpy is a state function, so you may calculate it along any "path" that you choose; and usually want to choose the one that makes calculation easiest. Vapor enthalpies are typically calculated using heat capacities and the latent heat of vaporization:

Mixture values are usually hard to find, so as an approximation, assume the unmixed liquids are heated separately to the dew point, vaporized at that temperature, and then mixed:

There is a direct relationship between Hxy and xy diagrams. If you have tie lines connecting equilibrium liquid and vapor compositions on an Hx diagram, you can use it to construct an xy diagram. The endpoints of the tie lines are at the equilibrium compositions needed/used on the xy diagram. Constructing an equilibrium curve from an Hxy diagram (or adding tie- lines to an Hxy diagram from data on an Hxy diagram) is a nice construction. Place the equilibrium diagram below the Hxy and line up the scales. Starting from a tie line liquid endpoint on the Hxy, drop a vertical line onto the equilibrium plot. This will be an x-coordinate. Next drop a line from a vertical endpoint on the Hxy. When you hit the 45 degree line, turn at a right angle. The intersection between this line and the previous line will give you one (x,y) point on the equilibrium curve. You can reverse the construction and use an xy diagram to get the tie lines on your Hx plot. Take x coordinates straight up and mark the intersection with the bubble point curve. Take ycoordinates over to the 45-degree line, turn and carry them up and mark the intersection with the dew point curve. Connect the two intersections to get a tie line.

21

5.5. Rearrangement of relative volatility expression

References: 1. Kister, H.Z., Distillation Design, McGraw Hill, 1992, p. 14. 2. McCabe, W.L. and J.C. Smith, Unit Operations of Chemical Engineering (3rd Edition), McGraw-Hill, 1976, pp. 486-488. 3. McCabe, W.L., J.C. Smith, and P. Harriott, Unit Operations of Chemical Engineering (5th Edition), McGraw-Hill, 1993, pp. 508, 553-554. 4. McCabe, W.L., J.C. Smith, and P. Harriott, Unit Operations of Chemical Engineering (6th Edition), McGraw-Hill, 2001, pp. 628, 672-674. 5. Seader, J.D. and E.J. Henley, Separation Process Principles, John Wiley, 1998, pp. 36-42, 173-176, 207-211. 6. Treybal, R.E., Mass-Transfer Operations, 3rd Edition (Reissue), McGraw-Hill, 1987, pp. 343-350, 357-360.

22

6. Flash distillation
6.1. Definition and purpose

Flash distillation (sometimes called "equilibrium distillation") is a single stage separation technique. A liquid mixture feed is pumped through a heater to raise the temperature and enthalpy of the mixture. It then flows through a valve and the pressure is reduced, causing the liquid to partially vaporize. Once the mixture enters a big enough volume (the "flash drum"), the liquid and vapor separate. Because the vapor and liquid are in such close contact up until the "flash" occurs, the product liquid and vapor phases approach equilibrium. Simple flash separations are very common in industry, particularly petroleum refining. Even when some other method of separation is to be used, it is not uncommon to use a "pre-flash" to reduce the load on the separation itself. Flash calculations are very common, perhaps one of the most common chemical engineering calculations. They are a key component of simulation packages like Hysis, Aspen, etc. When designing a flash system it is important to provide enough disengaging space in the drum. Drums can also be designed as cyclone separators.

6.2. Modeling equations

Modeling assumptions: 1. No heat losses to surroundings 2. Ideal gas behavior for vapor 3. Perfect mixing F, V, and L are the volumetric flows of the feed, vapor product, and liquid product, VT the total volume of the tank. Total material balance:

23

In many cases, the total mass in the vapor phase is much less than that in the liquid phase. One can then approximate the total mass of the system by the liquid phase mass. Component balance (for a binary system):

Enthalpy balance:

6.2.1. Steady state model These modeling equations can be reduced to their steady state form by setting the time derivative to zero. In the following, let F,V, and L be mass flow rates.

We'll consider a couple different rearrangements of the steady state model. These will be useful in some of the solution methods we might try. First, define the fraction vaporized of the feed as f = V/F so that the material balance can be rewritten as:

and consequently L/F = 1 f. The component balance can then be written as:

or

which are Equations 21.1 and 21.2 form Ref. [2]. This is an "operating line" equation plotted on a graph whose coordinates are vapor composition vs. liquid composition, it yields a

24

straight line. To plot the line, you only need to know the fraction vaporized (to get the slope) and the feed composition (to get the intercept). Later in the term, we will define a "quality" variable q to be equal to the fraction liquid in the feed, so that f + q = 1. Our second rearrangement will enable us to examine the relationships between the material flows and the compositions or enthalpies. First, solve the material and component balances simultaneously:

to get

This is a line when y is plotted against x. This formula can be rearranged to get:

The material and enthalpy balances can be combined in the same fashion to obtain:

These equations thus define operating lines in terms of the compositions and the enthalpies, respectively.

6.3. Calculation techniques

To solve a flash distillation problem, one simultaneously solves the operating and equilibrium equations. Flash calculations can be solved directly, but usually require an iterative solution. Graphical techniques are also common. Often, the choice of technique depends on the available form of the equilibrium relationship. Example: A mixture of 50 mole percent normal heptane and 50 percent normal octane at 30 degrees C is continuously flash distilled at 1 standard atmosphere so that 60 mole percent of the feed is vaporized. What will be the composition of the vapor and liquid products?

25

Given: xF = 0.5, f = 0.6 Find: x, y Basis: F = 100 mols Applying the mass balance yields:

The solution method really depends on what form the equilibrium data takes. If you have an equilibrium xy diagram, the problem can be solved graphically by plotting the operating line on the equilibrium diagram. The operating line is:

and since both slope and intercept are known, the line can be plotted. Sometimes, the scale of the equilibrium diagram is such that it is tricky to locate the intercept. In that case, it is usually easier to use the slope and some other point. The easiest other point to find it that where the 45 degree line is crossed (y = x) and x = xF, or the point (xF, xF). It is easy to show that this point satisfies the operating equation (just substitute in x F for both x and y). If you then plot the operating line on the equilibrium diagram, you can read the coordinates where the two curves cross for the solution

and estimate the solution to be x = 0.39, y = 0.58.

26

If an analytical expression is available for the equilibrium curve, one simply needs to set the two equations equal to each other. Say for instance that the system has a constant relative volatility of 2.16. Then the solution occurs when y = yeq, or

Which can be solved to get x = 0.386. This can then be substituted back into either original equation to get y = 0.576.

6.4. Water-hydrocarbon systems

Hydrocarbons and water usually can be assumed to be completely immiscible for the purposes of flash calculations (one exception: high temperature systems with small amounts of water). In this case, each liquid phase acts independently of the other and each immiscible phase (HC, W) has its own vapor pressure. The total vapor pressure is thus the sum of the vapor pressures of each phase:

If the total vapor pressure is less than the system pressure, there will not be an appreciable vapor phase unless a noncondensible gas is present to make up the difference. Consider a stream containing water (xW = 0.1) and mixed hydrocarbons (xHC = 0.9) at 80 Celsius. This will split into two immiscible phases. The vapor pressure of the water phase will be determined by the temperature only (so it will be about 355 mm Hg). If a second stream with xW = 0.7 and xHC = 0.3 at the same temperature is considered, the vapor pressure exerted by the water phase will be the same as well, because the water phase composition is essentially the same (all water). Consequently, we can see that the bulk feed composition doesn't really effect the water vapor pressure. Now, say that you are considering a case where the hydrocarbon mixture is ternary and equimolar. The vapor pressure exerted by the hydrocarbon phase will be

where the hydrocarbon vapor pressure is a function of the temperature and the composition of the hydrocarbon phase. This means you can do bubblepoint calculations for the hydrocarbon phase as if there is no water present as long as you use the total hydrocarbon pressure instead of the system pressure in the calculations. Also note that the calculation will be the same for both cases (xHC = 0.9 and xHC = 0.3) considered above because the composition of the hydrocarbon phase is the same in both despite the different amounts of hydrocarbon present.

27

Dew point calculations depend on partial pressures more than vapor pressures; consequently, they don't benefit from the immiscibility. When a hydrocarbon-water mixture is cooled, a temperature will be reached where one component will begin to condense. Note however, that typically only one component will condense initially the water and the hydrocarbon mixture must be checked separately when determining dewpoints. To reiterate, the relative amounts of hydrocarbon and water are unimportant in a bubble point calculation, because they depend on the vapor pressures of the immiscible phases, not on the bulk composition. Dew point calculations, however, are effected by the bulk composition. References: 1. Smith, B.D., Design of Equilibrium Stage Processes, McGraw-Hill, 1963, pp. 105106, 108-110. 2. McCabe, W.L., J.C. Smith, P. Harriott, Unit Operations of Chemical Engineering, 5th Edition, McGraw-Hill, 1993. pp. 521- 525. 3. Treybal, R.E., Mass-Transfer Operations, 3rd Edition (Reissue), McGraw-Hill, 1987. pp. 346, 348-349, 357-360, 363-365.

28

7. Distillation I: Principles
7.1. Definition and purpose
Distillation is the most widely used separation process in the chemical industry. It is also known as fractional distillation or fractionation. It is normally used to separate liquid mixtures into two or more vapor or liquid products with different compositions. Distillation is an equilibrium stage operation. In each stage, a vapor phase is contacted with a liquid phase and mass is from vapor to liquid and from liquid to vapor. The less volatile, "heavy" or "high boiling", components concentrate in the liquid phase; the more volatile, "light", components concentrate in the vapor. By using multiple stages in series with recycle, separation can be accomplished.

7.2. Operating principles

The feed to a distillation column may be liquid, vapor, or a liquid-vapor mixture. It may enter at any point in the column, although the optimal feed tray location should be determined and used. More than one stream may be fed to the system, and more than one product may be drawn. A column is divided into a series of stages. These correspond to a cascade of equilibrium stages. Liquid flows down the column from stage to stage and is contacted by vapor flowing upward. Traditionally, most columns have been built from a set of distinct "trays" or "plates", so these terms end up being essentially interchangeable with "stages". Each tray in a distillation column is designed to promote contact between the vapor and liquid on the stage. Distillation can be conducted in a packed column (just as absorption can be done in a trayed column), but we will focus on trayed columns for the present.

29

Stages may be numbered from top down or bottom up. When analyzing a stage, flows and compositions take the number of the stage they leave. The text for this class calls the top tray of the column "Tray 1" and numbers downward this is the convention we will use. Ref. [2] also denote the streams between the column top and condenser with an "a" subscript and those at the bottom with "b". Personally, I generally prefer to let "Tray 1" be the bottom tray of the column, the reboiler "Tray 0" and number upward (so if you catch me doing this, don't panic). I like this way of numbering because it tends to simplify computer based calculations. The product leaving the top of the column is called the overhead product, the "overhead", the "top product", the distillate, or "distillate product". Distillate product may be liquid or vapor (or occasionally both) depending on the type of condenser used. Most of the time the distillate flow rate is assigned the symbol D, and the composition xD or yD. The product leaving the bottom of the column is called the bottom product or "bottoms", and given the symbol B, with composition xB. In some situations, notably petroleum refining, one or more intermediate or "sidedraw" products may be removed from the column. 30

Vapor leaving the top of the column passes through a heat exchanger, the condenser, where it is partially or totally condensed. The liquid which results is temporarily held in the "accumulator" or reflux drum. A liquid stream is withdrawn from the drum and returned to the top tray of the column as reflux (R or L) to promote separation. The portion of the column above the feed tray is called the rectification section. In this section, the vapor is enriched by contact with the reflux. The portion of the column below the feed tray is called the stripping section. The liquid portion of the feed serves as the reflux for this section. The operating pressure of the column is typically controlled by adjusting heat removal in the condenser. The base of the column is typically used as a reservoir to hold liquid leaving the bottom tray. A heat exchanger, the reboiler, is used to boil this liquid. The vapor which results, the "boilup" (V) is returned to the column on one of the bottom three or four trays. In normal operation, there are five "handles" that can be adjusted to manipulate the behavior of a distillation column the feed flow, two product flows, the reflux flow, and the boilup flow (or reboiler heat input). A normal column has a temperature gradient and a pressure gradient from bottom to top.

7.3. Ideal stages

Stages are built to maximize contact between the incoming vapor and the incoming liquid. During the contact, some of the liqht component in the entering liquid is vaporized and leaves with the vapor; some of the heavy component in the entering vapor condenses and leaves with the liquid.

By definition, an ideal stage is one where the vapor and liquid leave the stage in equilibrium. Consequently, the vapor composition functionally depends on the liquid composition. Ideality is an approximation, but stage efficiencies can be used to account for real cases. A key result of the ideal stage assumption is that liquid streams leaving an ideal stage are assumed to be at their bubble point. Vapor streams leave at their dew point.

31

When no azeotropes are present, both top and bottom products may be obtained in any desired purity if enough stages are provided and enough reflux is available. In practice, there are limits to the number of stages and to the amount of reflux, so not every separation can be accomplished. Theoretical limits on performance are imposed by total reflux (minimum stages) and minimum reflux (infinite number of ideal stages).

7.4. Condensers and reboilers

There are two main categories of condenser, differentiated by the extent of condensation. In a total condenser, all of the vapor leaving the top of the column is condensed. Consequently, the composition of the vapor leaving the top tray y1 is the same as that of the liquid distillate product and reflux, xD.

In a partial condenser, the vapor is only partially liquefied. The liquid produced is returned to the column as liquid, and a vapor product stream is removed. The compositions of these three streams (V1, D, and R) are different. Normally, D (composition y D) is in equilibrium with R (composition xD).

A partial condenser functions as an equilibrium separation stage, so columns with a partial condenser effectively have an extra ideal stage. The "reflux ratio" is an important parameter in column operation. It is normally defined as the ratio of reflux to distillate (L/D), although other formulations (L/L+D, etc.) are occasionally used. Most reboilers are partial reboilers, that is they only vaporize part of the liquid in the column base. Partial reboilers also provide an ideal separation stage. Reboilers take several forms: they may be "thermosiphon" types that rely on the thermal effects on density to draw liquid through the heat exchanger, "forced circulation" types that

32

use a pump to force liquid through, or even "stab-in" types that come through the side of the column into the liquid reservoir. In large, complex columns, sidestream reboilers can be used. These draw liquid off a tray, heat it, and then return the vapor liquid mixture to the same or a similar trays.

7.5. Feed condition

The thermal condition of the feed determines the column internal flows. If the feed is below its bubble point, heat is needed to raise it to where it can be vaporized. This heat must be obtained by condensing vapor rising through the column, so the liquid flow moving down the column increases by the entire amount of the feed plus the condensed material and the vapor flow upward is decreased. If the feed enters as superheated vapor, it will vaporize some of the liquid to equalize the enthalpy. In this case, the liquid flow down the column drops and the vapor flow up is increased by the entire amount of the feed plus the vaporized material. If the feed is saturated (liquid or vapor), no additional heat must be added or subtracted, and the feed adds directly to the liquid or vapor flow. Feed effects are important enough that a variable, q is assigned as a descriptor. Subcooled liquid q>1 q = 1 + cpL(Tbp-Tf)/ Saturated liquid (bubble point feed) q=1 Partially vaporized 0<q<1 q is the fraction of the feed that is liquid. It can be found by doing a flash calculation and then q = (L/F) = (1-V/F) Saturated vapor (dew point feed) q=0 Superheated vapor q<0 q=-cpV(Tf-Tdp)/

References: 1. Foust, A.S. et al., Principles of Unit Operations, 2nd Edition, John Wiley, 1980, pp. 13-14. 2. McCabe, W.L., J.C. Smith, P. Harriott, Unit Operations of Chemical Engineering, 5th Edition, McGraw-Hill, 1993.

33

34

8. Distillation II: Modeling

When modeling a distillation column, one can draw balances on the entire system, the column base, the accumulator, each tray (or a group of trays). Mass and component balances are always required. In special cases, the energy balances can be neglected. The overall material balance on a distillation column with NT trays is:

where F, D, and B are mass flow rates, MB is the mass in the column base, MD the mass in the accumulator, and Mn the mass on tray n. Often, the mass of the vapor in the column is much smaller than the mass of the liquid in the column; in such cases, it is often a reasonable approximation to write the mass accumulation terms for the liquid phase only. The overall component balance is:

This can also be written with accumulation terms representing the liquid alone, if the appropriate assumptions are made. The balances required for the accumulator are:

where V1 is the vapor flow off the top tray to the condenser, and

for a total condenser or

for a partial condenser with a vapor product. The base and reboiler balances are

35

for a partial reboiler with boilup rate VB (LNT is the liquid flow off the bottom tray of the column), and

Don't forget that xB and yB are in equilibrium with each other. For each tray, we can write:

in a general form that allows one feed (Fn) and one liquid product (Pn) on each tray. Usually, only one tray will have a feed stream, and there will be only a few, if any sidedraw products. You can also have vapor products, which will require replacing Pnxn in the component balance with Pnyn.

8.1. Steady-state model

The column mass and component balances in their steady state forms are:

These are equivalent to Equations 21.3 and 21.4 in Ref. [2]. These two equations can be solved simultaneously to eliminate any one variable. For example:

Similar rearrangements yield:

and

36

Some authors like to use these ratios in their calculations. I usually find it makes more sense to work directly from the balance equations.

8.2. Feed tray

It also may be useful to examine the steady state balances on a feed tray. The total material balance on feed tray n is

To capture all necessary detail, one must recall that the feed can be both vapor and liquid. As described by the feed variable q, the fraction liquid is Fq, so the vapor rate above the feed tray and the liquid rate below the feed tray will change. The new values will be

An equivalent analysis can be made for draw trays. It will be necessary to specify whether the draw is made from the liquid space or vapor space on the tray. This defines a q value for the draw. The equations will then be essentially the same, but keep in mind that the effect of a draw is to reduce the traffic in the column while a feed increases traffic. References: 1. Foust, A.S. et al., Principles of Unit Operations, 2nd Edition, John Wiley, 1980, pp. 13-14. 2. McCabe, W.L., J.C. Smith, P. Harriott, Unit Operations of Chemical Engineering, 5th Edition, McGraw-Hill, 1993.

37

38

9. Distillation III: Operating equations

For a binary system, we can make a plot of the stage vapor composition vs. stage liquid composition, yn vs. xn. If the points for all the stages are joined, the plot represents the operating path of the system. If the liquid and vapor rates are constant through a section of the column, the operating curve will be a straight line. In the following analysis, we will assume our column has a single feed and no sidedraw products, but the equations are easily adjusted should this change.

9.1. Rectifying section

Writing steady state balances over the entire rectifying section, including the accumulator produces:

The latter can be rearranged: yn + 1 = Ln D xn + xD Vn + 1 Vn + 1

and the total mass balance used to eliminate the vapor rate: yn + 1 = Ln D xn + xD Ln + D Ln + D

This equation is the operating equation for the rectifying section of the column. Some authors prefer to call this the material balance equation.

9.2. Stripping section

Next, follow a similar procedure for the stripping section, including the reboiler:

Ln 1 B x n 1 xB Vn Vn Ln 1 B yn = xn 1 xB Ln 1 B Ln 1 B yn = and you have the operating equation for the stripping section.

39

9.3. Equimolal overflow

Calculations using these equations are much more convenient if the two operating equations are lines. This is true only if the liquid and vapor flows do not change in a given section of the column. What is required for them to be constant? Constant Molal Overflow (also called equimolal overflow) is what is needed. This occurs when the molar heat of vaporization of the liquid phase is essentially equal to that of the vapor phase. That is, the heat needed to vaporize one mole of liquid is roughly the same as the heat released when one mole of vapor is condensed. Consequently, any condensation on a stage is balanced out by vaporization and flow rates within the column are changed solely by feed or product streams. The quickest way to check the validity of an assumption of equimolal overflow is to compare the heats of vaporization of the components. If their ratio is roughly 1:1, the assumption is probably acceptable. When equimolal overflow is present, Ln Ln+1 and Vn Vn-1 for any given section of the column. (Remember: the only things causing L and V to change are feeds and products). L/V, D/V, and B/V are all constants within a section, so the operating equations are lines:

which can be expressed in terms of the reflux ratio

yn + 1 =

L D D D xn + xD L D+ 1 L D+ 1 RD 1 yn + 1 = xn + xD RD + 1 RD + 1

If you examine the equations, you see that when x = x D that y = xD as well (prove it by substituting xD into the equation). This means that the point (xD, xD) lies on the rectifying line. Thus, if we assume equimolal overflow, the rectifying operating line can be drawn using only this point and the slope. Similarly, the stripping operating line runs through the point (xB, xB). If there is a secondary feed or a sidestream product, there are more than two regions in the column; consequently, an additional operating line is required for each sidestream. The forms of the the equations don't change, but the numerical values do. The slope of line through the region will be the (L/V) ratio for that region.

40

9.4. Feed line

It would be nice to know where the rectifying line and the stripping line intersect. The vapor and liquid flow rates will be different in the different sections, so (still assuming equimolal overflow!) the rectifying and stripping section steady state balances are:

Subtracting the stripping balance from the rectifying balance yields

(VR VS ) yn = ( LR

LS ) xn 1 + DxD + BxB

In order to rearrange this to a convenient form, refer to the steady state model for the feed tray, where we showed how the flow rates were changed by the feed:

These can be rearranged to get the flow difference terms we'd like to replace:

and we can substitute them back into the balance equation to get

(1 q ) Fyn =

qFxn 1 + DxD + BxB

Similarly, we can substitute the column overall material balance

to obtain

This is a third operating equation. The feed variable q and the feed composition xF are constants so this equation is called the feed line. It can be plotted from xF and q alone. It intersects the diagonal at x = xF.

41

The slope and position of the feed line depend upon the thermal condition of the feed as described by the parameter q. The line will have positive slope and lie to the right of the vertical for cold feed. The line will be vertical for saturated liquid and horizontal for saturated vapor. For a mixed vapor-liquid feed, it will lie between the horizontal and vertical (negative slope). Superheated feed will produce a line below the horizontal. The rectifying and stripping lines intersect on the feed line. If the column has an intermediate feed or product, the same rules apply. A feed/product line, depending only on the composition of the stream and its thermal condition, can be constructed to serve as the set of possible intersections of the operating equations for the regions above and below the sidestream. References: 1. Foust, A.S. et al., Principles of Unit Operations, 2nd Edition, John Wiley, 1980, pp. 13-14. 2. McCabe, W.L., J.C. Smith, P. Harriott, Unit Operations of Chemical Engineering, 5th Edition, McGraw-Hill, 1993.

42

10. Distillation IV: Calculations

Before beginning most distillation calculations, a decision must be reached: does equimolal overflow apply? If so, the operating equations are lines and you have one set of options notably the McCabe-Thiele method. If not, energy balances must be explicitly considered. There are several ways of incorporating the energy effects. The Ponchon-Savarit method is a graphical approach that does not require an assumption of equimolal overflow. Graphical construction is done on the enthalpy composition diagram. Your text does not include this method, but it may be useful. In all cases, one can use a "stepping" approach. Starting from one end of the column, the component, material, and energy balances can be solved simultaneously. After a stage is determined, you step up (or down) to the next and calculate that stage. Depending on what information is known, the form of the equilibrium relation, etc., this approach may require an iterative solution.

10.1. McCabe-Thiele method

The graphical McCabe-Thiele Method can be used to determine the number of ideal stages and feed tray location. To do this, you make a plot showing the equilibrium curve, feed line, and operating lines for the rectifying and stripping sections (all on the same axes), and then find answers by graphical construction. A standard (beginner) distillation problem provides you with xF, q, xD, xB, and RD; although not necessarily directly. You may need to use the overall material balances to find some of the compositions; to calculate the q factor from temperature and composition data; and/or determine the reflux ratio based on the limiting minimum value. Once you have these values, the solution procedure is: 1. 2. 3. 4. 5. 6. Plot the equilibrium curve. Calculate the slope q/(q-1) of the feed line. Plot the feed line using x = xF and the slope. Calculate the y-intercept of the rectifying line xD/(RD+1). Plot the rectifying line using (xd, xd) and the intercept. Draw the stripping line connecting the intersection of the rectifying and feed lines and the point (xb, xb). 7. Correct for stage efficiency by drawing the "effective" equilibrium curve between the equilibrium curve and the operating lines. 8. "Step off" equilbrium stages. The number of stages is found graphically by constructing triangles on the diagram. You can start from either the top or the bottom. From a product composition on the operating line (either (xd, xd) or (xb, xb)) move horizontally to the equilibrium curve, then vertically back to the operating line, horizontally to the equilibrium curve, etc., constructing triangles along the way. Always make the steps between the equilibrium curve and the lower of the two operating lines (the rectifying line in the top half of the column, the stripping line in the bottom half if you're looking for the bottom line, the switchover point will be clear). The number of triangles you draw is the number of stages in your column. If you are using the equilibrium curve to step from, you are determining ideal stages. 43

If you don't want to draw, you can do the same thing, iteratively solving for the various equation intersections. You'll need to be pretty careful with your "bookkeeping" if you try this. To apply Murphree tray efficiencies, construct an effective equilibrium curve between the equilibrium and operating curves, and step using the effective curve to determine actual separation stages. Remember that a partial reboiler or partial condenser is by definition an ideal stage, so you use the ideal equilibrium curve (not the effective) for these stages. The optimum feed tray is the triangle with one corner on the rectifying line and one on the stripping line. Putting the feed anywhere else increases the number of stages needed to make the separation. To visualize this, notice that the closer the operating line is to the equilibrium curve, the smaller the stepping triangles become. Introducing the feed at the intersection of the rectifying and stripping lines maximizes the size of the triangles and so leads to the fewest steps. When analyzing existing distillation systems, the actual feed entry point may not be at the optimum (the tray where the operating lines intersect). In this case, the tray stepping should switch from the rectifying line to the operating line at the actual feed tray location.

10.2. Limiting cases

Frequently, when analyzing or designing a process, it is useful to look at limiting cases to assess the possible values of process parameters. In distillation analysis, separation of a pair of components can be improved by increasing the number of stages while holding reflux constant, or by increasing the reflux flow for a given number of stages. This tradeoff sets up two limiting cases: 1. Total reflux (minimum ideal stages) 2. Minimum reflux (infinite ideal stages) The design tradeoff between reflux and stages is the standard economic optimization problem chemical engineers always face balancing capital costs (the number of trays to be built) vs. the operating cost (the amount of reflux to be recirculated). A good design will operate near a cost optimum reflux ratio. 10.2.1. Total reflux The total reflux condition represents operation with no product removal. All the overhead vapor is condensed and returned as reflux. Consequently, the reflux ratio (L/D) is infinite. This, in turn, makes the operating lines the 45 degree line (prove it to yourself by setting D = 0, and noticing that consequently L = V). With the operating lines on the diagonal, they are as far as they can get from the equilibrium curve, so if the number of plates are stepped off using the diagonal and the equilibrium curve, the number of theoretical stages will be a minimum. Often, columns are operated at total reflux during their initial startup, and product is not withdrawn until a separation close to that desired is achieved. The Fenske Equation is another method for determining the minimum number of trays required for a given separation. It is an example of a "shortcut" distillation method. There are

44

a number of these approximate methods available to get initial estimates of distillation requirements. The Fenske equation applies to distillation systems with constant relative volatility. Note that the form of the Fenske equation shown calculates the minimum number of plates; it does not include the reboiler (hence the -1 on the right hand side). Other texts may use a form for the minimum number of stages and not subtract the reboiler.

If the relative volatility varies through a column because of temperature effects, it is possible to use a geometric mean value of the relative volatility (as is done for multicomponent distillation) and the Fenske equation to get an approximate value for the number of stages. 10.2.2. Pinch points The intersection of an operating line and the equilibrium curve is called a pinch point. A simple column will have two pinch points (because there are two operating lines). The points change when the operating lines do. An existing column can "pinch" if its operating line is too close to its equilibrium curve. This means that there are several stages doing very little separation and wasting resources. To cure a pinch, the most direct solution is to move the feed entry point. This is often an expensive proposition. In such cases, the reflux and boilup ratios can be increased to change the operating lines. This will increase operating costs and energy consumption, but may be the only realistic option. A pinch at the intersection of the feed line and the equilibrium curve indicates that the column is operating at minimum reflux. 10.2.3. Minimum reflux The minimum reflux condition represents the theoretical opposite of total reflux -- an infinite number of ideal separation stages. In this case, the intersection of the operating lines lies on the equilibrium curve itself. Thus, the distance between the equilibrium curve and the operating lines is at its minimum, the stepping triangles become very small, there is no gap between the equilibrium curve and the intersection point, so you cannot step past the feed point. The minimum reflux rate can be determined mathematically from the endpoints of the rectifying line at minimum reflux the overhead product composition point (xD, xD) and the point of intersection of the feed line and equilibrium curve (x', y').

45

The derivation of this formula is given later. One important thing to realize: the formula only applies when the feed line is the breakpoint for the operating curve in the top portion of the column. If there are intermediate product draws between the reflux and the feed, the formula does not apply. In this case, you must calculate the liquid flow down the column at the pinch point, and then work it back up the column to find the reflux flow at minimum reflux conditions. If the equilibrium curve has an inflection point, it may not be possible to construct a line between the overhead product point and the feed/equilibrium intersection without passing outside the equilibrium envelope. Operating curves must always intersect within the equilibrium envelope and cannot cross outside (in either half of the column). In this case, minimum reflux occurs at a tangent pinch and the operating line is tangent to the equilibrium curve. Calculations are based upon the intersection of the tangent operating line and the feed point. When designing columns, it is common to define the design reflux ratio as some multiple of the theoretical minimum reflux. The cost optimum reflux ratio is typically in the 1.1 to 1.5 range depending on energy costs, condenser coolant, and materials of construction. The rule of thumb reported most often suggests that a reflux ratio of about 1.2 times the minimum is a good design value. It may also be necessary to distinguish between returned reflux (the reflux stream flowing from the accumulator to the column) and the effective reflux flowing down the column. This is a concern if the reflux is subcooled. In this case, the effective reflux will consist of the returned reflux plus whatever additional liquid is condensed when the cold liquid contacts vapor on the reflux tray. It is probably best to use effective reflux in minumum reflux calculations. 10.2.4. Derivation of minimum reflux formula Minimum reflux corresponds to a pinch at the intersection of the feed line and the equilibrium curve. From its formula, the rectifying line has slope R D/(RD + 1) and will connect the intersection point (xint, yint) and (xD, xD). Consequently, we can express the slope in terms of "rise over run", or (xD yint)/(xD xint). Algebraic rearrangement gives the desired formula:

46

Clearly, this formula doesn't apply if there are more than two operating regions. In these cases, it is probably smarter to calculate the reflux ratio from the ratio of the liquid and vapor flow rates.

10.3. Condenser and reboiler loads

The heating and cooling loads in the condenser and reboiler can be calculated from straighforward energy balances. The chief difficulty is in getting good values for the heats of vaporization an enthalpy-concentration diagram is very useful! For the reboiler, energy must be added equal to the sum of the sensible heat needed to raise the liquid to its boiling point and the latent heat of vaporization. The steady-state energy balance on the process side of the reboiler is then:

This equation is expressed in terms of average heat capacities and an average latent heat and depends on the vapor boilup rate. Usually, the sensible heat transfer in a reboiler is relatively small, so that the heat load can be calculated from

The heating medium requirements can be calculated from an energy balance on the heating side of the reboiler. If saturated steam is the heating medium, then

so that (neglecting thermal capacitance in the reboiler and heat losses) the steam rate can be obtained from

Similarly, if a liquid heat transfer fluid (hot oil, etc.) is used, the equations

(htfs = heat transfer fluid supply, htfr = heat transfer fluid return) or

47

may be used. A similar analysis provides the condenser load as

(neglecting any subcooling of reflux), so that when cooling water is used in the condenser

(cws = cooling water source, cwr = cooling water return) and since the specific heat of water is 1.0 for common units

When calculating cooling loads, you may need to adjust the vapor rate or condenser duty to account for vapor condensed by direct contact with cold reflux on the reflux return tray. Watch for this whenever the condenser temperature is significantly below the expected tray temperature. It is possible to imagine a case where QB QC; that is, when V1 VB. The case when the vapor rates at the top and bottom of the column are most likely to be the same occurs when the feed is at its bubble point, and q = 1. In this situation, the condenser and reboiler loads will be approximately equal.

10.4. Stage efficiencies

Mass transfer limitations prevent the vapor leaving a tray from truly being in precise equilibrium with the liquid on the tray; consequently, the assumption of ideal stages is only an approximation. An efficiency is used to represent the deviation from equilibrium. There are three types of efficiencies we will consider: 1. Overall efficiency 2. Local efficiency 3. Murphree efficiency An overall efficiency is the simplest choice. It is the ratio of the number of ideal stages to the number of actual stages.

48

A single efficiency can thus be used for the entire column, but is only accurate enough for prelimary design. Some improvement can be achieved by using separate efficiencies for each section of the column. Accuracy is limited because effectiveness of mass transfer is constrained by geometry and design of the trays, flowrates and paths of all streams, compositions, etc. The problem is really too complex to lump into a single parameter, so when overall efficiencies are used they should be based on performance data from similar columns or laboratory tests. The local efficiency is the most accurate option, but also the most difficult to use. It is defined at only a single point on a specific tray.

It is most necessary on large diameter columns where position dependence is significant. A Murphree efficiency is probably the most common choice, since it represents a workable compromise between accuracy and ease of use. It has the same form as a local efficiency but is based on tray average compositions.

Values between 0.6 and 0.75 are common for sieve trays. We know that the liquid leaving a tray is not really the same as the tray average, so a Murphree efficiency effectively assumes perfect mixing on the tray. In practice, we normally measure the liquid composition and get the vapor composition from an equilibrium calculation or diagram. In the case of multicomponent systems, the efficiencies are different for each component. The overall efficiency and the Murphree efficiency are not directly related. You cannot use an average Murphree efficiency in place of an overall value. To use a local or Murphree efficiency with a graphical method, the true equilibrium curve is replaced with an effective equilibrium curve located between the true curve and the operating curves. The effective curve is used to count stages. Note, however, that the efficiency doesn't apply to the reboiler, so the true equilibrium curve should be used for the last stage of the stripping section. To construct an effective equilibrium curve is not difficult. The effective curve is given by:

49

References: 1. Foust, A.S. et al., Principles of Unit Operations, 2nd Edition, John Wiley, 1980, pp. 13-14. 2. McCabe, W.L., J.C. Smith, P. Harriott, Unit Operations of Chemical Engineering, 5th Edition, McGraw-Hill, 1993. 3. Seader, J.D. and E.J. Henley, Separation Process Principles, John Wiley, 1998, pp. 292-294, 299.

50

11. Distillation V: Enthalpy balances

Variations in internal vapor and liquid flows inside a distillation column depend on the enthalpies of the mixtures. Equimolal overflow assumes the rates are independent of the energy balance. To remove the assumption we must consider the enthalpy balances explicitly. The overall enthalpy balance on a distillation column with NT trays is:

where the Q terms are the condenser and reboiler loads. This equation reduces at steady-state to:

For a given feed and products, only one of QC and QR may be independently set. Normally, QC is chosen to get the desired operating pressure, product and reflux rates, and QR found by balance. In operation, however, the reboiler duty is usually varied to control an inventory or product composition.

11.1. Rectifying section

Balances on the rectifying section are:

These can be combined with the steady state balance on the accumulator:

to eliminate the condenser duty:

In cases where constant molar overflow is not valid, this equation may be used to calculate vapor flowrates. 51

11.2. Stripping section

The balance for the stripping section:

is combined with the steady state balance on the reboiler:

in order to obtain:

11.3. Calculations
In order to solve problems when constant molar overflow does not apply, one must simultaneously solve the set of equations consisting of: Component balance

Overall balance

Enthalpy balance

Thermodynamic property equations

Equilbrium equations

52

These equations are linked with enough complications, that an iterative solution is typically required. An alternative to iterative solution is to solve the system graphically using an enthalpy concentration method. References: 1. Foust, A.S. et al., Principles of Unit Operations, 2nd Edition, John Wiley, 1980, pp. 13-14. 2. McCabe, W.L., J.C. Smith, P. Harriott, Unit Operations of Chemical Engineering, 5th Edition, McGraw-Hill, 1993.

53

54

12. Distillation VI: Enthalpy-concentration methods

Distillation calculations can be performed graphically on an enthalpy-concentration (Hx) diagram. This approach is sometimes called the Ponchon-Savarit Method. Working on the Hx diagram is more general than a McCabe-Thiele construction, because it takes direct account of the thermal effects and does not require an assumption of equimolal overflow. The very shape of the Hx diagram provides a clue as to the importance of the energy balances. If the dew point and bubble point lines are more or less straight and roughly parallel, it indicates that the latent heat of vaporization is basically constant with respect to composition. This is the prerequisite for assuming equimolar overflow, and so the energy balances may be neglected. If the saturation curves show significant changes in curvature or separation, it suggests that to assume equimolar overflow will introduce error. Points for the feed and product can be located on the Hx diagram; for our purposes we'll call them the F, D, and B points. The coordinates are their composition and enthalpy. If the products are saturated liquid, as is the case for total condensers without subcooling and for partial reboilers, these points will lie on the bubble point curve on the Hx diagram.

12.1. Overall enthalpy balance

The steady-state enthalpy balance for a distillation column is:

One of the tools of graphical solution is the notion of colinearity. This has been used before if you have used lever arm principles. For an adiabatic process, the feed and products will be colinear on an Hx diagram. Thus, it is useful to redefine our distillation system to be adiabatic, by bringing the condenser and reboiler inside the system boundary. Rearranging the enthalpy balance gives: Q Q FhF = D hD + C + B hB R D B If we then define:

The enthalpy balance becomes:

Substituting from the overall material balance: 55

After this modification, the system is adiabatic, so a line can be drawn through the feed point, F, and the points (xD, hDp) and (xB, hBp). This line represents the system enthalpy balance, and so is called the overall enthalpy line. Remember: for a given separation only one of the reboiler and condenser duties is independent. So, you will probably want to pick one duty and then construct the line through the feed point to determine the other duty.

12.2. Reflux ratio

As before, the reflux ratio can be determined from the L/V ratio, and for this formulation is given by:

where hD is the enthalpy of the overhead product, Hy1 the enthalpy of the vapor entering the condenser, and hDp the adjusted enthalpy of the overhead. Notice that this represents the ratio of distances on the Hx diagram: the numerator is the vertical distance between the h Dp point and the dew point saturation curve, while the denominator is the distance between the saturation curves. A calculation often begins by using the overhead product composition and temperature to obtain hDand Hy1. These in turn are used with the reflux ratio to get hDp. Then the overall enthalpy line is drawn from hDp through the feed point, and the intersection with xB gives hBp.

12.3. Stepping off stages

In the McCabe-Thiele procedure, operating curves were constructed to represent the component balances for the column and to relate the liquid composition on a stage to the composition of the entering vapor. These were then paired with the equilibrium curve, which relates the composition of liquid on a stage to the vapor leaving the stage. A similar procedure can be followed on an Hx diagram, except instead of using component balances, enthalpy balances are used. The equilibrium data is represented by the equilibrium tie lines on the Hx diagram. The operating lines are developed from enthalpy balances on the rectifying and stripping sections (just as in the McCabe-Thiele approach operating equations were developed from the equivalent component balances). The operating lines will connect the point representing the liquid on a stage (with coordinates xn, hn) to the point representing the adjusted enthalpy at the appropriate end of the column. It will cross the saturated vapor line on the Hx diagram at the point corresponding to the vapor leaving the stage (coordinates yn+1, Hn+1).

56

12.3.1. Summary of procedure 1. Obtain enthalpy-composition diagram. 2. Fix the feed point F, and product points D and B using stream compositions and enthalpies. 3. Use the overhead product enthalpy and the reflux ratio to find the adjusted enthalpy of the overhead. Plot it as point D', on a vertical line with point D. 4. Construct the overall enthalpy line from point D' through the feed point. It intersects a vertical line drawn through point B at point B'. 5. Plot point V1. For a total condenser, the composition entering the condenser is the same as the overhead product, so this point will be vertically above point D on the saturated vapor curve. 6. Follow the tie line from point V1 to the saturated liquid curve. This intersection will be point L1. 7. Construct an operating line connecting points D' and L1. The intersection of the operating line with the saturated vapor curve will be point V2. 8. Repeat the two preceding steps until one of the V or L points is to the left of the overall enthalpy line. Once it is crossed, construct operating lines using points Li and B'. 9. When xi is less than xB, construction is finished. The number of stages can be read as the number of complete triangles.

12.4. Limiting conditions

At total reflux, operating lines are vertical (infinite slope). This can be used to determine the minimum number of stages. Not that operating curves are not required to do this only the endpoint compositions. While doing constructions on the yx diagram, a "pinch" was defined as the intersection point between the equilibrium curve and the operating curve. On an Hx diagram, there isn't an equilibrium curve it has expanded to a region, and each point from the xy equilibrium curve is represented by a tie line. The "pinch point" also expands, resulting in a single line where the operating and tie lines overlap. Minimum reflux still corresponds to a pinch at the feed conditions, so to determine the minimum reflux a line must be constructed so that the overall enthalpy line coincides with the tie line that runs through the feed point. Example: A feed containing 40 mole percent n-hexane and 60 percent n-octane is fed to a distillation column. A reflux ratio of 1.2 is maintained. The overhead product is 95 percent hexane and the bottoms 10 percent hexane. Find the number of theoretical stages and the optimum feed stage. Assume that a total condenser is used. The column is to operate at 1 atm. Step 1: Equilibrium data is collected. VLE data, Mole fraction hexane, 1 atm x (liquid) 0.0 0.1 0.3 0.5 0.55 0.7 1.0 y (vapor) 0.0 0.36 0.70 0.85 0.90 0.95 1.0

57

Enthalpy-concentration data Mole fraction hexane Enthalpy cal/gmol Sat. liquid Sat. vapor 0.0 7000 15,700 0.1 6300 15,400 0.3 5000 14,700 0.5 4100 13,900 0.7 3400 12,900 0.9 3100 11,600 1.0 3000 10,000 This information can be used to create enthalpy-concentration and equilibrium diagrams.

Step 2: Plot the feed and product points. All three will lie on the saturated liquid line, B at xB = 0.1, F at xF = 0.4, and D at xD = 0.95. Step 3: From the data tables (or from the Hx diagram) find the enthalpy of the distillate, h D = 3050 cal/gmol. Because a total condenser is used, the vapor leaving the top stage will have concentration y1 = 0.95. Consequently, it will have enthalpy HD = 10,800. These values and the reflux ratio can be used to find the enthalpy coordinate for the D' point, HDp.

and so HDp = 20,100. The D' point on the Hx diagram can then be placed at (0.95, 20100). Step 4: The overall enthalpy line is then drawn from D', through F. Its intersection with the line x = 0.1 is the point B'. Step 5: Because of the total condenser, the V1 point will lie on the saturated vapor curve at x = xD.

58

Step 6: Follow the tie line that passes through the V1 point back to the saturated liquid curve. The intersection is the point L1 (liquid on tray 1). If tie lines are not available, they may be constructed using the xy diagram.

59

Step 7: Construct an operating line through both the L1 point and the D' point. Its intersection with the saturated vapor curve will be the point V2. Step 8: Continue the construction, alternating between tie lines and and operating lines until you have moved to the left of the overall enthalpy line at point L3.

60

Once the overall enthalpy line is crossed, construction continues, but operating lines are now drawn between the Li point and the B' point (instead of the D' point). Construction continues until L5 which is almost directly on top of xB.

Consequently, there are 5 ideal stages required for this separation with a partial reboiler, that means 4 ideal trays. The optimum feed tray is number 3. References: 1. Foust, A.S. et al., Principles of Unit Operations, 2nd Edition, John Wiley, 1980, pp. 13-14. 2. McCabe, W.L., J.C. Smith, P. Harriott, Unit Operations of Chemical Engineering, 5th Edition, McGraw-Hill, 1993. 3. McCabe, W.L., and J.C. Smith, P. Harriott, Unit Operations of Chemical Engineering, 3rd Edition, McGraw-Hill, 1976. pp. 571-579. 4. Treybal, R.E., Mass-Transfer Operations, 3rd Edition (Reissue), McGraw-Hill, 1987.

61

62

13. Distillation VII: Equipment and column sizing

In order to have stable operation in a distillation column, the vapor and liquid flows must be managed. Requirements are: vapor should flow only through the open regions of the tray between the downcomers liquid should flow only through the downcomers liquid should not weep through tray perforations liquid should not be carried up the column entrained in the vapor vapor should not be carried down the column in the liquid vapor should not bubble up through the downcomers

These requirements can be met if the column is properly sized and the tray layouts correctly determined. Tray layout and column internal design is quite specialized, so final designs are usually done by specialists; however, it is common for preliminary designs to be done by ordinarily superhuman process engineers. These notes are intended to give you an overview of how this can be done, so that it won't be a complete mystery when you have to do it for your design project. Basically in order to get a preliminary sizing for you column, you need to obtain values for the tray efficiency the column diameter the pressure drop the column height

13.1. Tray construction and hydraulics

Three main types of trays are to be discussed: Bubble cap trays Sieve trays Valve trays

Typically, the liquid flow between trays is governed by a weir on each tray. The flow depends on the length of the weir and how high the liquid level on the tray is above the weir. The Francis weir equation is one example of how the flow off a tray may be modeled.

63

13.2. Tray efficiency

Ideally, tray efficiencies are determined by measurements of the performance of actual trays separating the materials of interest; however, this is usually not practical in the early phases of a design. Consequently, some form of estimation is required. Estimates can be based on theory or on data collected from other columns.

The O'Connell correlation is based on data collected from actual columns. It is based on bubble cap trays and is conservative for sieve and valve trays. It correlates the overall efficiency of the column with the product of the feed viscosity and the relative volatility of the key component in the mixture. These properties should be determined at the arithmetic mean of the column top and bottom temperatures. A fit of the data has been determined:

This, or a similar data set, can be used to get preliminary estimates of efficiency numbers.

13.3. Column diameter

Column diameter is found based on the constraints imposed by flooding. The number of ideal stages isn't needed to find the diameter only the vapor and liquid loads. You do need the number of actual stages to get the column height. Before beginning a diameter calculation, you want to know the vapor and liquid rates throughout the column. You then do a diameter calculation for each point where the loading 64

might be an extreme: the top and bottom trays; above and below feeds, sidedraws, or heat addition or removal; and any other places where you suspect peak loads. Once you've calculated these diameters, you select one to use for the column, then check it to make sure it will work. Some columns will have two sections with different diameters consider this possibility if you end up with regions where the estimated diameter varies by 20% or more, but realize it will be more expensive than a column that is the same all the way up. One issue that ought to be considered is the validity of your design numbers. If you are following the "traditional" approach, you've probably designed your column for reflux rates in the range of 1.1 to 1.2 times the minimum. This may not give you a column that can handle "upsets" well, so you may want to design for a capacity slightly greater than that increasing the flows by about 20% might be wise. 13.3.1. Flooding Downcomer flooding occurs when liquid backs up on a tray because the downcomer area is two small. This is not usually a problem. More worrisome is entrainment flooding, caused by too much liquid being carried up the column by the vapor stream. A number of correlations and techniques exist for calculating the flooding velocity; from this, the active area of the column is calculated so that the actual velocity can be kept to no more than 80-85% of flood; values down to 60% are sometimes used. A force balance can be made on droplets entrained by the vapor stream (which can lead to entrainment flooding). This balance yields an expression relating the vapor and liquid densities and a capacity factor (C, with velocity units) to the flooding velocity:

13.3.2. Capacity factors The capacity factor can be determined from theory (it depends on droplet diameter, drag coefficient, etc.), but is usually obtained from correlations based on experimental data from distillation tray tests. Depending on the correlation used, C may include the effects of surface tension, tendency to foam, and other parameters. A common correlation is one proposed by Fair in the late 50s - early 60s. The version for sieve trays is available in a wide range of sources (including Figure 21.28 of Ref. [6]). The correlation takes the form of a plot of a capacity factor (which must be corrected for surface tension) vs. a functional group based on the liquid to vapor mass ratio:

65

Enter the plot from the bottom with this number, and then read the capacity factor from the left. This capacity factor applies to nonfoaming systems and trays meeting certain hole and weir size restrictions. It will need to be corrected for surface tension:

where the surface tension is in dynes/cm. Other correlations for the capacity factor are also available. Several are based on more recent information, and may well be more accurate than the Fair plot; however, they also tend to be less broadly known and often require more a priori information on the system. You should use a correlation that is acceptable for your problem. 13.3.3. Diameter Once you have the capacity factor, you can readily solve for the flooding velocity:

(this solution is for the Fair correlation, and adds the surface tension correction). We know that flow = velocity area, so we can calculate the flow area from the known vapor flow rate and the desired velocity (a fraction of flood). This area needs to be increased to account for the downcomer area which is unavailable for mass transfer. The resulting tray area can then be used to calculate the column diameter. So, with everything lumped together, we have:

The only "new" term is the ratio of downcomer area to tray area. This should probably never be less than 0.1, and probably seldom will be greater than 0.2. Trays probably aren't a good idea for columns less than about 1.5 ft in diameter (you can't work on them) these are normally packed. Packing is less desirable for large diameter columns (over about 5 ft in diameter).

66

13.4. Pressure drop

There is a pressure gradient through the column otherwise the vapor wouldn't flow. This gradient is normally expressed in terms of a pressure drop per tray, usually on the order of 0.10 psi. The best source of pressure drop information is to measure the actual drop between trays, but this isn't always feasible at the beginning of a design. Detailed calculations are possible, but these depend so much on the actual tray specifications that final values are usually obtained from experts, but approximate methods can be used to get values to put in your design basis. There are two main components to the pressure drop: the "dry tray" drop caused by restrictions to vapor flow imposed by the holes and slots in the trays and the head of the liquid that the vapor must flow through.

13.4.1. Dry tray losses The dry tray head loss can be related to an orifice flow equation:

This equation determines the dry tray drop in inches of fluid (your text has a similar equation in SI units). The constant 0.186 takes care of the units and is appropriate for sieve trays. The orifice size coefficient Co depends on the tray configuration and will usually fall between 0.65 and 0.85. The hole velocity can be obtained by dividing the vapor flow rate by the total hole area of the tray. 13.4.2. Liquid losses The liquid head pressure drop includes the effects of surface tension and of the frothing on the tray. It is typically represented as the product of an aeration factor and the height of liquid on the tray:

Correlations are available for the aeration factor ; a value of 0.6 is good for a wide variety of situations. The height of liquid on the tray is the sum of the weir height and the height of liquid over the weir. The total height can be calculated directly from the volume of liquid on the tray and its active area. Another approach is to back the height out of a version of the Francis weir equation (which relates flow off a tray to liquid height and weir length). One version, for a straight weir, in units of inches and gal/min is:

67

Realize that these equations depend on the size and shape of the weir.

13.5. Column height

The height of a trayed column is calculated by multiplying the number of (actual) stages by the tray separation. Tray spacing can be determined as a cost optimum, but is usually set by mechanical factors. The most common tray spacing in 24 inches it allows enough space to work on the trays whenever the column is big enough around (>5 ft diameter) that workers must crawl inside. Smaller diameter columns may be able to get by with 18 inch tray spacings. In addition to the space occupied by the trays, height is needed at the top and bottom of the column. Space at the top typically an additional 5 to 10 ft is needed to allow for disengaging space. The bottom of the tower must be tall enough to serve as a liquid reservoir. Depending on your boss's feelings about keeping inventory in the column, you will probably design the base for about 5 minutes of holdup, so that the total material entering the base can be contained for at least 5 minutes before reaching the bottom tray. The total of height added to the top and bottom will usually amount to about 15% or so added to that required by the trays. You rarely will see a real tower that is more than about 175 ft. tall. Tall, skinny towers are not a good idea, so watch the height/diameter ratio. You generally want to keep it less than 20 or 30. If your tower ends up exceeding these values, you probably want to look at a redesign, maybe by reducing the tray spacing, or splitting the tower into two parts. References: 1. Foust, A.S. et al., Principles of Unit Operations, 2nd Edition, John Wiley, 1980, pp. 13-14. 2. McCabe, W.L., J.C. Smith, P. Harriott, Unit Operations of Chemical Engineering, 5th Edition, McGraw-Hill, 1993. 3. Douglas, James M., Conceptual Design of Chemical Processes, McGraw-Hill, 1988, pp. 453-457. 4. Kister, Henry Z., Distillation Design, McGraw-Hill, 1992, pp. 275-282. 5. Luyben, William L., "Introduction" in Practical Distillation Control (W.L. Luyben, ed.), Van Nostrand Reinhold, 1992, pp. 10-11. 6. McCabe, W.L., J.C. Smith, P. Harriott, Unit Operations of Chemical Engineering, 5th Edition, McGraw-Hill, 1993, pp. 560-568. 7. Seader, J.D. and Ernest J. Henley, Separation Process Principles, John Wiley, 1998, pp. 305-312.

68

14. Batch distillation

In batch distillation, a tank is charged with feed and then heated. Vapor flows overhead, is condensed and collected in a receiver. The liquid remaining in the tank is generally called the residue. The composition of the material collected in the receiver varies with time, so the composition of the product is an average of all the material collected. Often, the receiver will be emptied or switched several times during a distillation to collect separate cuts of product.

A batch process is inherently dynamic it cannot be modeled steady state. Batch distillation can be conducted with or without reflux. When reflux is used, any of several different operating policies may be used you might use a constant reflux rate, you might vary it, etc. Batch distillation is most common: in small capacity plants when feed or products vary widely and frequently (as in specialty chemical production) for test runs on new products when the feed is the result of batch processing when the process requires frequent cleaning which would interrupt continuous processing

We want to consider two main variants with and without reflux.

14.1. "Differential" distillation

Batch distillation without reflux is often called differential distillation. Because there is no reflux, the vapor product is in equilibrium with the liquid residue in the tank at any given time. There is one product stream, D, leaving the system, so the total material balance is:

69

where n(t) is the total moles of material in the distillation vessel. For a binary system, the component balance is:

This may be expanded via the chain rule

combined with the total material balance

multiplied through by dt

and rearranged to obtain

Integrating yields

This is the Rayleigh Equation (Ref. [2] Eq. 21.75) which relates the amount of residue to the composition. Any of the four variables n0, n1, x0, x1 can be found, provided the others are known. If there is no reflux, there is a single equilibrium stage, and yD is in equilibrium with xD (yD=yequilibrium(xD)). The right hand side integral of the Rayleigh equation can usually only be evaluated for constant pressure systems. Often, it is necessary to solve the integral numerically or graphically. The latter is done by making a plot of 1/(yD x) vs. x and finding the area under the curve between the initial and final concentrations.

70

The Rayleigh equation can also be used for any two components of a multicomponent system. Determination of the residue amounts or compositions using the Rayleigh Equation is one of the more common batch distillation calculations. Another is to determine the average overhead product composition. The composition of the overhead varies with time, so the average composition is that of a mixing tank that collects all of the distillate. It can be found from material balance to be:

14.1.1. Special case applications of the Rayleigh equation Consider the case where equilibrium constants are available and the K values are relatively constant (as for close-boiling mixtures with limited T). Then, the equilibrium relationship is y = Kx, and

Another case of interest is that where the relative volatility is constant. Then, for a binary system:

which can be substituted into the Rayleigh equation and integrated to get

which can be useful for some calculations. Your text prefers to do a further rearrangement to get the equation in terms of the two components, A and B.

71

(Ref. [2] Eq. 21.87).

14.2. Batch distillation with reflux

To use the Rayleigh equation when a column is refluxed, you generally must do one or more iterative calculations. Typically, a McCabe-Thiele diagram is used to obtain some of the information needed in the calculation. A key factor in how you approach the problem is the operating policy chosen for the system. We will consider two such policies: vary reflux rate to hold xD constant during the run fix the reflux rate and allow xD to vary

14.2.1. Variable reflux policy (constant xD) Variable reflux is generally more expensive to implement, since it requires composition measurements. Typically, a calculation follows 1. For a known charge (n0, x0), fix the desired product composition xD and the heat input (vapor rate). 2. Determine the number of ideal stages in the system this can be done from a total reflux construction. 3. Use McCabe-Thiele plots, and trial and error to determine the L/V ratio for the system. As the residue composition changes, the operating line will have to change it will rotate through the equilibrium envelope with its top end fixed at xD. You will get a fan shaped array of operating lines. 4. The L/V ratio can be used in subsequent calculations. Two formulas of interest (they will not be derived here) are:

which relates the amount of residue and the various compositions at any time t during the run (xt is the composition and nt the amount of residue at time t), and (Ref. [3] Eq. 13.16)

72

to calculate run length. 14.2.2. Variable composition policy (constant reflux) In other circumstances, it makes sense to operate with a constant reflux rate. In this situation, you fix the molar vapor rate to avoid flooding. The operating curves shift downward in parallel to keep the slope and number of stages constant. The batch time is given by (Ref. [3] Eq. 13.11):

References: 1. McCabe, W.L., J.C. Smith, P. Harriott, Unit Operations of Chemical Engineering, 5th Edition, McGraw-Hill, 1993, pp. 576-580. 2. McCabe, W.L., J.C. Smith, and P. Harriott, Unit Operations of Chemical Engineering (6th ed.), McGraw Hill, 2001, pp. 505-506, 986. 3. Seader, J.D. and E.J. Henley, Separation Process Principles, John Wiley, 1998, pp. 681-691. 4. Treybal, R.E., Mass-Transfer Operations, 3rd Edition (Reissue), McGraw-Hill, 1987, p. 367-371.

73

74

15. Multicomponent distillation

Rigorous computer methods for solving multicomponent distillation problems are available, but the approximate, or "shortcut", methods described here are common for preliminary design, examining the relationships between design parameters, process synthesis, etc. As we did with binary columns, we'll work with ideal stages which can be converted to real stages using an efficiency factor. The limiting cases of total and infinite reflux apply to multicomponent columns just as they do to binary systems. The overall approach to solving multicomponent problems is the same we use for all equilibrium stage systems use the equilibrium relationships and the operating relationships. The equilibrium relationships are more complex for multicomponent systems; in particular, the identity of the most volatile component may change with temperature in the system. You may wish to review multicomponent bubble and dew point calculations. These were likely covered in your thermodynamics and material balance classes.

15.1. Multicomponent flash distillation

For a multicomponent flash distillation, the operating equation

is written for each component in the system. The f factor is the fraction of the feed which vaporizes, just as it was for binary flash distillation. Because a flash is a single equilibrium stage, equilibrium fixes another relationship between y and x. If this is represented by an equation, using distribution coefficients for instance, the two equations (equilibrium and operating) can be combined and rearranged

This expression can be summed over all components to obtain:

75

which can be iteratively solved in the same fashion as a bubble-point or dew-point calculation. 1. 2. 3. 4. assume a flash temperature determine K values at that temperature compute the summation if the summation is not equal to 1.0, adjust the temperature and repeat

The final values of T and the Ks are used to determine the product compositions.

15.2. "Key" components

The first step in setting up multicomponent distillation problems is to select two components to serve as the heavy key and the light key. Key components must be present in both the overhead and the bottoms, so that by specifying the recovery of the keys, you specify the extent of the split. Product specifications for these calculations are generally in the form of recovery specifications for example, of the butane in the feed, 75% exits in the overhead rather than mole fractions, since to calculate mole fractions requires prior knowledge of how all the components distribute between the products. Non-key components (everything but the keys) are classed as distributed if they occur in both products, or non-distributed if they appear in only one product. Remember, keys must be distributed. Non-keys may be distributed when they have volatilities very close to the keys or between the keys, and when the desired separation is sloppy. The Shiras equation can be used to predict component distribution at minimum reflux:

Components are distributed when DR is between zero and one.

15.3. Minimum stages

The Fenske equation can be used to determine the minimum number of stages theoretically necessary to make a given separation at total reflux. The equation can be written for any two components. Typically, you will initially apply the equation to the key components and solve for the number of stages:

76

This is reasonably reliable except when the relative volatility varies a lot or when the liquid mixtures are not ideal. The mean relative volatility is used when alpha isn't constant normally a geometric mean value is best. As a minimum, a three point mean evaluated at the feed and product temperatures, can be used:

Once Nmin has been established, you can use the same equation to determine the splits of the other components in the mixture.

15.4. Minimum reflux

Minimum reflux calculations are based on invariant zones around the feed where the compositions stop changing. These are similar to the pinch point idea used for binary columns. One way to determine an approximate minimum reflux ratio is to represent a multicomponent mixture as a pseudobinary system. This is done by creating a hypothetical feed made up of the two key components only. A McCabe-Thiele construction can then be used to determine the pinch and minimum reflux. The accuracy of the approximation depends heavily on how large a portion of the material is made up of the keys. Considering this limitation, and the amount of work required to recast the calculations, other methods are usually preferred. Often, minimum reflux is estimated using the Underwood method. This assumes equimolar overflow and defines relative volatilities for each component relative to some reference component, usually the heavy key:

The equation

is solved for . Acceptable roots will fall between

and there should be one less root than the number of distributed components. The values are then plugged into

77

where the summation is over all components in the distillate, and solved for the minimum reflux. The approach is fairly straightforward as long as the keys are the only distributed components. If, however, there are distributed non-keys, the second equation must be written for each viable value of . These are then solved simultaneously for xDi and Rmin. It should also be noted that the Rmin determined by this approach is the "internal" or "effective" reflux, not necessarily the "external" or "returned" reflux. The latter must be obtained by enthalpy balance. Fortunately, the difference is sigificant mainly if the mixture is "wide boiling". The Underwood method assumes constant molar overflow and relatively constant relative volatilities. If these do not apply, caution is necessary when applying the results.

15.5. Stages, reflux for a given separation

Gilliland plots can be used to find the actual stages required for a given reflux rate or vice versa. The plots relate two variables:

which ranges from zero at R = Rmin to infinity at N = Nmin, and

which ranges from 1.0 at R = Rmin to zero at N = Nmin. The plots available on either linear or log axes. To use them, calculate either X or Y, read the other from the graph, and get your answer. The Eduljee equation is a numerical fit of the charts:

Although valid for X values between 0.01 and 1.0, it increases the degree of approximation and should only be used when high accuracy is not important. Erbar and Maddox have also prepared plots for determining the number of stages. These show curves of Rmin/(Rmin + 1) on axes of R/(R + 1) vs. Nmin/N.

15.6. Feed tray

The Kirkbride equation allows you to predict the optimal feed stage, by determining the relative number of trays in the rectifying and stripping sections:

78

References: 1. Kister, H.Z., Distillation Design, McGraw Hill, 1992, pp. 79-81, 110-113. 2. McCabe, W.L., J.C. Smith, P. Harriott, Unit Operations of Chemical Engineering, 5th Edition, McGraw-Hill, 1993, pp. 588-610. 3. McCabe, W.L., J.C. Smith, P. Harriott, Unit Operations of Chemical Engineering, 6th Edition, McGraw-Hill, 2001, pp. 717-728, 732-734. 4. Seader, J.D. and E.J. Henley, Separation Process Principles, John Wiley, 1998, pp. 492-514.

79

80

16. Leaching
Leaching is the preferential solution of one or more compounds from a solid mixture by contact with a liquid solvent. The solvent partially dissolves the solid material so that the desired solute can be carried away. Typical users include: the metals industry for removing mineral from ores (acid solvents) the sugar industry for removing sugar from beets (water is solvent) the oilseeds industry for removing oil from soybeans, etc. (hexane or similar organic solvents)

The basic concepts of leaching also apply in the environment, where materials can be leached out by rainwater and carried into the groundwater supply. A simple, everyday example of a leaching process is making your morning coffee.

16.1. Liquid-solid equilibrium

Liquid-solid phase equilibrium is important in understanding leaching, crystallization, and adsorption. Diffusion through solids is slow, even through pores in the substance, and so equilibrium is harder to achieve.

16.2. Principles
Leaching can be batch, semibatch, or continuous. It usually operates at an elevated temperature to increase the solubility of the solute in the solvent. Calculations involve three component (solid, solvent, solute) systems. Feed to a leaching system typically is solid, consisting of basically insoluble carrier material and a (usually desirable) soluble compound. The feed usually must be prepared by grinding or chopping. It is then mixed with a liquid solvent. The desired material dissolves (to some extent) and so leaves when the liquid is drawn off as overflow. The solids are then removed as underflow. The underflow is wet, and so some of the solvent/solute mixture is carried out here as well. Flow through a leaching system may be crosscurrent or countercurrent.

16.3. Modeling assumptions

Modeling a leaching system requires several assumptions to make the system "ideal". The solubility of solute may have an upper bound, limiting how much solute the solvent may hold. Ideally, the carrier will not dissolve, and so will not be present in the overflow. This is a generally safe assumption, although allowance should be made for entrainment of solid flakes, etc., in the overflow from the first stage. Mixing of the solid and the solvent is critical. Typically, "perfect mixing" is assumed, as is the idea of an equilibrium stage (the solid and liquid phases on each stage are in equilibrium).

81

These assumptions imply that all the liquid within the stage has the same composition and so the overflow and the liquid carried in the underflow are identical. This is known as the uniform solution assumption and will result in a linear equilibrium curve. One also needs to determine how much liquid leaves entrained with the solids in the underflow. The simplest option is the assumption of constant solution underflow which means that every stage has the same, fixed ratio of solution to solid in the underflow stream. Like the equimolar overflow assumption in distillation, this will produce a linear operating curve. The first stage is again problematic, since it must "wet" the solid (fill pores, etc.) and so will typically pick up more liquid than subsequent stages. More generally, the amount of solution in the underflow depends on the properties of the solution, which are dependent on its composition. The amount of solute present seems to effect the "stickiness" of the solution. As a result, "draining data" is typically collected. This relates the solution/solid ratio to the composition of the solution and is used to determine the nature of the operating curve.

16.4. Operating equations

As has become customary in this course, we develop the operating equations from the balances describing the system. The steady state material balance over any stage in a countercurrent flow system is written as:

Following McCabe et al. (1993), V is used for the liquid solvent phase and L for the entrained liquid phase. The subscripts indicate the stage in which the stream originates, with a used for the fresh feed and b for the final product. Similarly, the component balance is:

The operating equation is obtained by rearranging the component balance:

which can be shown to run through the points (xa, ya) and (xb, yb). In some circumstances, it may be useful to use the total material balance to eliminate Vn+1 from the equation:

82

In the general operating case, the density and viscosity of the solution change with the solute concentration and the mass retained by the solid phase changes from tank to tank. Consequently, the ratio of entrained liquid to solution varies from stage to stage. In this case, the operating curve is not linear. In the special case of constant solution underflow, the mass retained by the solid does not depend on the concentration and the operating line is linear after the first stage.

16.5. Calculations
Two main types of calculations are usually performed: 1. The extent of leaching is determined, usually by balance calculations. Efficiency depends on contact time and the liquid-solid separation efficacy. The limit on extent is imposed by equilibrium constraints. 2. The number of stages required to reduce the solute content to a specified value is determined. As with distillation, the relationship between the equilibrium and operating lines will be used to determine the number of stages. Solution techniques are algebraic or graphical. Graphical solutions may be set up using the actual compositions or on a "solid-free" basis. Leaching calculations are almost always based on the principle that the solid will not dissolve into the solvent. Thus, the concentrations x of solute in the solvent entrained in the slurry and y of the solute in the solvent liquid phase can be expressed on a solid free basis without significantly changing the calculation. The equilibrium behavior of the system establishes key behaviors. Begin with the assumptions that each stage has enough contact time that the system can reach equilibrium, that there is enough solvent present in each stage to allow equilibrium removal of solute from solvent, and that the solute will not absorb on the solvent. The uniform solution assumption then applies, and the equilibrium x-y curve will be a straight line, xe = ye. Two subsets of the uniform solution case should be noted. In the first, the solute is infinitely soluble in the solvent. In this case, all values of x and y from 0.0 to 1.0 may be obtained. In the second case, the solubility is limited to some maximum value, xs. The x-y diagram will be a straight line, but will not go all the way from 0 to 1. Instead, at xs it will become vertical. If the conditions (contact time, adequate solvent mass, no adsorption) that determine uniform solution do not hold, the x-y diagram may be curved (often, it will look like that used in distillation).

83

16.6. Constant solution underflow

In the case of constant solution underflow, both the equilibrium and operating curves are linear and it is possible to solve for the number of stages directly. The general solution developed using the Kremser equation is:

For leaching, the equilibrium line is xe = ye, so the equation becomes:

To use this, you need to make sure that La = Ln (that the flow of liquid entrained with the solid is the same entering and leaving the first stage constant solution underflow is specified starting with the first stage) or calculate the first stage separately.

16.7. Variable solution underflow

If solution underflow is not constant, the operating equation for leaching is not linear. In this case, an approach similar to the McCabe-Thiele method for distillation columns can be used. 1. Use system component balances to determine xa and xb. 2. Construct operating curve. In general, the operating curve will not intersect the equilibrium curve as it did for distillation. Select an xn somewhere in the middle. This gives an equivalent Ln. Use Ln and a material balance to obtain Vn+1. Find yn+1 by a component balance. Plot (xn, yn+1) as a point on the operating curve. Repeat as needed to obtain the full operating curve. Often a single point is adequate. 3. Operating stages may then be stepped off as triangles between the operating and equilibrium curves in the same fashion as was used for distillation. Normally, you start with the dilute product composition.

16.8. Solid-free calculations

Another approach to solving these problems uses a graphical method similar to the PonchonSavarit method for distillation columns; however, a composition-composition diagram is used for construction rather than an enthalpy-composition diagram. The method assumes that all streams are a mix of solid and solution and that the ratio of solid to solution can be calculated.

84

The working plot is constructed using modified compositions. Define a to be the mass (or concentration) of solute, b the mass of solid (zero for an insoluble carrier), and s the mass of solvent, then calculate: X= a b mass solute/mass solution, and Y = mass insoluble/mass solution a+ s a+ s

and plot Y vs. X. Since there are two phases (overflow and underflow), there will be two curves. The upper curve will be a line if solution underflow is constant. The lower curve will collapse to the X axis (Y=0) when the carrier is completely insoluble. The curves are connected by tielines, based on the relationship between the phases. If the uniform solution condition holds so that xe = ye, these will be vertical; if not, they will be slanted and must be determined. Generally speaking, the equilibrium data must be obtained experimentally. Once the composition diagram is constructed, a "J point" may be determined. This is useful in some, but not all problems. J is defined as the in terms of the total amount of solute entering (or leaving, they're equal at steady state): J = Vb + La or J = Va + Lb Graphically, J is the intersection of the line connecting Va with Lb and that connecting Vb and La. The J point is not used in the actual construction of stages; it can be used for lever arm calculations. The "P point" is used in the constructions. It is given by the equations: P = Va La or P = Vb Lb P represents the "net flow" of material through the system, so that

Effectively, this point pretends that there are mixers on either end of the leaching process, combining both the L and V flows into a single stream. This is thus a fictitious value, and one coordinate must be negative. Operating lines are constructed by connecting an L point with the P point. The intersection with the axis will be the V point for the next stage. Begin construction at Va point. First trace upward to the equilibrium curve along a tie line (remember!, the tie lines are vertical for when the uniform solution conditions hold). This will be the L1 point. Next construct a line by connecting L1 to P. The V2 point will lie at the intersection of this line and the X-axis (the operating curve). Trace up a tie line to L2, construct a line to find V3, etc. Continue to alternate between the tie lines and constructed operating lines until the Vb point has been passed. The number of triangles formed will correspond to the number of leaching stages required.

85

References: 1. McCabe, W.L., J.C. Smith, P. Harriott, Unit Operations of Chemical Engineering, 5th Edition, McGraw-Hill, 1993, pp. 614-623. 2. McCabe, W.L., J.C. Smith, P. Harriott, Unit Operations of Chemical Engineering, 6th Edition, McGraw-Hill, 2001, pp. 742-747. 3. Seader, J.D. and E.J. Henley, Separation Process Principles, John Wiley, 1998, pp. 198-201. 4. Treybal, R.E., Mass-Transfer Operations, 3rd Edition, McGraw-Hill, 1987 (reissue). pp. 717-761.

86

17. Liquid extraction

Liquid extraction produces separation of the constituents of a liquid solution by contact with another insoluble liquid. If the components of the original solution distribute differently between the two liquids, separation will result. The component balances will be essentially identical to those for leaching, but there are two major differences that complicate the calculations: the carrier phase is a liquid, not a solid, so the physical separation techniques will change, and two distinct phases develop, so the simplicity of uniform solution is lost.

Common applications of liquid extraction include: the separation and purification of lube oils, separation of penicillin from fermentation broth, etc. Extraction is driven by chemical differences, not by vapor pressure differences, and so can be used in situations when distillation is impractical. For instance, it can be used to separate materials with similar boiling points (so that distillation is impractical) or mixtures containing temperature sensitive compounds. Distillation and evaporation produce finished products; liquid extraction generally does not. The products are still mixtures, although with new compositions, and these must be separated to obtain final products. Secondary separation often requires distillation or evaporation. The overall process cost thus must be considered when choosing extraction. Extraction may become economical for dilute aqueous solutions when evaporation would require vaporization of very large amounts of water.

17.1. Terminology
Certain terms are commonly used when describing extraction processes. The solution to be extracted is called the feed, the liquid used in contacting is the solvent. The enriched solvent product is the extract and the depleted feed is called the raffinate. Extraction processes may be be single stage, multistage crosscurrent, or countercurrent. Cocurrent extraction offers no advantages over a single stage (convince yourself of this!). This class will primarily be concerned with countercurrent systems.

17.2. Equilibrium
Extraction calculations require an understanding of ternary equilibrium. You probably should refresh your memory on how ternary diagrams are read and used. I would anticipate that you learned this in your material balance course.

87

One new term that may not be familiar is the plait point. This point is located near the top of the two-phase envelope, at the inflection point. It represents a condition where the 3component mixture separates into two phases, but the phases have identical compositions. (Compare this with an azeotropic mixture of liquid and vapor.) There are two main classes of liquid-liquid equilibrium that occur in extraction. A Class I system is the one I expect you are familiar with; it has one immiscible pair of compounds and produces the familiar envelope. Class II mixtures have two pair of immiscible compounds, and so the two-phase envelope crosses the triangular diagram like a bridge. Class I mixtures are the most common and are preferable so if you can pick a solvent to get a Class I, you usually want to do so. Classes can change with temperature, so that is also a concern.

17.3. Solvent selection

88

One of the key decisions when designing an extraction process is the choice of the solvent to be used. Issues include: Selectivity compare the equilibrium ratio of solute in each phase Distribution coefficients y/x at equilibrium; large values preferable Insolubility solvent should not be soluble in carrier liquid Recoverability consider constraints (i.e. azeotropes) Density must be different so that phases can be separated by settling Interfacial tension if too high, liquids will be difficult to mix Chemical reactivity solvent should be inert and stable Viscosity, vapor pressure, freezing point low values make storage easier Safety toxicity, flammability Cost

17.4. Calculations
As with the other separations we discuss, there are two primary calculations: the number of stages needed to make a separation (extent) the amount of solvent needed to make a separation (rate/capacity)

Since LL equilibrium is seldom available in algebraic form, the calculations tend to be iterative or graphical. You have a choice of graphical approaches depending on the type of equilibrium diagram you have available (or choose to construct): A modified McCabe-Thiele approach can be used if y vs. x data is available. The coordinates for the diagram are the mass fraction of solute in the extract phase and the mass fraction in the raffinate for the other. The curve is typically concave downward, begins at the origin and ends with the plait point composition. When one has a convenient equilateral triangle diagram, construction can be done directly on the triangle. Some authors refer to this as the Hunter-Nash method. Rectangular equilibrium diagrams can be constructed. These look a lot like the enthalpy-composition diagrams from distillation calculations or the "solid-free" diagrams used in leaching.

17.4.1. Solvent-to-feed ratio For a given feed mixture, required degree of extraction, operating pressure and temperature, and choice of solvent, there is a minimum solvent-to-feed ratio which corresponds to an infinite number of contact stages. As with the other separations we have studied, this corresponds to a "pinch" between the equilibrium and operating curves at the feed composition. Algebraically, this corresponds to an extract phase in equilibrium with the entering feed. The pinch can also be found graphically on a McCabe-Thiele type construction, minimum solvent ratio corresponds to a pinch (curves intersecting) at the feed composition. During a triangular construction, a feed pinch is represented by the operating line overlapping a tie-line and running through the feed point.

89

A theoretical upper limit or maximum solvent-to-feed ratio also can be determined. If you visualize the ternary diagrams, you notice that if enough solvent is added, the equilibrium curve is crossed and the single phase region is entered. Once this happens, it is impossible to divide the mixture into different phases, hence no separation is achieved. The maximum solvent-to-feed ratio is thus that which puts the mixture on the phase boundary. 17.4.2. McCabe-Thiele method A modified McCabe-Thiele approach is probably the most straightforward graphical technique for solving extraction problems. As always, the main constraint is the equilibrium data. When the data is given in a tabular form, it isn't difficult to construct the needed y (solute in the extract phase) vs. x (solute in the raffinate phase) diagram; however, it is a bit of a chore to pull the points off of a ternary diagram. In the latter case, it may may sense to construct directly on the triangle. Once you have the y vs. x plot, the component and material balances can be used to set the endpoints of the operating curve. Interior points can be found by selecting an intermediate value of x, and calculating the appropriate y (this typically is an iterative calculation). You want to find enough interior points to be sure of the shape of the curve, but shouldn't have to calculate too many of these points. The operating curve that results will typically be curved. For extraction calculations, both the equilibrium and the operating equations will be typically curved. Once the curves are available, they can be "stepped off" into triangles, just as one would expect from McCabe-Thiele. 17.4.3. Construction on ternary diagram Construction on a ternary diagram is a little messier. The diagrams are typically much more crowded and so counting stages is more complicated. You also typically need substantial extra space on the side of the diagram; often you want to tape a spare sheet of paper in place to get the workspace. Pocket rulers end up being too short, so make sure you have a longer (~ 2 ft) straightedge around. The first step is to locate the known endpoints. The fresh solvent point will typically lie on a side of the triangle (it has little or no solute and minimal raffinate). If pure, the fresh solvent point will be at an apex. The feed point will lie on the solute-carrier axis; only rarely will the feed contain solvent. Products result from equilibrium stage separation, so both will lie on the phase envelope. Often, only one of these is given in the problem statement.

90

The fundamental idea of all constructions is that a single line connects points made from "mixing" two streams. The endpoints are thus connected two different ways. First, a segment is drawn connecting the "entering" streams (La, Vb) and one between the "leaving" streams (Lb, Va). These will intersect in the middle of the diagram at a "mixing point", M. Since the two streams leaving an ideal stage, are in equilibrium, this point is related to the equilibrium curve. Lever arm principles apply, so that the M point splits the line segments proportionately to the solvent/feed ratio, so that

An "operating" point, P, is located by connecting the "sides" of the cascade: L a to Va and Lb to Vb. The P point can lie on either side of the triangle, depending on the slope of the tie lines. (This is where that extra piece of paper comes in handy!) All possible operating points must pass through the P point.

With the endpoints and the P point, the stages can be stepped off. Begin at the Va (extract product) point. Trace down a tie- line to the raffinate side of the envelope; the intersection will be at the composition of the stream leaving stage one, so this is the L1 point. Next,

91

construct a line connecting L1 with P and extend the line back across to the extract side of the diagram. This is an operating line, and the intersection is the V2 point. The triangle that has been formed represents one ideal equilibrium stage. This procedure down a tie line, up an operating line is repeated until the feed point is reached/passed. The total number of triangles created is the number of stages. Often the construction gets crowded and may be hard to read.

Problems can be varied by changing the given information. You then will need to think through the relationships (solvent/feed ratio, equilibrium tie-lines, operating lines, etc.) and adapt the procedure. Minimum solvent-to-feed is determined by extending the tie-line that runs through the feed point until it intersects the segment connecting Vb and Lb. This intersection is the point Pmin, an operating point at minimum S/F. The line is then extended back across the envelope to the extract side. The intersection is the point Vmin, corresponding to the extract product at minimum S/F. (Remember that a "pinch" on this type of diagram means that the operating and equilibrium (tie-lines) overlap.) A mixing point "Mmin" can be determined using the feed point, Vb, Lb, and Vmin, and the lever-arm rule will provide a value for the S/F ratio. Maximum solvent-to-feed is found by locating a point "Mmax" on the line connecting the fresh solvent and the feed, at the point where it intersects the extract side of the envelope. S/F is then calculated by lever arm. 17.4.4. Note on tie lines Rarely do you have all the equilibrium tie lines you want. It is thus good to know that there is a fairly easy way of generating additional lines. To do this, you construct a "conjugate curve" from the existing tie lines. Take each endpoint and draw a line from it downward, perpendicular to base of the triangle. The extensions from the raffinate side will intersect those from the extract side, and each pair forms a point on the conjugate curve. The final point is the plait point. When a new tie line is needed, one composition is chosen, a line is traced down to the conjugate curve, and then back up to the envelope on the other side. This is the other end of the tie line.

92

17.4.5. Construction on rectangular equilibrium diagram Graphical solution can be readily done on rectangular equilibrium diagrams, but this method will not be discussed here. This method is usually the first choice only if the equilibrium data is already plotted in rectangular form. References: 1. McCabe, W.L. and J.C. Smith, Unit Operations of Chemical Engineering, 3rd Edition, McGraw-Hill, 1976. pp. 619-637. 2. McCabe, W.L., J.C. Smith, P. Harriott, Unit Operations of Chemical Engineering, 5th Edition, McGraw-Hill, 1993. pp. 623-641. 3. Seader, J.D. and E.J. Henley, Separation Process Principles, John Wiley, 1998, pp. 436-438, 443-448. 4. Treybal, R.E., Mass-Transfer Operations, 3rd Edition, McGraw-Hill, 1987 (reissue). pp. 477-553.

93

94

18. Adsorption
A component can be separated from a mixture if it selectively adsorbs onto a solid surface. This is the basis of the adsorption unit process. Adsorbents are usually porous solids, and adsorption occurs mainly on the pore walls "inside" particles. Examples of adsorbents include: activated carbon (adsorbs organics) silica gel (adsorbs moisture) activated alumina (adsorbs moisture) zeolites and molecular sieves synthetic resins

Ideally, one would be able to construct a continuous countercurrent system, but moving solids is tricky. Instead, most commercial applications use small particles of adsorbent in a fixed bed. Fluid passes down through the bed (down instead of up to avoid fluidization) and components adsorb onto the solid. The steps can be summarized: 1. 2. 3. 4. solute diffuses through the fluid to an area near the solid particle surface solute diffuses into the pores of the particle solute diffuses to the pore wall solute adsorbs to the pore wall surface

Ion exchange is a similar process; however, in this case ions create complexes with the solid instead of adsorbing. When a bed nears saturation, the flow is stopped and the bed is regenerated to cause desorption. The adsorbate can thus be recovered and the adsorbent reused. Regeneration can be accomplished in several ways, and these lead to the "cycle type": Temperature swing Pressure swing Inert/purge stripping Displacement purge

Temperature swing is usually the slowest of these (since the bed has to heat/cool before reuse).

18.1. Adsorption equilibrium

Adsorption equilibrium data is typically plotted in the form of an adsorption isotherm (i.e. at constant temperature) with the mass adsorbed on the y-axis and the mass in the fluid on the xaxis. The shape of the curve is significant and factors heavily into design. "Favorable" isotherms permit higher solid loadings at lower solution concentrations. These tend to start out steep and level out. Isotherms which start out flat are "unfavorable", since they only work well at high concentrations of solute. Usually, as temperature increases the amount adsorbed decreases (permitting thermal regeneration).

95

Several fits have been proposed for isotherms. A linear isotherm seems to work for very dilute solutions, but not for many others. The Freundlich isotherm describes physical adsorption from liquids and can also be used for the adsorption of hydrocarbon gases on activated carbon. It is a two parameter model, of the form

Maybe the best known is the Langmuir isotherm, given by

which assumes that the number of adsorption sites is fixed and that adsorption is reversible.

18.2. Breakthrough curves

Adsorption is a transient process. The amount of material adsorbed within a bed depends both on position and time. Consider the time dependence. As fluid enters the bed, it comes in contact with the first few layers of absorbent. Solute adsorbs, filling up some of the available sites. Soon, the adsorbent near the entrance is saturated and the fluid penetrates farther into the bed before all solute is removed. Thus the active region shifts down through the bed as time goes on. The fluid emerging from the bed will have little or no solute remaining at least until the bulk of the bed becomes saturated. The break point occurs when the concentration of the fluid leaving the bed spikes as unadsorbed solute begins to emerge. The bed has become ineffective. Usually, a breakpoint composition is set to be the maximum amount of solute that can be acceptably lost, typically something between 1 and 5 percent.

96

As the concentration wave moves through the bed, most of the mass transfer is occurring in a fairly small region. This mass transfer zone moves down the bed until it "breaks through". The shape of the mass transfer zone depends on the adsorption isotherm (equilibrium expression), flow rate, and the diffusion characteristics. Usually, the shape must be determined experimentally. The wave front may change shape as it moves through the bed, and the mass transfer zone may broaden or diminish. Unfavorable and linear isotherms tend to broaden. Favorable Langmuir and Freundlich isotherms may broaden at first, but quickly achieve a constant pattern front, an asymptotic shape. This means that the mass transfer zone is constant with respect to both position and time. When dealing with a constant pattern front, one can make measurements on a small scale apparatus and scale-up the results to a full-size adsorber bed.

18.3. Calculations
When scaling up an adsorber, the key design parameter is the length of the bed. The total length is split into the "required length" of an "ideal" fixed bed process and a segment of "unused bed" that is the length leftover at breakthrough. By adding these together, you obtain a bed that can achieve the needed removal, but not waste solute. The diameter of the bed is calculated from the fluid flow rate and the desired cycle time. Usually, superficial velocities on the order of 0.15 to 0.45 m/s are targeted. Capacity calculations are made based on plots of the composition vs. time (usually near the exit of the bed). Curves are integrated (analytically, numerically, or graphically) to obtain capacities (measured in time units, or how long a bed can run).

97

The time required for a bed to become totally saturated is obtained by integrating as time goes to infinity:

In operation, you want to stop the process before solute breaks through, so integration to the breakpoint time gives the "usable" capacity:

Most of the time the breakthrough time is very close to the time elapsed at usable capacity. The capacity times are directly related to bed length:

The unused height can also be readily measured by experiment. The total design height of a bed is determined by adding the required usable capacity to the unused height

Simple ratios allow direct scaleup to a new bed size. References: 1. Geankoplis, Christie J., Transport Processes and Unit Operations (3rd Edition), Prentice-Hall, 1993, pp. 697-704. 2. McCabe, W.L., J.C. Smith, and P. Harriott, Unit Operations of Chemical Engineering (6th Edition), McGraw-Hill, 2001, pp. 812-821. 3. Seader, J.D. and E.J. Henley, Separation Process Principles, John Wiley, 1998, p. 841.

98

19. Centrifugal separation

Mechanical separation based on particle travel can be based strictly on gravitational forces (sedimentation) or can include centrifugal forces. The latter are particularly useful for fine particles when gravity alone is inadequate for sharp separation. There are two broad categories of centrifugal separators: those where the fluid moves through a fixed device (cyclones, hydroclones, etc.) those where the system moves to create the centrifugal force (centrifuges)

These are all applications of "pressure diffusion" (if you remember that from Transport). Before analyzing these systems, take a second to review the force balance on a particle:

and notice that the external force term includes either gravitational forces or centrifugal forces or both (also remember that the angular velocity has dimensions of radians/sec).

19.1. Cyclones
Cyclones (or cyclone separators) are typically used to remove mist or small particles from gas streams. Dust removal is a common application. When particles are entrained in a liquid, the same principles are used in a device called a hydroclone. The principles can also be used to separate two liquids in a centrifugal decanter.

99

Cyclones are typically conical in shape. Gas enters near the top through a tangential nozzle and moves in a spiral. As it moves the particles are thrown against the wall where they slide down for collection, while the gas escapes through the top of the device. The separation factor of a cyclone is defined as the ratio of centrifugal to gravitational forces:

In most cyclones the particles being separated are small enough that Stokes' Law can be used to determine the drag force. This means that the force balance on a particle under centrifugal force becomes

Since the acceleration phase for the moving particle is fairly brief, the velocity can be treated as constant with respect to time (though NOT with respect to position) and the force balance solved for the radial velocity

which can in turn be expressed in terms of the gravitational terminal velocity and the tangential velocity

From which it can be seen (as expected) that the higher the terminal velocity the higher the radial velocity, and thus the easier the separation.

100

19.2. Centrifugation
In a centrifuge, the container spins to impart centrifugal force to the system. Although they have long been used for applications including cream separation and oil dehydration, they have become more important of late because of applications separating macromolecules, proteins, nucleic acids, etc. In particular, very high acceleration ultracentrifuges have been developed for extremely small particles. In a typical centrifuge, the fluid enters from the bottom and leaves at the top. As the fluid moves, particles settle out. Particles near the wall are then removed. If the diffusivity is negligible, the limiting behavior of a centrifuge is similar to gravity sedimentation with one big exception. In a centrifuge, the particle concentration is not the same throughout the fluid phase. Instead, it changes with time, since both the volume and acceleration increase radially. This phenomenon is called "radial dilution". Since the acceleration increases as the particle moves away from the axis of rotation, so does the force. As a consequence, terminal velocity is never achieved. However, the radial "drift velocity" of a particle is relatively constant, so calculations are generally based on this. The force balance and radial velocity calculations illustrated for cyclones also apply to centrifuges. Four radial distances will need to be defined. First, set ra as the point where a particle begins to settle, and rb as the point it has reached after a time interval tr, which in turn is defined as the average time an element of fluid remains in the centrifuge. This will be recognized as the residence time, which can also be expressed as (Volume)/(Flow). The other two distances needed describe the size of the fluid layer in the centrifuge. Define the inner radius of the fluid layer as r1 and the outer radius r2 (which is usually the wall). Now, return to the radial velocity expression derived above. The velocity is the time derivative of the radial distance, so the expression can be integrated over the thickness of the fluid layer and the residence time to obtain:

This expression can then be solved for the residence time, or for the flowrate:

101

Separation in a centrifuge can be quantified in several ways. First, define a diameter "cutpoint". This is the diameter of a particle which travels halfway through the fluid layer and thus halfway to the wall. To be removed, a partical must be at the wall, so r b = r2. A particle of cutpoint diameter must therefore have begun to settle at the halfway point, ra = (r1+r2)/2. Then,

and so if the flow through the centrifuge is greater than qcut, almost all particles larger than the cutpoint diameter will be removed, while the smaller particles will remain. As a final calculation, we will derive the "Sigma" value for the centrifuge. This corresponds to the cross sectional area of a gravity settler that would make the same separation as the centrifuge at the same volumetric feed rate. Hence,

if two centrifuges are to perform the same task. Consider a system where the liquid layer inside the centrifuge is thin when compared to the total radius. This means that r1 and r2 are roughly the same, so the expressions for residence time, etc., are undefined. But we can still use the expression for the velocity at a given position. This radial distance of interest is that corresponding to the cutpoint, or half the liquid thickness s, so the velocity must also be equal to this distance divided by the residence time, or

102

which has been rearranged in terms of the desired "Sigma" value and the terminal velocity under gravity. References: 1. Foust, A.S. et al., Principles of Unit Operations (2nd Edition), John Wiley, 1980, pp. 620-629. 2. McCabe, W.L., J.C. Smith, and P. Harriott, Unit Operations of Chemical Engineering (5th Edition), McGraw-Hill, 1993, pp. 31-32, 37-39, 1060-1072. 3. Probstein, R.F., Physicochemical Hydrodynamics: An Introduction (2nd Edition), John Wiley, 1994, pp. 142-147.

103