Coal Gas Cleaning and Purification with Inorganic Membranes

Transcription

1 THE AMERICAN SOCIETY OF MECHANICAL ENGINEERS 346 E. 47 St.. New York. N.Y GT-359 The Society shall not be responsible for statements or opinions advanced in papers or in discussion at meetings of the Society or of Its Divisions or Sections, or printed in its publications. Discussion is printed only If the paper Is published in an ASME Journal. Papers are available from ASME for fifteen months after the meeting. Printed in USA. Copyright 1992 by ASME Coal Gas Cleaning and Purification with Inorganic Membranes DOUGLAS E. FAIN and GEORGE E. ROETTGER Oak Ridge K-25 Site* Martin Marietta Energy Systems, Inc. Oak Ridge, Tennessee ABSTRACT The economic viability of coal gasification could depend on the ability to clean and purify the coal gases at elevated temperatures. Inorganic membranes have the potential for being used for that purpose. Efforts have been undertaken at the Oak Ridge K-25 Site to develop membranes that would be useful for separating hydrogen from the coal gas at the high operating temperatures. This paper will give a brief review of some fundamentals of gas separation with membranes. Also, a brief discussion of the theory derived to guide the development process will be given. The theory can be used to indicate the pore size needed to achieve good separation. In addition, some experimental results that have been obtained with some of the membranes that have been fabricated will be discussed. INTRODUCTION The economic viability of coal gasification could well depend on the ability to clean and purify the coal gases at elevated temperatures. Inorganic membranes have the potential for being stable and useful at very high temperatures. Therefore, they have the potential for being used to clean and purify the gasified coal. Under the sponsorship of the U.S. Department of Energy (DOE) Office of Fossil Energy's Advanced Research and Technology Development (AR&TD) Materials Program, a project has been undertaken at the K-25 Site to develop membranes that would be useful in the coal gasification process (Fain et al., 1991a, Fain et al., 1991b, Egan et al., 1991). This project has been directed toward developing a membrane that can separate hydrogen from coal gas at operating temperatures. Since hydrogen has economic value as a pure process gas, separation of some of the hydrogen from the coal gas for resale will significantly improve the economics of the process. A second phase of the development project will be to develop membranes to remove acid gases from the coal gas. FUNDAMENTALS OF GAS SEPARATION BY MEMBRANES Separation of a gas mixture using a membrane is accomplished by passing the mixture (feed stream) across one side of the membrane. A fraction of the gas (enriched stream) passes through the membrane and the remaining fraction of the gas (depleted stream) is collected and not allowed to pass through the membrane. The ratio of the enriched stream flow rate to the feed flow rate will be referred to as the cut. A driving force is needed to produce transport through the membrane. That driving force is a difference in the partial gas density of any given component on the two sides of the membrane. The transport will be in the direction of decreasing partial density. The difference in partial density can be achieved by a difference in pressure, temperature, or mole fraction. Controlling the boundary conditions of any of these three variables on the two sides of the membrane will provide a driving force and produce a transport of different gases through the membrane. The membranes themselves affect these variables by influencing, differently, the rates at which different gas molecules are transported through the membrane. The influence of the membrane depends on the physical and chemical interaction of the gases with the membrane. *The Oak Ridge K-25 Site is managed by Martin Marietta Energy Systems, Inc. for the U.S. Department of Energy under contract no. DE-ACO5-840R Presented at the International Gas Turbine and Aeroengine Congress and Exposition Cologne, Germany June 1-4, 1992 This paper has been accepted for publication in the Transactions of the ASME Discussion of it will be accepted at ASME Headquarters until September 30, 1992

2 With organic membranes, the transport appears to be influenced mostly by a solution diffusion model. It appears that both the solution and diffusion are affected by the chemical interaction of the gases with the membrane, the free volume in the membrane, and the flexibility of the organic matrix. In organic membranes, the free volume is usual fairly small, 15% or much less, but it can vary depending on the flexibility of the material. There is little or no interest in pores or pore size distributions. Inorganic membranes are usually porous with much more variable free volumes, perhaps 20% to 70%, and relatively little or no flexibility. The pore size and pore size distribution of the free volume is of fundamental importance. There are potentially many different transport mechanisms in such a membrane. The transport by the different mechanisms is greatly influenced (in different ways) by one or more of the mean dimensions of the distribution of path lengths in the free volume. The chemical interaction between the gas molecules and the membrane surfaces also influences the transport. The best known and understood mechanism is free molecule diffusion or Knudsen flow. This type of flow occurs when the molecular density is such that the mean free path between collisions with a membrane wall are small in comparison to the mean free path for collisions between molecules and when the pore size is large in comparison to the size of the molecules. Under these conditions, the diffusion coefficient for transport through the membrane is proportional to the mean molecular velocity and the mean path length between collisions with the membrane walls. The separation factor is equal to the square root of the ratio of the molecular weights. For most commercial applications, a square root of molecular weight separation factor is too small to be economical. The challenge is to design membranes that will increase the transport of the desired gas and decrease the transport of the other gases and provide effective separation factors much larger than the square root of the molecular weight ratio. The most common way that membranes are used to separate a gas mixture is to fix the pressure on the two sides of the membrane. The ratio of the pressures on the two sides of the membrane is a most important parameter. The largest change in mole fraction, AC*, will occur when the ratio of the low side pressure to the high side pressure, W, approaches zero and is zero when the ratio approaches unity. The change in mole fraction across an ideal membrane is where AC = AC (1 - W) AC = C1 - C2, C1(1 - Ci)(a - 1) AC = a (1 - C1) + Even when the separation factor is infinite (the membrane passes only the desired gas), the partial pressure on the high side of the membrane can not be reduced below the pressure that is fixed for the low side. The amount of recovery of the desired gas depends on the separation factor, the pressure ratio, and the cut. The economics of a separation system depend on the separation factor, the operating conditions for the process, and considerable ingenuity in optimizing the design of the separation unit to achieve maximum economy. The design of a unit for a binary mixture will be simpler than the design of a unit for a mixture with many gases. The separation factor tends to lose its meaning for a mixture of many gases. For a multi-component mixture, it is more meaningful to consider the relative gas velocity instead of a separation factor. The relative gas velocity can be defined as the ratio of a gas velocity to that of the velocity of the slowest gas molecule in the mixture that passes through the membrane. If a gas is not transported through the membrane, its relative velocity will be zero. The gas components in a representative coal gas mixture are shown in Table 1. The mole fractions of the enriched and depleted streams for the maximum mole fraction change that can be obtained with an ideal Knudsen membrane are shown in Table 2, and corresponding values for a perfect membrane that passes only hydrogen are shown in Table 3. The individual gas cut is the ratio of the amount of that gas in the enriched stream to the amount of that gas in the feed stream. Table 1. Typical Coal Gas Mixture Gas Mole Fraction Molecule Diameter Hydrogen (F12) Water (H20) Carbon Monoxide (CO) Hydrogen Sulfide (H2S) Methane (CH4) Carbon Dioxide (CO2) Table 2. Mole Fractions for an Ideal Knudsen Membrane With a 25% Total Volume Cut Mole Fraction Relative Individual Gas Velocity Enriched Depleted Gas Cut H H CO H2S CH, CO

3 - Gas Table 3. Mole Fraction Change for a Perfect Membrane That Passes Only Hydrogen with 50% Recovery of Hydrogen Mole Fraction Relative Velocity Enriched Depleted Individual Gas Cut H H CO H2S CH, CO O Knudsen Membrane A Pore Diameter O 3.7 A Pore Diameter A Perfect Membrane Figures 1 and 2 provide graphical comparisons of model calculation for several different types of membranes. The enriched stream mole fraction for various amounts of hydrogen recovery are shown in Fig. 1. The cut, fraction of the feed gas that passes through the membrane, needed to achieve a fractional hydrogen recovery is shown in Fig. 2. Two curves are shown for each membrane; in Fig. 1, the upper curve is for zero pressure ratio and the lower curve is for a pressure ratio of (high side pressure of 430 psia and low side pressure of 100 psia); in Fig. 2, the curves are oppositely oriented, the upper curve is for a pressure ratio of and the lower curve is for a pressure ratio of zero. These figures show the importance of a large separation factor in achieving a highly enriched recovery of hydrogen. A Knudsen membrane can produce only a moderately enriched product stream of hydrogen. The separation factor for hydrogen and carbon dioxide with a 5.0 Angstroms (A) pore diameter membrane is about 40 and infinite for a 4.0 A pore diameter. The small deviations between the 3.7 A pore diameter and the perfect membrane shown in Fig. 2 are caused by the transport of small amounts of carbon monoxide and water O Knudsen Membrane A Pore Diameter O 3.7 A Pore Diameter A Perfect Membrane A Fractional Recovery of Hydrogen Fig. 2 Fraction of Feed Gas Passing Through the Membrane to Achieve a Fractional Hydrogen Recovery If membranes that pass only one specific gas could be produced, then a separation unit using these specific membranes in series could be used to remove the individual gases one at a time. It is, therefore, very important to have some guidelines or a theoretical model to provide a basis for designing and developing membranes appropriate for separating individual gases. A SIMPLE MODEL FOR GAS SEPARATION BY INORGANIC MEMBRANES The usefulness of any model will depend greatly on the accuracy of the assumed mechanisms for the transport process. The difficulty of the task is emphasized by the fact that there may be a multiplicity of transport mechanisms and the degree of multiplicity may vary among the different gases. The modeling should start with and use as much as possible that which is already known. The problem of viscous laminar flow and free molecule diffusion or Knudsen flow through porous materials with pore radii down to about 100 A is approximately solved by assuming the additivity of Poiseuille and Knudsen flow (sometimes referred to as Maxwell slip flow in this approximation) (Pollard and Present, 1948). There are other more sophisticated solutions, but this is generally adequate for porous materials. The membranes of interest will generally have pore sizes much less than 100 A pore radii. Thus, in the development of a model, viscous flow will be ignored. Viscous flow can be simply added later if it is needed. The equation for free molecule diffusion of a gas in a circular capillary in standard cc/sec is F = 2/3v - nr 3 ToAP TPol (1) 0.4 Of Fraction Recovery of Hydrogen Fig. 1 Mole Fraction of the Enriched Stream for Various Amounts of Hydrogen Recovery and the separation factor for a binary mixture is given by the ratio of the flow per unit pressure difference for each gas v,. M,. a = =. M (2) 3

4 iidscn ot tiee molecule flow is generally considered to represent the mechanism for binary gas separation factors that approach the square root of the gas molecular weight ratio. Present and de Bethune have developed an excellent theory for this type of separation (Present and debethune, 1949). Experimental validity of the essentials of this theory have been demonstrated (Fain and Brown, 1974). This type of separation has been demonstrated repeatedly in membranes that have effective pore sizes that are very large compared to the size of the gas molecules. However, when the pore size of the membrane approaches the size of the gas molecules other considerations become important. The diffusion process associated with Knudsen flow considers the gas molecule as a point with no physical dimensions. In fact, molecules do have a finite size. As a first approximation we may consider a membrane as made up of capillaries with smooth walls and the gas molecules as hard spheres with a fixed diameter. Diffusion is a random walk process in which molecules move a given (variable) distance, then change direction in a random way independent of the direction from which it came. The transport through the membrane by Knudsen diffusion is the product of the diffusion coefficient, the area and the density gradient. The diffusion coefficient is proportional to the product of a mean distance between collisions with the capillary walls and the mean molecular \,elocity. The mean distance between wall collisions is calculated by averaging the distance from a point on the wall of a capillary to all visible points on other walls weighted by the cosine law distribution of paths from the wall's surface. A. leas:n the first approximation, the closest approach to the capillary wall by the hard sphere molecule is the radius of the hard sphere molecule. Therefore, the actual mean distance between wall collisions for a molecule is proportional tc ti ; iifferene between the mean distance to a wall and the diameter of the hard sphere molecule. In the case of a capillary, it would be the difference between the capillary diameter and the molecule diameter. In addition, the area open to flow is the area calculated by using the difference between an equivalent diameter equal to the diameter of a capillary and the diameter of the hard sphere molecule. In this approximation, the diffusion coefficient and the area depend on the diameter of the gas molecule. The equation for Knudsen flow is then given by F, = 2/3 vn(r-8,)3 To AP. (3) TPot For a binary mixture, the separation factor is the ratio of the specific flows of the two components. For point molecules, the separation factor is simply the square root of the ratio of the molecular weights. For hard sphere molecules, the separation factor is, (r-6,)3 (Afir2 a - (4) 0-931m,) Thc sztalation factor calculated for hard sphere molecules, usir, Eq. (4) for several binary mixtures is shown in Fig. 3. The separation factor can become very large as the pore size approaches the size of the larger molecule. However, this figure assumes only the free molecule diffusion component of transport. The effect of the other components of transport must also be taken Into account. :5; Classical Knudsen Separation Factor He N A 112 0O X He 0O V H Fig. 3 Separation Factors for Ideal Knudsen Flow of Hard Sphere Molecules As the pore radii decrease below 100 A, other transport components become significant. One of these components is usually referred to as surface flow. Surface flow is associated with the adsorption of gases on the membrane surfaces and the transport is thought of as diffusion of molecules on the surface. In addition to the enhanced flow due to the diffusion on the surface, the molecules adsorbed on the surface change the effective radius of the pores and, therefore, the amount of transport from other flow mechanisms. As the pore radii become still smaller, as indicated above, the size of the molecule itself becomes significant. The effective pore radius is the pore radius minus the effective thickness of the adsorbed layer and minus the radius of the molecule. This makes the effective pore radius different for different size molecules. We can use statistical arguments to infer the effect on the transport diameter by adsorbed molecules. In the hard sphere approximation, the effect of one monolayer of adsorbed molecules will be simply to decrease the radius of the capillary by the diameter of the adsorbed molecules. If less than a monolayer is adsorbed, sometimes the jump distance will not be affected, other times the jump distance will be decreased by one adsorbed molecule diameter and other times the jump distance will be reduced by two adsorbed molecule diameters. On the average the jump distance will be decreased by an amount equivalent to twice the average thickness of the adsorbed layer. The average thickness will be equal to the product of the fraction of a monolayer and the thickness of a single monolayer. It follows that the effective thickness is the product of the fractional number of monolayers and the thickness of a single monolayer. The adsorbed molecules can move around on the surface in a two-dimensional sense. This surface diffusion will contribute to the transport and for a first approximation may be considered simply additive to the gas phase diffusion. Surface diffusion is considered a two dimensional transport with the diffusion coefficient proportional to the product of the mean distance between direction changes on the surface and the mean velocity on the surface. The flow rate is proportional to the perimeter of the flow passage and the

5 driving force is the surface density gradient. For the current data of interest, the adsorption of gases is much less than a monolayer. Therefore, only Henry's Law adsorption is considered here. That can be modified when the adsorption is much higher. It has been well recognized in the literature (Kennard, 1938) that there is often a variation in the momentum accommodation coefficient when molecules collide with different surfaces. This phenomenon is observed with both Maxwell slip flow and with Knudsen flow. It has generally been concluded that the effect is the same for both types of flow. However, there are theoretical and experimental data that suggest the effect of accommodation coefficients less than unity is different for the two types of flow (Fain, 1981). In the transport equations the accommodation coefficients should be taken into account. The accommodation coefficient is simply a multiplicative factor on the Knudsen diffusion coefficient. The functional form of the equation which takes all of these transport mechanism into account is )-2 0.7AP, (r-5,-c-tg)3 [fi + A (1- I. There are several constants in Eq. (5) that have been combined into the single constant C. These constants can be estimated from various types of data, such as adsorption isotherms and single component transport measurements. These constants have been evaluated from such measurements with helium and carbon dioxide. The separation factors calculated using Eq. (5) are shown in Fig. 4 plotted against pore diameter. The initial decrease in separation factor as the pore size decreases is due to an increase in the surface flow of carbon dioxide. However, as the temperature is increased, the surface flow and adsorption of carbon dioxide should decrease and the separation factor should approach the high temperature limit estimated by Eq. (4) and shown in Fig Classical Knudsen Separation Factor = o Average Pressure O 0 psi (25 deg C) psi (25 deg C) O 200 psi (25 deg C) A 400 psi (25 deg C) X Ideal (High Temperature Limit) Mean Pore Radius, Angstroms Fig. 4 He-0O2 Separation Factors Calculated from the Transport Model for Individual Pure Gas Flows. (5) 5 EXPERIMENTAL RESULTS To date in this project, a large number of alumina membranes have been fabricated and characterized. As part of the characterization, pure gas flows rates have been measured on some of these alumina membranes having mean pore diameters between 10 and 50 A. The flow measurements were made at room temperature using helium, nitrogen, and carbon dioxide as test gases. Adsorption isotherms were also measured at room temperature. The constants needed in Eq. (5) have been approximated from these data. The separation factors calculated with these data are shown in Fig. 5 for direct comparison with the experimental data. As can be seen, the agreement between the calculated and measured separation factors is quite good Average Pore Radius, Angstroms Fig. 5 Comparison of Separation Factors Calculated from He and CO2 Pure Gas Flow Measurements with Calculations from the Transport Model The separation factors calculated from the model for the individual gas flows and for several pressures are again shown in Fig. 6 on a different scale. It is common practice to use the method of measuring pure gas flow rates to estimate the separation factor for a binary mixture. However, the adsorption of one gas on the walls of a membrane will also affect the transport of the other molecular component. This will be the case when an actual binary separation is measured. The separation factors for the same conditions shown in Fig. 6, but with the flow of a binary mixture, are shown in Fig. 7. The calculations in Fig. 7 take into account the effect of the adsorption on both gases, i.e., the effect of the adsorption of the carbon dioxide on the transport of the helium, which would not occur for the individual gas transport. However, no interaction or momentum exchange between the two species was considered. It is likely that momentum exchange and momentum accommodation coefficients less that unity for either or both gas molecules will occur in practice. Therefore, the separation factors in Fig. 7 show only the effects of surface flow and adsorption. A comparison of the results shown in Figs. 6 and 7 indicate the importance of making actual separation measurements at an early date to verify the assumption. Momentum exchange and variations in accomodation coefficients could make the differences even larger than shown in Figs. 6 and 7.

6 Classical Knudsen Separation Factor = size further. A particular problem is appropriately measuring the membrane pore size to continue to track the fabrication methods and control the development process. High temperature flow and separation measurements are expected to confirm the higher separation factors before the end of this year. Symbols used in equations I I f I Mean Pore Radius, Angstroms Mean Pore Radius, Angstroms Average Pressure o 0 psi (25 deg C) + 40 psi (25 deg C) o 100 psi (25 deg C) A 200 psi (25 deg C) X 400 psi (25 deg C) Fig. 6 He-0O2 Separation Factors from the Gas Transport Model Using Individual Pure Gas Flows. 20 Average Pressure O 0 psi (25 deg C) + 40 psi (25 deg C) O 100 psi (25 deg C) A 200 psi (25 deg C) x 400 psi (25 deg C) Fig. 7 He-0O2 Separation Factors from the Gas Transport Model for Gas Mixture Flows (Without Momentum Transfer) CONCLUSIONS Much progress has been made in reducing the pore size of inorganic membranes and learning about the physical characteristics of these membranes. However, the experimental data alone might have suggested that the development effort would not be successful since the separation factors have been decreasing with smaller pore radii. But the mathematical model has indicated that the development work is proceeding as should be expected. With the mathematical model the data are encouraging and we have incentive and direction to vigorously continue the development effort. To achieve high separation factors for hydrogen separation from coal gas, the model clearly indicates the need for smaller pores than have been achieved to date. We believe there are techniques for reducing pore M, = molecular weight of the ith gas molecule, F; = flow rate in standard cc per second, 7; = molecular velocity of the ith gas molecule, AP; = partial pressure drop of the ith gas molecule, Po = standard pressure, T = temperature, To = standard temperature W = pressure ratio, f; = momentum accommodation coefficient of the gas molecule, = capillary length, r = capillary radius, (5; = radius of the jth gas molecule, t, = the adsorbed film thickness for the ith gas molecule, A and C = temperature dependent constants, a = separation factor, = feed mole fraction, C, = enriched mole fraction, AC = C, Co, and AC* = maximum ideal value for Co. REFERENCES Fain, D. E., Roettger, G. E., and White, D. E., "Development of Ceramic Membranes for High-Temperature Hydrogen Separation," Proceedings of the Fifth Annual Conference on Fossil Energy Materials, Martin Marietta Energy Systems, Inc., Oak Ridge National Laboratory, Oak Ridge, Tennessee, ORNL/FMP-91/1, pp , September Fain, D. E., Roettger, G. E., and White, D. E., Development of Ceramic Membranes for Gas Separation - FY 1990 Development Activities, K/QT-413, Martin Marietta Energy Systems, Inc., Oak Ridge K-25 Site, Oak Ridge, Tennessee, June Egan, B. Z., Fain, D. E., Roettger, G. E., and White, D. E., "Separating Hydrogen from Coal Gasification Gases with Alumina Membranes," IGT&A Congress, June 1991, to be published. Pollard, W. G. and Present, R. D., Phys. Rev., 73, 762 (1948). Present, R. D. and debethune, A. J., Phys. Rev., 75, 1050 (1949). Fain, D. E. and Brown, W. K., Neon Isotope Separation by Gaseous Diffusion Transport in the Transition Flow Regime with Regular Geometries, K-1863, Union Carbide Corporation, Nuclear Division, Oak Ridge Gaseous Diffusion Plant, (1974). Kennard, E. H., Kinetic Theory of Gases, McGraw-Hill Book Co., Inc., New York and London, Fain, D. E., "A Comparison of the Effect of Nonrandom Surface Scattering on Free-Molecule and Slip Flow Through Capillaries," Rarefied Gas Dynamics, 74 (Pt. 1), pp ,