Air separation by pressure swing adsorption with a high capacity carbon molecular sieve

Molecular sieve abstract

Nitrogen is produced by separating air in a pressure swing adsorption process using a high capacity kinetically O.sub.2 selective material as the adsorbent. An example of such an adsorbent is a carbon molecular sieve made from coconut shell char. Such a coconut shell char having a high oxygen volumetric capacity is provided by crushing and sizing coconut shells to form granules which are then heated in flowing inert gas at a temperature ramp rate of about 2 to 12.degree. C. per minute to a peak temperature of 775 to 900.degree. C. which is then held so that the total heating time is up to 8 hours and thereafter the granular char is cooled in an inert gas atmosphere. The granular char thus produced is kinetically selective for O.sub.2 over N.sub.2 without further modification to narrow the openings of its micropores and has an oxygen volumetric capacity in excess of 8.0 cc/cc at 1 atm. and ambient temperature. Further modification of this char is provided by contacting it with an oxidizing atmosphere of carbon dioxide or a mixture of inert gas and carbon dioxide, H.sub.2 O or O.sub.2 at temperatures ranging from 650.degree. to 900.degree. C. until the gasified char has been altered so that its volumetric oxygen capacity is greater than 9.0 cc/cc at 1 atm. and ambient temperature.

Molecular sieve claims

We claim:

1. In a process for separating air for nitrogen recovery by pressure swing adsorption, the improvement for achieving enhanced nitrogen recovery which comprises using as an adsorbent in said process a carbon molecular sieve having an oxygen volumetric capacity of at least 9.0 cc/cc at 20.degree. C. and 1 atmosphere which sieve is made from a granular coconut shell char formed by:

(a) crushing and sizing raw coconut shells to form shell granules having a size suitable for use in an adsorbent bed for pressure swing adsorption,

(b) heating said granules in a flowing stream of inert gas at an average temperature ramp rate of about 2.degree. C. to 12.degree. C. per minute to a peak temperature of 775.degree. to 900.degree. C,

(c) holding said peak temperature for a period of time so that the total heating and holding steps total not less than 1 hour nor more than 8 hours to produce granular char,

(d) cooling said granular char in an inert gas atmosphere, and

(e) recovering said product of step (d).

2. The process of claim 1 wherein said carbon molecular sieve has an oxygen volumetric capacity of at least 10.0 cc/cc at 20.degree. C. and 1 atmosphere.

3. The process of claim 2 wherein said carbon molecular sieve has an oxygen uptake rate at least 15 percent slower and a kinetic selectivity (O.sub.2 /N.sub.2) at least 15 percent lower than a carbon molecular which has an oxygen volumetric capacity of 8.5 cc/cc or less but performs about the same in terms of productivity and recovery under equivalent process conditions with optimized cycle times in air separation by pressure swing adsorption.

Molecular sieve description

FIELD OF THE INVENTION

This invention relates to a method of separating air for nitrogen recovery by pressure swing adsorption using a carbon molecular sieve adsorbent having high volumetric capacity for oxygen.

BACKGROUND OF THE INVENTION

Nitrogen Pressure Swing Adsorption (N.sub.2 PSA) is a separation process commonly used for the production of nitrogen. The feed to this process is typically air but may be nitrogen enriched air. It is typically a two bed process operating with a relatively simple cycle that includes pressurization, product draw at high pressure, pressure equalization of the beds and depressurization. Other steps such as product purge may be used. The total cycle time is on the order of minutes. The high pressure is typically 4 to 8 atmospheres and depressurization is typically to atmospheric pressure though vacuum could be employed.

N.sub.2 PSA is a kinetics-based process, i.e.; the separation of O.sub.2 and N.sub.2 occurs because of a kinetic rather than an equilibrium selectivity of the adsorbent. During the high pressure feed step, O.sub.2 is selectivity adsorbed and an N.sub.2 enriched product is withdrawn. During the low pressure blowdown step, the O.sub.2 -rich adsorbed phase is desorbed and removed from the bed. The adsorbent commonly employed in commercial units is carbon molecular sieve (CMS).

Carbon molecular sieves are usually prepared by treating a carbonaceous material (for example: coal, coconut shell char, peat, pitch, carbonized polymers, and the like) with additional carbon-containing species. U.S. Pat. No. 3801513 Munzner, et al., (1974) describes obtaining carbon molecular sieves (CMS) for oxygen separation by treating coke having volatile components of up to 5% with a carbonaceous substance that splits off carbon at 600.degree. to 900.degree. C., thereby narrowing the pores present in the coke. The starting coke can be derived from coal, peat, coconut shell, wood or plastics. It has been stated by others that the average pore size of the adsorbent must be below 3 angstroms to effect oxygen separation from nitrogen. The average pore diameter can be adjusted by changing the intensity of the treatment. Example 6 of U.S. Pat. No. 3801513 describes coconut shell material having a particle size of 1 to 3 mm which is heated at 3.degree. C. per minute to 750.degree. C., the volatiles being equal to 4.5%, where it is held for 30 minutes while ethylene gas is introduced during the holding period. The material is then cooled under nitrogen. In an evaluation test, a gas product was reported containing 49.5% nitrogen and 50.5% oxygen.

At about the same time, attention was directed to the use of other materials as the base material for making carbon molecular sieves. Japanese Publication No. Sho 49-37036 (1974) describes making a carbon molecular sieve by condensing or polymerizing a phenol resin or furan resin, so that the resin is adsorbed on a carbon adsorbent and thereafter carbonizing the product by heating. The carbonizing can be carried out at 400.degree. to 1000.degree. C. in an inert gas. The heating rate given is 50.degree. to 400.degree. C. per hour (0.8.degree. to 6.7.degree. C. per minute) and an example is given of heating at 6.7.degree. C. per minute to 650.degree. and 800.degree. C. where the material is held for 1.5 hours. The operation is said to reduce the pore diameter of the carbon adsorbent.

Coconut shell char is a commodity material readily available commercially and is often cited as a suitable base material for preparation of carbon molecular sieves by various modifications. Very little description has been provided, however, about the preparation of the coconut shell char itself. Shi Yinrui, et al., "Carbonization of Coconut Shells", Forest Products Chemistry and Industry Institute, Chinese Academy of Forestry, Vol. 6 No. 2 pages 23-28 (1982), describes making coconut char for the production of activated carbon by heating the shells to 720.degree. C. Carbonization is said to be complete at 550.degree. C., using heating rates of 10.degree. and 20.degree. C. per minute. At lower carbonization temperatures, it is stated that the rate should be less than 10.degree. C. per minute.

U.S. Pat. No. 4594163 Sutt, Jr., (1986) describes a continuous process for making a CMS beginning with a charred naturally occurring substrate and using non-activation conditions, i.e. non-oxidizing, and no addition of pore-constricting or blocking materials. The process is stated to involve heating the char, e.g., coconut shell char made by the process described in U.S. Pat. No. 3884830 to 900.degree. to 2000.degree. F. for 5 to 90 minutes. The examples heat to 1800.degree. F. (982.degree. C.) and higher. The CMS product is said to have improved oxygen capacity at 25.degree. C. of 4.00 to 6.00 cc/cc, and average effective pore diameters of 3 to 5 angstroms. As an example, representing prior art, charred coconut shell is prepared by heating at 5.degree. C. per minute to 500.degree. C., crushing and sieving the char to obtain 20.times.40 mesh (U.S. sieve) material and then treating in nitrogen by heating at 5.degree. C. per minute to 950.degree. C. and holding for 2 hours. The product had a volumetric oxygen capacity of only 0.8 cc/cc. For the preparation of the coconut shell char, reference was made to U.S. Pat. No. 3884830 Grant (1975), which describes preparing activated carbon from starting material such as bituminous coal and charred materials such as coconut char. The coal or char is crushed, sized and mixed with a binder and either agglomerated or compressed into shapes which are then crushed and screened. Activation proceeds by air baking at 300.degree. to 400.degree. C. and calcination at 850.degree. to 960.degree. C. No information is given on preparation of the starting charred materials.

U.S. Pat. No. 4627857 Sutt, Jr., (1986) describes preparing a CMS for oxygen/nitrogen separation by continuous calcination of agglomerated non-coking or decoked carbonaceous material, such as coconut char. The agglomerated substrate includes a thermal binder and is sized and screened or pelletized. Calcining is carried out under inert gas purge at 250.degree. to 1100.degree. C. for at least one minute, preferably 10 to 60 minutes. Examples give oxygen capacities at 25.degree. C. for the product CMS of 2.25 to 4.44 cc/cc. For information on the starting char material, reference is made to the above mentioned U.S. Pat. No. 3884830.

U.S. Pat. No. 4629476 Sutt, Jr., (1986) describes making a CMS said to have improved selectivity for gas or liquid separations by impregnating a carbonaceous substrate, e.g. coconut shell char, with an organic polymer having a molecular weight of at least 400 or an inorganic polymer at a dosage rate of at least 0.001 wt. %. Further modification of the impregnated sieve by charring at 250.degree. to 1100.degree. C. is disclosed.

It is common in duscussions of preparing CMS from coconut shell charcoal to direct the preparation of the char into a pellet for use in separation processes. U.S. Pat. No. 4742040 Ohsaki, et al., (1988) describes making CMS by combining coconut shell charcoal with a binder of coal tar or coal tar pitch, pelletizing and carbonizing the pellets at 600.degree. to 900.degree. C., removing soluble ingredients from the pellets with a mineral acid, drying the pellets, adding a distilled creosote fraction and reheating to 600.degree. to 900.degree. C. for 10 to 60 minutes. Oxygen capacities at 25.degree. C. of about 6.0 to 7.0 milliliters per gram are disclosed for the product CMS and 8.0 milliliters per gram for the raw carbonized charcoal which is non-selective. A similar approach of converting the carbonized material into pellets is given by U.S. Pat. No. 4933314 Marumo, et. al. (1990) which describes making CMS from spherical phenol resin powder mixed with a binder, pelletized and heated to carbonize the pellets. In making the CMS, various materials such as finely divided cellulose, coconut shell, coal, tar, pitch or other resins can be added in small amounts to improve workability, e.g. in pellet molding. The use of pelleted CMS, besides involving expensive processes for forming the pelletized material, invariably suffers from residual binder material or its decomposition products in the pores of the CMS, thereby reducing its overall capacity. It is highly desirable, therefore, to be able to develop a carbon molecular sieve which is granular and can be used directly in an adsorbent bed for separations without going through a pelletizing process.

Modifications of carbonized materials are described to involve various steps other than the deposition of carbonaceous materials to narrow sieve pore openings. For example, Wigmans, "Industrial Aspects of Production and Use of Activated Carbons", Carbon, Volume 27 1 pages 13-22 (1989) describes activation of carbonized residues of coal, wood, coconut shell and the like using agents such as steam, carbon dioxide and air to expose internal porosity. Above 800.degree. C., oxygen reacts 100 times faster with carbon than do steam or carbon dioxide, so that activation is possible only under mass-transfer-limiting and product-inhibiting conditions. Pore volume and pore enlargement occurs with increasing burnoff, but an optimum in surface area and micropore volume is observed. Temperatures of 800.degree. to 850.degree. C. are said to seem to be optimum without notable pore shrinking behavior.

The value of using carbon molecular sieves for air separation in pressure swing adsorption (PSA) is documented in Seemann, et. al., "Modeling of a Pressure-Swing Adsorption Process for Oxygen Enrichment with Carbon Molecular Sieve", Chem. Eng. Technol., 11 pages 341-351 (1988). This article discusses PSA cycles for separating oxygen from nitrogen and argon using a CMS (commercial CMS N2 material manufactured by Bergwerksverband GmbH, Essen), for which structural data are given as are adsorption equilibria of oxygen, nitrogen and argon at 30.degree. C. It is pointed out that at equilibrium these gases are adsorbed in similar amounts, but oxygen is adsorbed considerably faster because its effective diffusion coefficient is more than 8 times those of nitrogen and argon. Consequently, an almost oxygen-free nitrogen-argon mixture can be recovered during adsorption, and on depressurization of the adsorbent bed, a gas containing over 50 volume percent oxygen may be obtained.

The use of zeolites as kinetics-based adsorbents for this application has been suggested in the literature (D. W. Breck, J. Chem. Education, 41 6781964; E. J. Pan et. al. in "New Directions for Sorption Technology", G. E. Keller and R. T. Yang (ed.), Butterworth, 1973; H. S. Shin and K. S. Knaebel, AlChE J., 33654 1987; H. S. Shin and K. S. Knaebel, AlChE J., 34 1409 1988). While it is recognized that the equilibrium selectivity of zeolites for N.sub.2 over O.sub.2 generally detracts from the kinetic selectivity for O.sub.2 over N.sub.2 the cited references show that a kinetics-based separation process using zeolites is possible. The emphasis of the work reported in these references was on process variables and cycle development. The zeolites employed were small pore materials with oxygen capacities of about 5cc/cc (at one atmosphere and ambient temperature).

The cost of N.sub.2 produced by N.sub.2 PSA is a function of the process productivity (SCFH N.sub.2 product/cu ft of adsorbent) and air recovery (moles of N.sub.2 produced/mole air feed). The recovery and productivity of a unit can be varied by changing the process cycle time and/or other operating conditions. Except in undesirable operating regions, however, increasing one parameter usually results in a decrease in the other.

Since N.sub.2 PSA is a kinetics based process, it was anticipated that the most significant process improvements would be obtained through improvements to the kinetic uptake rates and/or kinetic selectivity of the adsorbent. Generally, faster uptake rates yield higher productivity; and higher selectivity results in higher recovery. It is very difficult to increase both the uptake rates and selectivity of an adsorbent simultaneously, or even to increase one while holding the other constant.

Clearly, the potential for use of kinetic-based adsorbents, particularly carbon molecular sieves, in PSA is very high, but the prior art appears to focus on improving the selectivity and adsorption rates of the adsorbents with little or no attention to enhancing the O.sub.2 capacity of the starting material.

SUMMARY OF THE INVENTION

We have discovered that, quite surprisingly, N.sub.2 PSA process performance is much more sensitive to the O.sub.2 equilibrium capacity of the adsorbent used than to the gas uptake rates or kinetic selectivity of the adsorbent. Additionally, increasing the equilibrium capacity of the adsorbent increases both recovery and productivity. The increase in recovery with capacity is particularly surprising.

According to our invention a process is provided for separating air for nitrogen recovery by pressure swing adsorption, using as the adsorbent in the process a carbon molecular sieve having an oxygen volumetric capacity of at least 9.0 cc/cc at 20.degree. C. and one atmosphere. Preferably the carbon molecular sieve has an oxygen volumetric capacity of at least 10.0 cc/cc.

DETAILED DESCRIPTION OF THE INVENTION

A carbon molecular sieve based on coconut shell char and having an exceptionally high oxygen volumetric capacity is described in U.S. Ser. No. 07/644711 filed Jan. 23 1991 now U.S. Pat. No. 5164355 of which this application is a continuation-in-part. The parent application also discloses a method of making such a high capacity carbon molecular sieve (CMS). By modeling the use of this high capacity CMS for air separation by pressure swing adsorption (PSA), we have discovered that the exceptionally high oxygen volumetric capacity of the CMS enables higher recovery and productivity in PSA than commercial CMS, even if the high capacity CMS has a slower oxygen uptake rate and a lower O.sub.2 /N.sub.2 kinetic selectivity than the commercial CMS. Moreover, it has been found that such improvements are realized without depending upon the adsorbent. In other words, a superior PSA process for air separation is available through this invention using any adsorbent of sufficiently high oxygen volumetric capacity. This has been established because the parameters entered into the model for evaluation of the adsorbent in PSA air separation include only physical, equilibrium and kinetic properties of the adsorbent and do not specify the chemical nature or origin of the adsorbent.

The computational method that was used to rank the relative performances of an adsorbent in a N.sub.2 PSA cycle, based on the fundamental properties of the adsorbent and process conditions, involved calculating the overall cyclic steady state O.sub.2 and N.sub.2 working capacities of the system at each end of an N.sub.2 PSA bed. These working capacities were calculated using a computer model that accurately accounts for mass transfer within the macropores and micropores of kinetics-based adsorbents. We have found that O.sub.2 and N.sub.2 uptake by CMSs are best modelled by defining two mass transfer resistances; a surface barrier resistance at the entrances to the micropores in series with a homogeneous resistance along the micropores. The calculations were for adiabatic adsorption beds and employed a multicomponent dual site Langmuir equation to represent equilibrium capacity.

Once the overall working capacities were obtained, recovery and productivity were calculated by the following equations, using arithmetic averages of the overall working capacities at the end points of the bed. This method was not used to accurately predict the absolute values of recovery and productivity; instead it has been shown to track pilot and commercial scale N.sub.2 PSA performance data quite well and to provide reliable predictions of the relative performances of various CMSs. ##EQU1##

The computational method described above was used to rank adsorbents and define the properties of a superior adsorbent for N.sub.2 PSA. It was found that N.sub.2 PSA process performance is most sensitive to increases in the volumetric equilibrium capacity of the adsorbent (moles/cc of pellet or particle). An increase in capacity on a volumetric basis can be achieved by increasing the inherent sorption capacity of the adsorbent and/or by increasing its particle density.

High capacity CMSs can be derived from coconut shells. Such a CMS is a high density carbon molecular sieve material which can be used as a host material for further modification to produce an oxygen selective CMS. Not only does the material have enhanced volumetric capacity which can be carried forward through various modifications to the final CMS, but the material can also be produced in granular form without need for pelletizing. The capacity and the oxygen adsorption rates of this char can be increased by post treating with an oxidant without enlarging the micropores of the carbon beyond 8 angstroms. The method of making such a CMS includes (a) crushing and sizing coconut shells to form shell granules which have a size that would be suitable for use in an adsorbent bed for PSA, (b) heating said granules in a flowing stream of inert gas at an average temperature rate of increase of about 2.degree. to 12.degree. C. per minute to reach a peak temperature of 775.degree. to 900.degree. C. and (c) thereafter holding the peak temperature for a period of time so that the heating and holding steps total up to about 8 hours to produce the granular char and (d) cooling the granular char in an inert gas atmosphere.

The granular coconut shell char exhibits oxygen selectivity in air separation without modification to narrow the openings of its micropores and has an oxygen volumetric capacity in excess of 8.0 cc/cc at one atmosphere and ambient temperatures. A gasified form of the coconut char is provided by contacting this high capacity char with carbon dioxide or a mixture of carbon dioxide, water, or oxygen in inert gas at a temperature of 650.degree. to 900.degree. C. The gasified char thus formed has a volumetric oxygen capacity greater than 9.0 cc/cc. This gasification treatment is carried out for a time sufficient to increase the oxygen capacity of the char but reduces its oxygen selectivity with respect to nitrogen. The coconut char can also be modified either directly as formed by the original carbonization step or following the gasification procedure by reducing the effective pore openings of the high capacity char by contact with a volatile carbon-containing organic compound under pyrolysis conditions.

This invention makes a significant contribution to the technology of air separation with carbon molecular sieves (CMS) by providing increased capacity of the adsorbent. More gas capacity per unit volume of bed of adsorbent in pressure swing adsorption separation of air components leads to both increased recovery and productivity. While it is recognized that selectivity and rates of gas sorption are important, the value of developing increased capacity in the CMS has attracted insufficient attention. Readily available oxygen selective CMS materials exhibit an oxygen gas capacity at 25.degree. C. of less than 8 cc/gm at 1 atm. of oxygen. Much of the porosity of these commercial materials is made up of both macropores and mesopores generated during the steps of pelletizing and binder burnout while making these CMS materials. Although these macropores provide gas transport, they are typically non-selective for air separation.

A review of available materials and descriptions in the prior art reveals that an upper limit in the gravimetric oxygen capacity of CMS has been about 8 cc/gm at 1 atm. and ambient temperature. Although activation procedures have been developed to increase the gravimetric capacity (cc/gm) of such carbons, they also decrease the density and, therefore, lower the volumetric capacity (cc/cc). Conversely, increasing the density of the carbon by depositing polymer-based pyrolytic carbon generally reduces volumetric capacity by closing micropore structure. In addition, this supplemental treatment tends to slow gas adsorption rates excessively.

A method of producing a high capacity, high density CMS material which can be used as a host material for further modification to produce an efficient oxygen selective CMS is now possible. The CMS thus produced has the additional advantage of being a granular material so that there is no need for further pelletizing. Furthermore, the capacity and oxygen sorption rates of the base material can be raised by post-treatment with an oxidant without enlarging the micropores beyond 8 angstroms. The resulting product is a valuable material suitable as a base for subsequent micropore narrowing by hydrocarbon pyrolysis, either in one or two steps, to convert the host material to an oxygen selective CMS. A suitable two-step procedure is disclosed in copending U.S. patent application Ser. No. 575474 filed Aug. 30 1990 now U.S. Pat. No. 5098880.

The literature is replete with procedures designed to improve the efficiency of carbon molecular sieves by treating a carbonaceous material such as coal, peat, pitch, charred polymers, coconut shell char, or charred shells from other nut sources, with additional carbon-containing species which are pyrolyzed and deposit carbon on the starting carbonized base material. Referring to the references discussed in the foregoing Background of the Invention, it is known that such treating procedures employ very controlled temperature ramp rates and specific upper temperatures. Little attention has been paid, however, to the importance of the starting char or carbonaceous material. The further treatment of the host material depends to a great extent upon the character of the starting char which must be clearly defined. In developing a superior starting char, coconut shells from a cost/performance point of view provide the most promising source material.

Coconut shells are lignocellulosic material consisting of varying percentages of two major organic components. Cellulose and hemicellulose, collectively considered as holocellulose, are linear polymers of glucose and comprise approximately 62% of the shell. Lignin, a three dimensional polymer of aromatic alcohols, makes up 35%, while the remaining 3% is derived from other intracellular substances (McKay and Roberts, Carbon, Vol. 20 No. 2 page 105 1982). Thermal decomposition is most intense for these components below 500.degree. C. and there is little further decomposition above this temperature. Pyrolysis does not, however, destroy the natural cellular structure of the coconut shell.

A process is described herein for preparing a dense, high capacity, host material whose adsorption rates can be readily altered. Starting with fresh coconuts, the hard shells are removed and then crushed and sized in order to obtain a granular material which is suitable in size to be used directly in an adsorbent bed for pressure swing adsorption. The size can vary considerably, for example from 0.5 in. chunks down to 200 mesh material or smaller, but uniformity of size is desirable. Mesh sizes (U.S. Sieve) of 18-25 40-60 60-80 80-100 100-140 and 150-200 are suitable; however fine carbon powders (>60 mesh) are not preferred for PSA units; very fine powders are to be avoided.

This granular shell material is then heated under carefully controlled conditions in a flowing stream of inert gas, preferably nitrogen, at an average temperature rate, referred to as the "ramp" rate, of about 2.degree.-12.degree. C. and, preferably, 2.degree.-10.degree. C. per minute until a peak temperature is reached in the range of 775.degree.-900.degree. C., preferably 775.degree.-825.degree. C. [The set point thermocouple used to control temperature was mounted just outside the metal sleeve used to contain the rotating quartz tube. The temperature within the tube was quite close to this set point temperature]. If the ramp rate has been sufficiently slow and the peak temperature is in the higher end of the range, carbonization of the coconut shell is complete on reaching the peak temperature, but normally it will be desired to hold the peak temperature for a period of time so that the total heating and holding steps together total up to 8 hours. Preferably, the holding time at peak temperature is from about 15 minutes to 1 hour.

The carbonized granular char is then cooled in an inert gas atmosphere. The coconut shell char made in this manner not only has an unusually high volumetric oxygen capacity, but it is also slightly kinetically oxygen selective. The capacity of the char can be improved still further, although destroying the oxygen selectivity, by a post-treatment which involves gasification at elevated temperatures with oxidants, such as carbon dioxide, H.sub.2 O, oxygen. The rates of gas adsorption and density of the CMS thus produced have been controlled by pyrolysis parameters, including the heating rate, the atmosphere, the upper temperature limit, and the post-treatment with oxidizing gases or, either alternatively or successively, the deposition of pyrolyzed hydrocarbon to narrow the micropore openings.

Factors which help define the quality of the coconut shell char are its oxygen capacity, rate of oxygen sorption, pore size distribution (percentage of micropores) and density. The oxygen capacity of this material is crucial and can be expressed volumetrically (cc/cc of adsorbent) or gravimetrically (cc/g of adsorbent). All capacities are measured at ambient conditions (.about.23.degree. C., .about.1 atm O.sub.2). The granular coconut shell char described exhibits kinetic oxygen selectivity in air separation, even without modification to narrow the openings of its micropores, and it has an oxygen volumetric capacity in excess of 8.0 cc/cc. The volume of gas absorbed is measured at 1 atmosphere, with pure oxygen at room temperature. It is possible to have a high gravimetric capacity, but an undesirably low volumetric capacity because of a low density of the carbon material. It is important to realize that specifying capacity in cc/g without specifying the density (and how it is measured) is meaningless. Specifying volumetric capacity and Hg pellet density defines a new and preferred regime.

The material is also granular so that it does not need to be pelleted hence, saving an additional process step in the adsorbent manufacturing process. Because large beds of adsorbent are involved in pressure swing adsorption, it is necessary to use either an extruded or pelleted material if the CMS is not granular, because a fine powder is unacceptable, for pressure drop considerations through the bed. Also a fine powder as the starting char is of disadvantage in further hydrocarbon pyrolysis so that such powder must first be bound with an agent which permits it to be formed into an extruded pellet.

Although one may have a high capacity base material, it could be inferior to other CMS materials if it did not also have a suitable rate of gas adsorption; that is, sufficient to yield, after post treatment, a CMS with gas uptake rates comparable to (though not necessarily equal to) those of commercial CMS. Material which sorbs gas too slowly forces an increase of cycle time in PSA operations, thereby reducing productivity.

We had discovered, however, that if a CMS of sufficiently high oxygen volumetric capacity is used in PSA for air separation, substantially lower gas uptake and selectivity can be tolerated. For example, a CMS having an oxygen volumetric capacity of 10.0 cc/cc or higher at 20.degree. C. and 1 atmosphere will demonstrate a productivity and recovery in a PSA N.sub.2 process equal to or better than known commercial CMS having 15% lower oxygen volumetric capacity (e.g. about 8.5 cc/cc) even though both oxygen uptake rate and selectivity for the higher capacity CMS are each 15% lower than the corresponding properties for the lower capacity CMS. It could not have been foreseen that this could be the case, especially in view of the emphasis heretofore placed on uptake rate and selectivity. These values have been very difficult to raise together. Usually one performance property is improved at the expense of the other. Now, by selecting a CMS which has very high oxygen volumetric capacity of 9.0 cc/cc and above, preferably at least 10.0 cc/cc, equivalent or superior performance in PSA nitrogen production can be achieved with significantly lower uptake and selectivity than possessed by CMS adsorbents typical of those used commercially in this service.

In addition to the importance of adsorption rate and gas capacity, the pore size distribution of a CMS is important. It is desired that a host material with as much micropore volume as possible be used, yet with sufficient meso- or macropores for the transport of gas to the micropores. It is desired for the hydrocarbon deposition procedures described, that the host materials should have micropores below 8 angstroms. Micropores larger than 8 angstroms are more difficult to trim down to the critical 3.8 to 4.2 angstrom size which have been found to be effective gates for oxygen separation from air.

Coconut shells available commercially in the United States originate from several sources, including the Caribbean (primarily Costa Rica and the Dominican Republic) as well as from Hawaii and Singapore. Several varieties of coconut are common, including those having a thin shell of about 1/8 inch thick and a thicker shell of about 1/4 inch, from a football-shaped coconut of the MayPan palm. It has been found that with the proper crushing and sizing, the thickness of the original coconut shell has very little effect upon the quality of the final product. Oxygen adsorption (measured by the CAU method described subsequently) is found to be similar for both the thin shelled coco char and the thicker shelled material, indicating that species and regional variations are of minor consequence to the finished carbon. Dramatic seasonal weather changes (typhoons, droughts, etc.) can impact the quality of the coconut. This can influence the density of the char which is produced. Another parameter of importance, however, which could be affected by the source of shell if purchased as a commodity, is the moisture content. For example, with a moisture content of about 20%, it appears that the water driven off during the heating step reacts with off-gases or impedes their removal during the early stages of pyrolysis, which then permits the off gases to crack and restrict micropore openings. Prior air drying at 110.degree. C. eliminates this difficulty.

During the heat up or ramping stage of the carbonization process, an adequate inert purge gas rate is required in order to form a carbon which has its maximum potential for adsorptive capacity and rate. Nitrogen is the inert gas of choice because of its availability and cost. The flow rate will, of course, depend upon the configuration of the furnace and the amount of shell which is being pyrolyzed at one time. The flow rate must be sufficient to carry away from the granules any pore-plugging decomposition products from the volatilized organic material. With the amounts of shell used for pyrolysis in the Examples, nitrogen flow rates ranging from 0.5 to 7.5 liters per minute were examined and an indication of preferred rate under these conditions is about 3 to 7 liters per minute, enabling the production of a fast selective adsorbent. A flow rate below 1 liter per minute was found to be insufficient to remove off gases produced during pyrolysis. Failure to effectively remove hydrocarbons probably permits them to crack and fill the micropores, and this accounts for the inability of the resulting CMS to reach equilibrium within a reasonable time in gas separation operations. For the most part, in the Examples given, a purge rate of about 6.5 liters per minute was maintained. Care should be taken to sweep the entire rector volume.

As pointed out above, a ramp rate of about 2.degree. to 12.degree. C. and preferably 2.degree. to 10.degree. C. per minute is desired. This is an average ramp rate and the actual rate at which temperature is increased from ambient temperature to the peak pyrolysis temperature can be varied. In fact it is possible to practice the invention by heating the raw shell material to a temperature of about 500.degree. C., holding it at this temperature for a period of time and then increasing the temperature at a suitable ramp rate to the peak pyrolysis temperature. Many combinations of step-wise heating are acceptable. For practical purposes, however, and for ease of control, a steady temperature increase or ramp rate is the most feasible way of operating. If a ramp rate below 2.degree. C./minute is used, the overall heating period becomes inordinately long, while ramp rates much above 12.degree. C. per minute reach the peak pyrolysis temperature too soon and run the risk of pyrolysis of off-gas products during the heating step. To minimize this possibility, the purge rate of inert gas and the heating ramp rate should be coordinated to avoid the buildup of off-gases which could decompose to fill the micropores of the char. In addition to a minimum purge required to remove off-gases produced during pyrolysis and a suitable temperature ramp rate, the final pyrolytic temperature influences oxygen and nitrogen adsorption rates to the greatest extent. Chars produced at temperatures over 900.degree. C. adsorb oxygen and nitrogen more slowly and possess less capacity than those prepared within the selected temperature range, regardless of the purge rate.

Coconut chars prepared as described in the Examples were analyzed by mercury porosimetry and helium pycnometry to obtain pore volume and "pellet" (granule) density. Gravimetric oxygen capacity were converted to volumetric oxygen capacities by using Hg pellet density as determined by mercury porosimetry. A trend was observed toward obtaining greater pellet density and higher capacity as the pyrolytic temperature increased above 650.degree. C. However, this trend appears to plateau at 775.degree. C. so that the superior products were made at temperatures from 775.degree. to 850.degree. C. Such carbons have a higher mercury pellet density coupled with a high gravimetric oxygen capacity; the volumetric capacity has been observed as high as 40% above that for readily available commercial CMS carbons. It is recognized of course that low pellet density in commercial sieves results from pelleting powdered char and as much as 66% of the porosity in a pelleted carbon is macropore volume. Coconut base chars prepared as described below, on the other hand, retain micropore volume similar to those of readily available commercial sieves while containing .about.60% less macropore volume.

The Hg pellet density of the coconut shell chars is considerably greater than the density of commercial CMS or of activated carbons and ranges between about 1.15 and 1.2 grams per cc. Some variation in char density resides with the coconut itself and apparently subtle changes in the coconut char density can reflect climatic conditions during the time the coconut is maturing. Traditional values for bulk (tap density) do not provide sufficient distinctions between materials. On the other hand, Hg pellet density provides an extra measure of distinction and we use this to calculate volumetric O.sub.2 capacities.

After the pyrolysis period the granular coconut shell char is cooled in an inert gas atmosphere. It is desirable to stabilize the char after it has been cooled by passivation by heating in a dry, synthetic air mixture (prepared from H.sub.2 O-free and CO.sub.2 -free O.sub.2 and N.sub.2) at 150.degree. C. for about 15 to 20 minutes. These coconut chars have a highly reactive surface and liberate heat when exposed to ambient air unless passivated. Passified chars stored under dry air or nitrogen are subject to only a slight loss of rate of nitrogen adsorption, whereas if the chars are stored in a humid atmosphere (for example, 50% relative humidity), even though adsorbed water should be completely removed by out-gasing prior to air adsorption during CAU analysis, the rate of nitrogen adsorption can be reduced as much as 60%.

Coconut derived chars prepared at temperatures between 650.degree. and 900.degree. C. exhibit varying degrees of kinetic oxygen selectivity and gas adsorption rates which are attributed to the pyrolytic temperature. As these temperatures are raised, the pore sizes of this char shift below about 4 angstroms and O.sub.2 uptake rate (via CAU) becomes slower. A char prepared at 900.degree. C. while being the most selective, still retains excess adsorption capacity in pores of 4 to 4.3 angstroms compared to a 3.5 angstrom CMS. Oxidative treatments of such chars can then be used to enhance the oxygen capacity although reducing significantly the inherent oxygen selectivity of the original char. Subsequent modifications through hydrocarbon cracking in either a single cracking step or a combination of two steps using two hydrocarbons of different molecular dimensions restores oxygen selectivity and improves the percentage of selective pores. CAU adsorption rates for oxygen and nitrogen on thus modified chars are similar to those of readily available commercial materials. One of the features of the coconut chars described herein is that the predominant micropores are generally not over 8 angstroms and usually not larger than the 4 to 8 angstrom window that is particularly amenable to modification by two-step hydrocarbon pyrolysis.

The pore size of the coconut char can also be modified by varying the time at which the char is held at the peak temperature. In general, shorter hold times on the order of 15 minutes to 1 hour result in faster uptake rates but at approximately the same selectivity as the chars made using a hold time of 4 hours. This would indicate that chars held for less than 1 hour are likely to have smaller effective micro-particle domains (micropore diffusion paths). On the other hand, changing from 2.degree. to 10.degree. C. per minute in ramp rate produces only a marginal increase in adsorption rates. This effect is consistent with the shorter hold times. The use of either nitrogen or argon as the inert gas produces no apparent difference in the chars. Under the same conditions, helium produces a slow but selective adsorbent. The use of carbon dioxide as the purge gas, on the other hand, eliminates oxygen selectivity, although this severe oxidative treatment causes the char to lose only about 5% of its density while increasing its capacity.

After the coconut char has been prepared as described above, it can be further modified by gasification in the presence of either pure carbon dioxide or a mixture of an inert gas and either carbon dioxide, H.sub.2 O or oxygen. In this step, while exposed to the oxidizing gas, the char is heated to above 650.degree. C., preferably 750.degree. C. to 900.degree. C. and held at this temperature for a time sufficient to increase its oxygen capacity while reducing its oxygen selectivity with respect to nitrogen. Treating the chars with a mixture of carbon dioxide and helium increases the adsorption capacity but makes them unselective. This step further increases the oxygen capacity of the coconut chars by about 10% over the original material. The improvement in capacity is about 30% above readily available commercial materials.

The gasification procedure can be modified by impregnating the char beforehand with solutions of materials which serve as catalysts, such as potassium hydroxide, calcium nitrate, calcium acetate or nickel acetate. Additional capacity increase can thereby be obtained after reacting with carbon dioxide. The combination of nickel acetate impregnation and carbon dioxide gasification at 800.degree. to 900.degree. C. is very effective in increasing capacity and speeding adsorption rates for the chars. The use of nickel acetate for gasification at 650.degree. C. with 25% carbon dioxide in helium was also effective. Both capacity and adsorption rates increased. Gasification at 800.degree. C. of the char in pure carbon dioxide to increase its capacity is quite feasible and can be performed in about half the time required using a mixture of 25% carbon dioxide in helium. For example, gasification yields very good results at 800.degree. C. in 25% carbon dioxide in helium for 1 hour or with pure carbon dioxide for 1/2 hour. The addition of inorganic salts to accelerate gasification has a moderate positive impact on the resulting capacity.

After CO.sub.2 oxidation, which imparts additional capacity to the original char, selectivity of the CMS can be restored by exposure of the char to a volatile carbon-containing organic compound, preferably a hydrocarbon such as trimethylcyclohexane for 90 to 135 minutes at 590.degree. to 625.degree. C. A secondary treatment with a similar but smaller compound, such as isobutylene at 500.degree. to 530.degree. C. for 15 to 60 minutes provides further selectivity improvement.

As pointed out above, the granular char from coconut shells is much more economical to produce than the pelleted carbon molecular sieves. In pressure swing adsorption, the adsorbent chamber is subjected to repeated pressurizations which create impacts between bed granules and against the chamber walls during the PSA process. The char is very hard and attrition resistant but it is advantageous to grind the char by tumbling, with or without grit, or a comparable technique, in order to smooth sharp edges that might abrade into powder in adsorbent beds.

Oxygen and nitrogen adsorption properties were determined using a Circulating Adsorption Unit. The Circulating Adsorption Unit (CAU) had a Servomix oxygen monitor, 570A with 311 cell and bypass plumbing to allow 0.5-8 liters per minute flow. This was connected to a Cole Parmer pump, N-7088-48 with the head modified with a controller and high torque motor, (G. K. Heller, GT 21), allowing circulation rate to be varied at varying pressures (0.2-1.0 atm.) while maintaining consistent pump speed at any given rate and pressure. The pump led to a glass cell adsorption unit equipped with a thermocouple. The glass cell, in turn, was connected to the oxygen monitor through an MKS barometer, pressure transducer (#127AA001000A), power supply (#PDR-C-1C).

The response time of the O.sub.2 monitor was 7 seconds to 90% of scale, and the pump was sized to allow circulation rates of 150-7000 cm.sup.3 /min. A compression wave does result from the operation of the single diaphragm pump, therefore it is important to record data at a rate which is fast relative to the pump rate. This was accomplished using a MACSYM computer, Model 120 which was programmed to collect data with adjustable frequency throughout the adsorption run.

The CAU pressure transient is the summation of pressure uptake transients for the individual gas components. Using equations for gravimetric uptake, equations were derived which describe the pressure and % O.sub.2 traces measured on the CAU. System pressure as a function of time is given by the expression:

where;

P.sub.i =initial system pressure

P.sub.O2 =oxygen pressure sorbed at equilibrium

P.sub.N2 =nitrogen pressure sorbed at equilibrium

L,m are mass transfer coefficients for O.sub.2 and N.sub.2 respectively

The % O.sub.2 measured versus time for air (21% O.sub.2) is given by the expression:

Note that P.sub.O2 P.sub.N2 and P.sub.i are measured at t=0 and t=infinity, and can be obtained from the CAU data. The mass transfer coefficients can therefore be obtained by fitting equation 1 to the pressure data or by fitting equation 2 to the % O.sub.2 data. The kinetic selectivity is the ratio of the mass transfer coefficients, L/m. For attractive O.sub.2 selective material for use in PSA operations L should be>3 and selectivity>20.

The amount of O.sub.2 sorbed at short times (1 min) exceeds the equilibrium amount of O.sub.2 sorbed, and gradually decays back to the equilibrium value as N.sub.2 slowly diffuses into the micropores and displaces oxygen. This behavior is not accounted for by eqs. 1 and 2 and they therefore predict a working selectivity that is higher than the actual value. The observed "overshoot" of O.sub.2 adsorption above the equilibrium value, which occurs in the kinetic region of the experiment is a competitive adsorption effect. At short times, when O.sub.2 has largely saturated the adsorbent but N.sub.2 has yet to permeate the adsorbent and approach its adsorptive capacity, O.sub.2 will cover adsorption sites over the entire range of energetics. As N.sub.2 permeates the adsorbent, it displaces much of the O.sub.2 that was sorbed. This occurs owing to the higher heat of adsorption of N.sub.2 over O.sub.2 on CMS carbons at low pressure (.ltoreq.1 atm), and results in the lowest energy state of the adsorbate/adsorbent system at equilibrium. The net effect is that the apparent equilibrium constant for O.sub.2 adsorption is higher in a non-competitive experiment than when O.sub.2 competes with N.sub.2 for sites (which occurs as equilibrium is approached).

An additional term can be added to eqs. 1 and 2 to account for this behavior. Now:

where P.sub.ex is the pressure of O.sub.2 sorbed at short time which exceeds the equilibrium pressure of oxygen sorbed. For attractive O.sub.2 selective CMS materials for use in PSA operations P.sub.ex is usually 3-10 torr. When this additional term is added an excellent fit is obtained, and the selectivity value is in excellent agreement with values determined gravimetrically and volumetrically. P.sub.f, the final pressure reading should be <300 torr for desirable O.sub.2 selective CMS materials in a PSA unit.

The pressure we measure as a function of time reflects all adsorption which occurs, whereas the % O.sub.2 we measure reflects only the selective sorption which occurs. The difference between these measurements represents the unselective adsorption which occurs. By comparing the actual amounts of O.sub.2 and N.sub.2 sorbed at equilibrium (quantities determined by final experimental conditions) with those calculated by fitting the experimental O.sub.2 adsorption data using equation 4 we can quantify the amount of gas sorbed in kinetically selective pores versus the amount of gas sorbed in non-selective pores.

CAU RUN DESCRIPTION

The CAU unit had a total volume of 106 cc of which 27 cc comprised the adsorption cell. To obtain the most accurate results the cell was fully loaded with the carbon (typically 11.5 to 13.9 g depending on the pellet density and size) and outgassed at 110.degree. C. under vacuum until the pressure was less than 0.01 torr. Evacuation was continued for an additional hour. The sample was cooled to room temperature (.about.23.degree. C.) under dynamic vacuum, sealed against ambient atmosphere using a stopcock and transferred to the CAU. The pressure gauge and oxygen monitor were linked to a computer which acquired the raw data. The dead volume of the system was purged with dry air (21.1% O.sub.2 78.9% N.sub.2) for five minutes prior to connecting the cell. The gas mixture was blended to insure that the initial composition was always the same. This is important since these values are fixed in the CAU data reduction program, zero air should not be used since its composition varies. After the cell was connected using compression "O" rings and clamps, the pump speed was set to .about.250 RPM and the pump was started. After three seconds the stopcocks to the absorption cell were simultaneously opened, the operator monitored O.sub.2 composition and decreased the pump rate to <60 rpm when % O.sub.2 stabilized. The computer took 20 readings per second for the first 30 seconds, two per second for the next 90 seconds, one per second for the following eight minutes, and one every four seconds for the balance of the run, which was typically one hour total. A high pump rate, while the adsorbent was adsorbing oxygen rapidly, compared to nitrogen, allowed rapid meter response and did not significantly shift the % O.sub.2 trace. After the maximum O.sub.2 adsorption occurred, the pump rate was decreased, which allowed the adsorption of nitrogen to be monitored with high accuracy and signal/noise ratio. The "noise" in the data was not an artifact of the experiment, rather it showed how pump cycling affected pressure in the different components of the system. One can smooth these points out of the trace, but the raw data give a more true representation of the composition profile. As long as the true position of the trace was evident from the unsmoothed data, the data were fit and used for the plots and calculations.

While not to be bound by theory on how our invention works, we believe we have shown that N.sub.2 PSA process productivity is primarily a function of the overall O.sub.2 working capacity and that recovery is primarily a function of the ratio of O.sub.2 to N.sub.2 overall working capacities. Increasing the equilibrium sorption capacity increases the O.sub.2 working capacity of the CMS and thus the productivity.

Increasing the equilibrium sorption capacity also increases the ratio of the sorbed material to that in the void spaces in the bed (between pellets, in the macropores of the pellets, in the vessels, piping, etc.). Since void space is non-selective capacity, increasing this ratio increases the ratio of the O.sub.2 to N.sub.2 overall working capacities, which in-turn increases the recovery of the process. This is an effect of adsorbent/process interaction, not adsorbent properties alone.

This invention is based on our discovery that the process performance of the N.sub.2 PSA process is more sensitive to the equilibrium capacity of the CMS adsorbent than to the gas uptake rates or the kinetic selectivity. This is unexpected because CMSs separate air on a kinetic, not equilibrium, basis. The sensitivity of recovery to capacity is particularly surprising.

U.S. Pat. No. 4526887 states that capacity is an important factor in the effectiveness of a CMS but makes all claims with regard to rates and selectivity and none on capacity. It would appear that rates and selectivity were viewed as the most important features. In order to recognize the premiere importance of capacity, one must consider the N.sub.2 PSA process as a whole, not just the properties of the CMS.

Other advantages and embodiments of our invention will be apparent to those skilled in the art from the foregoing disclosure and the following claims without departing from the spirit or scope of the invention.