Y = adsorbent 2 = position in column, ft q” = concentration of solute on adsorbent which is in equilib-
rium with Ca References
Chen, J. W., Belter, P. A., French, G. H., “Simulations of Resin Ion Exchange Processes in Agitated Beds,” reprint, AIChE annual meeting, Boston, MA December 1964.
Chen, J. W., Buege, J. A., Cunningham, F. L., Northam, J. I., Ind. Eng. Chem. Process Des. Develop., 7, 26 (1968). Levenspiel, O., “Chemical Reaction Engineering,” Wiley, New York, IVY, 1962. Levenspiel, O., Smith, W. K., Chem. Eng. Sci., 6 , 227 (1957). RECEIVED for review September 9, 1971 ACCEPTEDApril 17, 1972 Presented at the Division of Industrial and Engineering Chemistry, 154th Meeting, American Chemical Society, Chicago, IL, September 1967.
Removal and Recovery of NO, from Nitric Acid Plant Tail Gas by Adsorption on Molecular Sieves Winfried Joithe,’ Alexis T. Bell,’ and Scott Lynn Department of Chemical Engineering, University of California, Berkeley, C A 94720
Adsorption of NO, on molecular sieves offers a convenient method for removing and recovering NO, from the tail gas of a nitric acid plant. This work examined the adsorption of NO, NOz, and NO/N02 mixtures from a dry carrier gas. The results show that molecular sieves have a high capacity for NO*. Mixtures of NO and NO2 can also b e adsorbed provided that oxygen is present in the carrier gas to oxidize the NO. After exposure of the sieves to NO, in a carrier gas saturated with waier they can b e reactivated wiihout a loss of loading capacity for adsorbing NO, from a dry carrier gas provided the reactivation process has desorbed all of the water vapor.The introduction of an adsorption unit into the flow sheet of a nitric acid plant has been considered and the design of such a unit for a 12O-ton/day plant is discussed. Rough economic esiimates suggest that the costs for the unit can b e recovered in two to four years through increased production.
D u r i n g the production of nitric acid (Chilton, 1968), nitrogen dioxide is absorbed in water by the process 3x02
+ H20
+
2HK03
+ NO
(1)
The nitric oxide formed is reoxidized to nitrogen dioxide by air introduced into the system for this purpose. Absorption and oxidation continue up the absorption tower until the gas stream contains only a few tenths of 1% of the oxides of nitrogen (NO,) a t which point further recovery becomes uneconomical. The tail gas, having a composition similar to that given in Table I, must be treated to recover or destroy the SO, before the gas is released to the atmosphere. The economics of tail gas treatment suggests t h a t recovery of the nitrogen oxides in a form which could be recycled to the absorber would represent the most favorable process. Absorption and adsorption are the only feasible means for achieving a recovery of SO,, and absorption does not appear to be practical a t the concentration levels in question. -4 number of solid materials have been reported in the literature as adsorbents for oxides of nitrogen (Bartok et al., 1969). Those materials on which a reversible adsorption has been obtained include silica gel, alumina, char, and molecular sieves. Of these materials only char and molecular sieves have Present address, Deutsche Shell A.G., Hamburg, West Germany. To whom correspondence should be addressed. 434
Ind. Eng. Chem. Process Des. Develop., Vol. 1 1 , No. 3, 1972
Table 1. Characteristic of Nitric Acid Plant Tail Gas Physical conditions: Pressure = 7.26 atm; temp = 30°C; mass flow rate = 41,200 lb/hr Species
NO NO2 HzO 0 2 h’2
HiiOo
Concn, vol
yo
0.1 0.15 0.6 3.0 96.15 =o. 00
a sufficiently high adsorption capacity to warrant serious consideration. The use of char as a n adsorbent for nitrogen oxides has not been considered extensively because of potential fire or explosion hazards (Ganz, 1958). Molecular sieves have shown a high adsorption capacity for nitrogen dioxide but essentially no capacity for adsorbing nitric oxide by itself. I n the presence of oxygen, though, molecular sieves can catalyze the oxidation of nitric oxide to nitrogen dioxide which will subsequently adsorb. A recent report of Sundaresan et al. (1967) has discussed the use of molecular sieves for the treatment of nitric acid tail gas. I n these studies the tail gas was simulated by a nitrogen stream containing 3.5y0oxygen to which was added 1800-2600 ppm of nitric oxide. h synthetic zeolite was used
as the adsorbent. The performance of the adsorbent was tested by cycling it through periods of adsorption followed by desorption using either hot air or steam and hot air as a purge. Nitric oxides were recovered as nitric acid and nitrogen oxides enriched in NO?. The experimental results showed a n initial adsorption capacity of 5-6 Ib NO, removed/100 Ib bed which decreased after 11 cycles to 2-3 lb NOz removed,' 100 lb bed. Regeneration of the bed with steam enhanced the recovery of nitrogen oxides as nitric acid but did not curtail the falloff in activity. Several more recent studies have been carried out on the adsorption of nitrogen dioxide alone on a variety of molecular sieves. Krasnyy et al. (1969) examined the effect of SiOz/ A1203 ratio on the adsorption of NO*. Dealuminated mordenite with a n Si02,'h1z03 ratio of 13: 1 was found to have the highest adsorption capacity a t lorn relative pressures of NOz. More highly dealuminated mordenites and commercial acid-resistant zeolites had lower capacities. Similar observations were obtained in a study by Lewis (1970). The present studies were initiated to examine t h e adsorption of mixed nitric oxides from a dry carrier gas. Primary objectives were to determine whether the absence of water vapor would lead to higher adsorption capacities as well as a greater stability in the capacity after repeated cycles. 4 n additional objective was to determine the relative merits of different types of molecular sieves. Finally, the results of the investigation were used to design a system for recovering NO, from the tail gas of a nitric acid plant.
U
Figure 1. Schematic of experimental apparatus
Experimental Apparatus and Procedure
The experimental apparatus used in this study is shown in Figure 1. The system consists of a gas supply manifold, a n adsorption bed, and provisions for sampling the gases entering and leaving the bed. The adsorption column was made from a 1-in. i.d. glass tube 3 ft in length, and was filled to a height of 1 ft with l/Io-in. pellets of molecular sieves. Linde 13X Molecular Sieves were used in most of the experiments. Norton Zeolon 200 Molecular Sieves were used in one. During the desorption of the nitric oxides the column was heated externally with an electrical heating tape. The carrier gas for the nitrogen oxides was a mixture of dry nitrogen and oxygen. The moisture content of both gases was about 3 ppm. To this stream were added measured amounts of nitric oxide and nitrogen dioxide. The nitric oxide was supplied directly from a tank of the compressed gas, whereas the nitrogen dioxide was introduced by passing a portion of the nicrogen stream through a bottle containing liquid Nz04a t 0°C. For those runs in which it was desired to produce a moist gas stream, the carrier gas was saturated with water vapor before being mixed with the nitrogen oxide streams. The flow rate of the nitrogen was measured with a capillary flowmeter while the flow rates of oxygen and nitrogen oxides were measured n i t h rotameters. The rotameters on t h e nitrogen oxide streams were calibrated by analyzing the gas collected in a n evacuated bottle over a fixed period of time. The temperatures a t the bed inlet and outlet as well as the humidifier outlet were measured by thermometers. The gas pressures upstream of the nitrogen flowmeter and a t t h e inlet to the adsorption bed were measured by manometers. Determination of the total nitrogen oxides content of the inlet and outlet streams was performed by the hydrogen peroxide method (Faucett et al., 1966). For this purpose a known volume of the gas to be analyzed was absorbed in a
Figure 2. Gas sampling assemhly
3 wt % solution of hydrogen peroxide. Both of the nitrogen oxides were thus converted to nitric acid by the reactions
+ H202 2HN03 + 3H202 2"03 + H20
2N02 2N0
-+
-+
The nitric acid produced was then determined by titration. Samples mere withdrawn into evacuated bottles containing a small amount of H202using the device illustrated in Figure 2. When the sample bottle had been filled, the temperature and pressure of the sample were recorded. The accumulated gas sample and the hydrogen peroxide solutions mere shaken and subsequently left to stay in contact overnight to assure complete absorption of the nitric oxides. The nitric acid formed was titrated with a 0.005N sodium hydroxide solution using a p H meter to detect the end point. After each adsorption run, the bed was reactivated by passing air preheated to 130°C through the bed. Additional heat was supplied to the bed directly from heating tape wrapped around the column. The exit gas temperature was allowed to rise to the inlet temperature. =It this point the inlet temperature was increased to 15OOC and the hot gas flow continued until the outlet gas again reached the inlet temperature. Heating was then stopped and the column was allowed to cool down in the presence of a flow of pure nitrogen. Results and Discussion
Six adsorption runs were performed using Linde 13X Molecular Sieves and one using Norton Zeolon 200 Molecular Ind. Eng. Chem. Process Des. Develop., Vol. 1 1 , No. 3, 1972
435
Table II. Conditions and Results of Experimental Runs Run lb
II
111,
lllb
IV
V
VI
VI1
Adsorption temp Tb, "c 21.2 23.5 25.0 22.5 24.3 22.8 23.7 23.9 23.7 ildsorption press. p , , torr 756.7 751.1 753.4 752.8 754.7 757.9 751.5 758.9 762.3 Gas velocity u9, ft/min 47.06 47.43 47.88 47.52 47.69 47.25 47.79 47.40 47.12 Inlet concn Y N O ~ % , 0.371 0.362 0.257 0.211 0.199 0.517 0.192 0.192 0.198 Oxidation ratio r, Yo 0.00 0.00 100.0 37.0 36.8 28.0 36.6 36.8 36.8 Loading capacity q, lb KO,/lOO lb ads 7.59 7.41 17.50 11.71 10.49 ... 8.23 10.15 5.83 Lengt.h of unused bed LUB, ft 0.780 0,780 0.424 0.413 0.411 ... 0.518 0.391 0.424 Temperature differential across bed AT,, "C 2.6 2.6 2.2 1 .o 1.0 0.8 1.6 1.2 1.2
Tlma (hr)
Figure 3. Breakthrough curves for the adsorption of NO and NO2in the presence of O 2
Figure 4. Breakthrough curves for the adsorption of NO/ NOzmixtures in the presence of 0,
Sieves. The experimental conditions for these runs as well as the derived results are summarized in Table 11. The oxidation ratio r is defined as the mole fraction of entering XO, present as SOz whereas the length of unused bed is defined as half the distance of the adsorption front in the bed. Breakthrough curves for all of the runs are illustrated in Figures 3-5. Runs I and I1 were performed to examine the extent of adsorption of nitric oxide and nitrogen dioxide, respectively, from a carrier gas of nitrogen containing 3% oxygen. The breakthrough curve for nitric oxide shown in Figure 3 (run I ) displays a n abrupt beginning followed by a period of decreasing slope and finally a n approach to saturation. This unusual shape was verified in a second run under essentially identical conditions. Visual observation of the effluent gas showed a distinct brown color indicative of nitrogen dioxide. Since Lewis (1970) has shown t h a t nitric oxide in the absence of oxygen nil1 not adsorb on l\lolecular Sieves, it is felt t h a t the observed retention of nitric oxide is due to its oxidation and subsequent adsorption as NO2. I n contrast to nitric oxide, nitrogen dioxide (run 11) gave a sigmoid breakthrough curve characteristic of Langmuir adsorption. The coadsorptioii of nitric oxide and nitrogen dioxide (run 111) was performed to simulate the treatment of dried
tail gas obtained in a nitric acid plant. T n ~ oruns were performed to verify the results. Figure 4 shows that the breakthrough curves for the two runs are essentially identical as are the loading capacity and LUB tabulated in Table 11. The shape of the breakthrough curves closely resembles t h a t for the adsorption of nitrogen dioxide alone even though only 37'70 of the NO, in the adsorbing gas was XO2 in this case. Visual observation of the bed showed t h a t a yellow-green band formed at the bottom of the bed and traveled slowly in the direction of flow. The length of this band was about 7.5 in. with the brightest color appearing in the middle of the band and diminishing a t its boundaries. The part of the bed directly above the band showed a slight darkening while the part passed by regained its original color. Breakthrough occurred at essentially the same time that the upper boundary of the band reached the end of the bed. The results of run 111 clearly indicate that the presence of XO2 aids the adsorption of XO. The process by which this takes place might involve the adsorption of NO2 which in turn acts as a solvent for NO, binding it in the form of N203 which is subsequently oxidized. The presence of a yellowgreen band suggests t h a t Nz03 is indeed formed and is the precursor to the final adsorption of nitric oxide. It should be
436
Ind. Eng. Chem. Process Des. Develop., Vol. 1 1 , No. 3, 1972
0
,I
2
3
4
5
6
7
8
9
IO
11
12
Tim (hr)
Figure 5. Breakthrough curve for the adsorption of a NO/ NO, mixture in the absence of O2
noted that Foster and Daniels (1951) have noticed a similar effect of NO2 on the adsorption of NO on silica gel. T o examine further the interaction between NO and KOZ, their adsorption was studied in the absence of oxygen (run IV), Figure 5 illustrates the effluent composition curve obtained in this case, which is characterized by a rather sudden breakthrough followed by a leveling off to 67y0 of the inlet concentration. h yellow-green zone formed a t the bed inlet and moved through the bed as a continuous front. The intensity of the color was weaker than in run 111, and the velocity of the front was slower than in the previous experiments. I n this experiment, because of the lack of oxygen in the carrier gas, it is assumed t h a t only X02 itself and NO in the form of N203 could be adsorbed. At its plateau the breakthrough curve indicates t h a t 33% of t h e inlet nitrogen oxides are adsorbed. If all of the SO2 is adsorbed during this period, 28% of the XO, is accounted for. The remaining 5% must be due t o NO weakly bound as N z O ~ . Runs V and VI were undertaken to examine the reduction in loading capacity which might be caused by the exposure of the molecular sieves to a stream containing nitrogen oxides and water vapor. Each of the experiments consisted of three parts: (1) Adsorption of water vapor and nitrogen oxides from a moisture-saturated carrier gas (2) Reactivation of the bed (3) Adsorption of nitrogen oxides from a dry carrier gas as in run I11
During run V the bed was reactivated t o a maximum temperature of 150°C while in run VI this was increased to 250OC. Exposure of the sieves to water vapor caused a dark grey zone to form at the bed entrance and then spread into the bed. I n a short layer above the bed entrance the pellets developed small cracks. The extent and number of these cracks increased during reactivation, but the layer containing these pellets never exceeded "4 in. The adsorption of nitrogen oxides from a dry carrier gas after reactivation of the bed produced breakthrough curves (Figure 4) and visual effects similar to those in run 111. The results of run V showed a 21.5% reduction in the loading capacity and 28.5y0 increase in the LCB compared to run 111. This suggests t h a t the desorption temperature of 15OOC was insufficient to remove the last traces of adsorbed water. The equilibrium isobar for water on molecular sieves a t 15OOC and 1 a t m is 6 wt %, a relatively high value in comparison to the capacity for nitrogen oxides. During the reactivation of the bed for run VI the desorption temperature was maintained a t 250OC. Figure 4 and Table I1 indicate t h a t this higher temperature was successful in restoring the initial adsorption capacity. Discrepancies between the results for runs VI and I11 are well within the limits of experimental error. R u n VI1 was performed to compare the loading capacity of Sorton and Linde Molecular Sieves. Figure 4 and Table I1 show the results of this experiment. The shape of the breakthrough curve for Norton Sieves is similar to t h a t for Linde 1 3 X . However, the loading capacity is 437, lower and the LUB is about 3 7 . higher. Visual observation of the bed showed a red-brown front developing a t the entrance of the bed l / 2 hr after the initiation of the run. This front moved u p the bed relatively rapidly, forming a continuous zone with a n increasing coloration toward the entrance of the bed. The decreased adsorption of the S o r t o n Sieves in comparison to the Linde Sieves is consistent with the earlier observations of Krasnyy et al. (1969) and Lewis (1970). Two factors might be responsible for the observed difference. The first is the smaller effective pore diameter of the Norton Sieves (8A instead of lOA). The second factor is their lower pore surface area. The desorption of nitrogen oxides from Linde Molecular Sieves was studied in one experiment to determine the time and temperature dependence of desorption. The conditions
Figure 6. Effluent concentration of NO, during the regeneration of the adsorption column
Ind. Eng. Chem. Process Der. Develop., Vol. 1 1 , No. 3, 1972
437
Figure 7. Flow sheet for a 120-ton/day nitric acid plant
U slack
of adsorption were those for run V. Figure 6 shows the concentration vs. time profile which was obtained. The most striking features of the desorption curve are the two sharp peaks reflecting changes in the temperature of the bed. The first peak is owing to the initial contact of the desorbing airstream a t 130°C with the cold bed. After a period of 1 hr, the rate of gas desorption decreases and levels t o a steady value. At this point heat was added to the column from the heating tape to obtain a n exit gas temperature of 130°C. This procedure produced the second desorption peak. Evideiice of an almost complete desorption is the absence of a third 1)eak \\-hen the exit gas temperature was raised to 150°C. A determiiiatioii of the amount of gas desorbed shows that 90.27, of the initially adsorbed gas can be accounted for in the desorptioii experimeiit. Design of a n Adsorption Process for the Recovery of NOn
The flowsheet for a modern nitric acid plant is shown in Figure 7 (Chiltoii, 1968), and is largely self-explanatory. The system is operated at a pressure of about 7 atm to facilitate the absorptioii step. Uefore discharge to the atmosphere the tail gas is heated by exchange with the combustion gas from the ammonia burlier and expanded through a turbiiie to provide the energy for compressing the air used in the process. T o incorporate ail adsorption process to recover S O , from the tail gas, one must provide for (a) a n adsorption step in which the tail gas is dry and cool (