The Progress of the Reactions in a Portland Cement Kiln, and A Method for the Production of Anhydrous Sodium Sulfate Citation Shirley, Howard Emerson (1931) The Progress of the Reactions in a Portland Cement Kiln, and A Method for the Production of Anhydrous Sodium Sulfate. Master's thesis, California Institute of Technology. doi:10.7907/vzj2-x573. https://resolver.caltech.edu/CaltechTHESIS:09062024-185453476 Abstract No abstract. Item Type: Thesis (Master's thesis) Subject Keywords: (Chemical Engineering) Degree Grantor: California Institute of Technology Division: Chemistry and Chemical Engineering Major Option: Chemical Engineering Thesis Availability: Public (worldwide access) Research Advisor(s): Lacey, William Noble Thesis Committee: Unknown, Unknown Defense Date: 1 January 1931 Record Number: CaltechTHESIS:09062024-185453476 Persistent URL: https://resolver.caltech.edu/CaltechTHESIS:09062024-185453476 DOI: 10.7907/vzj2-x573 Default Usage Policy: No commercial reproduction, distribution, display or performance rights in this work are provided. ID Code: 16699 Collection: CaltechTHESIS Deposited By: Benjamin Perez Deposited On: 06 Sep 2024 21:43 Last Modified: 06 Sep 2024 22:00 Thesis Files PDF - Final Version See Usage Policy. 14MB Repository Staff Only: item control page

THE PROGRESS OF THE REACTIONS IN A PORTLA...111D CID.mTT KILN and A :METHOD FOR TEE PRODUCTIOlil' OF ANHYDROUS SODIUM SULFATE Thesis by Howard E. Shirley In ~artial fulfillment of the re4uirements for the degree of l{aster of Science in Chemical Engineering CAI,IFORlifIA INSTITUTE OF TECirn'OLOGY Pasadena, Cali~ornia 1931 . I. PART A TEE PROGRESS OF THE REACTIONS IN A PORTLAND CEMENT ICTLN -2.- The Progress o! the Reactions in a Portland Cement Kiln The scientific investigation o! cement may be said to ( 1) ( 2) have begun in 181:, with the work o! Oollet-Descotils . who showed that the burning 0t the mix converts the silica into a .soluble torm. Le Chatel1er(:,)1n 1887 was the first to try to classify the definite chemical compounds ot importance in Portland Cement. He advanced the theory that Portland Oement 1s composed essentially of two compounds: :,oao.A.12 0, and :,cao.s10-z • Using his experimental facts as a basis he calculated the ratio between Lime and Silica-Alumina th.at would give a good cement. Though Researchers since have shown that his ratio gives too high a Lime content for soundness, most author~ties of the present agree th.at :,oao.s10 2 ia the most important single component of Portland Cement. Other Investigators fro~ time to time rejected old theories and advanced new ones among whioh lDS3' be mentioned the following: Tornbohm (l897)(l)tound four distinct mineral constituents ~ 1n Portland Cement which he called Al.it, Belit, Felit, and Oelit. He identified these by means of individual oolor and their refractive power on polarized light. composition. He did not determine their chemical o. Richardson in l904~ 4>< 5lnrougn mie~o- scopic studies on the clinker, suggested the theory that it was a · solid solution composed chiefly ot 3CaO.Si0 2 dissolved in 3CaO.AJ. 2 0~. this 1s not in accord with the present theories. Rank1n~ 6 in 1915, approached the problem in a ditterent way. Since S102 , OaO, and A.1 1 0~ are the most important oonstituente ot Portland Cement, he and his oo• wo:rkers undertook to study the phase relations of these compounds and to determine the various ratios in which thes~ constituents interacted. the results ot tbe various investigations may be summed up as tollows: The important compounds present in colll1Jlerc1al Portland Cement clinker are, . J(gO, 40aO.Fei,03 .Al 2 0 3 , 20a0.Si0 2 , 3Ca0. s102 , :,aao. Ali 03 , and CaO. (6) Rankin and Wrigh~ have identified all these except the Iron OoDlpound by means of their optical properties, •ven in the presenc~ ot 1Dlpur1t1es such as )(gO and Fe2 ~ In the formation ot the olinke:r, the following reaotions are the most important and are assumed to take place roughly iii the order given. Undoubtedly they overlap and several may be proceeding at the same time. 1. . Removal ot tree mo1eture and tb.e. dehydration ot the. combined water in the clay. 2. Decomposition ot the Mg003 and OaO~ to MgO and oao. 3. Reaction ot Cao with ~ 0.3' and Fe;i03 to give, 40aO. Ali 03 .Fez 03 • 4, Reaction of OaO with 8102 to give 2CaO.SiOz• -4- 5. Reaction ot Alz.03 and Cao to g1ve 30a0.Al.i0:3. 6. Reaction ot 20aO.S10 2 and Cao to give 3eao.s102 • Th1s order 1s 1n accord with the latest f1nd1ngs ot the investigators in the Department ot Commerce who further find that MgO 1a present 1n clinker uncombinea. The reactio~ ot D1•Calc1um Silicate and Lime to give Tri• Oalc1wn S111~ate 1s slow and the eharge must be held at a h1gh temperature tor considerable time 1n order that the reaction may go to ~ompletion since the tree lime in the finished cement a . dttrimental to the soundness ot the concrete. The 40aO.Fe 2 ~Al~03 acts as a flux for the fusion or the minerals and lowers the temperature at which th.e reaction takes place. The present study was undertake~ 1n order to follow the progress ot these reactions in a commercially operated Portland Cement Kiln of the rotary type using the dry proceas. A JIBthod. hara recently been worked out by w. N, Lacey an~ Hubert Woods< 7>at the Riverside Cement Oompa.ror ot Los Angeles whereby samples mq be taken from tile kiln while it 1s 1n continuous operation. Using samples obtained trom a klln of this Oompal\V, determination~ were made ot the water, 002 , and Free Lime (uncombined OaO) content ot the charge t~ougb.out the length ot the kiln. From thee~ Data reactions land 2 of the above 11st are obtained dir,ctly. The progress ot 3, 4, 5, and 6 was followed by calculating the ut111zat1on at L1meft-$l)l!l the oo, and the -s- Free Lime data. A close visual examination of the samples was made to determine at what place in the kiln the mix began to 11 ball up" or form lumps. The spee1fie ·gravity ot the samples was al.so determined. While this method has been used in the laboratory tor following the reactions, it is the writer's opinion that this 1s the first tim.e it has ever been applied to a eont1nuously operating kiln. Figure .1 gives the dimensions ot the kiln and the · locations ot the sampling Jtat1ons. These samples are composites of ones taken every halt hour ove~ a five nour period. As they were taken from the kiln they were placed in large tin sample containers provided with tightly fitting covers. After the run these covers were sealed with pa~affin until the sample was removed to be ground~ After grindi:cg·.to pass a sixty mesh screen it was well mixed, quartered, and ene quarter placed in a small., tightly covered, tin sample can, which, together with th~ remaining sample, was returned to the large can and ag.a in sealed with paraffin. After a sampl_e was with- drawn tro~ the small can for analysis the 88.!Jlple container was agai?l returne4 to the ·large oan and .resealed. Experimental MethodJ The determ1pat1ons of water, CO2 , and loss on ignition we:re made in a gas i'ireq. Qombustion tube furnace previded with necessary purifying train and absorbe:rs. 1s shown in Figure 2. The set up The combustion tube was of silica, 7/8" inside diameter, l & 1/an outside diameter, It was heated with. three Fisher burners using a natural gas and air mixture. zone was 9 inches long. This heated The tube was protected from the direct flame 'by a closely fitting cover of sheet nickel. However, in spite ot this, the tube slowly devitrified and after six months o-t lized. continuous use was badly crystal- Des1cchlera (anhydrous Barium Perchlorate) was used to absorb the water. Ascarite (asbestos- impregnated with NaOH) was the absor.'bent tor CO 2 • The absorbing train was composed ot four Fisher Absorption bottles, the first two filled witll deaicchlora followed 'by two more conta1n-1ng a mixture ot ascarite and desicchlora. The union between the first absorber and the silica tube was made by pulling ·down the end ot the large tube and sealing a small clear quartz tube to this. The end of the small tube was ground to fit into the side tube of the first absorber. A tight fit was not desired, the union being made air•·t ight by a piece of pure India rubber tubing. The other end of the silica tube was connected to the purifying train by a long rubber stopper, one end of this titting the end of the tube, the other end titting into a pieoe of large·.· i-ubber tubing which was taped around the silica tube. Runs made to determine the temperature ot the end of the stopper exposed to the radiated heat from the interior of tube showed that 1t attained a maximum temperature ot 40• o. -e· The temperature ot the air coming trom the exit end of the . tube was found to be 38 C for the rate of approximately 2 liters per hour used in the determinations. Since this is safely below the decomposition temperature of the Hydrate of Barium Perclorate (120 C) no difficulties were experienced from this source. Blanks were run at frequent intervals and it was 1'011:nd that using a preignited Alundum boat and an occasional shaking of the purifying train to prevent channeling, no corrections were necessary for foreign water or CO 2 • The method ~f making a Determination was as follows: The absoi-bers were carefully wiped with a clean, dry, lint• less towel and weighed. ignition train. They were then inserted in the A sample of the clinker was then weighed into a previously ignited Alundum boat and placed in the heating zone of the combustion tube with the help of a long wire. The tube was closed, stopcocks opened, and the water aspirator adjusted so that air at the rate of approximately 2 liters per hour was pulled through the train. of 15mm of Mercury was necessary for this. A vacuum The tube was then brought up to lOOC>O in the course of an hour. Temper- atures were measured with a ohromel- aluiuel couple. It was found.- that four hours were necessary for the CO 2 to be completely liberated and swept out of the tube. The Aspirator was then disconnected, the absorbers allowed to come up to -9- atmospheric pressure and then taken down and reweighed, the increase in weight being the C02 and the R2 0. was allowed to cool in the silica tube. The Boat It was then re- moved, placed in its weighing bottle and weighed. The loss in weight was reported as Loss on Ignition. Specific " gravities were determined with a standard Le Ohatelier Specific Gravity Bottle. The determinations of the free .. lime were kindly carried out by the Riverside Cement Company. In order to calculate the Recombination af'C~ o . . it was necessary to determine the COz which came from - MgOO~ • Using a method described by Mead.e(!)the mean of several determinations gave a value of 3.9% MgCO, in the feed. The OOz from this amount of Mg00.3 isl.~. In the eal• . culations of Cao liberated from the carlonates decomposed, this amount of 09z has been taken into account as coming from the Mg00 3 before ar11 CaCO, decomposed. Since the relative amount of Ca003 to MgOO~is large this assumption causes little error in the calculations. - ,oResults and Discuss-ion The experimental results are given in TaQle I. The Loss on Ignition, 002 absorbed, and HzO data were checked to the nearest o.1%. Tne check Dl1nat1.o ns agreed to 0.003. specific gravity deter- Figure :, shows .g raphically the changes in these values th.rough the length at the kiln. · The H2 0 content ot the mix is small and th.is data has been replotted in Figure 3b on a different scale in order that the variations may be more clearly seen. It is seen that there is a close agreement . between the 00 2 Absorbed ancL the Loss on Ignition at all sampling points of the kiln. This is to be expected near the dis- charge end of the kiln where little water is left, but since it was thought that the ignition loss was merely the sum of the H2 0 and 001 driven ott, a greater difference was expected at the head of the kiln. Oolumn 7 of Table I shows how much variation there is between the Ignition lose and the total ot the CO 2 and H2 0. The faot that this difference is negative in the ease of the teed and positive elsewhere leads to the supposition that there was some material in the feed taking up Oxygen. A cylinder of Nitrogen was secured and two determinations were made on the feed and on sample eight using Nitrogen instead of Air to sweep out the C01 and H1 0. in Table II. The results are shown TABLE I ~ ... -~ '\l Description of General Appearance QI St1e~1lil Los::i "" r 9a;l,on ~ I... ~ ~'o "' ~ ~~l.... ---., (:) ~ i -<:) ~ ~ Fr~e.. c.l) O]IJ ~~ ~ Feed Light gray, soft, not gritty. 2.744 31.8 31.3 1.4 -Oo9 "'" 0.9 Light gray, soft, fine, not gritty. 2.744 30.7 30.3 0.4 o.o 1.4 Light tan, definite change from 12, not gritty. 2.772 26.3 26.1 0.2 o.o 6.7 Light t .an srmll lumps present, slight tendency to form balls. 2.793 23.1 22.7 0.1 0.3 11.7 Slightly darker than 10. Larger lumps, very slight difference in color between t hese lumps and mix. 2.888 15.4 14.8 0.1 0.5 19.6 First visible evidence of clinker formation. Large lumps quite dark brown in center. Mix has brown specks. Gritty. 2.890 9.4 8.9 0.2 0. 3 26.l Large lumps of brown,gritty material. No evidence of fusion. Lumps easily broken. Tan color present. 3.040 6.4 6.0 0.2 0.2 26.3 Black specks plentiful. Tan color predominant. Lumps easily broken. Center of lump dark brown. No evidence of :fusion. 3.080 3.2 2.6 0.1 0.5 28.2 Definite colo r change :from 6. Mix dark gray. Lumps hard and homogenous. Apparently partially fused. Some tan lumps still present. 3.202 0.4 0.2 - 0.2 5.1 Slightly darker than 5. Tan particles covering some of the softer lumps. Very few fines. Particles mostly show no evidence of fusion. 3.202 0.3 0.1 - 0.2 4.2 More fine than 4. Particles crumble under pressure. More evidence of fusion. Slight tra ce of t an. 3 . 178 0. 3 0.1 - 0.2 2.. 7 I)_ £ l: ~ C,,-417;7 """Z 12 11 10 9 8 7 6 5 1 4 3 + ...... <:::S H\ h v ~~ . Lint{! ., 2 ' •• \. "' . -- . .. ' Q.. I: ...E f Cl) 2 1 I Sp,c., ·-.;c L,ss -::,.Lo~ H~o GNvity IjT ni+,o,, ' (.01, Lim~ I .,, '4 Description of General .A1.:,r~ ea ranee ,.. 9 cSJt> ir<-c. m:n ....s Of( i '?o . I '](> \ ~o..?1(0 Particles are fused. Fracture on applied pressure. ,Vecy hard. Quite dark in color. Frantures present uniform appearance 3.158 0.3 0.1 - 0.2 0.3 • Very hard. Tan particles not present. Particles fracture on applied pressure. Color is greenish gray 3.187 0.2 - - 0.2 0.1 ; - - 0.2 0.1 : -l. Particles scratch glass. q.. ~ . t" ~ ~ Fracture shows uniform oolor. Very small pores present . Color is dark p;reen. 3.190 - 0.1 - ,s- Table II Sample Ig,Loss in Air lg.Loss in Nitrogen 00z Hz.O Feed 31,Tf, 32.2% 31.3% 1.1% -0.2 Feed ,1.8 32.2 :,1.2 1.2 -0.2 8 9.4 9.3 8.9 0.2 0.2 a 9,4 9.2 8.8 0.2 0.2 Ditferenee between Ig. Loss & 001 -t Hz.O It is seen that using Nitrogen, the difference in the total 00~ and H1 0 and the Loss en Ignition for the feed is reduced to -0.2. This tends to strengthen the supposition that Oxygen is absorbed by the feed if the heating is carr~ed out in air. Since the Nitrogen was obtained from liquid air, the negative difference here m.ay be due to some Oxygen present as impurity, The positive difference is again obtained with Sample 8, although less than before. A series of determinations of Ignition Losses were then made in air in order to study the rate of loss over various periods of time. Platinum crucibles were used and the blast burners were not changed after the run was started. are given 1Q Table III. The results -/6- Table III sample Time of Heati;pg Loss in 12 12 hours 10 12 8 6 Feed " " It " Sample Ttme ot Heating Lose 1n 31.2 6 15 minutes 2.48 " 23.5 6 30 " 2~79 12 " 9.9 6 l hour 3.04 12 H 4.1 6 1-i hours 3.12 20 mi~utes 31~3 6 2i " ,.14 l hour 32.0 6 It½ " 3.20 2t hours :,2.2 4 15 minutes 0~215 5i It 32.:, 4 45 " 32.4 4 it ~o~e o.46:, 4 ,t II 0.491 4 4¼ " 0.500 16 '/. 1o - - It - 0.229 These data indicate that the ignition loss may be increaa.ed greatly by increasing the time of ignition. Figure 4 gives graphically the rate at which the various samples lose. weight on ignition. The loss after 4 and i hours is taken as 1001', 1n each case. The H2 0 absorbed curve in figure 3b shows that water 1s driven out of' the mix very rapidly. The material con• tains very little free moisture; holding a sample of the feed at 105°0 tor 5 days gave a loss of' 0.091<,. The slight rise at Stations 8 and 7 1s attributed to moisture taken up after the samples had been ground • .work done on samples which had been in, tightly covered tin sample cans (not sealed) for a year show that 1n this region ot the kiln the eamples have a water content ot around 9.0% while the teed runs only 1.5%. The data 1n Table IV indicate that the charge is highly n,grosc.op1c throughout the length of the kiln. Table IV Sample Initial Losa Loss on Ign1on Ignition tion 4 month• ,, D1tferencse late~ Free Lime Content ,, F~ed :,1.8 ,1..1~ -0.1 0.9 10 23.1 2:,.,4 11.7 9 15.4 17,2 0.3 1,8 8 9.4 12 .. 0 2.6 26.l 7 6.4 10.:, :,.9 26.:, 6 :,.2 8.2 s.o 28.2 4 Q.:, 1.,4 2 o.:, 1~7 o,4 0.1 4.2 o.:, Clinker 0,1 o.6 0.5 0.1 19.6 -19- Table V give$ the experimental data converted into pounds of material at tl::B different stations of the kiln based on 100 pounds of teed. This conversion is made on the assumption that the non-volatile material, as determined in the feed, travels through the kiln with practically no loss or gain in weight. Thus, the weight of the original 100 pound charge of teed at any point in the kiln can be calculated from the ignition loss at that point and the ignition loss of the teed. Direct multiplication of this weight by the percentage of the COt and the CaO present gives the quantities of these compounds present at the various stations. The decrease in OOz content b'etween the teed and the charges at the stations corrected for the 1.9 pounds attributed to th~ JlgCO is converted into CaO and this quantity plus the initial OaO determined in the teed is tabulated as "cao Liberated". The difference between this figure and the 11 0a0 Determined" at that station is given as "oao Recombined". The last three columns of the table give the per cent ot the total CaO from Carbonates that has been liberated, the per cent of the OaO liberated that has recombined and the per cent total CaO from Oarponate that has recombined. In Figure 5 and 6 these re- sults have been plotted against length of kiln. It is seen from Figure 5 that the rate of OaO liberation is rapid and practically constant until it is coiqpleted about twenty feet from the discharge end of the kiln. The recom• TABLE V Feed 100 31.3 12 98.5 29.8 11 92.7 24.2 7.8 6.2 1.6 20.4 20.5 10 as.a 20.5 13.5 10.4 3.1 35.2 22.9 8.l 9 80.7 12.4 22.5 15.8 6.7 58.8 29.8 17.5 · 8 75.4 7.1 29.3 19.7 9.6 76.5 32.8 25.1 7 73.0 4.7 32.3 19.2 13.1 84.4 40.7 34.2 6 70.6 2.3 35.4 19.9 15.5 92.5 43~8 40.5 5 68.6 0.14 38.1 3.5 34.6 99.5 90.7 90.4 4 68.5 0.1 38.2 92.5 92.1 3 68.5 0.1 38.2 1.0 36.4 99.7 95.3 95.0 2 68.5 38.3 0.2 38.1100.0 99.5 99.5 1 68.4 38.3 0.1 38.2 100.0 99.8 99.8 Clinker 68.3 38.3 0.1 38.2 100.0 99.8 99.8 "."' -22- bination of the lime 1s slower at the start but the reaction increases more rapidly, much like an autoeatalytio one. It slows down after about 91% of the Cao has reacted but continues at an appreciable rate until within seven feet of the mouth of the kiln. This final twenty feet is evidently the portion of the kiln in which most of the reaction between 2cao.s1O2. and Cao takes place. The specific gravity change~ ot the o~ge are also shown in Figure a. The curve follows the Liberation ot Cao, reaching a maximum at the point where the CaO is a maximum. The following sharp and decided fall in the eurve has not been explained.. It may be due to the conversio~ of the "p" 20aO.SiO1 into the "y" 20aO.SiO2 when the sample cooled after being withdrawn from the kiln. This change is accompanied by an increase in volume of about 10%. In Figure 6, the. per cent of the total Cao liberated from CaCO3 and the per c~nt of this quantity to have reco~bined is shown. As in figure 5 the amount of Cao formed is seen te vary as a straight line function of the distance the charge has traveled down the kiln. 'fhe "per cent 0aO • Liberated. Recombinedtt curve e1:11phasizes the fact that the initial,_ recombination. of the free Lime 1s rapid. !he reeom- binat.ion then . follows mucll the same course as shown in Figure 5 f' or the remaining leng.t h of the kiln. The "% of Charge Remaining" curve is theLoss on Ignition curve of Figure:, shown on a flattened scale. It is seen that the clinker comprises 68~ of the original charge4' Summary In summarizing it 1s seen that during the first fifteen feet of the kiln, most of the free and combined water has been driven off. Evidence is given that indicates an oxi- dation takes place in this portion of the kiln. The decom- position of the carbonate then progresses practically linearly with respect to the .distance travelled down the kiln until the reaction is completed about 20 feet from the discharge end. The recombination of the lime formed is slow at first but increases rapidly to within 18 feet of the kiln discharge where 90% of the reaction has taken place, 50% of it occurring in the last third of this distance. The reaction then proceeds more slowly and is complete about 7 ft. from the discharge point. The charge is hygroscopic through- out the length of the kiln and the samples must be hermetically sealed if they are to be preserved. Due to the presence of volatile materials (presumably largely K10) in addition to OOz and H~O, loss on ignition should be made under carefully controlled conditions as to temperature and time ot heating 1f the results are to be comparable. These data in themselves are too meager to warrant any change in kiln design or firing conditions 1n the production of Portland Cement. It was the plan of this research to study the effect of various changes in the raw mix and firing conditions, but owing to the impossibility of securing samples this procedure had to be abandoned. The data given can only be taken as representative of the conditions prevailing at the time ~f sampling, namely of normal conditions ot production as to firing, rate of burning, and composition of the raw mix. -26· Acknowledgment: Acknowledgment is due the Riverside Cement Company for their oo8peration in securing these samples and tor making certain determinations which this laboratory was not equipped to carry out and to Dr. W. N. Lacey for his assistance and suggestions during the course of these investigations. Literature Cited (1) Portland Cement,R. K. Meade, Cbamical Publishing Co. 1926 (2) Collet-Desootils, (3) Le Chatelier, Jour. des Mines, 34, 308, 1813. Ann. des Mines, p. 345, 1887. (4) Richardson, Papers Assoc. Port. Cem. Mfgs.June 15, 1904 (5) Richardson, Papers Assoc. Port.Cem. Mfgs. Dec.14, 1904 (6) Rankin and Wright, Amer. Jour. Sci., 39, 1•79, 1915 (7) w. N. Lacey and Hubert Woods,Ind. and Eng. Chem., 21: 1124-29. -Z1· PART B A METHOD FOR THE PRODUCTION OF ANHYDROUS SODIUM SULFATE A Method for the Production of Anhydrous Sodium Sulfate A chemical that is becoming of increasing importance for commercial use is Anhydrous Sodium Sulfate. Laury (l) gives a survey of the amounts used by various industries in 1925 as follows: Glauber's Salt (Na2 S04 .10Ha0) 57,000 tons Sodium Sulfite (from NaHS03 ) 18,660 Wood pulp requirement in u.s. 22,000 to 56,000 tons Glass manufacture 60,000 tons " In addition Canada requires about 50,000 tons annually and is constantly increasing the demand. At the present time this demand is almost entirely supplied by salt cake obtained as the by-product 1n the manufacture of hydrochloric acid by sulfuric acid and sodium chloride. In 1925, 139,000 tons of salt cake were sold at an average price of $17.00 per ton. This does not, however, represent the potential market for Na, so~ . The product obtained as salt cake ie very impure and not suitable ae a source of Nai O for many of processes in glass and paper manufacture. These industries need an anhydrous sodium sulfate of approximately 96% purity with certain foreign oonstituente, such as chloride, below definite limits. The sale of the salt of this grade in 1925 amounted to - 29- 2,100 tons at a.n average price of #61.00 per ton. Ob.viously at this pr1oe it cannot compete with Sodium Carbonate as the source of Na1 0. In view of the fact that there are numerous natural salt deposits in the United States, particularly in the western section, that run high in Na 2 S0+ , the present study was undertaken in order to determine a feasible way of manufacturing anhydrous sodium sulfate of suitable purity from this source of raw material. These deposits sometimes run as high as 98% Na 2 SO+ (l) with the chief impurities consisting of sodium carbonate, sodium chloride, and dirt with some organic matter. Usually, however, the Sodium S~Ifate is present 1n these deposits as the deoahydrate Na.eSOi- .lOHi O and this form, known a.s Gla.uber' s Salt, was chosen as the -!'aw material for this work. Figure 1 gives the equilibrium relations between Sodium Sulfate and Water over the temperature range from the freezing point to the boiling point of the saturated solution. It is seen that a transition occurs near 33° 0 where the hydrated form changes into the anhydrous, and that the solubility curve reverses. It is this reverse solubility that makes the preparation of the anhydrous phase by evaporation of a saturated solution~ery difficult. At the point of oontaet of the solution and the heating surfaces in an evaporator, a higher temperature -31· 1s maintained than in the main body of the liquor. If this liquor is saturated with Na2 S04 above the transition temperature, Sodium sulfate is precipitated as a hard scale on the heating surfaces and this scale greatly diminishes the rate of heat transfer to the main body of the solution. Obviously, unless scrapers are provided for the surface, this scale will continually increase in thickness as evaporation proceeds. Badger, (2) 1n the development of apparatus for the evaporation of solutions of salts with reverse solubility curves, found that the rate of heat transfer dropped in a few hours to a value that was much too low for economical production. He suc- ceeded in designing equipment for the evaporation of such solution by using a tubular preheater and then flashing this heated liquor in a double effect evaporator heated by a steam basket. It was necessary to provide two pre- heaters for each effect because scale formation was so rapid that the tubes had to be washed out every hour. The double effect evaporator was washed out every twelve hours. As it was desired to study various possibilities for the production of Sodium sulfate, the set-up used in these experiments included a single effect Vacuum evaporator provided with closely scraped, machined, heating surfaces, a centrifugal pump to circulate the liquor and sludge between the evaporator and settling vat, a melting vat provided with steam coils, and a centrifugal separator for the separation of the sludge and liquor. It is seen from the solubility curve that the anhydrous salt may be obtained by melting the decahydrate and separating the liquid and solid phases above 33 c. 0 1n Run l. This method was used Three hundred and fifty pounds of technical Glauber 1 s salt were added in the melting vat to a system containing a saturated solution of sodium sulfate contaminated with sodium carbonate and chloride. This resulting sludge and liquor were drawn into the vacuum evaporator and circulated to the settling vat by means of the pump. The liquor was then returned to the melting vat while the solid was recovered by means of the centrifugal separator. The run lasted about six hours; samples of the product were taken from the centrifugd.at intervals during this time. Analysis of these are given in Table I. Table I Sample No. Sodium Carbonate Sodium Chloride l. 2, 0.024% 0.016% 3. 0.16% 0.14% 0.28% 4, 0.22% 0.032% 0.055% It is seen that the samples are of about the same composition throughout the run. The product was good from a physical standpoint; 85% of the crystals were retained on an 80 mesh screen. dust. There was no tendency for them to There was a slight brown color which probably was due to iron picked up in the system. Very little water was evaporated in this run; the purpose of the evaporator was to keep the liquor well abOle the transition temperature. A temperature of 90° c was maintained in the settling vat. The temperature of the liquor dropped about le° throl..\?;h the separator and was returned to the evaporator at about 7o•a. A hard scale deposited on all heating surfaces. The scrapers were binding badly in the evaporator by the time the run was concluded. Fifteen pounds of impure sodium carbonate were added to the solution obtained from Run I which had increased considerable in volume due to the water from the Glauber's salt. The solution was quite dark in color and foamed badly 1n the evaporator. Sodium Sulfate was produced by evaporation of the water from this solution. The evaporation was slow; the deposited scale very effectively reduced heat transfer and the tendency of the solution to foam made it necessary to keep the temperature of the solution low so that the froth could be controlled by breaking vacuum. The samples taken and their analyses are given in Table II. (Sample six is the unwashed product as it came from the centrifugal. Number seven was thor- oughly washed with hot water.) Table II Sample No. Identification Sodium Carbonate Sodium Chloride 1.2% 0.01% 6 Solution in settling vat at beginning of run Unwashed product 0.5% 0.04% 7 Hot water washed product 0.5% 0.02% 8 Solution from settling vat at end of run 1.7% 0.08% 5 The product obtained by evaporation was composed of larger crystals than that made by melting Glauber's salt. Wash- ing with hot water improved the handling properties of the crystals in that the cake was not so tightly packed in the centrifugal or so hard to get loose from the aides as it was when not washed. As in Run I the scrapers rubbed quite hard against the deposited scale, the entire evaporator vibrating from the violence of the action at times. As a final run a Sodium Sulfate solution was nearly saturated with Sodium Chloride and this solution circulated through the system. by evaporation. 1n Run II. The NazS04 was precipitated entirely The solution did not foam so badly as The results are given in Table III. The samples are from successive centrifugal loads. Table III Sample No. Centrifugal Load Sodium Carbonate Sodium Chloride 0.2% second o.6% o.6% 11 third 0.6% 0.7% 12 fourth 0.8% 13 fifth o.6% o.6% 9 first 10 0.4% 1.6% All these samples were well washed with hot water. The crystals diff-e red little from those obtained in Run II. As usual, evaporation was slow, about six hours being required to collect the five oentrifUgal loads of product. A survey of these runs gives the following results: Sodium Carbonate 1s the hardest to get out of the finished product. Even in a nearly saturated solution of Sodium Chloride, Sodium Sulfate will precipitate with little sodium chloride impurity; but a solution containing only 2% Sodium Carbonate will give a product with about 0.5% of the Carbonate present. This is thought to be due to the similarity in the solubility curves of the two salts, Sodium Carbonate and Sodium Sulfate, and the fact that a double salt orystalizes from a solution of the two when concentrated above 35° 0. The data agree with the -36· findings of other workers in regard to the salting out of the sulfate on the heating surfaces. The success of the evaporation in these runs is attributed to the scrapers in the evaporator. However a oomrnerc1al size evaporator with machined heating surfaces scraped to within a sixteenth of an inch would be prohibitive in coat. The very satisfactory results obtained by melting the Glauber's Salt 1n a hot saturated solution of Sodium Sulfate indicate that the most satisfactory way to manufacture the. anhydrous material is to heat a saturated solution by direct means and to use this liquid as the medium in which to melt the solid Glauber's Salt. The use of live steam as a means of direct heating has several disadvantages, among which are: The steam must be made entirely from make-up water~ Extra water is added to the solution. The use of hot, permanent gases as the source of heat eliminates the above mentioned difficulties and also offers a means of using effectively the waste heat in flue gas. Likewise this method has its disadvantages, among which the inefficiency and the large volume of gas necessary to handle are the most important. However the latter method was thought to offer the best possibilities and the following plant was designed on this basis. Figure II is a schematic flow sheet of the plant showing the course of the sludge and liquor in the process. It is based on 100 pounds of Sodium Sulfate fed into the melting vat. The process consists essentially of cir- culating enough liquor at a sufficiently high temperature to melt the Glauber's Salt and then evaporating the water carried in as water of crystalization in a spray tower evaporator. If Sodium Chloride is present in the raw material, then the system will eventually become saturated with it and a small amount of the liquor will need to be discarded continually so that no Sodium Chloride will be precipitated in the tower. This flow sheet assumes a saturated solution of Sodium Chloride circulated to the melting vat, but since the amourt.· discarded will depend on the analysis of the raw materials no attempt has been made to estimate this quantity. Likewise the free moisture coming in with the Glauber's Salt has been considered as balancing the water carried out by the Na 2 SO+ to the drier. Data and Assumptions: Heat of fusion of Na 2 SO~ 10H 1 0 Specific Heat of Sol. 0.8 92.3 Ave. Specific Gravity of Sol. 1.3 B.T.U./# Assume flue gases into tower at 800" F and out at 300° F 0 Circulate liquor into tower at 100 F -39- Liquor withdrawn from tower at 215° F Temperature of Raw Material 50° F Circulating liquor for melting, worktng between 215 ° and 130° F Liquor sat. at 215°.F with NaCl + Na~S04 contains 24.7% NaCl and 5.1% Na 1 S0 4 Liquor . at 130'l' with 18% NaCl contains 10.6% Na SO Basis: Heat required to melt the Glauber's Salt and bring it to 130° F 224 x 92.3 • 20,700 B.T.U. (melting) o.e x 224 (130-50) = 14,300 B.T.u. for heating Liquor necessary to circulate to supply this heat: 35,000 (215-130) X .8 -- 485 # required BALANCES ON SYSTEM Melting Vat 1n Out ·485# liquor (120# NaCl (circulating)( 25# Na~S04 (340# H2 0 224# Glauber'a Salt 709# total 58# Na2 SO~ (Sludge) ~:81 651# liquor ~~6f 709# total ~ 6~ Na2S04 Settling Vat ·60# liquor to centrifugal 58# Na l so~ " " 591# to tower(llO# NaCl ( 63# Na 1 S04 (418# Hll. 0 ·709# Centri:f'UJ3al In 60# liquor 58# Na2,S0 4 50# liquor :rrom settling vat from tower 5L_wash water 217# 217# Tower I!! 591# liquor (110# NaCl from settling( 63# Na ~S04 vat ( 418# H1 O 117# liquor f'z,om centrifugal ·708# 485# liquor 11,- NaClSO Naa recir~u- 3 2 H O ~ lated a. 44# Na2.SO.,_ to NaCl cen50# liquor ~ 12#.. 3# Na~S0. trit3511 H2 o ):fugal 129# HaO evaporated 708# • ! TOWER DESIGN Basis: 5 ton per hr. Liquor in at l00°F out at 215°F 708 x 20 x 5 = 70,800# liquor to tower Liquor required. hr. 70,800 x 8 x 115 • 6,810,000 B.T .U. per hr. to heat liquor - 12,450,000 B.T.U. per hr. 12,900 X 965 Total B.T.U. supplied in tower per hr. 19,260,000 Flu gases 1n at 800° F out at 3oo" F Ciroulating liquor out at 215 585 At = 585 - 200 2.3 log in at : ~gg to evaporate H2.0 100 200 385 2.3 (.463) h :. (H) (from W.L.& Mc.) : IS 357 0 1.5 B.T.U. /min/F /f"C = 90 B.T.U. /hr./F /f"C 3 90 X 357 : 32,100 B.T.U. /hr./ft 19,260.000 32,100 = 600 cubic ft. required in tower Assume factor of 3 giving 1800 cu. ft. Assume gas velocity through tower as 5 ft/sec. From W.L. & Mc. MCp 800° • 5200 B.T .U. MCp 300° = 1800 B. T.u. 3400 B.T.U. 19,260,000 3400 - - 5660 # Mol s /hr. :: l.,'1.. X 359 X 1010 92 x 5 : 1.57 # Mol/S eo. 232 sq. ft. area of tower cross section Th.us a 17' tower with an effective height of 15 ft. provides ample safety factor. Assuming waste heat is utilized from boilers, which an absorb 25% of the heat derived from the fuel oil and 60% ef ficiency is secured in the tower: 19,260,000 l 7,000 X .75 X . • 6 X 10,000 = .25 # o11/#Naf so4 Produced A plant producing 5 tons of Na 2 S04 per hr. would then require 2500 # oil/hr or 2~gg : 7.14 BBl 7ir at #1.50 per BBl the fuel cost per ton of Na 2 SO+ would be 7.14 X 1.5 5 • $2 • 12 The power plant would have available (assuming 25% efficiency) 2500 x .25 x.25 x 17,000 3411 = 7800 K.W. Hours. General Discussion. Figure III is a flow sheet of a plant capable of · producing five tons of Anhydrous Sodium Sulfate per hour. Centrifugal pumps are recommended for handling the sludge and liquor. Badger (2) found that the Na~SO+ crystals were hard on the impellers and packing glands of the pump. He designed a centrifugal pump with extra large impellers and a liquid sealed packing gland and experiences no further difficulties from this course. This process is capable of several variations. It large quantities of Sodium Chloride are available, the tower may be used as a heat interchanger and the Na 2 S0 4 salted out of the melted solution by the addition of Sodi~~ Chloride to saturation. The main part of the liquid can then be run to solar ponds, while only that necessary to furnish heat for the melting of the Glauber's Salt need be circulated. If the Sodium Sulfate is to be obtained by evaporation there is no advantage in having a saturated solution of NaCl, since its presence only changes the quantities that precipitate in the melting vat and the tower and not the total amount precipitated. Naturally the chance of contamination of the product will be less, the lower the concentration of the foreign salts. Na 2 S04 has a specific gravity of 2.67 and settles very rapidly. This would facilitate the separation of the sludge and liquor in the melting vat in case the latter had to be filtered to remove suspended matter. A SWeetland filter press is suggested as suitable for this filtration. It is noted that a circulating system is shown in figure 3 for the tower. Due to lack of time no data was obtained on the efficiency of this means of evaporation. However recent works (3) by R.H. Newton and T. C. Lloyd on evaporation of water by hot, dry air indicate that by circulating sufficient quantities of liquor the exit air approaches 100% relative humidity. While this degree of efficiency 0 is not expected with flue gas working between 800 F and 300° F the figure of 60% used in these calculations is thought to be conservative. Large scale experiments in a pilot plant using this procedure of manufacture would give the desired information and such a study must necessarily be the first step if this method were to be developed commercially. Acknowledgments: Grateful acknowledgment is accorded Dr. w. N. Lacey for his suggestions and interest in this study and to Bruce H. Sage for his very material aid in the development of the method. Literature cited: (l) N. A. Laury, "Hydrochloric Acid and Sodium Sulfaten A.O. Society Monograph Series No. 36 (2) Badger, Trans. of Amer. Institute of Chem. Eng. 16, II 131-134 (1924) (3) Newton and Lloyd Ind. & Eng. Chem 23,530 ( 1931)