Thermodynamic Properties of Ammonia Lester Haar and John S. Gallagher onl Messurement Laboratory, National Bureau of Standards, Weshington D.C 20234 Amani thermody sa es he esperineotl data or mani for he temperate ange eeadiog rom the triple point 0750 avin and forte premure range extending fom he te ato vdynamic rope ae ablated at Sosy sped intra. A mejor ‘ua thems. As consequence, we ate ble tag le olerances tothe abated thermody pep ‘rie oe the range the url, ning proper forthe conning pares and even ce othe had ‘aor ee ont Keywords Ammonia creation: equation of ate; gu iu phae ule hemedyeamieproperi 1. Thermodyn Li Tntroduetion 2... 12. Theoretical Approseh 5. Reference Values 7 IL. Thermodynamic Data and Analysis | UL. Historical Background M2. P.e.T Liquid-Vapor Coexistence ... 13. Calorimetrie, Insluding Ideal Cae MWA. Pra Single Phase; Liquid and Gas 15. Reference List of Historieal Da ML, Thermodynamic Surface HLL. Introduetion L2. Derivation of Surface LB. Comparison with Date... ts 8. Symmetry and the Critical Region TLS. Evaluation ofthe Surface, MLS. Summary Acknowledgements References Appendices Physical Date and Reference Ststes Appondin A. Tablo of Thermodynamic Prope the Coexisting Phases of Liquid and Vapor Appendix B. Table of Thermodynamic Properties of the Liquid andthe Gas... ee Appendix C. Table of Second-Viral Coeficiente. Appendix D. Table of Joule‘Thomeon Coefficients Appendix E. Engineering Drawings Uist of Tebles 1. Coethcients of Vapor Pressure Equation, eg (0). « 2. Gaeieen of forthe Dome of Coxining Phases... © 197 bythe US, Secretary of Commerce om behalf ofthe Used Sate. “Tis coptgh x ssigned to th American Inte of Phys and the ‘mein Oni Sty oar. 268578 0435 $08.00 Contents 686 3. Cuefficiemt uf the Pree Bucngy Poetion fo the 686 “Ideal Gas . eee 646 (636 4, Cocffisients for Mini"Thermodynamic Surface... 652 688 5, Values of G,/R for Gas Phase to 125 Bar and S75K 65¢ S38 6 Coss for Thermodynamic Suse, Li and 688 Casto 5,000 Bar 661 640 7, Gibbs Function Residuele - 663 61S 8 Bnimeted Uncertainticein Tabulated C, Valuce.. 67H 650 9 List of Figures 660 660 rye te 00 Gy, Thermodynamie Consistency of 663 60 on eo 2, PT Schematic of Selected RT Date... 650 om 3, G,, Thermodynamie Consistency of ou Calorimetrie Date with P.aT Data 653 6s 4, re" Data, 655, 619 5a,Sb,Se. Comparison with Data of Caleulated 9 Values of Density. 6,665,666 ‘ Comparison mith Data of Calvulatcd Valves a0 ‘of Calorimetrie Properties for Coexsting Phases. sevens oor 907, Comparison with Data of Calorimetcie 79 ‘Values forthe Dilute Gas... oor 0 8 ‘Comparison with Reference Valuee oy of Cslorimetric Properties Caleulated for isc mae pee ee Second Viral Goeticien, BT 669 1. PeqPlot of Coexistence Dome and the oe Critical Region. . 670 62 ou, ‘Compressibilty vs Temperature on the Critical Isochore; Comparison with 045 Critical Sealing. acl 12 Compreasibilty vs Density on the Critical, Isotherm; Comparison with Critial Sealing en 134135, C, lobars ve Temperature. oz 635 hp. ete, Vol.7.No. 3, 1978636 1. Thermodynamic Features 1. Introduction ‘The standard reference data for ammonia have been thos provided by NBS Circular 142 "Tables of Thermodynamic Properties of Ammonia” which was published in 1923. I is tuibute tthe quality of te mork that went fay Clavules 142 that it has remained in current use for over 50 years. How: ‘ever, the principal technological application for Circular 142 was the strongly stated need for data relevant to the use of Smmonia as a refrigerant. Today this application is dwarfed hy the growth in utes for ammonia in agriculture and in- dustry. New applications in agriculture, plastics, .. almost tunknown in 1928, now require enormous quantities of am ‘monia (production in 1976 was 165 million tons compared to 11. thousand tons in 1923) Present and proposed uses for ammonia require date in temperature and density regions much beyond the range covered by Circular 142. It is our ob- jective with this new compendiam to satisfy modern techno: logical needs with st of thermodynamic tables that greatly cntend the rene of Gisslas 148. ‘This compendium contains a correlation of thermodynamic ta for gaseous and liquid ammonia from the triple point value of temperature to near twice the value at the criticel point and for values of pressure 10 5000 bar.’ The reeulte are lables of properties thst are thermodynamically consistent and in most part agree with the existing data to within the assigned eror tolerance of thene data, To aocomplieh thie, we developed several innovative procedures that fciitete analy ss of data for thermodynamic consistency. Certain features ofthis error analysis have been previously reported{1, 2,3}? Especially important was stateofsheart analysis of the thermodynamic properties forthe ideal-gas state (4), from hich the zero pressure reerence properties for this corel: tion were taken and which serves an sandard sot of data for smuch of the error analy ‘The compendium contains three sections. Ths frat section includes a discussion of the philosophical arguments underly is out appiowcl wo abe problem of correlating thermody namic data and the methods we employ to avoid thermody- namic inconsistencies. The second section contains an in depth examination of various reported thermodynamic data for ammonia. The Helmholtz free energy surface for am ‘monia js derived inthe third section. Included therein are ‘comparisons with the various experimental data and with previous corrclations for aunaui (Sy 6). Alsi cluded in the thd section are accuracy estimates for the overall correlation, Tables and charts of thermodynamic properties caleulated from the derived surface ate given in Appendices A-E, 2, Theoretical Approach ‘The general approach used in this correlation involves the ‘onstruction of « Helmholt free energy function (or surface) for the entte temperature and density range ofthe correla "Paar nace nde rata ern 1LPhy. Chem. Re. Oo L HAAR AND J. S. GALLAGHER ‘ion. To construct this surface the fice energy for the ideal fas is combined with the free energy contribution for the lemperature-density surface, the later being obtained by a least squares fit (with no arbitrary constrains) of an ans representation fo selected P, @, T equation of state thermodynamic properties are then ealeuated from this su face by differentiation. ‘The vuly date wihes Usen 25 T deta weed i the devivar tion of the thermodynamic surface are the thermodynamic properties calelated vi statistical mechanics forthe ideal- sas sate. These properties were calculated by us [4] to state fof the art accuracy, so that over most of the temperature range ofthe thermodynamic measurements for ammonia, the largest contribution to the uncertainty for the ideal 3 values arises from the uncertainty inthe valu stant, an uncertainty much smaller than that associated with, the experimental thermodynamic date. Thus, inthe context ‘ofthe P,¢, T and other data, the idesl gus values can be con- sidered to be exact and afford a fundamental anchor line to Which the other data can be compared on an absolute basis. All the Medal ‘This section includes a discussion of certain thermodynam: ic inconsistencies generally associated with derived thermo- dynamic surfaces and the methods we use either 10 avoid them of (o limit the extent of the degradation. These incoa- sistencies result more from how data are used, than from crrors in the data themaclver. For example, a very seriou in consisteny results from use of an analytic surface to repre: sent data in the vicinity of the eritical point. Others result from: use of arbitrary constraints to impose exact agreement ofthe surace with certain preselected values, overiting (or tunderfiting) of the data, and inconsistencies between data sets, This ection alo containe discussions of how we use the ralrimetrie dats to ery the arourary of the surfzce and 3 ‘comparison of our method of single property Biting with the currently popular method of mult-property fitting (sing r= dundamt thermodynamic data where available, eg... @ T 21e, cic Region Iti generally believed that P,o isotherms in the vicinity of the eitical point re flatter than those which ean be obtained ite thermodynamic surface. (One consequence is that an unconstrained analyte surtace tends to pro duce 2 value for the critical temperature somewhat higher than that obtained experimentally). Thus an analytic surface, though it may contain many terms and many adjusable con. stants and though it can represent data accurately over large part of the surface, hat only « limited capability for representing data in the vieinty ofthe eritcal point. If letept this as genera Feaue UP aire, I fll ‘consistencies must result when we require an analytic surface to accurately represent dats close to the critical point. Fur. thermore, because the critical point is e point of thermo: dynamic symmetry, thermodynamic inconsistencies in the critical region tend to cause very serious degradation over large regions of the derived surface. Our remedy for thisTHERMODYNAMIC PROPERTIES OF AMMONIA situation i very simple: We exclude from the least squares fit all constraints atthe erticel point (ie. conditions tobe exact ly satisfied there by the surface) and we exclude also all Pg, T data points inthe immediate vicinity of the eritieal point. ‘With this simple procedure a very good representation of the data can be achieved, even extending quite close to the ert cal point. We shall show that the effective extent of the critical region, that ie the region in which the analytic surface cannot describe the effet of the non-analticcritieal anoma- ly, is very small, covering range for ammonia which is within 1% ofthe eritical temperature 2th, Acitrory Constraints and Phase Equilibria Constraints that impose exact accord with preconceived values eause inconsistencies in the derived thermodynamic surface. Since constraints are usually used to force exact con- sistency between data and loci of symmetry forthe surlace, ‘eg the line of coexisting phases, the critical po small erzors in the dats chosen to be fit ex ly can cause in a given tolerance, the line of coexisting phases predicted by the surface willie within that tolerance without Fecourse to exact constraints, In this correlation the Teature of note is that explicit use ie made of the data for the vapor pressure of the saturated liquid in the least squares fit, but fly with »sttitienl wright consistent with the accuracy of those data, We shall show that based on this procedure the thermodynamie surface derived for ammonia yields accurate values forthe properties ofthe evexisting phases, including latent heats of vaporization and values for the heat capacity forthe saturated liquid and eaturated vapor. ‘The thermodynamic validity ofthe derived free energy su face depends strongly on tolerances imposed in the Biting process. For example, a close fit to accurate Fg, T data does not guarantee that the derivatives obtained from that fit wll be accurate, Thus if such data are overftted, inthe sense that palynnmials of pawer higher than necessary nerd, net rect derivatives must result. An optimum fit would include only the general trends of the dats and would ignore the detail of the scatter. Such a result is dificult to achieve, since distinguishing between random error and systematic trends isnot a simple matter. We have been guided in our analysis by the belief that errors in the derivaties from an fvertt af the data are sally larger than are those frm an _underfit. Therefore, as fo the coexisting phate region, were {quire ofthe derived P, g, T surface that it be consistent with the data only to within our estimate of the sccuracy (not the seatter)of the Perhaps the largest errore from overfiting resull from the very nature of the mathematical representation ofthe ther modynamie surface. These errors area direct consequence of the fact thet there ie an inherent tendency of the thermody: namie surface to overit particular regions. For example, the representation for the superheated vapor in most part re quires polynomials of low degree in density, On the other 637 hhand, the equation for the thermodynamic surface must in- clude terms of comparatively high degree in density in order to represent the high density regions adequately). The extent to which these higher degree erm contribute w the dynamic properties in the superheated vapor region is a ‘measure of the emount of error from ths type of overfit. To reduce these errors we made detailed analyses of the data in thie region. This method included the derivation of sub-sur- faces that were fited to those data alone. Special care was taken so that these sub-eurfaces included only terms to the lowest degree in density (and temperature) compatible with {good representation ofthe data. Thereby we could be confi dent that the subvsurfaces exhibited minimum overfit and that thermodynamic properties calculated therefrom would he the most accurate values we could obtain from the data ‘The selection ofthe number and type of terms in the overall thermodynamic surface was in large part based on comps sons of the thermodynamic values calculated trom such sub- surfaces with the corresponding values from the overall sur- face. Thus the criteria for the selection was thatthe overall yield values in elace agreement with those from the subsurface. Large degradation in the accuraey of the derivatives can also result from quite small inconsistencies between different Gala ets. Ihe extent of the degradation in queliy of the resulting free energy surface depends on how the free energy surface compromises inconsistencies between data sets. (Us: ally such inconsistencies tend to eause lage vs derivatives of the surface in the region of the data overlap) [An important part of our approach consists of analyses made for each date set fret by itself and then in combination with those sets in its immediate neighborhood on the thermody: namie surface. Based on these analyses, we have been able to avoid (ora least to reduce) the effect of such inconsistencies, Property Versus Mull. Property iting In our approach the ealorimetrie data are not used in the leastsquares determingtion of the thermodynamic surface. The calorimetric data are ueed only in an independent test 0 vetfy the thermodynamic consistency and accuracy of the surface. Calorimetric values are obtained from the surface by differentiation. Because errors in the derived thermodynamic surface tend to be greatly amplified in thei effect on the se- curacy of the derivatives, our approach is # very sensitive test, That we are able 10 assign small error bounds 10 the calorimetri values ealeulated from the thermodynamic sur face attests not only to the accuracy of the data, but to the fof our methods. But more important. those error ‘hounds apply to regions of the surface considerably removed from where experimental calorimetric data exit! In mult-property fitting, calorimetric (and other) redun- nt date are included with the P,g.T data in the (east “squares) derivation of the thermodynamic surface. With such method it is comparatively easy to achieve reasonable sgreement between calorimetric values calculated from the ‘surface and experimental calorimetric data. This is because the thermodynamic relation between, say, the heat ci and the Helmholtz free energy surface includes functions of density (or pressure). Thus, inthe mult-property Phy. Chem Re Date, Vo. 7, No, 3, 1978638 iting approach there is some Mexbility available to force the ‘surface 10 accommodate to inconsistencies that may exist, ‘among the several kinds of date. As a consequence, mult property iting offers no assurance thatthe resulting surface vil yield accurate derivatives, excep in thos limited regions for which calorimetric data exist, No information on accuracy i aveilahle where anh date da not eviet OF corer. i the calorimetric data and the P,qT data are thermodynamicaly consistent their combined use would not seriously compro- rise the overall quality ofthe surface and (the use of redun ‘dant data) should simplify the derwation. But to establish such @ consistency, as we shall show, i itself « major tesk Finally, multipropery fitting techniques do ot yield calor retrievers that are in heer agreement with experimental data than are those calevlated ky us for ammonia for which deviations from the idesl gas are obtsined from the oT data alone .e. the single property approach) 2.2. Analysis of the Date ‘The NBS program for ammonia of half «centruy ago pro: duced a large body of thermodynamiealy consistent data for the eoexiting phases and for the superheated vapor, in cluding hoth calorimetvc and Pg, Tata, Much of the erat ‘analysis diseusted in section 1, 2. is based on previous ‘studies we have made with these dats. Thus, we have already established in reference 2 the accuracy of the heat capacity ‘measurements of Usborne etal for the dilute gas, In tha ‘work, the ideal gas tate values forthe heat capacity which we ante in eference yee compared ih he cored the range ofthe data, -15 ©C&< t-< 180" Cit wae found ‘that, when corretions for the differences im temperature scale were included, the agreement approached the accuracy ff the calculated ideal gat values, that is agreement was within 1 or 2 parts in 10,000! (Such is the accuracy estab Tished for these data tha i was suggested in reference 2 that probes ofthe thermodynamic temperature eal (8). ‘We also have completed a detailed theoretical study 1} 10 correcta certain type of systematic error that was present in all the P, , T data then available for ammonia, In the ex- periments which produced such dats, meteury wa used as a confining fluid and wes in physical contact with the am: mona. However, the effeste ofthe gae interactions hetween, the ammonia and the mercury vapor were not considered in the analysis of those measurements. Corrections for such effects were therefore absent from the published dat. It had been conjectured [9-13] that systematic errors from this ‘omission could be large, particularly atthe higher pressures, In our study we ealelated the magnitude of the errors and, Actually very small, We showed that such errors can, in fact, be ignored in this work. In reference 1, this conclusion was Shown to hold even forthe bulk of published Pg, T data for cother substances wherein the experiment mercury was used as 6 confining Aid. All such data had previously been con sidered suspect becauce of the omission of corrections due to Ae aocte ee sein Bud 1.Phys. Chem. Ra. to, Vol. s.t978 L. HAAR AND J. $. GALLAGHER In reference 3, other important results that can be ob- tained from careful analysis ofthe data were presented. Thus, the overall quality for the free energy sub-surface was ob- tained by combining two P, qT data sets (16,19) with the results forthe deal gas, Vales for the constant pressure heat capscity calculated from this sub-surface for ammonia vapor “ners chown in he arrrate to within about several tenths of percent over most ofthe range, even close to the saturation line! Since direct calorimetric measurements of high quality (7 exist over only a very small part of this temperature- ‘density region, the results described in reference 3 extended the range of the high quality experimental heat capacity measurements tworfold in temperature and over fivefold in i y- These results attest to the extraordinary quality of the dats inthis region (thats, to the extraordinary quality of measurements carried out sunder the old NBS program). These results also are a strong Alliemation ofthe power of our approseh, particulary that of single property iting. ‘These theoretical data analyses previously published by us ars further davalaped ond stlved thennghaw tion, The elace tolerances we obtain for he aceuracy of the Uubulated thermodynamic properties in large part area direct, consequence of these analyses. ‘wth very litle loge in ae 3, Reference Values The values af temparstrs far al thormadynamia data ame ployed in this work have been converted tothe IPTS-68 sale [1]. The caleuleted properties for the ideal gas, however, ‘refer tothe thermodynamic sale, Important differences have been shown fo exist Between the two scales tn the neighbor- thood ofthe steam point (17, 18], end in reference 2 we have shove that these differences can result i significant errors in derived enthalpy and heat eapacty values. In reference 3 we presented arguments tothe effect thatthe errors would be ‘minimal over most of the surface if the differences between the IPTS-68 and thermodynamic scales are ignored and if ll resus are relerred to the ftest estimate of the thermody: namie tempersture scale. Thus, the results of the correlation refer to values of temperature on the 1968 thermodynamic tale [16 The Helmholts free-energy surface we derive refers to pure ammonia. All chemical reactions such as dissociation are ig nored. Dissociation in the absence ofa catalyst could be im- portant ony for the dilute gas at temperatures above 300% [4]. The reference state forall thermodynamic ealeulations at 0 K. The physical constants used in this smmended by Eahen and ‘Taylor {19}. The mass of « mole of ammonia is taken to he 17.08026 grams (20), Il, Thermodynamic Dota and Anolysis 1. Historical Background ‘The thermodynamic date for ammonia include 2 number of measurements that are now of only historical interest. The‘THERMODYNAMIC PROPERTIES OF AMMONIA teenth century: Bunsen [21], Faraday 22, 28), Regnaul [24-26], Ledoux [27], Dewar [28], Moller [29], and Dieterici 20); and also many ofthe ery 20h century: Nernst (31, 82), Lange (88), Lewis and Kendal (54), Haber [59], Berthoud [86], end Gibbs [37]. The discussion inthis correlation is Tinted to data reporied subsequent tothe start ofthe org ral NRS program (1913), and only those data which hat a bearing on our correlation wil be discussed in detail those data that were used in the derivation or evaluation of the fre energy surface. The experimentalist is usualy inthe best postion to evalu ste or estimate the quality (or accuracy) of his own work. However, such estimates of quality by the individual scien ties often are not ueful different dats sets, This is due tothe fact that estimates of quality by the experimentalist are often ambiguous (and Sometimes even omitted) even in work of otherwise high ‘quality. In this section we attempt to establish the relative order of accuracy ofthe various sts of date. With the very great advances in instrumentation achieved years ome might onpost that «similar improvoment in the state of the art for thermodynamic measurements Would also have occurred. To the contrary, our experience with the data for ammonia leads us to believe that eccura Fesults are associsted more with the ingenuity and artistry the experimentalist then with the use of sophisticated i ttrumentation, We have found that reeent data are not neces tarily better than olddata. As ilutraton, inthe past 50 there has been almost « revolutionary improvement in auto- matic control equipment and in measurement instrumenta- tion. Yet many ofthe messurements of $0 years ago on which [NBS Circular 142 [38] is based rank in quality wih the best ‘of current research. The principal effects of technological a ‘vances over the past 50 years appear to have been to reduce random scatter and to make experimental measurement eat ‘er to achieve. But, ax then, the challenge to the experimen: talist today is in the reduction of sources of systematic error, ‘and the ultimate quality of his work depends on the degree to hich this effort ix successful It follows then that sophis cated instrumentation improves precision but hes only « sec ‘ondary effect on accuracy tion are the set of measurements made over 50 years ego for [NBS Circular 142. Circular 142 was based primarily on the measurement of seven basic properties of ammonia, the spe cifie heat of the saturated liquid [39], the latent heat of vaporization [40], the vapor pressure [41], the specific vol ume of the saturated liquid [42], the specific volume of the saturated vapor [1] the apecifc heet of the superheated vapor [Ta], and the specific volume ofthe euperhested repor 24]. In addition, supplementary results having somewhst larger uncertainties were obtained for the lent heat of pres sure variation [44], the Joule-Thomson coefficient for the vapor [Ta], and the epecitic hee ofthe saturated vapor [40] The purification process fo the ammonia was by itself ma jor part of the program and § differnt sets of purification procedures were used [45]. Several of the purification pro- cedures produced samples with no more than about I part in UW of non-condensable gases, and with other impurities 639 ‘reduced to below the limits of detection, that isto lees than (0.008% water and les than 0.001% all others, which at shat time was a major achievement. For the entire set of measurements over the range from the ammonia triple point temperature to 300 °C the practical temperature scale emploiyed was nearly identical to the LYS. 1927 seale (wo within about 0.01 K)[46, 47) Au wld ‘ional uncertainty arises from the variation found among thermometers [48] calibrated to the Callender equation as ‘modified for LT.S. 1927, The entice set of data can be re- ferred to the present temperature scale LP-T'S. 1968 [16] With an uncertainty of no more than * 0.015 K. This is no small accomplishment, since these messurements were made sta time when the International Temperature Seale was not defined for values of temperature below the freezing point af ice. Finally, there is an uncertainty of perhape I part in 10" in the interpeetaton of the enerey data in terms af present dy physical units, In several cass, to ensure consistency and accuracy, prop ‘erties were measured in more than one way. Thus the hest capacity of the saturated liquid was determined by two meth ‘ods, the specific volume of saturated vapor by three methods. Important physical constants, such as the triple point temp- by two methods. Experimental results that were deemed be- low the standards of quality set hy the experimentalist were discarded and, in some cates, entirely new experiments then designed, even though the discarded date were known to hove been within engineering tolerances. The basic philoto- phy that prevailed is clear from the statements by the then Director of the NBS: "The date resulting from this series of investigations have heen made mutually consistent, not so much by mutual ad- iustment of values or arbitrary choice as by refinement of ex- perimental work o the point where thermodynamic diserep- ancies are negligible... . the data avsilable appear to be sufficiently reliable to meet all possible requirements for ‘many years ro come” [9]. Emphasis supplied by the present authors) The error estimates originally reported for the various experiments were realistic and conservative, and throughout ther range the entire set af primary data are till believed to be the most accurate measurements available for ammonia. Where more recent measurements of high quality indicated inconsistencies or small errors, careful investigs ‘on in all cases to date nas tenaea to resolution of aitterences in favor of the older NBS data, as we shall illustrate in this section. includes ve subsections, Scetion He 2 a discussion ofthe Po, data for states of liquid-vapor coex- istence, Section I. 3 contains a discussion ofthe calorimetric dota, including the statistically ealeulated properties for the deal gas and the measurements by Osborne etl forthe heat capacity forthe dilute vapor and the heat espacity and the enthalpy associated with liquid wapor coexistence. In eection UL 4 we diseues the P, gs T da including the liquid snd gas. for the single yese tegious ection I, 5 includes a brief summary of measurements that are now only of historical in 978640 L. HAAR AND J. S. GALLAGHER 2. P,.T Data for Coexisting Phases Measurements of the vapor pressure of the liquid and of the specific volumes of the eaturted liquid and saturated vapor are discussed in this section. Also discussed are equ tions derived to fit the selected data. The vapor pressure is, Alseued Sn section 215 the volume dat is seats 2.2, and the dome of coexisting phases, n section 23, 2.1, Vavor Pressure of the Saturated Liquid 2 ropes, Mayers od Tyler 41] ‘Thece measurements extend in temperature from lout 645 °C to +10°C, over which temperature range one hun red fifty measurements were taken inthe first of two exper vents, including 11 determinations for the normal boiling point. The data were taken at near uniform intervals except near values of temperature sescisted vsith important fixed points, for which, date were taken over very shal intervals measurements based on a completely independent experi- rental procedure. Values of temperature were measured with platinum re: sistance thermometers. Pressures were measured with mer- ‘eury manometers and a piston gauge: below S atmospheres? ‘each of two mercury manometers was used; between 5 and 15 stmngpheres « mercury manometer and a piston gauge were used; above 15 atmoepheres only the piston gauge wes used. ‘The pressure measurements were consistent over the range, including the pressure regione of instrument overlap. ‘The seater ofthe data was in large part essciated with the pres: son ofthe pressure measurement, which was about I part in 5000 for pressures above several atmospheres, ‘The experimental method seed forthe main hady of dat is referred to as the static method. The ammonia is confined to one arm of a manometer, with mercury as the confining uid. At coexistence, measurements were made of the total Pressure of tne vapor in equininium with iguid at var ous temperatures. An elaborate system of baths was used to reduce thermal gradients. However, certain modifice tions were required below ~85 °C for which region measurements were made for only three values of tempera ture. (A more accurate treatment of the triple point region was then in progress at NBS [45], Throughout the experi iments values of temperature were believed to be held con- stant to within 0.01 °C. The measurements over the entire range were repeated after time intervals spanning months, and with several different ammonia «aml ples obtained by several different purification procedures ‘The second set of measurements was limited to determing: tion of the normal boiling point. These measurements were Dbwved on an experimental procedure reterred to as the ‘dynamic method, which is comparatively insensitive to the presence of impurities or to the attainment of equilibrium. In thie experimental procedure. one arm of the manometer vas open to the atmosphere, but with a NaOH trap. be- inaleding com Tame 15 be, Phys chem Ra Date, Vel.7, No.3, 1978 tween the sample and the atmosphere, so that the volatile impurities would not accumulate. Only the values of tem: perature for the vapor was measured, so that the effects of any non-volatile impurities which remained in the liquid were also relatively small. To the existence of thermal gradients, the measurements were made with the thermometer at different p ive te the guid vapor interface. The results for the dynamic method were selfconsistent, with 2 random seatter near that obtained for the stati experiment. The mean of 17 dynamic measure ‘ments agreed with results for the determinstione by the static method to within ebout 0.01 °C and the authors sug gested a value for the boiling point of ~83.88 °C (PTS 68) ‘which ie the average forthe two methods. If we assign an overall uncertainty of 20.010 °C to the measurements ebove ~50 °C, the resulting uneertainy in the values for the vapor pressure isles than 0.1% at ~50 C, but this decreases with increasing value of temperature, becoming less than 0.05% at 0°C and less than 0.03% at +70 °C. The pressure measurements carried outwith the various 0.02% for pressures above several atmospheres, 2b, Mekelvey and Tylor [8] Measurements of the triple point values of temperature and pressure were reported by McKelvey and Taylor. The rethod shey weed is basicaly the same ae veed in the atic determinations of Cragoe et a. setion 2.18, but with im- provements to ensure achievement of equilibrium, The pres: sure and temperature instrumentation were alo similar to ‘those employed by Cragoe etal, with presures readable to £02 mm mercury A catastrophe ia the laboratory which caused the death ofthe senior scientist resulted in an sbrupt termination of the experiment. However, two triple point vapor pressure measurements and four triple point tempers ture measurements had already been made. The spread in the four temperature measurements was less than 0.01 °C, win ane average value ot ~ 11.01 °U UPI 1968) reported a= the value of the triple point temperature. The two vapor pressure measuremenis in this paper differed by 0.2 mm mer- ceury. The mean ofthe two, 45.2 mm, was reported as the tr ple point pressure. This is 0.3 mm higher than the value ‘reported in Cragoe etal. (Ine private communication to us Cragoe indicates thatthe vale listed in hi paper is in error) Ir seems reasonable to assume that the triple point was in- deed achieved inthis experiment. If the principal uncertainty in the presure measurements is associated with the precision ‘of the manometer. then a 0.2 mm sestir in the data 9 tents nearly @ 0.5% error in the measured values forthe ti ple point pressure. The value reported for the triple point temperature would appear to be uncertain by no more than avons 0.010 "C10 0.013 "e. ‘Also included in this paper were results of two measure: ments ofthe density for solid armmonia, 0.817 g/em? at ~79 °C and RAG gem? at —1R5 °C. ‘The accuracy ls 202%, sod we Ta Hew npTHERMODYNAMIC PROPERTIES OF AMMONIA 2, Keyes (0.81) Keyes [50] reported vapor pressure measurements for values of temperature above 0 °C, which corrected earlier results of Keyes and Brownlee (SI). The date reported in ref- erence 51 have very large scatter, nearly 0.3% atthe higher saluos of tomporsture, These me he range. of the NBS values to near the critical temperature. The data obtained in reference 50 agree with those of Cragoe et a., section 1a, in theit region of overlap, 0° < ¢< 70°C, with ‘maximum deviation of 0.05% and an everage devition of (0.03%. (This is within the precision of either set of meat urements) Volatile impurities were concluded to be absent, bbesed on the observation that the pressure meacuremente ‘were insensitive to the volume of the vapor phase. The temperature scale employed was very nearly idential to LTS. 1927 above 0 °C. The experimental apparatus was similar to that used by Keyes [52] to measure isometries for ammonia, section I], 44. Keyes [50] regarded his results as confirmation of the accuracy of cragoe et al. and recom tended thee date over thei range 214. Bom Lewrence (18) Beattie and Lawrence measured values of vapor from 30°C to near the temperature of the ertieal point fan experimental procedure similar to that employed by Keyeesenction 2s. tthe lover vale of tomperatar results ae in excellent agreement with those of Cragoe ets, section Il, 21a, and with Keyes, section I, 2.le. The agree ment with Cragoe etal. is within 0.088 at 80 °C, but the difference increases slightly with increasing Values of tem perature to just under 0.1% at 70°C, with an average devia- tion in this range of 0.05%. (The agreement is within the curacy they claim for their measurements.) Above 70°C the data agree with those of Keyes to within 0.1% up to 110 °. Differences incresse to~0.2% at 120°C and then uniformly 00.8% at 182°C, the highest value of temperature reported, the Beattie and Lawrence values being consitenty higher throughout. Beattie and Lawrence made tests similar to those of Keyes for the presence of volatile gases: the cm? to about 80 cm" at each temperature, over which range they found the vapor pressure to be insensitive to the vapor volume. Their paper also includes P, g, T date of very high ‘quality for superheated emmonis. The experimentel ap tus and procedures are discussed in section II, 4.1. ste, Crarteet end Cheegee [68] The vapor pressure mestrement of Orerstreet and include the temperature range from =258 °C <1 < SE The date for ne iid icied resale at I valves of temperature spaced at nearly uniform intervals, including data forthe triple point, Also measured were the hest ea ie forthe saturated solid and liguid, the heat of fusion, and the heat of vaporization st the normal boiling point. These are discussed in section , 3.34 Values of temperature were measured with a special sold resstance-thermometersheaer, but the primary standard was 641 ‘thermocouple, which was calibrated with a hydrogen gas- thermometer, and which was compared from time to time with the gas thermometer. Such a comparison wat also made immediately following thie experiment and the values of lemperature measurements were adjusted accordingly. The precision ofthe gold resistance thermometer was 0.00! K, as thermocouple, The pressure mearuremente were made with a high precision manometer which could be read to within about 0.01 mm. The precision achieved for the pressure ‘measurements is at least an order of magnitud greater than had been achieved inthe earlier NBS ammonia work. Despite the fact that the measurements were of very high presionyit oul to determine the overall acursey ofthis experiment, This is becauee the values of temperature to which the measurements of pressure correspond cannot be precisely related to values on any of the defined practical temperature scales. The values of temperature reported by Overstreet and Giaugue forthe triple point and the normel boiling point are respectively 0.04 °C and 0,07 °C lower than, tore reported for the NDE weacurcuntoy sections Hy ly 2b If we adjust the values of temperature reported by Over- street and Giauque to be consistent with the IPTS 68 value for the ice point 273.15 K (the value they used was 273.10 K) and convert the NBS data to IPTS 68, then values of Over- street and Giaugue ate lower than the NBS values by about 0.01 °C and 0.05 °C atthe triple point and the boiling point, respectively. Ths difference is consistent with the authors” ‘observations that their temperature scale canbe in error ( Js differ from the thermodynamic scale) by es much as 0.05 “C2 Furthermore the primary temperature standard forthe experiment was a thermocouple, and, from the data in their table Tit appears that this stendard had « precision of be tween 0.01 °C and 0.02 °C. Thus, even though their ance thermometer could be reed to higher precision than this ite, nevertheless, the thermocouple with which the gas ‘thermometer was compared and to which absolute seeuracy must be referred. (The thermocouple was an integral part of the experiment and the resistance thermometer was contin- ‘ally calibrated to it.) Note that near the triple paint an error oF 0.04 °C rorulte nan oor of 07% in vapor pressure, and near the boiling point an error of 0.05 °C in Temperature results in an etror of about 0.2% in vapor fhe vapor prestre t the ammonia triple point was re ported 10 be 45.58 mm Hg. The value reported by McKelvey and Taylor was 45.2mm Hg. The difference of epproaimately 0.4 mm Hg ie outside the sombined press ‘ment, and is indicative ofa small systematic descrepancy be- teen the two experiments. An eor in pressure of 0.4 mm Hg corresponds to an error of 05% in the vapor pressure at the boiling point of ammonia. 2, Det [54-86] Date reported measurement for the vapor preseure of am: rmonia for the temperature range 25 °C to the critical he ve of ete ut fet Ort ad Gian IPS Sie rcv irene tee pa 2.Phys. Chem Ra Det, Vl.7,Ne. 3, 1978642 temperature. These measurements were past ofa program to obtain isometries forthe vapor and the liquid [54,55] and for the critical region [56]. A discussion of the experimental methou is given in section H, 43. The vapor pressure data contain larg differences from those reported by Cragoe et al, Keyes, and Beattie and Lawrence, discussed in setions 11, 21a, 2e, 214, forthe range 25 °C < « < 70°C. (In this range these other data are in excellent agreement) The ma imum differences occur at 25°C for which the results of Date are about 1.0% too high. At 100°C the Date measurements are 00 low by about 0.25% (as compared to Keyes and Dea tie and Lawrence). At values of temperature near the erica) point the Date measurements tend to average the differences Detween those of Keyes and Beattie and Lawrence. 2g. Gores (57 Recenily, a part of sure eavensive set of Iuehurle Fy T measurements, Carnjost reported a st of very precise data for the vapor pressure for the range 828.495 < T'< 403.588 K APTS 1968). The desis ofthis anperatus and the exoer ‘mental techniques are given in section Il, 42 ‘These data in their region of overlap are in excellent ae- cord with the results of Cragoe et al, section Il, 2, 10 . The maximum difference is about. 0.08%. They tend to somewhat lower values than the of Beattie and Lawrence, section II, 2d. Near the al point the difference is about 0.19%. This same aopa- us has also been used to measure the vapor pressure of water in this pressure range, and these are consistent with the very accurate dats of Osborne et.al. [58] for water to within 0.007%. We assign aus urerall accuracy vo these data of 0.05%. 2h, atop ect Finally, vapor pressure measurements have aso been made to study effets of isotope substitution. Such experiments are eigued primarily w iehd vapor promure Ulfferenes be tween particular isotopic modifications rather than to be primary data for the individual species. Among these are ‘measurements for the effect of deuterium substitution [59-64] and N‘** substitution [65]. 2.1, The Vapor Pressure Equation for he Lgsd-Vepor For this correlation, for values of temperature from sbout 10 °C above the triple point to about +70 °C, we have selected the data of Gragoe, section Il, 21a, The value we recommend for the triple point temperature is 195.48 K (PTS 1968), which is the value reported by McKelvey and ‘Taylor seston Il, Alby and the value forthe correaponding pressure is 45.58 mm Hg, which is from Oversireet and Giaugue, section II, 2.1e. Above +70 °C the selected values are those reported by Garnors, section II, 21g. at ited he etd ale apt Free othe ion recommended by Bachr et a. [66]. This equation tas only adjustable premeters not incadg the Sloe temperature and pressure atthe ert pot 1Lhys. Chem. Ret Date, Vol. 7,Ne.3, 1978 L. HAAR AND J. $. GALLAGHER loePP=P = PeLtadmoradinor +A(l-wF*+4Al-oF |, Qa) wi ern. ‘The least squares fit to the selected data yields the coei= cients shoven in table 1. ‘Toate. Coeficien 4m i055 ae-aniene ‘These coefficients differ slightly from those obtained by Bache et al ina fit by them to much the same dats. (They used the value given by MeKelvey and Taylor for the triple point pressure.) The differences in values forthe coefficients in most part result rom the fact that our choice for the value of temperature at the ertieal point is To=405.4 (PTS 1968), s2e section Il,2.. This is 01 K lower than the value used by Bachr et al. Extrapolation of the Garnjorst data to this value of temperature yields P=111.85 atmospheres. Equation (I)is oneiotent wth the selected date to i their accuracy. 2.2. Specific Volumes for the Saturated Liquid ‘end Vapor 2.20. Crogoe ond erp [42] (Sturted Liquid) ‘The main body of these measurements incudes 58 deter: minations ofthe specific volume forthe saturated liquid from the values of temperature atthe triple point to 100 °C. Meas lurements were concentrated mostly in the region 50°C < ¢ ‘< +50 °C, Six measurements were taken above +50 °C and only a single measurement below $0 °C (near the te point). The measurements wese uiade with imo type of spe cially designed glass pycnometers, one designed for use a low values of temperature and the other for high values, but with considerable overlap. The measurements include deter ‘inations of the otal mass of ammonia, the ttal volume of the pyenometer, the volume occupied by the liquid and the value of temperature. The pyenometer designs were such that Se worsein for the presente uf ante vapor map sll ciently small that only approximate values of the vapor volume were requited forthe correction. ‘Thete measurements were taken over a period of several years, with a considerable amount of data overlap. The re ‘uls are grouped in eleven sets, with each set corresponding to a particular ammonia sample used in particular pyenom- lets and to « pasteulas set uf expenivetal procedases. IeTHEDMO AYN AMIE PROPFETIEG AF AMMONIA aan all, seven different ammonia samples were used, the purifice- tions of which were achieved by four entirely diferent proce- dares. Au lalate temperature contol ited th mal gradients and assured a temperature control to within 0.01 °C for periods of half an hour (the time allowed for equilibration in most cases). The effect of temperature on the velume of the pyenometer wae determined by two corollary ‘experiments, on for the bulk expansion of the actual pye- rometer and the other for the linear expansion of the glass taesial ow which Ie was comrucied. Gompressblly effects forthe pyenometer itself were determined by experi ments in which the “known” value for water was employed standard, Al data obisined are remarkably consistent, with « scatter of about 7 parts in 10%. The authors claim an ‘overall accuracy of one or two pars in 10,000 forthe entire ev of data. This estate we quality to the eeglon ~90°C = ‘<+50°C. At higher values of temperature (above +50 °C) we vould suggest an uncertainty of 3 parts in 10,000 and at val ues of temperature below ~50 °C, about 5 parts in 10,000, ‘These estimates are based on the fact thatthe specific volume corrections forthe vapor present are less certain at the higher values of temperature, where the results are most sensitive to nts Aw at gh values uf seasperasare she seauls seid so have a grester sensitivity to temperature control. An uncer iy of O01 °C in temperature by itself can produce an ‘uncertainty of over Y part in 10,000 at 100 °C. Finally, the region above 50°C included only 2 few data points. The pau city of data below ~50 °C and the fact thatthe experimental procedure ihe bath) had to be modified somewhat for the one tive estimate forthe tabulated results below ~50°C. 2.20, Kayes ond Seowales [67] Seturoted Liquid) Values for the specific volume of saturated liquid are re: ported for values of temperature inthe range ~50 °C < 1 < $190 °C. Only smoothed soruhe ot 90°C fn tabulated. These were obtained as part of s set of measure iments of isometries forthe liguid by an extrapolation of the isotherms forthe liquid to the saturation pressure. Discussion ‘of the experimental technique is given in section Il, 44. The values for specific volume agree to within 0.2% with those of| CCragoe and Harper, section Il, 2.2; the maximum deviations ‘eur atthe high and low value of temperature, with the Keyes and Brownlee results consistently lower. Keyes in a subsequent review [68] accepts the results of Cragoe and Harper throughout their range. 2.26. Cragoe, Mekalvy, end ©'Comna 4] (Stvrated Vapor) Specific volumes f for values of temperature inthe range ~50°C < 1 < 450°C. ‘The experiments for this research included two quite differ ent approaches, The measurement results were used to es: tablish the valiity of values calculated from the Clapeyron equation. ‘Two different experiments were deseribed; one is a direct method using many of the ame techniques employed in the the eaturated vapor were determined measurement of volumes for the saturated liquid, seeion 2.2; the other isan optical method based on the relation be- tween the fan of vel ‘The sess of the two experiments at low values of temperature tend to differ by large amounts, the difference being several percent at -50 °C, but Cragoe et al. argue that the two methods tend to have errors in opposite directions, and that these errors should be 4uite small at the higher temperatures. The experimental re- sults for the two methods do agree to within 0.1% near 50°C. ‘The values obtalued Grom she Clapeyron equailon wn vapor pressure, latent heat and liquid volumetri data nearly split the difference between those for the two experiments lover the experimental range. The difference between the caleulated values and those of either of the experiments in- ‘creases slowly with decreasing temperature to approximately (03% at 0°C and to more than one percent at ~50 °C. The ex: fecllent agreement for te three memods at high values of temperature, where the experimental errors were claimed to be small, is an indication of the accuracy of the results calculated from the Clapeyron equation. The date for the ‘vapor pressure and specific volume ofthe liquid used inthis ion have already been discussed, sections 21a and 2.25 Values forthe latent heat of vaporization are from reference 50, ue of she elasleeaperlaets of se wld NBS program. 9 discussion ofthat experiment is contained in ection Il, 3.3 As indicated therein these data are believed accurate 0 with- jn #0.15% over the entire range. The derivative ofthe vapor pressure with respect to temperature is assigned an uncer- ‘inty of about #0.2% for the range ~80°C << $0°C. The specific volume for the saturated liquid is uncertain by about (2O.02 oie sete TD 28a dg In addition tothe above listed uncertainties in the data, another possible error arises fom the fact thatthe Clapeyron ‘equation i stietlycorreet only when the measurements refer to thermodynamic values of temperetore. That is, the Clapey- ron equation isnot an exact relation for practical temp ture scales. The Clapeyron equation is a direct consequence ofc eve a day be ven tow vse e here Lis the latent heat of vaporization, @ ie the thermody- namic (shsolut) value of temperature, u’ and u refer t0 the specific volumes for the saturated vapor and the saturated liguid, reapectively, and P is the vapor pressure of the saturated liquid, Let us conser the transformation required to convert eq 2)10 the temperature scale ITS(1927), ‘The phase change associated with liguid-vapor equilibrivm is independent of the temperature scale, that is, there is value of temperature T on the practical sale (TTS 27) that is represented by Teen, 6 such that LORLD, PORAN, Ga) u@=u(N, and (Ow (D. 4.Phys Chem. Rel Det, Vo-7, No.3, 1978oa From eq) and eqs Gwe obtain 4 pon(2.r) 1-H af ron( gan) ® where the approximation du/d6¥du/dT hasbeen employed wihervrof the order (4°). Combining eq 2.3, we ob Ure Tw Tuya 8), ® where CCragoe etal. used eq (5) with AO. In general y and du/dT, are not zero, and so therefore, B40. Asan illustration, suppose that IPTS 68 is nearly identical with the thermodynamic scale and take ITS 27 as the prac- tical scale used by Cragoe etal. We evaluate A for this hypo thetical situation. From eq (3) “WT Bs and TyeTirtt © and Bam Estios where the superscripts at the ice point. Combi “o" refer othe value of temperature ng eqs (6, we Giada o Alo, ‘The value used by Cragoe et al forthe ice point was 278.10 K. The IPTS.60 value forthe fee point 279219 K. Values fos the differences fy—tuy and their derivatives are tabulated by Douglas [69]. (Note thatthe quantities tabul refer to differences involving ta. However, tay and tar are nearly identical in this range.) The maximum correction in percent occurs st about ~25 °C for which value A&—0.04% ‘This correction decrease with incre ture to about ted by Douglas ing values of tempers 0.09% at the ie point and is negligible at about +25 °C where the error undergoes a sign reversal, At 8373 K the equation isin error by about +0.04% Recent results [17,18] suggest important differences exist between IPTS 68 and the thermodynamic sale inthe range 0 $1 150°C. Ifwe refer to reference 28 an indication of the Jifference between ITS 27 snd the thermodynamic tempera tare senley then the correction required would be approx: imately 0.10% at 373 K. Fortunately, the results in reference 2 suggest thet over most of the range ~50 °C < 1 < +50°C the difference between the scales is comparatively mall, 0 ‘hat the correction inthis range is only about ~0.05% ‘The combined experimental errors arising from the values used forthe latent heet of vaporization, the derivative ofthe vapor pressure dP/dT, and the specific volume for the 2.Phys. chem, ete, Vl.7.No, 31978 L. HAAR AND J. 5. GALLAGHER saturated liquid contribute an uncertainty in the value for the specific volume ofthe saturated vapor u', of about 0.20% for values of temperature 0°C < £ < +50 °C. The uncertainty from this souree increases below 0° to about 0.030% at ~50 C. Because the distinct possibilty exists that the difference between IPTS 68 and the thermodynamic scale are atleast as large as those between ITS 27 and IPTS 68 (and may even be of different sign), corrections owing to the error in the Clapeyton equation were not made by vs. An additional un- retain of approximately 0.05% should. therefore. be assigned. The Critical Temperature and the ‘The value for the critical temperature is a key parameter in ing law analysis. I is unfortur large one in reported experimental measurements for the J tempersture of ammonia, All such measurements were made many years ago, with the most recent more than fifty years old, The value most often quoted is from a review made by Pickering [70] in which results ofall experiments subsequent to s particular date near the turn ofthe century were included in a numerical average, The measurements in- ‘ued a dias veview Indl a apres ie evial semper tures reported of over one kelvin. We decided to retain the value ofthe erticl temperature as a parameter, and to at tempt to ascertain « "bes value from the eiteris associated with the “goodness of fit” ofa scaling law equation (71, 72) to the experimental specific volumes for the coexisting haces. sion for the reduced density as a function of non-integral powers of a reduced temperature, From referenges Tl, 72 Ae IAT BEHBOTIM+BSIATR MS BNATI-M), hore Ag= Yoke, OT THT wal ey Ty ste values or ae critical density and critical temperature, respectively. The superserits to the coeicients in eg (refer tothe two phys Jeal branches, corresponding to above (+) and below ( eal density. The quantities 8 and e that the ex ponents are universal sealing law parameter, taken for this ‘work as P=0.35 and e=1.85. Continuity imposes the condition Bisals, (G2) Equation ®)has been successfully employed to corelate date temperatures as low a4 AT=-0.5 fr the liquid branch and to temperatures below AT==0.3 for the vapor branch [73} ‘We applied eq (8) using forthe liquid branch data from Gragoe and Harper, and using for the vapor branch data from Cragoe, MeKelvey, and O'Connor. The data for the ‘equation were restricted to values of temperatures 0°C and above. In preliminary set of ealelations we determined the oeficents Bs by least squares ft for each ofthe two branches of the equation to the experimental data. A family of solu- ‘ions (101 in total} wes obtained, exch corresponding to pat ticular selection of values fr the parameters Tos @e- The =THERMODYNAMIC PROPERTIES OF AMMONIA 645, tions were arranged to form a closely spaced grid covering the range of reported values for the critical constants. The tans an the efficients eqs a) were then impos ‘these resulls, and a best ft wat selected baced on the statis- tical data for the standard deviations ‘Aweak preference was indicated forthe values T=405.4K 3.90 mol/L Ch 1 is interesting 10 note that this value for the critical temperature is about 0.] K lower than the value recom- mended by Pickering in hs eviee. In order to obtain smooth values at the ertical point, the Jneture ofthe two branches of eq (final fit was made in eae (@s) were in fitting process. In ‘cluded as data were the values for the critical constants Te, listed in eg) above, Table 2 contains values ofthe cet ‘cients o obtsine AML Coen of rth dome f coring pes = Hi z r [an | mtaor 28 ae esta ee 2s On the gas side the root mean square ofthe fractional de vation of calculated densities from the experimental is ap- proximately 0.14%. On the liquid side, it is about 002%, The raeulte are within the experimental accuracy of the vapor data, section 21, and within the experimental acuracy of the liquid data for temperatures above +50 °C, section II, 2.2 ‘The calculated value forthe liquid density at 120°C agrees with the measurement of Keyes and Brownlee, section II, 2.2, almost to within 0.1%, which datum was not included in the fi. 3. Calorimetrie Data Included in ths section are direct calorimetric measure: ‘ments and statistical mechanical calelations for the proper ties of ammonia inthe ideal gas state. The statistical marhan- ical calculations are discussed in section 3.1. Section 3.2 con tins & discussion of the experimental measurements for the constant pressure heat capacity for the vapor phase. Calori- reirie data for the coexisting liquid and vapor phases are discusted in section 3.3, ‘We note that the statistical mechanical calculations neces- sarily refer to the thermodynamic temperature seale (16) ‘he effect on the thermodynamic functions due to differ- ences between the thermodynamic seale and the practical scale IPTS (1968) discussed in section Il 3.2 and in section mi. 43.1. Colorimetric Properties forthe Ideal Gos ‘This deqwasion i limited tothe work reported hy Haar [4]: 1 review of earlier work i given therein. The properties al- culated in reference 4 relate to the ideal gas state atone at mosphere pressure, and refer to the ideal ga state at 0 K as the 2ero of energy. The results include the Gibbs (ree energy) function, the enthalpy function, the heat capacity at constant pressure and the entropy, all ealeulated at closely spaced temperature intervals. The tables are given in dimensionlase units for the temperature range 50 K < T'< $000 K. A high degree of accuracy was achieved for properties calculated in the temperature range of interest for this correlation by means ofa very detailed analysis for the highly anharmonie outofplane vibrational mode. As part of this analysis {generalization was made ofthe representation for the it feasible to include the effects of coupling ofthe individual vibrational stats of this mode with the other internal degrees of freedom. ‘A detsiled examinetion was made ofthe various sources of error and it was concluded that for values of temperature below 1000 K the ideal gas properties of ammonia are a ae curate ar are the ideal gen properticn for any polyetomnic ‘molecular substance for which calculations are avsilale. The principle sources of uncertainty aree from lack of structural formation for the high energy states and from simplfice tions employed inthe statistical analysis, both of which tend {orbe important atthe very high values of temperature. (The simplification employed in the statistical analysis also com: ‘romizee the rerults at very ow values af tempereture) Thus the heat capacity value at $000 K is uncertain by about 3%, but the uncertainty falls sharply with decreasing values of temperature to about 0.05% at 1000 K. At the very low tem: perature limit the uncertainty is also comparatively large at 50 K itis perhaps 0.3% to 04%, but this unce creases sharply with increasing values of temper about 0.08% at 100 K, and less than 0.01% at 250 K. For values of temperature in the range 250 K to 425 K, the overall uncertainty is no more than 0.019%. The uncertainty tends to increase slowly with temperature for values of temperature above 425 K, but i still no more than 0.08% at 750 K, In the temperature range of major importance for this corrl 200 K to 600 K, the ideal-gas values approach the accuracy by which the gas contant ohne Reference 4 contains comparisons with de obtained in previous calculations. Several of these have had wide use in engineering applications [74, 75] Important differences (er ofreference 4even at ordinary temperatures. ‘The Gibbs (ree) energy values forthe ides! gas at 1 atmos- pore preseute were fted, in the least squares sense, to the following equation over the temperature range 100 K to 1000 Ke oe ge os Fp mend+ Dow, ao) Th Ser te wed he te nea wes The ae {1.Phys Cham Ref. Dote, Vo. 7, Mo 3.1978646 LL. HAAR AND J. $. GALLAGHER ‘where @ isthe value of thermodynamic temperature in kl ‘ins; Ris the gas constant; Eis the energy forthe ideal gas ‘OK. The coefficients a, are listed in table 3, The heat capacity values and the other thermodynamic functions cal culated from eq (10) for the temperature range 100 K < 8 < 1000K agree with those tabulated in reference 410 within the aeuracy of tose values. ‘Tobe Coffee ibe re nea fection for theese asssae10 13.2. Constont Pressure Heat Capacity Measurements {or the Vapor ‘This discussion ic limited tothe date of Osborne, Stimson, Sligh and Cragoe [7] which extend from sbout ~15 154+150 °C ant from shout 1/2 atmosphere to appron: rately 20 atmoaphcren except atthe lower values of tempera ture where they extend in pressure to within a few degrees of superheat of the saturated vapor states. This is one of the classic experiments inthe history of experimental thermody ‘namics. Systematic errors for these measurements have heen shown [2, 3} to be in the neighborhood of 2 or 3 parts in 10,000, making these data probably the most accurate of, ‘hee kind available for any tubetance. Infact, the random scatter in the measurements are the major source of exper: imental error Also included in reference 7 are 20 measure rents of the Joule-Thomson coefficient and 44 determina tions of the heat capacity using an estlier prototype of the calorimeter. These latter data were labled by Osborne eta. as discredited and were not inchuded in their final results (I is interesting to note that those “iseredited™ data ate of ‘quality that would satisfy ll but the most demanding of modern scientists.) ‘The experiment was performed with «flow calorimeter, In such a device a steady stream of vapor flows through the calorimeter at a steady and accurately determined rate ‘Measurements are made of the amount of increase in the value of temperature af thie vapor dete the sddition of hea ‘which is also added at a steady, acurately measured Elaborate procedures were employed to maintain the steady stale conditions required. For exemple, the temperature con- {col bath in which the eslorimeter was immersed (one of three baths, the other two being used to contol conditions up- stream and downstream from the calorimeter) was main- tained constant to within 0.001 K for time interval 9 ong a8 1/2 hour. (This wat achieved without the benefit ofthe ex tremely sensitive automatic contol equipment that is stan dard in laboratories today.) 4LPhy. Chem, Dota, Vel 7, No. 31978 ‘The calorimeter was about 15 em in length overall Gy standards of that day, it was a minature calorimeter). The small size facilitated repid equilibration. The design also per mitted considerable flesibility in the range of operating con ditions. Thus the amount of ammonie required for an experi ‘ment was small, which in turn permitted wide variation inthe ‘ow rate between diferent experiments, andy faethe Now rates were varied by more than a factor of 10. This greatly fe ciltated the analysis and correction for flowratedependent best leaks. The miniaturization was not without its draw: backs, since the amount of material on which the measure: ments were based wes aso small, and this tended to increase the uncertainty in the flow rate, But these stat-stop errors, tended primarily increse ihe raunows seater of the me lrements, This eatter could have been reduced by increas ing the run times, but compromises in this direction would aost likely have been atthe expense of increased uncertainty in the steadiness ofthe experimental condition, thereby in- ‘creasing the likelihood of systematic erors. As a consequence of the particular set of compromises made, the experimental large a© 0.2%. But, as we shall show, the system: were by comparison almost negligible. Perhaps the most interesting feature ofthe experimental technique is associated withthe analysis of heat leaks, (nei dentally, the principal reason results obtained with the earlier prototype calorimeter were rejected was the uncertainty in foseming this rource of error) The heat loco in the final ‘design rarely exceeded 0.2%. On the basis of considerable preliminary experimental measurements, Osborne etal cm: ‘cluded that the loss wae independent of flow rate, and they derived provi an expression o correct their measurements for these losees, The general design and measurement tech niques established by them for this purpose are standard todays however, their achievement in reducing thece heat leaks to such emall amounts and in rendering them indepen- dent of flow rate to our knowledge has not since been ex- celled, ‘The temperature messurements were made with platinum resistance thermometers and, a8 indiated in section I, 1 refer tothe LTS. 1927 temperature scale. The ammonia sa ple tuficcntly pure that etrore from contaminante were igible. (See the discussion of the purification program in section Il, 1) There is, however, an uncertainty of about 1 part in 10" in the conversion ofthe data to presen accepted nits of energy (due largely tothe fact shat the instruments vused are no longer available for direct calibration; even if they were, there would sil be uncertsinties due to various period Inference 2 these data were compared with the ideal statistical ealeulations of Haar discussed in section Tl, 3.1, tnd in igure I we include the results ofthis comparison. In the figure the fractional difference (in percent in values for the constant pressure heat capacities ae ploted as ordinates eras valine tomperatine ax aberiees At nach pit shan in the figure, 4C+ refers to the ealeulated ideal gas value Ce ft & particular value of temperatute on the thermodynamic scale minus the analogous measured value extrapolated to‘THERMODYNAMIC PROPERTIES OF AMMONIA a7 Pent 1, The ration dite oxreposting sperine nein theta capcy ve tempertureetwen the ea aolitediosero presse Case sion 1,32. Te major pt oh tea the ema difereces are cid odiferece between he practi tmpersoe wl ted in he me ‘ements and th thermodvnamietmpersae ele ueediathe exactions zero pressure Cy fr the ssine numerical value of tempers ture on the practical scale used in the experiment. As stated Inseetion Ih 1.0, this practical seal is nearly dential to 113 1922. A smooth eurve was drawn through these diferenees. (The Cyao values used are the smoothed extrapolations given in One a a) As indicated in figure 1, for values of temperature below #30 °C the Cyan data agree with those for the ideal gas to within 2 parts in 10,000 tha is, to almost within the accuracy ofthe ideal gas results) Note that the experimental values below “15 °C are extrapolations in a region where the io: therms exhibit sharp curvature, and s0 the agreement below oC ie lbely af temperate fe 430°C to +150 °C, i is apparent that small systematic di ference exists which is monotonic in temperature, with = max imum difference of 1S parts in 10,000 t 150°C. ‘tas shown inreterence 2 hati the small systematic dt ferences indicated in figure 1 are ascribed to the difference between the thermodynamic scale (lo which the statistics ealelatinns necestavily refer) and the practical tempersto scale (to which the measurements refer) then at 100 °C the thermodynemie scale would be about 0.035 °C hotter than is the ITS 1927 practical scale. This result also indicates that the thermodynamic scale is hotter than the IPT 1968 scale by this amount. This conclusion is consistent with results of recent experiments in gas thermometry (17, 18] in which it seas found that the dhermodynamia cea at 10 °C wae hatter than IPTS 1968 by about 0.080 °C. (Section II, 4.1 and sec tion IL contain discussions of the effect on the overall core lation of differences between the thermodynamic scale and ITS 1968) It would appear From this analysis that a major part of the small systematic differences indiceted by figure ) could indeed be esribed to differences between the tempers: ture ccalas, Thos the experimental measurements of Osborne, Stimson, Sligh snd Cragoe and the caleulation of Haar for the ideal gas would eppear to agree almost within our ability to interpret the 50 year old calorimetric data, that is, 10 near. ly 2 parts in 10,000: 3.3. Measurements for the Coexisting Phases ‘3.36, Lent Hea of Voperzation fom ~A2to +80°% Calorimetric measurements relating 10 coexisting liquid vapor phases were made at the NBS with an aneroid calor meter specially designed for the ammonia measurements by Osborne [76]. Such a device has heen ueed previouly at NBS for latent heat measiements fr ie, and that ealorimeter appropriately modified, was employed in preliminary mese urements for ammonis. But the data so obtained were ‘thought to have large uncertainties, As remarked by Osborne, ‘While the results ofthese experiments would pethaps have met the immediate needs ofthe engineering profesion, they didnot promise to yield final values which would bear retest srietifie aerating...” ‘The Oshorne calorimeter was designed to operate in the range 40 °C whee 2090, 2% siete 22 ae Pee 28, ad 2Q ag Beetaglg Soe ae with Q given by equation (14. ‘teach experimental C data point, the corresponding con- tribution of the P.e,T surface tothe ealeulated value of Cy aiven by eg (15) was calculated. This was then combined with the Ideal gas Cz values to obtain calculates neat capacity values C, corresponding to the temperature and pressure of cach experimental C, data point. Thus Cac, a9) was calculated fo each experimental state ‘We note that over the range of the FQ. data the L, are not sensitive to small differences in the scales. (This conclu- tion is reedily deduced aso from the anslysis in reference 2) Since the CF refer to the thermodynamic seal, this ix the scale to which the ealeuated C, should refer, We now compare the calculated C, with the experimental data by examining the difference ACEC HC a) ‘The sesult are illustrated in figure 8. The ordinate is the %% fractional difference (AC, /C, *100) and the abeciss is the value ofthe pressure in bars. The various straight line seg- ments connect the A C, isotherms. Each line includes points near (fo within a fraction of a degree) to the temperature value designated in the figure. In those eases where two or more measurements were made at temperature-pressure ints veuy clove wogeter, the plowed polms are averages. For several of these, scatter hars are shown, A statistical analysis of the results for AC, shows that the fractional diference is appronimetaly 0.20%, and the rootmean square fractional difference is 0.28%. This is essentially identical to the scatter of the experimental C, hich is 20.20%. The largest individual difference is 58% which occurs for s date point very close tothe satu ration curve, which point is beyond the range of the P,gT lata (being closer to the saturation curve by over an phere than are the data in eats 1 and 2) and icin a here the contribution ofthe P,g,7 surace is extremely ten- sitive tothe pressure. There are few P,g,T data points fr iso- theems below 50 °C, henee the devition found at low tem- peratures are of litle consequence inthis comparison. In the pressure-lemperature region for which the experimental PT data are sufficiently dense to fix” the surface, we note thatthe magnitude ofthe deviatio are in moot eases emallerTHERMODYNAMIC PROPERTIES OF AMMONIA 653, "oun 8. The ection ference ve rere between the ele bat eapsces Co ‘te the conepenngesperimental nts pim. The tempertre tle indi minal vale iin ere of the ata exe Tow tempertce ee tribe {mal emai ted forthe eens ey emi rl large deviation fer the tbr erase scales ee gr than the experimental scatter ofthe C, data. There is, how- ‘versa definite bias to positive deviations. But even this smal systematie difference Is nearly eliminated if we take account of differences in temperature ecale. We can approximate this effect by correcting the AC, values in accordance with the <6C2in figure 1 Exeept forthe love temperate di rection brings all deviations within the C, cea We have thas demonstrated thet the Pa surface we have derived, eq (14) and table 3, is thermodynamically consistent {or accurate) to within the satter of the experimental C data ‘The scatter ofthe C, data is thus a measure ofthe accuracy of thik ee likewise, the scatter of the da (14) indicates a seatter (and thus an erzor) of less than = ‘UU3% for these 1) dat, Values for C, are tabulated in table 5 at coarse intervals over the range of the Pe.T data. Consistent withthe forego: The diseusion following eq ing srgronente, the seenwary stint fon approximately equal tothe seatter in the experimental C, data. We assga to these data an uncertainty of 20.3% forthe region 50 °C < ¢ < 300°C and preseure less than the maxi ‘mm tabulated to each wotherm. Ihe remainder ofthe table is asigned an uncertainty of 20.5%. This higher uncertainty is asigned atthe lower temperatures up to 50°C, due to the pancty af Po.T in that reginn The route given in tahle § willbe used as basic “ata” to aid in the evaluation of ,g.T ata in subsequent sections, particularly in section I, 42.In section IL the overall results of the correlation wil be tested 1 0 Pihael 15 20 (1) of etn ach aterm i ts) Toe rate age evan tthe 0 alae cif PT dts. Tae i ecu ake of he illeence in em 4.2, Dota Set 3 CGarnjost [37] made measurements of very high accuracy for values of temperature inthe range 60°C << 300°C and for values of pressure from below 50 atmospheres to 750 at- chores, for specific volumes from 1.8 em per gram to 25.5 cm! per gram. A total of 148 date points were reported. A ‘numberof the data points were taken near the liquid wapor tical pomt and, for the lower densities, near the lie of coexisting phases. Vapor pressures were reported for values of temperature from near 70°C to within several kelvins of, ‘The region of the thermodynamic surface and the setusl deta points included by these measurements are shown in fig. 4, which is « PLT plot with the origin taken at the critical point (Also outhine in the tigre are the regions covered by ata sets 1 and 2) As can be seen, the Garnjost date overlap those of data set 2 and extend the range of those dats to ‘The apparatus used included a specially designed eylindr- cal pressure vessel which was immersed in « carefully con- twolled bath. The temperature ofthe fluid was messured with 4 platinum resistance tnermometer positioned along the axis fof the pressure vessel. A second platinum resistance ther- rmometer was used inthe control nd regulation of the bath temperature Presi ra ade with w dead eight gauge. Special care was taken for the specifi volume determinations to ensure complete filing of the pressure vessel. (Preliminary results had been compromised by in- ‘complete wetting ot the walls of the preseure veel, which 4. Phys. Chem. Rel-Date, Vo. 7,No.3, 1978656 L. HAAR AND J. $. GALLAGHER TABLES. Vales for Cfo gu phate 125 bar nd 328°C Pia we a = 7 2 z z © =o. 71395 “Tas Tae as ae 45% a3 ° aus 1309 133 43800 sa ‘sm aan rr ‘os6 ‘e108 ‘9s ‘sor 75 asst 903 ‘sm ‘5082 6562 98 52048 108, 45508 ast sso se 50908 ‘$700 ses 135, ‘iva ‘658 a2 6 42509 ‘a? soot 150. as ata ams ‘hoo ‘ee ‘08 S0ai6 1, 59 ‘ass 4296 ‘908 souis 19563 S649 200, ‘ows 4950 5688 ‘se soot Sona Sis 2 S098 5% Sse S078 50936 51228 S198 40 S192 51s E166 sim Sees S2u9 52199 m5. s24%6 52590 5.2508 52693 5.2002 53021 53576 0, 534 505 5342 53638 53138 539% Sasi ES so sata 589 56573 54658 5.4828 55258 Pos) ec 1s = % a 7 106 i. % 5 is Sta e908 ous 1. S382 S708 10907 ‘905 15, 5268 sa 558 75036 150, sit S30 ss 22 0 $2193 sa sae St ease 2 S282 S37 5456s 5104 68108 10499 0. 5345 snes sam Sau 250 e708 ae Saot6 S08 ss018 480 8 rea, 3, 55605 Salas S658 est ome est resulted in erratic values forthe epecifie volume.) From an analysis ofthe possible sources of error, Garnjost estimated the maximum error to be 5 parts in 10,000 for measurements Inthe hguid phase and & parts in 10,000 for measurements in the gas phase. Using this same spparatus Garnjst (also in reference 57) reported measurements for water fr the above Iamperatira and prawn cenge (The mon water ae all in the liquid phase) The results fr water appear to be in agreement with existing data for water within the above accuracy estimates Ie isoenores cortesponding 10 the higher values ot densi: ty can be extrapolated to within several parts in 10,000 ofthe values of density reported for the saturated liquid, section 1, 2.2.On the other hand the values of density forthe ixnchore ‘at 26.5 em'/g are in secord with values calculated from the thermodynamic surface forthe vapor phase, section I, 4.1, ‘only 10 within about 0.2%, The differences tended 10 be largest at the higher values of temperature, with the Germ jst measurements systematically lower than the values of| ensity cleulated from the surface. Based on the analysis in ‘errors in the Garnjost data, To study the thermodynamic consistency of the Garnjost measurements and alto of the other high quality data further, vee perlormed an anelysis similar to that described in section 1, 4.1. We derived a much larger surface extending to 750 bar and including the thermodynamie region of date sets 1,2, {Phy Chem. Ral. Det, Vol. 7,No 3, 1978 data set 4 forthe saturated liquid, section 1, 22a, date set § for the ssturated vapor, section Tl, 2.2b, and the selected ‘values forthe vapor pressure, section II, 2.1. The thermody hammie surtace yeludes the data schematically shown on the PL -diagram, fg. 4 ‘The density, temperature and pressure range of the data 4 threefold inecesse in flexibility aver that forthe suztace i: cluding just the low pressure vapor-phase data, section II, 4.1. Thos eg (14) was modified by extending the range of the summation to sncuge values ofthe mndices given By r=9 and {F°6, but with 4, values set to 2er0 fori and j values given by i j and j> 4. Thus the surface includes 44 terms. The least aquares iting procedutes was further modified to include conditions for phase equilibrium. However, as we have stated in section I, 2.1b, the Gibbs condition was not in- ‘eluded as 2 constraint, but was used to produce values 10 be ‘compared with appropnately weightea dala forthe exper ‘mental vapor pressures. The result is that the Gibbs condi tions for phase equilibria ere satisfied, but only to an extent cancstent with the aronrany nf the vapor presae datn De tails of this procedure end the method for deriving the very small corrections to the vapor pressure to exactly satisfy the Gibs conditions are described in seetion I], 24, ‘The thermodynamic surtace 18 sensitive 10 the statistical weights used for the various datasets. As part of the deriva: tion we calculated several surfaces thet differed only in the la lod ta the Carnjoet data in the lenet‘THERMODYNAMIC PROPERTIES OF AMMONIA 655 rein squares fit, The weights ranged from a factor of 1/5 10 fac- tor of Sas compared with the weigh! of uaity assigned tod fete Land 2. The selasive weights for the saturated liquid, saturated vapor, and the vapor pressure values were and I respectively. Several festures of the results are noted 1) Except for the 2isochores at 26.5 and 15.2 em?/g, the surface fits all deta to within the Garnjost estimate for ac curacy, independent of the weight assigned to the Garnjst ‘data, The fit forthe saturated liquid data and the Garnjot data near the orticl point did no dege for eigniticant lation with increased weight forthe Garnjost dats 2) The two poorly fit sochores tend toe better fit with in creasing statistical weight for the Garniost data. but this ix at the expense of a slight degradation of the ft to the Garn jst isochore at 10.8 cm?/g. But more important the fit for datasets I and 2suffered major degradation. '3) The best overall result was obtained with Unity weight for the Garnjost data 4) The deviations for the two poorly fit isochores tended to be largest at the higher values of temperature. The values of ‘density forthe two ischores are systematically lower than the corresponding values forthe fitted surface, The isochore at 15.2 em*/g, which is just outside the region of dataset 2, ap pears to suffer she largest degradation, about 307% larger than that forthe 265 em?/g. '5) Significant degradation alco occur forthe fit to dataset 2 over that achieved in section Il. 4.1. This, however. is limited to the date points close to the Garnjost isochore at 15.2em/g ‘Based on observations | and 2 we conclude the data for the two SSochores at 20.5 and 15.2 em*/g are slightly inconsistent 1+ Meyers and Jessup 2- Beattie and Lawrence 3 Gamjost eo ate tat thee very hgh guy de entend ay clo the erie with the data of sets 1 and 2, but also with the other Garnjst data. However, the Garnjost data in most part are consistent sith the date forthe eaturated liquid and with data sete Land 2. Since the thermodynamic surface is very sensitive 10 in ‘consistent data, particulary to data forthe coexisting phases tnd near the critical point, the excellent fit for this surface to the data in these regions, strongly implies high degree of thermodynamic consistency for these dats. It is apparent from this analysis that the accuracy estimates given by Garn- joet are concervative, except for the 2 ieochoree 265 and 15.2, for which the inconssteney ean be several tenths of a percent. We conclude that datasets 1, 2,3.and 4in most part are thermodyaamically consistent within about this seme {general tolerance, that sto within 210 3 parts in 10,00. 4.3. Dota Set 6 Date [54] reported P, ¢, T measurements forthe iguid for S isotherms at 25 °C intervals from 25°C < ¢ < 125 °C. The pressure mesturements extend from near the presets of the ‘sturated liquid to 500 atmospheres, The experimental d were presmoothed and results were tabulated for the com: pressbilty factor and the specific volume at uniform pres- fures for each of the 5 Isotherms. Included by means of @ small extrapolation are values for the saturated liquid. This paper also contains Pg, T measurements forthe vapor, and measurements ofthe vapor pressure of the saturated liquid ate used a variable volume isometric experimental meth- od discussed in section II, 41. (A moveable piston varies the volume of the piezometer, For each volume the pressure is ‘measured st various temperatures.) The pressure measure: J.Phy. Cham, Ret. Dota, Vo. 7, No, 9, 1978656 ments were made with « mercury in glass manometer which could be read to £0.01 atmospheres. The temperature mess- ‘urements were made with mercury thermometers. Though these would be veal ty 20.01 K, hey were elaine to be ac> curate only to +0.05 K at 125 °C and 0.08 K at 25 °C. Date ‘estimated his semple to be 99.9% pure, with no more than 0.04% volatile impurities. The calibration forthe volume of the piezometer for various piston positions was established by preliminary measurements with nitrogen and comparison with the published nitrogen data of Michels (81, 82). Date sssigued au overall uncertainy for hie measurements of 120.2%. He indicated that his smoothed results agree with his measured values to within 20.1%. Date noted very large discrepancies between hie results and those of Keyes [52]. At 100 atmospheres he reported dif- ferences for the specific volumes of 0.94% at 80°C, 1.02% at 75°C and 1.55% at 100 °C. It appears thet these differences are mostly due to errors in the Eecond (smoothed) table ot Keyes. As we note in section Il, 44, this second table has large errors due to smoothing to uniform values of tempera i Aahle Antally the data ‘of Keyes and of Date agree to within about 20.3% to 20.4%. ‘We compare these date in the region of overlap with data ‘et 3, using the thermodynamie surface derived in section Il, 442, which closely fits the very quality data of set 3. The difference between values of density calculated from the sur face and the experimental values indicate a systematic trend of these dat to lower values of density. The maximum differ- fence (error) in the region of overlap is about 0.25%, and about half the measurements are within an error band of 01%. ‘Though these daue are of secondary quality comparea to data set 8, they appear to he of higher accuracy than the data ‘of Keyes, dataset 7. However, ite somewhat troubling that at the lower values of temperature, outside the region included bby data 4et 3, thatthe measurement appears to be inconsit ent wth the date of Cragoe and Harper forthe saturated lig: uid. As noted in section II, 22s those values are believed to be aecurawe 10 several paris in 10,000. The inconsistency at the lower values of temperature was found tobe about 0.3%. The results reported for the vapor phase differ in ceveral cates by more than 0.5% from those of Meyers and Tessup tnd those of Beattie and Lawrence in section Tl, 4]. Ashort discussion of Date's vapor pressure measurements i given in section H, 216 44, Dato Set 7 ‘These measufements By Keyes {92] extend trom about 29 °C to 210 °C and from the pressure ofthe saturated liquid to 1100 atm. Approximately half the data points refer to tem: peratures helow the eritial Several nf the maacnremente refer to temperature end density values in the neighborhood ofthe critical point. The results are presented in two tables of smoothed values of pressurevolume products. In the first table, the date are tabulated for uniform values ot specitc volume and the actul temperatures of the experiment; the second, the same dats are presented at uniform values of| temperatura and provera The dete im the ascend table are 1LPhy. Chem. Re Dot,Vl.7.Ne. 3, 1978 L. HAAR AND J. S. GALLAGHER not thermodynamically consistent at the lower values of tem: perature, and in several other places are inconsistent with the results given inthe first table. The second table also includes extrapol ‘ment. Since the second table ie more convenient to use, sev- eral investigations (5, $4, 83} apparently overlooked the first table and their analyses ofthis experiment are somewhat mis- leading. A further diseussion of this situation is tions 44nd 45 which follow. ‘The details of the experimental procedure were described by Keyes and Brownlee [91]. A premeasured (by weighing) quantity of ammonis was placed inastel “bomb”. A closely Stted, calibrated and threaded piston varied the bomb di- mensions. Pressures were measured with a piston gauge, Keyes published these results in 1981; however, the experi: rental measurements date back to 1913. The experimental data had been withheld from publication in anticipation of re-doing the experiment, primarily, i appears, Decause of concern shout sorption. However, st the time of publication, Keyes ecemed to fel thatthe principe limitation on accuracy won nt the afert of xorption bt eather was due tothe unce inty in the elastic constants for the steel used in the ‘bomb”. The estimate of the error due to thie uncertainty was given as 14% forthe volumetric determinations. Keyes sso Indicated an uncertaimy in the temperature scale rela tive to the values on ITS 1927, which he gave as 40.05 K 200°C. Davies [5] encountered difficulties ia his attempt to recon ile some ofthe Keyes data in the critical region withthe rest ‘of is surface by a graphical method. He suggested that there are errors in the data which in some cases, are ‘or 9 percent. Razarnowsny [es] compared res measurements (data set 8) with Keyes results for the 200 °C isotherm. The agreement was o within about 0.6%. We com: pared values of the density and compressibility for the ‘ated liquid obtained from a small extrapols eta to the coesistence curve with values for these properties reported by Cragoe and Harper, section I, 2.2a and Osborne ana Van Dusen, sections Hi, 3Se,b. Ihe agreement 1n most part is consistent withthe Keyes volumetric uncertainty esti: rate of 4%. ‘To investigate the thermadynamie sonsstenry af these data further we used the thermodynemie surface derived in section Il, 42. (The surface fits the very high quality dat shown in figure 4 quite closely.) We ealeulated values of den- sity or each of the experimental temperature and. pressure ata pointe of dataset 7 for the overlap region. The most in- teresting feature of the differences is a systematic trend of| the Keyes data a highor vale af density, Par mat ane the differences are less than 1% with about half of the dats points lying within a 0.5% error bar. Very large diferences, some in excess of 5%, occur inthe criticl region, Because of Ines large errors and bectuse the errors tend to be systemat- ic, these data should only be used outside the region where there are high quality date, and then only with very low sta tistical weit. is to temperatures below the sange of the exper 45. Dota Set 8 sontal measurementsTHERMODYNAMIC PROPERTIES OF AMMONIA bby Kazarnowsky [64]. The results are tabulated for five iso- therms et 25 °C intervals forthe temperature range 200 °C < 1 300°C, Values of pressure are given for each isotherm for each of 15 values of specific volume. The pressure range 100 < P< 1400 atm. Thus, these data overlap the very high ty measurements of Garnjost, see section Il, 42. The tabulated values appear to have been smoothed in both tem- perature and pressure. The text indicates that measurements ‘were also made at 150°C, but no reaton was offered as to why these date were not listed. In subsequent publication Kazar- nnowsky and Karapetyans [6] refer to this paper asthe souree for low temperature data (below 200°C) and include tables of thermodynamic properties and figures based on those “data to 160 °C. However, the figures in reterence 0, par- ticularly the inset in figure I for the 140 °C isotherm, ae in- consistent with results in the accompanying tables. We limit oe Banthace date actually Hetod in reference 84, that is, to data forthe temperature range 200 "CK 1< 300°C. ‘The method of measurement is based on a technique em: ployed by Michel [9] in which pressures inside and outside the piezometer are constrained to he equal, so that pressure deformation of the container is reduced. The results are fected hy sorption and hy imp sample, with such effects most serious atthe lower values of temperature and for values of density in the neighborhood of the critical density. The pressure measurements were made with @ piston gauge and were estimated by Kazarnowsky {0 be sccurste 10 within 0.35%. He also suggests an overall uncertainty of +0.5% forthe tabulated values. It is apparent thatthe overall quality ofthese data ix much Inwer than that oftdata ets 1,2, and 8 ‘The thermodynamic consistency for these data was invert gated using the same procedure described for data sets 6 and 7. We used the thermodynamie subsurtace derived in section I, 42 to derive values of density, corresponding to experi- mental pressure and temperature data points ofthis dataset, for the region of oxerlap with the differences ar less than 0.25%, with no important systematic trends, except that atthe low pressures and along the lowest isotherm the differences tend to be large, with the maximum aterence approaching 2%. These data should be uselul at pressures above the range of the Garnjost data (et 3); how: ever, below 750 atmospheres thei inclusion with the Garn- Jct data wld mainly anntete date ea tae. in meet eaten the the eurface 46, Data Set 9 Tn thie worl Date [84] reparted meaeuremente fr 100. T data points very lose to the gasiquid critical point. Ex: perimental pressuredensity pairs were obtsined at nearly uniform intervals for seven closely spaced isotherms: 131,05°, 131.58", 18208, 18225", 18258, 133.00", 138.96". The density range is 0.65 < elo, < 1.35. Ako included were date in the two phase region for the sub-rtial isotherms. Date 32.30 °C, P1115; atm. and (@13.8; moles per liter. (In section 2.3 the values adopted for this correlation are 1=1322, °C, P=11156 alm and (018.80 moles per liter) 657 ‘The experimental apparatus was the same as that used by the author for his measurements of the equation of tate of the liquid, (ection 4.4). The experimental sample wae slr similar to that used inthe earlier experiment. The sample was stated tobe 99.9% pure with no more than 0.04% volatile im- purities. Date suggested the overall accuracy of the mea ‘urements to be £0.27. In eection 4.4 measurements with this apparatus for io- therms at 25 °C to 125 °C were ascertained to he uncertain by ahout 0-25 49 +0.4%, Measurements in the eritical region are generally more dificult than are thote for the compressed liguid. Also, there is increased sensitivity of such measure- rents to the presence of impurities. That impurities are present in this experiment is clearly indicated by the dats. ‘There is a consistent postive slope to the isotherms in the two phase region with (22) =1 atm/g - em. We have cal: ‘culated the effect on the critical temperature of 0.04% of nitrogen (we thus assumed the volatile material present to hhave heen mostly at). The calculation was based on a theo- 1 anlysie raported hy Keacem [96] und indicates thet the sample were contaminated by 0.04% nitogen, the value observed for the critical temperature would be 0.15 °C t00 low. In addition to the suspected impurities, visual inspection of the graphical representation of the experimental data in- dicates an inconsistency between the slopes of neighboring isotherms immediately above the critical temperature for densities near eiteal Final ‘hat the temperature measurements were claimed accurate only to 0.05 °C, and close to the critical point small errors in the value of tempers ture ean cause large error ‘We atiempted to correlate these data using the methods of critical scaling. The mejor part of this work was carried out for us at the University of Maryland by .V. Sengers and T.A. Murphy as part of coreelation af arveral zuhatanoas near the critical point. In this work they utilized correlation tech- niques recently developed by them in collaboration with MAH. Levelt Sengers [87] of thie laboratory. The results ob- luined show the experimental data tobe compatible wth eit ical scaling theory to better than 0.2%, provided the eitcal temperature of ammonia i taken tobe close to 129.80°C and rrowided only the data for the single phece region ore ine cluded. The data are clearly not compatible with scaling theory forthe value (=13225 °C, the value we have recom- ‘mended (section 2.3) for this correlation. That the experimen- {al data are compatible with ertial scaling theory strongly suggests that the systematic errors are largely due to the presence of impurities (or to other errors associated with the Integrity ofthe comple) Since the tompers is 0.45 K the presence of 0.04% of nitrogen would only ae- count for a third of this critical temperature uncer Even if we could account for the entire temperature differ: fence, its not clear how we would correct this body of exper- imental data for the effect of errors resulting ftom the presence of impurities. We assign an overall uncertainty of st least several percent forthe volumetri: 47. Date Set 10 Kumagai and Toriumi [83] reported measurements for the 4.Phy. Chom. Rel Det Vol. 7, No.3, 1978temperature range ~20 °C 4000 alm at interval of 100 at. ‘The experimental method is similar, in principle, to that used by Beattie and Lawrence (ection 4.1). The volume of the pievometer was varied by injecting how amounts of Ly. Chem. Rel. Det, Vol. 7,3, 1978 L. HAAR AND J. $. GALLAGHER mercury. The pressure measurements were made with pis ton gauge. The author refers to an earlier publication [91] for detailed description of the apparatus, but that reference, in fact, does not contain sufficient detail. Included in reference ‘91 is a comparison of experimental results using this apps tus with measurements reported by Benedict [92] for nitro ‘gen. The comparison was made at 50°C for preseures in the range 3000 atm < P< 6000 atm. The average difference was about 19% with « maximum deviation of about 3.8%. (The Benedict values tend tobe larger.) “Tells correlated his results for ammonia with those of Keyes, section Il, 43, by means ofthe Tait equation, which, relate the speifie volumes at any pressure along an isotherm to thote of a reference apecific volume and preavre. The Keyes values eta pressure of 1000 atm were chosen as the reference states for each of three jotherms. A maximum Aiference of 0.5% was obtained between values calculated with the Tait equation end the smoothed experimental e- sults, We infer from the foregoing that the results for amimo- nia tabulated by Tsiklis atthe lower pressures appear to br ‘consistent with Keyes data to within about 0.5%. Since the Keyes date are of questionable quality, particularly above 100°C, an uncertainty estimate for the Tsiklis measurements ‘ofa least afew percent would appear reasonable 49. Data Set 12 Recently, Takis, Semenona, and Tsimunerman [95] re- ported volumetric measurements at 500 bar intervals for three isotherms 100 °C, 150°, and 200 °C in the pressure ange 2000 < P < 9500 bars. The authors state that the ‘measurements were made wsing a diplacement method, and that this method yields grester accuracy than achieved in their older work, dataset 11 of this correlation. Homever ex perimental details are not included in this paper, and severe, references to earlier papers yielded only sketchy details ‘The authors compared the volumetric data with date set 11. For the to isotherms 100° and 150° in the region of ove lap, they report mean differences of 1.2% and 0.755%, respec tively. ‘An overall accuracy of 0.5% is claimed by the authors for ata set 12, We feel tat eatinte fs snewlaa aubitvus, Sy tematic deviations of these data from those of Harlow and Franck (94], data set 13, are suggestive that the absolute Uncertainty is probably in excess of 1.5% 4.10, Data Set 13 ‘This dataset was obtained from a graphical representation of measurements made by Harlow and Franck [94]. The details ofthis experiment have yet tobe published: though a Figure containing plots ofthe data was nde avallble 1 ua (ASTD, we have not been able to obtain + copy ofthe seus] data or ofthe experimental details needed for eritial evalu tion of this work. Recently, Profesor Franck sesred one of ur that ¢ publication is fortheoming and that the delay in publication had nothing todo withthe dat. The graph eon: tains esl for 10 isotherms for the range from 25 °C 1 450 8C.Tievesllswhuve 60 °C ave presente at fntervl f 30THERMODYNAMIC PROPERTIES OF AMMONIA 659 °C. The pressure range ie 100 bar < P< 10,000 bar at the lowest temperature and 1800 bar

200 °C; #5 har for the pressure measurements. The errors in- ‘woduced in interpreting the drawing ere probably leee than +£0.25% overall. This estimate is based on two separate inter- pretation, with «two year lapse between, and several spot or partial interpretations. A different investigator was wsed for tach interpretation. ‘A modification of the apparatus used for the ammonia work has been used by Professor Franck to make measure- tents for water [95], ialudiag the same preseute range, but extending to somewhat higher velues of temperature. Those data are believed to be accurate to within 1% over their range, and this may be a reasonable estimate for the accuracy oftheir experimental measurements for ammonia. 4.11, Data Set 14 Volumetric messurements have been made by Lichtblau, Bretton, and Dodge [96] at 50 °C intervals along isotherms including 300 °C < T'< 450 °C. The pressure range includes 500 atm < P< 2500 stm. Only smoothed values of the com- pressbility factor for each isotherm are reported. The ex: perimental procedure included a constant volume apparatus in which the peesure was transmitted tow deadweight gauge via & transducer. It is apparent from the article that very careful design and thermodynamic analyses had been used in this work The authors claim an absolute accuracy of about 0.9%. If this claim were valid, then these measurements would extend the region of quality data for ammonia to an important part, of the thermodynamic surface. Unfortunately results of our analyses are nol in accord with the authors claim of sccurscy. We have found the dats to be inconsistent withthe very high ‘quality data of set 3 as well as with the data of sets 8 and 13, ‘The differences are systematic, withthe data set 14 values of density consistently low. The deviations are in excess of several percent. Deta sets 3 and 8 overlap date set 14 along, 21 900 °C. Date mt 19 orelaps dete oot 14 for several isotherms at the higher values of pressure. To ascribe the inconsistency to data eet 3, 8, and 13 isnot feasible, pa ticulerly aftr the analysis for dataset 3 in section II, 4.2. We therefore must conclude that the data of ect 18 are in error by at least several percent over their entire range. Itis unfortunate thatthe actual P.g,T data points taken in Ais eaperimeut mete uot seposted in the publication, Ie ie conceivable that the unsmoothed values could yield some ra tionale for the inconsistency, and perhaps a bess for corre The authors also report values for the compressibility fa tor for the binary mixture contsining 81% smmonia end 19% nitrogen forthe P,T region of dataset 14 5. Other Data of Mostly Historical interest ‘hie section has included a discussion of the thermody: namie data of importance for this corelation. The order of the discussion indicates our estimate of the relative impor- tance ofthe regions ofthe surface, Thus the discussion ofthe data began in section 2 with the equstion of state for the ‘eoesisting phases of vapor and liquid. Next considered in sec- tion 5 were calorimetric measurements, such as specific hests tnd latent heats and the properties for the ideal gas state. ‘The Pye? data for the single phate rogion forthe gas liquid phases was discussed last in section 4. An important pat of the discussion was devoted to the establishment of ac- ‘curacy estimates forthe data. In most cases the estimates ob tained are conservative that is, the estimates we propose for the uncertainties in the data tend to be larger than those given by the original authors ‘The following list includes date that are mainly of only Iristorical interest. Also included are some Pia measure- tents that apply to very limited regions adequetely covered by other data. The order of presentation of the properties that given in the previous sections, 5.1. Coexistence Data Vapor Pressure: Bunsen [21], Faraday 52], Regnul (24, Bincke [97], Bril 98), Davies 95}, Holst (100), Bure and Roberton (01, al of which are discus in reference 4: Bergstrom (103) reported mearuremente in the range fom the triple point to 30°C, which have an aerege deviation from the NBS date of 0.5%; Henning end Sock [13] made ‘meararements over the sume range and thoe heve an eter tage deviation from the NBS values of 03% Specific volime of satureted ligui: Faraday (104), D’Andreet (105) oly (16), Lange and Hert [33], Usher {h07}, Lange [16], Diteri f0} nd Keyes and Brownlee ist) Specific volume of satus {002} and Berthoud [24 itical point: Dewar [28], Vincent and Chappins [110], Gentnergiver [111] Jaequerod[112}, Scheffer [113], Ca dora and Giay (14), Esreicher and Sehnert (118), Bes ‘thoud [37], Mathias [116), Postma (117]. Triple point: Ruff and Hecht [119} Bil (8, Flo (120), Bergstrom [102,Eucken and Kerwt (7) Deg psn Dunns [2] bared rion (121), Regul {26} Jownni 122), Ladenburg [123 Lange {108 Dickerson {124}, de Foreand (125), Gib (31), Bri [6], Davies [9], Burre] and Robertson [101], Keyes and Brownlee [51]. sd vapor: Dieterii and Drews 5.2. Colorimetric Date cat capacity: Gy of vapors Regneult [96, Wiedman (296) [Nernst (31), Haber end Tamara [38}, Giacomini (127), C, of vapor, Masson {128}, Cozin (129), ‘Wallner [130]. Muller [131], Keate [132] Schoer [133], Schweiker [134], Perting- ton and Cant [135], Dixon and Greenwood [13], Pertington and Schilling [187], Budde [138], Voller {135}, Nerast (32) . of saturated liquid, Drewes [140], Dieteriei [30], Beau tnd Ennis [141}, Ludcking and Starr [142], Von Strombeck [143], Keyes and Brownlee (67), Keyes and Babeock (144), Babeock (145), Eucken and Karwat [78]: C, of saturated vapor, Babcock [145]. Latent hea of vaporization: Regoaul [28], Von Strombeck em Rel Det, Vol 7, No.8, 1978660 143], Bstreicher and Schnert [146], Franklin and Keauss hia Joule-Thomson coeficent: Wobes {148} 7 Date Experimental P, qT. data for the single phase regions hich are now mostly of historical interest include the work cof Roth [149] and of Keyes [50]. Other data for which the range is very limited and adequately covered by more recent nembuteatents iuclude weesureneuts by Hele [190], Drow lee, Babcock, and Keyes (151, 152], Bridgman [153], Diet richson, Bircher, and O° Brien (154), Ditrichson, Orleman, and Rubin [155], Moles and Batuecas (156l, and Moles and Sencho [157] and Mayan [158]. References 154-157 are mesturements of the normal density for ammonia vapor at standard conditions. Reference 153 includes data along one Iwoderm at 90 “C for the pressure range 1000 atm Sr % 12,000 atm, Reference 158 includes only data for thermal ex: pansion and compressibility for the liquid. The text indicates these were derived from P.o.T’ measurements taken in the ange from ~20 °F << 120 °F and at prestures to 100, rmospheres. Based on the limited description given, the quali- ty of the messurements ae adjudged to be somewhat below shove of references 92, 94, T8 whieh Include te range of reference 158. I, Thermodynamic Surface Introduction In this section we develop an analytic thermodynamic sur face for ammonia and derive the relations for the thermo- ‘dynamic properties to be calculated from it. We aleo include detailed comparisons between thermodynamic properties calculated from the surface and the corresponding exper: imental dats, The reference state forall properties is taken to be the ideal gas at zero kelvin, ‘This section comtans six subsections. The derivation of the surface and of the expressions used to calculate the vari- ous thermodynamic properties are given in section 2. Section onesies eal the surface and experimental date. A discussion of physical features including the validity ofthe thermodynamic surface close tothe critical point ae given in section 4. Section 5 on- alos eslmates (OF ime overall accuracy of the ealcalated properties. A short summary of the results and some general ‘comments ate contained in setion 6. tated from 2, The Derivation of the Thermodynamic Surface Tn thi ewction we isin the lect squares sense, sclected P, @, Tdata schematically represented in figure 2 to an snalytic equation of state. No mathematical constrains were imposed 3m, end only P eT data were used in the least (except thatthe equation is consistent with the eal gas in the limit of zero density). The equation is a 44 term double power series function oftempereture and de L. HAAR AND J. §. GALLAGHER mental accuracy nearly all the available P, o, T data as well as 10 produce limited extrapolations (bated on thermody- namic arguments) of the surface into important regions where data are sparse. The range ofthe equation is bounded a Jow values of temperature by he triple point temperature (195.48 K) and the meting curve for the liquid, and et high temperatures bythe iotherm at 750 K (hich is epproximate- ly 5/3 the eritical temperature). The pressure renge extends 1 5000 bar. ‘This section includes four parts, The general relations are piven i section 2.1. The derivation uf dhe eting curve fs given in section 2.2. Certain features of the fitting process concerning the selection of statistical weights are given in section 23, Conditions and relations associated with phase equilibria are discussed in section 2.4 2.1. General Relations Following Keenan et al. {80} the Helmholtz free energy function is represented as the sum of two term: th First is the contribution from the equation of state: the second is @ function of temperature only and refers to the properties of the ideal gas. Thus we write forthe Helmholtz free energy, Me Die TD, o where °(7) is the contribution of the ideal gas. We define, Ala, PFRT{n e+. 7, @ andsince Pre'adive, qs (and (sel, ProRT{I+e0+e'2 0/30), ° seth Qe=OB:, (Ba) ‘where By isthe second virial coefficient ‘The form we choose for Qi given in section Ieq (14), OTT ctw, @ where r= 500/T, T is the value of temperature in kelvins, 111,2883498, gis the density in g/em and Ris the gas cons tant, Equation 3 for the pressure has been fitted in the least squares sense to the experimental P, 0, Tdata. The results of values for the constants wy lated fa table 6, By ions of eg (1) we obtain fr the entropy, So, T-—Rln e+ eQ-er 90/9 x14S4(N, o the internal energy, Ble, PHReP9 0 4E%N), o‘THERMODYNAMIC PROPERTIES OF AMMONIA “That 6 Cosine for theradyamis rare gud ad get te S00 er Tea ia rissneros0s 0020815713 Rone rs6618 6345520616 ‘06004367698 a seenmneas a c5202690428 379050220018 —sozonansss732 Laseonaouras s.549s5000532 anneeso 682 ranesssppa72905 “Sosresis2tes 74350041622 ~oseae4ss200 mre rTTa135086 Arana en8si6768 17796860140 38 7460958505, “Tasooasrissa0 masssTise02 2130461497509 2999901550901 09.005 7579758 sono 0881791548 9065788510971 241168510882 93 35479900199, ‘rvanarasosee27| 16 1037459 alsnansrn3es 2458 710606858190 rr aaieso93685 -mrmassrsszs6i12 eseonasr94sa0 the constant volume heat capacity, Clo. T-RoPa°0/0P+C° the emhalpy function, Mg MATER Qe! e19 QI TEM, the constant pressure heat capacity, C0, THAR. where, om 4+e0+e8Q/9e- 18 W/97-e'x9°9/2720, and Bo1+2eQ+4e' 0/0 ute'9"Q/OK", (6) oo thet capacity forthe strated Hi, nc, % (2PM. dee ® SBPlBeh | dT" where Ps the vapor pressure ofthe liquid given by Eq (of tection I and where Pis given by eqs (3,4). [SE and Cy are the corresponding contributions obtained from 4° APC -ES-RTU“In 481807) @ so fare, de BATT fe 4°) (00) e TEAM), where G°—E3 is given by eq 10 of section I. Equations (1-10) are all that are required for calculating ll thermodynamic properties forthe fluid. It is convenient for calculating prop> fries of coenisting phases to use the explicit vapor pressure versus tempersture equation, eq (I)of section Il As wil bei: Justrated in section II, 3 the difference between calculations sing the vapor preseure equation (lo abtain the anavieting phates) and calculations using only eq (1) (with the usual eon- ditions for phase equilibrium) is smaller than the uncertainty assigned tothe experimental data 2.2. The Melting Curve ‘The low temperature Louudary of the erect melting curve. The relationship between the pressure and temperature of the melting solid was ealeulated by means of the Clapyron equation, using th latent heat of fusion and the specific volumes for the saturated lgu for the latent heat the value reported by Over Giauque discussed in section I, 3.3ds forthe specifi volume for the solid at the normal melting point, we used the value reported by MeKelvey and Taylor, discussed in section TI, 2b; forthe specific volume of the liquid we used the value revorted by Cragoe and Harper. discusred in section I. 22a. ‘The Clapeyron equation isthe relat 48 oT where the quantities ae as defined following eq (in section M, except that wand w now refer tothe specific volumes of the liauid and solid, respectively, and L refers to the latent heat of fusion, From the above data, the quantity =" =pxio and eq (I) integrates to yield a, ay 0=Bexp[sx10-4P-P)), cr) ‘where 8, and P, are triple point values. (Bq (12) ignores the small differences between the triple point values for the spe: cifie volumes end latent heat and the corresponding values at the normel melting point) Since P, atthe triple point is very small the selatioushiy es be simplified wo Phys Chem. Rel. Date, Vol.7, Na. 3, 1978662 L HAAR AND J. $. GALLAGHER (6195.48 exp (10*Platm). a3) Equation (18) is the relation we use to represent the coer: isting phases of solid and liquid, which isthe low temperature ‘boundary for the corelation. 2.3. Stotistical Weights ‘The overing erierin ued forthe selection of weights was the condition tha the derived thermodynamie wrfce fare withthe daa within he accuracy eatimater signed in section Tl For this purpose, the root mean squares of the fractional deviation forthe vale of presure and deni sore compared withthe szpenve experimental uncertain tes for these quate, Weighs wee tsgned 2 ato yield rma, vats which ate abou equal to the curacy asig Tn sections I, 41 and T, 42, i was established tht the “core® data shown in igre 4a of considerable higher ac cacy than any ofthe other Pe T dala reported for thot dscssed in section Hl, 42. Al other data have been assigned relatively lw ststial weight, Dat that ovrap these "core valet have been given sero weigh Tom lcontensy ves ned in sson fi 42, betwen the data of st 3 andthe data of st 2 near their region of overap (The inonistency exits only forthe ochre cor- Terponding to the lowe! slun of densi for dat et 3 an seems to be largest at the higher values of temperature) In consistencies of this nature could seriously compromise the sccuracy of the derivatives of the thermodynamic surfac even beyond the region of the inconsistency. To reduce the effect of the inconsistency we adjusted the statistical weights 0.8 to smooth the transition between the two data rets, Thus we eet to a6r0 the eatintcal weight for the sponding to the lowest value of density for dataset 3 Simi larly, we se the statistical weight to zero forthe isochore of highest density of date set 2. The experimental values of den sity for the next lowest density isochore of set 3 were adjusted to split the difference between them and the results for the trial surface derived in section Il, 4.2. Finally, the statistical (for there adjucted data” woro reduced by from those assigned to the main body of date set 3, In this way a comparatively smooth surface has been achieved, but atthe expense of some degradation in the fit to data set 2 at the higher values of temperature in the region ofthe overlap Excluded from the least squares fit were data that ex tended closer tothe crtial point than the data of set 3, Data thin ahout 20% ofthe critical density neat the eriticl isotherm and to within 3% ofthe eritiel temper: ature near the critical isochore, The data closer tothe critical point were omitted, principally to remove a logial inconse: teney associated with analytic thermodynamic surfaces (These data also have certain questionable features; see rec: tion TI, 46.) As was explained in section I, 2.1a, attempts to reprecent the region close tothe critical point with an analyte je equation can seriously compromise the representation not only in the critical region but even in regions considerably removed from the critial point. Thus all of dataset 9 was ex cluded as were data associated with the values of the thermo 1.Phy. Chom. Re. Vol.7.No. 3, 1978 dynamic parameters atthe critical point, Features ofthe ther smodynamie surface near the critical point are di section III, ‘The evteria for the last squares fit included tests for over- fit, As we noted in section I, 2c, an analytic sueface that yields the pressure within the accurecy of measurement would yield an equation that seriously overfite the dencity values in the compressed liquid region. On the other hand an analytic equation that yields values of density within the ae curacy of measurement would seriously overft values of pres sure in the eritical region. We rejected equations of state that produced such overfit. 2A. Coenisting Phases The liquid~vapor saturation boundary is defined by the Gibbs conditions for equilibrium between the coexisting phases rer, a3) Preps, ay) C=, as) where the superscripts fg refer tothe liquid and gas phase, respectively. isthe Gibbs funetion defined by caatPie, 16) Analytic representations for A(T) and Plg,7) are given by eas (la) ‘The Gibbs condition for phase equilibrium was included by a procedure that makes explicit ue in the least aquares fit of values for the vapor pressure. In this procedure the Gibbs conditions ate satsted only to within the accuracy assigned Jn seetion II, 2.1 0 the vepor pressure dal, employed ita follows: Step A tral surface war derived by least the P.,T data, without the Gibbs conditions Step 2. The vapor pressure equation given by eq (I) of ec tion Is used with the tial surface of step Ito ealelate trial values for the densities of the coesisting phases. Thus we solve Plo TPT) ay for values of density for the coexisting phases, where PLT) is the vapor pressure equation and P(g,7)is the tral surface of sep 1 Step 3. Using eqs (15,16) the Gibbs condition is written Aloe, AUC TT e as) e where ¢ and of are values for the liquid and vapor coex- isting Jensties, enpectively, as calculated in sep 2. Thus P is formulated in terms of the 4 parameters of the equation of state (end the preliminary estimates for gt and gf) Step 4. We now consider an expanded data set which in- cludes the data used in step 1 plus values for the vaporTHERMODYNAMIC PROPERTIES OF AMMONIA 663 pressure, which for convenience have been included at 10 K using eq (1) of section IL. A new thermodynamic surface is then derived by least squares fit to the expanded dataset. In the least squares fit we now include the dtterence between the vapor pressure ‘“date” points and values for P riven by eq (18) ‘Stop &. A-cacend sot of trial values of density for the coos isting phates for the new surface is determined, using the ‘equation of state derived in step 4 and the vapor pressure ‘equation. Corresponding to these, the Gibbs energy condi- tion eq (15) is then tested. In what follow, we shall use 4610 represent the small deviations from exact agreement. Step 6. The dG values obtained instep 5, are now used to stermine the very amall corrections to the vapor pressure "data" required to achieve exact agreement for eg (15). This done by Taylor series expansion. Retsining only the leed terms in such an expansion we obtain the thermodynamic re lation, 1 (E-3) eee, 09) where Pin the mall coreton tothe vapor presare “ta” Teed tatty the Gibs condone 5 eae. Step 1, The final values for coring dense ate then deterbined eeoring to sep 2aning eosin I crementd bythe AP ales bined in ep 8 andthe egos tion of tae derived in step 4 the dimensiones quantity AC/RT aha step 5 are listed in table 7. The maximum is less than 0.0006, which value corresponds to less than 0.05% uncertainty in pressure, In practice differences are negligible between ‘alues forthe thermodynamic properties obtained by the pro cedure consistent with using steps I through 7 and the much simpler procedure in which the procedure is terminated after sep §. However, for theriendynamie consistency, the valine for the coexisting phases in Appendix A were determined us- ing steps 1 through 7 “Tate, Gibbs function esos Te) won mas Scv00rs rg 000133 mais “ones ms ‘0019 2818 ovo2te 21s acosrs aes case mais 00s3) aes Sovosi2 201s 002s as 0000T7 3315 ones EBs Soon 315 Faoni0s GIS o.omnss ‘5015 0088 aus ‘coors 015 ‘oeo14s 815 00061 3. Comparisons of Derived Properties with Date In this section the thermodynamic properties calculated from the derived surface are compared with experimental deta. Comparisons with P.@,T data including data for phase ‘syle ‘wompainon calorimetri data in section 8.2 and with values reported for the second virial coefficient in section 3.3. ‘ne inured in weet: an. T data, Including Phase Equilibrio ‘The comparisons are presented es fractional differences in percent between the ealeated and experimental values of density. The results are illustrated forthe fit vo Pa,T data in figures Sa, 5b, and Se. It is important to appreciate thatthe ‘uncertainties associated with various P,g.T data sets differ from each other by up to several orders of magnitude, Fur ‘thermore, measurement errors are largely systematic, 0 that the low quality data are, in general, thermodynamicallyin- ‘onsiotent with the quality date nd with the de face). Thus the closeness ofthe fit for the vstious Po,T dat sets isnot the important eriterion. Rather, what is important is that each dats set be represented within the accuracy sesigned that set. In most part an examination of the Figures show that ths criterion is satisfied Deviation plots for the “core” data are shown in figure Se. ‘Thee very high quality data extend over the region ofthe surface (ge figure 4) that isthe most important technologi cally. In section I, 23 the small inconsistency ofthe data of set 3 in the region of overlap with the data of set 2 was resolved by a small compromine i the fit for data set 2 and a some- what greater compromise forthe fit to dataset 3. The devia tion plots forthe suspect data pointe of et 3 are indicated in Figure Sa (ond figure 4) by the citeles. The amount of the in- ‘consistency approaches 0.2%, withthe largest degradations ‘occurring at the higher values of temperature. tis noteworthy thatthe one data point of data set 3 with the largest fractional deviation (ymbol Zon the Figure) isnot an outlier. This data point is very close tothe eriteal point parts in 10,000. Infact, s can be seen in figure 4 the data of 1 Sinclude a numberof data points tha extend close tothe critical point. The thermodynamic surface was able to fit all ofthese data within the accuracy of measurement, tht is to vrithin several parts in 10,000, as was the cate for the large rajorty of the datapoints forthe data of sets Ito 4. "The somperion shown in igure Sh and Be tand to mush larger deviations than those forthe core data. Thus, forthe data set 6, the deviation plot indicates these data are in consistent with the core dats of sets 3 and 4 by about 0.4% ‘The very large differences shown for dataset 7aceur near the al point, where these data are inconsistent with those of. set 3. The largest differences forthe data of set 8 oceur for the lowest icotherm at 473 K for the loweet densitie. In this region the data are clearly inconsistent with the data of set Sand withthe data of set 2. Except for that isotherm, the fgreement with the data of set 3 in the region of overlap in ‘most parti within about 14% {Lys Chem Ral. Dat,Vl.7,Ne. 3, 1978664 L HAAR AND 3. Data set 1 4a 7 ae A aa Presa, bat Data set 2 +2 0 = 4 Data set 4 (Saturated liquid) +3 0 -al on 03 10 Pressure, Data set § (Saturated vapor) ‘hermodyoumicsuroce and ie eperineml dtl igure are pote "core data For eons givens veces 1,2 apd 8 the trae thet he hermdynami rae consent withthe POT Data set 9 includes measurements made very close to the critical point. These date were nol included in the least, squares fit forthe reasons given in sections 1, 2.la and HI], 23, It so happens, st noted in section Il, 4.6, that these di ‘lso exhibit large thermodynamic inconsistencies. However, the vory large deviation shown in figure Sb for thece dat primarily due to the inadequacy of the derived thermody humic surface as a representation of data very close to the critical point. (A discussion of the fit fr the critical region, ‘J.Phye. Chom. Rel Dat, Vol, 7,No, 9, 1978 8. GALLAGHER 0} — ent panning giana ten ws wey ae FF Pressure, bor Pe a np ee teeter 30 sapere gre Sincere very bgh uly dat refered om te ere ot ted in he derivation, Encephale gre, tin tbe acorary signed them inet including the critical region boundary, is given in sections Mi, 4.1 and 1,42) ‘The largest deviations for dataset 13 (e fig. Se) occur for ‘the isotherm st 298 Kin the region of overlap withthe data of sets Tand 10, The deviations forthe data of sets 11 and 12in- dicate that thove data are inconsistent with each other and with the data of set 18. Inconsistencies are even more ap- paren for the data of set 14. These date are inconsistent wi the data of sete 3,8, 13.THERMODYNAMIC PROPERTIES OF AMMONIA 665 Data set 6 490 = 90 Pressure, bar = 3 4d Data set 8 2a Soe cter, © fee 0 ser 7000 rr Press -20) +5) . Data set 9 eee ee a at : a° ee ee 120 ° + Pressure, bar 3.2. Comparisons with Calorimetrie Dato ‘The calorimetsie data included inthe comparisons are the ‘measurements for the coexisting phases and for the super heated vapor, discusted in sections Il, 3.3 and II, 3.2, respec- tively. The comparisons for the ovexning phases are shows in figure 6. The comparisons for the superheated vapor sre {given in figures 7 and 8. The calorimetric values calculated fom the thermodynamic hhave been combined with the contributions forthe ideal gas state, eqs (10) The values for the ideal-gas state have been shown to be of very high se curacy. section I. 4. ‘As we have stated in section J, 2.1d no ealorimetric data used in the derivation of the thermodynamic surface, so these comparisons are a severe test forthe derived 5 face Figure 6 includes percentage deviation plots for calori- metric properties for the coexisting phases, Included in the fieure ate deviation nlts forthe Istent heat of vaporization the heat capacity forthe saturated liquid, the heat capacity {Lys Chem. Re. Dat, Vl. 7,No 2, 1978666. L HAAR AND J. 5. GALLAGHER saat set 10 of mae age i . : : 2 foe 0 Data set 1 +20] ° i Press, br : cee is : 7 eee uly . 2 . 30000 : + -20 = < Data set 12 ad : 3 cae Prasae bar ee q eee ee . pee TS Data set 13 “20 : pres tees 7 elec Te, > “20 Data set 14 20 Presse, bar ° 1090 apo 2.0) = zs . 3 oi Se, Secapton for gue Se for the saturated vapor, and the heat capacity of the liquid at ‘constant presure equel tothe saturation pressure. ‘The experimental values forthe latent hea of vaporization re those discussed in section Il, 3.38, and have been as signed an overall aouracy of about 01%. The ealelat values were obtained from the equation LT HQs.T- He!.2) 2) 4LPhys. Chem. Ra. Vol.7, No.3, 1978 ‘The deviation for the latent heat are shown in the fist curve ‘on figure 6. The maximum deviation is less than 0.2%, which ‘occurs at the lowest value of temperature. The average of the bsolute deviations ie lees than 0.1%. ‘The experimental valuee forthe hest capacity ofthe satu rated liquid are discussed in section II, 3.3b. These date are believed to he accurate to within about 0.2%, The ealeulated values were obtained using eq (9 (with eq (1) of section Il forTHERMODYNAMIC PROPERTIES OF AMMONIA 667 +0 1 {LATENT HEAT OF VAPORIZATION Bhan AL a DeVATIONS (96) Sore TEMPERATURE (*¢) PURE 6 Clrimeti omparons fe the coesiting pases. The percent ‘elton are ped eran temperature forthe lect het, the heat ‘pec for the strated gba apc fhe nataated operand ‘hehe! eapaciy at eontnt presto the started giao seen IL, 32 The mort important fer ie hat the eelted rales agree ith enpernent amen to ihn the arercy of te experimental mee “erent eacepl avalos of temperature sow about ~20 "for which the pay of POT date has compromised the accuracy ofthe derived sre, aepep te CURE 1. Calorie compares for sopeheted vapor. The perent ‘deviation vera pesre Between th aur auld fom the derived Uhermodynanie sufoce andthe experimental date se section MI, 33, ‘Te fete of ot tha the average dria lean 035, ih ‘he meine than 1% occuring 0°C er satration, in ego lor ich the Tate are ose q Se é = —— ce 7 oe a __—_— we $ a ——— wwe | « ~— ve Pune 8. Deviation plots betneen ves of G, alee ‘meena nef un ar Ted be a seston fret illerences ae plated vera press others 30°C i Teale forts of temperature nthe ange O° 1300. la ely ‘cate the agrerment within the uncertainty aie or the ihe Salen af empertre athe higher vrs of pes: Noe tha he Tower valaesof temperate he pure iced eins of he sac ey ear tothe strated ape sales In thee reo the contain {fom he thermodynani rac ger than ha ofthe eal a jon between the calculated and the experimental values ie about 1.0%, and this occurs at the lowest value of temperature. The average of the absolute deviations is less than 04%. For values of temperature greater than ~20 °C the maximum deviation does not exceed 04%. ‘There arene di the saturated vapor. The values diseuseed in section Il, 33a ‘were calculated from the equation raciso-2. crecio. an where Ci and L are measured quantities. Because equation 21) involvee the differentiation of the latent heat and in volves a difference between large numbers, we assigned the comparatively large uncertainty of 1.5% to these data. The values calculated from the thermodynamic surface were de- termined from eq (9). The figure shows a monotone decrease 1.Phps Cham. Ral. Date Vol. 7.No. 8, 1978668. in the fractional differences for C¥ ass function of temperature. The largest deviation, about 2.0%, occure at the lowest value of temperature. For values of temperature great ferthan 0°C the deviations do not exceed 1.0% Experimental values for G, forthe saturated liquid are dis- cussed in section II, 8.8b. The ealeulated values are derived fiom the thetmodysenie sunface using eg Os). The constant pressure heat capacities are very nearly equal tothe corre sponding C, values, Thue the deviation plot forthe C follows the trend forthe C, values, except thatthe magnitude of the eviations ae slightly smaller. ‘The most important conchision from figure 6 is thet, ex cept at low values of temperature where thee is « paucity of ccurate experimental Pig T data for the single pase regions the calorimetric values calculated from the surface are in accord with the experimental data for the coexisting phases almost to within the accuracy of measurement Figure Tis comparison between calculated and measured values for the heat capacity ofthe dilute vapor. The exper rental valves are diteuseed in section TI, 82. 1 was noted ‘Heri tha Ue angen utee uf eto i these easurenicnts is from the random seater of the data, approximately 0.2%. ‘The caleulated values have been obtained using eq (8). The ordinate inthe figure is the percent fractional difference and the abcissa the pressure in har. The straight line segments join the individual points. Bach line includes points near (o within degree) the value of temperature designated inthe Tiguc. In cues where several dt together, the ndvidcal deviations were averaged. The max imum deviation approaches 1% and this our cle 10 the saluted vapor sate a OC. The average of he magnitude At the deviations isles than 039% Figure is ery similar o figure 3, In tha igure the same calorimetric data are compared with values calelated forthe nd see section I, 1. The net signin comparison vith figure 3 is the ietene i the maximum devintion, However, except fr one orto pint the map tudes of the deviations are about the same, Pariclrl for telus oftemperture above 35°C the compan indvates ‘ery small degradation ofthe accuracy of he derived surface 4. ” Figure Bisa comparison of the C values given in table Sof section Tl and those calulted from the thermodynamics fee. The fractional differences are plated a function pressure in bars forthe values of temperature OF < «300° “The presue ange extends, atthe lw value of temperature, Aerner the coturaed vapor sate and atthe higher alee temperature. to 125 bare. The principal featre of the com parison is that over most of the range the deviations are less ‘than several tenths percent. The notable exceptions are atthe highest values of pressure forthe isotherms at 250° and 300 C, where the large deviations, approaching 2% at 300 °C and 125 bar, are direct consequence of the procedure we sets 2 and 3; see sections Il, 42; IM, 23; IM, 3.1, This com: promise is preferable to the situstion that would have ob: ‘ined if the small inconsistency had nor been smocthed. For then the degradation in the C, values would not only have 1. Phys: Chem, Rel-Dete, Vo. 7, Ho 2, 1978 LL. HAAR AND J. S. GALLAGHER been larger, but would have extended over much large re- ions ofthe surface, 3.8. The Second Viriol CooHficiont Figure 9 includes comparisons ofthe second viral coef late from oa) Galld curve) with value repo ‘ed by Hirschfelder et a. [159] (circled point) and Rowlinson [160] and Keyes (161) oxed points. For values of tempera than 423 K the agreement is ner within £1 atthe lower values of temperature the results of Rowlinson tend to be lower. At 273 K Rowlinson’s value is shout 40 em?/mol lower than the vale ealulaed from eq (Gx) and about 30 em?/mal lower than the value obtained by Hieschfeder et ‘The ress reported by Hirschfelder etal were derived from the data of Myers and Jessup, data set 1, using a taphical method. This involved an extrapolation to 2er0 resware of the volume resduslsesleulated from the dat for ‘ach isotherm, We have established in chapter Il section 41, thatthe date of Myero and Jesup are accurate almost to within their seater, that to about #0038, I we apply thi uncertainty tothe volume residuals used in the graphical analysis, we obtain an uncertainty of about Sem” ma forthe ‘xtrapolation for the isotherm at ¢=100 °C, The uncertainty for the extrapolations at lower values of temperature is a leat thi large Is apparent fom the igure therefore, that the reeulte of HireeMalder otal rein accord with values Calelated from eq () within the accuracy of the graphs! analysis by Hirschfelder et al ‘The valucs listed by Rowlinson were calculated fom an equation derived by Keyes [161] several years prior to the work of Hirschfelder et. Though details were not given it appears that the derivation i based on a least squares teat creme of volume residuals [168 for PT data of sts | and 2. Shorlly thereafter, Sockmayer {163 16] used those B, values im his approximation forthe Stockmayer potential for ammonia, Then, subsequent t the Hirschfelder work, Rowlinson (160), Based in part om unre: ported experiments at his laboratory, selected the results of Keyes for his further development ofthe Stockmayerpoten- ° he Blocknayer potent rmonia obtained by Rowlineon is recommended in the refer cence work by Hirschfelder eta [165 Even though there appears to have been a general consen- sus in its favor, there is serious defeet in the Keyes treat iment. We note that volume residuals forthe lower isotherms Ihave a great deal of curvature a a function of pressure, par ticularly inthe region near the eseniatemce bows aloo ‘90 degrees curvature for temperatures below 80°C), Thus the values for B, calculated by Keyes are sensitive to his (arbi- trary) choice for the pressure dependence fo the volume re siduals. (Keyes ignores the fourth virial snd al virils higher ‘than the fifth, except that he includes the fourteenth in hie representation of the volume residuals as a power series in premure:) I eppear 1 us that large par of th diffcsmee between the values for B; from Keyes and those ftom the araphical analysis are due to the particular representation used by Keyes. ‘As we have demonstrated, the results for By calculated‘THERMODYNAMIC PROPERTIES OF AMMONIA 669 -100- 7 700 [EUR 9, The eon vil uefiet vers tempertare see section My SA The sll ne repent sles elated frm the thermody orf The ced pointe ae tle repre by Miele et and tHebosed pint we valerie by Keyer and Relies from our thermodynamic surface are in excellent agreement with the very accurate heat capacity measurements for the dilute vapor. This ie strongly suggestive that our calclated second virals are quite accurate. We note that with inereat- ing values of temperature above the critical, the plots ofthe fet, and the results obtained by Keyes and Hirschfelder are in good accord with those values calculated from the thermo- dynamic surface. 4A, Thermodynamic Symmetry ond the Critical Region Certain symmetry features ofthe thermodynamic surface inthe eritcal region, including the critica iochore and the al isotherm are discussed herein. The section contains arts. Section 4.1 meiudes rome comments related 10 the extent of the critical region. In section 4.2 we discuss cer tin features relating to thermodynamic symmetry in the Vicinity af the evtial paint Camparicnne arr given with results of critical sealing, Section 4.3 includes a disevsion of| certain calorimetric symmetries. 4.1. The Critical Region ‘That the analytic surface was not coerced to be consistent with experimental data inthe immediate vicinity ofthe erti- cal point is an important feature of ths correlstion. As we hhave already stated, sections I, 21 namie inconsistency must and IIT, 2.3, «thermody, I when an analytic surface is coerced to fit such data. To verify this conclusion, we im> posed on the analytic surface our best estimate for the parameters of the critical point, the values given by eqs (1) and (9) of section II. We also included in the least squares fit {he thermoaynamie featares assorted with he dvergence ‘of the compressibility atthe critical point In this way we in veatigated the de funetion of how strongly it was coerced (ia weights) to be consistent with our best estimates for the values of the eritcal parameters for ammonia. As expected, ‘egradation did occur. The errore seemed to increase with in- ‘creasing satstical weights, that is with improvement inthe realization by the surface of the critical point. The quality of| the surface in those region of dataset 3 that lie closest tothe cites! paint cocmed tn euffor the grestect degradation However, the degradation extended to regions far removed from the critical point. The errors were found to be largest for the heat capacity valves. Even the very small amount of coercion needed to produce only a 30% "improvement im the Pa fit atthe crtiesl point caused serious degradation in values for C, for the dilute vapor at low values of tempereture. The devintions from experiment for this case were found to be about 100% larger than those shown in figure 8 ‘We now illustrate the extent of the critical region. We de fine the eriieal region a that region else to the critical, point for which the derived surface would yield errors primar- ily because the thermodynamic surface is analytic. The so Tine in Figs If rapreuante the liquicevepar {or the derived thermodynamic surface. The open point just below the dome of the curve is our best estimate for the loca sion ofthe critical point, eqs (1) and () of seetion I. It ies about 1 K below the entical point ofthe (uncoerced) derived surface. The symmetrical dashed curve above the dome en: closes what we believe to be the critical region for our sur- Anncity the dashed nye Hes about 10 bars above the critical pressure and about 4K above the eritical value of temperature. It intersects the dome at values of density approximately 20% larger and 20% smaller than the erieal density, corresponding toa value ot temper- ature about 1.5 K below the value forthe ertieal point of the derived surface. The solid points late those data of data set, in the general neighborhood of the ertieal region. Several of the date points le very close to the boundary one isin side). All ofthe data shown in the figure are fit by the surface to within their accuracy, that is to within several parts in 10.000 [eee section IIL, 31). A turther discussion ot the qual ity ofthe it i given in section IIL 42, immediately following, 42, Tharmadynamis Symmatry andthe Critical Baglon Tn this section we compare thermodynamic values caleu- lated from the derived thermodynamic surface with predic: tone of simple scaling [166]. The dached curve in igure 10 encloses region somewhat smaller than that usualy asoci- ated withthe renge of validity of simple linear sealing. Sim ple scaling has been used to represent thermodynamie prop- tries along the erticalisochore out toa value of temperature Phy. Chem Re, Dato, Vl.7.Ne. 3, 1978670 [eI 10, The dome fhe oenting phase satin IL, 1. The soit "inn retain he sateton sates elelted (om the theme {vace Toe ai aoe ate poof ota 8 et ee ‘ny ofthe eral pn. The open ce jt below the Top ofthe (dome ict the "perimental yale forthe ere ot, sets 25. The dae carve js aoe the to fhe dome eacoer wat Feed nthe tent abe real eon, To fee of ote ta he ‘hermodynani uface nace ih al the data pints Gli point) te wshin the ecurcy af mesnurement nat he qulty of the tap fears be compel neni tothe covets the eit pit "Tn together nu cei etre of gues an 12 oppo hs he thermodynamic ufc avatar representation or te Fepian eter ‘doth ere region bout 3% higher then the value of the critical temperature, tnd on the eitcal izotherm to a value of density #25% from the critical value, Thus thee isa region ofoverlap where i is believed simple sealing and the derived thermodynamic su. face are both valid, except that the former uses the correct ‘values ofthe ertial parameters while the later does not. We shall eompare the results of the two methods on the critical jsochore and onthe critical isotherm. ‘The thermodynamic quantity we compare isthe isothermal compressibility, The isothermal compressibility is defined by -3@) ‘The inthermalsompoosiy em seve pre for the thermodyoamie earface, Simple sealing ile, for esa plerefeence 167) k, oxen, (2) eo onthe ie hohe nd okra Se es ‘on the critical isotherm, Values for the “universal” exponents y and 6 have been reported by Sengers etal. [167], based on a correlation in- volving # number of substances. The constants A, and As; for smmonia lio ware obtained from parameters for ammanis 2.Phy. Chom. Re. De L. HAAR AND J. $, GALLAGHER reported in reference 107. As we indicated In secton Th, 6, the data available for ammonia in the immediate vicinity of the ctitiesl point are questionable, Therefore, we are more concerned with comparitons between the “universe!” con- stants reported by Sengers etal. end those predicted by the surface than in comparicons between the respective magni tudes of Ky. To accomplish this, we compared values forthe natural logarithm of @* Kr. Figures 11 and 12 compare, on a logarithmic scale, results calculated from our analytical surface with values caleula from eqs (23) and (28), In those igure, the slid lines repre sent values for @K; caleulated from the thermodynamic sur face at very close intervals. There ino detectable waviness in the curves connecting these points, and we conclude, there fore, that the thermodynamic surlace yields smooth erva tives, The results of sealing ate indicated by the dashed straight lines. In figure 11 itis rather remarkable (likely coin ial that the roclte for cealing ar with those from the esleulated thermodynamic surface from the critical region out to value ofthe reduced temperature ‘ear 0.1, The value forthe “universal” constant y obtained from the thermodynamic surface is y=1.16. The value trom reference 167 is y=1.19. This agreement is well within the estimate for accuracy ofthe scaling result. Figure IB llusratca significant differences ison om the eriticaljotherm. First ofall, the thermodynamic surface yields two different curves, corresponding to values of density greater than or less than ertical. Now, except in regions close to the eitical there are ample PT deta of very high quality, as ustrated in figure 4 and the derived thers modynamic surface is in very good accord with these quality PresT data. I is apporent, thorefore, that simple ceaing om the eriical isotherm results in serious oversimplifcation at large values of the reduced density. It is interesting to note that, as we approach the critical point, the values calevlated from the analytic surface tend to small differences forthe two branches. The two branches cross the resull for eitca seal ing ata reduced density near 0.17. ‘We cannot obs from the analytic surface an unambigu ‘ous result for the value of the constant 6. We have made several estimates, fist we considered the region including the values of reduced density from 0.19 to 0.30. The "best" Straight lin that spits the diference between the results for the two branches, ave 434.5. If we change the region slight- ly so as to include reduced densities from 0.20 to 030, we ob- tin obo. If'we conlder just tne region given by 0.19 3 0.21 we obtsin 24.7. The result from simple scaling is 3, Me af simple erating an the sritical isotherm tend to fail at values of density that differ from the critical by more than 20%. The results for the ther rmodynamie surface are in closest accord with those of sealing near the critical region boundary illustrated in figure 1U. On the critical isochore the agreement with scaling is within the accuracy ofthe sealing result, Ins a very interesting feature of figures 11 and 12 shat ‘over most of the region included in the figures the values calculated from the thermodynamic surface lie on curves in Aiainguishahle from steaight lines. Deviations occur only"WERMODYNAMIC PROPERTIES OF AMMONIA sr (bar) 10 ‘ane 1. The eompeniiliy om the etl nchores ae setin I, 62 "The recipes ofthe ithemal sempre ie pote vere the reduced temperature T==T/T~ The soil line represent vauee Calewned fe te thermodynamic surface, The dashed ugh line x {hereof fnew ealing. Te eid ine wan eppronate ih ing repreveiton oie vale cleused rm he herman re, ‘The gant the ale ofthe sles ofthe ight nes The featare cf notes the eel! seco betwen the peision fine sealing ‘eo the rel point. The ale ef rece temperate gil 0) {eat the boundary ofthe crea rein show in figure 0 se eritcal region and at very low values of density. (For the ilute gas a straight line representation cannot apply.) 4.3. Colorimetric Features In this section we describe certain qualitative features of ne ealorimerric values eeleulatea ftom the thermodynamic urface. Figures 13 and 13b contsin isobats of the heat ty at constant pressure as a function of temperature. “igure 18s inchudesisobars forthe pressure range 10< P< igure 13b features a closer spacing of isobar for vressure interval bracketing the eritcsl preseure, 50 < P< 50 bar. The vertical dashed lines atthe lower values of tem verature connect the two values of Cp corresponding 10 iquid-vapor coexistence. There is a cross-over near 50 bar velow which C, values forthe saturated liquid are larger and bbove which they are larger for the saturated vapor. Another feature ofthe figures isthe trend of the maxima toward high + values of temperature with increasing values of pressure These trends are in acord with results for other Huids; see ference 80 for water. “The reciprocal of umes the ‘he maghtude ofthe reduced deny, leer The values eee ‘rom the Sheedy sulce frm to separate ances, th soll ‘hare. The we branches rd to oer acer with Screg value of he redced density. The dated le ithe rer bined rom lear scaling ith the au ofthe slop. The etre nt tt esis of ie cannot be extended very ft rom he rip on the 5, Evaluation of the Thermodynamic Surface In this section we obtain estimates for the accuracy for values of calorimetric properties calculated from the thermo: dynamic surface. The accuracy depends on the smoothness of the surface, the quality ofthe fit to the Pj’ data, and the thermodynamic consistency ofthe Po,T data. The uncertain ties in ealeulated Gy values are appreciably larger than those 1.Phys: Chem. Rak Dot, Vol. 7, No.9, 1978m ‘omperature, K um pears bet cope are plted versus vales of tompreas fr avowsvaluen of restr, Ti gare nes mon of the ange he Served thermolysis srr. The vereal dashed nes its the ‘Senin t gidsepor conser For aero preemie ee than aout Sri wale forthe snared tu ages Bat sovethn vale f rene, Calero hatred apr ae ihe. for the eartespondine PT values: compare Resers SR Therefore the very small uncertainty inthe ideal-gas values in thie analysis can be ignored ‘As we explained in section I, 2.14, an important feature of four approaen is that calorimeinic data have not Bee” Yn ‘eluded in the derivation of the thermodynamic surface. Its this feature that enables us in section 5 to assign clos tl cerances forthe accuracy of the ealenlaed ealorimettie prop 1.Phys-Chem Rel: Date, Vo. 7,o.3, 1978 L. HAAR AND J. S. GALLAGHER vith those of erties. Comparisons of results of this corre other correlations are given in section 5.2. 5.1. Accuracy Estimetes for Cand H ‘We first consider the accuracy forthe region covered by Ue ve1y high quality “core” date shown schematically in figure 4, In sections Il, 42 and III, 3.1, 3.2 those data were shown tobe of uniform quality and in aceord with each other and with the thermodynamic surface, In other words, these data (and the region ofthe thermodynamic surface scheme ically lustrated in igure 4) are of uniform thermodynamic consisteney and accuracy (excep! in small region of overlap fof data sete 2 and 9 and except for values of density forthe saturated vapor). Tie feature greatly simplifies the analysis. ‘We note that different parts ofthe surface are more sen sitive to measurement error of a given fractional size than are ‘others. Thus calorimetric values ealeulated from the derived thermodynamic surface will vary in accuracy, even though the quality of the PT deta end the fit to them of the derived surface are uniform In particular, fr egions of the thermodynamic surface where the Pgisotherms tend to very large curvature or are very lat or very seep compared to a reference region, we would expect some degradation from the sceuracy achieved in the reference region. We take as refer tence those regions for which the very accurate calorimetric data exist. Thus for the gas phase the reference region isthe dilate vapor for the liquid, isthe satorated guid stats. ‘A useful thermodynamic function with which to quantize this intuitive analysis is the isothermal compressibili defined by eq (22) We see in figures 6 and 7 that, except where the P,o.T data for the single phase regions are very parse or nonexistent, the values of Cy caleulated from the thermodynamic surface agree withthe experimental data (0 thin 01190: Though she accursey af the caleulted values (our reference values) do not appear to be sensitive to the isothermal compressibility, there is @ trend in figure 710 in creased degradation with increased compressibility It is seen in the figure that along an isober the factional error tends to decrease monotonically with increasing values of tempers- ture, and so does the value of the isothermal compressibility. in accuracy from this comparison is appreciably less that linear. We assume, therefore, that linsr degradation with isothermal compressibility forthe calculated C, values yields seonservative estimate forthe gas phase. Based on this ana: (2/2) arane. 0 where (8Cy/C,P is the fractional uncertainty in Cy that we Wish o estimate andthe quantity (6C,/CyPny is the reference value of uncertainty, given by (86,6 Pry =0.008, 26) ig the feothermal eamprescihility unltated a the pint ofTHERMODYNAMIC PROPERTIES OF AMMONIA ora 1nd K," is the analogous reference value. For con- venience, ths reference is taken as the velue for th ideal gas, KseUP. @ We use eqs (25-27) 10 calculate unce of density in the range @ 1000baris taken as 25% ‘The uncertainty for values of enthalpy should be no more than 36 as large at fr the corresponding G, value, 5.2. Comparisons with Other Correlati Pethaps the most widely used correlation for ammonis ie [NBS Circular 142 (reference 38) It has been shown in section I, 3that there is excellent agreement between the results of tue presen corse aud she messurements upon which Circular 142 wat bated, Only relatively small extrepolstions beyond the range of those measurements are included in Cir- cular 142.0 that in most part agreement withthe present cor 4.Phys. Chem Rel, Dota, Vol 7, Mo 9.1978L. HAAR AND J. $. GALLAGHER Tame & Famed wceiits in abused Caen 100 666) ows | om | or | om | os | om ] wo | as | 1 | 2 ] a5 asm | suey | ase fame | aren | asian Jazzy arm | 6519 | aso Tesi] estes |r ests perialsee arab: || taivoee |tcver) 1 |aranaas create: | erase (axe a o os 100 o | os | os os |) a6 180 ou | oe | os | or | ww | 20 zo | or | os | os 0 os | oe | om | ao | os os_| on "Rest veo temperate en 101 4778157 here, alae fri em ‘Then enrich he aller oh oman between data set 2and Sin hereon of vel relation is within the accuracy of the basie measurements on ‘which Circular 142 was based wonily Devise (referonee 5) that extends in pressure 1o over 1100 bars and in temperature to about 312°C, and includes results for the compressed lig i. The Pq data used to generate this thermodynamic sur face includes the data sets 1.245,7, and 8 Davies made Ele, Aud Newt Belorn 88 F-17396) 7 ALT Bat of een tte Gravee by Gate poimomairhogoal on = uo pits Mh eth Teo! 3180565). (0TB) JJ. Manin an JB. Edvares, Creation of Latent Het af Vpor tin tee Chem yea 00) (0, Rt an si for preset 30 bas Beet Koos, as (090, H.W Copper an I-€Golranch BWR Benet. Web ‘tsi ard ar coratien, hasan ror 161567 Uu8n) P. W hogan Boperoes amon fom the fompater, Ta site 0h 1, 897%, [usa LN. Gana EK Plc. DE O'Brien and FS. Muosing Ther ‘nedrami ropes of amnni, Hyon Proce et [085 V.ARabaich wd G1 Tolman, The equi of tae and her ‘Redynamic papers af hud amon, nan Zh ‘Aad Nook Beene $3.4, 13) (96D) [184 6.1 Toman, Thrmedyeanic propre of lgié amnesia, ev. ‘yan cee caesen ete Fh fer [085] BM. Burmie,Aoption of Marni uation of ‘nia Chem Landon a0 1108971" [u84 Hansen, F jority P Rasmussen, B, Serene, eof ution of at: ore waar Dan Kemi G17 09701, ue. Gobn ENA Pe Kecherpen Volts of tmosia, aten ose, sad water Inform. Sates Ga Noctis, Pol ts aot, Prom. Prod Org. SNe. 1 (P song DC, Kasco, Zp: Obs Pe. 444198. (099) KF Kock Das Entulpe Entropy Dingeom sar exergetcken ‘Beatng vor Verrenangroginen, Brena WaermeKrle 18,6596 1.Phys: Chom, Rel. Date, Vo. 7.No, 3, 1978THERMODYNAMIC PROPERTIES OF AMMONIA 7 Appendices Physical Data and Reference States ‘The units of temperature are the degree Celi sive to the helvin). The IPTS 1968 temperature sale [16 ‘The unit of energy isthe Joule, the definition for which is ‘onsistent with the recently recommended values ofthe phys- cal constant[19} ‘The unit of density isthe gram per cubic centimeter, where the molar mass of ammonia is taken to be 17.08026 gram 19. ‘The zero of energy is taken as that of the ideal gas at OK Tn order to refer the tabulated values to the liquid at the tripe point, the following values should be subtracted from the respective entries: Entropy of liquid at triple point = 4.20248 / K. Enthalpy of liquid at triple point = ~1110.256/g. Internal Energy of liquid at triple point= ~1110.268./g. {Phys Cham. Rel, Data Vol 7. No.3, 1978630 LL HAAR AND J. §. GALLAGHER Appende A._Table of Therma ropetie Pressre Temp. Specie volime ‘Free Interal enay Enthalpy Tat Enepy Tar SC Ligid Vapor encigy Liquid Vagor Ligh Vapor best” Ligad Vapor cle cle “GRY ‘ie ie die le tle Sie Se a ee me ee eT eT $358 eee “esse Lament Gres 2649 —losssa Son7s | Chasse Sao Ser seo Teist Yarer Stag “850 Cisse ors) Mlusnss | Sas So Soe Tbs Toss fisrse | “9516 “teats Surge “haats) sas es ste “shoe has akse “505 “lua ses “iowse anal Sas) ae “ese Vital Heassh “bss ovet6 Sint ooen Sco Loses 47406 L950 Seesbhonsueneas Same Seo? “sei enna Sakeg © omeae Red lalzasio 47m 149% cas Boon? “oka ome.te Soett “owa0 Set? tara tes 1089 arse “atts “ere fanaa OTT) 4519 a0 aos 1.0686 chee Ise “ates ookas Saas —oohse 783 Laoazom e099 110890 “Sa se “ates “osrs0 Sanso —os7st aaas 13078643 4.0227 1.0083 8 Most “ohao “eokis Sera “sokoa iam lameamia Saa00 10.0001 ose Tsssl 2039 “ontis S90. “oatan 45259 IsTowose S089 0.8618 “3.60 fists “buss “olese Sam “plas 4552 Igrossre 51063108209 Lo “281 e536 242s “emns Sib asm? aslo IabDasTS «SRM 107610 eo ewig “ohis “eonsa Sinak “antl eset lamin S000 07071 Tass fast “akor | “gnrso Sse39 “enka tener | lownsad S22 1 Tavs Goss 201 aso S50 “esos 7325 Issam Samm Leo Thats BAST “29s em Sis “smol Mee imssam Same Ins a amas “a3 “ancer Seat “baat ase lame Sum loans oe Bice Tyee Faas teat ais iomseet Sout losers te amo CR Gam “ie @ro lasers Seo teas 1s an Tasos Sons? usnawe sna ozs em mo SEALS Saseocae Soe a 528 bios? Those eas Mantsae S900 0.190 na Toss These Sage iainsom Same t0.leer eo ela? Sooo S129 lowes Soni sass rs 150s “6B S82 Usha Som IaoasT 3 15099 Teme Sieg lensoe? Gams ior Be ee “ios Sikoe ma2tr Gaste Tease nas 1810 “Gie Sioat Mase Gorm inom aso iat TeS5s Smz Tiesese Glows 100182 a8 rst Testes $2125 Uisaso00 Gian "90% mas inn “ean ae S212 Lusadart Gls Sand 2) a0 eo 282 ilscooar GIT 99600 Ba 103.58 003s S639 lisse ame San) ae as “wer S409 lzrssio Ga 39185 38 coer} “630 5542 lial Ts Samm S50u aoa x0 =wo1l 5390 lui6oorr 62m S500 an air “smo Soe onmaer Gazt Baws ae a» coma Gro loam Gams See son ane “ste.ss 52733 los.tomb 68H San Ss wae TEsssi ELS ltenSioe Gitar o.atee as na S556 laasTT Toate ani2 “sates sao Istaasre Sumas Ines? “oi6% Gate “Sioa fgeo2 IsTISSN Some InsI9 Toons) Sins bore 4gt91 Asesesst | Sse 0.709 “onside Sess | “Gans) 459.39 gee TOO Sigel 10.7196 “ema 34 “emis 460.86 1256 109 S15 10.365 “Bie oot “ian ease mera Sim? Tae Test Senor “eases Kes20 sonora | S'2000 o.ti22 Tena noe “BHLOd ieee oerelm S21 Ines? Se6% SSLIS “sTess 46805 Aged seHs 5.2611 10.5095 Temae 3221 “gis opus sas) $2791 In6e9 cerat S82 “bors Moa) kins S50 Insts “esse 12) “aos finde Istana Sala) lols Testes 6882 “esas? 47334 Isazlise S.ssan 106158 “es4s4 S584 “Askoo STaD laza sets $505 Lem Spiess ds “hiss TTS aa. Sda00 loses Teioas 3935 “oan se armas 1519.3 S006 10.ssKe Tacs aosk “esse aanIe Jsl6Ias) Sand LOSE Term 36182 “esl46 asta isizooe Sasa? loses “ean Xa “ear eo? toss? Saree loess) Tee2s S6e19 “eive2 aas21 isososea Sas10 loaan6 2h soe Ae Test S683 “ak2T aeas v7) Ssomh Tae ooo rss 188 Toop'ts S05 won T0 aero sees S057 loasa0 Som Tess Sees “okt “eokad Seose “eons ane? I2oa0089 | S3u9 10.4599 Riss swore? “2872 “0005 oreo Toss somos mms 55001 10420 Son Tea Sal 270 “isea9 Son T9 T9490 491.25 Izme2IS6 S775 10.8100 EEERREREREREYRER EGR Sasa 210 1827 S101 ou sists S07 2363 sien a8 2a ano st Dow isms ties i 1L.Phy. Chem Rel Det, Vo. 7, No.3, 1978,‘THERMODYNAMIC PROPERTIES OF AMMONIA forthe Coie Phen of Lin nd Vapor ses. G 5 G Lothernel PAID, Density Temp isk ink ek “Tek ek Liga Vapor em’ lem r= bar ils a7aa6 Basse ooo Awe ST OMT ——To.00 Mao? Gowor Sst09 ooomso eset agit “aro “0 Tips 4599785178 omns) et ouase amse 3.00 Vast so “S.aer2 “ommss ican “toss “izse So Tin ewe “paned “mes Ieaae “aon “raeas cH Lime Gisss “Seas Stone ‘S52 See HE BR Tso $4o99 “83559 “ooomst some 20 9.0 Teel Gauge “E3199 looms sums ase 00 tame (aon “sar oomer 7818 nso fae “eon TSist Sess Stet omer Gk onoto “rtose sige S3es “Sheat amoss iol “ange “ast Tan 43610 “Santo “ooo © Ss) “oie “rheat Tsar dss “Saise “ooo et “olaat “aise Ys ¢3a © “A9t00 boom 4 “olsao “tine Ys 43a “howe looms ae “aise “tuse Tse $34 “AG0m omnes Sash “olen “ruoar ie 388i SRS Ra BEE late aim iam Get “keee “oom S02 ese “rama Sites Tse? 4302 “heme one 359 ims“) Site sso 3456 “A 7 “oooor? 3.0 nse “ae Sesriem 4301 “A7IS) ‘doom Sant ante ats Sera sean 43608 “A.r73 “oooom 2.518 amas “home Sigs 1s668 (Hat “Ast “oooost 2376 so “Toe Shes Tse 43591 “AIss “ooooel 2283 Oasad “eet Sitges sme Ciege “A'SsSs “ooooms 2.002 ones “oie ceo pe Sises seq ore “esi “aopooe 14 Os0a8 “eer Seo im Stow seo ¢3iss “apr “noposs 192 ees a0 ‘ae Sis 1sou Sone “Ata “ooooes 497 ast paoe a0 aes Sti Soot 438 Vor ‘Sonat “east er) tone Stes Teo 388 1593 ‘geod “ein air tae Sis reo 43050 Kis gna “eons Sos cow Stan Tense C300 TSn pe) “es aor act 2iem Sian Lem ares Tage “onsn “erren am Sig aos S158 ease ates 12% Biss “eae? Sm ie 3b Sasa esis alan Tits “ows “eons “36.00 sen 2218 Stam ams Star 15 “oso “0 “8 See dion Sum Less Caan ‘ios ‘baa ea set aie 230 Sit esse ae Sot kaos oro a) feos 23s ne Leen hae? Bie “ussis. “eras Sie sim 236m Same Leen Sas ts “osrss “erm “ooioas an 27re2 068 Lanka a8 “estes “ere2e“ooion 44a 23003 Somer err ees ‘st “Oo? “ers “tovlen ‘at 33im@ Soar Leon ase fas mets “sin? “toa See 2322 30m Los tao oo “owowe “srima“opi203 S350 238m 300m Lae ae fe “uraes “eaoos “Wolds tie 2383 Soa i ie fs tran? “eeoop “toes Sao 2385 Sort am Law sar “ores “eetaa “tna ess 236m 30m] rar Came ‘ssi “twos? “tosis “ovis Sos 230 3.0m La ame S51 “ogape “eos? “corse Sse fom See Lis18 sone St “tme29 “eease “Gonna asim 24s asia 7365 sion 60 “teper “esiox “ois fin Yam tous Len is ‘ro “tases “egons “onto ie ie one Limo Sgam ‘si “twee “esous “ose? fe thee ost om ‘is “toset “s6509 “ont ‘sue Beas Sioa Goer 0 ao “host “eoee? “one ‘sor 2st Seals Lome 455 ma “hase “sssa5 “ooo Sse 251m) Sime Law ae Seo “neee “egioe “ose Soa Bsn 30d Lae 3S Ss “now es Som sue Sout asia seco ee “izaze “oteuz “ose isu 23s Sono Lame 3a ‘amr ‘steno ome 1 Plye, Chen Ba. Bate Vol 7. Nov 8, TE