The Calculus ofThennodynarnics 5 In this chapter we develop functional property relationships for simple systems. These relationships are completely general and are a direct consequence of applying the postulates and laws of thermodynamics discussed in Chapters 2-4. They also reflect behavior consistent with the calculus of multivariable functions. These functional property relationships are fundamentally different than physical property models, such asPyrN equations of state, which represent practical constitutive equations specific for each pure substance or mixture composition. In order to obtain working physical property equations, experiments are performed and data correlated or suitable estimation methods are utilized to predict property values (see Chapters 8 and 13). Alternatively, molecular models using statistical thermodynamics may be used, albeit with some approximations (see Chapter 10). A major objective of this chapter is to introduce techniques for recasting derived properties and their derivatives into forms that can be directly evaluated in tenns of primitive properties which are measurable. Our starting point is to introduce a fundamentaJ operating equation for simple systems that establishes the functional dependence of internal energy U on the variables entropy§.. volume V, and moles (or masses) of all components present N1,. .. , N,. This Fundamental Equation has its origin from the combined first and second laws for open simple systems that was derived in Section 4.7. The S, V, N1, ... , N11 coordinates of that equation were used extensively by J.W. Gibbs in his classic work on thermodynamics in the late 1800's, and as a result they are commonly referred to as Gibbs coordinates. They represent a set of n + 2 natural or canonical variables that provide the necessary and sufficient information required to describe the thermodynamic state of a simple system. By employing methods from calculus and geometry 1 Gibbs was able to show, in a completely rigorous mathematical mannert how various thermodynamic properties were related. He clearly laid the groundwork for understanding the functional interrelationships that exist among the so-called potential functions. U, HlA, and G. Without the computational power and graphical display methods that we have today, Gibbs had to conceptualize the structure of these property inter-relationships in his mind without being able to illustrate them graphically. He used only narratives with equations in his writings to describe his findings. In part, this restriction has limited how fast we have been able to understand the details of Gibbs• work. It has taken over a century of study to fully appreciate the significance of his contributions. Interested readers should consult Section 5.1 The fundamental Equation in Gibbs Coordinates 125 the papers by K. Jolls and his co-workers listed at the end of the chapter for a comprehensive discussion of Gibbs' contributions in this area. In the sections that follow, we will employ graphical methods to illustrate some of the geometric aspects of the Fundamental Equation and its transformations. In addition, several methods of manipu1ating functions and partial derivatives will be described and applied to problems frequently encountered in classical thermodynamics. 5.1 The Fundamental Equation in Gibbs Coordinates As will be seen presently, the relationship U = /u(§_, V, Np· .. , N,J (5-1) completely describes all of the stable equilibrium states of a simple system containing n components. By solving Bq. (5-1) explicitly for~. an alternative form is obtained: §.=f"-U,V,N1, • ••• N,) (5-2) Either relationship is called the Fundamental Equation: Eq. (5-1) is termed the energy representation and Bq. (5-2) the entropy represen1ation. The Fundamental Equation is represented by a (hyper) surface in (n + 3} dimensional space. The points on this surface represent stableequilibriumstates of the simple system. Quasi-static processes can be represented by a curve on this surface. Processes that are not quasi-static are not identified with points on this surface. (Recall that derived properties such as U and§. are not defined for non-equilibrium states.) Then+ 2 ftrst-order partial derivatives or the Fundamental Equation correspond to traces in the respective coordinate walls of plan.es tangent to the thermodynamic surface 1• The significance of these tangent plane traces can be seen by expressing the Fundamental Equation in differential fonn. For the energy representation, Eq. (5-1), du;: as (dful - _')I (dful d~+ av - -~ ~ (dfu] aN. dV+ ~ 1=1 I dN; (5-3) ~~J[J] where subscript fvj[i] indicates that all N1 except j =i are held constant in the differentiation. If we compare Eq. (5·3) to the combined Jaw for a simple system (see Section 4.7). namely n dU=Td~ -PdV+ L Jl.;dN; (5-4) i=l it is clear that l The lenn plane is used in the generic sense here. Even for a pure fluid there arc three independent variables and three partial derivatives. ThuK the tangent plane is re.a.Uy a three-dimensional hyperplane. The Caladus of Thenno~amics Chapter 5 126 (5-6) and (Jful ( ()N, =lli =g~, V, Np ... , Nn) (5-7) _.X:. N1 [tl where the functions 8T• gp. and gi could be obtained directly from the Fundamental Equation if it were available. Equations (5-5) through (5-7) represent a particular set of equations ofstate expressed in Gibbs~ V N cooroinates. As shown later in Sections 5.2 and 5.3, only two of these three equations of state for a pure substance are independent. The second-order partial derivatives of the Fundamental Equation are also related to quantities that can be measured experimentally. For example, for a pure materialt there are four second-order partial derivatives at constant mass or moles: ~- = ;sfL(~~) ] =(~~lN -Y..NV' l ~ ~;=aav[(~~J~..N ]~,II =-(~~t ~ - :~y=:~[(~~L (5-8) (5-9) - L=-(~L a~~~= :v[(~~L L=(~~1 (5-10) (5-11) Of these four derivativ~ only three are independent because the last two are related by the reciprocity theorem of Maxwell (see Section 5.3, part 4). For an n-component mixture there are n + 1 ftrst-order and (n + 2Xn + 1)/2 secondorder independent partial derivatives of the Fundamental Equation given by Eq. (5-1) orEq. (5-2). These derivatives are particularly important because they form a basis for all other partial derivatives involving thermodynamic properties. That is, any partial derivative can be expressed in tenns of an independent set of first-order and second·order derivatives of any form of the Fundamental Equation. Proof of this statement fo1· a pure material is given in Section 5.3. At this point. it is interesting to consider what the shape of the internal energy sutface would look like in~ V N coordinates. Since even for a pure material this would entail a Section 5.1 The Fundamental Equation in Gibbs Coordinates 127 four-dimensional plot, one is forced to reduce the dimensionality of the system in order to visualize the surface in three dimensions. For a pure system, one can propose a 3-D plot of intensive U as a function of Sand V. In fact, James Clerk Maxwell in the late 1800's was so intrigued with Gibbs' work that he actually made a 3-D model of the USV surface for water. He sent a plaster replica of his model to Gibbs at Yale, where it remains on display today. Figure 5 .la is a photograph of a more recent 3-D USV surface for water constructed by Clark and Katz (1939) while Figure5.1bis a2-D projection of the surface showing several important features. Note, for example, the point of triple tangency that corresponds to the three-phase, solid-liquid-gas, triple point equilibrium for water. We will have much more to say about such geometrical aspects and their relationship to the characterization and limits of stable equilibrium states in Section 5.7. With more accurate thermodynamic data available and computational tools for correlating and manipulating these data, it is possible to generate USV surfaces as computer-generated, 3-D plots for a wide range of pure materials. Jolls (1990), Coy (1993) and coworkers (1991, 1992) at Iowa State havedonejusl this for a number of pure compounds using a well-tested PVTN equation of state developed by Peng and Robinson (see Section 8.4) and a temperature-dependent correlation for the heat capacity in an ideal gas state. The actual methods for generating values of U for specified va1ues of Sand V are straightforward and described in detail in Sections 8.1 and 8.2 [see also Coy (1993)]. Figure 5.2 shows two fundamental USV surfaces~ne for an ideal gas and the other generated by J olls and co-workers for ethylene using the Peng-Robinson equation of state as a modeJ. It is important to keep in mind that the values of S, V and U are not absolute, they depend on the specification and U 0 . of reference state conditions, E"ample 5.1 actually derives the governing equation for U for a monatomic ideal gas that is plotted in Figure 5.2a. The USV surface is convex with respect to variations inS as the second derivative is always positive for the exponential dependence of U on S shown in Figure 5. 2a. Later we will invoke even stronger arguments that this convexity must be true in general. Although the actual functional dependence of U on V will vary from substance to substance, the general power 1aw decrease depicted in Figure 5.2a will be followed for a11 fluids in the dilute gas region where nearly ideal behavior is expected. The important thing to remember is that the graphical representation of the USV surface encodes all the thermodynamic infonnation about a pure substance as it is equivalent to the Fundamental Equation itself. The geometry of the surface represents all single phase stable equilibrium states, aU two-phase coexistence points, and all triple points; it also can be used to identify limits of stability such as critical and spinodal points (see Chapter 7). A major difficulty, of course, is that a significant amount of physical property information is required to construct the Gibbs USV surface for a pure substance, e.g., the steam tables for water, or a robust PVTN equation of state that models the behavior of a compound adequately in all states of aggregation. sn The Calculus of Thermodynamics Chapter 5 128 u VAP'OR Figure S.la Three-dimensional model of the USV surface for pure water [from Clark and Katz (1939), with permission of the Royal Society of Canada]. u{; Ts.L < Ttp - Metastable region c::=J Unstable region p ""I Steble region Figure S.lb A two-dimensional projection of the USV surface for pure water shown in Figure 5. t a. Section 5.1 The Fundamental Equation in Gibbs Coordinates 129 t VIC!' \~ T=llopet, ~(()U1 as~,. -P=slo~ =(au~ iY J~ Figure S.2a The USV surface from the Fuodamental Equation for a pure monatomic ideal gas referenced to lJ>, ~.and V' (see Example 5.1). Figure S.2b A predicted USV surface for the fluid phases of pure ethylene. Volumetric PVT properties from the Peng-Robinson equation of state and heat capacities from a fitted ideal gas state equation. (Vnuu = 15 V,J [adapted from Jolls and Coy (1992)]. It should be clear that Fundamental Equations would be of great use if they were generally available. The problem is that the complete form of the Fundamental Equation is not specified by classical thermodynamics; each substance has its own peculiarities The Calculus of Thennodynamics Chapter 5 130 that are reflected in different functionalities of the Fundamental Equation. Thus, there is no single Fundamental Equation governing the properties of all materials. The postulates of classical thermodynamics place some restrictions on the form of the Fundamental Equation. Let us examine the differential form of the Fundamental Equation in the energy representation, Bq. (5-3). Since !f. must be first order in mass or mole number, we can apply Euler's theorem (see Appendix C) to obtain the integrated form of Eq. (5-3). In differential form, dU is given by: du = (oful as d§. + (Clfu] av - - ,N - ful 0 ~ ( dV + £..J aN. 'S,N i=l - 1 dN;. (5-12) V.N [i] -•-J or as we demonstrated earlier n dU = Til§.- PdV + L t.t,J.Nj (5-13) i::l which can be directly integrated using Euler's theorem recognizing, of course, that T. P. and ~are intensive properties and kept constant. The result is U= T~- PV+ ~ llf/; (5-14) Equation (5-12) is a linear partial differential equation of the first order. Therefor~ the solution must be of the form U=x[{;. >·] (5-15) where x, y, z, ... can be ~. V, N 1 N, or any permutation of these variables. For a one-component system, it is often convenient to choose x == N, y = ~. and z = V~ we then obtain p • •• (5-16) or, since U =NV. U=g(S, V) (5-17) The only other requirements our prior developments placed on the form of the Fundamental Equation are that U should be a single-valued function of~. V, and N (see Postulate I), and that. (()fuf()~Y...JJ::: Tshould be nonnegative. 5.2 Intensive and Extensive Properties At this point, let us digress to a special case of Postulate I. In Eq. (5-17), we see that for a one-component system. only two properties, Sand V, are required to obtain the specific energy of a system. This is in no way a violation of Postulate I; by delineating Section 5.2 Intensive and Extensive Properties 131 the specific properties (expressed in tenns of unit mass or mole number), we have determined the .. intensity" of the system but notthe "extent'' ofthe system. To specify the system completely (e.g., so that it can be reproduced by others), we must specify the mass of the system in addition to S and V. The variables that express intensity of the system are zero order in mass and are called intensive variables. Variables that relate to extent of the system are first order in mass and are called extensive variables. We now prove that for a single-phase system of n components, any intensive property can be defmed by the values of n + 1 other intensive properties.2 Let us call b, c 1, c 2, •• •, cn+1 intensive properties of a single-phase simple system containing n components. In general, we can express b as a function of n + 2 other properties according to Postulate I. Let us choose these n + 2 as ca. c2, •. . , en+ I• and the total moles (or mass) N. Thus, db = (::,l dc1 + ... + [a~:~l . [I], N J dcn+1 + . [n+I], N (~!l dN (5-18) J.L •... ,cn+l j Integrating Eq. (5-18) by using Euler's theorem (see Appendix C), we have (dNdbl N= 0 (5-19) L, ... ,c..l Since N can be nonzero, ("dbl'dN)c ..... c must be zero. Therefore, Eq. (5-18) reduces to a function of n + 1 intensive variab~es. course, these n + 1 intensive variables must be independent, so that we clearly cannot use all of then mole fractions x 1, • •. , xn. We could, however, use n- 1 mole fractions in addition to two other intensive variables to obtain the ~uired n + 1. Note that this result is valid because we Jirnited the original set of n + 2 variables to include only one extensive variable; if we had included two extensive variables in the original n + 2 set, no partial derivative would have to be zero. Thus. we could state as a corollary to Postulate t Ot For a single-phase simple sys'U~ the change ofan} intensive 'Hlrilzble can be expressed as a function ofany n + 1 other independent intensive Wlrilzbles. We shall use this corollary frequently in Chapter 9 in dealing with the properties of mixtures. At this point, a word of caution is in order when dealing with intensive and extensive variables in partial derivatives. For example. if for a pure material we express U as a function of Sand V, then using Eqs. (5-18) and (5-19), we find (oUiaN) 5 v= 0 . • But (CJ!l/'dN)s, vis not zero~ from Eq. (5-7) applied to a pure material. 2 The proof is restricted to a single-phase system to allow us to choose any n + 2 variables as an independent set The proof can be extended to specific cases of composite simp~e systems provided that the n + 2 variables chosen fonn an independently variable set. The Ca/wlus of Thermodynamics Chapter 5 132 aul =~ (C>N1¥ (5-20) = u + N(auj =J.L (5-21) Since U = NU, we also have (aul i)N _.f oN _k,y Since U is not equal to ~ (dU/dN)s vis not equal to zero. Thus, each of the three (iJUidN}s,v have different connotations. The derivatives, (i1Uii1N)s,v~ (C)U/i)N)£V• flrst represents the change in the -specific energy-as we add more material while maintaining constant specific entropy and specific volume. Since we are holding two intensive variables constant during the process, all other intensive variables (e.g., T, P, etc.) for the pure material must remain unchanged. The only way to conduct the process is to enlarge the system in direct proportion to the added mass. The second and third cases, however, represent changes in the total and specific energy during a process in which we ma]ntain constant total entropy and total volume. Since we are adding mass to the system, the only way to keep total entropy and total volume constant is to change the specific entropy and specific volume (e.g., by varying T and P during the addition of mass). Thus, the specific energy changes as the state of the system is varied. The total energy changes because both the specific energy and mass vary. and Emmple5.1 In the entropy representation, the Fundamental Equation for a monatomic ideal gas (such as He, Ne, Kr. or Ar at low pressure) is (5-22) where SO. r.f', and V' are constants representing values n a reference a- base state. From the energy representation in the fonn of Eq. (5-l)~ determine the three equations of state in the fonn ofEqs. (5-5) through (5-7). Solution r (~S-S'YR) Solving Eq. (S-22) explicitly for U yields U =lf ( ~ exp (5-23) and, thus, 213 U: Nlf ( ~) exp (~S--S"))R) (5-24) Section 5.3 Methods tor Transfonning Derivatives 133 Using Bq. (5-24). U= U/N was plotted in Figure 5.2a as a function of Sand V. The equations of state can be found directly by partial differentiation of Eq. (5-23) or (5-24): l (~S-sjl R) 3 (5-25) exp C1<S--S')/R) (5-26) 213 l(aul (au1 3l (V"J - P • (~~1 = (~ =- ~ lf~la 00 r= = =l ( ] CJ§. _.N as~ R v exp 213 ~=(~~t =a~[Nr'(~fexp (~rs;jL =1~r exp(fs:>) (i -;!) (5-27) The results given in Bqs. (5-25) through (5-27) may be simplified if Eq. (5-22) is used to evaluate the exponential terms. If thts is done, then T ~ 2U . P 3R ' =2U . and " = 3V ' t"' u (53 _32 Rs) ( 5. 28) In fact, the assumptions behind the development of Eq. (5-22) are that U = ~ RT and PV=Kl'. 5.3 Methods for Transforming Derivatives Often one is faced with the problem of evaluating the magnitude of a particular partial derivative or integral involving non-measurable, derived thermodynamic properties. For example, suppose we needed a specific value for the entropy change M of a pure fluid between two well-defined states (T1• P 1) and (T2~ P2). The calculus of continuous functions in two variables works to specify the intensive state for a pure component system with n + 1 = 2 degrees of freedom. Thust AS= *~tdP+ J(~}u (5-29) The second integral involves (as/a1)p which equals CIT. Presumably, values of cp may be tabulated for the substance of interest at particular temperatures and pressures. But what about the first integral that requires evaluation of (dS/aP)T which is not readily available? We need some convenient way to express this partial derivative in terms of measured or tabulated properties. Again, by adhering to the calculus of continuous functions, we have several tools for manipulating these derivatives. Some of the more important ones for applications in thermodynamics are listed below. For a general function F(~ y) of two variables~ The Calculus d Thermodynamics Chapter 5 134 1. DeritJative inversion (5-30) For example, asj 1 (5-31) (aP )r =(dPidS}r 2. Triple produ.d (xyz-1 rule) l (~l(~ 1~ (S-32) =-l For example, (t~l(~~u~~1 = - I 3. (5-33) Chain rule expanswn to add another independent variable Cl> (aFl = dy ((JFl (aq,l Oy (iJF!dci>)x = (dyldel>).x del> (5-34) For example, set <f)= T as 1= <as;anp = c;r =l (aH) (oHia1)p cp (5-35) r 4. MaxweU redprocizy rekdiotult.ip Maxwel1 's reciprocity theorem states that the value of 2nd or higher order derivatives is independent of the order of the differentiation for a smoothly varying, continuous function F(x, y) l (d(d:dy),l (d(d~~xly = (5-36) or representing these 2nd order derivatives in abbreviated form; FXJ=Fyx (5-37) The relationships cited above in (1) - (4) can easily be extended to larger functionalitiesl for example where n + 2 variables exist such as in the Fundamental Equation. Our original problem of evaluating the integral containing (oS/aP) 1 can be made tractable using a Maxwell relation in the appropriate thermodynamic variables. Later in this chapter we will estab]ish methods to select the appropri.ate variables for tl)ese transformations, but for now, leCs employ the Gibbs free energy function, G. For a pure system, G is written in intensive form as a function of Tand P, G = j{T,P), and dG=-SdT+ VdP (5-38) Section 5.3 Methods fOI' Transforming Derivatives 135 where Thus, _(acaa/dT)pl- (acac;aP)r1~ _ Grp- (Jp - dT -GPT (5-39) or (5-40) so we can use PVT properties or a fitted PVT equation of state to determine (CJSI'dP)T and then evaluate the integral directly. Earlier in Section 5.1, we stated that the 2nd derivatives of Uwith respect to§. and V[see Eqs. (5-10) and (5-11)] were related by Maxwell's reciprocity theorem. Now we can show that these 2nd derivatives are indeed equal to one another. uy_~= a2 u a2u asav = avas=~~ (5-41) where we have used the abbreviations introduced earlier for the 2nd derivative. By taking the reciprocal ofEq. (5-40), we immediately see that u~~=-(~lN =(~~t=~y (5-42) recognizing, of course, that we have returned to an extensive fonn for U. In addition to the derivative manipulations characterized above, it is also useful to illustrate some of the different properties of intensive and extensive variables. A partial derivative may involve intensive and extensive variables. For a pure material, only n + I = 2 intensive variables are independent; hence, a partial derivative involving only intensive variables b, c. and d can be expressed as (oblac)d, where it is implied that N is constant. That is, (5-43) We now show that for a pure material, any partial derivative involving extensive variables can always be reduced to expressions involving partial derivatives of entirely intensive variables. Now consider the derivative (()b/CJc)d. ~where one of the four variables is extensive. ) (i). If c is extensive, then (5-44) The Calculus of Thermodynamics Chapter 5 136 The proof follows froiD applying Euler's theorem ro b = f(t;_, d. e): 0= (db1 ac- -c .~ l(h). If b is extensive, then l. (iJc/~!!)d, (~~ = e ~ ~ = = (5-45) l(iii). If d (or~) is extensive. then (5-46) Since and (~~l J~~J l c,t. ,e then abde I1 ( (aN/oc)b.~ e:::- =d (ahl (iJN/ob)c, e = oc N= (abl de Thus, if only one variable is extensive. the partial derivative is ftnite and nonzero only if d ore is extensive. The extensive variable may be deleted to yield a partial involving only intensive variables. Now consider the three cases in which two of the four variables are extensive. 2(i). If band care extensive, then (5-47) which follows directly by applying Euler·s theorem to!!_= f(£, d, e): ~=(~!l} and !!.=Nb, ~:Nc 2(ii). If b aod d (or e) are extensive, then (~! Since, expanding !! == Nb, LN[(~~1-!(~:),] = (5-48) Section 5.4 Jacobian Transfofmations 137 (!~1. =N(~!L +b(~~L (5-49) Eq. (5-46) may be used to reduce (ob/CJc)~.- The last tennis reduced as follows: 2JN1 _- (i)gjoc)N (OC -' e- (d#ON)C, e = _ N(CJdloc)t! d t! 2(fu). If d aDd e are extensive, then (5-50) The numerator and denominator can each be reduced by applying the results of case 2(ii). Any partial derivative involving three extensive variables can now be reduced to partials involving two extensive variables by using Eq. (5-49) followed by one or more of the steps iHustrated above. Similarly, partials involving four extensive variables can be reduced to three, etc. The net result is that any partial derivati11e for a pure material can be expressed in terms ofpartials invol.ving only three inten.rille variables. 5.4 Jacobian Transformations A useful technique for manipulating thermodynamic properties and their derivatives involves a transformation using Jacobians or functional determinants. For the derivatives of functions of 2 variables./ (x, y) and g(x, y ), the ]acobiJln is defined as: a{f, g)= Cl(x, y) - (!), (~). -(211(E.s_l_(.?ti(M} (~) (~) J ay1 - )' ax Oy (5-51) dx % Certain properties of Jacobians make them particularly useful for transforming derivatives of thermodynamic variables. These include: 1. Transposition au; s) acg. I> (){x, y) =- d(.r. y) 2. ln11ersion d([. g) 1 O{x, y) - o(x. y) (5-53) CJ(f, g) 3. CluJin rule expllllsion C)(£ g) _ d(/, g) d(Z. w) a(x, y) - o{z. w) o(x, y) (5-54) The Calculus of Thetmodynamics Chapter 5 138 where z and w are two additional variables. Another important property is the simplification that occurs to Eq. (5-51) if we are only interested in evaluating the first partial derivative of a functionjtc~. g) with respect to z at constant g: (Etl- a(f; g) (5-55) dZ - o(z, g) By applying the chain rule expansion and inversion property [Eqs. (5-54) and (5-53)], Eq. (5-55) becomes: act&) (dZ ~l= d(x,y) (5-51) a(r, g1 acx. y) The properties of two-variable Jacobians are easily extended to functions in m variables xi(i=l, ..., m),j,{x1, x 2, ... , x,J. The Jacobian is O(fl,f2, ...• f,.) (5-58) acx •• ~····· xm) or in determinant form atl ax. att a~ ox2 axm (5-59) ofm dfm at"m axl ax2 oxm where each partial derivative holds all X; constant except the one involved in the differentiation. Let's consider a few examples to illustrate how Jacobians are used in thermodynamics. &ample5.2 Consider the isenthalpic (Jou]e-Thompson) expansion of a fluid across a well-insulated va]ve. Here, the derivative (~~l., Joule-Thompson coefficient= aH is of interest. Express a.H in terms of measurable properties. Solution For a pure fluid, the specific enthalpy, H = f(T. P), so we can use Eq. (S-57) wtth /= T. 8 = H, and z=P. Thus, Section 5.4 Jacobian Transformations 139 d(T, H) d(X, ~) ()(P, H) d(x. y) (5-60) Now we need toselectxandy. Although we have many choices, frequently.as in this case, we want to obtain expressions in terms of measurable properties like T, P. V,etc. A good first guess for x and y would be to use T and P since they are in the derivative of interest. Thus, d(T, H) d(T, P) d(P,H) (5-61) Cl(T, P) Using the tranrposition property [Eq. (5-52)] and Eq. (5-55), Eq. (5-61) is quickly simplified to: arj (~~~ (~~~ _ o(P, n _ )r ___ )r d(H, 7) (dP 1- -a(H, P)- joHj iJ(T,P) (5-62) -CP lar ~ In this instance, we could have used the triple product directly (Eq. (5-33)) to obtain Eq. (S-62). However, such direct lransformations are not always possible. For example, in Section 5.9 we will demonstrate how Jacobian transformations are of considerable use in manipulating complex partial derivatives of multivariable functions. Before we try ID evaluate (()H/c)Phin Eq. (5-62) in terms of measurable properties. let's consider another example. Example 5.3 The jsentropic (reversible. adiabatic) expansion of a pure fluid iovolves the derivative: ((}TloP)s = isentropic coefficient =0-s Express the isentropic coefficient in terms of measurable properties. Solution The as derivative can also be reconstructed using 1acobians arj = acr. P) (aP) o(P, ~ a~.~ (~I aP _ ~ _ -T(as( _ as( - -cp - cp aP )r as o(T, S) :. ar> r (5-63) The Calculus of Thermodynamics Chapter 5 140 In order to express cx5 (or o.H} in tenns of measurable properties we must convert (CJS/aP)T [or (i)H/clP}r] to another form. To start with. consider the closed system combined Law equation which is equivalent to the intensive fonn of the Fundamental Equation in differential fonn (5-64) dU=TdS-PdV By rearranging, apd differentiating with respect to P at constant T. we get asj 1 (au( P (av( (aP )r =T l ()p )r + T ()p )r (5-65) Again. we end up with a cumbersome derivative @UioP)T' Some readers may recognize that we need to introduce an auxiliary potential function that has as its natural variables P and T. This is the Gibbs free energy G =f(T. P) wtUch is defined as: G= U+ PV-TS So by chain rule expansion of the differentials: dG=dU + PdV+ VdP- (TdS+Sd1) Thus. by substituting the expression for d.U given in Eq. (5-64}~ (5-66) dG becomes (5-67) dG=-SdT+ VdP And, by usiog Maxwell's reciprocity theorem, we obtain (!!1 =-(:~t (5-68) Now we can revise Eq. (5-63) for as to ~s=(~!l:+ ~(~~1 (5-69) Let"s see if we can use this result to help us evaluate the derivative (iJH/aP) 1 in the O.n expression [Eq. (5-62)]. Here we begin with the definition of H U + PV, differentiate. and use the closed system combined taw as we did before to obtajn = dH == TdS + VdP (5-10) Thus. by differentiating with respect toP at constant T, we get (~~1 =r(~~1+ v (5-71) Because we already have an expression for (iJStaP)r in terms of measurable PVT properties (Eq. (5-68)), we can express the Joule-Thompson coefficienl CJ"H as ijT -[T(- dVIo1)p + V] ~n:~Pl:-----c-p_____ where only PVT properties and c, values are needed to evaluate et.H- . (5-72) 141 Most vapor compression refrigeration systems in domestic use employ Joule-Thompson expansions to cool the circulating refrigerant. Thus. in these practical situations. ~ must be> 0 to obtain cooling as the pressure is lowered across the valve. What happens if the fluid being expanded is an ideal gas? In this case. a,;.= 0; thus. fluids whose PVJ' properties closely approximate ideal gas behavior would not be good choices for refrigerants when Jou]e..Thompsen expallsions are used. An alternative is to use an adiabatic turbine to expand the fluid to a lower pressure. Here. a 9 is the appropriate sealing parameter as it provides the maxbnum cooling effect for the limiting c:ase of a reversible, adiabatic or isentropic expansion. Now. even an ideal gas will cool upon e:xpansion. 5.5 Reconstruction of the Ftrtdamental Equation The Gibbs Pun.damental Equation where U=/~. V.N., ...,N,J provides the necessary and sufficient information needed to describe all the stable equilibrium states of any simple system. Unfortunately.~ V N coordinates are not completely amenable to direct measurement-for example. nodired.-reading entropy meters e:xistl Thus aforma1 transformation is required that preserves the encoded information content of the Fundamental Equation while expressing the functiona1 depeodence in variables other than§.. v. and.N1 (i=l .... , n). Once we have this procedure, we can be sure that all other infonnation of interest to us in classical thermodynamics ean be obtained from the transformed equation. The Legendre tnnsfonnation provides a rigorous mathematical route to achieve the desired reconstruction of the Ftmdamental Equation. As shown in Section 5.1. if the Fundamental Equatioo were known. the properties T. P, and J.a.,. could be determined by partial differentiation as expressed in the equations of state in Bqs. (5·5) through (5-7). Altemativdy. the Pondamental Equation c.an be recovered, if alJ the equatiODS of state are known, by substituting these equations into Bq. (5-13) or the common integmtedform (Bq. (S-14)] (5-73) As shown in Section S.2. then+ 2 intensive variables. T. P, and 1Ji (i:=l ..... n), which are expressed explicitly by the equations ofstate, are not all independently variable. Any one of these variables can be expressed in diffes:ential fonn in tenns of the other n + 1 variables and. upon integration, an expression connecting the n + 2 variables can be determined to within an arbitruy constant. It thus follows that only n + 1 equations of state are necessary to determine the.Pundamenta1 Equation to within an arbitrary constant (equivalent to specifying one or more reference state conditions). In Bxample 5.1. ifEq. (5-28) is combined with Bq. (5-22) to eliminateS, p. can be expressed in terms of reference state constants ll'. V' and S'; and variables T. V. As an example, let us consider a pure material. The intensive Gibbs free energy or chemical potential, G = JL, can be expressed as a function of T and P: The Calculus of Thermodynamics Chapter s 142 d~=(~}iT+(~}lP (5-74) =-Sa nd (dJ,l./aP)r= v. If In Section 5.4 [see Eq. (5-67)], it was shown that (dWOT}p ltaneously to obtain Eqs. (5-5) and (5-6) were known, we could solve these simu (5-75) S = g(T, P) and (5-76) ly in analytical form, but These equations are usually available. although not necessari us assume that we have the entropy is known only to within an arbitrary constant. Let available Eqs. (5-76) and (5-77): (5-77) V=g '(T, P) ition. Substitution into Eq. (5-74) T=T', P =po, V = VO, and and integration from an arbitrary reference state for which 0 f.L = f.L leads to (5-78) J.l = J.L0 - S"(T - T')- IT g''(T, P) dT + I' g'(T. P) dP 'f P. Although this equation which is the desired relationship for J.1 as a function ofT and , (S-6), and (5-78) are contains two arbitrary constants (S' and J.L~. when Eqs. (5-5) tal Equation, these two substituted into Eq. (5-73) in order to obtain the Fundamen find arbitrary constants appear as a sum; in particular, we would (5-79) where~ is an arbitrary constant or reference state cond r or (5-80) JJ.".>. it is clear that only Of the three arbitrary constants in Eq. (5-80) (i.e.• UO, S", and s of U and Sat the reference two can be chosen independently. Thus. we can set base value base value for J.l is uniquely state for which T=T', P;;; pt', V =V', but having done so, the specified by Eq. (5-80). 5.6 Legendre Transfonnations Eq. (5-1), the properties In the energy representation of the Fundamental Equation, always an appropriate §.. V, N1, . .. , Nn are treated as independent variables. This is not re can be measured much set of independent parameters. For example, since temperatu , ••.• Nn as the independent more conveniently than entropy. we might like to use T, V. N 1 express a propertysuch as variables. For a single-phase simple system, we can always , we know that a U in terms of n + 2 other properties such as T, V, N" .. ., N,. Thus func tion / exists such that (5-81) I I Section 5.6 Legendre T1ansfonnations 143 and dU =(~l d:r + '£N (£G- I.fr dV + .N i (!kl .r=l ' .Y...N, [•] (5-82) tiN; Given the Fundamental Equation, the function of Eq. (5-81) can be found by differentiating Eq. (5-1) to obtain Eq. (5-5), T= g-M_, V, N1, ... , N,J and then solving Eqs. (5-1) and (5-5) simultaneously in order to eliminate J... The result is an equation of the form U=f(T, V, N,N,.)={(~) , V ,N1, .. . 1, ... ,N.] (5-83) Y..N Although Eq. (5-83) is cf the fonn desired (i.e.• Eq. (5-81)), the information content ofEq. (5-83) is less than that of the Fundamental Equation. Equation (5-83) is a partial differential equation that can be integrated to yield the Fundamental Equation only to within an arbitrary function of integration. We must now ask whether or not there are other functions with the same information content as that of the Fundamental Equation but with independent variables different than §_, V, N1, •• •1 Nn. The answer is that there are such functions if we are willing to restrict ourselves to a set of independent variables in which we choose only one from each of the following (n+2) pairs: {~. T), { V. P]. (Ni, P,;} for i=l, n. These pairs of variables are usually referred to as conjugate coordinates. Note that there is one extensive (e.g.,§., V, Ni) and one intensive (e.g., T, P, Jli) variable in each conjugate pair. This grouping of variables is a natural consequence of the original formulation of the Fundamental Equation in Gibbs coordinates. As mentioned in Section 5.3, conjugate pairs of additional natural variables may be added to the Fundamental Equation to account for non-PV work effects. Table 5.1 lists the pairs of conjugate coordinates that are commonly encountered in problems of classical thennodynamics. Table 5.1 Coojugate Coordinates Type &tensive Para01eter Intensive Parameter Reversible heat flow T Pressu~ Voaume work -P Mass flow enthalpy and entropy Stufacedefonnation work Generalized work a The Calculus of Thermodynamics Chapter 5 144 To formulate these functions, a Legendre transform is employed. Such a transformation stems from a basic theorem in line geometry, and although the rigorous proof is no simple task, the results are easy to apply. The basic principle is that a curve consisting of a locus of points can be described completely by the tangent lines that form the enve]ope of the curve. Lef s consider a simple case to illustrate what Legendre transfonns are and how they are used. We define a basis function yC0>==f(x) with only one independent variable. The graph of / 0) versus x shown in Figure 5.3 is indicative of a well-behaved, continuously differentiable function. For any value of x~ a straight line tangent to the curve has a defined slope and intercept with the y-axis as shown on the figure. If we call the slope; =dymldx and the intercept yCI>. we can construct an infinite set of line tangents to the original curve yCO> = f(x) such that: y(O) = 9: + yC I) ( 5-84) for all values of x. By knowing values of y<J) and~ for every value ofx we can reconstruct the original function y<O>. The Legendre transformation is carried out by solving Eq. (5-84) fory<•> and inverting the functional relationship between~ and x. Thus, yO>= yro> _ = f[~] (5-85) x; Now y<•> can be represented as a function of~ with its derivative slope given by, -x= #') ~ and tangent intercept equal to y<0>. The function yO> is the Legendre transform in one dimension. It clearly contains the same information content as our original basis function y<O> since one can reproduce the original function plotted in Figure 5.3 from the set of line tangents. An analogous approach can be used for functions in two variables, y<O> = f(x 1, x2). Here there are two characteristic slopes corresponding to partial derivatives: ~. =(:i~l and ~=~=l (5-86) i The transforms y(l) = y(O)- x1 ; 1 and 2>~ yCO>- Xt l;l -~~are shown in Figure 5.4. In this 2-D case the transforms are obtained from planar tangents to the basis function surface rather than from line tangents to a curve. Now we can generalize the results tom variables, where the basis function, y0) = f(x 1, .. •, x,J (5-87) represents an m-dimensional surface. There are m f.trSt-order partial derivatives of ;/0> with respect to each of the m independent variables, x1, ••• , xm. Defming these derivatives as ~i• (5-88) Section 5.6 Legendre Transformations 145 Family of line tangents ~ y.o> = ;x + yl} t y.o> { yeo>= f(x) yO>= y.o>-;x !~"' df lldx "'f'(x) 0 ~ = f'(x al'\ at Xa I 1 Figure 5.3 Reconstructing a function y =j(x) using a family of ltne tangents-the one-dimensional Legendre Transform concept. or (5-89) where the symbol [x;] in the subscript of the partial derivative indicates that X; is not held constant. It follows that the variation of yCO) with x 1 could be described by the envelope of tangents in the / 01 - x 1 plane. If yC11 is the intercept of the tangent corresponding to ~~, Y(l)(~ 1 • Xl•· ..• xm) = / 0 1 ) - ~ x, (5-90) The function / >(~ 1 • x 2, . ..• x,J is called the first Legendre transform of / 0) with respect to xa. In other words, a Legendre transform results in a new function in which one or more independent variables is replaced by its slope. There are obviously m different first transforms, depending on the ordering of the variables x 1, .•• , x111• Higher·order transforms are defined in a similar manner; thus, y(t)(~J~· .. , ~ xk+l•· ... x,J is the kth Legendre transfonn: 1 k Yc~r.> =,CO> - L ~ri (5-91) i=l The total differential of the original basis function as 1ft dy(O) =I, ~i dx.i i=l y(o) can, of course, be expressed The Calwlus of Thennodynamlcs Chapte1 5 146 E 2 Figure 5.4 Geometric construction for a two-dimensional Legendre transformation using y(O) =f[.xp -s]. Two first orrler transforms yU> =IJ~1 , xJ at© and y<ll f[x 1, ~]at (M) and one second order transform y<1 = /[~~· ~] at ®, are shown. = and the differential of the kth transform can be obtained by differentiating Eq. (5-91) using the chain rule and substituting Eq. (5-92) to give /c m. /c ~ ~ dx- ~X~.-~ ~.flx. dy(i)_ - ~ .,; i ~ .~. ~ ~. ' pl i=I i•) which simplifies to k dy(/c) m =- ~ xi~~+~ ~; dxi l::l (5-93) r-kt-1 Since y<lcl is a function of~~~· ..• ~"' xk+b· . . , xm, it follows from Eq. (5-93) that: For transformed varUlbles (1~ i 'Sf< )1 (ak:) l.. . r;J,.... ~,.x..,..... x. ; -x, •(5-94) Section 5.6 Legendre Transtormations 147 (5-95) Equation (5-94) is sometimes called the inverse Legendre transform. Equations (5-95) and (5-94) clearly show the canonical relationship between conjugate coordinates {x;. ~t} that is similar to forms found in classical mechanics. Eq. (5-95) is applicable for all cases where i > k, and thus one may generalize the result as a,<'-•> ay<i-'-> == ... ity<O> == ; . '::! dXI dX; (5-96) 1 dX; The partial derivatives in Eq. (5-96) were expressed without indicating the set of variables to be held constant. However, it is clear from the discussion above that the degree of the transform determines the set. and the only exception would be that variable used in the actual differentiation. For example, the restraints on the term ayO-l>/c)x; would be that ~ 1 , ••. , ~-~, xi+l , . ..• x, would be held constant. To illustrate the application of these relations. letyC«J be the total internal energy of a simple system U; then the Fundamental Equation would be given in Bq. (5-1). Suppose that we desired the transform yC2>. Using the ordering of variables given in Eq. (5-1), we can organize the transformation using the table below: 'YP,))= ..ti I .,,., =G u ~; ~ X; T T -s I 2 s v -P 2 -P -V 3 Nt J.L, 3 N• J.l, n-t-2 N11 J1 n-t-2 N,. 11 with this ordering of variables: (2)_/(r. ~ N - ~l• ~' 1•· .. , Nn) Y {5-97) with (5-98) (5-99) The Calculus of Thermodynamics Chapter 5 148 Then, with Eq. ( 5-91 ), y<2) = U- ~- (- PYJ::: G (5-100) where G is the total Gibbs energy. The analogs ofEqs. {5-92) and (5-93) are Eqs. (5-4) and (5-101). n dyo.) == dG ==- §. ar + v dP +I, J.L, dN, (5-101) i=l where the chemical potential ll; can be defined in several ways as lli =(:~]~f.N, I (5-102) •... ,[Ni), •.. .N,. cr equivalently as ~i=(:~.l (5-103) r :,P.Nt·· .•,[N~ •.. -.N, The significance of the Legendre transfonn is thus evident. The important thennodynamic property G is simply a partial Legendre transform of the energy U from (§_, V, N 1, ... , N,J space to (T, P, N 1, ..., N,J space. Equation (5-101) is also a Fundamental Equation and no loss in infonnation content has resulted in going from the U to the G representation. We could have started our transformation in reverse. Referring to the table of x; and ~; given above, we could have defined a new basis function as G = y<2>~ y•(O) and transformed back to U = y *(Z) :: y<0>, that is, to the old basis function. Again, there would be no loss of information content. The totm Legendre transform ofEq. (5-1) is n yCrt+l) = U- T§. + PV- 2, 11/Ji =0 [by Eq. (5-73)] (5-104) i=l Thus, Eq. (5-93) becomes n dy(n+l) = 0 =- §. dT + v dP - L Ni $; (5-105) i=l Equation (5-105) is known as the Gibbs-Duhem equation. Later in Chapter 9 on mixtures and in those chapters that follow, we will utilize the Gibbs-Duhem relationship extensively. It is important to point out certain generalities in the use of Legendre transforrnations. We introduced a general functional equation [Eq. (5-87)] with arbitrary ordering of x 1,.. ·~ xm. We noted that for each xi there was a conjugate coordinate variable, ~i [Eq. (5-88)]. We then illustrated that one could readily derive a functional relation with the same information content wherein we replaced independent variables x 1•••• , xk by Section 5.7 Graphical Representations of Thermodynamic Functions and Their Transforms 149 ~ 1 .... , ~. This was the ktb Legendre transform defmed in Eq. (5-91) and shown in differential form in Eq. (5-93). As we demonstrated in the example above, one may also redefine this kth Legendre transform as a new y(O) basis function if care is taken in defining the correct independent variables and conjugate coordinates. For example, we have shown that beginning with Eq. (5-l), we obtained the Gibbs energy potential function by a Legendre transform of U into T, P, N 1, .•. , Nn space. We could now use the Gibbs energy function as our basis function/0>, but the independent variable set (x,, ... , xnJ would be (T, P, N 1, ••. , N ,J with arbitrary ordering. The conjugate coordinate variables ; 1, . .. , ~m would still be defined by Eq. (5-88); for example, for the variable T, ~r= (dGid1)p N;:;;; -S~ for P, ~P = V, and for Ni, ~i ~ J.l.J In fact, we may select any Legendre t~sfor~ as th~ y<O) basis function by redefming the independent variable set. Examples illustrating the use of Legendre transforms for manipulating partial derivatives are found at the end of Section 5.8. If we had started the mathematical transform development with the entropy fotm of the Fundamental Equation ~ = fi!:!.. V, N 1, ••• , N11 ), we would have produced an equivalent set of transfonns that are called Massieu-Planck functions, which are sometimes easier to use than Legendre transformations, e.g., when working in r 1 coordinates [see Debenedetti (1986)]. Additional treatments of Legendre transformations are given by Callen (1985), Alberty (1994), Aris and Amundson (1973) , and Tisza (1966). 5.7 Graphical Representations of Thennodynamic Functions and Their Transforms Beginning with the Fundamental Equation for U [Eq. (5-1)], we can now transform one or ali of the independent variable set~. V, N1, •• ., N,.. Let us choose only the variable V and reorder so that V represents x 1• Then y<O::::y< 0)-~x =V-(-P)V=U+PV:H 11- ---- (5-106) where this particular Legendre transform is called the enthalpy. We note that H = j(P, §., N 1, .. . , Nn) (5-107) We call Eq. (5-107) a Fundamental Equation in the same way that we refer to Eq. (5-1)~ then P is x 1, §.is~. etc., and ~ 1 = (dHI()P) ~ V. We can recover Eq. (5-1) by carrying out a Legendre transform assuming that Eq. (5-1 07) is the yCO) function; that is, y<'> == y<O)- ~ 1 x 1 =H- (YJP =U (5-108) This transform can be readily shown in Figure 5.5 for a common pressure-enthalpy diagram. If a curve of constant entropy is considered, the slope is V. The intercept of this tangent of the enthalpy axis is, as shown, equal to the internal energy V. The CaJrulus of Thermodynamics Chapter 5 150 t I I •• I I -~--------J ~ { 0 p----. Figure S.5 With internal energy as the basis function, there are n + 2 pennutations of first Legendre transforms: the two common potential functions, H(P. §.. N 1•... , Nn) and A(T, V, N 1,. .. , N,.). and n other functions for the independent variable set of ~. V, N 1, .... Ni-h J!;, N;+ 1, ... , N11 • Since the ordering of components is arbitrary, we shall refer to the n functions as U'(§., V, Jl.h N2o···· N,J . In a similar manner, there are (n + 2)(n + 1)/2 second Legendre transforms: one is another potential function, the Gibbs free energy G(T, P, N 1••.., N,.); there are n of the formA'(T,IJ. 1, V.N2,•.. ,Nn) and n of the fonn H'(P, IJ.~o ~. N'b···· N,J, and (n)(n - 1)12 of the form U"(~ 1 .1J.2,~, V, N3, ... ,N,J. Third Legendre transforms would involve G',A".H" , and U"' potential functions. Table 5 .2lists the important potential functions used in chemical thermodynamics. Again it is instructive to re-e:umine the plots of the Fundamental Equation to show how various Legendre Transfonns are geometrically related to the USV surface. Fundamental equations are plotted in Figures 5.1 and 5.2 for pure water and ethylene. Multiple phase equilibrium conditions, such as the triple point of pure water and the liquid-vapor coexistence envelope occur at specific temperatures and pressures for each particular substance. At first glance, one might expect that the thermodynamic potential that involves the natural variables of temperature T and pressure P would yield useful relationships. This function is, of course, the Gibbs free energy G which is the second Legendre transform of U and can be visualized geometrically as the intersection point of a tangent plane to the U ~ V surface with the U axis. Section 5.7 Graphical Aapresentations of Thermodynamic Functions and Thei' T1ansforms 151 Table S.2 Thermodyna mic Potential Functions and the Gibbs·Duhem Relation for Simple Systems Symbol Function1 Ca.o.onk.al Coordinates' !l'"" T~-PY+L,J11 N, v yfO> §., y, N1 (i=l, ..., n) -H=U+PV - - H y(J) ~~ P, A ·l> T, V, N. (i=:l ,... , n) G yfl) T, P, ~ (i=1,... 1 .n) Total frr+1) T, P, )J.i (i=l,... , n) Type Functional Equation~ lntemal&.ergy • ;.J Enthalpy Helmbolt:z Free Energy G=U+Pf-~=H-~ Gibbs Free Energy Gibbs·Duhem Relation N, (i=l,..., n) - I Transform Type Total Differential II Internal Energy dU =Td~- Pdf+ L llflN, i=l II Enthalpy dH '=' Td§, + Y..dP + L JJ.tlN; i•l Helmholtz Free Energy " ~ = -§.dT- PdY, +L ll,dN1 ial Gibb5 Free Energy Gibbs-Duhem Relation 1 y<0>::: basis function 2 y(l >=First Legendre transform, Obtaioed by Euler integration 3 " dG=-~dT+ Y.dP+ l:wm, b:1 dftt+'l) =""§.tlf + 'f.dP - • L N,dfl, = 0 ~I yrl) =Second Legendre transfonn, ... The canonical coordinates td'er to the set of natural variables that preserves the infonnational content of the Fundameoral Equation, yr!J) = U. The Calculus of Thermodynamics Chapter 5 152 For a pure component system, the intensive form of the Fundamental Equation can be used as we did with Fi~s 5.1 and 5.2 where U is plotted as a function of S and V. 0 In this case, -y' >= U andy >= G =Jl ani points d multiple tangency with a plane rolling over the USV surface will have the same intersection point on the U axis and thus will 2 have identical values of,< >= G = j.l. Later in Chapter 6 we will show that this geometric condition rigorously describes the mathematical criteria for phase equilibrium. Thus, it is no accident that the triangular plane in Figure 5.1b corresponds to the triple point where solid ice. liquid water. and water vapor or steam coexist in equllibrium. The triple point is equivalent to a unique condition where triple planar tangency exists on the USV surface of Figure 5.1a. Similar common points for double tangency on the USV surface can be used to characterize two-phase equilibria, such as the liquid-vapor coexistence region for a pure substance. 5.8 Modifications to the Fundamental Equation for Non-simple Systems Additional variables given in the original expression for the Fundamental Equation [Eq. (5-l)] may be needed to describe the behavior of systems in many practical situations in chemical thennodynamics. Some of these cases have already been introduced in Chapters 3 and 4. For example, there is frequently a need to account for gravitational or inertial forces- potenti al and kinetic energy contributions to the total energy E. Systems with these effects, of course, are not simple systems as we have defined them in this text; but nonetheless, their behavior can be represented by modifying the Fundamental Equation given in Eqs. (5-l) or (5-2). A common approach is to int oduce additional generalized work terms. such that m dE or dU =Til§. - PdV + n L F1 · dx1+ L Jl,d.Nt j•l (5-109) i=l where the 1ast summation involving tenns Fj · ~represents all non-PdV work effects that may be important. A number of these are listed in Table 3.1 in Chapter 3. These effects typicaliy include: • Surface defonnation work- ada • Electric charge transport - E dq • Electric or magnetic polarization - EdD or HdB • Linear elastic deformation· Fx ~ The addition of these other work or energy effects does not violate the four postulates that we have set forth to develop the laws ofthermodynamics. In particular, Postulate I still hoJds in that two independently variable properties plus the masses of all components present are sufficient to specify the equilibrium states of a simple system. Non-simple systems with these non-Pd V work effects, with potential or kinetic energy effects, or with other constraints (such as semi-penneable boundaries) can be treated. Section 5.8 Modifications to the Fundamental Equation for Non-simple Systems 153 With sufficient information we can specify the equilibrium state of these systems or the path of a particular quasistatic process. Assuming that Eq. (5-109) can be integrated using Euler's theorem, we can write: E or U = fC§., V, N1•.•• , Nn, z 1, •••, !m) (5-110) which should be a well-behaved continuous function. Then all of the relationships developed earlier in this chapter can be utilized for manipulating partial derivatives and the like. For example, Maxw ell reciprocity relationships can be applied: (a:S•l =(:~I l - -· l,y, Nl' %i (5-lll l Nl ' xi [k) Legendre transfonnations can also be developed to interrelate thermodynamic properties. The following example problem illustrates the use of Legendre Transforms in manipulating partial derivatives. Example 5.4 5.4{a) Express (d~laP>r. N. in terms of PYTN properties and/or their derivatives. • Solution The variables held constant (T. N;) and variable involved in lhe differentiation (P) sugges t that y(J.) = G should be used dG=-§.dT+ VdP + l:" JJ.•.dN; i=l and we see that a Mu.well relationship for y~>: yJfl gives the desired result y<?l =a ==11 ~ )] 5.4(b) Express [ d{g_/T d(l/7) (a§_l (avl aP ar = ·~ - (5-112) .~ in terms of a Legendre transform of U. P.N; Solution = Because Q .,<2>is involved with its natural coordinates of P. T, N , first try to simplif y by 1 expanding the derivative, recognizing that Gr == (OGioT)p, N. =~ and that TS. -G=y(2)=U-T - -S+P'V - ~H-- d(l / 1))] [d(QIT I (aol =G- T oT P, N1 , = G + T§.=H - T§. +T§.= H (5-113) N1 Eq. (S-113) is the famous Gibbs-Helmholtz relationship which will be used frequently to show the temperature dependendes of various derived properties. The Cslculus of Thermodynamics Chapter 5 154 5.4(c) Express (ac,,aV)T in terms of PVTproperties and/or their derivatives. Solution We start by using the definition of Cv and the basis function of y<0>= U in intensive form. c."'(~~l and dU=T&-PdV=Cft+(~~ldV Therefore. with (aulaV),-::; T(i:)SiiJV)T-P and a Maxwell relation (~;1 = [ a[i~) -PJ a; Expanding. acv[ =(as! (accas;av)r)l-(aPI ( av )r av )r + T ar ar} With the V. T variable set, use ytl) =A to see if a suitable Maxwell relation can be developed to express (aS!oV>r differently dA=-SO-PdV and A1v={~~1 ={~~l=Avr and we can simplify the equation for (aC/dV>r to (~~1 ~+T(~l (5-114) 5.4(d) For systems where surface forces are important. the Fundamental Equation can be redefined by adding a term + a to account for the reversible work due to surface deformation, where a--surface tension in J/m2 and a= area in m2 • Develop an expression for (a¥oa)r, y, Ni in terms of properties that can be measured experimentally. a Solution The modified Fundamental Equation in differeotiaJ and integrated fonn is: n dll. =Ttl§.- PdV +ada+ L ~idN; Fl and IJ u = T§.- PV + a a + I, J.1.1 N1 i=. Section S.9 Relationships Between Partial Derivatives of Legendre Transforms 155 Again by inspecting the variable set 0', T, V, and Ni involved in the derivative, a yC2> transform is suggested that yields that set, so two lransformatjons from S to T and a to a coordinates are needed: n ttyC2l ~ -§.dT- ada - PdV + L, J.Lj tiN; i=J A Maxwell relation on the fU"st two tenns gives the desired result: y~=-(;~1.fr. =-[~l ~. !• N; , Ni (5-115) =f: 5.9 Relationships Between Partial Derivatives of Legendre Transforms Although not obvious at this stage of our theoretical development, it is extremely helpful to be able to express derivatives of Legendre transforms. For example, later in Chapter 7 we will introduce criteria that wilJ determine the limits of phase stability for single and multicomponent systems. Often the functional forms cf various PVTN equations of state and other models of non-ideal behavior, such as activity coefficient models, require that certain variables be used to facilitate calculation of the phase stability criteria. This situation leads directly to the manipulation of partial derivatives of Legendre transformations. The presentation below follows earlier work by Beegle, Modell, and Reid (1974) which has been updated by Kumar and Reid (1986) who applied Jacobian transformations to calculate partial derivatives of Legendre transforms. Single tJarUlbk transforms. Starting with Eq. (5-87), we wish to investigate the relations between derivatives of yCOl and the first transform y(l>, where y (I) ~ (5-116) = /(..., 1, ~·· ••, xm) ~~= a (~l !X) (5-117) 2'"""' XIII There are several ways to obtain the desired results, but the most expeditious involves the use of the derivative operators (fori> 1) (_£._1 o.l'; ,.,..... lx~ •...,x,.. and (_l__l a~. lx, ,.... Y)!:J (aa r Ytt and oz·· ··· :em (i-;t: 1) x "' (5-118) The Calculus of Thermodyntmics Chapter 5 156 ? 1.. . ~ Y~~> (a:1. . .. (a~ (5-119) The terms ft~> and >{ are second-o rder derivatives: 0 co>,. a2 y<o> (5-120) Yti - dXI dXi co>- a2fo> a.J Yu - (5-121) 1 Equation s (5-118} and (5-119) are of little value to obtain first derivatives since, in view of Eqs. (5-94) and (5-95). all)l ( x. = ct) __ -Y, - a~~ ,....x. (~1 =y~t) = y~O) = ,_. dX· ' 1 l (5-122) (i > 1) -,; ,.xl,..., lx;], .. •, x. YW to second-order derivatives of Howeve r, to relate, for example , employ the operator equation (Eq. (S-119)] on y~ll. Thus for i# 1: ayp)l (d~l - co - _!_ (ay~ul """Y1i - Y\of 'd.xl I" • o .(Ill -y~~) [ay~o'l -Y~~>y\0( I - - a.%t 2'"" "I .x. I" • .. ycm, one can (5-123) .t,.. where Eq. ( 5-96) was used to simplify the third step. In a similar manner, other secondand third-order derivatives may be readily transformed. A list of these is presente d in Table 5.3. Example 5.5 Assume that the basis function is U derivatives of this basis function. =f (§, y, N1, .••, N ). Determine yW2 in teems of 11 Solution 0 Withfl) =f(T, V, N 1, ... ,N11) and -/ ) =/(§_, y, N 1, ..•, N11), yt l) =A = U- TS - - - a is the Helmholtz energy. From Table 5.3, 3 iPsA a - -A yCI) -A 112 - - TTY- rrv- afl av - - arav (5-124) Section 5.9 Rela1ionstips Between Partial Derivatives of Legendre Transforms 157 (O) YU2 - (y~~~2 _ Ussv - u2 - Usss Usv ss u3 ss where the underbars have been omitted for simplicity. Now we can write - 2 (J s arav = - (a Pl 2 ar ~N = n;asav- (riTirJ§?)v.JdTI'dYJs N (fJTI~~ N ~ ('dTifJ$l (5-125) ~N Table 5.3 Second- and Third-Order Derivadves ofy(l) in Terms ofy<O> Derint:fve ln _yCI) Quantity operated upon: ~II :Equadoo Used (5-119) (S-119) (S-118) .111 Yu1 Yu .lll ylll YII . JI) YuJ •.II> (5-ll9) (t) (S·Il9) (S-119) (5-118) The Calculus ol Thermodynamics Chapter 5 158 Multipk Varillble Transforms. The development shown above was limited to the case where only a single variable was transformed. Should one wish to transform more than a single variable~ it is always possible to proceed a step at a time and transform each separately as was shown above. It is also possible to develop a more general technique to allow one to express the partial derivatives of a Legendre transform -f)(~1 ,... , ~i' -Xj+1, ... , Xm~ tenns of the basis function y<o) or, in general, some other Legendre transform y ~. where q = 1, 2, ...,j. The equations to obtain these secondorder derivatives are shown in Table 5.4 for the basis function and in Table 5.5 for other functions. In Table 5.4 there are three cases for~: that is,j > i~ k; lc ~j > i; andj s; i. k. Various cases are illustrated in Examples 5.6 and 5.7. Note that we can make use of Maxwell's reciprocity relation -/N. = y<fj to simplify the calculations. Table S.4 Relations Between Second-Order Derivatives of thejth Legendre Transfonn and the Basis Function D~O) B D F A 0 ~- =.1_ - 0 YW = :...___AD~O):....____ j > i. k y~ =- - - - J_>< ,k. D~O) J 1 J- D F n<o> ~ /J)_ ik- D(O) C1E G "<" . <k 1- ) _ l D~O) J 0.11: Kronecker delta, = 1 if i = j, = 0 if i ~ j. A: j terms; each with a value of (-Ow·). where m = 1. 2, ... ,j. B: j terms; each with a value of (~,4-)~ where m 1, 2, ... , j . C: (j- l) terms; each with a value of (1- ~;)Y;. where m = 1, 2, ... , (j- 1). D: (j- l) terms; each with a value of (1- SiJym! where m = 1, 2,... , (j- 1). E: One term; (1 - 51 8j;· F: One term; (1- Sfllk - 1"' = ;).Yt- G: (1- oi;X1- B,>y:>. o and ,co)11 ,co> 12 (0) YJi ,co) 21 D~O)= J (0) Yjl (0) . Y·· }} Section 5.9 Relationships Between Partial Derivatives of Legendre Transforms 159 Table S.S Relationships Between Second-Order Derivatives of thejth Legendre Ttansform and the (j- q) Transfonn" B' n~> Ir' Y&' A'IA" H' ---:-:-~j > i, k D(j-q) = q n<r> D' p A1 IA" f j >i Y(Jik~ = -~~-(j-q) • 1Dq <" J"<. _l "<k 1and U-q) {j-q) Y(i-q+l)(j-q+l) Y<f-qf-l)(i-qt2) nN> = q U-q) · · · Y(j-q+I)j .li-4) (j-q) Y~l)(;--q+l) >'(i-q+2)~2) (J-q) Yj(j-q+l) = a Zn 0 if r S s :: 1 if r> s A': (q-l)terms;each with avalueof[Z(j-q+l)iY~~-8cf.t,)i]wherep=(q-1),(q-2),... ,1 A": Z(}-q+-l)iY~-q) . B': (q- 1) ~s. each with a value of [Z(;-q+I)k y~~k- Su-p)k] where p = (q-1), (q-2), ... , 1 B'': Z(j-.q+l}JtYjtil> . H': ZU-q+I);Z{j-q+J)l;)'~) . D': (q- 1) ten:nsS each with a value of (1 - Ojk)Y~~ where p = (q-1), (q-2), ...~ 1 P: (1 - Oj.t)Y~-q - Ojft f: (1- Ojk}Z(j-q+t)iY}k-q) J': (q- 1) te~s. each with a value of (1- Oji)~t· where p =(q-1), (q-2)..... 1 K~ o- Oji)yJtti>- s-. L': ( 1 - OJ;)(l - Ojk)y~> Example 5.6 Rerate the derivative yfR to second-order derivative ofy!0 >for four cases: (a) j= 3. i = l, k:=::2 {b) j=3, i= l,k=3 The Calculus of Thermodynamics Chapter 5 160 =2, i =3, k =4 (c) j (d) j = 3, i = 3, k ;::: 3 Solulion Table 5.4 is employed. The results are: (a) j=3, i= L,k=2 --S12 - 022 D(30) - -- - - y<GJ 12 - 032 3 --S,, --812 --813 0 y( ) - ------:-::-;---- - = (b) j y~ D~O) 12- y<O> 13 yCO) 33 D{O) 3 =3, i = L. k =3 i 0 (l--S33)Y\ ( 1--SJ3)y~~ <1-o33)Y ~~ (3) _ _ -_s_. _l Y13 - (c) j ((Jj Y12 o33 _--S_2:....:.•_~_3_1~--o___ == _ D~O) 13 y~~ y~ D~O) =2, i =3, k = 4 (1 - ~24)y~~ ( 1 - ~2,)}'i~ - riP> (0) 2 (0) (d) j =3, i = 3. k = 3 rf.O) 2 Y33 =- D(O) (3) 3 Bxtlmple 5.7 Using Table 5.5, express: Y'ti in terms of derivatives of yCll. (b) ~~in tenns of derivatives ofy<2>. y<O) fO) y~ y<O) y(O) '3 y<O> y<O) II 024 21 24 = DiO) y<O} y<O) (1 - ~23)y n (1 - 023)Y23 - 023 ( 1- On)(l - o )f~ (a) y(O) 12 23 D(O) 2 14 24 34 Section 5.9 Relationships Between Partial Derivatives of Legendre Transfoons 161 Solution (a) In this case j = 4, i =2, k =3, q = 3. We will use the case where j > i, k. Notice that Z(;-q+t)i = ZC4-3+l)i = ~i· Since 2 =i, then Zu = 0. Also, Z(;-q+t)t: = 0. Dt''=D~ ) 1 where (1) Y22 nc•>3 - (l) Y23 y<i) y<f] y~ y<l) y(O) 33 34 yO> y<i} 34 Then -&23 -~33 DCO y(4}_ 0 --~n --032 0 0 23- y~ y~ - D<•> 3 y!}j lll 3 D(ll 3 (b) Here j>i j<k j = 5, i =2, k =6, q =3, yC2) y<2) 33 ~2) 34 yf2) 3S y(S)- 26- 34 y<}l y(2) -45 y<2) 35 ~2) 4S y<fj z(i-q+t); = ~z y<l) y(2) y<2> 46 y(2) 56 y!l) yfl) y<2> y~ 42 32 52 n<2> 3 yc;; y<l> 33 yC2) 24 y(2) 34 iii -).2> 35 yc;) y(2) 45 Y46 y(2) 45 y(2) y(2) 23 36 (2) Y35 ==- =I 55 y(2) 26 yC2) 36 (2) 56 v<2> 3 Let us now consider derivatives in intensive variables only and show that they can be reduced to expressions involving only S, T. P, V. In general, a partial derivative of intensive variables may involve U, H. A, G, S, T, P, V. We have excluded N because it is extensive;~ is also excluded because Jl is equal to Gfor a pure material The first step in reducing the general derivative (CJb/()c)d is to eliminate any of the four potential functions U, H. A. or G, which may appear as b, c. or d. The three possibilities are treated as follows: 1. If b is U, H, A, or G, eliminate it by using the differential form of the transform. For example, with Eq. (5-101), for a closed, non-reacting system, 162 The Calculus of Thermodynamics Chapter 5 s (5-126) 2. If c is U, H. A, or G, invert the derivative using Eq. (5-30): 1 (~ = (db)dc)4 (5-127) and then proceed by step 1. 3. If dis U, H, A. or G, bring the potential function into the brackets by using the triple product relation of Eq. (5-32): (abl ac = .... (ad/Clc)b (5-128) (od/ob)c and then proceed by step 1. Using these three steps, we can reduce any partial derivative involving only S, V. T, and P to a second-order derivative c:i one ri tbe four fonns of rhe Fundamental Equation. This is illustrated in Example 5.8 below. EmmpkS.B For a pure material there are 24 partial derivatives involving S, T, P. and V. Suggest a techoique to relate any such derivative to second-order derivatives of the Gibbs energy and, therefore, to CP' 1C7' and a, where ~,.-~(~:1 and a,s~(~l Solution Of the 24 derivatives, 12 are simply inverses. The others are as foJlows: ( 1) (oS/ilT)p• (2) cas;anv. (3) (iJSioP>']\ C4> cas;ap)\h C5) cas;aV)7\ (6) cas;aV)J?, (7) cavloP)p (8) (dV/dP)s. (9) (aVIa1)P' (10) (oV/C)T)St (11) (aP/oT)s- (12) (dPidT)v. Number (3) is equal to the negative of (9) from the Maxwell reciprocity theorem; similarly for (4) and (10). Number (S) is equal to (12) and (6) to (11). Thus, there remajn (1) through (8). Three of these are second-order derivatives of the Gibbs energy: (1) (oSidT)p == -o.,~ c,1r (3) (oSidP)r==-GrP==-Gn=-(oV/CJ1), =-VtxP <7) <avldP)1 = GPI' =-Vte., In addition, derivatives (2), (5), and (lO)arerelated by theX-Y-Z.l rule, as are (4), (6), and (8). [The sets (1)~ (3), (11) and (71 (9). (12) are also so related, but chese derivatives have already been eliminated.] Thus, it is only necessary to relate (2), (41 and (6) to derivatives a Section 5.9 Relationships Betweert Partial Derivatives of Legendre Transf arms 163 in G to allow all derivatives to be so expressed. The reduction of (2) has been inlroduced as Problem 5.12. For (4) and (6), we define om basis function G =y(O) =f(T , P, N ) 1 where thex~ values arc shown below. For (6}, we define y<t)::: H =!~. P, N1, ...) and the values are also shown below. x--s ,, y<ll .X ~ 1. T -s 2. p 3. N• .X ; J. -s -T v 2. p v Jl,, etc. 3. Na . _v== Y2(1)• Thus, st.nce (ava~) - 'P,N == (as1 av = ~·etc. <-Yo>)-1 12 But, by Table 5.3. and (dSidV)p =Ciet.PVT. For derivative (4), using the same basis function. we need to use a doubl e transform; thus, y<2>= U -f~, ~. N 1....) with y(lj X 2.. -s v -P 3. N, J,1.1, etc. 1. ·T and (iJ§.IdP)v,N:N(as/CJP)v:N(y~~rt. s1nceP =-y~l. From Table 5.4, (0 YJl - (2)Y12 - .lOL(O ) _ 1,,(0)) 2YttY2 2 v t2 and GTP G G TI' PP _ G2 TP The Calculus of Thennodynamics Chapter 5 164 Although the procedure described above is always valid and unequivocal, there are shortcut methods such as those introduced earlier in Sections 5.3 and 5.4 which, when applicable, are usually less tedious. BXIlmple 5.9 Evaluate the following partial derivatives as functions of P. V, T, their partial derivataves, and CP: {a) (aS;aP)0 ; (b) (dA/aG)T Solution (a) Using Eqs. (5-126) and (5-128). we find that as! __ (oG/CJP)s (()p _1- (oG/c)S)11 -S(CJTioP)s+ v S(()T/dS)p (5-129) Using Eq. (5-128) to eliminate (CJTiaP)s, (}TJ = _ (()SI()P)r = (dVId1)p ( iJP (iJSioT)p C,IT (5-130) asj =~-(av( (()p} TS dT), (5-131) 1 so that The entropy, S, io Eq. (5-131) can be expressed as a function of the desired variables, as demonstrated jn Section 8.1 [see Eq. (8.23)]. (b) This partial deTivativecan be reduced by the following shortcut procedure aA I (aa caAiiJV)r P )r = caG!dV)r =- V(oPioV)r in which Eq. (5-34) was employed. The insertion of V was a convenient but not an arbitrary choice. Although P could have been used in place of Vwithequal simplicity, S would have been less convenjenL The choice was guided by the fact that A =/(T, V) and G =f(Tl P); since Twas the constant in the differentiation, either T, V or T. P would be a convenient set of independent variables. As an alternative, Jacobian transforms could be used. Section 5.10 SUmmary 165 5.10 Summary Starting with the Fundamental Bquation U = f(S. V. N) as a basis function, this chapter introduces a number ofmathematical techniques formanipulating variables. "''he important methods discussed were: l. Euler integration of the Fundamental Equation 2. Maxwe1rs reciprocity theorem for relating 2nd and higher order derivatives 3. Derivative inversion, triple product_ and chain rule expansion for manipulating variables 4. Jacobian transformations for derivative manipulation S. Legendre transformations 10 express the U-/<$... V. N) Fundamental Equation in terms of different variables without losing any iof'onnation content 6. Modifications to the Fundamental Equation to account for non-PV work effects 7. Derivation of relationships involving partial derivatives of Legendre transforms In addition. graptUcal techniques were used to illustrate the geometric relationships of the USV surface and various Legendre transfomled functions. References and Suggested Readings Alberty, R.A. (1994), "Lecendre transforms in c:bemical thermodynamics." CMmical Revi~s. 94(6). p 1457-1482 (Chemist's pcnpcctivcon Legendrelransforms] Aria, R. and N.R. Amwtdson, (1973). Ma~Mmarical Mtlhods in Cht~mi&al Engineering, Prentice HaJJ, Eng~wood Cliffs, NJ. Vol2, p 197-201. {Iattoduetion to Legendre transformations} Beegle. B.L., M. ModelL and R.C. Reid (l97.ot). "Legendre t.ransfonns and their application in thermodynamics." .AICitE J. 20 (2). p 1194-1200. Cafteu. H.B. (1985), Tlumnodynamics and an /IIU'odlltfion to TMrmo.rtatistict, 2nd cd, Wiley. New York. [Legendre transfonnation.s) Clark, A.L and L. Katz ( 1939). 'Thermodynamic surfaces of lb.O," TJWU Royal Soc. CtJit/JdD. 3rd series, Section III. 33. p. 59-72 [Figure S.la taken from p 72]. Coy, D.C. (1993), "Visualizing lhennodynalnic stabilily and phase equilibrium through computer graphics. PhD dissertation. Dept. of Otem. Bng., Iowa State Univ., Ames, Iowa [3--0 images of variously ttansformtd and scaJed fundaDWIC.alsurface.s of pure. binary, and ternary aystems] DebenedeU.i, P.G. (1986), "Generalized Massieu-Planck functions.: Geometric representation exttema and uni~tueness properties," J. Clumt. Plrys. 85(4), p 2131-2139. Gyftopoulos, E.P. and O.P. BereUa (1991), ~s: F~ andAppliaUions, MaeMillanf New Yort. {Mechanical engineering porspec:tive] The Calculus of Thennodynamics Chapter 5 lolls. K.R. ( 1990). "Gibbs and the art of thermodynamics", Proceedings ofthe Gibbs Symposium, published in G.D. Mostow and D.G. Caldi (eds.) Amer. Math. Soc. and presented at Yale University, New Haven, CT(May 15-17, 1989) p 293-321. [computer generated fundamental surfaces for a pure fluid] Jolls, K.R., M.C. Schmitz, and D.C. Coy (1991)J "A new look. at an old subject." The C~mical. Engineer, [nstitution of Engineers (UK), No. 497, May 30, p 42. lolls, K.R. and D.C. Coy (1992), ..Gibbs's models visualized," Physics Today. March, p 96. Kumar, S.K. and R.C. Reid. (1986), "Derivation of the relationships between partial derivatives of Legendre transforms'', AIChE 1. 32 (7). p 1224-1226. Tisza, L. (1966), Generalized Themwdynmnics, Mrr Press, Cambridge, MA, Ch. 2 in general and specifically, p 61, 136, 236. [Transformation methods] Problems 5.1. (a) Demonstrate the utility of Eq . (5-96) fer a system with six components and where i 4. Order the Fundamental Equation as = U == f@_, V, N 1, ... , N6) (b) If one were to write the Fundamental Equation as ~=f(U, V, N1, ... , N") (c) What would the derivative of the tota) Legendre transform be? Suppose that one were write the Fundamental Equation as yCO) = G =J(T, P~ N 1, •••~ N,) (d) Prepare a table showing the various conjugate coordinates, ~;. X;. Next, write the third Legendre transform of the basis function yC0>shown above. Prepare a~~. x 1 table for this transform. What generalization can you infer from this exercise? Choose the basis function for the Fundamental Equation to be / 0 ) =A= f(J_, N 1, ...• N11, T) Obtain an expression for y1}d in terms of derivatives y(O}. S.l. Given y<O> =f(§., Y, N1N2••.• , Nn) obtain the Legendre transform and its differential if one wished to work in the fonowing coordinate systems: (a) (b) 5.3. f(T. y, ~~~· .., J.tJ f[{l/1), V, N1, N1 , .. ., N,J Express the following partial derivatives in an equivalent form using the Fundamental Equation Problems 167 and Maxwell's relations. Indicate, if possible, how they might be measured experimentally. (a) ayC2ll ( av- S' SA. (b) (c) (d) (e) (f) S.6. N (b) laQ;' (c) ,N l(~l dV soN (d) (a~ l S (1) • N Carry out the following transformations. (a) s.s. (.al&.l Express (CJ~/()_!07, N,,.•., N11 as a function of P, ~. T. and their derivatives. Express @H/dP)T, N...h. N as a function of G and its derivatives and show how these may be given in tenns o(P, ~. T, N, and c,. Express [()~/7)/0(1 /1)]y, N...... N,. as a function of U and its derivatives. Express (dHiafh. N••..., N as a function of G and its derivatives and show how these may be given in terms o(P, f, T, N, and CP. Express ((}T/(}NA)¥... !· p,N" ... as a function of U and is independent derivatives. Express (aTlaNA)~. P, "• Nco ... as a function of u and its independent derivatives. For a one-component system, show: (a) v ~ -~JN av ()N .f- (b) v (Jp --N -av ~~L= The fa) lowing discussion is limited to one-dimensional motion along the x coordinate and for a constant-mass system. Newtonian mechanics relates the force on a particle to the mass and acceleration; that. is, Another way in which to study the dynamics of motion is with Lagrangian mechanics. In this case, a function Lis defined as L = g - fl), where ~ =the kinetic energy = ~l/2 and Cl> is the potential energy= j{x). Newton's Law in Lagrangian mechanics is given as l (:t)(~tl (~: L=f(x.X.) = . dx x= dt . =vel octty Another branch of mechanics uses a function (-H). which is a function of x and the momentum of a particle, p; that is, -H=f(p,x) with The Calculus of Thermcxlynamics Chapter 5 168 - (aLl_(asl_ ax - a.x - mx· P_ (a) (b) x. Using the concepts of Legendre transfonns, define -H in terms of L. p, and In the latter branch of mechanics. H is called the Hamiltonian; with your definition of H complete the following equations: (~~l = 5.7. and (~l = Express the following in terms of CP, P, V, T and derivatives of these variables. (a) (b) (dSiona. N (()AiaGh,N 5.8. Prove that for an n-component mixture there are (n + 2)(n + l)n independent second-order derivatives of the Fundament.al Equation. 5.9. A spherical tank contains l mol of heliwn gas at 10 bar and 300 K (see Figure P5.9). We would like to carry out an experiment in which helium is released to the atmosphere, but at the same time, the remaining contents of the sphere maintain a constant tot.al energy. U. Heating or cooling coils may be used to keep U constant during venting. Helium behaves as ao ideal gas with a constant value of Cv 12.6 J/mol K. Choose a base state where the specific enthalpy is zero at 300 K. = (a) When one-half of the helium has been vented, what is the temperature and pressure of the residual helium? What is the heat interaction? (b) (c) Detennine (oT/dP)Q.~ at the instant when venting begins. Repeat parts (a) and (b) if the base state were chosen so that the specific internal energy, were zero at 300 K. For part (a), what would be the residual helium temperature when 60% of the gas had been removed? (d) u. !·Heating coils Ot" •Cooling coils as clesired Figure P5.9 Problems 169 5.10. Choose as a basis function YW in terms of derivatives of the basis function. Discuss how experiments and obtain could be designed and conducted to obuun numerical values of the / 0) derivatives. Using your result, consider the foHowing problem. We have a system containing, initially, N; motes of a material. We wish to remove N/2 moles under conditions where the total entropy and the pressure remain constant. If the initial temperature is 400 K, what is the final temperature? Asswne that the base state for entropy is such that at 400 K and the system pressure, the specific entropy is 10 J/mol K.. Also assume that the heat capacity is 10 J/mol K. independent of temperature. 5.11. If the basis function is chosen as show that by performing two single-step transforms from / 2) to /I) and then to / 0 >. 5.12. Assume that the basis Legendretransfonn is/0>= f(P, T, N1, N2,1 ... ). Express y~in terms of derivatives of y<O) and interpret tbe results on a physical basjs. 5.13. Express y£{ in terms of derivatives of / / 0 } 1 ) for a ternary mixture when the basis function =U-=- f(~. V, NA• N8 , Nc) 5.14. Express y~ in terms of derivatives of the basis function y!O) = U =f(§_, V, NA, ... ) 5.15. Show that 5.16. y<m- 2} is a Legendre transform of a basis function y<0), where y(0) -j( - x 1, x2, ... , xm) /m-2) = f(~l' ~2'"'' ~m-2' Xm-1• xm) The Calculus of Thennodynamics Chapter 5 170 Derive a general equation to express the derivative y~==~~m-l) in terms of derivatives of a Legendre transform y<rl, where 0 ~ r< m -2. Define any derivatives and show what variables are held constant. 5.17. Show that "== S. =[av] [()P] and check to see if the relation holds for an ideal gas. c.., aPT av s 5.18. (a) Derive Eq. (5-118), for the derivative operator; ( l.,. .~J.~. l L-~)-~. (b) (c) a- a- dXi = ax; (0) ( aYli yft> - OxJ = L.~. Show that y~1l y<f:l /y~o; fori~ 1 as given in Table 5.1. Express the derivative (CJPidns.tA Jl. .-~ in lenns of derivatives of the basis function (0) -11 y =G. Show that the result given in part (c) above reduces to jl (d) (~~t :(~~t fcupure~rW 5.19. (a) Develop a suitable expression for ('iJTioP)ni(ClT/ClP)s in terms of PVI' properties and their derivatives. (b) Starting wi1h the basis function: y<O) = rf =f<§., V, a, N1, N2 p .. , N;J wheTe new conjugate coordinates {x;. ~}=(a, cr}have been introduced to account for surface effects. a for area and a for surface tension, develop a suitable expression for: (a~lda>r. v, N in terms of the appropriate 2nd derivative of a 1st order Legendre transfonn and its derivatives and show that it can be equated to a temperature derivarive of 0'. 5.20. The thermodynamics of rubber subjected to uniaxial tension is described by a fundamental equation given in differential form with all extensive quantities reduced by mass as, dU= TdS + V0 'td£ = e (L- LJ!Ln == Unear tensile strain "t =FI A0 == uniaxial tensile stress L == length in the direction of tension F =tensile force exerted on the rubber a L o> _o• Vo =length. area, and volume in the unstressed state. Problems (a) (b) (c) 171 Starting from this fundamental equation, express the stress t as a linear combination of the isothermal derivatives of U and S with respect to strain. For understanding rubber elasticity, it is important to determine the relative contributions of internal energy and entropy to 't. To achieve this, we perform the following experiment: Keeping the ends of a stretched rubber band fix.ed, we measure the stress 't as a function of temperature T as indicated in the figure. Show how the internal energy and enttopy contributions to stress can be separately evaluated on the basis of these T measurements. Give a graphical interpretation Figure PS.20 on the plot. A 1935 experiment by Meyer and Ferri indicated that, for a typical rubber extended to 4.5 times its original length at temperatures T> 210 K, the function depicted in the figure is adequately deseribed by the equation t = TflOO~ with Tin Kandt in MPa. What conclusion can you draw about entropic contributions to the elastic response? S.lt. Rocky and Rochelle Jones, while backpacking near the Presidential Range of the White Mountains of New Hampshire, were trapped in a severe thunderstorm. Rocky was very worried that they would be struck by lightning. As thunder and lightning bolts crashed along the mountain ridges, he remembered that one ought to be able to estimate the distance to a lightning strike by counting the seconds between when you first see the strike and when you first hear the thunder. Rochelle recalled her 10.40 training and remembered that the speed of sound could be related to certain thermodynamic properties. Deep jn her photographic memory she recalled that: v~ = (ClP/ap)swhere vc is the sound speed, Pis the pressure) p is the fluid density~ and Sis the entropy. But neither Rochelle nor Rocky can quite figure out how to evaluate the isentropic derivative (dP ;ap)s in terms of measurable properties. Please help them out! Rocky just timed the last strike and it was only 2 seconds away. State and justify all assumptions in your analysis. N.B. Rocky's portable solid-state weather computer which fits into his backpack has provided some va1uable meteorological data: air temperature = 0°C (32op) relative humidity = 80% barometric pressure =0.99 bar and falling rapidly! wind speed= 50 mph (80 kmph) from theSE In addition. Rochelle had a copy of Perry•s Handbook with her which provided some other {hopefully useful) thennophysical parameters for air. The C8Jculus of Thetmodynamlcs Chapter 5 cP = 29.3 J/rnol K molecular wt = 29 gfmol Tc::: -140 .7°C Pc= 37.2 bar 3 Pc = 1IVc = 350 kg/m And at 25°C, l bar p = 1.17 kg/m3 r Joule-Thompson coef4ficient= a.H = (OT /aP)H =0.23°Ciba viscosity :: 0.19 X l 0" Pa-s Prandtl number = Pr :: 0. 70 2 thermal conductivity= 1= 2.6 x 10" W/m K of CP and C.,: 5.22. (a) Stllrtin.g with the following ahernative definition CP =T(()S/iJ1),. C.,= T(CJS!iJ1)v (b) erties only. Using lhe show that CP- C'll can be expressed as a function of PVI prop result of part (a) show that CP- C.,= R for an ideal gas. ible to express the ratio Given that c,- C.,= f(P. V, 1) from part (a). is it poss erties and theirdmvatives? Explain K = C,IC"' as an explicit function of PVT prop your answer. 5.23. A basis function is defined as follows: O=. f(T, V, N 1, N2, ~1) With conjugate coordinate variable pairs given as .XI ~ T -s v -P Jll Nt N2 fJ-2 Ft ~I order in mass) and F 1 is a where z1 is a generalized extensive displacement (first other terms have their usual generalized intensive force (zero order in mass) and the mean.ing. Provide expressions for the following: (a) dyco> (b) yeo> (c) dyCSl (d) ( e) [()Fl/dJ.L1 1r.t.N1'~ in tenns of a derivative of N1 . (I ) (I) 1n terms o Ysss Ysss= f d r.. ~an You may want to use Table 5.3. . . . denvatiVes 1ts Problems 5.14. 173 Ify<0l ~ U= /~. V, N1,... , N,J, are y<n+2) and dyCn+2) always equal to zero? Answer yes or no and e:JC.plain briefly. lf Y(l) = j[t11, P. T, N2, N3,•.. , Nn], what are yCO) and dy(0)? What derived thermodynamic property does yC0>correspond to? yjf) Give one example of y~> = for any lah Legendre transform ofy!0> ~ U and any two independent variables z and J. Express your answer as an equality of partial derivatives involving derived properties (U, ~. !:f., G, ~and/or primitive variables (T, P, V, N1). 5.25. A basis function for a modified Fundamental Equation that includes reversible electromagnetic work effects in a two-component system is given as: y!O'J = lf =/[§., V, N1, N2• B] -j(x 1• X2..... x 5) where the conjugate pair of variables added for electromagnetic work effects is {x5, ~s} == {y B, H} and H = magnetic field strength (amp/m) V = tota1 volume subject to the magnetic field, includes the system volume and any free space outside the boundaries of the system (m3) B =magnetic induction (Volt·slm) For all cases of interest, you can assume that the volumetric lll8gnetic permeability is constanl (a) What are yes> and t!f>? (b) Derive an expression for (CJJ.L1/oBh:x. M.N2 At constant T, P, and Bcould you calculate J1l as a function of mole fraction of component (xi) from knowledge of how JJ.1 varies with x1 ? If so. explain your answer. (c) 5.26. According to a recent anicle in the March. 1989 issue ofScientific American. harbor sea)s are believed to have sophisticated acoustic transmission and detection systems that are similar to sonar. For example, mother seals locate their pups by measuring the difference in reflected acoustic wave arrival times in air and in water. Given that the speed of sound (vc) can be related to an isentropic derivative of pressure with respect to density: v:=(dPJi)p)s The Calculus of Thermodynamics Chapter 5 174 estimate about bow far a mother seal js from her pup if the time difference is 3 seconds. Air can be assumed to behave as an ideal gas with CP = 5/2 R. The properties of liquid water can be obtained from the steam tables. The air and water temperatures are about 2°C. S.27. A non-ideal gas of constant heat capacity Cv = 12.56 J/mol K undergo a reversible adumatic expansion. The gas is described by the van derWaals equation of state (P + a/V2)(V- b) =RT where a= 0.1362 J m3/mol2 and b = 3.215 x 10-s m3/mol. Derive an expression for the temperarure variation of the gas internal energy. and calculate its value when the gas volume of 400 moles is 0.1 m3 and its temperature is 2.94 K. S.28. The basic thermodynamic relationships for an axially stressed bar can be written as: dQ,,.v = Td§., dWte\1 =- 'td§, ~ =N£ where 't is the stress and £ is the srmin. Derive the Fundamental Equation for a one-component bar and show that (~I~.N--{a~l oT (JN .t 5.29. A very useful property is the partial molar volume of one component in a mixn.are. It is defined fer an n-component case as = ~ (dV/CJ~)T.PJJ, (J1 f If we have a basis function 0l, y<O> = f(§_. V, NA• N 8 ,... /'I,J we can see that, if we desired V8 , then - (2) VB =-y24 where "J• However, to evaluate we normally only have property relations which express P as a function of the independent variable set Vl T, NA• N8,.... N11 Use this variable set to define a new basis function y<O>, and express ~. T, NA• N 8 ,•••, N,. using the foil owing equation of state. P= NRT V- NAbA -NfiJ8 -NdJc·-· -N.,./JN f/l in zenns of • Prablams 17S where bA• b8...., bNare conatants and N =VIP with i =a. b,... .N S.30. The extensive variabJ~ that characterizes magnetic behavior is the magnetic dipole momentlln this situation, the fundamental equation is rewritten as U= V(§_, V,[,N) with the added conjugate variable set £r;1, ~~a[!, H) where H is the magnetic field strength. For a pure component system ck.vclop exptessions for; U (a) (b) f-4> and d:,'.., (e) (~/aB>gt, N