You are on page 1of 13

arXiv:physics/9704006v1 [physics.

ed-ph] 8 Apr 1997

Ideal Gas in a Finite Container∗
M. I. Molina

Facultad de Ciencias, Departamento de F´ısica, Universidad de Chile
Casilla 653, Las Palmeras 3425, Santiago, Chile.

To appear in Am. J. of Phys

1

is to replace the relevant statistical sums by integrals over a quasicontinuum of levels. but at the same time obscures the manner in which size effects enter when dealing with finite systems. when doing that. 2 . they are really assuming an infinite system. In this Note. This. leading to finite size effects in the thermodynamics of the system. simplifies the problem at hand. Sufficient conditions under which it is valid to replace sums by integrals have been examined by Stutz2 using the Euler summation formula. and washing away any finite size effects.1 Since the interlevel separation is inversely proportional to the system’s size. container volume. A different approach will be followed for the case of an impenetrable spherical container. yet finite. where we will expand the density of states in powers of the energy. and will explicitly derive the main finite volume corrections to thermodynamical quantities. Surface terms appear in the expansion coefficients. In the spirit of Fox. right at the outset. of course. For the case of the one-particle. Fox3 rederived Stutz results by using the Jacobi transformation. we illustrate the role of finite size effects by evaluating the main corrections to the canonical partition function of the ideal gas due to a large. we will express the partition function of a three-dimensional ideal gas enclosed in a impenetrable box in terms of complete elliptic integrals of the first kind.A common practice in many statistical thermodynamics textbooks when calculating the partition function. whose argument obeys a transcendental equation. one-dimensional partition function.

However. have the same functional dependence on the ratio (thermal wavelength× area/volume). for both. there are situations. In those conditions. We will assume a gas temperature (density) high (low) enough so that the Boltzmann statistics is applicable. according to Eq. a2 and a3 . the partition function is given by4 Z = [z]N /N! (1) with z= 3 X ∞ Y exp(−σi n2i ) (2) i=1 ni =1 and 3 σi = 8 λ(T ) ai !2 (3) √ where λ(T ) = h/ 3mkB T is the de Broglie thermal wavelength and kB . the spacing between two consecutive arguments in the exponential is very small (of the order of 10−10 for m = 10−22 g. can grow considerably (of the order of 10−3 for T = 1 K and ai = 10−5 cm). since σi depends on the inverse of the product temperature × (linear size of the system)2 . At room temperature. h are the Boltzmann and Planck constants. Ideal gas in a box: Consider an ideal monoatomic gas of N identical particles of mass m. while still small. at temperature T . enclosed in a box with sides a1 . the box and the spherical container. such as small volumes (“molecular cavities”) where such spacing. respectively. 3 .We will find that. the finite volume correction terms to thermodynamical quantities.(3). ai = 10 cm and T = 300 K).

The function we are interested in is the infinite sum s(σ) = ∞ X exp(−σn2 ) (4) n=1 which is a special case of the more general type of sum P∞ n=1 2 q n with |q| < 1. we obtain 1 s(σ) ≈ 2 "  1/2 π σ 4 −1 # (8) . we have  2K(k) 1 s(σ) =  2 π !1/2  − 1 (6) where K(k) is the complete elliptic integral of the first kind K(k) = R π/2 0 dα(1 − k sin(α)2 )−1/2 and the argument k satisfies the transcen- dental equation7 K(1 − k) − (σ/π)K(k) = 0 (7) Fig. At high temperatures (or large container dimensions) (σ/π) << 1 and the root of eq.(7). which implies.(7) lies close to k = 1. This sum can be expressed in terms of the theta function5 θ3 (0. the intersection of the two curves tends to occur very close to either k = 0 or k = 1.. As soon as (σ/π) deviates slightly from unity.(6). which justifies a perturbation expansion around those points. (5) From the identity6 (2K/π)1/2 = θ3 (0. according to eq. After inserting this into eq. q). q) = 1 + 2q + 2q 4 + 2q 9 + .1 shows the plot of K(1 − k) and (σ/π)K(k) for different values of (σ/π). K(k) ≈ (π/σ)K(0) = π 2 /2σ..

pressure P = kB T (∂ log(Z)/∂V )N. Entropy: S ≈ S0 − 5 s 3 NkB 128π λ(T )A V ! (13) . Pressure : P ≈ P0 " 1 1+ √ 96π λ(T )A V !# (11) where P0 = NkB T /V . specific heat C = (∂U/∂T )N.This approximation differs slightly from the standard textbook8 expression s(σ) = (1/2)(π/σ)1/2.V .T .(2) and (1). We see that eq. We obtain : Energy: U ≈ U0 " 1 1+ √ 96π λ(T )A V !# (10) where U0 = (3/2)NkB T and A = 2(a1 a2 + a1 a3 + a2 a3 ) is the area of the box and V = a1 a2 a3 is its volume.V and entropy S = kB log(Z) + kB T (∂ log(Z)/∂T )N. Specific Heat: C ≈ C0 " 1 1+ √ 384π λ(T )A V !# (12) where C0 = (3/2)NkB .(8) shows better agreement up to values of σ as high as σ ∼ 1.(8) into Eqs.  π3 σ1 σ2 σ3 !1/2  π3 σ1 σ2 σ3 !1/2 1 log(Z) ≈ N log  8 1 ≈ N log  8  3 X e σi +N log 1 − N π i=1   3 X e σi −N N π i=1  1/2 1/2 ! (9) We are now in position to evaluate the main corrections to the energy U = kB T 2 (∂ log(Z)/∂T )N.V . Table I shows a comparison of both expressions with the exact result. we have. After inserting Eq.

where k = (8π 2 mE/h2 )1/2 and jl (x) is a spherical Bessel function. we have an infinite number of solutions. n) we have an angular momentum degeneracy of (2l + 1)... The eigenvalue equation is : 9 jl (kR) = 0 with l = 0. 1. For each l. instead of using Eq. we can approximate the Xln using McMahon’s expansion: 10 (µ − 1) 4(µ − 1)(7µ − 31) 32(µ − 1)(83µ2 − 982µ + 3779) − − 8b 3 (8b)3 15 (8b)5 64 (µ − 1)(6949µ3 − 153855µ2 + 1585743µ − 6277237) + . (15) − 105 (8b)7 Xln = b − with µ ≡ (2l + 1)2 and b ≡ (n + (l/2))π.. 2.where  1 S0 = (3/2)NkB + NkB log  8 π3 σ1 σ2 σ3 !1/2  e . Ideal gas in spherical container: Let us now consider the same ideal gas of the preceding section enclosed in a spherical container of radius R.. Now. 2. indexed by an integer n = 1. .(15) to calculate directly the atomic partition function. In the limit of large R.. and for each (l. we will use it to evaluate the density of states. The energies 2 are Eln = (h2 /8π 2 mR2 )Xln . where Xln is the nth root of jl (x). For a container of “simple” shape it is natural to expect that the main contributions to the density of states will have the form: ρ(E) ≈ ρ1 E 1/2 + ρ2 + ρ3 E −1/2 6 (16) . the main correction due to the finite volume of the container has the form (thermal wavelength× area/volume). which is all we need to calculate the thermodynamic properties of the system. .. N (14) Thus.

R2 . The approximate density of states is then given by Eq. we perform a least-squares fit of the resulting histogram to the smooth function φ(E) = R1 h2 8π 2 mR2 !−3/2 E 3/2 + R2 h2 8π 2 mR2 !−1 E (18) obtaining the optimal values for R1 .(16) will depend upon the “dimensionality” of the container.(17). R2 . it is convenient to rewrite Eq.(17). for our spherical container where “volume” predominates over “area” and “length”. (19) 0 By using Eqs.where each term has the typical form of a three. with a small (constant) correction term.(16) in terms of dimensionless parameters R1 .(19) and the numerical values obtained for R1 .(15). the main contribution comes from the term proportional to E 1/2 . The relative importance of each term in Eq. and one dimensional density of states. Thus. defined as the number of states with energies less than E. R2 : ρ(E) ≈ (3/2)R1 h2 8π 2 mR2 !−3/2 E 1/2 + R2 h2 8π 2 mR2 !−1 (17) where we have kept only the surface correction term. Then. We proceed as follows: Starting from Eq. In this case. we can write the main corrections to several thermodynamics quantities: 7 . two. The atomic partition function is given by z= Z ∞ dE ρ(E)e−βE . we evaluate numerically the integrated density of states φ(E).

V are the area and volume of the sphere.0461 V !# (20) where U0 = (3/2)NkB T and A.0693NkB where  S0 = (3/2)NkB + NkB log 2. To its conceptual simplicity we have to oppose its rather slow numerical convergence. Pressure : P ≈ P0 " λ(T )A 1 + 0. requiring very many energies to obtain coefficients with a given accuracy. the main finite volume corrections for the sphere have the form constant × (thermal wavelength× area/volume). As in the case of the box. respectively.0461 V !# (21) where P0 = NkB T /V . Specific Heat: C ≈ C0 " λ(T )A 1 + 0.804 2 λ(T )A V 8mR kB T h2 !3/2 ! (23)  e .02305 V !# (22) where C0 = (3/2)NkB . Entropy: S ≈ S0 − 0. For 8 . N (24) The method outlined above for the sphere can be used for any other container of “simple” shape. provided the eigenergies are known either analytically or numerically.Energy: U ≈ U0 " λ(T )A 1 + 0.

the box and the spherical container. P0 . Tabensky for very stimulating discussions. R¨ossler and Professor R. The main finite size corrections have the same functional dependence on the ratio (thermal wavelength× area/volume) for both shapes. ACKNOWLEDGMENTS I wish to thank Professor J. C0 and S0 do not depend on the shape of the container. the thermodynamical quantities U0 . 9 .both.

10 Reference 5. Phys.A. 1965). New York. 116-117 (1971). 10 . “Comment on: ‘On the Validity of Converting Sums to Integrals in Quantum Statistical Mechanics’ ”. Abramowitz and I. 111. Am. p. Statistical Mechanics. Stutz. p. 81. 3rd ed.D. J. Lifshitz. 82 9 L. 371. p. 82. 39. Phys. Donald A.. for instance. p. 7 Reference 5. (Harper and Row. Handbook of Mathematical Functions. (Dover. 3 K.1 See. 579. Oxford. Stegun.M. p. 36. New York. New York. McQuarrie. p. 826-829 (1968). 5 M. Am. 4 Reference 1. Fox. p. p. 1977). 1976). 6 Reference 5. Quantum Mechanics (Pergamon. “On the Validity of Converting Sums to Integrals in Quantum Statistical Mechanics”.576. J. 8 Reference 1. Landau and E. 2 C. 591.

Table I Particle in a box: Comparison of the exact atomic partition function s(σ) (second column) with approximation (8) (third column) and with the standard approximation for high temperatures (fourth column). 1 and 2. while the dashed. 11 . dotted and dot-dashed curves are (σ/π)K(k) for (σ/π) = 0. The continuous line is K(1 − k).1 : Complete elliptic integral of the first kind.5.Captions List Fig. respectively.

249561 .001 0.749561   1/2 π σ −1 0.1226926 279.5249561 88.30249561 8.86226926 28.749561 12  1 2  1/2 π σ 0.25331414 2.00001 −σn n=1 e P∞ 2 1 2 0.σ 1 0.75331414 2.1 0.36226926 27.01 0.6226926 280.36226926 27.80024956 8.75331413 2.5 0.38622693 0.38631860 0.30249561 8.0249561 88.88622693 1.5249561 88.1226926 279.0001 0.

2 0.0 .8 1.6 0.5 2 K (k) 4 3 K (1-k) 13 K (k) 2 1 (1/2) K (k) 0 0 0.4 k 0.