Heat and Entropy Van der Waals-Zeeman Institute, University of Amsterdam Jacques C. P. Klaasse Lecture on Heat and Entropy, according to Clausius and Boltzmann Presentation IMNU, Hohhot, China June 2008 Contents of the presentation • • • • • • • • • Internal energy U Heat, Work, and Entropy Specific heat Entropy and statistics Boltzmann entropy Entropy and ideal gas Boltzmann distribution Specific heat revisited Conclusions Contents of the presentation • • • • • • • • • Internal energy U Heat, Work, and Entropy Specific heat Entropy and statistics Boltzmann entropy Entropy and ideal gas Boltzmann distribution Specific heat revisited Conclusions Internal energy U For an ideal gas, the “internal” energy, U, is nothing more than the total kinetic energy of the N particles: Ekin = (½).N.m.vav2 with m the mass and vav2 the average squared velocity. This kinetic energy shows to be only dependent on the temperature of the gas. A simple derivation, showed on the next sheets, gives for a mono-atomic ideal gas the simple relation: U = Ekin = (3/2)NkBT with kB the so-called Boltzmann constant. Internal energy U Take a particle of the gas with x-component of the velocity vx. On each collision with the yz-wall, an impulse of 2mvx is transferred. This happens vx/(2.lx) times per sec. => Total force for this single particle: F = m.vx2 / lx . For all particles together holds: F = m.(Σ vx2) / lx . With Σ vx2 = N.vx,av2 , follows for the pressure, p : p = F / lylz = (N/V).m. vx,av2 , with vx,av2 the “mean square” value of the velocity in the x-direction. All 3 directions are equivalent, thus: vx,av2 + vy,av2 + vz,av2 = vm2 is the total mean square velocity, with vm2 = 3 . vx,av2 . Internal energy U We have seen that for the pressure holds: p = (N/V).m. vx,av2 . Because of the equivalence of the three directions follows: vm2 = 3 . vx,av2 . Thus for p follows: p = (1/3).(N/V).m.vm2 . We know: Ekin = (1/2).N. m.vm2 , and p.V = N.kB.T . By eliminating m.vm2 and p follows Ekin = (3/2).N.kB.T => U = Ekin = U(T) . Contents of the presentation • • • • • • • • • Internal energy U Heat, Work, and Entropy Specific heat Entropy and statistics Boltzmann entropy Entropy and ideal gas Boltzmann distribution Specific heat revisited Conclusions Heat, Work, and Entropy For an ideal gas, energy can be added or removed by - mechanical work: ΔW = -p.ΔV - heat: ΔQ . We have, by consequence, dU = dQ + dW = dQ - p.dV The total internal energy, U, is, for the same amount of gas, only a function of T. Two isotherms in a pVdiagram, with Th > Tl. The internal energy U is constant along the isotherm. Heat, Work, and Entropy Let us go, in the pV diagram, from state 1 to state 2. ∫dU (1->2) is only a function of T. => independent of the path. ∫ However: dW (1->2) is different for the two path’s 1 -> A -> 2 1 -> 1’ -> B -> 2’ -> 2 ∫ (The pressure in – p.dV is much lower in the second case.) ∫ By consequence: also dQ depends on the path that is followed in the pV diagram. Two isotherms in a PVdiagram, with Th > Tl.. Going from 1 to 2, both Q and W are dependent on the path. Heat, Work, and Entropy We have seen that ∫dW is dependent on the path that is followed. However: ∫(1/p).dW = ∫dV is independent on the path: V(2)-V(1) is the same for both cases. Question: how to make ∫ dQ independent on the path? The related variable is T, so, we look for dQ = T d “?” . Clausius S, “entropy”. called this “missing variable” named It can be shown that ∫dS = ∫(1/T).dQ is independent on the path that is followed,…. ….. provided that during the process thermal equilibrium is never broken. Heat, Work, and Entropy We have defined dS = dQ/T . Q is dependent on the number of particles, N, so is S. That means that for the two boxes here S = S1 + S2 . If we open the valve, heat will flow from 2 -> 1. That means, if that heat is dQ, that dS1 = + dQ / 99.5 dS2 = - dQ /100.5 < - dS1 !! By the spontaneous temperature egalisation, the total entropy is increased. It can be shown that for any spontaneous process the entropy will increase. Entropy is, considering the whole system, always developing to a maximum value. Contents of the presentation • • • • • • • • • Internal energy U Heat, Work, and Entropy Specific heat Entropy and statistics Boltzmann entropy Entropy and ideal gas Boltzmann distribution Specific heat revisited Conclusions Specific Heat A method to determine the entropy content of a system (for instance a piece of material) is to determine the Specific Heat (this is the heat capacity per mass unit) as a function of T from T=0 up to the wanted T. For the heat capacity C holds C = dQ / dT . By consequence: (C/T).dT = dQ / T = dS. For S follows S = ∫ (C(T) / T) .dT from T=0 up to the final temperature. NB: for solids it is not very important whether we keep V or p constant. Specific Heat Typical heat-capacity curve for a solid from low T up to about half the “DebyeTemperature”, most of the order of about 100 K. If we make a plot of C/T versus T, the area under the curve is the entropy. Problem: we cannot measure from T=0 … Fortunately: the S contributions are low there. Specific Heat For mono-atomic noble gases we have seen that Ekin = (3/2).N.kB.T That means for the specific heat: C = (3/2).N.kB . This is in good agreement with experimental values for noble gases. This C is taken at constant volume: CV . You can also measure at constant pressure: Cp = 5/2 N.kB . Solids: constant volume is very difficult. But at low T for solids CV ≈ Cp At 300K the difference is only a few percent. Specific Heat For a two-level system the specific heat has a typical shape, called a “Schottky peak”. The maximum of the most simple two-level system (without degeneracy) is at T ≈ 0.3 * Δε / kB , with Δε the distance between the levels. Later in this course we will learn how to calculate this Schottky peak. Typical Schottky peak in the specific heat. The left part of the C/T curve increases roughly exponentially, the right part decreases as 1/T 3 . Contents of the presentation • • • • • • • • • Internal energy U Heat, Work, and Entropy Specific heat Entropy and statistics Boltzmann entropy Entropy and ideal gas Boltzmann distribution Specific heat revisited Conclusions Entropy and statistics Boltzmann: nature is ruled by statistics. Energy is distributed randomly over the given particles. Any possible configuration at given energy has equal probability. A possible configuration is often called a micro-state. A micro-state is also called a quantum state (individual energy levels are given by quantum mechanics). A certain solution for the energy distribution over the particles is (shortly) called a distribution. Entropy and statistics The distribution of highest probability (containing the highest number of possible configurations) is the equilibrium distribution. The equilibrium distribution is adopted most of the time, BUT NOT ALWAYS. Distributions far away from equilibrium have, in case of a large number of particles, a negligeable probability, but not zero probability. Entropy and statistics • Look at this system of 6 places. Particles can be distributed randomly with sign “+” or “-”. • Here we present a “distribution” with two “+” and four “-” particles. • In how many ways this distribution can be realized? The first + can be put at 6 positions, the second at 5. But we can change the two “+” particles without seeing the difference: the total amount of really different configurations is 6*5/2=15. Entropy and statistics The number of possible configurations, Ω, for arbitrary distributions n(+), n(-) equals: (6*5*4*3*2*1) / [n(+)*..*1]*[n(-)*..*1] . With the notation n*(n-1)*..*2*1 = n! follows: Ω = 6! / [n(+)! * n(-)! ]. Results for all 7 distributions +,- in case N=6: 0,6 1,5 2,4 3,3 4,2 5,1 6,0 Ω= 1 6 15 20 15 6 1 We see that 3,3 is most probable but others are quite well possible. If we take N = 60, the 30,30 is most probable but the curve is more sharply peaked. NB: the sum over all configurations equals 64 = 26. Entropy and statistics We repeat the results for all 7 distributions in case N=6: 0,6 1,5 2,4 3,3 4,2 5,1 6,0 Ω= 1 6 15 20 15 6 1 We see that the Ω for 2,4 (or 4,2) is 75% of the 3,3 value. During 78% of the time the system is in one of the three central configurations. If we take N = 60, the Ω for 20,40 is only 3% of the 30,30 value … Entropy and statistics Here we see plots for three numbers of N. The larger N, the more peaked is the curve around 50 %. For N= 6.1023 the peak is so sharp that the system is nearly all the time in the 50 % state. However, all other distributions are possible!! Ω N=6 N=60 N= 6.1023 Contents of the presentation • • • • • • • • • Internal energy U Heat, Work, and Entropy Specific heat Entropy and statistics Boltzmann entropy Entropy and ideal gas Boltzmann distribution Specific heat revisited Conclusions Boltzmann entropy Boltzmann realised that entropy has something to do with Ω. For two systems we know that the number of configurations is the product of the two: Ω = Ω1 * Ω2 . However, the entropy, being directly related to the heat content, has to be the SUM of both entropies. By consequence: S = const* log(Ω) . The const is called kB , the “Boltzmann constant”. Boltzmann entropy Some mathematics in order to calculate S for large N. For large N holds: log (N!) ≈ N.log(N) - N (Stirling formula). NB: with “log” we mean the natural log. For large N, counting is not possible. But if x = n(+) / N , and thus n(-) / N = 1-x , we can write log (Ω) = log { (N!) / [ n(+)! * n(-)! ] } = N*log(N)-N –Nx*log(Nx)+Nx-N(1-x)*log(N(1-x))+N(1-x). This can be simplified to: log (Ω) = - N { x*log(x) + (1-x)*log(1-x) } Boltzmann entropy We have found that for large N: S/kB = log (Ω) = - N { x*log(x) + (1-x)*log(1-x) } with 0 < x < 1 . For equilibrium we are looking for the maximum in this function. That means differentiating and zeroing. After some straightforward calculations one finds: log (x / (1-x) ) = 0 , or x = 1-x , or x = 0.5 … …. Just what we have seen for small numbers. This is not a sharp maximum, but going from S to Ω, you will find the sharpness presented before. Boltzmann entropy We have found : log (Ω) = - N { x*log(x) + (1-x)*log(1-x) } = - N * f(x) . Let the function between parentheses, f(x), differ only 1/1000 from its equilibrium value fmax, which is of the order of one. Then Ω / Ω(max) = exp ( N.[ fmax - 0.001 ] ) / exp ( N.fmax ) = 1 / exp ( 0.001*N ) . For N = 1023 , this means Ω / Ω(max) = 1 / exp (1020) .. and this is a tremendously small number… Boltzmann entropy We have seen that a collection of particles with two possibilities with equal probability is most in a 50/50 distribution. Think for instance on particles without interaction with spin up or down in zero field. But if we apply field, and we let the spins freely interact with a large system at temperature T, the distribution will change, dependent on the temperature and the magnetic field. We will generalize this to a system with a finite number of energy levels at εi, in thermal contact with a large system with energy E, and we will look for the probability for each level to be occupied. Contents of the presentation • • • • • • • • • Internal energy U Heat, Work, and Entropy Specific heat Entropy and statistics Boltzmann entropy Entropy and ideal gas Boltzmann distribution Specific heat revisited Conclusions Entropy of ideal gas But first something different. Let us calculate the S for a vessel filled with ideal monoatomic gas. All energy is kinetic energy. Counting microstates is no option. We know dS = dQ/T = dE/T , thus T = dE/dS , with E = (3/2) . N.kB.T . This gives dS = (3/2) . N.kB. dE/E . Integrating this gives: S = (3/2) . N.kB. log(E) + const. With const = log(a), and S = kB. log(Ω), we have log(Ω) = (3/2).N.log(E) +log(a) = log (a.E (3/2).N) , or: Ω = a.E (3/2).N Entropy of ideal gas We have seen: Ω = a.E (3/2).N For small E, however, in particular E<1, the probability seems to go to zero for large N. So, we have to think on the constant a. We know that at T=0, the Ω has to be =1. All particles are in the ground state (or zero point energy) E0, so a.E0 (3/2).N = 1, or: a = E0 - (3/2).N , or Ω = (E/E0) (3/2).N For simplicity we will continue with writing “a” in our formulas. But we can ask for a good guess for E0. On the next sheet we calculate this energy for helium gas. Entropy of ideal gas Quantum physics learns: for an ideal mono-atomic gas holds S = NkB [ log (nQ/n) + 5/2 ] , with n the density, nQ = (2π mkBT / h2)3/2 with m the mass of the particles. Calculating the S for helium at 1 atm (105 Pa) and 100K gives S = 8.314 [ log (667) + 5/2 ] ≈ 75 J/K. That means: S = kB log [ (E/E0) (3/2).N ] ≈ 12.5 log (E/E0). => log (E/E0) ≈ 6 , and (E/E0) ≈ 400 . => E0 / (N kB) ≈ 0.37 K . We find for the ground state an energy roughly equivalent to 0.4 kelvin, which is anyhow the correct order of magnitude. Below 1 K we are in the quantum regime. Entropy of ideal gas If we consider two vessels with ideal gas, and if we keep the valve open, equilibrium will be attained. That means Ω = Ω1 * Ω2 has a maximal value: dΩ /dE1 = Ω1 (dΩ2 / dE1 ) + Ω2 (d Ω1 / dE1 ) = 0 . Now dE1 = - dE2 (!!) and thus Ω1 . (d Ω2 / dE2 ) = Ω2 . (d Ω1 / dE1 ) . With Ω = a.E (3/2).N , and thus dΩ/dE = a.(3N/2).E ((3/2).N) -1 , it follows (after some simplification) that E1 / N1 = E2 / N2 , or T1 = T2 !! Contents of the presentation • • • • • • • • • Internal energy U Heat, Work, and Entropy Specific heat Entropy and statistics Boltzmann entropy Entropy and ideal gas Boltzmann distribution Specific heat revisited Conclusions Boltzmann distribution Without exact counting, we can say that for a large system holds: Ω = Ω(E). We now start with system 1 filled at total energy E, and system 2 empty. We now allow one particle going from 1 to level εi in 2 and we define the probability for attaining just level i as Pi . This Pi can be written as being proportional to the total Ω, being Ω1.Ω2 . Now Ω1 = Ω1(E – εi), but Ω2 = 1 !! Thus: Pi = Ω1(E – εi) / ΣkΩ1(E – εk) . Boltzmann distribution Thus: Pi = Ω1(E – εi) / ΣkΩ1(E – εk) . Now Ω is a steep function of E, but log(Ω) is smooth. That means we can approximate log(Ω1(E – εi)) in a first order Taylor expansion: log(Ω1(E – εi)) ≈ log(Ω1(E)) – εi .d(log Ω1(E))/dE . ⇒ Ω1(E – εi) ≈ exp {log(Ω1(E)) – εi .d(log Ω1(E))/dE} = Ω1(E) . exp(– β.εi ) , with β = d(log Ω1(E))/dE . The formula above now reduces to: Pi = Ω1(E).exp( – β.εi) / Σk { Ω1(E).exp(– β.εk) } = exp( – β.εi) / Σk { exp(– β.εi ) } Boltzmann distribution We have : Pi = Ω1(E).exp( – β.εi) / Σk { Ω1(E).exp(– β.εk) } = exp( – β.εi) / Σk { exp(– β.εk ) } We can calculate β for a mono-atomic ideal gas where Ω = a.E (3/2).N and E = (3/2).N.kB.T : β = d(log Ω1(E)) /dE = d( log(a)+ (3/2).N.log(E)) /dE = (3/2).N. (1/E) = (3/2).N. / (3/2).N. kB.T . Conclusion: β = 1/kB.T This also shows to hold for all other systems. Boltzmann distribution We have now: Pi = exp( – εi / kBT ) / Σk exp(–εk / kBT ) . If we have two levels, i and j, at energy difference Δε, the relative occupation number follows from (NB: the denominator is the same for both states) Pi /Pj = exp(– εi /kBT ) /exp(– εj /kBT ) = exp(–(εi –εj )/kBT ). => Pi / Pj = exp( – Δε / kBT ) And in this form, in most cases, the “Boltzmann factor” will occur in physics and chemistry. Contents of the presentation • • • • • • • • • Internal energy U Heat, Work, and Entropy Specific heat Entropy and statistics Boltzmann entropy Entropy and ideal gas Boltzmann distribution Specific heat revisited Conclusions Specific heat revisited We will learn how to calculate the specific heat of a simple two-level system, energy Now the specific heat C : difference Δε . C = dE /dT = We know now that 2 e+δ/T / [1+e+δ/T ]2 . R(δ/T) n(+) / n(-) = exp( – Δε / kBT) , We see that at low T the and with n(-) = N – n(+) follows exponentials dominate, with δ = Δε / kB and R = N kB : where at high T they are E = n(+) Δε = about one, leading to the R δ e-δ/T / [1+ e-δ/T ] or previously mentioned behaviour. E = R δ / [1+ e +δ/T ] . Specific heat revisited For low T we have, thus: C/T ≈ R (δ2/T3) e - δ/T , where the exponential dominates at low T. At high T we have C/T ≈ (R/4). ( δ2/T3 ) . Question: what is the entropy of the system (thus the red shaded area) for T very high ?? Remember what is the equilibrium at high T, and how we calculate the entropy …. Specific heat revisited At high T, sufficient energy is available for each particle to make a free choice between + and – , leading to Ω = 2N possible configurations. For S follows S = kB log (Ω) = kB log (2N) = = N.kB log (2) = R log 2 . And 2 is exactly the number of levels… For m levels, degeneracy one, we will find at sufficiently high T S = R log m . In case the degeneracy is >1 the formulas become some more intricate, but the principle doesn’t change. Contents of the presentation • • • • • • • • • Internal energy U Heat, Work, and Entropy Specific heat Entropy and statistics Boltzmann entropy Entropy and ideal gas Boltzmann distribution Specific heat revisited Conclusions Conclusions • Entropy is controlled by statistics. • Equilibrium is nothing more than the most frequently occurring distribution. • Equilibrium can be calculated by maximizing probability, and thus entropy. • The “Boltzmann factor” can in special cases directly be derived from the maximum in the probability. • However, the factor exp(-Δε/kBT) has a universal validity. Acknowledgements • A considerable part of this course is based on a first year tutorial given by professor dr. D. Frenkel (now at Cambridge University in the UK) during the last few years in Amsterdam. • Many others have contributed during many discussions on entropy, thereby giving me a deeper understanding of the topic. Thank you all. End of the presentation. Thank you for your attention. Stirling approximation N=100 N= 500 N= 1000 10^4 10^5 10^6 10^7 log(N!) = 363.74 2611.3 5912 82109 1051299 12815518 151180965 NlogN-N = 360.52 2607.3 5908 82103 1051292 12815510 151180956