Fall 2001 Problem Set 5 Solutions - McQuarrie Problems 3.20 MIT ...

131 downloads 13191 Views 279KB Size Report
Problem Set 5. Solutions - McQuarrie Problems. 3.20 MIT. Dr. Anton Van Der Ven . Problem 3-4. Fall 2003. We have to derive the thermodynamic properties of an ...
Problem Set 5 Solutions - McQuarrie Problems 3.20 MIT Dr. Anton Van Der Ven Fall 2001 Fall 2003 Problem 3-4 We have to derive the thermodynamic properties of an ideal monatomic gas from the following:  = eq  3 2mkT 2   = e and q = V h2  is the partition function for the grand canonical ensemble, where T, V,  are fixed. The characteristic potential for the grand canonical ensemble is the ’’grand canonical potential’’  = E  T S  N = pV

(since E = T S  pV + N )

d = SdT  pdV  N d

The thermodynamic properties for the grand canonical ensemble are:   ∂ S= ∂T N,V   ∂ p= ∂V T ,   ∂ N = ∂ T ,V The grand canonical potential is related to  according to  = kT ln  but  = e

q

so,



 = kT  q = kT 

2mkT h2

 32

V e



(see table 3-1)

=  (kT )

5 2



2m h2

 32

V  exp

   kT

Starting with N, N =



∂ ∂

 T ,V

=

 2mkT  32 h2

V  exp



 kT



and p p=



∂ ∂V

 T ,

= kT

 2mkT  32 h2

exp



 kT



Putting those together we can get the ideal gas equation of state, namely N p = kT V Now S,    3  3   3 5 5 2m 2 2m 2  ∂  2 = k  (kT ) 2 + (kT ) S= V e V e 2 ∂T N,V h2 h2 kT 2 1

((3-8-1))

3  3 2mkT 2  2mkT 2  Ve  V e h2 T h2     3 5 2mkT 2   S= k Ve 2 T h2

N   5  S= k N 2 T

5 S = k 2



((3-8-2))

But from (3-8-1) we can get  32 2mkT e = NV h2  3  2mkT 2 V  = kT  ln h2 N 1





Putting this into (3-8-2) we get

    3   3   5 5 2mkT 2 V 2mkT 2 V S= k + k ln N = N k ln e 2 + ln 2 N h2 N h2 S = N k ln



2mkT h2

 32

V N

5

 e2



This is the same expression as that obtained in the canonical ensemble (see Chapter 5). This is due to the equivalence of ensembles when N is very large. Problem 3-10 We are dealing with the isothermal-isobaric ensemble this time, with the partition function for an ideal monatomic gas given to us in the problem as  N 3 5 (2m) 2 (kT ) 2 = ph3  The isothermal-isobaric is for fixed (N, T, P ).  The characteristic potential for this ensemble is the Gibbs free energy G = E  T S + pV dG = SdT + V dP + dN  The thermodynamic properties are:



 ∂G ∂T p,N   ∂G V = ∂p T ,N   ∂G = ∂N p,T

S=

 G is related to the isothermal-isobaric partition function  according to (see Table 3-1):   3 5 (2m) 2 (kT ) 2 G = kT ln  = N kT ln ph3 2

  3 5 3 2 G = N kT ln (2m) + ln (kT )  ln h  ln p 2 Starting with V, V =



∂G ∂p

 T ,N

=

NkT p

which is the ideal gas equation of state. Now for  :   3 5   ∂G (2m) 2 (kT ) 2 = = kT ln ∂N p,T ph3   3 2mkT 2  =kT ln kT +kT ln p h2

o

And now for S,

 3 5 (2m) 2 (kT ) 2 5 S= = N k ln + Nk 3 ph 2 p,N   3  5 2mkT 2 V S = N k ln + N k ln e 2 2 h N 

∂G ∂T





S = N k ln



2mkT h2

 32

V N

5

 e2



These are the same expressions as obtained in the grand canonical ensemble (Problem 3-8) and in the canonical ensemble (see chapter 5). This is due to the equivalence of ensembles in the thermodynamic limit, i.e N is very large such that fluctuations are negligible. Problem 3-12 When looking at fluctuations, we derived

  2  E E 1 P (E) ≈ √ exp  2kT 2 Cv 2kT 2 Cv

For an ideal monatomic gas (derived explicitly in chapter 5) 3 E = N kT 2 ∂E 3 Cv = = Nk ∂T 2 We are now asked what is the probability that the N -particle system will sample an energy that differs by 104 % from the average energy, E = 32 N kT ? (We can let N = NA = 6.022  1023M )  4 

 3 10 E = 106 E = 106 N kT E = 100 2  6 2 9 2 2 2 2 10 N k T (E) 3  6 2 4 = = 10 N 3 2 2 4 2kT Cv 2kT  2 N k So now we can go back to P (E ),

  1 3  6 2 √ N P (E) ≈ exp  10 4 kT 3N 3

  1 √ exp 4.5  1011

kT 3N This is an extremely small probability which validates our earlier assumptions. P (E) ≈

Problem 3-18 Derive an expression for the fluctuation of the pressure in the canonical ensemble..

We know the pressure in state i is

  ∂Ei

pi =  ∂V and by definition     ∂Ei  i

 ∂V exp E kT   p = i  i

exp E kT i

2

Fluctuation is defined as  p2 = (p  p) = p2  p2 . Using the methods developed in class:

Step 1: Multiply both sides by the partition function

   Ei

pQ = pi exp kT i

Step 2: Get derivative with respect to mechanical variable’s conjugate.

   Ei

∂ ∂ pi exp (pQ) = ∂V ∂V kT i

   ∂Q ∂p ∂ Ei

p pi exp +Q = ∂V ∂V ∂V kT i



      ∂ ∂p ∂ Ei Ei

p pi exp exp +Q = ∂V kT ∂V ∂V kT i i

                 ∂pi  ∂Ei ∂p ∂Ei Ei Ei Ei 1 1

p pi exp     exp +Q = exp + kT ∂V kT ∂V ∂V kT kT ∂V kT i i i

Step 3: Divide through by the partition function        ∂E  1   E       ∂pi i i i i i pi exp E p      ∂E  exp E exp ∂p ∂V kT kT ∂V kT ∂V kT Q ∂V i

i i + = + Q Q Q Q

  2 p ∂p ∂pi p2

+ = + kT ∂V ∂V kT Rearranging a little...     ∂p ∂p i  p2 = p2  p2 = kT  ∂V ∂V 

∂pi ∂V

1 kT





has no immediate macroscopic interpretation, it must be calculate in any specific case and depends  2  i on the particular spectrum of ∂∂VE2i . This conclusion holds for all generalized forces in the form Ai = ∂E ∂a , where a is an extensive displacement conjugate to A. Hence, we cannot make an unqualified assertion that fluctuations in all kinds of external forces will be small. Note:

4

Compare this with fluctuations in extensive quantities such as E, H, or N which can be expressed in terms of thermodynamic response variables suchs as heat capacities or compressibilities. Specific calculations of the fluctuations in p of a perfect gas by Fowler is estimated as (p  p2 ) ≈ 5  1012 p2 for a cubic centimeter of gas under standard conditions. This is approximately molecules in the gas.

1 2 n3

, where n is the number of

(Source: The Principles of Statistical Mechanics, Richard C Tolman, Oxford University Press, first edition 1938)

Problem 3-24 2

Show that H 2  H = kT 2 Cp in an N, p, T ensemble. N, P, T fixed means we are working in the isothermal-isobaric ensemble..  The partition function in this ensemble is  = eEj pV V

j

Where the Ej ’s are the energies of the system when it has volume V. We also remember that H = E + pV. Using the methods developed in class: Step 1: Multiply both sides by the partition function  (Ej + pV ) eEj pV H = V,j

Step 2: Get the temperature derivative at constant (N, P ) (The conjugate variable to H in this case)      1  1 ∂H 2 +H 2 (Ej + pV ) eEj pV  = (Ej + pV ) eEj pV ∂T N,P kT kT 2 V,j

V,j

Step 3: Divide through by the partition function      Ej pV 2 Ej pV pV e (E + ) j V,j ∂H 1 1 V,j (Ej + pV ) e + H = 2 ∂T N,P kT 2  

kT

H2

H

or 2

H 2  H = kT 2  but we know

∂H ∂T

 N,P



∂H ∂T



= Cp . So, 2

H 2  H = kT 2 Cp 5

N,P

Problem 3-26  Show that

∂ ∂N



2

V,T

V = N 2



∂p ∂V

 .. N,T

From Gibbs-Duhem we have SdT  V dp + N d = 0 At constant T , we then get



We can use the chain rule and get:



d dp



d dp

 =

 =

d dN

V N



 V,T

dN dp

 V,T

Using partial derivative manipulation       dN dN dV = dp V,T dV P,T dp N,T  dV  V But for a single component system dN p,T = the molar volume = N and we get     dN N dV = dp V,T V dp N,T Putting this all together we get



d dN

 V,T

V2 = 2 N



dp dV

 N,T

Problem 4-2 Show that

6N V



h2 12mkT

 32

given in table 4-1 is very large for electrons in metals at T = 300K.

 Take Na-metal having the following properties - stable in the bcc crystal structure with lattice constant a = 4.23  1010m - two Na atoms per bcc unit cell - number of valance electrons per Na atom = 1 - valance electrons in Na can be considered nearly free So we can get the following values to substitute into the original equations: N 2 = 3 V (4.23  1010 ) h = 6.6262  1034 J  s J k = 1.3807  1023 K me = 9.1095  1031 kg Putting those all together we get 6N V



h2 12mkT

 32

= 1524 >> 1

Therefore Boltzmann statistics cannot be applied to electrons in metals. Must use Fermi-Dirac statistics. 6

Problem 4-6 S=

∞   

xn1 1 xn2 2

N=0 {nj }

with n1 and n2 = 0, 1, and 2. The Let’s consider

  {nj }

 



means

{nj }

n1 n2

with the restriction that n1 + n2 = N.

x1n1 xn2 2 for several values of N.

N = 0 → possible combinations of n1 and n2 are 0 and 0.   =⇒ =1 {nj }

N = 1 → possible combinations of n1 and n2 are n1 1 0  

=⇒

n2 0 1

N 1 1

x1n1 x2n2 = x1 + x2

{nj }

N = 2 → possible combinations of n1 and n2 are n1 0 1 2 =⇒

 

n2 2 1 0

N 2

2

2

xn1 1 xn2 2 = x22 + x1 x2 + x12

{nj }

N = 3 → possible combinations of n1 and n2 are n1 1 2 =⇒

 

n2 2 1

N 3 3

x1n1 x2n2 = x1 x22 + x21 x2

{nj }

N = 4 → possible combinations of n1 and n2 are n1 2 =⇒

 

n2 2

N 4

x1n1 x2n2 = x21 x22

{nj }

For N > 4 →

  {nj }

x1n1 x2n2 = 0 because n1 + n2 � 4

Putting everything together we get   x1n1 x2n2 = 1 + x1 + x2 + x22 + x1 x2 + x21 + x1 x22 + x12 x2 + x21 x22 N {nj }

7

Now lets consider S=

2    1 + xk + xk2 k=1

   S = 1 + x1 + x21 1 + x2 + x22 = 1 + x1 + x2 + x22 + x1 x2 + x21 + x1 x22 + x21 x2 + x21 x22 This expression contains the exact same terms as that obtained with



  N

{nj }

xn1 1 x2n2 .

Problem 4-8 We need to show this ( remember the upper (lower) signs is for Fermi-Dirac (Bose-Einstein) : S = k



[nj ln nj  (1 nj ) ln (1 nj )]

j

Start with the partition function and go from there. Becuase of the equivalence of ensembles in the thermodynamic limit, we can calculate the entropy using the ensemble that offers the most mathematical convenience. For Fermions   ∂ ln    or Bosons, this is the grand canonical ensemble. S = k ln  + kT ∂T V, =

 j

ln  = 

1 + e



 1

� εj kT

 � εj  ln 1 + e kT

j

∂ ln  = ∂T



� εj e kT



∂T

1  e

j

� εj kT

=





εj  kT 2



exp



(εj ) kT



� εj

j

   � εj  S=k  ln 1  e kT +  j

1  e kT   ε  eεj jkT  1  eεj



To make this a little easier to manipulate we can write this shorthand by making the following substitutions. Let: u = 1  eεj and v = eεj u = 1  v and u v = 1 v nj = u So we now have S=k



{ ln u  nj ln v} = k



j

S=k



{ ln u  nj ln nj  nj ln u}

j

{(1  nj ) ln u  nj ln nj } = k

j



 (1  nj ) ln

j

Remembering that u v = 1 S =k

 j

S = k

   1  nj ln nj u

{ (1  nj ) ln (1 nj )  nj ln nj }

 j

{nj ln nj  (1 nj ) ln (1 nj )} 8

Problem 4-12 N-distinguishable independent particles, each of which can be in state +εo or -εo . N+ = number of particles with energy +εo

N = number of particles with energy εo

with N+ + N = N

The total energy is given as: E = N+ εo  Nεo = 2N+ εo  N εo   We now have to evaluate the partition function Q by summing exp E over levels and compare it to the result kT Q = q N ..  We know    Ej Q= exp kT j where j labels a state in which the system can reside.  Instead of summing over the states that the system can be in, we can also sum over the possible energy levels, making sure we take account of the degeneracy of each energy level, i.e.    E Q= (E) exp kT E

where (E) is the number of states with energy E.

 For this system we have already stated the allowed energy levels are

E = 2N+ εo  N εo where N+ can vary from 0 → N.  For each allowed energy level E, there are (E) possible states compatible with this energy. Since the N particles are distinguishable, N! N! (E) = = N+ (N  N+ )! N+ !N! This represents the number of ways that N+ particles out of N can be in the + state.

 So now we can write

  N! Q= (E) exp (E ) =⇒ exp [ (2N+ εo  N εo )] Since E is uniquely a function of N+ N+ (N  N+ )! E

N+

This can be rewritten in a nicer form if we remember the binomial expansion n  n!  x n 1 (1 + x)n = n !(n  n1 )! n =0 1 1

So we have

 Q = eNεo 

 N+

 2εo

if we let x = e or

  2ε N+ N! o  e N+ (N  N+ )!

and rewrite using the binomial expansion we have N  Q = eNεo 1 + e2εo  N Q = eεo + eεo

Now we need to compare this to Q = q N , where q is the single particle partition function defined as 9

q=



eεl

l

where l labels the single particle states. In this example there are two single particle states, with energy -εo and +εo . =⇒ q = eεo + eεo and  N Q = q N = eεo + eεo The last part of this problem asks us to calculate and plot the heat capacity for this system. We know Cv = and E is given as   ∂ ln eεo + eεo ln Q 2 = kT N E = kT ∂T ∂T   εo εo εo εo εo  2e e  eεo 2e N ε E = kT 2 N kT εo kT = o e + eεo eεo + eεo 2∂

E = N εo tanh (εo ) and Cv is then Cv = Plotting

Cv Nk

 vs.

kT εo

ε   ε 2 ∂E o o 2 = Nk sech ∂T kT 2 kT



10

∂E ∂T

Problem 5-4 Calculate the entropy of Ne at 300K and 1 atm.  The entropy of an ideal gas (eqn 5-20):



S = N k ln

2mkT h2

 32

5

V e2 N



Note this is neglecting electronic excitations (see Chapter 5). Some data: m = 3.351  1026 kg k = 1.3807  1027 Jk h = 6.6262  1034 J  s

p = 1 atm = 1. 013  105 P a Putting that together we can get V kT = = 4.0889  1026 N p  3 2mkT 2 = 8.852  1031 h2 and

  S = N k ln 4.41  107  N  S = k ln 4.41  107

where N = the number of Ne atoms. We are now asked to estimate the translational degeneracy  -From our study of fluctuation theory, we found that the fluctuation in energy of a thermodynamic system (with N very large) is exceedingly small. -Therefore, the energy of the gas is essentially always very close to E (see discussion on page 63 of McQuarrie) and we can use the expression for the entropy in the microcanonical ensemble. S = k ln   Compare this with where  is the degeneracy at fixed energy E.  N  S = k ln 4.41  107 and we get

 N  = 4.41  107

which makes sense since it should be large because N is on the order of 1023. Problem 5-9 What is the DeBroglie wavelength of Arÿ at 298K?  = Use:

h2 2mkT

m = 6.634  1026 kg k = 1.3807  1027 Jk h = 6.6262  1034 J  s 11

 12

and we get  = 1.6  1011 m  Now compare this with the inter-atomic distance The volume per Ar atom is V kT = N p V



= 4.06  10 26 m3 . The interatomic distance ∼ with p = 1.013  10 5 Pa. So N (See page 83 on the relevance of this result.)

12

 V  1 3

N

= 3.0  10 9 >> 

Problem Set 5 Additional Problems Solutions 3.20 MIT Dr. Anton Van Der Ven Fall 2001 Fall 2002

Problem 1

(a) A system of N non-interacting particles with two possible states either 0 or ε. A good rule is too assume ’’particles’’ (e.g. atoms, electrons, etc) are indistinguishable unless they are localized in a crystal or on a surface. The number of atoms in the excited state can be determined using Boltzmann statistics under the assumption that we are working at high temperature and/or low density:: Nε = N  ε where  ε is the probability an atom will be in state ε. This probability is determined using the single particle partition function and can be written as  ε  exp kT   ε =  exp ε; kT

(McQuarrie 4-14)

i

But our system can be in only two states, so the sum in the denominator can be found explicitly:  ε  exp kT  ε  ε = exp [0] + exp kT So Nε can be written as  ε  exp kT N  ε  =  ε  Nε = N  ε = N 1 + exp kT 1 + exp kT Nε =

(b) The total energy is simply U = N ε = N



N ε 1+exp[ kT

]

 i εi (McQuarrie 4-12 and 4-13)

i

U =N



 i εi = N ( 1 ε1 +  2 ε2 ) = N [(1   2 )  0 +  2  ε]

i

U = N ( 2 ε) = N U=

 ε   exp kT  ε  ε 1 + exp kT

Nε ε 1+exp[ kT

]

Note: Since many particles will occupy the same state (either 0 or ε) these particles must be Bosons. At lower temperatures we would have to use Bose-Einstein statistics (McQuarrie 4-26) which would lead to a much more complicated problem since we would have to determine the chemical potential . 1

Problem 2 The magnetization is given as M =

N 

ni o

i=1

This is basically saying that we have atoms localized in a crystal and the magnetic moment at each site can be either up or down. The problem asks us to determine the thermodynamic properties as a function of T, N, H. Let us also assume for simplicity that we can work at constant volume. Therefore, our controlling variables are T, N, V, H. We need to make the appropriate Legendre transform to the entropy. Remember the entropy can be written starting from E: E = T S  pV + HM + N rearranging to get things in terms of S and  we get S = E + pV  HM  N k

Legendre transform such that our controlling variables are V, N, T, H

S  E + HM =  (T S  E + HM ) =  = ln  k where is the characteristic potential for this ensemble with V, N, T, H constant and  is the partition function.  can be written as  = exp [ (Estate  Mstate H)] states

where we sum over all possible energy states and magnetizations Mstate. Since the particles are non-interacting, the energy at N,  = constant and H = 0 is constant. E is independent of the number or arrangement of up versus down spins. Since the absolute scale of energy is not important for thermodynamics, we can arbitrarily set the constant energy equal to zero giving us  N  N      ni o H = exp  exp [nio H ] = exp [Mstate H] = states

n1 ,n2,...n N

=



N 

n1 ,n2 ,...n N i=1

i=1

exp [ni o H ] =

N 

+1 

exp [ni o H ]

i=1 ni =1

n1 ,n2 ,...n N i=1

We can evaluate the sum since ni = 1 so,  = (exp [o H ] + exp [o H ])

N

We know the characteristic potential of an ensemble is related to the partition function for that ensemble according to  = ln  → = kT ln  Furthermore, we know from thermo that d = SdT  pdV + dN  M dH which gives us the following relationships for the properties of the system:  S

=

 

M

=

 2

∂ ∂T ∂ ∂H

 V,N,H



V,N,T



 = p =

 ∂ ∂N V,T ,H   ∂  ∂V T ,N,H

and from stat mech we have from above = kT N ln [exp [o H ] + exp [o H ]] Lets get the entropy, S

 S=

∂ ∂T

 V,N,H



S = kN ln (exp [o H ] + exp [o H ]) + kT N

 o H (exp [o H ]  exp [o H ])  kT 2 exp [o H ] + exp [o H ]

S = kN {ln (exp [o H ] + exp [o H ])  o tanh (o H )} Now for the magnetization     ∂ exp [o H ]  exp [o H ] M = = kT N o = N o tanh (o H ) ∂H V,N,T exp [o H ] + exp [o H ] M = N o tanh (o H ) The energy E = E  T S  HM E = + T S + HM = 0 However if you define a quantity called the internal magnetic energy (which is a quantity analogous to the enthalpy in the T, p, N ensemble) EH = E  HM you can get EH = N o H tanh (o H ) The last part of this problem asks you to determine the behavior of the energy and entropy as T → 0. EH (T → 0) = N o H lim S = lim kN {ln (exp [o H ] + exp [o H ])  o tanh (o H )} = 0

→∞

→∞

S (T → 0) = 0 is in accordance with the third law of thermodynamics. Problem 3 (a) M =

N 

o ni

i=1

(see solution to Problem 2 - Method 1)     exp [o H ]  exp [o H ] ∂ M = = N o tanh (o H ) = kT N o ∂H V,N,T exp [o H ] + exp [o H ] M = N o tanh (o H ) 3

(b) The partition function for the N, V, H, T fixed ensemble is 

=

exp [ (Es  Ms H )]

s

where s = states

We want to determine the fluctuations in the extensive quantity M. Use the 3-step procedure developed in class.

Step 1: Multiply both sides by the partition function



M =

Ms exp [ (Es  Ms H )]

states

Because Es is always constant (see comment in Problem 2) we can therefore arbitrarily set it to zero and write

N 

M =

n1 ,n2,...n N

N 





exp o H

o ni

N 

 ni

i=1

i=1

Step 2: Get derivative with respect to mechanical variable’s conjugate.

∂M +M ∂H



2 

N  N N    ∂M ∂ +M = exp o H ni  o ni ∂H ∂H n ,n ,...n i=1 i=1 N 

n1 ,n2 ,...n N



N 



1

N

2



exp o H

o ni

i=1

N 



N 

=

ni

n1 ,n2 ,...n N

i=1



o

N  i=1

ni

2

 exp o H

N  i=1

Step 3: Divide through by the partition function

∂M 2 + M = M 2 ∂H or we can write it like this &

2

( M )

'

2

= M2  M =

1 ∂M  ∂H

with: 

∂M ∂H



) ( = N 2 1  tanh2 (H ) ,N

& ' 2 = 2 {1  1} = 0. In other (c) As  → ∞, tanh (H ) → 1. Therefore M T →0 = N  and ( M ) T →0 words, the ground state with all the spins aligned has no fluctuations. 4

 ni

Problem 4 From the first and second law we have dE = T dS  pdV + HdM + dN (Euler Form) E = T S  pV  HM + N We are working with fixed N, T, V, M in this problem. We need to make a Legendre transform because T is a controlling variable instead of E. S  E = F = ln Q k  Q= exp [Estate ] states

F = kT ln Q From the differential form of F, namely dF = SdT  pdV + HdM + dN we get that

 H=

∂F ∂M

 T ,V,N

We are working under constant magnetization so M = o

N 

ni = o (n+  n)

i=1

where n+ = number of up spins and n = number of down spins. That means that the sum in the expression of Q must be performed over only those states with fixed M (i.e. fixed n+ and n) Also the atoms in the system do not interact, meaning that the energy is independent of the number and arrangement of up/down spins and is therefore constant = Eo.   Q= exp [Eo ] =  exp [Eo ] exp [Estates] = states with magnetization M

states with magnetization M

where  is the number of states that are consistent with a magnetization M = o (n+  n) . N! n+ ! (N  n+ )! N! Q= exp [Eo ] n+ ! (N  n+ )!     N! N! F = kT ln exp [Eo ] = Eo  kT ln n+ ! (N  n+ )! n+ ! (N  n+ )! =

using Stirling’s approximation, F = Eo  kT {(N ln N  N )  n+ ln n+ + n+  (N  n+ ) ln (N  n+ ) + (N  n+ )} but we know from above that M = o (n+  n) = o (2n+  N ) =⇒ n+ =

M + o N 2o

we can substitute this back into our expression for F to get F as a function of M        M + o N M + o N o N  M o N  M F = Eo  kT N ln N  ln ln  2o 2o 2o 2o 5

 ∂F  Remembering that H = ∂M T ,V,N         1 1 1 1 ∂F M + o N o N  M  H= = kT  ln + + ln 2o 2o 2o 2o 2o 2o ∂M T ,V,N      kT M + o N o N  M ln H=  ln 2o 2o 2o   M + o N 2o H = ln o N  M and it is easy to show that M = o N tanh (o H ) which is the same result as in the other ensemble of Problems 2 and 3. This is due to the equivalence of ensembles in the thermodynamic limit.

Problem 5 (a) Grand canonical ensemble: =

∞ 



N =0

j

exp [Ej ]

Q (N, V, T ) exp [N ]

N =0



This sum extends over allstates with total # of particles = N 

∞ 

exp [N ] =

exp [Ej ] = Q(N, V, T ) =

j

qN N!

for a one component gas of non-interacting particles. We can write q in closed form as we have done before as  3 2mkT 2 q= V h2 So now  can be written as: =

∞ ∞   N   qe qN exp [N ] = N! N!

N=0

N=0

But if we let a = qe then we can write this sum in a form for which the solution is know, namely ∞  aN = ea N!

N=0

If we now substitute in for q we can write  as  = ezV

z=

with

e h2



2m 

(b) The characteristic potential for the grand canonical ensemble is pV =  = kT ln   = E  T S  N d = SdT  pdV  N d 6

 32



   ∂ ∂ (zV ) = kT = kT z p= ∂V T , ∂V     ∂ ∂z = kT N = V = zV ∂ T ,V ∂ putting those two parts together and eliminating z gives us the familiar ideal gas equation p =

kT N . V

(c) *

*

2

( ) = 2

2

( N ) N2

(because V fixed)

2

2

We need to get ( N ) = N 2  N . Follow the 3-step procedure. Step 1: Multiply both sides by the partition function N =  =

∞ 

N

N=0

q N N e N!

∞ ∞   q N N q N N = N e N e N! N! N=0

N=0

Step 2: Get derivative with respect to mechanical variable’s conjugate. 

∂N ∂

 +N 

∞  N=0

∂N ∂



q N N N e N!

 =

∞ 

N 2

N=0

 + N  N  = N 2 

Step 3: Divide through by the partition function   ∂N 2 + N = N 2 ∂   ∂N 2 2 N  N = kT ∂ 3  2  from (b) we have that N = zV with z = eh2 2m .      ∂N ∂z = V = zV ∂ ∂ 2

N 2  N = zV divide by V to get the densities z N = 2 V V 2  2 N = 2 2 N + * 2  2 1 kT = √ = pV N 2  2 =

7

q N N e N!

(d) To second order in N

 ∂ 2 ln P (N ) ∂N 2

N = N  2

assuming a Gaussian distribution around N with variance N we get  

7 P (N ) 1012  N = exp = e10 which is very unlikely. 2

P (N ) ln P (N ) = ln P (N ) +

1 2 ( N ) 2

8