Boltzmann Distribution Derivation Question

4 replies · 2K views · last reply
kidsasd987
Messages
142
Reaction score
4
Hello, I have a question about Boltzmann Distribution.

I wonder why partial N of Nj is 1 and partial U of Nj=Ej. because N is constant, partial N of Nj has to be 0 and Partial Nj of U has to be 0 as well.

They are constants so, to make sense of the equation, alpha and beta have to be 0 but this is still absurd.

Can anyone tell me why this is true?
 
Attachments
  • Slide titled Maxwell-Boltzmann Distribution showing derivative of ln Q with respect to Nj set to zero, then the simplified result ln gj - ln Nj + alpha - beta Ej = 0, labeled ECE6451-14.
    스크린샷 2016-10-17 오후 3.05.58.png
    74.7 KB · Views: 592
  • Lecture slide 13 of 28 deriving the Maxwell-Boltzmann distribution: Q(N1..Nn) equals the factorial N! over the product of Ni!, times the product of gi to the Ni; then ln Q expansion, Stirling's approximation, and Lagrange multiplier condition with partial derivatives set to zero.
    스크린샷 2016-10-17 오후 3.05.47.png
    75 KB · Views: 533
  • Lecture slide titled Maxwell-Boltzmann Distribution showing the total multiplicity function Q(N1,...,Nn) = N! divided by the product of Ni!, times the product of gi^Ni, and two constraints: sum of Ni = N and sum of Ei Ni = U.
    스크린샷 2016-10-17 오후 3.05.43.png
    77.3 KB · Views: 478
  • Typeset equations: with physical constraints; 1) phi = sum over i=1 to n of N_i = N, total number of particles, conservation of mass; 2) psi = sum over i=1 to n of E_i N_i = U, conservation of energy; then d(phi)/dN_j - beta d(psi)/dN_j = alpha - beta E_j, but this implies dN/dN_j = 1 = 0, dU/dN_j = 1 = 0??
    Quest.png
    11.7 KB · Views: 1,078
kidsasd987 said:
Hello, I have a question about Boltzmann Distribution.

I wonder why partial N of Nj is 1 and partial U of Nj=Ej. because N is constant, partial N of Nj has to be 0 and Partial Nj of U has to be 0 as well.
With partial N of Nj you seem to mean ##\partial N\over \partial N_j##, right ?

Well, if ##\ \ A = x+y+z\ \ ## then what is ##\ \ {\partial A\over \partial x}\ \ ## according to you ?
 
BvU said:
With partial N of Nj you seem to mean ##\partial N\over \partial N_j##, right ?

Well, if ##\ \ A = x+y+z\ \ ## then what is ##\ \ {\partial A\over \partial x}\ \ ## according to you ?

if A is const, isn't it 0 where x is 0? or y,z is depedent on x where partial_x (y+z)=-1.
N is not a function. it is a constant, according to the second and third screen shot. on the second screen shot, it says we are multiplying constants by zero.Well I thought, it would make sense if N is a function because we are dealing with some unknown N number of particles. then, partial_Nj (N)=1. but since the slide says Partial_Nj(N) is 0, so.. I got lost.
 
Last edited:
kidsasd987 said:
N is not a function. it is a constant
N is ##\sum N_j## and as such has a partial derivative with respect to ##N_j##. Look up the definition of partial derivative. That doesn't change because a constraint (##N## is constant) is imposed.
kidsasd987 said:
on the second screen shot, it says we are multiplying constants by zero.
That would be the sheet with "13/28" and it's nonsense. (I understand your question a lot better now :smile:)

If you look up the method of Lagrange multipliers perhaps it'll become a bit clearer, in the sense that at an extremum a necessary condition is that the gradients are linearly dependent. Hence these Lagrange multipliers ##\alpha## and ##\beta##.
It certainly isn't true that all ##\ \ {\partial \phi\over \partial N_j} \ \ ## and ##\ \ {\partial \psi\over \partial N_j} \ \ ## are zero.
 
BvU said:
N is ##\sum N_j## and as such has a partial derivative with respect to ##N_j##. Look up the definition of partial derivative. That doesn't change because a constraint (##N## is constant) is imposed.
That would be the sheet with "13/28" and it's nonsense. (I understand your question a lot better now :smile:)

If you look up the method of Lagrange multipliers perhaps it'll become a bit clearer, in the sense that at an extremum a necessary condition is that the gradients are linearly dependent. Hence these Lagrange multipliers ##\alpha## and ##\beta##.
It certainly isn't true that all ##\ \ {\partial \phi\over \partial N_j} \ \ ## and ##\ \ {\partial \psi\over \partial N_j} \ \ ## are zero.

Thanks a lot. I spent quite a lot of time on figuring out what went wrong in the note and what I did wrong.
Thank you!