Tuesday, 12 May 2015

combinatorics - Algebraic proof of a binomial sum identity.



I came across this identity when working with energy partitions of Einstein solids. I have a combinatorial proof, but I'm wondering if there exists an algebraic proof.
Nq=0(m+q1q)(n+Nq1Nq)=(m+n+N1N)

I've tried induction, but Pascal's Identity cannot simultaneously reduce the top and bottom argument for an inductive proof.



For those interested, a combinatorial proof of the identity can be given as follows: Consider the ways of distributing N quanta of energy to a system of n+m oscillators (where each oscillator can have any number of quanta). This is equivalent to the question of asking how many ways of putting N objects into n+m boxes. From the traditional stars and bars method, the total is given by
(m+n+N1N)
which is the right-hand side. Alternatively, consider partitioning the units of energy between the first m and the last n oscillators. Give q units of energy to the first m oscillators. Then there remains Nq units of energy for the latter n. Together, the number of states for this particular partition is
(m+q1q)(n+Nq1Nq)
Summing over all partitions gives the left-hand side.



Thanks for your time.


Answer




Let fn(z)=q=0(n+q1q)zq



Claim: This is (1z)n.



Then fn(z)fm(z)=fn+m(z). But the left hand side of your formula is the coefficient of zN in fn(z)fm(z), and the right hand side is the coefficient of fn+m(z).



The proof of the claim is the generalized binomial expansion, and uses the fact that (nq)=(1)q(n+q1q)



Alternatively, you can rewrite your statement as:




Nq=0(mq)(nNq)=((m+n)q)


No comments:

Post a Comment

real analysis - How to find limhrightarrow0fracsin(ha)h

How to find lim without lhopital rule? I know when I use lhopital I easy get $$ \lim_{h\rightarrow 0}...