Finding coefficients

This has to be true for all t, no? So that means you can let t be anything you want. If the coefficients are constants (not functions of t), there should obviously be no dependence on t. 0 comes to mind, but upon inspection, doesn't seem that useful. 1, however... This is assuming b_n = constant. If that's not the case, this doesn't work so well. Considering the language they use is "coefficient", my guess is that it's only a function of n.

 

It will be messy, but in principle it is do-able: just expand ##(4a-t)^n## using the binomial expansion, so that
[tex] \frac{1}{n!} \left( \frac{1}{3}t (4a-t) \right)^n = \sum_{k=0}^n {n \choose k} (-1)^k \frac{(4a)^{n-k}}{n!\: 3^n} t^{n+k} = \sum_{k=0}^n c_{n,k} x^{n+k} .[/tex]
For given ##N## the coefficient of ##x^N## is given by summing all the ##c_{n,k}## for which ##n+k=N##.