Polynomial expansion of a linear transformation (alternative proof)

102 Views Asked by At

This is the problem 35.3 from Halmos's 'Finite-Dimensional Vector Spaces'.

Let $\mathscr{P}_n$ be the set of all polynomials of degree less or equal to $n-1$, having complex coefficients (together with the identically zero polynomial, $\mathscr{P}_n$ forms a vector space).

Let $A$ be a linear transformation on $\mathscr{P}_n$, such that $Ax(t) = x(t+1)$ for every $x \in \mathscr{P}_n$. Prove that if $D$ is the differential operator, then \begin{gather} \tag{$\star$} 1+\frac{D}{1!}+\frac{D^2}{2!}+\cdots+\frac{D^{n-1}}{(n-1)!}=A \label{eq} \end{gather}

Proof

The proof follows by induction. The base case for $n=1$ is straightforward to verify. Suppose that \eqref{eq} holds for $n$. Then, for $n+1$, observe that any polynomial $x$ in $\mathscr{P}_{n+1}$ can be written as $x=y+\alpha_n t^n$, with $y \in \mathscr{P}_n$. Then, \begin{align} Ax=A\big(y+\alpha_n t^n\big)&=y(t+1)+A(\alpha_nt^n)\\ &=y(t+1)+\alpha_n\big(t^n+\frac{n!}{1!(n-1)!}t^{n-1}+\frac{n!}{2!(n-2)!}t^{n-2}+\cdots+1\big)\\ &=y(t+1)+\alpha_n(t+1)^n\\ &=x(t+1) \end{align}

I) Is the proof correct?

II) I would be interested in an alternative -more intuitive- proof for the above problem. For example, the polynomial expansion of the operator $A$ resembles the one of the exponential function. Is there anything in common there? What if we had a more general translation, like $x(t+c)$? Again, I can see what's happening when I carry out the computations, but I wonder if there is a more intuitive proof, unveiling something more interesting.

Thanks in advance!

1

There are 1 best solutions below

1
On BEST ANSWER

It is perhaps slightly more intuitive to use the fact that these are linear operators and so we only need prove identities for individual powers of $t$.

Consider the infinite series for $e^D$. Acting on $t^k$ it is immediate that this produces $(t+1)^k$.

Re. the question about $x(t+c)$, the same argument proves this transform is $e^{cD}$.

Hope this helps.