Ler $V=P(\Bbb R)$ be the vector space of the polynomials with real coefficients, on the field of real numbers $\Bbb R$. For $i \geq 1$, let $T_i(f)=f^{(i)}$ the $i$th derivate of $f$. I have to show that for any $n \in \Bbb N$, $\{T_1, T_2,..., T_n\}$ is a linearly independent subset of $L(V)$ (the set of oprators of $V$.)

$\textbf{Attempt:}$ Suppose there exist $c_1,...,c_n \in \Bbb R$ such that $$c_1T_1 + \cdots c_n T_n = 0.$$ Then, for any $f \in P(\Bbb R)$, we have $$(c_1T_1 + \cdots c_n T_n)(f)=c_1 T_1 (f) + \cdots c_n T_n (f) = 0.$$

We must show that $c_1 = c_2 = \cdots = c_n =0.$ Any $f \in P(\Bbb R)$ is of the form

$$f=a_m x^m + \cdots + a_1 x + a_0.$$

for some $m \in \Bbb N$. Also, we have that

$$f^{(k)}=\sum_{i=k}^m \bigg( \prod_{j=0}^{k-1}(i-j) \bigg)a_i x^{i-k}.$$

If $m \geq n$, then $T_n(f)=f^{(n)}$ is a linear combination of $1, x, x^2, ..., x^{m-n}$, which is a linearly independent set, and then their coefficients are all zero. Then, we substitute this coefficients in $f^{(n-1)}$, which is a linear combination of $1, x, x^2, ..., x^{m-n}, x^{m-n+1}$ and we obtain that the coefficient of $x^{m-n+1}$ is zero. We continue with this process and we obtain $c_1=c_2=\cdots =c_n = 0$.

I think this proof is not complete or even correct because I only handled the case $m \geq n$, and I don't know what to do with the case $m<n$, and also $m-n$ could be greater than $n$ and then we'd have that the coefficients of $1, x, x^2, ..., x^{m-n}$ are more than $n$ coefficients, so I'm really confused and I don't know what else I could do. I hope you can help giving some idea to solve this problem. Thank you in advance.


A slightly simpler proof than the others is to evaluate $(c_1T_1+\cdots+c_nT_n)=0$ at $x^n$. You get the polynomial $$c_1nx^{n-1}+c_2n(n-1)x^{n-2}+\cdots+c_nn!$$ Since this is the zero polynomial, all the coefficients are zero in a single step.


A much more direct proof is to just consider a few specific polynomials, namely $x^n$ for various $n$. We see that $x' = 1$ and $x'' = 0$, so

$$(c_1 T_1 + ... + c_n T_n) x = c_1 \cdot 1 + 0 + ... + 0 = 0$$

implies that $c_1 = 0$.

Next, we have

$$(c_1 T_1 + c_2 T_2 + ... + c_n T_n) x^2 = 0 + c_2 (x^2)'' + c_3 (x^2)''' + ... + c_n (x^2)^{(n)} = 2c_2 = 0$$

So $c_2 = 0$. Now proceed similarly to show that each $c_k$ is $0$.

  • 2
    $\begingroup$ $4$ seconds. ${}{}{}$ $\endgroup$ – Pedro Tamaroff Oct 12 '13 at 1:59
  • $\begingroup$ @PedroTamaroff Ha, indeed. $\endgroup$ – user61527 Oct 12 '13 at 2:00

I think this is simpler. Suppose that $$\alpha_1T_1+\cdots+\alpha_nT_n=0$$

Then evaluate at $P(X)=X$. Since $P'=1$, $P''=0$, we get that $$\alpha_1=0$$

Now evaluate at $P(X)=X^2$. Then $P''=2$; but $P'''=0$, so...?

(Observe you have a proof that $(\Bbb R[X])^\ast$ cannot be finite dimensional!)


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.