I have a question and please I need help. First all, this is the context:

Let $p : T\rightarrow\mathbb{R}$ a polynomial of degree $\leq k$ with null measure condition onto T, that is $$\int_T p\ =\ 0,$$ where $T$ is some region in $\mathbb{R}^n$. I want to construct a basis of the subespacio where $p$ lies, namely $Q_{k,0}$.

Now suppose the I have $\{\varphi_0,\varphi_1,\ldots,\varphi_N\}$ an arbitrary basis for $P_k(T) = $ polynomial space of degree $\leq k$ onto $T$. Then, in order to obtain a basis for $Q_{k,0}$, I defined $$\psi_i\ =\ \varphi_i - \frac{1}{|T|}\int_T\varphi_i,\quad i = 0,1,\ldots,N.$$ Observe that $\{\psi_0,\psi_1,\ldots,\psi_N\}\subseteq Q_{k,0}$, but is not a basis, because they are linearly dependent.

So, my question is: how I can get a basis of $Q_{h,0}$ from the set $\{\psi_i\}$? In other words, how I know WHICH $\psi_i$ must be delete, to get a basis?

Remember, $\{\varphi_i\}$ is an arbitrary basis, because if we now that one $\varphi_i$ is constant, then the respective $\psi_i$ must be delete. The problem is, for example, for the Lagrange basis.

Thanks so much in advance for any hints.


Assuming that $\varphi_0 = c \neq 0$, i.e. that $\deg \varphi_0 = 0$, for your basis $(\varphi_i)$, the $\psi_i$ are linearly independent of you leave out $\psi_0$. By construction of the $\psi_i$, you have $$ \psi_i = \varphi_i + \lambda_i\varphi_0 \quad\text{where } \lambda_i = -\frac{1}{c|T|}\int_T \varphi_i(t) \,dt $$ Assume that for some linear combination of the $\psi_i$, i.e. for some $\mu_1,\ldots,\mu_N$, you have $$ \sum_{k=1}^N \mu_i \psi_i = \sum_{k=1}^N \mu_i\varphi_i + \varphi_0\underbrace{\sum_{k=1}^N \mu_i \lambda_i}_{=:\mu_0} = 0 \text{.} $$ Since the right-hand side is a linear combination of the $\varphi_i$ with coefficients $\mu_0,\mu_1,\ldots,\mu_N$, this implies $\mu_i = 0$ for all $i$, which shows that the $(\psi_i)_{1 \leq i \leq N}$ are linearly independent.

Thus the $(\psi_i)_{1 \leq i \leq N}$ generate a least a subspace of $Q_{k,0}$. In particular, for $N=k$, the generate a $k$-dimensional subspace. Now you just need to argue that $\dim Q_{k,0} = k$. This follows from the fact that the integral of a constant polynomial $p$ over $T$ is only zero if $p$ is identically zero, i.e. the zero polynomial.

In the general case, i.e. if nothing is known about the $\varphi_i$ except that they form a basis, you always can select $k$ linearly independent $\psi_i$, but I don't think you can describe which those are except by, well, saying "pick them linearly independent".

  • $\begingroup$ Thanks for your answer. But, is a easy way to do the case deg $\varphi_i \neq i$ ? $\endgroup$ – FASCH Apr 28 '14 at 17:48
  • $\begingroup$ @FASCH I've updated my answer to say $\varphi_0 = c$ - that's in fact the only one of the degree constraints that the rest of the answer used. Without any such constraint, I don't think there's a better characterization of the $\psi_i$ you have to pick except saying "Pick them linearly independent". For a specific basis $\varphi_i$, however, better characterizations might exist - but to explore that you'd first have to state the basis you have in mind. If you don't have a particular basis in mind, then why not just use $\varphi_i = x^i$? $\endgroup$ – fgp Apr 28 '14 at 18:56

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.