1

Suppose $$ f(x)= a_0+a_1x^{-1}+\mathcal O(x^{-2}),\quad x\to\infty $$ and $$ g(y)= b_0+b_1y^{-1}+\mathcal O(y^{-2}),\quad y\to\infty. $$ How would we quote the error term of the product $f(x)g(y)$ for large $x$ and $y$?

Performing term-by-term multiplication one has $$ \begin{align} f(x)g(y) &= (a_0+a_1x^{-1}+\mathcal O(x^{-2}))(b_0+b_1y^{-1}+\mathcal O(y^{-2}))\\ &= a_0b_0+a_0b_1y^{-1} +a_1b_0x^{-1}+a_1b_1x^{-1}y^{-1} +\underbrace{\mathcal O(x^{-2})+\mathcal O(y^{-2})+\mathcal O(x^{-1}y^{-2}) +\mathcal O(x^{-2}y^{-1})+\mathcal O(x^{-2}y^{-2})}_{\mathcal O(???)}.\\ \end{align} $$ If we suppose $x=y$ the answer is obvious: $$ f(x)g(x) = a_0b_0+(a_0b_1+a_1b_0)x^{-1}+\mathcal O(x^{-2}). $$ This shows that the product is $\mathcal O(x^{-2})$.

I did have (the almost certainly incorrect) thought to define polar coordinates $x=r\cos\theta$ and $y=r\sin\theta$ so that we consider the expansion $r\to\infty$: $$ \begin{align} f(x)g(y) &= a_0b_0+a_1b_0x^{-1}+a_0b_1y^{-1}+\mathcal O(r^{-2})\\ &=a_0b_0+a_1b_0x^{-1}+a_0b_1y^{-1}+\mathcal O((x^2+y^2)^{-1}), \end{align} $$ since $r^2=x^2+y^2$. But again...this seems almost certainly incorrect.


EDIT:

In this answer I was able to find a definition for multivariable big-$\mathcal O$ that may be helpful here:

We say $u(x,y)=\mathcal O(v(x,y))$ if $\exists c>0$ such that $$ \lim_{\max\{x,y\}\to\infty}(cv(x,y)-u(x,y))\geq 0. $$

Another related question can be found here.

  • I do think you can simply write the error term as $O(x^{-2}+y^{-2})$ – JetfiRex Jan 25 '22 at 15:38
  • if they are both sufficiently large and you are only interested in this large limit behavior, then there is little advantage gained by precisely defining them. Just say the error term past the linear has quadratic decay. – dezdichado Jan 25 '22 at 16:10

0 Answers0