The Novikov ring (over a ground field $k$) is defined to be $$ R:=\Big\{\sum_{i=0}^{\infty} a_i T^{\lambda_i} \mid a_i\in k , 0\le\lambda_0 < \lambda_1 < \cdots, \lim \lambda_i = +\infty \Big\} $$ Here $T$ is a formal variable (sometimes we may assume all possible values of $\lambda_i$ belongs to a discrete subset $\Gamma\subset \mathbb R$). This ring is kind of similar to $k[T]$ or $k[[T]]$. Now that it is known $k[T]$ is a UFD, is it true that $R$ is also a UFD?


I do not believe that under this definition the Novikov ring is in general a PID. Let $\Gamma=\langle 0,1,\sqrt 2\rangle$, the closure of the set under addition. This continues to be discrete (check). Consider the ideal $I$ of $R$ which is formed of elements with no constant coefficient. Define the degree operator as below. $\partial(pq)=\partial p\cdot\partial q$. Assume $I=(p)$. If $\partial p>1$, then $(p)\neq I$ by sizing considerations. But if $\partial p=1$, $T^{\sqrt 2}\notin(p)$ because that would imply the existence of an element of degree $\sqrt 2-1$, impossible under our definition.

The above $\Gamma$ is discrete: $|(a-c)+(b-d)\sqrt 2|$ can be bounded in four cases, based on whether $a-c$ is positive or negative and whether $b-d$ is positive or negative. Clearly if they're both positive or both negative the distance is at least $1+\sqrt 2$. If, say $a-c> 0$ and $b-d<0$, there are only finitely many choices for $c$, and, fixing $c$, there are only finitely many choices for $d$ which make the error less than $M$ for some large $M$, and none of them give error $0$ by the irrationality of $\sqrt 2$. Similarly for the other cases.

If $\Gamma$ is not discrete, I don't think we can say anything meaningful. Consider $\Gamma=\mathbb R^{+}$, and the ideal $I$ of elements $\sum a_iT^{\lambda_i}$ which have no constant term -- that is, whenever $\lambda_i=0$, $a_i=0$ as well. I claim that $I$ is not generated by any single element -- in fact, that it's not generated by any finite number of elements. Assume that it were -- say $I=(p_1(T),\ldots,p_n(T))$. Let $\partial p$ be $\lambda_0^P$ for each $p$ - the smallest term for which the polynomial has a nonzero coefficient. Let $m=\min\{\partial p_1,\partial p_2,\ldots,\partial p_n\}$. Then for each $q\in I$, $\partial q\ge m$, because every element is the $R$-linear combination of these, and multiplication and addition of elements in $R$ can only increase degree (there are no elements of negative degree). This contradicts $R$ being a PID, or even Noetherian. Moreover, the same argument holds for any additive subgroup which fails to have an element of smallest norm, or if there is an element of smallest norm and $m$ is a limit from above of things in $\Gamma$.

  • $\begingroup$ If $\Gamma\subset\mathbb R$ is discrete it is not necessary that $\Gamma$ is isomorphic to $\mathbb Z$ as a group, we can only construct a bijection between them. $\endgroup$ – Hang Jul 16 '18 at 0:11
  • $\begingroup$ Sorry, I made some rather strong assumptions. I'll see if I can fix them. $\endgroup$ – Ashwin Trisal Jul 16 '18 at 0:19

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.