Upper bound on the colength of the trace of the canonical module in dimension one (Q2166347)

For an additive semigroup of \(\mathbb{N}\) with \(0 \in H\) such that \(\mathbb{N} \setminus H\) is finite, denote by \(F(H)\) the largest integer in \(\mathbb{N} \setminus H\), \(g(H)=|\mathbb{N} \setminus H|\) and \(n(H)=|\{ h \in H \mid h < F(H) \} |\). Given a field \(K\), the numerical semigroup ring of \(H\) is defined to be the \(K\)-algebra \(R=K[H]=K[t^h \mid h \in H] \subseteq K[t]\). As proved by \textit{J. Herzog} et al. [Isr. J. Math. 233, No. 1, 133--165 (2019; Zbl 1428.13037)], if \(P \in \operatorname{Spec} (R)\), \(R_P\) is Gorenstein if and only if \(P\) does not contain the trace ideal \( \operatorname{tr}_R(\omega_R)=\sum_{f} \operatorname{Im} f\) where \(\omega_R\) is the canonical module of \(R\) and the sum is taken over all \(f \in \operatorname{Hom}_R(\omega_R, R)\). In particular, \(R\) is Gorenstein if and only if \( \operatorname{tr}_R(\omega_R) = R\). As a way to measure how far \(R\) deviates from being Gorenstein, \textit{J. Herzog} et al. [Semigroup Forum 103, No. 2, 550--566 (2021; Zbl 1494.13003)] considered the length of the \(R\)-module \(R/ \operatorname{tr}_R(\omega_R)\) and asked whether or not \[ \ell(R/ \operatorname{tr}_R(\omega_R)) \leq g(H) - n(H). \] In this paper the authors show that the question has a positive answer when \(R\) has Cohen-Macaulay type at most 3. They also show that there exist numerical semigroup rings with Cohen-Macaulay type 5 for which the answer is negative. Closely related to the above situation, if one considers the completion \(R=K[[H]]\) of \(K[H]\), \textit{T. Kobayashi} [Ill. J. Math. 64, No. 3, 349--373 (2020; Zbl 1451.13039)] asked whether or not \( \ell(R/ \operatorname{tr}_R(\omega_R)) \leq bg(R)\) where \(bg(R)\) denotes the smallest possible length \(\ell_S(R/S)\) for a one-dimensional Gorenstein ring \(S\) such that \(S \subseteq R\) is a birational extension. The authors prove the general inequality \( \ell(R/ \operatorname{tr}_R(\omega_R)) \leq 2bg(R) -1\) when \(R\) is not Gorenstein and show that for each \(l \geq 2\) there exists \(R\) with \(bg(R)=l\) such that equality holds. Consequently, the answer to Kobayashi's question is negative if \(bg(R) \geq 2\), but is positive if \(bg(R) \leq 1\).