betaFeedback

WorksMathematics

Hilbert-Kunz Multiplicity and Smirnov's conjecture

Abstract

In this paper, we study Smirnov's conjecture: Let (R,m)(R,\mathfrak{m}) be a Noetherian local ring of characteristic p>0p>0 and dimension dd. Let II be an m\mathfrak{m}-primary ideal. If RR is weakly FF-regular, analytically unramified and Cohen-Macaulay, then L1(I)=e(I)−eHK(I)\displaystyle L_1(I)=e(I)-e_{\mathrm{HK}}(I) if and only if II is stable. Here, e(I)e(I) and eHK(I)e_{\mathrm{HK}}(I) denote the Hilbert-Samuel multiplicity and the Hilbert-Kunz multiplicity of RR with respect to II, respectively. Moreover, L1(I)=lim⁡q→∞e1(I[q])qd,\displaystyle L_1(I)=\lim_{q\to\infty}\frac{e_1(I^{[q]})}{q^d}, where q=peq=p^e and e1(I[q])e_1(I^{[q]}) denotes the first Hilbert coefficient of RR with respect to the qqth Frobenius power of II. We prove the conjecture when RR is either excellent and strongly FF-regular, or excellent of dimension 33.

Introduction

Let (R,m)(R,\mathfrak{m}) be a Noetherian local ring of prime characteristic p>0p>0 and dimension dd. Let ℓR(M)\ell_R(M) denote the length of an RR-module MM. For an m\mathfrak{m}-primary ideal II, the Hilbert-Kunz multiplicity is defined by

eHK(I)=lim⁡q→∞ℓR(R/I[q])qd.e_{\mathrm{HK}}(I)=\lim_{q\to\infty}\frac{\ell_R(R/I^{[q]})}{q^d}.

where q=peq=p^e and I[q]=(xq∣x∈I)I^{[q]}=(x^q\mid x\in I) denotes the qqth Frobenius power of II. The Hilbert-Samuel function of RR with respect to II is defined as the function HRI(n)=ℓR(R/In)H_R^I(n)=\ell_R(R/I^n). It is well-known that for large values of nn, this function can be expressed as a polynomial PR(n)P_R(n) of degree equal to dim⁡(R)\dim(R), known as the Hilbert-Samuel polynomial of RR with respect to II. The polynomial can be written in the form

PR(n)=e0(I)(n+d−1d)−e1(I)(n+d−2d−1)+⋯+(−1)ded(I),P_R(n)=e_0(I)\binom{n+d-1}{d}-e_1(I)\binom{n+d-2}{d-1}+\cdots+(-1)^d e_d(I),

where ei(I)e_i(I) is the iith Hilbert coefficient of RR with respect to II. We write e(I)e(I) instead of e0(I)e_0(I). In particular, for fixed q=peq=p^e, ei(I[q])e_i(I^{[q]}) is the iith Hilbert coefficient of RR with respect to I[q]I^{[q]}. Moreover, e(I[q])=e(Iq)=qde(I)e(I^{[q]})=e(I^q)=q^d e(I). From the expression of PR(n)P_R(n), we can conclude that e(I)=lim⁡n→∞ℓR(R/In)nde(I)=\lim_{n\to\infty}\frac{\ell_R(R/I^n)}{n^d}.

The two invariants e(I)e(I) and eHK(I)e_{\mathrm{HK}}(I) exhibit several common properties. For example, if RR is a formally unmixed ring, then e(m)=1e(\mathfrak{m})=1 is equivalent to RR being regular and also eHK(m)=1e_{\mathrm{HK}}(\mathfrak{m})=1 equivalent to RR being regular. The first equivalence is due to a classical result established by P. Samuel in the equal-characteristic case [9] and by M. Nagata in the general case [7]. The second equivalence was established more recently by K. I. Watanabe and K. I. Yoshida [13], Theorem 1.5.

Since e(I[q])=e(Iq)=qde(I)e(I^{[q]}) = e(I^q) = q^d e(I) for all q=peq = p^e, we have lim⁡q→∞e(I[q])qd=e(I)\lim_{q \to\infty} \frac{e(I^{[q]})}{q^d} = e(I). This naturally leads to the study of the asymptotic behavior of the Hilbert coefficients ei(I[q])qd\frac{e_i(I^{[q]})}{q^d} as q→∞q \to\infty. We set, whenever the limit exists,

Li(I)=lim⁡q→∞ei(I[q])qd.L_i(I) = \lim_{q \to\infty} \frac{e_i(I^{[q]})}{q^d}.

The existence of these limits and their relationship with the Hilbert-Kunz multiplicities of ordinary powers of II have been investigated in recent work. V. Trivedi first established the existence of L1(I)L_1(I) for standard graded algebras and equigenerated homogeneous ideals [11], while I. Smirnov later established the existence of L1(I)L_1(I) for every m-primary ideal in a local ring [10] and formulated the following conjecture.

Conjecture 1.1. Let (R,m)(R,\mathfrak{m}) be a Noetherian local ring of characteristic p>0p > 0 and dimension dd. Let II be an m\mathfrak{m}-primary ideal in RR. If RR is weakly FF-regular, analytically unramified and Cohen-Macaulay, then L1(I)=e(I)−eHK(I)L_1(I) = e(I) - e_{\mathrm{HK}}(I) if and only if II is stable, i.e., I2=JII^2 = JI for a minimal reduction JJ of II.

Note that the conjecture was originally stated without the weakly FF-regular assumption, but it is known to be false without this hypothesis (see [2]). It is also known that if II is stable, then L1(I)=e(I)−eHK(I)L_1(I) = e(I) - e_{\mathrm{HK}}(I), see [14 Theorem 1.8(2)]. De Stefani, Masuti, Rossi, and Verma [3] proved the conjecture for dimension 1 and 2-dimensional excellent local rings (see [3 Corollary 2.2, Theorem 4.9]). In this paper, we establish an affirmative answer to the conjecture 1.1 for d≥3d \ge3 under certain mild conditions. We prove

Theorem 1.2 (Theorem 3.3 and 3.2). Assume the same hypothesis as in Conjecture 1.1. Then the conjecture holds if either of the following conditions is satisfied.

(1) RR is an excellent and strongly FF-regular ring.

(2) RR is an excellent ring with d≤3d \le3.

Organization of the paper: This paper is structured into three sections. Section 2 introduces the required preliminaries. In Section 3, we first establish a lower bound for L1(I)−e(I)+eHK(I)L_1(I) - e(I) + e_{\mathrm{HK}}(I) and then give a proof of the main result (Theorem 1.2) using the bound.

Preliminaries

Throughout this paper, all rings are Noetherian commutative with unity and modules are finitely generated. In this section, we discuss some definitions and preliminary results that we need.

We first recall the definition of FF-signature.

2.1. Let (R,m,k)(R,\mathfrak{m},k) be a Noetherian local ring of prime characteristic p>0p > 0, where k=R/mk = R/\mathfrak{m}. For each e≥1e \ge1, we denote by F∗e(R)F_*^e(R) the additive group of RR equipped with the RR-module structure r⋅F∗e(x)=F∗e(rpex)r \cdot F_*^e(x) = F_*^e(r^{p^e}x) for all r,x∈Rr,x \in R. The ring RR is called FF-finite if F∗e(R)F_*^e(R) is a finitely generated RR-module for all e≥1e \ge1. We write F∗(−)F_*(-), when e=1e = 1. In this case, set

α=log⁡p(dim⁡kF∗(k))=log⁡p[k:kp].\alpha= \log_p(\dim_k F_*(k)) = \log_p[k : k^p].

Let (R,m,k)(R,\mathfrak{m},k) be an FF-finite local ring of dimension dd. For every e≥1e \ge1, write

F∗e(R)≅R⊕ae⊕Me,F_*^e(R) \cong R^{\oplus a_e} \oplus M_e,

where MeM_e has no nonzero free direct summand. The integer ae=ae(R)a_e=a_e(R) is called the ee-th Frobenius splitting number of RR; it is the largest rank of a free RR-module occurring as a direct summand of F∗e(R)F_*^e(R). The FF-signature of RR is defined as

s(R)=lim⁡e→∞aepe(d+α).s(R)=\lim_{e\to\infty}\frac{a_e}{p^{e(d+\alpha)}}.

The existence of this limit was proved in full generality by K. Tucker [12].

We now list some properties that are preserved under faithfully flat extension and useful for our main results.

2.2. Let ϕ:(R,m)→(R′,m′)\phi:(R,\mathfrak{m})\to(R',\mathfrak{m}') be a local ring homomorphism. Assume R′R' is a faithfully flat RR-algebra with mR′=m′\mathfrak{m}R'=\mathfrak{m}'. Set I′=IR′I'=IR' for an m\mathfrak{m}-primary ideal II. Let MM be an RR-module. Set M′=M⊗RR′M'=M\otimes_R R'. In these cases, it can be shown that

  1. ℓR(M)=ℓR′(M′)\ell_R(M)=\ell_{R'}(M'). In particular, ℓR(R/In)=ℓR′(R′/I′n)\ell_R(R/I^n)=\ell_{R'}(R'/I'^n) for all n≥1n\ge1. Thus, ei(I)=ei(I′)e_i(I)=e_i(I') for all i≥0i\ge0.

  1. dim⁡RM=dim⁡R′M′\dim_R M=\dim_{R'}M' and grade⁡R(K,M)=grade⁡R′(K′,M′)\operatorname{grade}_R(K,M)=\operatorname{grade}_{R'}(K',M') for any ideal KK of RR.

  1. For R′=R^R'=\widehat{R} (the completion of RR with respect to the maximal ideal) with characteristic p>0p>0, the following properties hold. (a) I[q]R′=I′[q]I^{[q]}R'=I'^{[q]} for all q=peq=p^e, e∈Ne\in\mathbb{N}. (b) [4 Corollary 7.28] RR is excellent weakly FF-regular iff R^\widehat{R} is excellent weakly FF-regular. (c) [5 Theorem 2.23] RR is strongly FF-regular iff R^\widehat{R} is strongly FF-regular.

Next, we list some cases where weakly FF-regular implies strongly FF-regular. Note that a local weakly FF-regular ring is a domain.

Proposition 2.3. Let (R,m)(R,\mathfrak{m}) be a Noetherian local ring of characteristic p>0p>0. Let RR be a weakly FF-regular, FF-finite local ring of dimension d≤3d\le3 and a homomorphic image of a Gorenstein ring. Then RR is strongly FF-regular.

Proof. For d=2d=2 and 33, the result now follows from [15 pg-736, 742]. Let d=1d=1. By [4 Theorem 3.4(c)], RR is normal and hence regular. By [4 Theorem 5.5(c)], the result follows. □\square

main results

In this section, we first establish a lower bound for L1(I)−e(I)+eHK(I)L_1(I)-e(I)+e_{HK}(I) in terms of the FF-signature and Hilbert coefficients of RR with respect to II. We then use this bound, together with the Γ\Gamma construction (see [4]), to prove Theorem 1.2.

We begin by establishing a lower bound for L1(I)−e(I)+eHK(I)L_1(I)-e(I)+e_{HK}(I).

Theorem 3.1. Let (R,m,k)(R,\mathfrak{m},k) be an FF-finite Cohen–Macaulay local ring of characteristic p>0p>0 and dimension dd, and let I⊆RI\subseteq R be an m\mathfrak{m}-primary ideal. Then

L1(I)−e(I)+eHK(I)≥s(R)(e1(I)−e(I)+ℓR(R/I)).L_1(I)-e(I)+e_{HK}(I)\ge s(R)\bigl(e_1(I)-e(I)+\ell_R(R/I)\bigr).

Proof. Let q=peq = p^{e} and set α=log⁡p[k:kp]\alpha= \log_{p}[k : k^{p}]. For all n≥1n \ge1, we have InF∗e(R)=F∗e((I[q])n)I^{n}F^{e}_{*}(R) = F^{e}_{*}((I^{[q]})^{n}). Since F∗e(−)F^{e}_{*}(-) is an exact functor of RR-modules, we have

InF∗e(R)≅F∗e(R(I[q])n).I^{n}F^{e}_{*}(R) \cong F^{e}_{*}\left(\frac{R}{(I^{[q]})^{n}}\right).

We also have, for a finite length RR-module NN, ℓR(F∗eN)=qαℓR(N)\ell_{R}(F^{e}_{*}N) = q^{\alpha}\ell_{R}(N) (see [12 2.1]). It follows that

ℓR(F∗e(R)InF∗e(R))=qαℓR(R(I[q])n)\ell_{R}\left(\frac{F^{e}_{*}(R)}{I^{n}F^{e}_{*}(R)}\right) = q^{\alpha}\ell_{R}\left(\frac{R}{(I^{[q]})^{n}}\right)

for all n≥1n \ge1. Comparing the Hilbert polynomials, we obtain ei(I,F∗e(R))=qαei(I[q])e_{i}(I,F^{e}_{*}(R)) = q^{\alpha}e_{i}(I^{[q]}) for all 0≤i≤d0 \le i \le d. In particular, we get

e(I,F∗e(R))=qαe(I[q])=qd+αe(I)ande1(I,F∗e(R))=qαe1(I[q]).(1)e(I,F^{e}_{*}(R)) = q^{\alpha}e(I^{[q]}) = q^{d+\alpha}e(I) \quad\text{and}\quad e_{1}(I,F^{e}_{*}(R)) = q^{\alpha}e_{1}(I^{[q]}). \tag*{(1)}

For a maximal Cohen–Macaulay RR-module MM, set

δI(M)=e1(I,M)−e(I,M)+ℓR(M/IM).\delta_{I}(M) = e_{1}(I,M) - e(I,M) + \ell_{R}(M/IM).

By the module version of Northcott’s inequality (see [8], Proposition 12), we have δI(M)≥0\delta_{I}(M) \ge0. Since RR is Cohen–Macaulay, F∗e(R)F^{e}_{*}(R) is a maximal Cohen–Macaulay RR-module. Indeed, if x1,…,xdx_{1},\ldots,x_{d} is an RR-regular sequence, then x1q,…,xdqx_{1}^{q},\ldots,x_{d}^{q} is also an RR-regular sequence, and multiplication by xix_{i} on F∗e(R)F^{e}_{*}(R) corresponds to multiplication by xiqx_{i}^{q} on the underlying copy of RR. For all e≥1e \ge1, write

F∗e(R)≅R⊕ae⊕Me,F^{e}_{*}(R) \cong R^{\oplus a_{e}} \oplus M_{e},

where MeM_{e} has no free direct summands. Since F∗e(R)F^{e}_{*}(R) is maximal Cohen–Macaulay, so is MeM_{e}. Since the Hilbert coefficients and lengths are additive on direct sums, we have

δI(F∗e(R))=aeδI(R)+δI(Me)≥aeδI(R).\delta_{I}(F^{e}_{*}(R)) = a_{e}\delta_{I}(R) + \delta_{I}(M_{e}) \ge a_{e}\delta_{I}(R).

On the other hand, from (1), δI(F∗e(R))=qα(e1(I[q])−qde(I)+ℓR(R/I[q]))≥aeδI(R)\delta_{I}(F^{e}_{*}(R)) = q^{\alpha}(e_{1}(I^{[q]}) - q^{d}e(I) + \ell_{R}(R/I^{[q]})) \ge a_{e}\delta_{I}(R). Dividing by qd+αq^{d+\alpha} yields

e1(I[q])qd−e(I)+ℓR(R/I[q])qd≥aeqd+α(e1(I)−e(I)+ℓR(R/I)).\frac{e_{1}(I^{[q]})}{q^{d}} - e(I) + \frac{\ell_{R}(R/I^{[q]})}{q^{d}} \ge\frac{a_{e}}{q^{d+\alpha}}\left(e_{1}(I) - e(I) + \ell_{R}(R/I)\right).

Taking the limit as q→∞q \to\infty, we obtain

L1(I)−e(I)+eHK(I)≥s(R)(e1(I)−e(I)+ℓR(R/I)).L_{1}(I) - e(I) + e_{\mathrm{HK}}(I) \ge s(R)\left(e_{1}(I) - e(I) + \ell_{R}(R/I)\right).

Note that if s(R)≠0s(R) \ne0 (for example, RR is strongly FF-regular, see [1 Theorem 0.2]) and L1(I)=e(I)−eHK(I)L_{1}(I) = e(I)-e_{\mathrm{HK}}(I) in Theorem 3.1, then e1(I)=e(I)+ℓR(R/I)e_{1}(I) = e(I)+\ell_{R}(R/I) which is equivalent to II being stable (see [8 Theorem 16]). We now prove the Conjecture 1.1 when RR is an excellent and strongly FF-regular ring. Since stabile ideal II implies L1(I)=e(I)−eHK(I)L_{1}(I) = e(I) - e_{\mathrm{HK}}(I) in general (see [14 Theorem 1.8(2)]), it remains only to establish the converse.

Theorem 3.2. Assume the same hypothesis as in Conjecture 1.1 and let RR be an excellent and strongly FF-regular ring. Then L1(I)=e(I)−eHK(I)L_{1}(I) = e(I) - e_{\mathrm{HK}}(I) if and only if II is stable.

Proof. Assume that L1(I)=e(I)−eHK(I)L_1(I)=e(I)-e_{\mathrm{HK}}(I). By 2.2, we may assume that RR is complete. We first reduce the case to FF-finite ring by the well known Γ\Gamma-construction and then use Theorem 3.1. By Cohen’s structure theorem, we have R≅T/aR\cong T/\mathfrak{a}, where T=k[[x1,…,xn]]T=k[[x_1,\ldots,x_n]] and kk is a coefficient field of RR.

Let Λ\Lambda be a pp-basis of kk (see [6 3.1]). Choose a sufficiently small cofinite subset Γ⊆Λ\Gamma\subseteq\Lambda and set S=RΓS=R^\Gamma. By [4 6.11, 6.13, 6.6(d)], the map R→SR\to S is faithfully flat and purely inseparable, and SS is FF-finite local excellent domain with maximal ideal n=mS\mathfrak{n}=\mathfrak{m}S. Now by [6 Theorem 3.4(ii)], SS is strongly FF-regular.

Since the map (R,m)→(S,n)(R,\mathfrak{m})\to(S,\mathfrak{n}) is faithfully flat with mS=n\mathfrak{m}S=\mathfrak{n}, by 2.2 and the above argument, we may assume that RR is FF-finite. Therefore, by [1 Theorem 0.2], s(R)>0s(R)>0 and by Theorem 3.1, we have e1(I)=e(I)−ℓR(R/I)e_1(I)=e(I)-\ell_R(R/I). It follow that II is a stable ideal, see [8 Theorem 16]. □

As a consequence of Theorem 3.2, we prove the Conjecture 1.1 when RR is an excellent local ring of dimension at most 3. Note that De Stefani, Masuti, Rossi, and Verma [3] proved the conjecture for dimension ≤2\leq2. We now give a proof for dimension at most 3, which is different from their proof.

Theorem 3.3. Assume the same hypothesis as in Conjecture 1.1 and let RR be an excellent ring with d≤3d\leq3. Then L1(I)=e(I)−eHK(I)L_1(I)=e(I)-e_{\mathrm{HK}}(I) if and only if II is stable.

Proof. Assume that L1(I)=e(I)−eHK(I)L_1(I)=e(I)-e_{\mathrm{HK}}(I). By 2.2, we may assume that RR is complete. We first reduce the case to FF-finite ring by the well-known Γ\Gamma-construction and then use Theorem 3.1. By Cohen’s structure theorem, we have R≅T/aR\cong T/\mathfrak{a}, where T=k[[x1,…,xn]]T=k[[x_1,\ldots,x_n]] and kk is a coefficient field of RR.

Let Λ\Lambda be a pp-basis of kk (see [6 3.1]). Choose a sufficiently small cofinite subset Γ⊆Λ\Gamma\subseteq\Lambda and set S=RΓS=R^\Gamma. By [4 6.11, 6.13], the map R→SR\to S is faithfully flat and purely inseparable, and SS is FF-finite local excellent domain with maximal ideal n=mS\mathfrak{n}=\mathfrak{m}S. In particular, mS\mathfrak{m}S is radical. Therefore, [4 Theorem 6.17(b)] shows that SS is weakly FF-regular. By [4 Lemma 6.6(a,b)], we also have S≅TΓ/aTΓS\cong T^\Gamma/\mathfrak{a}T^\Gamma, where TΓT^\Gamma is a regular local ring. Thus, SS is a homomorphic image of a Gorenstein local ring. Hence, by Proposition 2.3, SS is strongly FF-regular.

Since the map (R,m)→(S,n)(R,\mathfrak{m})\to(S,\mathfrak{n}) is faithfully flat with mS=n\mathfrak{m}S=\mathfrak{n}, by 2.2 and the above argument, we may assume RR is FF-finite and strongly FF-regular. The result now follows from Theorem 3.2. □

Acknowledgment: The author thanks Dr. Shreedevi Masuti for suggesting the problem and for her valuable comments and discussions. The author also thanks ChatGPT (version 5.6, Sol) for its assistance in working out the proof of Theorem 3.1. Email address: [email protected], [email protected]

Indian Institute of Technology Dharwad, Dharwad, Karnataka, India, 580011

References

  1. [1]I. M. Aberbach and G. J. Leuschke, The F-signature and strong F-regularity, Math. Res. Lett. 10 (2003), no. 1, 51-56.DOI
  2. [2]A. Banerjee, K. Goel, and J. K. Verma, On the Hilbert-Samuel coefficients of Frobenius powers of an ideal, Proc. Amer. Math. Soc. 152 (2024), no. 4, 1501-1515.arxiv.org/abs/2203.10173
  3. [3]A. De Stefani, Shreedevi K. Masuti, M. E. Rossi, and J. K. Verma, Hilbert-Kunz multiplicity of powers of ideals in dimension two, Bull. Lond. Math. Soc. 57 (2025), no. 7, 2155-2176.arxiv.org/abs/2412.07546
  4. [4]M. Hochster and C. Huneke, F-regularity, test elements, and smooth base change, Trans. Amer. Math. Soc. 346 (1994), no. 1, 1-62.
  5. [5]M. Hochster and Y. Yao, Frobenius splitting, strong F-regularity, and small Cohen-Macaulay modules, Trans. Amer. Math. Soc. 376 (2023), no. 9, 6729-6765.DOI
  6. [6]T. Murayama, The gamma construction and asymptotic invariants of line bundles over arbitrary fields, Nagoya Math. J. 242 (2021), 165-207
  7. [7]M. Nagata, Local rings, Robert E. Krieger Publishing Co., Huntington, NY, 1975. Corrected reprint.
  8. [8]T. J. Puthenpurakal, Hilbert-coefficients of a Cohen-Macaulay module, J. Algebra 264 (2003), no. 1, 82-97.DOI
  9. [9]P. Samuel, La notion de multiplicité en algèbre et en géométrie algébrique, J. Math. Pures Appl. (9) 30 (1951), 159-205.
  10. [10]J. Simnory, Hilbert-Kunz multiplicity of the powers of an ideal, Proc. Amer. Math. Soc. 147 (2019), no. 8, 3331-3338.
  11. [11]V. Trivedi, Asymptotic Hilbert-Kunz multiplicity, J. Algebra 492 (2017), 498-523.DOI
  12. [12]K. Tucker, F-signature exists, Invent. Math. 190 (2012), no. 3, 743-765.DOI
  13. [13]K.-i. Watanabe and K.-i. Yoshida, Hilbert-Kunz multiplicity and an inequality between multiplicity and colength, J. Algebra 230 (2000), no. 1, 295-317.DOI
  14. [14]K.-i. Watanabe and K.-i. Yoshida, Hilbert-Kunz multiplicity of two-dimensional local rings, Nagoya Math. J. 162 (2001), 87-110.DOI
  15. [15]L. J. Williams, Uniform stability of kernels of Koszul cohomology indexed by the Frobenius endomorphism, J. Algebra 172 (1995), no. 3, 721-743.

Paper details

vlnc:W_hrcx9rzg
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

GPT-5.6 Sol (bare)

View statement
Looking for related papers…
Contents