Quantitative injectivity of the Fubini--Study map

Yoshinori Hashimoto

REVIEW 3 major objections 4 minor 12 references

Quantitative injectivity of the Fubini--Study map

T0 review · 3 major / 4 minor · reviewed 2026-08-08 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read For any two metrics on H^0(X,L^k), the inverse-matrix distance is controlled by a Sobolev norm of their induced function, with a k^n bound.

desk verdict The main theorem is not established; a rank obstruction in Lemma 2.2 makes the central inequality impossible for large k. read the letter →

arxiv 2502.08038 v1 pith:R4H6EAHB submitted 2025-02-12 math.AG math.DG

Yoshinori Hashimoto This is my paper · ORCID

classification math.AGmath.DG MSC 14C2032Q1553C5532A2514N05

keywords Fubini-StudymapquantitativeinjectivityKodairaembeddingBergmankernelexpansionsecondfundamentalformHilbert-SchmidtnormSobolevW^{22}Kählergeometry

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper proves a quantitative version of the injectivity of the Fubini–Study map: for a polarised smooth projective variety (X,L) with ample L and a fixed hermitian metric h, the difference $A^{{-1}}$-$B^{{-1}}$ between the inverse matrices of any two Hermitian forms on $H^{0}$(X,L^k) is bounded above by a constant times k^n times the $W^{{2,2}}$ norm of the function f_k(A,B;H_k). The exponent n is the dimension of X, so the estimate is polynomial in the level k, not exponential. This matters because the Fubini–Study map connects Hermitian metrics on the space of sections to Kähler metrics on X, and a quantitative injectivity statement with polynomial growth is precisely what is needed to compare geometric constructions across different powers k. The paper also corrects the arguments in the author's previous papers that relied on an erroneous surjectivity claim for the Hilbert map.

What carries the argument

The proof combines the Bergman kernel expansion, which gives ω_{h,k} = kω_h + O(1/k) and locates the centre of mass of the embedding near the identity, with a lower bound on the second fundamental form -A^*_{h,k} ∧ A_{h,k} of the Kodaira embedding (Lemma 2.1), the load-bearing nondegeneracy statement. It then follows the Phong–Sturm decomposition of the traceless part of Λ into the automorphism algebra and its orthogonal complement, bounding the normal component of the associated holomorphic vector field by its (0,1)-derivative through the second fundamental form. The trace part is handled separately by integrating the defining function against the Bergman kernel, yielding the final $W^{{2,2}}$ estimate with the k^n factor.

What would settle it

Find a polarized variety, for instance a smooth projective curve of genus at least two, compute the second fundamental form of the Kodaira embedding for a sequence of levels k and points x_k, and check whether the minimum eigenvalue of -A^*_{h,k} ∧ A_{h,k} stays bounded away from zero; a sequence tending to zero would falsify Lemma 2.1. A more direct test is to verify the perturbation assertion in the proof of Lemma 2.1: construct h_ε from the Kähler potential ε|z|^2|z_l|^2 times a cutoff function and check whether Hilb(h_ε^k) = Hilb(h^k) + O($ε^{2}$) for all large k; a counterexample to this identity would break the proof of the uniform lower bound.

Watch

Extended reading notes

Core claim

Theorem 1.2 states that there exist constants k_0 in N and C_h > 0, depending only on h, such that for every A,B in B_k and every k ≥ k_0, ||$A^{{-1}}$-$B^{{-1}}$||^2_{HS(H_k)} ≤ C_h k^n ||f_k(A,B;H_k)||^2_{$W^{{2,2}}$(ω_h)}. In words, the Fubini–Study map is injective with a quantitative bound, and the bound is polynomial of degree n in the exponent k. The result is stated with respect to the reference metric H_k = Hilb(h^k), and the proof shows that the trace-free part of Λ = $A^{{-1}}$-$B^{{-1}}$ is controlled by the first and second derivatives of f_k, while the trace part is controlled by the $L^{2}$ norm of f_k via the Bergman kernel expansion.

Load-bearing premise

The proof of Lemma 2.1 relies on an unproved assertion that the hermitian metric h can be locally perturbed to h_ε so that the curvature of the induced metric changes by -εδ at a point while the Hilbert functional Hilb(h_ε^k) changes only at order $ε^{2}$; if this perturbation property fails, the uniform lower bound on the second fundamental form, and with it Theorem 1.2, collapses.

Editorial extensions

If this is right

The distance between A and B in B_k is controlled by the W^{2,2} norm of f_k, not merely its C^0 norm, so quantitative injectivity holds with a fixed reference metric H_k = Hilb(h^k).
The bound is uniform in A,B and all k ≥ k_0, with constants depending only on the reference metric h, so it applies to arbitrarily large powers of the line bundle.
The trace part of Λ is controlled by the L^2 norm of f_k via the Bergman kernel, so the full Hilbert–Schmidt norm is equivalent to the W^{2,2} norm up to the factor k^n.
The correction in Section 3 yields a quantitative bound |d_i - 1| ≤ C k^{2n-m-2} for the eigenvalues in the extremal-metric quantisation argument, at the cost of a power of k.
Injectivity of the Fubini–Study map in the sense of Lempert is recovered in the limit: if f_k has zero W^{2,2} norm, then A = B.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

A natural test of sharpness would be to replace W^{2,2} by W^{1,2} for the trace-free part, since the control of the normal component of the vector field uses the (0,1)-derivative of the projected field, a second-order object; lowering the exponent to n-1 may fail.
The same strategy should give a uniform constant for families of metrics h varying in a compact set, since the compactness argument in Lemma 2.1 is the only place where uniformity in h enters.
The failure of the earlier C^0 estimate and the success of the W^{2,2} norm here suggest that the right function space is tied to the second fundamental form: quantitative injectivity holds exactly where the Kodaira embedding is uniformly non-degenerate.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Dear colleague,

The headline is that the paper's main theorem is not proved. The argument fails at a specific point: Lemma 2.2 asserts a pointwise lower bound on the second fundamental form that is algebraically impossible when the codimension exceeds the dimension. No repair of Lemma 2.1 can fix this.

What the paper does well: Theorem 1.2 is a clean, natural statement – a quantitative injectivity estimate for the Fubini–Study map with polynomial dependence on k. The author is transparent about the gaps in his earlier work, and the correction to [8] is a serious attempt to salvage a result that depended on a flawed claim. The writing is clear, and the use of Phong–Sturm's framework is standard.

The soft spot is load-bearing. The proof of Lemma 2.1 asserts, without proof, the existence of a local metric perturbation that changes the Hilbert functional only at order ε². That is already a nontrivial claim. But the deeper problem is Lemma 2.2. The inequality there says that for every point x and every vector b ∈ C^m (where m = codimension of X in P^{N_k-1}), we have |M b|² ≥ λmin |b|², with M_ij = ∂̄φ_ij(x). M is n × m, and for large k, m ≈ V k^n − n − 1 > n. A linear map from C^m to C^n cannot be injective when m > n, so M has a nonzero kernel, and the inequality fails for any λmin > 0. This holds even granting Lemma 2.1, because the obstruction is pure dimension counting. Consequently, Lemma 2.2 collapses, and with it the bound on ‖π_N ξ_β‖ that feeds directly into Theorem 1.2.

The final estimate does not go through, and the correction to [8] relies on Theorem 1.2, so it also loses its basis.

Who this is for: anyone working on balanced metrics or the Yau–Tian–Donaldson program will want to know about this paper, mainly as a cautionary example of how these injectivity estimates can fail. The author's honesty about his earlier errors is commendable.

My recommendation: I would not cite the main theorem. But I would still send it to a serious referee – a competent referee can catch this quickly, and the author might find a way to repair the argument by restricting the class of normal vectors or using a different norm. As it stands, the paper should not be accepted without substantial revision.

Referee Report

3 major / 4 minor

Summary. The paper claims a quantitative injectivity estimate for the Fubini–Study map: for a polarised smooth projective variety (X, L) with Hermitian metric h, there exist k0 and C_h > 0 such that for all k ≥ k0 and all positive definite Hermitian forms A, B on H^0(X, L^k), ||A^{-1} - B^{-1}||^2_{HS(H_k)} ≤ C_h k^n ||f_k(A,B;H_k)||^2_{W^{2,2}(ω_h)}, where H_k = Hilb(h^k) is the Hilbert functional and f_k is the ratio of Fubini–Study metrics. The proof uses the Bergman kernel expansion, a claimed uniform lower bound on the second fundamental form of the Kodaira embedding (Lemma 2.1), an estimate for the normal component of Hamiltonian vector fields (Lemma 2.2), and a comparison of the tangent projection of the vector field with derivatives of f_k. Section 3 applies Theorem 1.2 to correct an argument in the author's previous paper [8].

Significance. If the theorem were established, it would be a substantial quantitative refinement of Lempert's injectivity theorem, with an explicit polynomial dependence on the level k and a natural Sobolev norm. It would also repair arguments in the author's earlier work. The paper is well-motivated and the choice of the W^{2,2} norm is plausible. However, the proof as written contains a load-bearing algebraic impossibility in Lemma 2.2, so the main theorem is not established. The manuscript would require a fundamentally different estimate for the normal component of the Hamiltonian vector field, not a local repair.

major comments (3)

[Section 2, Lemma 2.2] The pointwise inequality proved in Lemma 2.2, namely sum_i |sum_{j=1}^{m} b_j dbar φ_ij(x)|^2 ≥ λ_min sum_j |b_j|^2 for all b = (b_j) in C^m, is algebraically impossible when m > n. Here M_ij = dbar φ_ij(x) is an n × m matrix, so for m > n the Hermitian matrix M^*M has rank at most n and therefore has a kernel; its smallest eigenvalue is 0, so no λ_min > 0 can satisfy the inequality for all b. Since m = N_k - 1 - n and N_k = V k^n + O(k^{n-1}), one has m > n for all sufficiently large k. This is not a missing detail in Lemma 2.1: a lower bound on the endomorphism -A^*_{h,k} ∧ A_{h,k} of TX does not imply injectivity of the normal-to-tangent map A: N → T^*X when the normal bundle has rank larger than n. The bound on ||π_N ξ_β|| used in the proof of Theorem 1.2 therefore does not follow, and the main theorem collapses.
[Section 2, Lemma 2.1] The proof of Lemma 2.1 asserts without proof that one can perturb the Hermitian metric h locally to h_ε such that h_ε(x)=h(x), g_ε(x)=g(x), the curvature tensor is perturbed by -ε δ^i_j (k g_h)_{l\bar m} at x, and Hilb(h_ε^k) = Hilb(h^k) + O(ε^2). This is a nontrivial statement: a local perturbation of order ε, even with a rapidly decreasing cutoff, will generically change the L^2 inner product Hilb at first order in ε unless the first moment of the perturbation against all products of sections vanishes. No such cancellation is demonstrated. The O(ε^2) property is load-bearing because the subsequent comparison of -A^*_{h,k,ε} ∧ A_{h,k,ε} with -A^*_{h,k} ∧ A_{h,k} assumes that the Fubini–Study metric and the orthogonal projection π_T are unchanged to first order. Without a proof of this perturbation claim, Lemma 2.1's uniform lower bound for all k is unsupported.
[Section 3, correction to [8]] The correction to [8] in Section 3 relies directly on Theorem 1.2 (see the estimate |d_i - 1| ≤ C_7 C_9 C_h k^{2n-m-2}), and the earlier step bounding |A^{-1} - B^{-1}|_{HS(H_{m,k})} also uses Theorem 1.2. Since Theorem 1.2 is not established due to the failure of Lemma 2.2, the claimed correction to [8] is likewise unsupported.

minor comments (4)

[Throughout] There are numerous OCR-type typographical errors, e.g. 'Fub ini' in the title, 'K¨ ahler' and 'ωn h' spacing issues, and 'λmin > 0 > 0' in Lemma 2.2.
[Section 2, Lemma 2.2 proof] The notation 'with N − 1 = n + m' is ambiguous; it should be 'N_k - 1 = n + m' to be consistent with the rest of the paper.
[Section 2, proof of Theorem 1.2] The constant C_4 is defined as 2 max{C_3, 2V}, but the bracketing of the preceding display suggests the coefficient of the L^2 term should be max{2, 2C_3} or similar; the stated choice of C_4 does not obviously match the displayed inequalities.
[Section 2, proof of Theorem 1.2] The notation '||∇∇f_k||' is not defined; presumably it denotes the L^2 norm of the Hessian (or of ∇^{0,1}df_k), but it should be made explicit.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the paper's main estimate is derived from external curvature and Bergman-kernel results, and its self-citations are used only to identify and correct prior errors.

full rationale

I find no circularity in the derivation of Theorem 1.2. The proof is a conventional chain: the curvature identity for the second fundamental form is cited from Griffiths–Harris and Phong–Sturm; the Bergman kernel expansion and Rawnsley's formula for the Fubini–Study metric are cited from external sources; and the Hamiltonian vector field construction is quoted from Fine and Phong–Sturm. The matrix difference Λ and the function f_k(A,B;H_k) are related by definition, but bounding one by the other is precisely the content of the theorem, not a tautology. Lemma 2.1 supplies a geometric lower bound that Lemma 2.2 then uses; this is a legitimate logical dependence, not a circular one. The self-citations [7,8] are explicitly described as containing errors and are used only to state what needs correction; they are not load-bearing for Theorem 1.2. The unproved metric-perturbation assertion inside Lemma 2.1 ('This is possible by considering the perturbation of the Kaehler potential...') is a gap or correctness risk, and the skeptic's rank-count objection to Lemma 2.2, if valid, would be a mathematical error, not a circularity. Since no step reduces by construction to its own inputs and no fitted parameter is renamed as a prediction, the appropriate circularity score is 0.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

The paper introduces no free parameters fitted to data and no new postulated entities. It relies on standard background results from the literature (Bergman kernel expansion, asymptotic Riemann-Roch, curvature identities, Phong-Sturm estimates). The main unproved step is the perturbation claim inside Lemma 2.1, which is better treated as a proof gap and is listed as a red flag.

assumptions (5)

standard math Asymptotic expansion of the Bergman kernel: rho_k(omega_h) = k^n + O(k^{n-1}) in C^m norm
Invoked in equation (2) to control omega_{h,k} = k omega_h + O(1/k) and the centre of mass condition; cited to Ma-Marinescu [10].
standard math Asymptotic Riemann-Roch: N_k = dim H^0(X,L^k) = V k^n + O(k^{n-1})
Used to bound the trace part c via N_k and to obtain the final k^n dependence in Theorem 1.2.
standard math Curvature identity: pi_T∘(tilde F_{h,k}|_{TX}) - F_{h,k} = -A^*_{h,k} ∧ A_{h,k}
Used in Lemma 2.1 to identify the second fundamental form with the difference of curvatures; cited to [6, page 78] or [11, equation (5.27)].
standard math Phong-Sturm estimate [11, (5.7)]: ||Lambda||^2_HS <= C_2 k ||xi_Lambda||^2_L^2 with the centre-of-mass condition (5.1) satisfied to order 1/k
A key external input; the paper verifies the condition via the Bergman kernel expansion.
standard math Orthogonal decomposition sl(N,C) = aut(X,L) ⊕ aut(X,L)^⊥ and the Hamiltonian/L2 duality pi_T xi_Lambda dual to d f_k
Follows from the linearisation and [3, Lemma 20]; used in the proof of Theorem 1.2.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Quantitative injectivity of the Fubini--Study map." pith.science (2026). https://pith.science/paper/R4H6EAHB

@misc{pith2026250208038,
  author       = {Pith},
  title        = {Pith review of: Quantitative injectivity of the Fubini--Study map},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/R4H6EAHB}},
  note         = {Machine review of arXiv:2502.08038}
}

read the original abstract

We prove a quantitative version of the injectivity of the Fubini--Study map that is polynomial in the exponent of the ample line bundle, and correct the arguments in the author's previous papers.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

12 extracted references · 10 canonical work pages

[8]

, Quantisation of extremal K¨ ahler metrics, J. Geom. Anal. 31 (2021), no. 3, 2970–3028. MR4225832 ↑1, 7, 8

work page 2021
[1]

S. K. Donaldson, Scalar curvature and projective embeddings. II , Q. J. Math. 56 (2005), no. 3, 345–356. MR2161248 ↑1 [2] , Some numerical results in complex diﬀerential geometry , Pure Appl. Math. Q. 5 (2009), no. 2, Special Issue: In honor of Friedrich Hirzebruch. Part 1, 571–618. MR 2508897 ↑8

work page 2005
[3]

Fine, Calabi ﬂow and projective embeddings , J

J. Fine, Calabi ﬂow and projective embeddings , J. Diﬀerential Geom. 84 (2010), no. 3, 489–523. With an appendix by Kefeng Liu and Xiaonan Ma. MR2669363 ↑4

work page 2010
[4]

Finski, On the metric structure of section ring , arXiv preprint arXiv:2209.03853 (2022)

S. Finski, On the metric structure of section ring , arXiv preprint arXiv:2209.03853 (2022). ↑2 8

arXiv 2022
[5]

, Submultiplicative norms and ﬁltrations on section rings , arXiv preprint arXiv:2210.03039 (2022). ↑2

arXiv 2022
[6]

Griﬃths and J

P. Griﬃths and J. Harris, Principles of algebraic geometry , Wiley Classics Library, John Wiley & Sons, Inc., New York, 1994. Reprint of the 1978 original. MR1288523 (95d :14001) ↑3

work page 1994
[7]

Hashimoto, Mapping properties of the Hilbert and Fubini-Study maps in K ¨ ahler geometry, Ann

Y. Hashimoto, Mapping properties of the Hilbert and Fubini-Study maps in K ¨ ahler geometry, Ann. Fac. Sci. Toulouse Math. (6) 29 (2020), no. 2, 371–389. MR4150546 ↑1, 2, 7

work page 2020
[9]

On the Bergman kernels of holomorphic vector bundles

L. Lempert, On the Bergman kernels of holomorphic vector bundles , arXiv preprint arXiv:2109.08593 (2021). ↑2, 7

work page Pith review arXiv 2021

Show all 12 references

[10]

Ma and G

X. Ma and G. Marinescu, Holomorphic Morse inequalities and Bergman kernels , Progress in Mathematics, vol. 254, Birkh¨ auser Verlag, Basel, 2007. MR2339952 (2008 g:32030) ↑3

2007
[11]

D. H. Phong and J. Sturm, Scalar curvature, moment maps, and the Deligne pairing , Amer. J. Math. 126 (2004), no. 3, 693–712. MR2058389 ↑3, 4, 5

2004
[12]

J. H. Rawnsley, Coherent states and K¨ ahler manifolds , Quart. J. Math. Oxford Ser. (2) 28 (1977), no. 112, 403–415. MR0466649 (57 #6526) ↑3, 6, 8

1977
[13]

Sun, On The Image Of The Hilbert Map , arXiv preprint arXiv:2208.13407 (2022)

J. Sun, On The Image Of The Hilbert Map , arXiv preprint arXiv:2208.13407 (2022). ↑2 Department of Mathematics, Osaka Metropolitan University , 3-3-138, Sugimoto, Sumiyoshi-ku, Os- aka, 558-8585, Japan. Email address : yhashimoto@omu.ac.jp 9

2022 arXiv

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.

[1] [8]

, Quantisation of extremal K¨ ahler metrics, J. Geom. Anal. 31 (2021), no. 3, 2970–3028. MR4225832 ↑1, 7, 8

work page 2021

[2] [1]

S. K. Donaldson, Scalar curvature and projective embeddings. II , Q. J. Math. 56 (2005), no. 3, 345–356. MR2161248 ↑1 [2] , Some numerical results in complex diﬀerential geometry , Pure Appl. Math. Q. 5 (2009), no. 2, Special Issue: In honor of Friedrich Hirzebruch. Part 1, 571–618. MR 2508897 ↑8

work page 2005

[3] [3]

Fine, Calabi ﬂow and projective embeddings , J

J. Fine, Calabi ﬂow and projective embeddings , J. Diﬀerential Geom. 84 (2010), no. 3, 489–523. With an appendix by Kefeng Liu and Xiaonan Ma. MR2669363 ↑4

work page 2010

[4] [4]

Finski, On the metric structure of section ring , arXiv preprint arXiv:2209.03853 (2022)

S. Finski, On the metric structure of section ring , arXiv preprint arXiv:2209.03853 (2022). ↑2 8

arXiv 2022

[5] [5]

, Submultiplicative norms and ﬁltrations on section rings , arXiv preprint arXiv:2210.03039 (2022). ↑2

arXiv 2022

[6] [6]

Griﬃths and J

P. Griﬃths and J. Harris, Principles of algebraic geometry , Wiley Classics Library, John Wiley & Sons, Inc., New York, 1994. Reprint of the 1978 original. MR1288523 (95d :14001) ↑3

work page 1994

[7] [7]

Hashimoto, Mapping properties of the Hilbert and Fubini-Study maps in K ¨ ahler geometry, Ann

Y. Hashimoto, Mapping properties of the Hilbert and Fubini-Study maps in K ¨ ahler geometry, Ann. Fac. Sci. Toulouse Math. (6) 29 (2020), no. 2, 371–389. MR4150546 ↑1, 2, 7

work page 2020

[8] [9]

On the Bergman kernels of holomorphic vector bundles

L. Lempert, On the Bergman kernels of holomorphic vector bundles , arXiv preprint arXiv:2109.08593 (2021). ↑2, 7

work page Pith review arXiv 2021