Equivalence of the pearly tree immersed Lagrangian Floer theory and the Hamiltonian immersed Lagrangian Floer theory

Zuyi Zhang

REVIEW 4 major objections 5 minor 13 references

Equivalence of the pearly tree immersed Lagrangian Floer theory and the Hamiltonian immersed Lagrangian Floer theory

T0 review · 4 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read Pearly-tree and Hamiltonian-flow constructions of immersed Lagrangian Floer chain groups are canonically identified at chain level, extending the unobstructed comparison to the obstructed case.

desk verdict A real idea and a plausible strategy, but the proof rests on an unproved and suspect invariance lemma, so the central identification is not established. read the letter →

arxiv 2501.04672 v1 pith:QR225C7B submitted 2025-01-08 math.SG math.DGmath.GT

Zuyi Zhang This is my paper · ORCID

classification math.SGmath.DGmath.GT MSC 53D4053D12

keywords LagrangianFloertheoryimmersedpearlytreediscslocalHamiltonianflowMorsefunctionWeinsteintubularneighbourhoodobstructedchain-levelidentification

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 aims to prove that two standard constructions of Lagrangian Floer theory for an immersed Lagrangian—one counting pearly trees, the other counting holomorphic discs with boundary on the Lagrangian and its image under a local Hamiltonian flow—are the same chain group. The identification is made directly on generators and boundary maps, so the result applies even when the Floer differential does not square to zero. This matters because the two formalisms have different strengths: pearly trees are Morse-theoretically computable, while Hamiltonian flows stay inside the standard Lagrangian-intersection setup.

What carries the argument

The machinery has three cooperating pieces. The Weinstein tubular neighbourhood theorem identifies a neighbourhood of the immersed Lagrangian with $T^*_\epsilon L$, so a Morse function $f$ on $L$ can be extended constantly in the fibre directions to a local Hamiltonian $H_1\in\mathcal{H}_{cf}$. Corollary 3.9 then reduces the comparison: for boundary-map contributions, the Morse trajectory parts of Type 2 and Type 3 pearly trees are constant, so the counts that must be matched are holomorphic discs. The transfer mechanism is Lemma 5.3, which claims that for a smooth family of transverse Lagrangian immersions, zero-dimensional moduli spaces of holomorphic discs are identified at the endpoints; this is what lets the boundary map of one theory be reinterpreted as the boundary map of the other. The four pearly tree types—Morse-to-Morse, Morse-to-self-intersection, self-intersection-to-Morse, and self-intersection-to-self-intersection—organize the domain of $\partial_P$.

What would settle it

Exhibit a smooth one-parameter family of Lagrangian immersions $L_t$, all transverse to a fixed $L$, such that the zero-dimensional moduli spaces for $L_0$ and $L_1$ have the same boundary intersection points but different signed counts (for example, a disc bubbles off and reappears with reversed orientation). Lemma 5.3 predicts equal counts; one such family would disprove the main theorem.

Watch

Extended reading notes

Core claim

The central discovery is that the pearly-tree immersed Lagrangian Floer chain group $(\mathrm{CF}_P(L), \partial_P)$ and the Hamiltonian immersed Lagrangian Floer chain group $(\mathrm{CF}_H(L, L_{\phi_\epsilon}), \partial_H)$ are canonically identified. Starting from a Morse function $f$ on the domain $L$ of a Lagrangian immersion $g:L\looparrowright M$, the paper extends $f$ constantly along the fibres of the Weinstein tubular neighbourhood to obtain a local Hamiltonian function $H_1$; the small-time fixed points of its flow are exactly the critical points of $f$, and self-intersections contribute matching generators in both theories. The boundary maps agree type by type: Type 1 Morse trajectories become holomorphic strips swept out by the flow, and Types 2–4, which reduce to holomorphic discs by Corollary 3.9, are matched by moving one boundary Lagrangian with the flow. In the converse direction, any non-degenerate local Hamiltonian flow is compared, through the constant extension of its restriction $H|_L$, to the pearly-tree group of $H|_L$. This establishes a chain-level identification that does not require $\partial^2=0$.

Load-bearing premise

Everything hinges on Lemma 5.3, which says that isolated holomorphic discs counted by the boundary map persist uniquely as one boundary Lagrangian moves through a smooth transverse family; if that persistence fails, the identification of boundary maps collapses.

Editorial extensions

If this is right

A computation of the pearly-tree boundary map $\partial_P$ automatically gives the Hamiltonian boundary map $\partial_H$ for the corresponding local Hamiltonian perturbation, and conversely.
The identification is canonical for all sufficiently small $\epsilon$, so the small-time parameter, the scaling factor $a$, and the regular almost complex structure do not change the resulting chain group.
The chain-level statement covers obstructed cases where $\partial^2 \neq 0$; the earlier unobstructed comparison arises as the special case where the complexes are genuine chain complexes.
In the two surface examples, the same boundary maps are obtained from Hamiltonian intersection diagrams and from pearly-tree diagrams, showing the correspondence diagrammatically.

Reading between the lines

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

If Lemma 5.3 holds up under rigorous checking, the same persistence mechanism should prove invariance of the Hamiltonian immersed Floer chain group under any sufficiently small Hamiltonian isotopy of one Lagrangian, not only the constant extensions considered here.
The chain-group identification suggests that higher-order structures—such as the $A_\infty$ operations or bounding cochains used in obstructed theory—could be transferred between the pearly-tree and Hamiltonian models; the paper does not construct this transfer.
On surfaces, the equivalence converts Hamiltonian intersection-counting into a Morse-theoretic picture reading directly off an immersed curve diagram, which may simplify explicit computations in examples.

Share X Bluesky LinkedIn Reddit HN

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

Zuyi Zhang's paper tries to show that the pearly tree immersed Lagrangian Floer chain group and the Hamiltonian immersed Lagrangian Floer chain group are the same object, not just quasi-isomorphic, and without assuming the Floer differential squares to zero. That is a genuine extension of Alston-Bao, and the paper deserves credit for stating the chain-level identification explicitly. The generator bijection via the constant extension of the Morse function is clear, and the calculation in Theorem 5.2 Case 1 showing that gradient trajectories become holomorphic discs under the local Hamiltonian flow is a nice, explicit check. The examples on surfaces, although they are in R^2 rather than a closed symplectic manifold, do show the two boundary maps matching in concrete cases.

The soft spot is Lemma 5.3, and it is load-bearing. The lemma claims that for any transverse one-parameter family of Lagrangian immersions L_t, the 0-dimensional holomorphic disc counts with boundary in L_0 and L equal those with boundary in L_1 and L. The proof asserts that the parametrized Cauchy-Riemann operator has 0 as a regular value and therefore its zero set is a smooth family of points, but neither the regularity of the parametrized operator nor the properness of the projection to [0,1] is shown. The 'path' J_λ is actually chosen as a loop J_0=J_1, which is not the right setup for a continuation, and orientations are never mentioned. In standard Floer theory a one-parameter family gives at best a chain map, and the counts can change by wall-crossing or bubbling. Without smallness or exactness hypotheses, the lemma is not a standard fact and is likely false as stated. Both directions of the main theorem reduce to this lemma: Theorem 5.2 Case 1 uses it to relate the rescaled Hamiltonian aH1 to H1, and Theorem 5.4 Step 2 uses it for the general Hamiltonian H. So the central identification is unsupported as written.

There are smaller issues: Theorem 2.5 cites the author's own preprint [13] for the core transversality result, and the Riemann mapping justification is terse because the boundary conditions do not obviously carry over from R×[0,1] to the disc. The paper also silently assumes orientations on the moduli spaces, which are needed even to define the boundary maps over Z.

For someone working in immersed Lagrangian Floer theory, this is a relevant paper to know about, but it is not ready to be used as a reference. I do not think this is a sloppy paper by an unserious author; the strategy is coherent and many pieces are correct. But the omissions in Lemma 5.3 are the analytic heart of the proof, not cosmetic details. I would send it to a referee because the question is real and the approach might be repairable—the author should either prove the lemma under the small-epsilon hypotheses actually used (the family is C^0 close to L, coming from a short Hamiltonian flow) or replace the claimed equality of counts with a continuation argument and see what remains. As it stands, I wouldn't cite it, and if I were the editor I'd expect the report to come back as 'reject, but there's a viable project here.'

Referee Report

4 major / 5 minor

Summary. The paper claims a chain-level equivalence between two versions of immersed Lagrangian Floer theory for a closed symplectic manifold: the pearly-tree complex generated by Morse critical points and self-intersections, and the Hamiltonian-immersion complex generated by intersections of L with a local Hamiltonian flow. Theorem 1.1 states that, for a Morse function f on a Lagrangian immersion L and a regular almost complex structure, a constant-in-fiber extension of f produces a local Hamiltonian flow whose Floer complex is identified with the pearly-tree complex, and conversely that any non-degenerate local Hamiltonian flow gives a Floer complex identifiable with the pearly-tree complex defined by its restriction. The proof proceeds by constructing explicit generator bijections and then identifying boundary maps through a parametrized-moduli-space lemma (Lemma 5.3) and an intermediate constant-extension Hamiltonian flow. Two planar examples are computed to illustrate the claimed equality of boundary maps.

Significance. If the main theorem were correct, it would give a canonical chain-level identification between two widely used flavors of immersed Lagrangian Floer theory, extending the Alston–Bao quasi-isomorphism to the obstructed case. The paper's explicit generator bijections and its two worked surface examples are useful and clearly presented. However, the central equivalence is not established in the manuscript: the proof of Lemma 5.3, which is the hinge of both directions of Theorem 1.1, omits the essential compactness, regularity, and orientation arguments, and the lemma is not a standard fact in the stated generality. Because both directions of the main theorem reduce to this lemma, the paper's central claim is unsupported as written.

major comments (4)

[Section 5, Lemma 5.3] The proof of Lemma 5.3 asserts that the parametrized Cauchy–Riemann operator has 0 as a regular value and that its zero set is a smooth family of points. This does not follow from the fiberwise regularity of the operators \bar\partial_{J_t}; the total linearized operator must be surjective, and the proof gives no argument for this. Compactness is also not addressed: the parametrized zero set could have boundary at a bubbling solution, which would destroy the claimed identification of the 0-dimensional moduli spaces at t=0 and t=1. Since Theorem 5.2 Case 2 and Theorem 5.4 Step 2 both reduce to this lemma, the boundary-map identification in the main theorem is not established.
[Section 5, Lemma 5.3 and Section 2, Theorem 2.8] Lemma 5.3 chooses a regular path J_\lambda with J_0=J_1, i.e., a loop, but the parametrized problem needs a path from J_0 to J_1 to compare the ends of the family; a loop does not define the required continuation. Moreover, Theorem 2.8, which is cited for the existence of such a regular path, is itself only sketched: the space W^{1,p}_{x^{[0,1]}_\pm}(D,M) is asserted to be a Banach manifold even though the Lagrangian boundary conditions vary with t, and the conclusion that a generic based loop has J_t \in J_{\rm reg} for all t is stronger than what the given Sard–Smale argument would imply.
[Definitions 3.7 and 4.6] The boundary maps in both theories are defined as sums #M(x,y)y, but no orientations or signs are introduced. In immersed Lagrangian Floer theory, even in the obstructed case, the differential is a signed count and an identification of boundary maps must compare signs. The proofs in Section 5 never discuss orientations of the moduli spaces, so the claimed equalities of boundary maps are not well-defined as stated.
[Section 5, Theorem 5.4, Step 1] The identification of generators in the converse direction uses the path P_{s,t}=\phi_{(1-s)t}\circ\phi^1_{st} and proves a local transversality statement for its tangent map. This does not by itself establish that the preimage P^{-1}_{\cdot,t}(L) is a compact 1-manifold whose boundary components are in bijection with the two generator sets, nor that the ordered-pair structure at self-intersections is preserved. A global argument for the fiber-product generalized intersections is needed; the compactness of L alone does not provide it.

minor comments (5)

[Throughout] The manuscript contains numerous typographical errors and misspellings, such as 'definde', 'Lagarangian', 'morse' (should be 'Morse'), 'manfold', 'theorme', 'Bananch', 'Consiquently', 'molduli', 'Hamiltonain', and 'immsersed'.
[Section 2, Definition 2.3] The notation u|_{\partial D-(\pm1,0)} is confusing; the boundary of the disc minus the two marked points is meant, but the notation should be clarified, e.g., as \partial D\setminus\{(1,0),(-1,0)\}.
[Section 6, Example 6.1] The displayed boundary-map formulas use arrows with unusual separators such as 'd // j // ...' which are hard to parse as text; the diagrams should be redrawn or the notation should be explained more fully.
[Section 6, Example 6.2] The text near Figure 7 contains a stray character '这' that appears to have been left in from editing.
[Theorem 1.1] The statement of the converse direction says 'canonical identification' without specifying whether the identification depends on the choice of Weinstein neighborhood or on the regular almost complex structure; the nature of the canonicity should be clarified.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the two chain complexes are defined independently, the identification is argued through an explicit Hamiltonian-flow construction, and the cited self-result is a separate transversality theorem rather than a restatement of the main theorem.

full rationale

Walking the derivation chain, I find no step in which a claimed output is equivalent by construction to its input. The pearly tree chain group (Definitions 3.4–3.7) is generated by critical points and ordered self-intersections of the immersed Lagrangian, with boundary map counting pearly tree discs; the Hamiltonian chain group (Definitions 4.4–4.6) is generated by generalized intersections of L and L_{φ_ε}, with boundary map counting holomorphic discs with boundary in the two Lagrangians. Neither definition invokes the other or the theorem. In the forward direction, H1 is explicitly chosen as the constant fiber extension of f; Step 1 identifies generators by a direct vector-field calculation, and Case 1 converts Morse trajectories into holomorphic strips by the explicit equation ∂_s φ^1_t(u(s)) + J∂_t φ^1_t(u(s)) = 0, so the identification there is computed rather than assumed. The remaining cases pass through Lemma 5.3, an invariance statement for 0-dimensional moduli spaces under a family of transverse Lagrangian boundary conditions. Lemma 5.3 does not mention pearly trees, Morse functions, or the chain-group identification; it claims only that disc counts for two Lagrangians connected by a transverse family agree. Thus the main theorem reduces to a separate, though under-proved, invariance lemma, not to its own conclusion. The only self-citation is Theorem 2.5, imported from the author's prior preprint [13]; this is a parameter-free regularity theorem for the Cauchy-Riemann operator with Lagrangian boundary conditions, distinct from the equivalence being proved, so it functions as external support rather than a circular premise. There are no fitted parameters, no data subsets, and no generator identified with itself by definition. The proof gap in Lemma 5.3—missing compactness, properness, and orientation arguments—is a soundness concern, not a circularity.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The paper has no fitted data constants, but it depends on several smallness parameters whose existence is asserted. The five axioms listed are imported from prior literature or the author's own preprint without proof in this manuscript; the last two are silently assumed and are especially critical in the obstructed setting.

free parameters (3)

epsilon (Weinstein neighborhood radius / flow time) = sufficiently small, no explicit bound
Both directions of Theorem 1.1 require ε > 0 small enough for the Hamiltonian immersion L_{φ_ε} to be transverse to L and for the identifications to hold. No effective bound is given.
a (scaling factor for H1) = small positive real, unspecified
In Theorem 5.2 Case 1, H1 is replaced by aH1 so that the C^2 bound in Equation 2 holds. The proof asserts existence of a and then transfers the result back via Lemma 5.3.
epsilon1 (C^2 smallness bound) = small enough, unspecified
Equation 2 imposes |H1| + |∂H1/∂x| + |∂^2H1/∂x^2| ≤ ε1. The bound is needed to invoke Floer's argument [8]; its compatibility with the given J and H is not checked.

assumptions (5)

domain assumption Transversality for immersed Lagrangians: the Cauchy-Riemann operator has surjective linearization for an open dense set of compatible almost complex structures (Theorem 2.5).
Used to define regular J and to ensure counted moduli spaces are manifolds. The proof is delegated to the author's earlier preprint [13].
standard math Sard-Smale theorem applies to the based loop space of almost complex structures (Theorem 2.8).
Used to claim generic paths J_t have every time regular. The loop space is not given a Banach manifold structure, so the application is not justified as written.
domain assumption Gromov compactness holds for moduli spaces of J-holomorphic discs with immersed Lagrangian boundary conditions, including in the obstructed setting.
Invoked to make 0-dimensional moduli spaces finite (Corollary 2.7). No energy bounds or bubbling analysis is provided, which matters because the paper works without assuming the boundary map squares to zero.
domain assumption Floer's Lemma 5.1 in [8] implies every sufficiently small holomorphic disc between fixed points of the local Hamiltonian flow has the form φ^1_t(u(s)).
Central to Case 1 of Theorem 5.2 to convert holomorphic discs to Morse trajectories. The argument is cited, not reproduced, and the smallness assumption is only removed by rescaling.
domain assumption Moduli spaces counted in ∂P and ∂H are oriented and their 0-dimensional parts are compact.
Definitions 3.7 and 4.6 count #M(x,y); orientations are never defined and compactness is never proved. These facts are prerequisites for integer-valued chain group boundary maps.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Equivalence of the pearly tree immersed Lagrangian Floer theory and the Hamiltonian immersed Lagrangian Floer theory." pith.science (2026). https://pith.science/paper/QR225C7B

@misc{pith2026250104672,
  author       = {Pith},
  title        = {Pith review of: Equivalence of the pearly tree immersed Lagrangian Floer theory and the Hamiltonian immersed Lagrangian Floer theory},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/QR225C7B}},
  note         = {Machine review of arXiv:2501.04672}
}

read the original abstract

The goal of this paper is to prove an equivalence relation between the immersed Lagrangian Floer theory, defined using pearly tree discs, and local Hamiltonian flows, i.e., Hamiltonian flows performed in the Weinstein tubular neighborhood. This is a generalization of Alston-Bao's work.

Figures

Figures reproduced from arXiv: 2501.04672 by the authors.

**Figure 2.** The four types of pearly tree discs. The left dots are inputs and the [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗

**Figure 3.** The left picture is the holomorphic disc for Hamiltonian immersed La [PITH_FULL_IMAGE:figures/full_fig_p019_3.png] view at source ↗

Figures from the paper (2 more)

**Figure 4.** Figure 4: Hamiltonian version k f e g h j a b c i d [PITH_FULL_IMAGE:figures/full_fig_p020_4.png]

**Figure 6.** Figure 6: Hamiltonian version f e d g h j c i b a 这 [PITH_FULL_IMAGE:figures/full_fig_p021_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

13 extracted references · 11 canonical work pages

[13]

Construction of holomorphic quilts in cartesian product of closed surfaces

Zuyi Zhang. Construction of holomorphic quilts in cartesian product of closed surfaces. arXiv preprint arXiv:2409.19744 , 2024. 22

arXiv 2024
[1]

On the Fukaya categories of higher genus surfaces

Mohammed Abouzaid. On the Fukaya categories of higher genus surfaces. Advances in Mathematics , 217(3):1192–1235, 2008

work page 2008
[2]

Immersed Lagrangian Floer theory

Manabu Akaho and Dominic Joyce. Immersed Lagrangian Floer theory. Jour- nal of differential geometry , 86(3):381–500, 2010

work page 2010
[3]

Immersed lagrangian floer cohomology via pearly trajectories

Garrett Alston and Erkao Bao. Immersed lagrangian floer cohomology via pearly trajectories. Journal of Geometry and Physics , 169:104335, 2021

work page 2021
[4]

Quantum structures for lagrangian submani- folds

Paul Biran and Octav Cornea. Quantum structures for lagrangian submani- folds. arXiv preprint arXiv:0708.4221 , 2007. 20 f e d g h j i c d a Figure 6: Hamiltonian version f edg h j icb a 这 Figure 7: Pearly tree version

arXiv 2007
[5]

A lagrangian quantum homology

Paul Biran and Octav Cornea. A lagrangian quantum homology. New per- spectives and challenges in symplectic field theory , 49:1–44, 2009

work page 2009
[6]

Introduction to the h-Principle

Yakov Eliashberg, Nikolai M Mishachev, and Susumu Ariki. Introduction to the h-Principle. Number 48. American Mathematical Soc., 2002

work page 2002
[7]

Morse theory for lagrangian intersections

Andreas Floer. Morse theory for lagrangian intersections. Journal of differ- ential geometry, 28(3):513–547, 1988

work page 1988

Show all 13 references

[8]

Witten’s complex and infinite-dimensional morse theory

Andreas Floer. Witten’s complex and infinite-dimensional morse theory. Jour- nal of differential geometry , 30(1):207–221, 1989

1989
[9]

Lagrangian intersection Floer theory: anomaly and obstruction

Kenji Fukaya, Yong-Geun Oh, Hiroshi Ohta, and Kaoru Ono. Lagrangian intersection Floer theory: anomaly and obstruction. Part i, volume 46 of AMS. IP Studies in Advanced Mathematics. American Mathematical Society, Providence, RI, 2, 2009

2009
[10]

J-holomorphic curves and symplectic topology, volume 52

Dusa McDuff and Dietmar Salamon. J-holomorphic curves and symplectic topology, volume 52. American Mathematical Soc., 2012

2012
[11]

Relative floer and quantum cohomology and the symplectic topology of lagrangian submanifolds

Yong-Geun Oh. Relative floer and quantum cohomology and the symplectic topology of lagrangian submanifolds. Contact and symplectic topology , 1994

1994
[12]

Invariance of immersed floer cohomology under maslov flows

Joseph Palmer and Chris Woodward. Invariance of immersed floer cohomology under maslov flows. Algebraic & Geometric Topology, 21(5):2313–2410, 2021. 21

2021

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.

[1] [13]

Construction of holomorphic quilts in cartesian product of closed surfaces

Zuyi Zhang. Construction of holomorphic quilts in cartesian product of closed surfaces. arXiv preprint arXiv:2409.19744 , 2024. 22

arXiv 2024

[2] [1]

On the Fukaya categories of higher genus surfaces

Mohammed Abouzaid. On the Fukaya categories of higher genus surfaces. Advances in Mathematics , 217(3):1192–1235, 2008

work page 2008

[3] [2]

Immersed Lagrangian Floer theory

Manabu Akaho and Dominic Joyce. Immersed Lagrangian Floer theory. Jour- nal of differential geometry , 86(3):381–500, 2010

work page 2010

[4] [3]

Immersed lagrangian floer cohomology via pearly trajectories

Garrett Alston and Erkao Bao. Immersed lagrangian floer cohomology via pearly trajectories. Journal of Geometry and Physics , 169:104335, 2021

work page 2021

[5] [4]

Quantum structures for lagrangian submani- folds

Paul Biran and Octav Cornea. Quantum structures for lagrangian submani- folds. arXiv preprint arXiv:0708.4221 , 2007. 20 f e d g h j i c d a Figure 6: Hamiltonian version f edg h j icb a 这 Figure 7: Pearly tree version

arXiv 2007

[6] [5]

A lagrangian quantum homology

Paul Biran and Octav Cornea. A lagrangian quantum homology. New per- spectives and challenges in symplectic field theory , 49:1–44, 2009

work page 2009

[7] [6]

Introduction to the h-Principle

Yakov Eliashberg, Nikolai M Mishachev, and Susumu Ariki. Introduction to the h-Principle. Number 48. American Mathematical Soc., 2002

work page 2002

[8] [7]

Morse theory for lagrangian intersections

Andreas Floer. Morse theory for lagrangian intersections. Journal of differ- ential geometry, 28(3):513–547, 1988

work page 1988