Birkhoff normal form via decorated trees

Jacob Armstrong-Goodall; Yvain Bruned

REVIEW 3 major objections 6 minor 30 references

Birkhoff normal form via decorated trees

T0 review · 3 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read For Hamiltonian PDEs, the Birkhoff normal form at every order is an explicit sum over decorated planar binary trees, and each symplectic generator is a tree sum too.

desk verdict A genuinely new explicit decorated-tree formula for the Birkhoff normal form, with a plausible but compressed proof; the stress-test coefficient objection does not hold up. read the letter →

arxiv 2505.04547 v1 pith:2OU5LDWP submitted 2025-05-07 math.AP math.DSmath.RA

Jacob Armstrong-Goodall , Yvain Bruned This is my paper · ORCID

classification math.APmath.DSmath.RA MSC 37K5537J4035Q55

keywords BirkhoffnormalformdecoratedtreesHamiltonianPDEsPoissonbracketssymplectictransformationsresonancecubicSchrödingerequationTaylorexpansionalongflows

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

The paper claims that the Birkhoff normal form reduction for Hamiltonian PDEs can be made fully explicit: after $m$ symplectic transformations, the Hamiltonian truncated at order $2\ell$ is $H_0$ plus three finite sums of decorated planar binary trees. The same ansatz expresses each symplectic generator $F_i$ as a tree sum over non-resonant iterated Poisson brackets. If the claim is right, a recursive procedure that normally generates an ever-growing list of terms becomes a combinatorial enumeration that can be written down to any order. The authors demonstrate the construction on the cubic Schr\"odinger equation, but the tree rules are independent of the equation.

What carries the argument

The load-bearing object is a planar binary rooted tree whose nodes carry one of four decorations: $k$ (the quadratic Hamiltonian $H_0$), $\circ$ (the quartic interaction $H_1$), $n$ (a non-resonant bracket divided by its phase), and $r$ (a resonant bracket). Two constraints organize the shape: $n$-nodes appear only on right branches, $r$- and $\circ$-nodes only on left branches, and $k$ appears only as a leaf below a non-root $\circ$. The map $\Pi$ recursively converts a tree into the iterated Poisson bracket it names, and the coefficient $S(T)$, defined recursively from factorial path counts, converts each bracket into its Taylor coefficient in the flow expansion. Assumption 1 fixes which trees are legal, and Definition 4.2 groups them into the classes appearing in Theorem 4.3.

What would settle it

Compute $H_m^\ell$ directly by the recursive algorithm for the cubic Schr\"odinger equation on $\mathbb{T}^1$ or $\mathbb{T}^2$ at an order beyond the worked examples, for instance $m=2$, $\ell=5$, expand every iterated Poisson bracket in Fourier modes, and compare the result with the corresponding tree sum over $\mathcal{T}_r^{<4}$, $\mathcal{T}_\circ^4$, and $\mathcal{T}_\circ^{4,5}$; any unmatched or extra term, or any mismatched coefficient, would disprove the formula.

Watch

Extended reading notes

Core claim

The central claim is Theorem 4.3: for integers $1 \le m < \ell$, the $m$-th Birkhoff normal form truncated at order $2\ell$, written $H_m^\ell = (H \circ F_1 \circ \cdots \circ F_m)_\ell$, equals $H_0$ plus the three tree sums over $\mathcal{T}_r^{<m+2}$, $\mathcal{T}_\circ^{m+2}$, and $\mathcal{T}_\circ^{m+2,\ell}$, with each tree $T$ contributing $\Pi_T / S(T)$. Moreover $F_i = \sum_{T \in \mathcal{T}_n^{i+1}} \Pi_T / S(T)$ and $\{H_0,F_i\} = -\sum_{T \in \mathcal{T}_\circ^{i+1}} (\Pi_T)^{\mathrm{non-res}} / S(T)$. The tree classes are the decorated-tree encodings of resonant brackets, degree-matched outer brackets, non-resonant generators, and higher-degree brackets; $\Pi$ maps a tree to the iterated Poisson bracket it encodes, and $S(T)$ supplies the factorial Taylor coefficient. The proof is by induction on $m$, matching each recursive step of the symplectic-reduction algorithm to a tree-building rule.

Load-bearing premise

The argument rests on the premise that the tree-formation rules in Assumption 1 and Definition 4.2 capture every term the recursive algorithm can produce, with no missing or spurious trees.

Editorial extensions

If this is right

At any fixed order, the normal form can be written down by enumerating the allowed decorated trees in the three classes, with no need to solve implicit equations for the symplectic transformations.
Each symplectic generator $F_i$ is itself a finite sum over $\mathcal{T}_n^{i+1}$, so the recursive definition of the transformations becomes a closed formula.
The tree-sum representation separates resonant contributions ($r$-rooted trees) from non-resonant corrections ($\circ$- and $n$-rooted trees), making the cancellation mechanism of each step transparent.
Because the tree rules are independent of the concrete equation, the same ansatz applies to any Hamiltonian PDE with the same algebraic structure; the cubic Schr\"odinger equation serves only as the worked example.
Higher-order truncations are organized by the growth of allowed tree sizes, giving direct combinatorial bookkeeping for remainder terms in the flow expansion.

Reading between the lines

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

Beyond the paper, the tree enumeration is algorithmic: a computer algebra system could generate $H_m^\ell$ for specified $m,\ell$ by producing all trees in $\mathcal{T}_r^{<m+2}$, $\mathcal{T}_\circ^{m+2}$, and $\mathcal{T}_\circ^{m+2,\ell}$, then evaluating $\Pi$ and $S$; the paper gives the rules but does not discuss implementation.
A speculative extension is that the tree spaces could carry a natural algebraic product, so that composing normal-form transformations corresponds to a tree operation; this would connect the formal normal form to numerical integration schemes for Hamiltonian PDEs.
The formula may make it possible to compare different dispersive equations at the level of their decoration rules, isolating which combinatorial features control resonance clustering and long-time energy exchange; that comparison is not attempted here.
A concrete test of the ansatz would be to verify the next-order identity $H_2^5$ for the cubic Schr\"odinger equation against a direct iterated-bracket computation; the paper stops at $H_2^4$.

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

Quick take: this is a genuinely new combinatorial object. The authors give an explicit decorated-tree ansatz for the Birkhoff normal form at arbitrary order, for Hamiltonian PDEs, with NLS as the worked example. The tree formalism is close to their earlier arborification work and to Magnus-expansion trees, but the specific ansatz for the normal form itself—the r, o, n, k decorations and the S(T) coefficients—appears to be new. That is the paper's real contribution, and it is useful for anyone who wants to see the recursive normal-form computation made fully explicit. The small examples compute out, and the recursive definition of S(T) is internally consistent. I checked the alleged contradiction in the stress test: it does not hold. The conditional in (4.2) depends on whether T1 is an o-rooted node whose right subtree matches T2 in size; it does not require the left child to be o. So the worked example S(r(o(k,n),n))=2 is exactly what the recursion gives, and Proposition 4.1 is not contradicted either. The citation pattern is fine; the self-citation to [10] is background, and the main formula is not lifted from it.

Where the paper is genuinely soft: Theorem 4.3's proof leans heavily on unproved set identifications. In the base case m=1, the claim that n-decorated nodes must be leaves is valid, but only after combining constraint (c) (k leaves attach only to o nodes) with the degree bound; the text just says "one first notices." In the induction step, "it is easy to see that it is a full description" is doing real work. I do not see a contradiction, but it is not a proof-complete argument as written. There are also typos in the example: T^{3,4}_o is listed but unused, and the same symbol T^3_o appears twice with different meanings in the proof. Those are fixable. The expansion is purely formal—no analytic estimates are claimed—so that is not a flaw if the theorem is read as a formal normal-form statement, but it should be stated more carefully.

Who is this for: people working on Birkhoff normal form combinatorics, resonance structures, or numerical integrators for Hamiltonian PDEs. Not for KAM experts expecting new stability results. It deserves a serious referee. I would send it out, asking the authors to fill in the induction details and clean up the typos. The core formula is worth publishing.

Referee Report

3 major / 6 minor

Summary. The paper introduces a decorated planar binary tree formalism to encode the iterated Poisson brackets that arise in the Birkhoff normal form reduction of Hamiltonian PDEs, with the cubic NLS as a running example. The main result, Theorem 4.3, asserts that the m-th Birkhoff normal form truncated at order 2ℓ, and the symplectic generators F_i, can be written as explicit sums over decorated trees, divided by combinatorial coefficients S(T). The proof is by induction, and small cases (m = 1, 2) are worked out in detail.

Significance. If correct, the result would give an explicit, order-by-order combinatorial description of the Birkhoff normal form, a quantity that is usually defined only recursively. The decorated tree framework is original and potentially useful for understanding the combinatorial structure of resonances. The manuscript also gives hand-checkable low-order examples, which is a strength. However, the central coefficient S(T) is defined inconsistently with the paper's own examples and with Proposition 4.1, so Theorem 4.3 is not well-defined as written; this must be repaired before the main claim can be assessed.

major comments (3)

[Section 4, Eq. (4.2) and the example after it] The recursive definition of S_j in Eq. (4.2) is internally inconsistent. For r- and n-rooted trees the rule is S_j(r T1 T2) = S_j(n T1 T2) = (j+1) S_0(T1) S_0(T2). For the displayed example T = r(∘(k,n), n), this gives S_0(T) = S_0(∘(k,n)) S_0(n). The subtree ∘(k,n) falls into the 'otherwise' branch because its left child is k, not an ∘ node, so S_0(∘(k,n)) = S_0(k) S_0(n) = 1. Hence Eq. (4.2) yields S(T) = 1, whereas the text states S(T) = 2. The same rule gives S = 6 for the stated example with a right subtree T3 satisfying S(T3) = 3!, not the claimed 2!·3! = 12. Proposition 4.1 is also contradicted: for p = 2 with T1 = ∘ leaf and T2 = n leaf, the proposition's formula S(T) = p!∏ S(T_i) gives S(∘(T1,T2)) = 2, while Eq. (4.2) gives 1 via the otherwise branch. Since every summand in Theorem 4.3 is divided by S(T), and the proof of that theorem invokes Proposition 4.1 for exactly such trees, the tree sums in Theorem 4.3 are not computable as written.
[Section 4, proof of Theorem 4.3, base case] The proof asserts without adequate justification that T^{3,ℓ}_∘ consists only of left combs with n-nodes as leaves. The text writes the forbidden condition as '|T1|+|T2|-2 ≥ 2m = 6' for m = 1, although 2m = 2, and the claimed characterization does not follow from Assumption 1 and Definition 4.2 without a case analysis that is not supplied. Because Theorem 4.3 is an exact set equality, this unproved identification is load-bearing.
[Section 4, inductive step of Theorem 4.3] The inductive step contains the assertion that the set T^{m+3,ℓ}_◦ \ T^{m+2,m+3}_◦ is fully described by the comb trees obtained from Proposition 4.1. This is stated with the phrase 'It is easy to see' and is not proved. The argument relies on an order relation on subtree sizes that is only informally justified via Definition 3.3. Since the theorem's conclusion is an exact combinatorial description, this step needs a precise proof rather than an appeal to intuition.

minor comments (6)

[Throughout] There are frequent typos, including 'Hamiltonain' for 'Hamiltonian' and 'corrolary' for 'corollary'; the paper would benefit from a careful proofreading pass.
[Section 2.1 and Section 3] The symbol H0 is used both for the full initial Hamiltonian H0 + H1 and for the quadratic part alone (e.g., in Eqs. (2.5), (3.1), and the bracket computations). This notational collision is confusing and should be resolved.
[Definition 3.3] The definition of the sequences s_m^n and the coefficients c_z is vague; the example clarifies the case n = 3, but a precise combinatorial definition of c_z in terms of repetition counts should be stated.
[Definition 4.2] The text reads 'One set the following subsets' and 'form∈N∗'; these should be 'One sets' and 'for m∈N∗'.
[Section 4, base case of Theorem 4.3] In the base-case proof, the threshold '2m = 6' is arithmetically wrong for m = 1; it should be 2m = 2, or the formula should be written in terms of m + 2 if that was the intent.
[Section 4, examples and notation] The plain-text tree notation, e.g., r(∘(k,n), n), is readable in principle, but a few figures or a more systematic bracketing convention would make the large tree sets in Definition 4.2 and Theorem 4.3 much easier to verify.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the main theorem is an inductive bookkeeping reformulation of the standard normal-form recursion, with no fitted parameters and no load-bearing self-citations.

full rationale

The paper's derivation chain is self-contained: Theorem 4.3 is proved by induction from Definition 2.3 (Hamiltonian Taylor expansion), the explicit solution of the homological equation {H0,Fi} = -H_non-res_i, and the recursively defined tree map Π and tree spaces in Section 4. There are no fitted parameters, no empirical benchmarks, and no load-bearing self-citations; the reference to the author's earlier decorated-tree work [10] appears only in the Introduction as motivation ('The former work motivated the decorated tree approach by one of the authors in [10]') and is never used in the proof. The proof checks set equalities between algorithmically generated iterated brackets and the tree spaces, which is a genuine representational theorem rather than a definitional equivalence. I therefore find no circular step. I do flag a serious internal inconsistency that is a correctness risk rather than a circularity: Definition (4.2) computes S0(r(∘(k,n),n)) = 1 by its recursive rule, while the surrounding example and Proposition 4.1 require S(T) = 2 for the coefficient of {{H0,F1},F1}; since every term in Theorem 4.3 is divided by S(T), the displayed formula is not computable from (4.2) as written.

Assumptions & free parameters 0 free parameters · 4 assumptions · 1 invented entities

No physical parameters are fitted and no numerical constants are introduced. The paper's input is a formal Hamiltonian PDE, and its main new content is a combinatorial encoding; the principal unstated premise is that the tree constraints fully and faithfully represent the normal form expansion.

assumptions (4)

domain assumption The phase filter F_{n,N}(k) = 1_{R^c_n(k)}/(2Φ_n(k)) is well-defined and formal Fourier sums can be manipulated termwise.
Used in Definition 2.5 and in the computation {H0,F1} = -H_non_res; requires a nonzero denominator on the non-resonant set and no convergence issues.
domain assumption All Taylor expansions along Hamiltonian flows are performed formally, without remainder estimates.
Definition 2.3 and the iterative construction in Section 3 treat infinite series and Poisson brackets of infinite sums without regularity assumptions.
ad hoc to paper The decorated-tree constraints in Assumption 1 and Definition 4.2 are complete: each term from the normal form algorithm corresponds to exactly one allowed tree.
This is the load-bearing combinatorial premise of Theorem 4.3; it is asserted rather than fully proved in the induction step.
standard math Standard properties of Poisson brackets and symplectic transformations are valid in the infinite-dimensional setting.
Section 2 introduces the Poisson bracket and Taylor expansion using standard symplectic geometry, without addressing the functional-analytic conditions required for infinite dimensions.

invented entities (1)

Decorated planar binary trees with node decorations k, ∘, n, r independent evidence
purpose: Bookkeeping device to encode iterated Poisson brackets and Taylor coefficients in the Birkhoff normal form
The translation map Π and the coefficient map S give a direct computational rule, so the formalism can be checked term by term; it is not used as an unexplained physical entity.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Birkhoff normal form via decorated trees." pith.science (2026). https://pith.science/paper/2OU5LDWP

@misc{pith2026250504547,
  author       = {Pith},
  title        = {Pith review of: Birkhoff normal form via decorated trees},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2OU5LDWP}},
  note         = {Machine review of arXiv:2505.04547}
}

read the original abstract

We derive an explicit tree based ansatz for the Birkhoff normal form up to any order in the context of Hamiltonian PDEs. To do so we make use of a tree based representation of iterated Poisson brackets to encode the nested Taylor expansions along flows of a sequence of symplectic transformations. As an example we consider the cubic Schr\"odinger equation.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

30 extracted references · 17 canonical work pages

[10]

Y. Bruned. Derivation of normal forms for dispersive PDEs via arborification. arXiv:2409.03642

arXiv
[1]

V. I. Arnol’d. Proof of a theorem of A. N. Kolmogorov on the preservation of condi- tionally periodic motions under a small perturbation of the Hamiltonian. Uspehi Mat. Nauk 18 (1963), no. 5(113), 13–40. doi:10.1070/RM1963v018n05ABEH004130

doi:10.1070/rm1963v018n05abeh004130 1963
[2]

D. Bambusi. Birkhoff normal form for some nonlinear PDEs Comm. Math. Phys. 234 (2003), no. 2, 253–285. doi:10.1007/s00220-002-0774-4

work page doi:10.1007/s00220-002-0774-4 2003
[3]

Bernier, S

J. Bernier, S. Blanes, F. Casas, A. Escorihuela-Tom´as. Symmetric-conjugate splitting methods for linear unitary problems . BIT 63 (2023), no. 4, Paper No. 58, 26 pp. doi:10.1007/s10543-023-00998-4

doi:10.1007/s10543-023-00998-4 2023
[4]

Bernier, E

J. Bernier, E. Faou, B. Gr ´ebert. Long time behavior of the solutions of NLW on the d-dimensional torus . Forum Math. Sigma 8 (2020), Paper No. e12, 26 pp. doi:10.1017/fms.2020.8. Decorated trees formulation 21

work page doi:10.1017/fms.2020.8 2020
[5]

Bernier, E

J. Bernier, E. Faou, B. Gr´ebert Rational normal forms and stability of small solutions to nonlinear Schr¨odinger equations. Ann. PDE 6 (2020), no. 2, Paper No. 14, 65 pp. doi:10.1007/s40818-020-00089-5

doi:10.1007/s40818-020-00089-5 2020
[6]

Bambusi, B

D. Bambusi, B. Gr ´ebert. Birkhoff normal form for partial differential equations with tame modulus. Duke Math. J. 135, no. 3, (2006), 507–567. doi:10.1215/ S0012-7094-06-13534-2

work page 2006
[7]

Bernier, B

J. Bernier, B. Gr´ebert. Birkhoff normal forms for Hamiltonian PDEs in their energy space. J. ´Ec. Polytech. Math. 9, (2022), 681–745. doi:10.5802/jep.193

doi:10.5802/jep.193 2022

Show all 30 references

[8]

Bourgain

J. Bourgain. Quasi-periodic solutions of Hamiltonian perturbations of 2D linear Schr¨odinger equations. Ann. of Math. (2) 148 (1998), no. 2, 363–439. doi:10.2307/ 121001

1998
[9]

I-method

J. Bourgain. A remark on normal forms and the “I-method” for periodic NLS. J. Anal. Math. 94 (2004), 125–157. doi:10.1007/BF02789044

2004 doi
[11]

Bruned, K

Y. Bruned, K. Schratz.Resonance based schemes for dispersive equations via decorated trees. Forum of Mathematics, Pi, 10, (2022), E2. doi:10.1017/fmp.2021.13

2022 doi
[12]

W. L. Craig, C. E. Wayne. Newton’s method and periodic solutions of nonlinear wave equations. Comm. Pure Appl. Math. 46 (1993), no. 11, 1409–1498. doi: 10.1002/cpa.3160461102

1993 doi
[13]

upside-down

J. Colliander, S. Kwon, T. Oh. A remark on normal forms and the “upside-down” I-method for periodic NLS: Growth of higher Sobolev norms . J. Anal. Math. 118, (2012), 55–82. doi:10.1007/s11854-012-0029-z

2012 doi
[14]

H. Cong, S. Li, Y. Sun, X. Wu. Birkhoff normal form and long time existence for d-dimensional generalized Pochhammer-Chree equation J. Stat. Phys. 192 (2025), no. 2, Paper No. 30, 27 pp. doi:10.1007/s10955-025-03409-w

2025 doi
[15]

Curry, K

C. Curry, K. Ebrahimi-Fard, B. Owren. The Magnus expansion and post-Lie algebras. Math. Comp. 89 (2020), no. 326, 2785–2799 doi:10.1090/mcom/3541

2020 doi
[16]

Gr ´ebert

B. Gr ´ebert. Birkhoff normal form and Hamiltonian PDEs . Partial differential equations and applications, 1–46, S ´emin. Congr., 15, Soc. Math. France, Paris. doi:hal-00022311v2
[17]

Z. Guo, S. Kwon, T. Oh. Poincar´e-Dulac normal form reduction for unconditional well-posedness of the periodic cubic NLS . Comm. Math. Phys. 322, no. 1, (2013), 19–48. doi:10.1007/s00220-013-1755-5

2013 doi
[18]

A. Iserles. Magnus expansions and beyond. Combinatorics and physics, 171–186, Contemp. Math., 539, Amer. Math. Soc., Providence, RI. doi:10.1090/conm/539/ 10634

doi
[19]

Iserles, S

A. Iserles, S. P. Nørsett. On the solution of linear differential equations in Lie groups. R. Soc. Lond. Philos. Trans. Ser. A Math. Phys. Eng. Sci. 357 (1999), no. 1754, 983–1019. doi:10.1098/rsta.1999.0362

1999
[20]

S. B. Kuksin. Hamiltonian perturbations of infinite-dimensional linear systems with imaginary spectrum . Funct. Anal. Appl. 21 (1987), no. 3, 22–37, 95. doi: 10.1007/BF02577134. Decorated trees formulation 22

1987 doi
[21]

S. B. Kuksin, J. P¨oschel. Invariant Cantor manifolds of quasi-periodic oscillations for a nonlinear Schr¨odinger equation. Ann. of Math. 143, no. 2, (1996), 149–179. doi:10.2307/2118656

1996 doi
[22]

Laurent, O

A.B. Laurent, O. Cosserat. Butcher series for Hamiltonian Poisson integrators through symplectic groupoids. arXiv:2503.05000

arXiv
[23]

W. Magnus. On the exponential solution of differential equations for a linear operator. Comm. Pure Appl. Math. 7 (1954), 649–673. doi:10.1002/cpa.3160070404

1954 doi
[24]

J. K. Moser. On invariant curves of area-preserving mappings of an annulus. Nachr. Akad. Wiss. G¨ottingen Math.-Phys. Kl. II 1962 (1962), 1–20

1962
[25]

H. Z. Munthe-Kaas, W. M. Wright. On the Hopf algebraic structure of Lie group integrators. Found. Comput. Math. 8 (2008), no. 2, 227–257 doi:10.1007/ s10208-006-0222-5

2008
[26]

T. Oh, P. Sosoe, N. Tzvetkov. An optimal regularity result on the quasi-invariant Gaussian measures for the cubic fourth order nonlinear Schr¨odinger equation. J. ´Ec. polytech. Math. 5, (2018), 793–841. doi:10.5802/jep.83

2018 doi
[27]

P´erez-Marco

R. P´erez-Marco. Convergence or generic divergence of the Birkhoff normal form.Ann. of Math. (2) 157 (2003), no. 2, 557–574. doi:10.4007/annals.2003.157.557

2003 doi
[28]

Krikorian

R. Krikorian. On the divergence of Birkhoff normal forms Publ. Math. Inst. Hautes ´Etudes Sci. 135 (2022), 1–181. doi:10.1007/s10240-022-00130-2

2022 doi
[29]

C. Sun, N. Tzvetkov.Quasi-invariance of Gaussian measures for the 3d energy critical nonlinear Schr¨odinger equation. arXiv:2308.12758

arXiv
[30]

C. E. Wayne. Periodic and quasi-periodic solutions of nonlinear wave equations via KAM theory. Comm. Math. Phys. 127 (1990), no. 3, 479–528 doi:10.1007/ BF02104499

1990

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.

[1] [10]

Y. Bruned. Derivation of normal forms for dispersive PDEs via arborification. arXiv:2409.03642

arXiv

[2] [1]

V. I. Arnol’d. Proof of a theorem of A. N. Kolmogorov on the preservation of condi- tionally periodic motions under a small perturbation of the Hamiltonian. Uspehi Mat. Nauk 18 (1963), no. 5(113), 13–40. doi:10.1070/RM1963v018n05ABEH004130

doi:10.1070/rm1963v018n05abeh004130 1963

[3] [2]

D. Bambusi. Birkhoff normal form for some nonlinear PDEs Comm. Math. Phys. 234 (2003), no. 2, 253–285. doi:10.1007/s00220-002-0774-4

work page doi:10.1007/s00220-002-0774-4 2003

[4] [3]

Bernier, S

J. Bernier, S. Blanes, F. Casas, A. Escorihuela-Tom´as. Symmetric-conjugate splitting methods for linear unitary problems . BIT 63 (2023), no. 4, Paper No. 58, 26 pp. doi:10.1007/s10543-023-00998-4

doi:10.1007/s10543-023-00998-4 2023

[5] [4]

Bernier, E

J. Bernier, E. Faou, B. Gr ´ebert. Long time behavior of the solutions of NLW on the d-dimensional torus . Forum Math. Sigma 8 (2020), Paper No. e12, 26 pp. doi:10.1017/fms.2020.8. Decorated trees formulation 21

work page doi:10.1017/fms.2020.8 2020

[6] [5]

Bernier, E

J. Bernier, E. Faou, B. Gr´ebert Rational normal forms and stability of small solutions to nonlinear Schr¨odinger equations. Ann. PDE 6 (2020), no. 2, Paper No. 14, 65 pp. doi:10.1007/s40818-020-00089-5

doi:10.1007/s40818-020-00089-5 2020

[7] [6]

Bambusi, B

D. Bambusi, B. Gr ´ebert. Birkhoff normal form for partial differential equations with tame modulus. Duke Math. J. 135, no. 3, (2006), 507–567. doi:10.1215/ S0012-7094-06-13534-2

work page 2006

[8] [7]

Bernier, B

J. Bernier, B. Gr´ebert. Birkhoff normal forms for Hamiltonian PDEs in their energy space. J. ´Ec. Polytech. Math. 9, (2022), 681–745. doi:10.5802/jep.193

doi:10.5802/jep.193 2022