Quantitative refinements of Hörmander's abstract Nash–Moser implicit function theorem: explicit smallness threshold, strong convergence, and uniform constants

Hörmander's five-page 1985 note derived an abstract Nash–Moser type implicit function theorem from Nash's telescoping-series iteration combined with nonlinear functional analysis, simplifying earlier treatments of Moser, Zehnder, and his own work on physical geodesy. The note's hypotheses and proof leave three quantitative gaps that it never resolves and never explicitly poses as problems: the structural constants appearing in the smoothing and interpolation estimates are not uniformly controlled as the two regularity indices approach one another; the existence theorem is stated only for data of 'sufficiently small norm' with no explicit threshold in terms of the structural constants; and the iteration converges only in a weak topology on a completion space, with no statement about strong convergence in the scale itself. Later work — most decisively Baldi and Haus (2017), building on Hörmander's 1976 scheme and 1990 weak-space reformulation — identified the weak completion with Besov spaces, and in a modified framework with an additional summability condition on the data obtained strong solutions and an explicit smallness threshold for Sobolev scales. The problem posed here is the surviving core: obtain these quantitative refinements within Hörmander's original abstract framework, or determine by counterexample that they are unattainable there.

Background

The classical implicit function theorem for a $C^1$ map between Banach spaces fails in the settings of differential geometry and analysis where the linearized operator loses regularity, i.e. where solving the linearized equation requires more differentiability of the data than the solution possesses. To overcome this, Nash (1956), in his isometric embedding theorem for Riemannian manifolds, introduced an iteration in which each Newton-type step is regularized by smoothing operators; Moser's iteration method was developed into a general 'Nash–Moser' implicit function theorem for scales of Banach spaces by Zehnder (1975), and Hamilton's 1982 survey standardized the 'tame' formulation. Hörmander had used the scheme in his solution of the boundary problems of physical geodesy (1976), where the relevant scale consists of Hölder spaces $C^{k,\alpha}$ (functions whose derivatives up to order $k$ are $\alpha$-Hölder continuous).

Hörmander's 1985 note reformulates the theorem in an abstract setting in the spirit of Zehnder. A scale $(E_a)_{a\ge 0}$ of Banach spaces is given, decreasing in $a$ (each $E_b$ injects continuously into $E_a$ for $b>a$), together with smoothing operators $S_\theta\colon E_0\to E_\infty$, where $E_\infty=\bigcap_{a\ge 0}E_a$ is the space of elements belonging to every level of the scale (a family of linear 'low-pass filters' approximating the identity, indexed by a frequency parameter $\theta\ge 1$), satisfying, with constants $C$ bounded when the indices $a,b$ are bounded:

• (i) $\|S_\theta u\|_a\le C\|u\|_a$ (boundedness in the base norm),
• (ii) $\|S_\theta u\|_b\le C\theta^{b-a}\|u\|_a$ for $b>a$ (smoothing gain),
• (iii) $\|u-S_\theta u\|_b\le C\theta^{b-a}\|u\|_a$ for $a>b$ (approximation),
• (iv) $\|\theta\,\tfrac{d}{d\theta}S_\theta u\|_b\le C\theta^{b-a}\|u\|_a$ (a strengthened smoothing bound, the strongest hypothesis).

Hörmander's remark, early in the note, is that (iv) implies (iii) and (ii) by integration in $\theta$, 'although the constants may become large as $b$ approaches $a$', and that in (iii) the strong norm $\|u\|_a$ may be replaced by the weak-space norm $\|u\|_{E_a'}$, giving the strengthened approximation estimate $\|u-S_\theta u\|_b\le C\theta^{b-a}\|u\|_{E_a'}$ for $a>b$ used later in the proof, which comes 'with another constant, which may tend to $\infty$ as $b$ approaches $a$'. Thus (i)-(iv) are the primitive hypotheses of the note, with constants bounded as stated, whereas the estimates obtained from (iv) — (ii), (iii), and this strengthened form of (iii) — carry constants that need not be bounded as $b\to a$. These are passing parenthetical remarks, not posed problems: the note never asks for uniform constants.

From (ii)–(iii) one obtains logarithmic convexity of the norms, $\|u\|_a\le C\|u\|_b^{\lambda}\|u\|_c^{1-\lambda}$ for $a=\lambda b+(1-\lambda)c$, the interpolation inequality on which the whole machinery rests. Hörmander then constructs, from a fixed sequence of smoothing frequencies, a discrete decomposition $u=\sum_i \lambda_i R_i u$ (a telescoping decomposition in the spirit of Littlewood–Paley theory, which represents a function as a sum of pieces $R_iu$ concentrated at disjoint frequency scales) and defines, for $a_1<a<a_2$, a 'weak' space $E_a'$: the set of all sums $\sum_i \lambda_i u_i$ with $\|u_i\|_{a_1}\le M\theta_i^{a_1-a-1}$ and $\|u_i\|_{a_2}\le M\theta_i^{a_2-a-1}$, normed by the infimum of such $M$. The weak space $E_a'$ and its topology are independent of the auxiliary choices; $E_a'$ contains $E_a$ with a weaker norm; and in the particular case of Hölder spaces (non-integer exponent) $E_a'=E_a$ with equivalent norms. A sequence bounded in some $E_{a'}$, $a'>a$, that converges in $E_0$ is Cauchy in every $E_b$, $b<a$; Hörmander calls such convergence to a limit in $E_a'$ 'weak convergence'.

With a second scale $(F_b)$ carrying smoothing operators with the same properties, with the embedding $F_b\hookrightarrow F_a$ compact for $b>a$, with $\Phi$ a $C^2$ map defined on a convex neighborhood of the origin in the first scale, with values in the second scale, satisfying tame estimates on $\Phi''$ (bounds of the form $\|\Phi''(u;v,w)\|\le C(1+\|u\|)\|v\|\|w\|$ with shifted indices), and with $\Phi'(u)$ admitting a right inverse losing a fixed number of derivatives, the theorem concludes: for every $f\in F_{\beta}'$ of sufficiently small norm there is a sequence $u_j\in E_0$ with a weak limit $u\in E_{\alpha}'$ such that $\Phi(u)$ converges weakly in $F_{\beta}'$ to $\Phi(0)+f$. The proof is, schematically, a Newton iteration $u_{j+1}=u_j+\Delta_j(\Psi(\Phi(u))g_j-\text{smoothing error})$ with summable steps, followed by a Leray–Schauder degree argument (a topological fixed-point principle exploiting the compact embedding) to pass from the weakly convergent scheme to the existence statement. Smallness enters only as 'if $\|g\|_1$ is small enough'; no threshold is computed, and the note explicitly avoids fixing constants to escape circularity in the induction.

Baldi and Haus (2017) revisited exactly this circle of ideas. They identify Hörmander's weak space $E_a'$ with the Besov space $B^a_{2,\infty}$ when $E_a$ is a Sobolev space (Besov spaces are function spaces defined by the behavior of the dyadic frequency blocks of $u$: $B^a_{2,\infty}$ requires only a uniform bound on the $L^2$ sizes of the blocks, whereas the Sobolev space $H^a$ requires those sizes to be square-summable; thus $E_a'$ is then a strictly larger set with a strictly weaker norm), they observe that Hörmander's theorems are the sharpest in the literature in terms of loss of regularity of the solution relative to the data, and they prove a modified 'Nash–Moser–Hörmander' theorem for Sobolev scales in which the solution belongs to the strong space $E_\alpha$, the iteration is Cauchy in $E_\alpha$, and the smallness threshold $\delta=1/B$ is explicit in the structural constants — at the price of a modified splitting of the telescoping series (a trick from paradifferential calculus, the PDE technique of splitting sums and products into frequency blocks and estimating each block at its own regularity) and an $\ell^2$ dyadic-block summability condition on the data. Hörmander's own 1990 chapter in the volume 'Analysis, et Cetera' reformulated the theory with sharp loss of regularity but within the same weak-space framework. Poppenberg (1996) showed by counterexample on Köthe sequence spaces (a class of sequence spaces with weights) and Beurling-type ultradifferentiable function classes that Nash–Moser schemes fail outright for some scales, delimiting how far the hypotheses can be weakened. A large applied literature (controllability of dispersive PDEs, KAM (Kolmogorov–Arnold–Moser) persistence of quasi-periodic motion, water waves, isometric embedding, Monge–Ampère equations) cites the 1985 theorem as a ready-made tool without addressing any of the three quantitative gaps.

Problem Statement

Work within the abstract framework of Hörmander (1985): a decreasing scale $(E_a)_{a\ge0}$ of Banach spaces (each $E_b$ injecting continuously into $E_a$ for $b>a$) with smoothing operators $S_\theta\colon E_0\to E_\infty$ satisfying the four estimates of the Background, namely $\|S_\theta u\|_a\le C\|u\|_a$, $\|S_\theta u\|_b\le C\theta^{b-a}\|u\|_a$ for $b>a$, $\|u-S_\theta u\|_b\le C\theta^{b-a}\|u\|_a$ for $a>b$, and $\|\theta\,\tfrac{d}{d\theta}S_\theta u\|_b\le C\theta^{b-a}\|u\|_a$, with constants $C$ bounded when $a,b$ are bounded, a second scale $(F_b)$ of the same type with compact embeddings $F_b\hookrightarrow F_a$ for $b>a$, a $C^2$ map $\Phi$ satisfying tame second-derivative estimates, and a right inverse of the linearization $\Phi'(u)$ losing a fixed number of derivatives; the theorem then produces, for data $f$ of sufficiently small norm, a sequence converging weakly (in Hörmander's sense) in the weak space $E_\alpha'$ to a solution of $\Phi(u)=\Phi(0)+f$. Determine the quantitative refinements of this theorem that the original proof leaves open: (1) give an explicit admissible smallness threshold for $\|f\|_{F_\beta'}$ in terms of the structural constants of the hypotheses (the constants $C$ in (i)-(iv), the tame bounds on $\Phi''$, and the norm bounds on the right inverse of $\Phi'$) under which the conclusion holds, or else demonstrate that no such explicit threshold can exist, as the published proof only asserts existence of some threshold; (2) decide whether the approximating sequence converges strongly in $E_\alpha$ and the solution belongs to $E_\alpha$ rather than only the weak completion $E_\alpha'$ — under the original hypotheses, or else under an explicitly identified natural strengthening of them — or else construct, in the same hypothesis class chosen for the affirmative branch, a scale for which the conclusion cannot be so strengthened; and (3) determine whether the constants in the estimates obtained from (iv) — the constants in the implications (iv)$\Rightarrow$(ii) and (iv)$\Rightarrow$(iii) and the constant in the strengthened approximation estimate displayed in the Background — can be bounded uniformly as the regularity index $b$ approaches $a$, or whether such uniformity necessarily fails for some scale satisfying (i)-(iv), possibly under additional hypotheses that prevent the blow-up. The three parts (1)-(3) are posed jointly, and each part is resolved by either a constructive result or a proof that the corresponding refinement is unattainable in the hypothesis class named for it. The target is Hörmander's original theorem, not its already-sharp Sobolev variant.

References

1. Lars Hörmander, "On the Nash-Moser implicit function theorem", Annales Academiae Scientiarum Fennicae. Series A. I. Mathematica 10 (1985) 255-259, DOI 10.5186/aasfm.1985.1028, https://doi.org/10.5186/aasfm.1985.1028; open-access full text at https://afm.journal.fi/article/view/134523
2. Pietro Baldi and Emanuele Haus, "A Nash–Moser–Hörmander implicit function theorem with applications to control and Cauchy problems for PDEs", Journal of Functional Analysis 273 (2017) 3875-3900, DOI 10.1016/j.jfa.2017.09.016, https://doi.org/10.1016/j.jfa.2017.09.016; arXiv:1609.00213, https://arxiv.org/abs/1609.00213
3. Lars Hörmander, "The Nash-Moser Theorem and Paradifferential Operators", in "Analysis, et Cetera", Academic Press, Boston, 1990, pp. 429-449, ISBN 9780125742498, DOI 10.1016/B978-0-12-574249-8.50024-9, https://doi.org/10.1016/B978-0-12-574249-8.50024-9
4. Markus Poppenberg, "Negative results on the Nash-Moser theorem for Köthe sequence spaces and for spaces of ultradifferentiable functions", Manuscripta Mathematica 90 (1996) 465-478, DOI 10.1007/BF02568319, https://doi.org/10.1007/BF02568319
5. Lars Hörmander, "The boundary problems of physical geodesy", Archive for Rational Mechanics and Analysis 62 (1976) 1-52, DOI 10.1007/BF00251855, https://doi.org/10.1007/BF00251855
6. E. Zehnder, "Generalized implicit function theorems with applications to some small divisor problems, I", Communications on Pure and Applied Mathematics 28 (1975) 91-140, DOI 10.1002/cpa.3160280104, https://doi.org/10.1002/cpa.3160280104
7. John Nash, "The Imbedding Problem for Riemannian Manifolds", The Annals of Mathematics 63 (1956) 20-63, DOI 10.2307/1969989, https://doi.org/10.2307/1969989
8. Richard S. Hamilton, "The inverse function theorem of Nash and Moser", Bulletin of the American Mathematical Society 7 (1982) 65-222, DOI 10.1090/S0273-0979-1982-15004-2, https://doi.org/10.1090/S0273-0979-1982-15004-2
9. Michela Berti, Philippe Bolle and Massimiliano Procesi, "An abstract Nash–Moser theorem with parameters and applications to PDEs", Annales de l'Institut Henri Poincaré C, Analyse non linéaire 27 (2010) 377-399, DOI 10.1016/j.anihpc.2009.11.010, https://doi.org/10.1016/j.anihpc.2009.11.010
10. Sanghyun Cho and Jaeseo Park, "A simple Nash-Moser implicit function theorem in weighted Banach spaces", Bulletin of the Korean Mathematical Society 40 (2003) 289-300, DOI 10.4134/bkms.2003.40.2.289, https://doi.org/10.4134/bkms.2003.40.2.289