betaFeedback

WorksMathematics

Smooth isometric bending of a closed sphere

Abstract

We construct a smooth embedded sphere Σ⊂R3\Sigma\subset\mathbb R^3 together with a smooth one-parameter family of isometric embeddings of Σ\Sigma whose images are not all congruent. The family is periodic; an inner disk and a complementary cap stay fixed pointwise; and the area, the enclosed volume and the total mean curvature do not change. This gives a flexible closed surface in the smooth category. The mechanism is elementary. A disk whose metric is rotationally symmetric, but whose embedding is not, can be turned inside a fixed shell provided it is exactly planar near its rim. The whole difficulty is to produce such a disk. This reduces to a scalar equation of Monge–Ampère type on an annulus, the Darboux equation, for a height function which must be nonradial and yet constant to infinite order at both boundary circles. We solve this equation by a Nash–Moser argument around an infinitesimal bending of a radial profile. The profile is forced to change the sign of its curvature infinitely often near each boundary circle, so the linearized equation changes type on infinitely many circles accumulating at both ends. We build the profile from explicitly solvable one-dimensional models, describe its accumulating ends by an exact three-term recurrence, and invert the linearized operator on the whole annulus. Two modifications of the linearization, each proportional to an unsolved equation, keep the type-change circles and the characteristic return maps fixed during the iteration, so that all linearized operators act on one domain.

Introduction

The result

A bending of a closed surface is a continuous family of isometric embeddings of one fixed Riemannian metric into R3\mathbb{R}^{3}. It is nontrivial when the shape of the surface changes, that is, when the images are not all congruent. Connelly’s flexible sphere [4] shows that nontrivial bendings exist for embedded polyhedra. Problem 1.1 of Ghomi’s list [11], which is attributed there to Euler and Maxwell, asks whether a closed C2C^{2} surface in R3\mathbb{R}^{3} can be flexible; for the history of the question see also [1, 49]. We construct a C∞C^{\infty} example.

Theorem A. There are a smooth surface Σ\Sigma diffeomorphic to S2S^2, a smooth Riemannian metric gΣg_{\Sigma} on it, and a jointly smooth map F:Σ×R→R3F:\Sigma\times\mathbb{R} \to\mathbb{R}^3 such that the maps Fτ=F(⋅,τ)F_{\tau}=F(\cdot,\tau) satisfy

Fτ is an embedding,Fτ∗gE=gΣ,Fτ+1=Fτfor all τ∈R,F_{\tau}\text{ is an embedding},\qquad F_{\tau}^{*}g_{\mathrm{E}}=g_{\Sigma},\qquad F_{\tau+1}=F_{\tau}\qquad\text{for all }\tau\in\mathbb{R},

where gEg_{\mathrm{E}} is the Euclidean metric. At least two of the images Fτ(Σ)F_{\tau}(\Sigma) are not congruent under the Euclidean isometry group. An inner disk and a complementary cap of Σ\Sigma are fixed pointwise throughout the family. The area, the enclosed volume and the total oriented mean curvature of Fτ(Σ)F_{\tau}(\Sigma) do not depend on τ\tau.

Three conventions are implicit in this statement. A closed surface is compact and without boundary. Nontrivial always refers to noncongruence of the images: relabelling the points of a fixed surface by a family of intrinsic isometries moves material points but does not change the shape, and is not a bending in our sense. Families of this kind are the slide bendings studied by [44, 48, 49]; they need not be induced by rigid motions, although the surface does not change as a set. The mean curvature is (k1+k2)/2(k_1+k_2)/2 for the outward normal.

The mechanism

The family in Theorem A is completely explicit once one particular disk is available. Let DD be a disk with a rotationally symmetric metric gg, and write ϱϑ\varrho_{\vartheta} for its intrinsic rotation by the angle ϑ\vartheta. Suppose F:D→R3F:D\to\mathbb{R}^3 is an isometric embedding. If RϑR_{\vartheta} denotes the rotation of R3\mathbb{R}^3 about the vertical axis, then

Fτ=R−2πτ∘F∘ϱ2πτ(1)F_{\tau}=R_{-2\pi\tau}\circ F\circ\varrho_{2\pi\tau} \tag*{(1)}

is again an isometric embedding of (D,g)(D,g) for every τ\tau, and its image is the rotated copy R−2πτF(D)R_{-2\pi\tau}F(D) of the original image. Two cases are uninteresting. If F(D)F(D) is a surface of revolution, the image does not move at all. If F(D)F(D) is not a surface of revolution, the image does move, but in general so does its boundary curve, and nothing can be attached to it.

There is a third possibility. Suppose that F(D)F(D) is not a surface of revolution, but that near the boundary of DD the embedding is the inclusion of a flat horizontal annulus, F(x)=(x,c)F(x)=(x,c). There the two rotations in (1) cancel, so Fτ=FF_{\tau}=F near the boundary for every τ\tau. All the maps FτF_{\tau} can then be glued to one and the same complementary cap. The result is a closed surface in which a wavy annulus turns relative to a stationary shell (Figure 1 and Figure 2). If the shell carries a marker which breaks its rotational symmetry, no rigid motion can account for the relative rotation, and the shape of the closed surface changes. In the terminology above, F∘ϱ2πτF\circ\varrho_{2\pi\tau} is a slide bending of the disk, and the rotation R−2πτR_{-2\pi\tau} in (1) is chosen so that the collar stays fixed pointwise instead of the image.

Schematic view from above showing the initial and later positions of a moving annulus, with a fixed inner disk, a fixed marker on the cap, and the same metric

Figure 1. The relative rotation, seen from above. The blue curves are schematic height contours of the moving annulus; they are not numerical data for the constructed surface. The dashed circle bounds the planar collar, which stays fixed pointwise, as does the green inner disk. The red marker on the cap also stays fixed while the nonradial part turns, so no ambient rotation accounts for the whole motion.

Schematic view of a closed surface in $\mathbb{R}^{3}$ at two values of $\tau$, showing the initial and later positions, moving ring, planar collar, fixed inner disk, and marker on the fixed cap

Figure 2. A schematic view of the closed surface in R3\mathbb{R}^{3} at two values of τ\tau. This is an illustration and not a computation of the surface of Theorem A: the nonradial part is magnified by many orders of magnitude, and the infinitely many small oscillations near the two edges of the moving ring are not drawn. The colours show the sign of the nonradial part of the height, which has angular period π\pi. Between the two pictures this pattern has turned about the axis; the inner disk, the planar collar outside the dashed circle and the cap with its marker have not moved.

The motion itself therefore requires no analysis; in particular, no infinitesimal bending has to be integrated in the variable τ\tau. Everything rests on the existence of the disk, and the difficulty lies in the requirement that it be exactly planar on a collar of its boundary while being nonradial inside.

The analytic core

Such a disk is produced as follows. Fill the middle of the disk by a flat piece, so that only an annulus (r−,r+)×T(r_-,r_+)\times\mathbb{T} remains, where T=R/2πZ\mathbb{T}=\mathbb{R}/2\pi\mathbb{Z}. On the annulus take the metric of a radial graph with slope ww,

g=(1+w(r)2) dr2+r2 dθ2,(2)g=(1+w(r)^2)\,dr^2+r^2\,d\theta^2, \tag*{(2)}

and look for an isometric embedding of the form (P,z)(P,z), where zz is a height function and PP takes values in the horizontal plane. Such an embedding exists locally exactly when the

remainder metric g−dz⊗dzg - \mathrm{d}z \otimes\mathrm{d}z is positive definite and flat, and flatness is a single scalar equation for zz, the Darboux equation. The disk of § 1.2 is then obtained from the following statement, which is the main analytic result of the paper.

Theorem B. There exist radii 0<r−<r+0 < r_- < r_+, a smooth function w>0w > 0 on (r−,r+)(r_-,r_+) and a smooth function zz on (r−,r+)×T(r_-,r_+) \times\mathbb{T} with the following properties, where gg is the metric (2).

(i) ∣dz∣g<1|\mathrm{d}z|_g < 1, and the metric g−dz⊗dzg - \mathrm{d}z \otimes\mathrm{d}z is flat.

(ii) z(r,θ+π)=z(r,θ)z(r,\theta+ \pi) = z(r,\theta), and zz is not independent of θ\theta.

(iii) ww extends smoothly by zero across r−r_- and r+r_+, and zz extends smoothly by a constant across each of them.

Condition (iii) says that ww, and zz minus its boundary value, vanish to infinite order at both boundary circles. After a slight enlargement of the annulus this gives open collars on which w=0w = 0 and zz is constant, which is the exact planarity needed for the mechanism. A small error at the boundary would be useless; this is why flatness to infinite order is the right target. The precise form of Theorem B, with quantitative information on ww and zz, is Theorem 10.3.

The passage from Theorem B to Theorem A is carried out in § 2.1 and uses only classical differential geometry. The rest of the paper proves Theorem B.

Three difficulties, and how they are met

We look for the height in the form z=h+quz = h + qu, where h′=wh' = w is the radial height whose graph has the metric gg, and qq is a small parameter. To first order in qq the Darboux equation is

Lwu=(wur)r+w′ruθθ=0.(3)\mathcal{L}_w u = (w u_r)_r + \frac{w'}{r}u_{\theta\theta} = 0. \tag*{(3)}

A nonzero solution of this equation which is constant to infinite order at both ends is an infinitesimal version of what we want, and the plan is to choose ww so that one exists and then to correct it to a solution of the nonlinear equation. Three features of the problem shape everything that follows.

Flat ends force infinitely many changes of type. For w>0w > 0 the operator (3) is elliptic where w′>0w' > 0 and hyperbolic where w′<0w' < 0. At a flat end ww tends to zero. If ww were monotone near that end, no nonzero angular mode of (3) could vanish there together with its flux (Proposition 3.2). So ww must oscillate: near each end there are infinitely many circles on which the equation changes type, and they accumulate at the boundary. We build such a profile from infinitely many copies of one explicitly solvable cell, with lengths and amplitudes decaying fast enough for smoothness (§ 3.1 and § 4.1).

Two unrelated resonance conditions. On each hyperbolic annulus, characteristics turn by a fixed travel angle TT, and solving along them involves dividing by sin⁡(jT)\sin(jT) at the angular frequency jj. Independently, the solutions on neighbouring annuli must be matched, and the matching fails at eigenvalue-like resonances. A finite model in which both sets of resonances can be computed exactly (§ 2.2.3) shows that neither condition implies the other. In the actual profile the travel angles are quadratic irrationals fixed by the choice of the cells, while matching resonances are avoided by adjusting the relative amplitudes of consecutive cells. The outcome is a two-sided inverse of Lw\mathcal{L}_w on the whole annulus, in spaces weighted towards the flat ends (§ 4.2).

The inverse has to survive perturbation. The Nash–Moser method needs an inverse of the linearization at every state near the reference one, not only at the reference. For a nearby state both the type-change circles and the characteristic return maps move, so its linearization does not even act on the same space. We remove this problem by subtracting from the linearization two terms, each proportional to an equation that is not yet solved: one fixes the type-change circles, the other fixes the return maps. Both terms vanish at a solution, and they produce exactly the quadratic error allowed by Hamilton’s form of the implicit function theorem [14] (§ 5.1–§ 6.2).

Why the problem was open, and what is new

What the existing theory provides. A closed surface which can be bent is not convex: two isometric closed convex surfaces are congruent [6, 16, 41]. Since a closed surface of nonnegative curvature is convex, the curvature of a flexible closed surface takes both signs, and the equation for its isometric embeddings is of mixed type. For such equations the existence theorems known to us are local or semi-global. The local ones construct an embedding of a neighbourhood of a point at which the curvature changes sign, under a hypothesis on the way it vanishes: cleanly across a curve [31, 15], to finite order across a curve [26], or to finite order on two curves which cross [19]. The semi-global ones construct an embedding of a narrow band along one closed curve on which the curvature vanishes: on both sides of the curve when the curvature changes sign cleanly [8] or to finite order [3], and on one side of it in [30]. The angular variable is periodic there, as it is here, but a single curve of type change is involved, and in [30, 3] the embedding has finite regularity. A theorem of either kind produces one surface, and a bending is a path of surfaces. In the other direction, an infinitesimal bending of a surface with a planar patch can be written down at once (§ 1.7), but nothing guarantees that it is the derivative of a bending.

What was missing. The mechanism of § 1.2 is elementary, and it turns one surface into a path. What was not available is a surface to which it applies: a nonradial isometric embedding of a rotationally symmetric disk which is exactly planar near the boundary. The existence of such a disk is a problem of a different kind from the local and the semi-global ones. The equation has to be solved in C∞C^\infty on a whole annulus, with conditions at both boundary circles. It changes type on infinitely many circles which accumulate at the boundary (Proposition 3.2), so no hypothesis of finite order holds there. And its linearization has the two families of resonances of § 1.4.

One circumstance works in our favour. In an embedding problem the metric is given. Here it is not: Ghomі’s question asks for some surface, so the profile ww is ours to choose, and we choose it so that the linearized equation can be solved explicitly.

The new steps.

  1. The profile. A profile w∗w_* with infinitely many changes of type at each end, for which the linearized equation has a solution ϕ∗(r)cos⁡2θ\phi_*(r)\cos2\theta flat at both ends, has no such solution at the higher even frequencies, and has, after bordering by one scalar, an inverse bounded in weighted spaces at all frequencies (Proposition 4.2). Its ingredients are classical: two solvable models, a chain recurrence, Schur complements; so is the design of a meridian on which only finitely many modes are admissible (§ 1.7). What is new is their assembly into one inverse on the whole annulus, and in particular the choice of the amplitudes of the cells which avoids the matching resonances at all frequencies simultaneously (Lemma 5.9).

  1. The second residual correction. Subtracting from the linearization a multiple of the residual of the equation is a device of Han and Khuri [19 pp. 650–653]; here it fixes the circles on which the type changes (§ 2.2.5). On an annulus of hyperbolic type which is traversed completely this is not enough, because the return map of the characteristics also depends on the state. We add an equation which says that the return maps are rotations in suitable charts, with the charts as new unknowns, and subtract from the linearization a second term, proportional to the residual of this equation (§ 5.1.5).

  1. The inverse at nearby states. After the two corrections all linearizations act on one domain, and they are inverted by a Neumann series on a single Hilbert space, although the reference inverse loses weight (§ 6.1).

The Nash–Moser theorem itself is used as a black box, in Hamilton’s form.

Organization of the proof

Figure 3 shows the logical structure, and Table 1 lists what each section contributes. The paper has three parts.

Logical structure of the proof

Figure 3. The structure of the proof. An arrow means “is used to prove”. The diagram contains: Theorem A: flexible sphere; Theorem 2.5: geometric closure (§ 2.1); Theorem B = Theorem 10.3: annular height with flat ends; Nash–Moser branch q↦x(q)q \mapsto x(q) (§ 6.2) and endpoint gain (§ 6.3); Proposition 8.1: inverse of the corrected linearization near the reference (§ 6.1); Proposition 4.2: reference profile w∗w_*, seed ϕ∗\phi_*, inverse of Lw∗\mathcal{L}_{w_*} (§§§ 3.1–§ 4.2); Corrected linearization on a state-independent domain (§ 2.2, § 5.1).

SymbolMeaningIntroduced
G∗G_{*}, D∗D_{*}, L0L_{0}reference inverse, its domain, its lossLemma 8.2
∣v∣F\lvert v\rvert_{\mathcal{F}}L1L^{1} norm of the Fourier transformLemma 7.2
gs,m\mathfrak{g}_{s,m}, ns,m\mathfrak{n}_{s,m}localized seminorms of the deformation and of the state§ 5.1.4
δnroot\delta_{n}^{\mathrm{root}}width of the root zones(147)
Rx\mathcal{R}_{x}relative operator§ 6.1.4
GxG_{x}inverse at a nearby stateProp. 8.1
R\mathbf{R}, V\mathbf{V}, Q\mathbf{Q}residual, approximate inverse, quadratic error§ 6.2
X\mathcal{X}, Y\mathcal{Y}spaces of increments and residuals§ 6.2
C∘C^{\circ}, R∘R^{\circ}, D\mathcal{D}, JJ, KxK_{x}, UxU_{x}blocks of the linear system(211)
(η,μ,κ,ψ)(\eta,\mu,\kappa,\psi)increments of (u,ν,κ,hL)(u,\nu,\kappa,h_{L})(213)
BβB_{\beta}space of control increments§ 6.2
Bw,U\mathcal{B}_{w,U}, Bv∘B_{v}^{\circ}, C\mathcal{C}, P\mathcal{P}, U\mathcal{U}remainders, mean return map, primitive, mean of the height§ 6.3

Table 1.

Part I (Sections 2 and 3) contains the geometry and the derivation of the scalar equation. It is self-contained and elementary.

Part II (Sections 4–6) is one-dimensional. After separation of angular modes, the linear equation (3) becomes a family of ordinary differential equations, and the whole part is devoted to the construction of one profile w∗w_* for which these equations have the right kernel and a controlled inverse. Its only output is Proposition 4.2; a reader willing to accept that proposition can go directly from Section 4.2 to Part III.

Part III (Sections 7–10) solves the nonlinear equation near the reference and proves the two theorems.

The estimates are arranged so that precision is spent where the decisions above are implemented: the matching of infinitely many intervals, the common domain of the perturbed operators, and the decay at the endpoints are proved in detail, while the spectra of the models and the elementary operator theory are quoted. The specific numbers that appear in the construction, such as the exponents 2n3/32n^3/3 and n5/2n^{5/2}, the power 12 and the weight exponent 3/43/4, are each introduced together with the inequalities they have to satisfy; none of them is optimal, and Table 2 collects them.

QuantitySizePurpose
Length of cell nnAn≲e−n2A_n\lesssim e^{-n^2}The lengths are summable, so the cells accumulate at a finite endpoint.
Amplitude of cell nnλn=e−n5/2+O(n)\lambda_n=e^{-n^{5/2}+O(n)}Beats the cost eCsn2e^{C_s n^2} of ss derivatives: the profile is flat.
Length of a join≲An12\lesssim A_n^{12}A large change of amplitude is crossed with negligible transfer error.
Matching angleδn=c0e−n2\delta_n=c_0e^{-n^2}The excluded sets of amplitude ratios are summable over all frequency bands.
Decay of the solutionHn=e−SnH_n=e^{-S_n}, Sn=23n3S_n=\frac{2}{3}n^3Produced by the recurrence; gives integrability and flatness of the selected solution.
Coefficient allowanceEn=en5/2log⁡(e+n)\mathcal{E}_n=e^{n^{5/2}\log(e+n)}Absorbs fixed powers of coefficients and of changes of coordinates, while staying below eεSne^{\varepsilon S_n}.

Table 2. The scales of the tail. Their compatibility is expressed by (36) and verified in this section.

Context

Continuous families versus pairs. Ghomi’s problem [11 Problem 1.1] asks for a continuous family. In his survey of the work of Cohn-Vossen, Alexandrov puts the question in the words “Do there exist closed surfaces that admit a continuous finite bending?” [1]. Pairs are easier to obtain than paths. Replacing a dimple by its reflection across a flat rim gives two different isometric realizations of one metric, and an infinitesimal bending zz of a closed surface xx gives the isometric pair x±tzx \pm tz [7, 1]; neither gives a path between the two surfaces. Sa Earp and Toubiana [51] construct smooth closed surfaces, some of genus zero, which have isometric and noncongruent smooth realizations arbitrarily close to them in every CkC^k topology, as well as families of isometric surfaces parametrized by a nowhere dense compact subset of [0,1][0,1]. Ghomi, Qiu, Raffaelli and Zhang [12] recently used the second observation to construct, for every genus g≥1g \ge1, a metric on the closed surface of genus gg with uncountably many smooth tight isometric embeddings no two of which are congruent. They state that it is not known whether these surfaces are flexible, and that no smooth counterpart of Connell’s flexible polyhedra has been found. Theorem A provides a path in the smooth category.

References

  1. [1]A. D. Alexandrov. On the works of S. E. Cohn-Vossen. Uspekhi Matem- aticheskikh Nauk, 2(3(19)):107–141, 1947. In Russian; English translation by A. Iacob, 2012, https://math.uni-koeln.de/sites/math_career/studium/ Weitere_Dokumente/Stefan_Cohn_Vossen/Alexandrov_Iacob.pdf.
  2. [2]F. J. Almgren, Jr. and I. Rivin. The mean curvature integral is invariant under bending. In The Epstein Birthday Schrift, volume 1 of Geometry & Topology Monographs, pages 1–21. 1998.
  3. [3]W. Cao. The semi-global isometric embedding of surfaces with curvature changing signs stably. Proceedings of the American Mathematical Society, 147(10):4343–4353, 2019.DOI
  4. [4]R. Connelly. A counterexample to the rigidity conjecture for polyhedra. Publications Mathématiques de l’IHÉS, 47:333–338, 1977.DOI
  5. [5]J. M. Combes and L. Thomas. Asymptotic behaviour of eigenfunctions for multiparticle Schrödinger operators. Communications in Mathematical Physics, 34:251–270, 1973.DOI
  6. [6]S. Cohn-Vossen. Zwei Sätze über die Starrheit der Eiflächen. Nachrichten von der Gesellschaft der Wissenschaften zu Göttingen, Mathematisch-Physikalische Klasse, pages 125–134, 1927.
  7. [7]S. Cohn-Vossen. Unstarre geschlossene flächen. Mathematische Annalen, 102:10–29, 1930.
  8. [8]G.-C. Dong. The semi-global isometric imbedding in R³ of two-dimensional Riemannian manifolds with Gaussian curvature changing sign cleanly. Journal of Partial Differential Equations, 6(1):62–79, 1993.DOI
  9. [9]N. V. Efimov. Qualitative problems of the theory of deformation of surfaces. Uspekhi Matematicheskikh Nauk, 3(2(24)):47–158, 1948. In Russian.
  10. [10]N. V. Efimov. Research on the deformation of surfaces containing points with zero gaussian curvature. Matematicheskii Sbornik (N.S.), 23(65)(1):89–125, 1948. In Russian.
  11. [11]M. Ghomi. Open problems in geometry of curves and surfaces, 2019.
  12. [12]M. Ghomi, H. Qiu, M. Raffaelli, and Y. Zhang. Nonrigidity of smooth tight surfaces, 2026. Preprint, arXiv:2611.17269v1, 12 November 2026.arxiv.org/abs/2611.17269
  13. [13]M. Ghomi and J. Spruck. Rigidity of nonnegatively curved surfaces relative to a curve. International Mathematics Research Notices, 2020(17):5387–5400, 2020.
  14. [14]R. S. Hamilton. The inverse function theorem of Nash and Moser. Bulletin of the American Mathematical Society (New Series), 7(1):65–222, 1982.DOI
  15. [15]Q. Han. On the isometric embedding of surfaces with Gauss curvature changing sign cleanly. Communications on Pure and Applied Mathematics, 58(2):285–295, 2005.DOI
  16. [16]G. Herglotz. Über die Starrheit der Eiflächen. Abhandlungen aus dem Mathematischen Seminar der Hansischen Universität, 15:127–129, 1943.DOI
  17. [17]Q. Han and J.-X. Hong. Isometric Embedding of Riemannian Manifolds in Euclidean Spaces, volume 130 of Mathematical Surveys and Monographs. American Mathematical Society, Providence, RI, 2006.DOI
  18. [18]Q. Han, J.-X. Hong, and C.-S. Lin. Local isometric embedding of surfaces with nonpositive Gaussian curvature. Journal of Differential Geometry, 63(3):475–520, 2003.DOI
  19. [19]Q. Han and M. A. Khuri. On the local isometric embedding in ℝ³ of surfaces with Gaussian curvature of mixed sign. Communications in Analysis and Geometry, 18(4):649–704, 2010.
  20. [20]Q. Han and M. A. Khuri. Smooth solutions to a class of mixed type Monge–Ampère equations. Calculus of Variations and Partial Differential Equations, 47:825–867, 2013.arxiv.org/abs/1204.6704
  21. [21]W. Hamouche, C. Maurini, S. Vidoli, and A. Vincenti. Multi-parameter actuation of a neutrally stable shell: a flexible gear-less motor. Proceedings of the Royal Society A, 473(2204):20170364, 2017. arXiv:1707.09013.DOI
  22. [22]L. Hörmander. The boundary problems of physical geodesy. Archive for Rational Mechanics and Analysis, 62:1–52, 1976.DOI
  23. [23]L. Hörmander. On the Nash–Moser implicit function theorem. Annales Academiae Scientiarum Fennicae, Series A I Mathematica, 10:255–259, 1985.DOI
  24. [24]L. Infeld and T. E. Hull. The factorization method. Reviews of Modern Physics, 23:21–68, 1951.DOI
  25. [25]I. Ivanova-Karatopraklieva and I. K. Sabitov. Deformation of surfaces. I. Itogi Nauki i Tekhniki. Seriya Problemy Geometrii, 23:131–184, 1991. In Russian; English translation in Journal of Mathematical Sciences 70 (1994), no. 2, 1685–1716.
  26. [26]M. A. Khuri. The local isometric embedding in ℝ³ of two-dimensional Riemannian manifolds with Gaussian curvature changing sign to finite order on a curve. Journal of Differential Geometry, 76(2):249–291, 2007.
  27. [27]I. S. Kac and M. G. Krein. On the spectral functions of the string. American Mathematical Society Translations, Series 2, 103:19–102, 1974.DOI
  28. [28]N. H. Kuiper. On C¹-isometric imbeddings. I. Proceedings of the Koninklijke Nederlandse Akademie van Wetenschappen, Series A, 58:545–556, 1955.DOI
  29. [29]R. E. Langer. The asymptotic solutions of ordinary linear differential equations of the second order, with special reference to a turning point. Transactions of the American Mathematical Society, 67:461–490, 1949.DOI
  30. [30]C.-H. Li. The semi-global isometric embedding of surfaces with Gaussian curvature changing sign cleanly. Science China Mathematics, 55(12):2507–2515, 2012.DOI
  31. [31]C.-S. Lin. The local isometric embedding in ℝ³ of two-dimensional Riemannian manifolds with Gaussian curvature changing sign cleanly. Communications on Pure and Applied Mathematics, 39(6):867–887, 1986.DOI
  32. [32]A. D. Milka. On points of relative non-rigidity of a convex surface of revolution. Ukrainskii Geometricheskii Sbornik, 1:65–74, 1965. In Russian.
  33. [33]P. M. Morse. Diatomic molecules according to the wave mechanics. II. vibrational levels. Physical Review, 34:57–64, 1929.DOI
  34. [34]J. Moser. A rapidly convergent iteration method and non-linear partial differential equations. I. Annali della Scuola Normale Superiore di Pisa, Classe di Scienze (3), 20(2):265–315, 1966.
  35. [35]T. Minagawa and T. Radó. On the infinitesimal rigidity of surfaces of revolution. Mathematische Zeitschrift, 59:151–163, 1953.DOI
  36. [36]B. Muckenhoupt. Hardy’s inequality with weights. Studia Mathematica, 44(1):31–38, 1972.DOI
  37. [37]J. Nash. C¹ isometric imbeddings. Annals of Mathematics, 60(3):383–396, 1954.
  38. [38]J. Nash. The imbedding problem for Riemannian manifolds. Annals of Mathematics, 63(1):20–63, 1956.
  39. [39]National Institute of Standards and Technology. Digital Library of Mathematical Functions: Chapter 9, Airy and related functions. https://dlmf.nist.gov/9.
  40. [40]F. W. J. Olver. Asymptotics and Special Functions. Academic Press, New York, 1974.
  41. [41]A. V. Pogorelov. Extrinsic Geometry of Convex Surfaces, volume 35 of Translations of Mathematical Monographs. American Mathematical Society, Providence, RI, 1973.DOI
  42. [42]Y. G. Reshetnyak. On non-rigid surfaces of revolution. Sibirskii Matematicheskii Zhurnal, 3(4):591–604, 1962. In Russian.
  43. [43]I. K. Sabitov. Possible generalizations of the Minagawa–Rado lemma on the rigidity of a surface of revolution with a fixed parallel. Mathematical Notes, 19(1):74–79, 1976. Russian original: Mat. Zametki 19:1 (1976), 123–132.DOI
  44. [44]I. K. Sabitov. Quasiconformal mappings of a surface generated by its isometric transformations, and bendings of the surface onto itself. Fundamentalnaya i Prikladnaya Matematika, 1(1):281–288, 1995. In Russian.
  45. [46]I. K. Sabitov. Infinitesimal and global rigidity and inflexibility of surfaces of revolution with flattening at the poles. Sbornik: Mathematics, 204(10):1516–1547, 2013.DOI
  46. [47]I. K. Sabitov. Second-order infinitesimal bendings of surfaces of revolution with flattening at the poles. Sbornik: Mathematics, 205(12):1787–1814, 2014.DOI
  47. [48]I. K. Sabitov. Bendable surfaces without visible deformations of their shape. Mathematical Notes, 110(1):126–134, 2021.DOI
  48. [49]I. K. Sabitov. Infinitesimal sliding bendings of compact surfaces and Euler’s conjecture. Siberian Mathematical Journal, 64(5):1213–1228, 2023.DOI
  49. [50]R. T. Seeley. Extension of C∞ functions defined in a half space. Proceedings of the American Mathematical Society, 15:625–626, 1964.
  50. [51]R. Sa Earp and E. Toubiana. Discrete and nondiscrete isometric deformations of surfaces in ℝ³. Siberian Mathematical Journal, 43(4):714–718, 2002.DOI
  51. [52]J. Sjöstrand and M. Zworski. Elementary linear algebra for advanced spectral problems. Annales de l’Institut Fourier, 57(7):2095–2141, 2007.arxiv.org/abs/math/0312166
  52. [53]G. Teschl. Mathematical Methods in Quantum Mechanics: With Applications to Schrödinger Operators, volume 99 of Graduate Studies in Mathematics. American Mathematical Society, Providence, RI, 2009.
  53. [54]F. Tricomi. Sulle equazioni lineari alle derivate parziali di secondo ordine, di tipo misto. Atti della Accademia Nazionale dei Lincei, Serie 5, 14:134–247, 1923.
  54. [55]D. A. Trotsenko. Nonrigid analytic surfaces of revolution. Siberian Mathematical Journal, 21(5):718–724, 1980. Russian original: Sibirsk. Mat. Zh. 21:5 (1980), 100–108.DOI

Paper details

vlnc:W_gw25zjr5
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

OpenAI ChatGPT GPT-6 Pro, Anthropic Claude Fable Claude Fable 5.2

View statement
Looking for related papers…
Contents