Home Confined elasticae and the buckling of cylindrical shells
Article
Licensed
Unlicensed Requires Authentication

Confined elasticae and the buckling of cylindrical shells

  • Stephan Wojtowytsch EMAIL logo
Published/Copyright: February 11, 2020

Abstract

For curves of prescribed length embedded into the unit disk in two dimensions, we obtain scaling results for the minimal elastic energy as the length just exceeds 2 π and in the large length limit. In the small excess length case, we prove convergence to a fourth-order obstacle-type problem with integral constraint on the real line which we then solve. From the solution, we obtain the energy expansion 2 π + Θ δ 1 3 + o ( δ 1 3 ) when a curve has length 2 π + δ and determine first order coefficient Θ 37 . We present an application of the scaling result to buckling in two-layer cylindrical shells where we can determine an explicit bifurcation point between compression and buckling in terms of universal constants and material parameters scaling with the thickness of the inner shell.


Communicated by Frank Duzaar


A Proofs of basic properties

Let us begin by proving the basic scaling and energy estimates.

Proof of Lemma 1.

First claim. Immediate from scaling properties.

Second claim. See proof of fourth claim. For embedded curves, it also follows from Hopf’s Umlaufsatz

( 2 π ) 2 = ( γ κ d 1 ) 2 1 ( γ ) 𝒲 ( γ ) .

Third claim. This is essentially a convenient way to phrase the second claim.

Fourth claim. Without loss of generality, we assume that t 1 = 0 . We can now decompose the curve γ into segments γ j = γ | [ t j , t j + 1 ] where we identify L = t k + 1 = t 0 modulo L – note that γ j is a curve segment, not a coordinate function. Then, applying Poincaré’s inequality to the coordinate functions γ i j , we find that

t j t j + 1 | ( γ i j ) | 2 d s | t j + 1 - t j | 2 π 2 t j t j + 1 | ( γ i j ) ′′ | 2 d s .

Equality is attained for eigenfunctions of the Laplacian. We note that the boundary conditions for γ j are γ j ( t j ) = γ j ( t j + 1 ) , so

t j t j + 1 ( γ i j ) d s = 0 .

This only leaves the option that γ i j ( s ) = α i j sin ( π ( s - t j ) t j + 1 - t j ) for all j = 1 , , k and i = 1 , , n . At this point, we make two observations.

  1. In the situation of the first claim, the curve t 2 = L (i.e. γ may not have any multiple points), and the curve is even C 1 -periodic. That excludes the first eigenfunction and allows only

    γ i ( s ) = α i cos ( 2 π s L ) or γ i ( s ) = α i sin ( 2 π s L )

    with the larger constant 4 π 2 which establishes the first claim.

  2. A curve γ such that all coordinate functions are multiples of the sin of the same argument cannot be parameterised by arc-length. Hence we conclude that in fact

    i = 1 n t j t j + 1 | ( γ i j ) | 2 d s < i = 1 n | t j + 1 - t j | 2 π 2 t j t j + 1 | ( γ i j ) ′′ | 2 d s

    for all admissible curves. We conclude that

    C := inf γ ( 0 ) = γ ( 1 ) = 0 𝒲 ( γ ) > π 2 1 ( γ ) .

From this we obtain

𝒲 ( γ ) = j = 1 k t j t j + 1 | ( γ ) ′′ | 2 d s
j = 1 k C | t j + 1 - t j | 2 t j t j + 1 | γ | 2 d s
= C j = 1 k 1 | t j + 1 - t j | .

The sum on the right becomes minimal for equi-distant points t j = j - 1 k L giving rise to the estimate

𝒲 ( γ ) C j = 1 k 1 L k = C k 2 L .

Referring the reader to a more classical treatment of the fact that the only closed elasticae are the circle, a figure eight curve and their periodic covers, we prove that only the once covered circle can be approximated by embedded curves.

Proof of Lemma 3.

The transversal self-crossing of the figure eight is easily excluded when writing the curves locally as graphs and using the intermediate value theorem. The multiply covered circle, which only has tangential self-contact, is slightly harder to exclude.

Any curve γ which is W 2 , 2 - or more generally C 1 -close to an m-fold covered circle can be written as a radial graph

γ ( s ) = r ( s ) ( cos s sin s )

for a 2 π m -periodic function r which is C 1 -close to the constant 1-function, applying a general statement about writing a surface as a normal graph over a C 1 -close surface. If m > 1 , we consider the shifted function r ~ ( s ) = r ( s + 2 π ) and pick an interval [ a , b ] [ 0 , 2 π m ] such that r ( a ) = max r , r ( b ) = min r . By the intermediate value theorem

r ~ ( a ) - r ( a ) = r ( a + 2 π ) - max r 0 , r ~ ( b ) - r ( b ) = r ~ ( b ) - min r 0

imply that there exists s [ a , b ] such that r ~ ( s ) - r ( s ) = 0 since r , r ~ are continuous. Then

γ ( s + 2 π ) = r ( s + 2 π ) ( cos ( s + 2 π ) sin ( s + 2 π ) ) = r ~ ( s ) ( cos s sin s ) = r ( s ) ( cos s sin s ) = γ ( s ) ,

which means that γ is not embedded. ∎

Remark 1.

While we chose an elementary argument, there are more powerful tools that would cover larger classes of curves. Assuming that a curve γ of length L is parametrised by unit speed, we can write γ ( s ) = ( cos ω ( s ) , sin ω ( s ) ) and compute that the curvature of γ is κ = ω ( s ) . It follows that

γ κ d 1 = 0 L ω ( s ) d s = ω ( L ) - ω ( 0 ) 2 π

since γ ( 0 ) = γ ( L ) . The function ω is called a Gauss representation of γ. The quantity ω ( L ) - ω ( 0 ) 2 π which measures how often the tangent turns is the Whitney index of the curve.

It is clear that for the circle and all curves C 1 -close to a circle, we have

γ κ d 1 = 2 π .

Since the space of embedded curves is connected (every embedded curve becomes a round circle under curve shortening flow), all embedded curves must have Whitney index 2 π , while an m-fold covered circle has Whitney index 2 π m and any figure eight curve has Whitney index 0.

The same result on the Whitney index of an embedded curve can be obtained by using the Gauss–Bonnet theorem on the disk bounded by the curve γ due to Jordan’s curve theorem instead of the connectedness of the space of embedded curves.

Acknowledgements

The author would like to thank Patrick Dondl and Matthias Röger for drawing his attention to the subject and the anonymous referees for their valuable feedback.

References

[1] S. Avvakumov, O. Karpenkov and A. Sossinsky, Euler elasticae in the plane and the Whitney–Graustein theorem, Russ. J. Math. Phys. 20 (2013), no. 3, 257–267. 10.1134/S1061920813030011Search in Google Scholar

[2] G. Bellettini and L. Mugnai, Characterization and representation of the lower semicontinuous envelope of the elastica functional, Ann. Inst. H. Poincaré Anal. Non Linéaire 21 (2004), no. 6, 839–880. 10.1016/S0294-1449(04)00018-6Search in Google Scholar

[3] P. W. Dondl, A. Lemenant and S. Wojtowytsch, Phase field models for thin elastic structures with topological constraint, Arch. Ration. Mech. Anal. 223 (2017), no. 2, 693–736. 10.1007/s00205-016-1043-6Search in Google Scholar

[4] P. W. Dondl, L. Mugnai and M. Röger, Confined elastic curves, SIAM J. Appl. Math. 71 (2011), no. 6, 2205–2226. 10.1137/100805339Search in Google Scholar

[5] G. Friesecke, R. D. James and S. Müller, A theorem on geometric rigidity and the derivation of nonlinear plate theory from three-dimensional elasticity, Comm. Pure Appl. Math. 55 (2002), no. 11, 1461–1506. 10.1002/cpa.10048Search in Google Scholar

[6] M. Goldman, M. Novaga and M. Röger, Quantitative estimates for bending energies and applications to non-local variational problems, Proc. Roy. Soc. Edinburgh Sect. A (2018), 10.1017/prm.2018.149. 10.1017/prm.2018.149Search in Google Scholar

[7] R. Levien, The elastica: A mathematical history, Technical Report no. UCB/EECS-2008-103, University of California, Berkeley, 2008. Search in Google Scholar

[8] S. Müller and M. Röger, Confined structures of least bending energy, J. Differential Geom. 97 (2014), no. 1, 109–139. 10.4310/jdg/1404912105Search in Google Scholar

Received: 2019-05-02
Revised: 2020-01-08
Accepted: 2020-01-14
Published Online: 2020-02-11
Published in Print: 2021-10-01

© 2021 Walter de Gruyter GmbH, Berlin/Boston

Downloaded on 23.9.2025 from https://www.degruyterbrill.com/document/doi/10.1515/acv-2019-0033/html?lang=en
Scroll to top button