betaFeedback

WorksPhysics and Astronomy

Carbon-made Lobachevsky surfaces

Abstract

Carbon-made surfaces of constant negative Gaussian curvature realize portions of the Lobachevsky plane in the laboratory, with motivations ranging from analogue gravity, where the elliptic, Beltrami and hyperbolic pseudospheres correspond to the three static phases of the Bañados–Teitelboim–Zanelli black hole, to water desalination. We review and extend their construction around a single principle, the balanced Gauss–Bonnet theorem: the curvature fixed by the topology is shared among disclination defects, elastically deformed hexagons and boundaries. Curvature without defects cannot change the topology: a sphere made of hexagons only is a punctured sphere, as an explicit polyhedral construction shows. The arrangements of the defects on the three pseudospheres stem from the elliptic, parabolic and hyperbolic one-parameter subgroups of SL(2,R)SL(2,\mathbb{R}): we construct them for the elliptic family, recover the known one of the Beltrami pseudosphere, and propose one for the hyperbolic pseudosphere. We compare these ideal arrangements with carbon structures obtained by design and by simulated nano-extrusion, and suggest that the infinitely many Fuchsian groups may correspond to the infinitely many Lobachevsky surfaces.

Introduction

The main message of this work is that a surface is defined primarily by its topology, and only to a minor extent by its local geometry, provided that “topology” is understood in the precise sense of the Gauss–Bonnet theorem, boundaries included. When the surface is to be built from a discrete material, such as graphene, this statement becomes very concrete: the total curvature required by the topology must be carried by something, and that something can be disclination defects, elastically deformed hexagons, or the boundaries. Which of the three carries what, and how, decides whether the structure we obtain is the one we wanted. We shall call this the balanced Gauss–Bonnet theorem, and we shall use it as the guiding principle throughout. A carbon structure that realizes this idea is shown in Figure 1.

A carbon-made hyperbolic pseudosphere joining two flat graphene sheets

Figure 1. A carbon-made hyperbolic pseudosphere (HYP) joining two flat graphene sheets. The structure contains exactly twelve heptagons, i.e., the twelve units of curvature −π/3-\pi/3 that the balanced Gauss–Bonnet theorem requires for HYP, −4π-4\pi in total (§ 5.1 and § 8), while all the other rings are hexagons. The same geometry is a natural candidate for desalination filters (§ 9.2).

Our main theoretical motivation is the use of graphene as an analogue of gravitational phenomena, both classical and quantum [1, 2, 3]. This includes the Hawking effect, noncommutativity, generalized uncertainty principles (GUPs), teleparallel gravity and many other phenomena [4, 5, 6, 7, 8, 9, 10, 11, 12, 13]. Those results are based on the fact that graphene low-energy quasiparticles are massless Dirac pseudo-relativistic fermions [14] (the matter field ψ\psi), propagating on a carbon two-dimensional honeycomb lattice that, for large electronic wavelengths, well approximate the (spatial part of) continuous three-dimensional spacetimes [4] (the metric gμνg_{\mu\nu}). The combined system, (ψ,gμν)(\psi,g_{\mu\nu}), realizes in the lab a quantum field on a curved classical background and, when higher-energy lattice effects are considered, quantum gravity scenarios too.

In particular, we are motivated by the recent theoretical results of Ref. [15], which present a novel approach (compared to the one that uses Weyl symmetry [1], employed earlier, see, e.g., the reviews [2, 3]) to extract information on gravitational physics from suitably shaped graphene membranes. This approach emphasizes the role of the three pseudospheres, the elliptic (ELL), hyperbolic (HYP) and Beltrami (BEL) pseudospheres, in relation to three (static and spinless) phases of the dynamics of the one physical system, that is the three-dimensional black hole of Ba~nados, Teitelboim and Zanelli (BTZ) [16]. A prominent role there is played by HYP [15]. Indeed, in Ref. [17] the first carbon-made HYP was realized in a simulation lab and studied with the aid of a supercomputer.

As ELL, HYP and BEL are three important representatives of the infinite family of “Lobachevsky surfaces” [18, 19] (that is, the surfaces of constant negative Gaussian curvature, KK), a general and mathematical motivation of this study is to establish a link between defects structure, arrangement and symmetry group and specific surfaces. In turn, given the correspondence between such surfaces and solitons [19], theoretical motivations lie in that area as well.

Motivations rooted in gravitational physics, nonlinear systems or pure mathematics are not the only ones. The management of the geometry and topology of carbon structures is important in a variety of other contexts too, both theoretical and practical. One example, of vital practical importance, is carbon desalination filters. Indeed, among the innumerable applications of graphene and of other allotropic forms of carbon, such as nanotubes, nanocones and others, their use as a “molecular sieve” for filtering polluted, fresh or marine waters, to obtain clean or drinkable water, is by now an established reality, as reported in a large literature; see, e.g., Ref. [20] and Ref. [21]. To this end, the geometry of such carbon structures plays a very important role; a particularly interesting case is the hourglass-shaped carbon structures [22, 23]. The carbon-made HYP, with its hourglass shape, is a natural candidate for a serious study of the role of defects in the desalination process [24].

The paper is organized as follows. In Section 2 we recall the key facts of the topology of surfaces and of their tessellations, from the Platonic solids to the hexagonal lattice. In Section 3 we discuss intrinsic curvature without topological defects, and show that it cannot change the topology: a sphere made of hexagons only is a punctured sphere. In Section 4 we introduce the Lobachevsky geometry, its symmetry group SL(2,R)∼SO(2,1)SL(2,\mathbb{R}) \sim SO(2,1), and its discrete subgroups, and we show how the elliptic, parabolic and hyperbolic one-parameter subgroups single out the three pseudospheres, whose geometry and Gauss–Bonnet balance are discussed in Section 5. Sections 6, 7 and 8 are devoted to carbon-made ELL, BEL and HYP, respectively: the ideal arrangements of defects and their discrete symmetries, and the actual carbon structures obtained by design or by simulated growth. In Section 9 we outline the use of these structures for analogue gravity (the BTZ spectrum and wormholes) and for desalination. Our conclusions are in Section 10. Two appendices collect the deficit-angle formulae for discrete structures and an explicit counting of the curvature of a simulated carbon HYP.

Topology of surfaces

Let us set up the stage, recalling key facts of elementary topology1. In particular, we shall discuss why a surface is defined primarily by its topology and only to a minor extent by its local geometry. An illustrative example, for instance, is the punctured sphere: it is topologically equivalent to a disk, not to a sphere, even though it is locally everywhere a sphere, except for a single point of zero measure. Details of this fascinating case are given in Section 3.

This fact carries on to the discrete structures associated with that given surface, e.g., the Platonic solids corresponding to the sphere. They are themselves characterized by the type and spatial distribution of their topological defects, in turn descending from the topology of the surface of which they are the “discrete” version. For instance, to stay with the previous example, a dodecahedron with a missing pentagon is a possible discrete version of the punctured sphere, not of the sphere.

When it comes to discrete structures, we shall mostly discuss hexagonal lattices, because we have graphene as our focus, but the main results apply to the general case.

The five Platonic solids are the only regular polyhedra with the same topology of the sphere. Their Euler characteristic, χ\chi, is the same as that of the sphere

χ(sphere)=2=χ(Plato).(1)\chi(\text{sphere}) = 2 = \chi(\text{Plato}). \tag*{(1)}

For the sphere, one can see this from the Gauss–Bonnet theorem [25] for a manifold M\mathcal{M}, that in its general form is

∫MK d2μ+∫∂Mkg dℓ=2πχ(M),(2)\int_{\mathcal{M}} K\,\mathrm{d}^{2}\mu+ \int_{\partial\mathcal{M}} k_g\,\mathrm{d}\ell= 2\pi\chi(\mathcal{M}), \tag*{(2)}

with KK the Gaussian curvature and kgk_g the geodesic curvature of the curve ∂M\partial\mathcal{M}, that is the boundary of M\mathcal{M}. The formula, when applied to a sphere (that has no boundary ∂M\partial\mathcal{M}) of constant radius RR, gives simply the total curvature

Ktot(sphere)=∫sphereK d2μ=∫02πdv∫−Rπ/2+Rπ/2du (Rcos⁡u/R)1R2=4π,(3)K_{\mathrm{tot}}(\text{sphere}) = \int_{\text{sphere}} K\,\mathrm{d}^{2}\mu= \int_{0}^{2\pi}\mathrm{d}v\int_{-R\pi/2}^{+R\pi/2}\mathrm{d}u\,(R\cos u/R)\frac{1}{R^{2}} = 4\pi, \tag*{(3)}

where K=+1/R2K = +1/R^{2}. Henceforth, being Ktot=2πχK_{\mathrm{tot}} = 2\pi\chi, we have

χ(sphere)=2.(4)\chi(\text{sphere}) = 2. \tag*{(4)}

For the polyhedra one can see this from the Euler–Poincaré formula for the tiling of two-manifolds

V−E+F=2=χ(Plato).(5)V - E + F = 2 = \chi(\text{Plato}). \tag*{(5)}

It is easy to show that the five Platonic solids are the only regular polyhedra satisfying this formula.

As ELL, BEL and HYP have boundaries, henceforth the full theorem (2) applies, it is useful to recall that

χ(M)=2−2g−b,(6)\chi(\mathcal{M}) = 2 - 2g - b, \tag*{(6)}

with gg the genus and bb the number of boundaries of the orientable manifold M\mathcal{M}. Examples are in Table 1.

SURFACEggbbχ\chi
BEL011
ELL011
HYP020
Disk011
Spherical cap011
Punct. Sphere011
Sphere002
Cylinder020
Annulus020
Torus100

Table 1. Euler characteristic for various examples.

Two remarks on the boundary term of (2) are in order, as it plays a major role in what follows. First, if ∂M\partial\mathcal{M} is only piecewise smooth, the boundary term also includes the exterior angles at its corners [25],

∫MK d2μ+∫∂Mkg dℓ+∑i(π−θi)=2πχ(M),(7)\int_{\mathcal{M}} K\,\mathrm{d}^{2}\mu+ \int_{\partial\mathcal{M}} k_g\,\mathrm{d}\ell+ \sum_i(\pi- \theta_i) = 2\pi\chi(\mathcal{M}), \tag*{(7)}
TetraHexaOctaDodecaIcosa
VV4862012
EE612123030
FF4681220
qq33435

Table 2. The five Platonic solids: X-hedra. Here qq is the number of edges meeting at each vertex (degree of the vertex).

with θi\theta_i the interior angle at the iith corner. For a flat polygon with straight sides, for instance, K=0K = 0 and kg=0k_g = 0, and (7) is the elementary statement that the exterior angles add up to 2π2\pi, with χ=1\chi= 1. Second, for a discrete structure, such as a carbon lattice, both terms become sums of angles: the bulk term is the sum of the deficit angles at the interior vertices, 2π−∑iθi2\pi- \sum_i \theta_i, that we shall use extensively from § 3.1 on, while the boundary term is entirely of the kind of the last term of (7), i.e., it is the sum, over the boundary vertices, of π−∑iθi\pi- \sum_i \theta_i, where the θi\theta_i are now the angles of the polygons meeting at that vertex (see Appendix § 13). The two are the discrete versions of KK and of kgk_g, respectively. As an illustration, in a flat flake of graphene the interior vertices carry no deficit angle, a boundary vertex belonging to one hexagon only contributes π−23π=+π3\pi- \frac{2}{3}\pi= +\frac{\pi}{3}, and a boundary vertex shared by two hexagons contributes π−43π=−π3\pi- \frac{4}{3}\pi= -\frac{\pi}{3}: whatever the shape of a simply connected flake, the former always exceed the latter by exactly six, so that the boundary term is 2π2\pi, just as the pentagons of a fullerene are always twelve.

The formula (5) is invariant under the duality transformation

V→F, F→V, E→E,(8)V \to F,\ F \to V,\ E \to E, \tag*{(8)}

for which Hexahedron ↔\leftrightarrow Octahedron and Dodecahedron ↔\leftrightarrow Icosahedron, with the Tetrahedron being self-dual.

The only regular polygons used for the five Platonic solids are: Triangle (p=3p = 3), Square (p=4p = 4) and Pentagon (p=5p = 5). The reason is in the rules that the tessellation/tiling of the surface must obey, as we now explain.

The simplest case is the tessellation of the plane. Say qq the degree of the vertices, pp the number of edges of the polygon, nn the number of polygons used in the tessellation. They must all be natural numbers. With these V=(np)/qV = (np)/q, E=(np)/2E = (np)/2 and F=nF = n, hence, using those expressions in the Euler formula for a periodic tessellation of the plane, i.e., of the torus, V−E+F=0V - E + F = 0 (equivalently, for a large patch, V−E+F=1V - E + F = 1, whose contribution per polygon, 1/n1/n, vanishes as n→∞n \to\infty), we have

npq−np2+n=0,(9)\frac{np}{q} - \frac{np}{2} + n = 0, \tag*{(9)}

Notice that the formula gives no constraints on nn (nn can be dropped), as it must be for the infinite plane, p/q−p/2+1=0p/q - p/2 + 1 = 0, hence

p=2qq−2,(10)p = \frac{2q}{q - 2}, \tag*{(10)}

from which Table 3 and Figure 2 descend

ppqqdual to
(a)63(c)
(b)44self
(c)36(a)

Table 3. Combinations of number of sides of the polygon, pp, and degree of the vertices, qq, for the three only possible tessellations of the plane with regular polygons.

Figure 2. The three cases of Table 3.

Thus, the tiling of the plane with regular polygons can only happen with hexagons, squares, triangles. No other choices of pp, q∈Nq \in\mathbb{N} are possible. Notice also that the hexagon is the only regular polygon made of regular (equilateral) triangles. That is the reason behind the duality, here being q↔pq \leftrightarrow p, between hexagonal and triangular tessellation of the plane. Clearly the square, p=4=qp = 4 = q, is self-dual.

The aware reader here might wonder how it is possible that a flat surface, the plane, can be tiled with squares and triangles. The reason is in the degree of the vertices of the tessellation. If one insists in having q=3q = 3 for the squares and the triangles too, then it would not be possible to tile the place with regular squares or triangles, but only with hexagons.

For the sphere

npq−np2+n=2.(11)\frac{np}{q} - \frac{np}{2} + n = 2 . \tag*{(11)}

The formula now depends on nn, i.e., only for specific numbers of the given regular polygon we can close a sphere. Simple algebra gives

n[(2−q)p+2q]=4q,(12)n[(2-q)p + 2q] = 4q , \tag*{(12)}

that gives Table 4

qqconditionppnnsolid
3n(6−p)=12n(6-p)=1234tetrahedron
3n(6−p)=12n(6-p)=1246hexahedron (cube)
3n(6−p)=12n(6-p)=12512dodecahedron
4n(4−p)=8n(4-p)=838octahedron
5n(10−3p)=20n(10-3p)=20320icosahedron

Table 4. The only possible tilings of the sphere with regular polygons, from n[(2−q)p+2q]=4qn[(2-q)p + 2q] = 4q: nn polygons with pp sides, qq at each vertex. Duality, p↔qp \leftrightarrow q, exchanges hexahedron and octahedron, dodecahedron and icosahedron, while the tetrahedron is self-dual.

Let us now take one more step towards the tessellation of the pseudospheres. Let us consider the tessellation formula, descending from the Euler formula, in this form

∑nnp(6−p)=6χ(M),(13)\sum_{n} n_p(6-p) = 6\chi(M) , \tag*{(13)}

where npn_p is the number of regular polygons of pp sides and, for simplicity, a definite degree of the vertices is fixed to be q=3q = 3.

We notice here three facts. First, for p>6p > 6 the polygons contribute with a negative factor to χ\chi, hence, through the Gauss–Bonnet theorem, to KK. Henceforth, pairs of “curvature charges” of the same absolute value but with opposite sign, e.g., 5 − 7 (“curvature charges”: + 1 - 1) or 4 − 8 (“curvature charges”: + 2 - 2) or 3 − 9 (“curvature charges”: + 3 - 3), are always topologically allowed for any tiling, since they contribute a net zero, overall. Second, while 3 is the lower limit for pp (if degenerate cases are not considered), there is no upper limit. Henceforth, the negative “curvature charge” carried by a polygon, with p>6p > 6, can be any.

The reason for that is the noncompact nature of SO(2,1)\mathrm{SO}(2,1), the symmetry group of the Lobachevsky plane (or of the Poincaré disk or of the hyperboloid model), that is the geometry of constant negative Gaussian curvature, K<0K < 0. This group is the negative curvature counterpart of SO(3)\mathrm{SO}(3), the symmetry group of the sphere, K>0K > 0. Besides the two infinite families of cyclic and dihedral groups, the latter has only three exceptional finite subgroups, the tetrahedral, octahedral and icosahedral groups, that are the symmetry groups of the Platonic solids. The former, being noncompact, has only cyclic finite subgroups (rotations about a point), but it has infinitely many infinite discrete subgroups, the Fuchsian groups, and that makes its “crystallography” (see § 4). Preliminary results in this comparison, within the carbon structure laboratory realization, can be found in Ref. [26]. Third, in the middle is the hexagon, that never carries any “curvature charge”. Henceforth, we can add any number of them to the tessellation of any surface.

To schematically summarize our discussion, we recall that there are three geometries: the spherical, K>0K > 0, the Euclidean, K=0K = 0, the Lobachevsky, K<0K < 0. The polygons that best represent each of those geometries, in a tessellation, are: the pentagon, for K>0K > 0; the hexagon, for K=0K = 0; the heptagon (and all polygons with p>6p > 6), for K<0K < 0. The two missing regular polygons, the equilateral triangle and the square, are the only two that can live in both the K=0K = 0 and the K>0K > 0 geometries, but not in the K<0K < 0 geometry.

Figure 3 shows the Platonic solids made with carbon. Apart from C20\mathrm{C}_{20}, the remaining molecules can only approximate the actual flat-faced polyhedra. Fullerene C80(Ih)\mathrm{C}_{80}(I_h) is a chamfered dodecahedron. Being the smallest example in an infinite series of icosahedral fullerenes - those equivalent to Goldberg polyhedra - it still appears rather spherical.

Platonic solids made with carbon

Figure 3. A look at a possible set of carbon-made X-hedra through one of their vertices: (a) tetrahedron, (b) cube, (c) octahedron, (d) dodecahedron (aka fullerene C20\mathrm{C}_{20}) and (e) icosahedron (aka fullerene C80(Ih)\mathrm{C}_{80}(I_h)).

Intrinsic curvature without topological defects

A perfect hexagonal lattice has zero intrinsic curvature, K=0K = 0. To cover with such a lattice a surface of nonzero total intrinsic curvature, when the hexagons are regular and rigid, the only way is to insert disclinations: pentagons (or squares, or triangles) for positive curvature, heptagons (or octagons, etc.) for negative curvature, see, e.g., Refs. [27, 28, 29] for the general case, Refs. [30, 31] for the positive curvature case and Refs. [26, 6] for the negative curvature case in graphene.

There is, however, a second, and somehow counter-intuitive, way to have intrinsic curvature [32, 33] on a hexagonal lattice: when the hexagons are elastic, i.e., their edges can change length and their angles can change amplitude, a hexagonal patch can carry a (vertex by vertex, or even continuous) distribution of Gaussian curvature without any change in connectivity. Such curvature is non-topological in two precise senses:

  • it is not produced by a change of topology (the Volterra cut-and-glue process), hence it cannot change the topology of the manifold/discrete structure/lattice, and its total contribution to KtotK_{\mathrm{tot}}, for a given boundary, is fixed by Gauss–Bonnet;

  • it fades away when the structure is relaxed, i.e., it is not a persistent property of the lattice, but rather of the stress state the lattice is in.

In contrast, the curvature carried by disclinations survives relaxations and, together with the boundaries, is the topology. In a way, this accounts for the name “topological” defect.

That Gaussian curvature can be sustained by elastic strain alone is well established in the elasticity theory of two-dimensional crystalline membranes: in the Föppl–von Kármán compatibility equation the Gaussian curvature and the density of disclinations enter as sources of opposite sign, so that defects “screen” the curvature, and they are needed only beyond a threshold, see Refs. [27, 30, 34, 35, 36, 37, 38, 39]. What we add here is the explicit topological bookkeeping of such non-topological curvature, through the Gauss–Bonnet theorem with boundaries, its extremal case (a sphere-like structure made of hexagons only, that is a punctured sphere), and its consequences for the counting of defects on the pseudospheres. We present several independent, although interconnected, arguments: discrete deficit-angle calculations, a continuum Gauss–Bonnet analysis, an explicit polyhedral construction, a nonlinear-metric computation and energetic considerations.

Discrete picture: three types of deformations

We use the deficit angle at a vertex, δV=2π−∑iθi\delta_V = 2\pi- \sum_i \theta_i, see Appendix § 13.

First type. We can deform the polygons keeping the deficit angle fixed, i.e., we change the individual θi\theta_is but not their sum, as in Figure 4(a) for the plane. These deformations carry no change in local or global curvature, nor in topology.

Figure 4. The three types of deformations of § 3.1: (a) the individual angles change, but not their sum, hence no curvature; (b) change of connectivity: three regular hexagons replaced by a regular pentagon, hexagon and heptagon, giving a small local deficit angle, while the total curvature of the 5–7 dipole is zero; (c) no change of connectivity: deformed hexagons open a deficit angle, all carried by the hexagons, and the gap closes only when the hexagons buckle out of the plane.

Second type. We can change the connectivity, changing the local curvature, but keeping the total curvature and the topology fixed, as in Figure 4(b). The vertex common to three regular hexagons carries δV=2π−3⋅23π=0\delta_V = 2\pi- 3 \cdot\frac{2}{3}\pi= 0, whereas the vertex common to a regular pentagon, a regular hexagon and a regular heptagon carries

δV=2π−∑n=57n−2nπ=2π−416210π≃0.02π.(14)\delta_V = 2\pi- \sum_{n=5}^{7} \frac{n-2}{n}\pi= 2\pi- \frac{416}{210}\pi\simeq0.02\pi. \tag*{(14)}

Thus regular (rigid) polygons induce a local curvature that must add up to zero when all vertices of the dipole are considered, whereas if the polygons are allowed to deform we can still lie in a plane.

Third type. Without changing the connectivity we can still change the local intrinsic curvature, as in Figure 4(c): deforming the hexagons opens a deficit angle (or produces an excess angle), and to close the structure the hexagons must buckle out of the plane.

Hence, a smooth pattern of bond-length and/or bond-angle variations induces a distributed Gaussian curvature without altering the network connectivity. Notice that keeping the edge lengths fixed does not forbid intrinsic curvature: the angles alone can do the job. This is the case of the Chebyshev (fisherman’s) nets, where all edges have the same length and only the angles vary, and that do cover curved surfaces, including those of constant negative curvature, whose Chebyshev nets are governed by the sine-Gordon equation [19]. In the discrete setting, an equilateral but non-equangular hexagon gives, in general, δV≠0\delta_V \ne0.

Continuum picture: from the disk to the spherical cap

The Gauss–Bonnet theorem (2) tells us that one can deform M\mathcal{M} very drastically, changing its local curvature at every point: as long as the deformations are homeomorphisms (no cuts, no tearing apart, no gluing), the Euler characteristic on the right-hand side is untouched. Notice that, if one needs to count how many defects are necessary for a given surface to have a given total curvature, the first term is the only relevant one. For compact surfaces without boundaries that term carries all the topology; for surfaces with boundaries the topology is shared with the boundary term [25].

Consider the planar disk of radius aa, M={(r,θ):0≤r≤a, θ∈[0,2π)}\mathcal{M}=\{(r,\theta):0\leq r\leq a,\ \theta\in[0,2\pi)\}, i.e., the base of the cap of Figure 5. It is flat, so ∫MK d2μ=0\int_{\mathcal{M}}K\,\mathrm{d}^{2}\mu=0, while its boundary, a circle of the plane, has kg=1/ak_g=1/a, hence

Figure 5. The spherical cap on the sphere of radius RR and center OO, obtained by deforming its base, the planar disk of radius aa, at distance Ra=R2−a2R_a=\sqrt{R^2-a^2} from OO. The point at distance rr from the axis and azimuth θ\theta is displaced by h(r)h(r) to the point PP of the cap: in cylindrical coordinates, P=(r,θ,h(r))P=(r,\theta,h(r)). The bulk and boundary terms of the Gauss–Bonnet theorem change, their sum, 2πχ=2π2\pi\chi=2\pi, does not.

2πχ=∫∂Mdℓa=2π⇒χ(disk)=1.(15)2\pi\chi=\int_{\partial\mathcal{M}}\frac{\mathrm{d}\ell}{a}=2\pi\quad\Rightarrow\quad\chi(\mathrm{disk})=1. \tag*{(15)}

We now deform the disk continuously, through the out-of-plane displacement h(r)h(r) of its points, into a spherical cap of radius RR and colatitude φ\varphi, with a=Rsin⁡φa=R\sin\varphi and Ra=Rcos⁡φR_a=R\cos\varphi, Figure 5 (regular shapes ease the computation, but any homeomorphic irregular variant gives the same final result). With K=1/R2K=1/R^{2} and area A=2πR2(1−cos⁡φ)A=2\pi R^{2}(1-\cos\varphi)

Ktot(φ)=∫MK d2μ=2π(1−cos⁡φ).(16)K_{\mathrm{tot}}(\varphi)=\int_{\mathcal{M}}K\,\mathrm{d}^{2}\mu=2\pi(1-\cos\varphi). \tag*{(16)}

The boundary is a circle of radius Rsin⁡φR\sin\varphi, whose total curvature is 1/(Rsin⁡φ)1/(R\sin\varphi) and whose normal curvature is 1/R1/R, hence, from k2=kg2+kn2k^{2}=k_{g}^{2}+k_{n}^{2},

kg=cos⁡φRsin⁡φ,B(φ)≡∫∂Mkg dℓ=2πcos⁡φ.(17)k_g=\frac{\cos\varphi}{R\sin\varphi},\qquad B(\varphi)\equiv\int_{\partial\mathcal{M}}k_g\,\mathrm{d}\ell=2\pi\cos\varphi. \tag*{(17)}

Therefore

Ktot(φ)+B(φ)=2π(1−cos⁡φ)+2πcos⁡φ=2π=2πχ(Sφ),(18)K_{\mathrm{tot}}(\varphi)+B(\varphi)=2\pi(1-\cos\varphi)+2\pi\cos\varphi=2\pi=2\pi\chi(S_{\varphi}), \tag*{(18)}

that is χ(spherical cap)=1=χ(disk)\chi(\mathrm{spherical\ cap})=1=\chi(\mathrm{disk}), for every 0≤φ<π0\leq\varphi<\pi, see Figure 6a. The bulk curvature grows from 00 to 4π4\pi, while the boundary term decreases from 2π2\pi to −2π-2\pi, and their sum is always 2π2\pi. At the north pole, φ→0+\varphi\to0^{+}, the area vanishes and the boundary circle collapses to a point with ∫kg dℓ→+2π\int k_g\,\mathrm{d}\ell\to+2\pi; at the equator, φ=π/2\varphi=\pi/2, the boundary is a geodesic and the whole contribution is from the bulk; at the south pole, φ→π−\varphi\to\pi^{-}, the area tends to 4πR24\pi R^{2} and the boundary circle collapses with ∫kg dℓ→−2π\int k_g\,\mathrm{d}\ell\to-2\pi, so that 4π−2π=2π4\pi-2\pi=2\pi.

Figure 6. (a) Total curvature, Ktot(φ)K_{\mathrm{tot}}(\varphi) (red), and boundary term, B(φ)B(\varphi) (blue), for the spherical cap of colatitude φ\varphi: their sum is 2π2\pi for all φ\varphi. At the north pole the boundary term makes the whole Euler characteristic, at the equator (a geodesic) the contribution is entirely from the curvature, beyond it the boundary contribution becomes negative and reaches −2π-2\pi at the south pole, where Ktot=4πK_{\mathrm{tot}}=4\pi. (b) Ktot(φ)K_{\mathrm{tot}}(\varphi) with the colatitudes (dots) at which the cap has collected nπ/3n\pi/3 of curvature, i.e., the curvature of nn pentagons, n=1,…,12n=1,\ldots,12 (right axis).

The single missing point is of paramount importance: it carries the contribution −2π-2\pi that keeps the surface a topological disk, not a sphere. To close it, a further +2π+2\pi of curvature must be added in a non-continuous way: that is, by the defects.

It is instructive to read Eq. (16) in units of the curvature charge of a pentagon, π/3\pi/3: the cap of colatitude φ\varphi has collected the curvature of nn pentagons when 2π(1−cos⁡φn)=nπ/32\pi(1-\cos\varphi_n)=n\pi/3, i.e., at φn≃34∘\varphi_n \simeq34^\circ, 48∘48^\circ, 60∘60^\circ, 71∘71^\circ, 80∘80^\circ, 90∘90^\circ, 100∘100^\circ, 109∘109^\circ, 120∘120^\circ, 132∘132^\circ, 146∘146^\circ, 180∘180^\circ, for n=1,…,12n=1,\ldots,12, see Figure 6b. This is the bridge between the defect-free and the defected descriptions of the same curved surface.

A sphere made of hexagons only is a punctured sphere

Topologically, building a sphere-like structure belt after belt of hexagons, starting from one hexagon at a pole, is equivalent to starting from a disk and adding to it, at each step, a strip, obtaining at each step another topological disk, now with nonzero curvature. The belts are necessarily made of smaller and smaller hexagons, so that their deficit angles distribute the curvature, Figure 7.

Figure 7. A sphere-like structure made of hexagons only, whose sizes decrease from the south pole towards the north pole, that is missing: topologically, a punctured sphere, i.e., a disk. It is the inverse stereographic image of the honeycomb lattice (rings 0 to 14 shown, the central hexagon is at the south pole, hidden in this view). Each hexagon is planar, all vertices lie on the sphere, and all deficit angles are positive. The hexagons accumulate at the north pole, the projection point, that is missing.

The combinatorial obstruction to closing the structure is immediate. For any finite trivalent cell complex made of hexagons only (however deformed), 3V=2E3V = 2E and 6F=2E6F = 2E, so that χ=V−E+F=0\chi= V - E + F = 0 if closed, see also Eq. (13), and χ=1\chi= 1 for any simply connected patch: never 2. Therefore, a hexagons-only “sphere” can only be an infinite complex, whose hexagons accumulate at a point, that is the missing point.

What is less immediate is the explicit computation of the curvatures carried by the various belts of hexagons, and the proof that they add up exactly to 4π4\pi, the whole curvature of the sphere, while the missing point accounts for the −2π-2\pi of the boundary. Here is a construction that makes it possible. The vertices of a regular hexagon lie on a circle. The inverse stereographic projection maps circles of the plane into circles of the sphere, and a circle of the sphere lies on a plane of R3\mathbb{R}^{3}. Hence, the inverse stereographic image of the honeycomb lattice is an infinite polyhedral surface made of planar (non-regular, cyclic) hexagons, with trivalent vertices, inscribed in the sphere, whose only missing point is the projection point (the north pole), where the images of the hexagons accumulate, see Figure 7. This surface is the image of the whole plane and the latter, being homeomorphic to the punctured sphere, has χ=1\chi= 1. Adding the point at infinity, i.e., the north pole, closes the plane into the sphere, χ=2\chi= 2: that single point carries exactly the curvature that the rings of hexagons accumulate only in the limit, and it is the discrete counterpart of the +2π+2\pi that, in § 3.2, closes the cap into the sphere. Moreover, as the circumcircle of a hexagon of the honeycomb contains no other vertex of the lattice, and the stereographic projection preserves this property, every face plane leaves all the other vertices on one side: the polyhedral surface is convex and all the deficit angles δV\delta_V are strictly positive.

The deficit angles can then be computed vertex by vertex, and summed ring after ring of hexagons, starting from the central hexagon at the south pole. The result of the numerical computation is in Figure 8 and in Table 5: the cumulative curvature approaches 4π4\pi, the curvature of ring nn decays as n−3n^{-3} and the curvature still missing after nn rings decays as n−2n^{-2}. The limit does not depend on the lattice spacing, while the approach to it does. No finite number of rings reaches 4π4\pi: the total is attained only in the limit, i.e., at the missing point. That the limit is exactly 4π4\pi can be shown without summing the series: for any finite patch, the discrete Gauss–Bonnet theorem gives ∑V∈intδV=2π−τ∂\sum_{V\in\mathrm{int}}\delta_V = 2\pi- \tau_{\partial}, with τ∂\tau_{\partial} the total turning angle of the boundary polygon, and τ∂→−2π\tau_{\partial} \to-2\pi as the boundary shrinks around the north pole. This is the discrete counterpart of the limit φ→π\varphi\to\pi of Eq. (18). How deformed hexagons produce a deficit angle was made explicit in [40], through a ring of hexagons, each with one side and five sides kℓk\ell: imposing the closure of each hexagon in the complex plane relates kk to the internal angles, and gives a deficit angle at the vertices that grows from 0 to π/3\pi/3 as the hexagons depart from regularity. Extending such a closed-form description to all the rings of the stereographic construction, where the deformation changes from ring to ring, is an open problem of discrete geometry.

Cumulative curvature of the inverse stereographic image of the honeycomb

Figure 8. Cumulative curvature (sum of the deficit angles) of the inverse stereographic image of the honeycomb, ring after ring of hexagons, from the south pole. The sum tends to 4π4\pi; the missing −2π-2\pi is carried by the boundary collapsing onto the north pole.

ring nncurvature of the ringcumulative curvature/π\pi
17.7×10−27.7 \times10^{-2}0.025
100.611.527
503.0×10−23.0 \times10^{-2}3.756
1004.0×10−34.0 \times10^{-3}3.936
2005.1×10−45.1 \times10^{-4}3.984
4006.4×10−56.4 \times10^{-5}3.996

Table 5. Curvature carried by the nn-th ring of hexagons and cumulative curvature, for the inverse stereographic image of the honeycomb lattice of side ℓ\ell, on the sphere of radius RR, ℓ/R=0.05\ell/R = 0.05. Computed with the script hexsphere.py (available with this work).

We conclude that we cannot make a sphere with hexagons only, but only a punctured sphere, and extreme conditions are needed for the hexagons, as they must become of infinitesimal size. Even in this extreme case, with one single point missing, we have not managed to change the topology through the sole deformation of the hexagons, but only the geometry.

Nonlinear metric and distributed curvature

Let us now look at the same issue from the continuum, metric point of view. An out-of-plane displacement h(x,y)h(x,y) of the material point (x,y)(x,y) induces the metric

gij=δij+∂ih ∂jh(i,j=1,2).(19)g_{ij} = \delta_{ij} + \partial_i h\,\partial_j h \qquad(i,j = 1,2). \tag*{(19)}

In the linearized limit, ∥∂h∥≪1\lVert\partial h\rVert\ll1, one recovers the flat metric (K=0K = 0), while retaining the nonlinear term ∂ih ∂jh\partial_i h\,\partial_j h produces a genuine intrinsic distortion. Notice that, by Gauss’ theorema egregium, such a displacement, at fixed (x,y)(x,y), necessarily stretches the material: lengths measured along the membrane change. This is precisely why it can produce intrinsic curvature, and why such curvature costs stretching energy (see Section 3.5).

For instance, for the spherical cap of Figure 5,

h(r)=R2−r2−R2−a2,r∈[0,a],(20)h(r) = \sqrt{R^2-r^2} - \sqrt{R^2-a^2}, \qquad r \in[0,a], \tag*{(20)}

and the first and second fundamental forms of the embedding (rcos⁡θ,rsin⁡θ,h(r))(r\cos\theta,r\sin\theta,h(r)) give

K=h′(r)h′′(r)r[1+h′(r)2]2=1R2,(21)K = \frac{h'(r)h''(r)}{r[1+h'(r)^2]^2} = \frac{1}{R^2}, \tag*{(21)}

as it must be: the nonlinear term ∂ih ∂jh\partial_i h\,\partial_j h, obtained through elastic deformations only, does produce a nonzero intrinsic curvature, while its linearization would give K=0K = 0.

Energetics: when do defects take over?

How much curvature can actually be stored in a hexagonal membrane without defects? In the small-slope Föppl–von Kármán description of a two-dimensional crystal of Young modulus YY, the Airy stress function χA\chi_A obeys [27, 34]

1Y∇4χA=s(x)−K(x),(22)\frac{1}{Y}\nabla^4\chi_A=s(\mathbf{x})-K(\mathbf{x}), \tag*{(22)}

where s(x)=∑αsαδ(x−xα)s(\mathbf{x})=\sum_\alpha s_\alpha\delta(\mathbf{x}-\mathbf{x}_\alpha) is the density of disclinations of Frank angles sαs_\alpha (s=±π/3s=\pm\pi/3 for a pentagon/heptagon) and KK is the Gaussian curvature. The two sources enter with opposite signs: curvature can be “screened” by defects, but it can also be sustained by stress alone. For a disk of radius WW with uniform KK and a disclination ss at its center, with stress-free boundary, solving Eq. (22) gives the elastic energy

E(s)=YW232πs2−YKW464s+πYK2W6384.(23)E(s)=\frac{YW^2}{32\pi}s^2-\frac{YKW^4}{64}s+\frac{\pi YK^2W^6}{384}. \tag*{(23)}

The last term is the cost of the defect-free curved state, E(0)∝YK2W6E(0)\propto YK^2W^6; the first term is the energy of a disclination in a flat disk, that grows as W2W^2 [27] (not logarithmically, as for a dislocation); the middle term is the defect-curvature interaction. A disclination of charge ss lowers the energy, E(s)<E(0)E(s)<E(0), when

∫K d2μ≃πKW2>2s,(24)\int K\,d^2\mu\simeq\pi KW^2>2s, \tag*{(24)}

i.e., for a pentagon, when the integrated curvature exceeds 2π/32\pi/3, that is one sixth of the sphere. This agrees with the threshold found in Ref. [39], where it is also shown that dislocation scars lower it to about one twelfth. The threshold is essentially geometric: it is set by the integrated curvature, i.e., by the ratio W/RW/R (or by the aspect ratio of a bump [35]), and only weakly by the lattice spacing, through the core energies. Below threshold, curvature is accommodated by elastic stress without defects; above it, disclinations (or scars) become energetically favorable, restoring the defect-based description.

When the membrane can also bend, the bending energy (Helfrich) Ebend≃12κ∫(2H)2d2μ∼κW2/R2E_{\mathrm{bend}}\simeq\frac{1}{2}\kappa\int(2H)^2d^2\mu\sim\kappa W^2/R^2 [41, 42] competes with the stretching energy E(0)∼YW6/R4E(0)\sim YW^6/R^4, and the two are comparable at W∗∼(κ/Y)1/4R1/2W^*\sim(\kappa/Y)^{1/4}R^{1/2}, the geometric mean of the elastic length κ/Y\sqrt{\kappa/Y} and the radius of curvature. For graphene the bending rigidity κ\kappa is not negligible: it stems from the misalignment of the π\pi orbitals of neighboring atoms, and it sets κ/Y\sqrt{\kappa/Y} at the sub-Ångström scale.

Experimental observations and consequences for the pseudospheres

Graphene nanobubbles grown on the Pt(111) surface are highly strained, curved, defect-free regions, where the strain produces pseudomagnetic fields above 300 T [43]; more generally, wrinkles, ripples and crumples of graphene carry curvature and strain without disclinations [44], and the metric induced by strain enters the Dirac equation of the quasiparticles [45]. On the other hand, in closed shells, or above the threshold (24), defects do nucleate, as seen in spherical colloidal crystals [31] and in crystals on negatively curved surfaces [37].

For the purposes of this work, the important lessons are two. First, graphene shaped as a curved surface between defects does not need to be flat there: it can keep some curvature, through deformed hexagons, e.g., when it is fixed on a substrate. Hence scars, that could be energetically costly, are not the only way to distribute curvature, which can be evenly spread even without defects. Second, such curvature cannot change the topology of the surface, i.e., it cannot change the total Gaussian curvature for given boundaries (4π4\pi for S2S^2, −2π-2\pi for BEL, etc.). If the defects necessary for the given topology are already there, the extra local curvature stemming from the deformed hexagons is bound to contribute zero to the total (global) curvature. A general proof of this fact, for arbitrary deformations of the hexagons in the presence of disclinations, is an open problem; for the scars it is a simple counting of 5s and 7s, that need be equal in number.

Lobachevsky surfaces

We know that there are only five Platonic solids, and that their symmetry groups are the finite subgroups of SO(3)SO(3), the isometry group of the sphere. What is their counterpart when the curvature is negative? This is the question that motivated [26], and we now set up the stage to answer it.

The Lobachevsky plane and its discrete symmetries

The Lobachevsky plane can be represented as the upper half-plane {(x~,y~)∣y~>0}\{(\tilde{x},\tilde{y})\mid\tilde{y}>0\}, with the line element

dl2=r2y~2(dx~2+dy~2),(25)d l^2 = \frac{r^2}{\tilde{y}^2}(d\tilde{x}^2+d\tilde{y}^2), \tag*{(25)}

of constant Gaussian curvature K=−1/r2K=-1/r^2. Its geodesics are semicircles orthogonal to the “absolute”, the x~\tilde{x}-axis, and vertical half-lines. Its (orientation-preserving) isometries are the Möbius transformations w↦(aw+b)/(cw+d)w\mapsto(aw+b)/(cw+d), w=x~+iy~w=\tilde{x}+i\tilde{y}, with real coefficients and ad−bc=1ad-bc=1, i.e., the group PSL(2,R)PSL(2,\mathbb{R}), locally isomorphic to SO(2,1)SO(2,1), the Lorentz group in 2+12+1 dimensions [26]. This is the negative-curvature counterpart of SO(3)SO(3).

The crucial difference is that SO(3)SO(3) is compact, while SO(2,1)SO(2,1) is not. The finite subgroups of SO(3)SO(3) are few (cyclic, dihedral, tetrahedral, octahedral and icosahedral) and they are the point groups of the Platonic solids (only some of them are crystallographic: the icosahedral group is not, as quasicrystals remind us). The finite subgroups of PSL⁡(2,R)\operatorname{PSL}(2,\mathbb{R}) are instead only cyclic, while its discrete subgroups, the Fuchsian groups [46, 47] (or, including orientation-reversing elements, the non-Euclidean crystallographic groups), are infinitely many, and infinite. The cleanest way to see the contrast is through the triangle groups Δ(l,m,n)\Delta(l,m,n) [47, 48], generated by the rotations about the vertices of a geodesic triangle with angles π/l\pi/l, π/m\pi/m, π/n\pi/n: they are finite, and spherical, when 1/l+1/m+1/n>11/l+1/m+1/n>1, i.e., for (2,2,n)(2,2,n) (dihedral) and (2,3,3)(2,3,3), (2,3,4)(2,3,4), (2,3,5)(2,3,5) (the rotation groups of the Platonic solids); Euclidean, when the sum is 11, i.e., for (2,3,6)(2,3,6), (2,4,4)(2,4,4), (3,3,3)(3,3,3) (wallpaper groups); and hyperbolic, infinite and Fuchsian, for the infinitely many triples with 1/l+1/m+1/n<11/l+1/m+1/n<1. For instance, the regular tilings {p,q}\{p,q\} (pp-gons, qq at each vertex) are spherical, Euclidean or hyperbolic when 1/p+1/q1/p+1/q is larger than, equal to, or smaller than 1/21/2, respectively. For trivalent tilings, q=3q=3, as for graphene, the spherical cases are p=3,4,5p=3,4,5 (tetrahedron, cube, dodecahedron), the Euclidean one is p=6p=6 (the honeycomb), and all p≥7p\geq7 are hyperbolic. The heptagonal tiling {7,3}\{7,3\} is the closest hyperbolic relative of the honeycomb, and its symmetry group, the (2,3,7)(2,3,7) triangle group, plays for K<0K<0 the role that the icosahedral group (2,3,5)(2,3,5) plays for the dodecahedron and for fullerenes. While the spherical (2,3,p)(2,3,p) groups stop at p=5p=5, the five Platonic solids, the hyperbolic ones go on for all p≥7p\geq7: there are infinitely many “Lobachevsky Platonic solids”, the regular tilings {p,q}\{p,q\} with (p−2)(q−2)>4(p-2)(q-2)>4, each with its own infinite Fuchsian symmetry group. These groups are the basis of the recent crystallography and band theory of hyperbolic lattices [49, 50], realized in circuit quantum electrodynamics [51], and were used earlier to describe liquid-crystalline and minimal-surface structures through the crystallography of the Lobachevsky plane [52, 53].

Hilbert theorem and the infinitely many Lobachevsky surfaces

There is, however, a fundamental obstruction to realize this geometry in our Euclidean space: by Hilbert’s theorem, no complete smooth surface of constant negative curvature can be embedded in R3\mathbb{R}^{3} [18]. Every surface of constant K<0K<0 in R3\mathbb{R}^{3} is only locally isometric to the Lobachevsky plane, it realizes only a portion of it (a “stripe”, or horocyclic sector), and it carries some unavoidable singularity: cusps, conical points, self-intersections, singular boundaries. Intuitively, these singularities stem from forcing an intrinsically non-compact structure, with symmetry SO⁡(2,1)\operatorname{SO}(2,1), into a space whose isometries form the compact group SO⁡(3)\operatorname{SO}(3) [26]. The portion realized is, however, larger than it looks. On BEL and HYP there are geodesics that wind many times around the axis: lifted to the Lobachevsky plane, each turn crosses a new copy of the fundamental strip of the surface, so that a geodesic with many windings explores a correspondingly larger piece of the Lobachevsky plane, one strip per turn. This is well known for BEL, whose isometric map onto a horocyclic strip of the upper half-plane makes it explicit [54, 2], and the near-critical trajectories that wind many times around the throat of the carbon BTZ wormhole [15] are of this kind. As there are infinitely many ways to cut a stripe out of the Lobachevsky plane, there are infinitely many such surfaces, the “Lobachevsky surfaces” [18, 19], related to the solutions of the sine-Gordon equation [19]: the pseudospheres, the Dini surfaces, the Kuen surface, the breathers, etc. The conjecture of Ref. [26] is that this infinity of surfaces is in correspondence with the infinity of discrete subgroups of SO⁡(2,1)\operatorname{SO}(2,1), suitably adapted to the embedding in R3\mathbb{R}^{3}.

For tessellations, all this means that there are infinitely many polygons that can carry negative curvature (any p>6p > 6), that different Lobachevsky surfaces have different Euler characteristics and boundaries, and, through the Gauss–Bonnet theorem, need different numbers and arrangements of defects. The boundary term, which for the sphere is absent, becomes here unavoidable, and it is one of the three actors of the balanced Gauss–Bonnet theorem.

Elliptic, parabolic and hyperbolic: the three pseudospheres

The elements of PSL⁡(2,R)\operatorname{PSL}(2,\mathbb{R}) come in three classes, according to the absolute value of their trace [46, 48]: elliptic (∣tr⁡∣<2|\operatorname{tr}| < 2), with one fixed point inside the Lobachevsky plane (rotations about that point); parabolic (∣tr⁡∣=2|\operatorname{tr}| = 2), with one fixed point on the absolute (translations along horocycles, e.g., x~↦x~+b\tilde{x} \mapsto\tilde{x} + b); hyperbolic (∣tr⁡∣>2|\operatorname{tr}| > 2), with two fixed points on the absolute (translations along the geodesic joining them, e.g., y~↦λy~\tilde{y} \mapsto\lambda\tilde{y}, x~↦λx~\tilde{x} \mapsto\lambda\tilde{x}). Correspondingly, there are three kinds of one-parameter subgroups.

A surface of revolution of constant K<0K < 0 has a one-parameter group of isometries, the rotations v↦v+ϕv \mapsto v + \phi about its axis. Lifted to the Lobachevsky plane, these rotations form a one-parameter subgroup of PSL⁡(2,R)\operatorname{PSL}(2,\mathbb{R}), and the three classes give precisely the three pseudospheres of the next section:

  • ELL: the line element du2+c2sinh⁡2(u/r) dv2du^{2} + c^{2}\sinh^{2}(u/r)\,dv^{2} is that of geodesic polar coordinates about a point, with polar angle (c/r)v(c/r)v. The rotation about the axis is elliptic, and the fixed point is the tip of ELL, a conical point of total angle 2πc/r<2π2\pi c/r < 2\pi.

  • BEL: the line element du2+c2e2u/r dv2du^{2} + c^{2}e^{2u/r}\,dv^{2} is that of horocyclic coordinates. The rotation is parabolic, and the fixed point, on the absolute, is the tip of BEL, at infinite distance (a cusp).

  • HYP: the line element du2+c2cosh⁡2(u/r) dv2du^{2} + c^{2}\cosh^{2}(u/r)\,dv^{2} is that of Fermi coordinates about a geodesic, with arclength cvcv along it. The rotation is hyperbolic, and the invariant geodesic is the waist of HYP, a closed geodesic of length 2πc2\pi c.

Thus the names elliptic, parabolic and hyperbolic of the classic differential geometry of pseudospheres are, after all, the names of the conjugacy classes of SL⁡(2,R)\operatorname{SL}(2,\mathbb{R}) to which their rotations belong, and the global invariants that distinguish the three surfaces are the cone angle (ELL), the cusp (BEL) and the length of the closed geodesic (HYP).

This gives the common structure of the discrete symmetries of the defects of the carbon pseudospheres: the rotations v↦v+2π/pv \mapsto v + 2\pi/p, that form a cyclic subgroup Zp\mathbb{Z}_p of the elliptic, parabolic or hyperbolic one-parameter group, combined with discrete steps along the meridian, that place the defects at the radii where the surface has accumulated an integer number of units of curvature. We shall see this at work for BEL (§ 7, [26]), for the family of ELLs (§ 6), and as a proposal for HYP (§ 8).

Pseudospheres

In § 4.3 the three pseudospheres appeared through the symmetry they keep: they are the Lobachevsky surfaces that are also surfaces of revolution, and their rotation about the axis is an elliptic (ELL), parabolic (BEL) or hyperbolic (HYP) one-parameter subgroup of SL⁡(2,R)\operatorname{SL}(2,\mathbb{R}). Among the infinitely many surfaces of constant K<0K<0 [18, 19] (see also Ref. [54] and, e.g., the discussion in Ref. [2]), they are the most precise counterparts of the sphere that the Hilbert theorem allows: whereas for the sphere of radius RR the Gaussian curvature K=+1/R2K=+1/R^2 is constant and positive, for a pseudosphere K=−1/r2K=-1/r^2 is constant and negative. We now give their explicit form, which is what is needed to place the defects and to compute the Gauss–Bonnet balance, and which is also the starting point of the applications of § 9. From here on, rr denotes the (fixed) radius of curvature of the Lobachevsky geometry, as in Eq. (25), while RR denotes either the radius of the sphere (§ 2 and § 3) or, from now on, the radius R(u)R(u) of the parallels of a surface of revolution.

Surfaces of revolution in R3\mathbb{R}^3 are typically defined by making a complete turn of 2π2\pi around, say, the zz-axis, of a curve γ:[a,b]→R2\gamma:[a,b]\to\mathbb{R}^2 in the, say, (x,z)(x,z)-plane. Such surfaces can be parameterized by the inner coordinates (u,v)(u,v), where v∈[0,2π]v\in[0,2\pi] is the angular coordinate and u∈[a,b]u\in[a,b] is the meridian coordinate along the surface (see e.g. Ref. [54]).

The embedding of the surface of revolution into R3\mathbb{R}^3 is then given as

x(u,v)=R(u)cos⁡v,y(u,v)=R(u)sin⁡v,z(u)=±∫u1−R′2 duˉ,(26)x(u,v)=R(u)\cos v,\quad y(u,v)=R(u)\sin v,\quad z(u)=\pm\int^{u}\sqrt{1-R^{\prime2}}\,\mathrm{d}\bar{u}, \tag*{(26)}

where R′=dR/duˉR^{\prime}=\mathrm{d}R/\mathrm{d}\bar{u}, and the function R(u)R(u), with its range and domain, fully identifies the surface. The line element is dl2=du2+R2(u) dv2\mathrm{d}l^2=\mathrm{d}u^2+R^2(u)\,\mathrm{d}v^2, and the Gaussian curvature is K=−R′′(u)/R(u)K=-R^{\prime\prime}(u)/R(u) [54]. For K=−1/r2K=-1/r^2 the general solution is R(u)=c1cosh⁡(u/r)+c2sinh⁡(u/r)R(u)=c_1\cosh(u/r)+c_2\sinh(u/r), and with an appropriate choice of the constants one finds the three pseudospheres [54]

ELL:R(u)=csinh⁡(u/r),R∈[0,rcos⁡θ],BEL:R(u)=ceu/r,R∈[0,r],HYP:R(u)=ccosh⁡(u/r),R∈[c,r2+c2],(27)\begin{aligned} \mathrm{ELL:}\quad R(u)&=c\sinh(u/r),& R&\in[0,r\cos\theta],\\ \mathrm{BEL:}\quad R(u)&=c e^{u/r},& R&\in[0,r],\\ \mathrm{HYP:}\quad R(u)&=c\cosh(u/r),& R&\in[c,\sqrt{r^2+c^2}], \tag*{(27)} \end{aligned}

where, for ELL only, c=rsin⁡θc=r\sin\theta, and the domains are u∈[0,rarcsinh⁡cot⁡θ]u\in[0,r\operatorname{arcsinh}\cot\theta] (ELL), u∈[−∞,rln⁡(r/c)]u\in[-\infty,r\ln(r/c)] (BEL), ∣u∣≤rarccosh⁡1+r2/c2|u|\le r\operatorname{arccosh}\sqrt{1+r^2/c^2} (HYP). Here cc is a constant that plays a different geometrical role in each pseudosphere (in the correspondence with the BTZ black hole of § 9.1 it is related to the horizon, hence to the mass). The minimum of RR is the tip of ELL (R=0R=0 at u=0u=0), the cusp of BEL (R=0R=0 at u=−∞u=-\infty) and the waist of HYP (R=cR=c at u=0u=0). The maximum is reached at the boundaries, that are necessarily present due to the Hilbert theorem [18], and that are named “Hilbert horizons” in Ref. [6]: ELL and BEL have one each, HYP, whose R(u)R(u) is even, has two. The three surfaces are shown in Figure 10.

Figure 10. The three pseudospheres: (left) ELL, (center) BEL and (right) HYP. Locally they have the same geometry, K=−1/r2K=-1/r^2; they are distinguished by global (topological) properties. In the correspondence of § 9.1 they are the time slices of the BTZ solution for M<0M<0, M=0M=0 and M>0M>0, respectively.

Limiting cases and Gauss–Bonnet balance

The three pseudospheres are related by limits, but at finite values of the parameters they are not isometric to one another: as seen in § 4.3, they are distinguished by global invariants (cone angle, cusp, closed geodesic), even though they are locally isometric, see Figure 10.

  • For c/r=sin⁡θ→0c/r=\sin\theta\to0, ELL tends to BEL: after a shift of the meridian coordinate, csinh⁡(u/r)→ceu/rc\sinh(u/r)\to c e^{u/r}, the tip moves to infinity and becomes a cusp, the cone angle 2πsin⁡θ2\pi\sin\theta goes to zero, and the area 2πr2(1−sin⁡θ)2\pi r^2(1-\sin\theta) tends to the area 2πr22\pi r^2 of BEL.

  • For c/r→0c/r\to0, HYP is often said to tend to two BELs joined by their tails (Figure 9). This is true only away from the waist2. Measuring the arclength ss from one Hilbert horizon, the meridian of HYP is

Figure 9. HYPs with parameters (left) r=1r = 1 and c=1c = 1, (center) r=1r = 1 and c=1/10c = 1/10, (right) r=1r = 1 and c=1/100c = 1/100: for c/r→0c/r \to0 the two halves approach two BELs away from the waist, but a neck of radius ≃c\simeq c and length ≃2.6r\simeq2.6r survives around it (see § 5.1).

R(s)=r2+c2cosh⁡sr−rsinh⁡sr=re−s/r+(r2+c2−r)cosh⁡sr,(28)R(s)=\sqrt{r^2+c^2}\cosh\frac{s}{r}-r\sinh\frac{s}{r} =r e^{-s/r}+\left(\sqrt{r^2+c^2}-r\right)\cosh\frac{s}{r}, \tag*{(28)}

where the first term is BEL, with its Hilbert horizon at s=0s=0. The correction, ≃(c2/2r)cosh⁡(s/r)\simeq(c^2/2r)\cosh(s/r), is negligible near the horizon, but it becomes as large as the BEL term at s≃rln⁡(2r/c)s\simeq r\ln(2r/c), that is precisely the distance of the waist, sw=rarcsinh⁡(r/c)s_w=r\operatorname{arcsinh}(r/c). The limit is then only pointwise: at any fixed distance from a Hilbert horizon each half of HYP tends to a BEL, but the waist runs away to infinite distance and is never replaced by a cusp. For any c>0c > 0, around the waist HYP keeps a neck, a nearly cylindrical tube where R′≃0R' \simeq0: the portion with R<2cR < 2c has length 2rarccosh⁡2≃2.63r2r\operatorname{arccosh} 2 \simeq2.63r, independently of cc, while for BEL the region R<2cR < 2c is infinitely long. The waist is a closed geodesic of length 2πc2\pi c, crossed by the geodesics that go from one half to the other, and each half of HYP is an annulus, χ=0\chi= 0, i.e., a truncated BEL rather than a BEL with its cusp (Table 6); also, HYP is finite, with height ≃2r[ln⁡(2r/c)−0.31]\simeq2r[\ln(2r/c) - 0.31]. Correspondingly, the hyperbolic rotation tends to a parabolic one only as a contraction. The limit of ELL is better behaved: there the tip, carrying 2π(1−sin⁡θ)2\pi(1 - \sin\theta), does become the cusp, carrying 2π2\pi, and χ=1\chi= 1 is preserved.

surfacebulk ∫K d2μ\int K\,d^2\mutip / cuspboundaries ∮kg dℓ\oint k_g\,d\ellχ\chi
ELL, with tip−2π(1−sin⁡θ)-2\pi(1-\sin\theta)+2π(1−sin⁡θ)+2\pi(1-\sin\theta)+2π+2\pi (Hh)1
BEL, with cusp−2π-2\pi+2π+2\pi+2π+2\pi (Hh)1
BEL, truncated−2π-2\pi—+2π+2\pi (Hh), 0 (cusp)0
HYP−4π-4\pi—+2π+2\pi, +2π+2\pi (two Hh)0

Table 6. Balanced Gauss–Bonnet for the three pseudospheres: bulk curvature, contribution of the tip (ELL) or cusp (BEL), boundary terms, and Euler characteristic. ELL is parametrized by c=rsin⁡θc = r\sin\theta, HYP by c>0c > 0; Hh stands for Hilbert horizon.

  • The sign ±\pm in z(u)z(u), Eq. (26), gives a mirror copy of each surface, and the two copies join at the boundary where R′=1R' = 1: this is the Hilbert horizon for all three surfaces; ELL too is thus made of two pieces, joined at its Hilbert horizon, R=rcos⁡θR = r \cos\theta.

The Gauss–Bonnet balance of the three surfaces is summarized in Table 6. The bulk curvature is −A/r2-A/r^2, with AA the area. For a surface of revolution d2μ=R(u) du dvd^2\mu= R(u)\,du\,dv, hence, integrating Eqs. (27)–(29) over their domains,

A=2π∫R(u) du={2πr2(1−sin⁡θ)(ELL)2πr2(BEL)4πr2(HYP)(29)A = 2\pi\int R(u)\,du = \begin{cases} 2\pi r^2(1-\sin\theta) & \text{(ELL)}\\ 2\pi r^2 & \text{(BEL)}\\ 4\pi r^2 & \text{(HYP)} \end{cases} \tag*{(29)}

and the areas of BEL and HYP do not depend on cc. For a parallel, the geodesic curvature is kg=R′/Rk_g = R'/R, so that the Hilbert horizon, where R′=1R' = 1, always contributes +2π+2\pi, and the waist of HYP, where R′=0R' = 0, contributes nothing (it is a geodesic). The tip of ELL is a conical point carrying the curvature 2π(1−sin⁡θ)2\pi(1-\sin\theta), exactly opposite to the bulk curvature: with its tip, ELL is balanced like a flat disk. The cusp of BEL, a point at infinity, carries 2π2\pi. Removing the cusp turns BEL into an annulus, χ=0\chi= 0, the missing curvature reappearing as the boundary term of a small circle around it. In what follows, ELL is always considered with its tip.

For carbon structures, the bulk term must be carried by defects and/or deformed hexagons, in units of π/3\pi/3: 6(1−sin⁡θ)6(1 - \sin\theta) units for ELL, 6 for BEL, 12 for HYP, independently of cc.

Carbon-made ELL

ELL is the pseudosphere whose rotation is elliptic: its tip is a conical point. It is the natural negative-curvature counterpart of the cone, and the first step towards BEL, that is its limit. We first recall the cone and its singularity, then discuss ELL as a K<0K<0 cone, and finally construct the arrangement and the discrete symmetries of its defects, extending the construction of Ref. [26].

Cones and the K<0K<0 cone

A cone of half-aperture θ\theta and generatrix dd (Figure 11, left) is a developable surface: cut open and laid flat, it is a disk of radius dd with a missing wedge (Figure 11, right). Equating the base circumference, 2πdsin⁡θ2\pi d\sin\theta, with the length of the arc, dαd\alpha, gives α=2πsin⁡θ\alpha=2\pi\sin\theta, hence the deficit angle

Figure 11. (Left) The cone in R3\mathbb{R}^3, with half-aperture θ\theta, generatrix dd, height h=dcos⁡θh=d\cos\theta, base radius Rcone=dsin⁡θR_{\mathrm{cone}}=d\sin\theta. (Right) The same cone cut open and laid flat, a disk of radius dd with a missing wedge of angle δ=2π−α\delta=2\pi-\alpha, the deficit angle.

δ≡2π−α=2π(1−sin⁡θ).(30)\delta\equiv2\pi-\alpha=2\pi(1-\sin\theta). \tag*{(30)}

With the line element dl2=dρ2+ρ2sin⁡2θ dφ2dl^2=d\rho^2+\rho^2\sin^2\theta\,d\varphi^2, φ∈[0,2π]\varphi\in[0,2\pi], the Gaussian curvature is zero everywhere except at the tip,

K(ρ,φ)=δ2πδ(1)(ρ)ρsin⁡θ,∫coneK d2μ=δ.(31)K(\rho,\varphi)=\frac{\delta}{2\pi}\frac{\delta^{(1)}(\rho)}{\rho\sin\theta},\qquad\int_{\mathrm{cone}}K\,d^2\mu=\delta. \tag*{(31)}

This is a quasi-regular singularity: a finite “curvature charge” sits at a single point, all curvature invariants stay bounded for ρ>0\rho> 0, and radial geodesics reach the tip in finite affine parameter, with no unique continuation.

ELL’s line element is dℓ2=du2+(rsin⁡θ)2sinh⁡2(u/r) dv2d\ell^2 = du^2 + (r\sin\theta)^2\sinh^2(u/r)\,dv^2, with K=−1/r2K = -1/r^2 [26]. For u→0u \to0 it reduces to du2+sin⁡2θ u2dv2du^2 + \sin^2\theta\,u^2dv^2: the tip of ELL is a conical point with the same deficit angle 2π(1−sin⁡θ)2\pi(1-\sin\theta) of the cone tangent to it. Cone and ELL share the conical singularity and the geodesic incompleteness, but differ in the distribution of curvature: zero everywhere, with a Dirac delta at the tip, for the cone; constant and negative everywhere for ELL, whose tip carries the positive curvature 2π(1−sin⁡θ)2\pi(1-\sin\theta) (Table 6). On a lattice: for the cone, hexagons everywhere and a (6−n)(6-n)-gon at the tip; for ELL, heptagons distributed over the surface, while the tip cannot host a net negative defect (a single heptagon there would be a saddle, spoiling the cone). The precise balance is discussed in § 6.3.

First attempts with graphene

The first five noticeable θ\theta angles are those obtained by the Volterra process of cutting wedges of π/3\pi/3 alongside the lattice directions and gluing together what is left [28], see Figure 12: removing nn wedges gives sin⁡θ=1−n/6\sin\theta= 1-n/6, the same values that we shall find for the family of ELLs in § 6.3, Eq. (32). A truncated nanocone is shown in Figure 13.

The Volterra process on the hexagonal lattice, removing wedges and gluing the edges

Figure 12. The Volterra process on the hexagonal lattice: removing n=1,…,5n = 1,\ldots,5 wedges of π/3\pi/3 (shaded), i.e., the deficit angle δn=nπ/3\delta_n = n\pi/3, and gluing the edges, produces cones with sin⁡θ=1−n/6\sin\theta= 1 - n/6 and a (6−n)(6-n)-gon, at the tip, in place of the central hexagon (highlighted in the first panel).

Nanocone with no tip, or a truncated nanocone

Figure 13. Nanocone with no tip, or a truncated nanocone.

Defects and discrete symmetries of ELL

In Ref. [26] (see its Supplemental Material) BEL was reached as the limit of a family of ELLs, constructed by accommodating one unit of negative curvature, −π/3-\pi/3, at a time. We now complete that construction and give the explicit arrangement and symmetries of the defects for the whole family.

The area of ELL is A=2πr2(1−sin⁡θ)A=2\pi r^2(1-\sin\theta), Eq. (29), hence its bulk curvature is −2π(1−sin⁡θ)-2\pi(1-\sin\theta). The member ELLk\mathrm{ELL}_k of the family carries kk units, −kπ/3-k\pi/3, i.e.,

sin⁡θk=6−k6,k=1,…,5,(32)\sin\theta_k=\frac{6-k}{6}, \qquad k=1,\ldots,5, \tag*{(32)}

with k=6k=6 being the limiting case of BEL. For a surface of revolution, dR/du=sin⁡α\mathrm{d}R/\mathrm{d}u=\sin\alpha, with α\alpha the angle between the axis and the tangent to the meridian, i.e., the half-aperture of the cone tangent to the surface along that parallel. The construction of Ref. [26] places the defects at the parallels where the surface is tangent to the cones of half-aperture

sin⁡αn=n6,(33)\sin\alpha_n=\frac{n}{6}, \tag*{(33)}

which are also the parallels where the surface, measured from the tip, has accumulated n−(6−k)n-(6-k) units of curvature. For ELL, R′=sin⁡θcosh⁡(u/r)R'=\sin\theta\cosh(u/r), hence the defects sit at

cosh⁡unr=n6−k,Rn=r6n2−(6−k)2,n=6−k,…,5.(34)\cosh\frac{u_n}{r}=\frac{n}{6-k}, \qquad R_n=\frac{r}{6}\sqrt{n^2-(6-k)^2}, \qquad n=6-k,\ldots,5. \tag*{(34)}

The first value, n=6−kn=6-k, is the tip, R=0R=0, while the value n=6n=6, that is not occupied, gives the Hilbert horizon, R=rcos⁡θkR=r\cos\theta_k: the boundary is the “sixth cone”, the plane. The k−1k-1 defects off the axis are related by the rotations of Zk−1Z_{k-1} (an elliptic subgroup, see § 4.3), combined with the discrete steps along the meridian un=rarccosh⁡[n/(6−k)]u_n=r\operatorname{arccosh}[n/(6-k)]. By analogy with FYF_Y, that is the hyperbolization of the icosahedral group YY, we call this discrete structure FCk−1F_{C_{k-1}}, as it is the hyperbolization of the cyclic group Ck−1C_{k-1}:

FCk−1≡{(u,v) | v=2π[n−(6−k)]k−1, u=rarccosh⁡n6−k, n=6−k+1,…,5}.(35)F_{C_{k-1}}\equiv\left\{(u,v)\,\middle|\,v=\frac{2\pi[n-(6-k)]}{k-1},\ u=r\operatorname{arccosh}\frac{n}{6-k},\ n=6-k+1,\ldots,5\right\}. \tag*{(35)}

In the limit k→6k\to6, arccosh⁡(n/ϵ)≃ln⁡(2n/ϵ)\operatorname{arccosh}(n/\epsilon)\simeq\ln(2n/\epsilon), and Eq. (34) gives Rn=nr/6R_n=nr/6 and the discrete “boosts” un=rln⁡(n/6)u_n=r\ln(n/6) of the structure FYF_Y of BEL, Eq. (36), with its Z5Z_5 symmetry. The construction is illustrated in Figure 14.

Figure 14. Arrangement of the defects (charge −1-1, curvature −π/3-\pi/3) on the elliptic pseudospheres ELLk\mathrm{ELL}_k: each defect sits on the parallel where the surface is tangent to the cone of half-aperture αn=arcsin⁡(n/6)\alpha_n = \arcsin(n/6). In the limit k→6k \to6 one recovers BEL, with the defects at R/r=1/6,…,5/6R/r = 1/6,\ldots,5/6 (and one at infinity).

The Gauss–Bonnet balance of Table 6 tells us what the tip must carry. With the tip, the kk units of negative bulk curvature are exactly compensated by the cone angle at the tip. If one of the kk units of negative curvature is placed at the tip, as in the construction above, the tip carries the net positive charge (k−1)π/3(k-1)\pi/3, compensated by the k−1k-1 heptagons off the axis, and the structure is balanced as a flat disk.

Carbon-made BEL

BEL is the case where both the group-theoretical construction and the simulations were carried out [26]; here we recall the main results, in the light of the balanced Gauss–Bonnet theorem.

Topology and number of defects. The area of BEL is 2πr22\pi r^2, Eq. (29), hence its bulk curvature is −2π-2\pi, i.e., six units of −π/3-\pi/3: six heptagons for an ideal carbon BEL. One of them sits at the cusp, at infinity, and is removed in any practical (truncated) realization, so that five heptagons are expected on the surface (Table 6).

The discrete group. The six defects are the vertices of a “Lobachevsky Beltrami polyhedron”, constructed by deforming the icosahedral group Y=(C2,C3,C5)Y=(C_2,C_3,C_5), oriented with one vertex at the south pole, mapped to the tip of BEL. The C5C_5 symmetry along the parallels is preserved, while the C5C_5 symmetry along the meridians is “hyperbolized” into a discrete boost, giving [26]

FY={(u,v)∣v=2πn5, u=rln⁡∣n∣6, n∈Z5},(36)F_Y = \left\{(u,v)\mathrel{\big|}v=\frac{2\pi n}{5},\ u=r\ln\frac{\lvert n\rvert}{6},\ n\in\mathbb{Z}_5\right\}, \tag*{(36)}

where n=0n=0 is the cusp, at infinity, and the angles are taken modulo 2π2\pi [26]. Strictly speaking, FYF_Y is not a group but a finite set of points, the positions of the defects: its angular part is generated by the cyclic group Z5\mathbb{Z}_5 of the rotations v↦v+2π/5v\mapsto v+2\pi/5, while the steps along the meridian, un=rln⁡(n/6)u_n=r\ln(n/6), are fixed by the curvature condition and are not uniform, so that the set is not closed under composition. We keep the name FYF_Y of Ref. [26], and use analogous names below, for these discrete structures, the “hyperbolized” counterparts of the corresponding finite groups; the genuine discrete groups behind them are discussed in § 10. The sign of nn gives the chirality of the helix on which the defects sit. The defects are at the radii Rn/r=n/6R_n/r=n/6, n=1,…,5n=1,\ldots,5, where BEL is tangent to the cones of half-aperture arcsin⁡(n/6)\arcsin(n/6), and where the surface has accumulated nn units of curvature from the cusp, see § 6.3.

Simulations. In Ref. [26] carbon BELs with 1146, 2146 and 5506 atoms were obtained by a triangulation of the surface with a Lennard-Jones potential (minimization with the fast inertial relaxation engine, FIRE), followed by Voronoi dualization, relaxation with a Stillinger–Weber potential and density-functional tight-binding calculations. The extrinsic curvature produces a diverging stretching at the Hilbert horizon and a contraction at the tip; the first is cured by pinning the atoms at the horizon, and by adding a flat ring of graphene beyond it. Heptagons and pentagons form charged and neutral scars, distributed along a helix; the number of charged scars converges to the expected five as the size grows, in agreement with FYF_Y. A large hexagonal circulation tangent to the Hilbert horizon appears only when no charged scar crosses the horizon, and it is the signature that a BEL has been obtained. The 1146-atom BEL is stable at the density functional theory (DFT) and density-functional tight-binding (DFTB) levels also when the constraint to the surface is released. In the language of this paper: the bulk charge is carried by five charged scars, the boundary term is controlled by the pinned Hilbert horizon and the flat ring, and only then the structure is a BEL.

Action of $F_Y$ on the vertices of the Lobachevsky Beltrami polyhedron and optimized carbon BELs

Figure 15. (Top) Action of FYF_Y on the vertices of the Lobachevsky Beltrami polyhedron, side and top views; the vertex at infinity (n=0n = 0) is identified with n=5n = 5. (Bottom) Optimized carbon BELs with 1146, 2146 and 5506 atoms (top views), with the charged scars and the hexagonal circulation tangent to the Hilbert horizon. Reproduced from Ref. [26].

Carbon-made HYP

HYP is the pseudosphere whose rotation is hyperbolic, and the only one with two Hilbert horizons and a closed geodesic, the waist. Its bulk curvature is −4π-4\pi, i.e., twelve units of −π/3-\pi/3, for any value of cc (Table 6), and each half, between the waist and one Hilbert horizon, carries six units, like a BEL. We discuss two complementary routes to carbon HYPs: by design, as polycyclic aromatic hydrocarbons with an ideal arrangement of heptagons, and by simulated growth, through nano-extrusion and annealing [17]. The first route places the twelve units of curvature in twelve isolated heptagons, symmetrically arranged; the second lets the carbon find its own arrangement. That the two lead to the same surface, and under which conditions, is a direct test of the balanced Gauss–Bonnet theorem.

Ideal HYPs by design

A HYP, being finite, can be realized as a single polycyclic aromatic hydrocarbon (PAH) containing heptagonal rings. Compared to fullerenes, which only contain carbon atoms and map to closed surfaces, PAHs possess a boundary where the hydrogen atoms are located. Moreover, since each heptagon carries a fixed amount of Gaussian curvature, only a homogeneous distribution of these—as in a heptagonal lattice—can realize an ideal surface with constant Gaussian curvature. As with Goldberg polyhedra, the arrangement of heptagons can be represented by indices (i,j)(i,j) along the two lattice directions. The smallest family, HYP(1,0)\mathrm{HYP}(1,0), has members made of heptagons only. It contains three molecules: C48H24\mathrm{C}_{48}\mathrm{H}_{24} (D6hD_{6h}), C48H24\mathrm{C}_{48}\mathrm{H}_{24} (D3hD_{3h}) and C44H20\mathrm{C}_{44}\mathrm{H}_{20} (D2dD_{2d}). In the family HYP(1,1)\mathrm{HYP}(1,1), the negatively curved counterpart of C60\mathrm{C}_{60}, each heptagon is surrounded by seven hexagons. The atoms can in principle still lie on the mathematical surface: C192H48\mathrm{C}_{192}\mathrm{H}_{48}, C228H48\mathrm{C}_{228}\mathrm{H}_{48} and C168H40\mathrm{C}_{168}\mathrm{H}_{40}, with the same three symmetries (Figure 16). Larger structures, such as the chiral trigonal HYP(2,1)\mathrm{HYP}(2,1) with 384 carbon atoms, are obtained by spreading the heptagons further apart. All these molecules contain exactly twelve heptagonal rings, as the total curvature, −4π-4\pi, requires. As for large fullerenes containing isolated defects, the larger the structure, the flatter the areas away from the defects will appear. In the limit, these PAHs would approach triangulated polyhedra (e.g., the lateral surface of a hexagonal antiprism for a HYP with a hexagonal symmetry). The curvature would be concentrated at the heptagons in the vertices of the polyhedra while the hexagons would both cover their faces and easily bend over their edges.

Carbon-made HYP(1,1)s with hexagonal, trigonal, and improper 4-fold symmetry

Figure 16. Carbon-made HYP(1,1)s, with heptagonal rings surrounded by hexagons: (a) hexagonal, (b) trigonal, and (c) improper 4-fold symmetrical.

Which arrangement of defects is preferred? Total energies of small PAHs containing only 5-, 6- and 7-membered rings, computed with the self-consistent-charge density-functional tight-binding method, show that heptagons are energetically more expensive than pentagons, and that the two signs of curvature behave differently: for K>0K > 0 a 5-7-5 scar is more stable than an isolated pentagon, while for K<0K < 0 an isolated heptagon is more stable than a 7-5-7 scar. This suggests a principle of curvature matching: to approach the ground state, the curvature produced by a local tiling must match the curvature that the surface would form there; scars, as discussed in Ref. [30], split and dilute the curvature of isolated defects. For HYP, it points to isolated heptagons, homogeneously distributed, as in the ideal molecules above.

On the experimental side, small molecules with one ([7]circulene) and two ([7.7]circulene) heptagonal rings have been synthesized [55, 56]. Heptagons can be formed by Scholl reactions of macrocyclic precursors [57], but each additional ring requires its own precursors and catalysts, and the strain grows with the number of heptagons. Larger molecules require multi-step processes, in which monomers are assembled through cross-coupling reactions (Suzuki–Miyaura, Yamamoto, Scholl); recently, a molecule with four heptagonal rings, a fragment of a Schwarzite, has been synthesized [58]. Monomers for the trigonal HYP(2,0) can be designed along these lines.

A proposal for the discrete symmetries of HYP

Applying to each half of HYP the rule found for ELL and BEL, i.e., placing the defects on the parallels where the surface is tangent to the cones with sin⁡αn=n/6\sin\alpha_n = n/6, and using R′=(c/r)sinh⁡(u/r)R' = (c/r)\sinh(u/r), one finds the radii

Rn=c2+(nr6)2,n=0,…,5,(37)R_n = \sqrt{c^2 + \left(\frac{nr}{6}\right)^2}, \qquad n = 0,\ldots,5, \tag*{(37)}

on each half, as for BEL in Ref. [26]: n=0n = 0 is the waist (the counterpart of the cusp of BEL), while the value n=6n = 6, that is not occupied, gives the Hilbert horizon, R6=c2+r2R_6 = \sqrt{c^2+r^2}. In the limit c→0c \to0, for n≥1n \ge1, these tend to the radii nr/6nr/6 of BEL, while the defect n=0n = 0 stays on the waist (see the discussion of § 5.1). The rotations are now hyperbolic, and the reflection u↦−uu \mapsto-u exchanges the two halves. On each half the six defects are related by the rotations of C6C_6, combined with discrete steps along the meridian, un=rarcsinh⁡(nr/6c)u_n = r\operatorname{arcsinh}(nr/6c), and the two halves are exchanged by a twofold rotation: by analogy with FYF_Y and FCk−1F_{C_{k-1}}, we call this discrete structure FD6F_{D_6}, the hyperbolization of the dihedral group D6D_6. In the limit c→0c \to0 the defect on the waist (n=0n = 0) of each half runs away to infinite distance, and, away from the waist, each half approaches the BEL of FYF_Y. This arrangement, one defect per level along a helix, is not the only one compatible with the integrated curvature: the ideal molecules of § 8.1, with symmetries D6hD_{6h}, D3hD_{3h} and D2dD_{2d}, place six, three or two heptagons on the same level, just as the icosahedron, along a fivefold axis, has the structure 1+5+5+11+5+5+1. Which of these arrangements corresponds to the minimal-energy carbon HYP, for given rr and cc, is an open question.

HYP by nano-extrusion and annealing

A completely different route was followed in Ref. [17], using molecular dynamics (with the LAMMPS code and the second-generation reactive empirical bond-order potential, REBO2). Carbon atoms are first forced, by a simulated extrusion press, down a thin volume shaped as a HYP (r=cr=c, axis length 28.5 Å, minimum and maximum diameters 40 and 56.5 Å), bounded by a strong Lennard-Jones wall and a repulsive harmonic wall. The extruded carbon has the areal density of graphene, but it is an unrealistic, highly defective and unstable precursor. It is then annealed, still constrained to the HYP shape, at temperatures up to 3500 K and for times up to 0.1 μ\mus: most defects annihilate in pairs, or migrate to the free edges, and the structure turns into polycrystalline, curved graphene, with grain boundaries made mostly of alternating heptagons and pentagons (Figure 17). During the annealing, the total Gaussian curvature fluctuates and develops towards the expected Ktot=−4πK_{\mathrm{tot}}=-4\pi.

Carbon HYP obtained by nano-extrusion and annealing, with free edges, before and after annealing, and a heptagon-pentagon grain boundary

Figure 17. Carbon HYP obtained by nano-extrusion and annealing, with free edges: (a) before and (b) after annealing at 3500 K, atoms colored by potential energy; (c) a heptagon-pentagon grain boundary. Reproduced from Ref. [17].

The resulting arrangement of defects is very far from the ideal one of § 8.1: in the structure with free edges there are more than fifty heptagon-pentagon pairs, a few isolated heptagons in excess, and a complex quad-vacancy cluster of three octagons and two pentagons, with a defect density about one order of magnitude higher than the minimal one. Nonetheless, the detailed counting of all the non-hexagonal rings, of the triangular rings, and of the spsp and sp3sp^3 carbons, reported in Appendix § 14, gives exactly twelve units of negative curvature, i.e., −4π-4\pi. The counting was carried out on one of the HYPs embedded between the two periodic sheets, and some of the scars extend across the Hilbert horizons into the sheets. For this closed structure (two tori joined by a neck, χ=−2\chi=-2) the total, −4π-4\pi, is fixed by topology; the non-trivial information is where the charge sits: on the HYP and across its junctions with the sheets, which are otherwise flat. The bulk term of the Gauss–Bonnet balance is thus fully accounted for by the defects, but distributed in a completely different way. Such a defect-rich curved graphene is interesting also beyond geometry: the magnetic moments carried by its defects, and by those of its hydrogenated counterpart, graphane, have been studied in Ref. [59].

What decides the shape is then the boundary term. When the structure is released, with free edges, it deviates considerably from the mathematical HYP. When the free edges are removed, by embedding the HYP between two periodic flat graphene sheets, so that the Hilbert horizons are no longer material edges (Figure 18), the carbon atoms sit on average 0.6 Å from the mathematical surface, and only 0.2 Å when the sheets are kept flat, as if they were stiff stacks of many sheets. The structure also proved stable against 20% elongation, 20° shearing and annealing at 1500 K. This is the balanced Gauss–Bonnet theorem at work: the same total curvature charge, −4π-4\pi, gives the same surface, whether carried by twelve isolated heptagons or by tens of scars, provided that the boundary term is the right one.

Annealed and relaxed carbon HYP embedded in flat periodic graphene sheets

Figure 18. Annealed and relaxed carbon HYP embedded in flat periodic graphene sheets: (a) without and (b) with flat walls on the sheets, atoms colored by potential energy; (c) deviation of the atoms from the mathematical HYP. Reproduced from Ref. [17].

Applications: analogue gravity and desalination

The applications of these structures can be many. Carbon is used today in a wide range of contexts, from electronics [14] to energy storage [60] and filtration [20], and curved carbon structures are certainly at work in many of them in ways that are not yet recognized. Much of what we discussed is also of purely mathematical interest: the surfaces of constant negative curvature [54, 18] and their link with the solitons of the sine-Gordon equation [19, 61], the discrete geometry of deformed lattices [40], the Lobachevsky geometry and its discrete groups [46, 47, 50]. Among all these possible directions, we focus here on the two that we have been pursuing: the realization of the BTZ black hole and wormhole in the laboratory, and water desalination.

The BTZ black-hole spectrum and wormholes

We briefly recall here how the three pseudospheres emerge from the spectrum of the BTZ black hole when the time is frozen (time-slice approach). The differences between the pseudospheres proper and the structures obtained from the limits of HYP, together with the classical traversability of the corresponding carbon BTZ wormhole, are discussed in Ref. [15].

The BTZ metric [16] for a non-spinning black hole is

dsBTZ2=(R2l2−M)dT2−(R2l2−M)−1dR2−R2dV2,(38)ds^2_{BTZ} = \left(\frac{\mathcal{R}^2}{l^2} - M\right)d\mathcal{T}^2 - \left(\frac{\mathcal{R}^2}{l^2} - M\right)^{-1}d\mathcal{R}^2 - \mathcal{R}^2d\mathcal{V}^2, \tag*{(38)}

with Λ=−l−2\Lambda=-l^{-2} and horizon at Rhor=lM\mathcal{R}_{\mathrm{hor}}=l\sqrt{M}. At fixed T\mathcal{T}, the spatial part is SO(2)SO(2)-symmetric [62], and identifying it with the line element of a surface of revolution, du2+R2(u) dv2du^2+R^2(u)\,dv^2, gives v=Vv=\mathcal{V} and R(u)=RR(u)=\mathcal{R}: the radial coordinate of the carbon structure coincides with that of the BTZ space (in the Weyl approach, instead, the two differ, and the BTZ horizon is mapped to the Hilbert horizon of BEL [6]). The meridian part then gives

dRdu=1lR2−l2M,(39)\frac{dR}{du}=\frac{1}{l}\sqrt{R^2-l^2M}, \tag*{(39)}

whose solutions, for the three signs of MM in the BTZ spectrum [63], are

M<0:R(u)=l∣M∣sinh⁡(u/l),(40)M<0:\quad R(u)=l\sqrt{|M|}\sinh(u/l), \tag*{(40)}
M=0:R(u)=αeu/l,(41)M=0:\quad R(u)=\alpha e^{u/l}, \tag*{(41)}
M>0:R(u)=lMcosh⁡(u/l).(42)M>0:\quad R(u)=l\sqrt{M}\cosh(u/l). \tag*{(42)}

Comparing with Eqs. (27)–(29), each pseudosphere realizes one part of the BTZ spectrum: ELL for M<0M<0 (no horizon), BEL for M=0M=0 (horizon at the cusp, at spatial infinity), HYP for M>0M>0, the only one with the horizon within the surface, at the waist, R=c=RhorR=c=R_{\mathrm{hor}} (Figure 10). This result is implicit in earlier investigations of the BTZ spectrum [64, 65, 66]; here we make it explicit. The parameters are identified as

r=l,c=rM(not for BEL),i.e.,Λ=−1r2,M=c2r2.(43)r=l,\qquad c=r\sqrt{M}\quad\text{(not for BEL)},\qquad\text{i.e.,}\qquad\Lambda=-\frac{1}{r^2},\qquad M=\frac{c^2}{r^2}. \tag*{(43)}

The whole spectrum [63] is sketched in Figure 19.

Figure 19. The spectrum of the BTZ solutions in the (J,M)(J,M) plane, with l=1l=1 (see Ref. [63]): black holes (M≥∣J∣M \ge|J|), conical defects (−1<M<0-1 < M < 0), the anti-de Sitter vacuum (M=−1M=-1, J=0J=0) and conical excesses (M<−1M < -1). The regions ∣J∣>∣M∣|J| > |M|, left blank, also correspond to naked singularities.

Desalination

Purification and desalination of seawater is a global issue of vital strategic importance, given the growing scarcity of drinking water caused by climate change.

Among the innumerable applications of graphene, and of other allotropic forms of carbon, such as nanotubes and nanocones, its use as a “molecular sieve” [20], for filtering polluted, fresh or marine waters, to obtain clean or drinkable water, is by now an established reality, as demonstrated by a large literature [67, 68, 69, 70, 71, 72, 73, 74, 20, 75], see also the review [21]. To this end, a very important role is played by the geometry of these carbon structures. A case of great interest is that of the theoretical study, computer simulation and laboratory realization of “hourglass”-shaped carbon structures, studied, e.g., in Ref. [22].

The rationale, behind those studies, is to create filtering structures, bio-inspired by aquaporins in biological organisms [76]. The latter are the so-called “channel proteins” which, also by virtue of their hourglass shape, facilitate the rapid passage of water, inside or outside the cells of specific tissues, while effectively filtering other molecules which are not water, as sodium ions or potassium.

The hourglass-shaped carbon nanopores studied in Ref. [22] consist of two carbon nanocones (CNCs) connected, by their truncated tips, to the two ends of a carbon nanotube (CNT), and bounded by two parallel graphene sheets, see Figure 20(a). With molecular dynamics simulations of a 0.5 M NaCl solution, nanotubes of chirality (6, 6), (8, 8) and (10, 10) and nanocones of half-aperture α=41.8∘,30.0∘,19.45∘,9.6∘\alpha= 41.8^\circ, 30.0^\circ, 19.45^\circ, 9.6^\circ (and 0∘0^\circ, i.e., the bare nanotube), the authors of Ref. [22] found that the conical entrances enhance the water flux by a factor up to about 4 with respect to the bare nanotube, the optimal half-aperture being 9.6∘9.6^\circ for all the nanotubes considered, in line with the hydrodynamic predictions of Ref. [76]. The narrowest pores, built on the (6, 6) nanotube, showed complete salt rejection for all half-apertures, while for the wider ones the salt rejection decreases slightly, remaining above 96%.

Figure 20. (a) The hourglass-shaped carbon nanopore of Ref. [22]: two carbon nanocones (CNC) joined to a carbon nanotube (CNT), between two graphene sheets (here with half-aperture α=19.45∘\alpha= 19.45^\circ). (b) The carbon HYP between two graphene sheets proposed in Ref. [24], with the same total curvature, −4π-4\pi, distributed over the whole surface. (c) A matrix of carbon HYPs between two graphene sheets, acting as a filter [24].

The half-apertures considered in Ref. [22] are those that preserve the hexagonal lattice of the nanocones, and they are precisely the angles αn=arcsin⁡(n/6)\alpha_n = \arcsin(n/6), n=4,3,2,1n=4,3,2,1, of Eq. (33): the Volterra cones of Figure 12, obtained by removing 6−n6-n wedges of π/3\pi/3, and the same cones tangent to ELL and BEL at the locations of their defects (§ 6.3). In particular, the optimal angle, 9.6∘=arcsin⁡(1/6)9.6^\circ= \arcsin(1/6), corresponds to the narrowest of these cones, tangent to BEL at R=r/6R = r/6, the last defect before the cusp.

The balanced Gauss–Bonnet theorem tells us something more about these structures. For a surface of revolution, the integrated Gaussian curvature of a band between two parallels is −2π-2\pi times the jump of R′=sin⁡αR' = \sin\alpha across the band (from K=−R′′/RK = -R''/R and d2μ=R du dvd^2\mu= R\,du\,dv). In the hourglass, nanotube (R′=0R' = 0), nanocone (R′=sin⁡αn=n/6R' = \sin\alpha_n = n/6) and graphene sheet (R′=1R' = 1) are all flat, and the curvature is concentrated on the junctions: −2πn/6-2\pi n/6, i.e., nn units of −π/3-\pi/3, on each CNT-CNC junction, and −2π(6−n)/6-2\pi(6-n)/6, i.e., 6−n6-n units, on each CNC-sheet junction. The total, −4π-4\pi, i.e., twelve units, is exactly that of HYP (Table 6), as it must be, since both connect a tube-like neck to two flat sheets. The difference is in the distribution: in the hourglass the twelve units must be carried by defects concentrated along four circular junctions, that in Ref. [22] are idealized as unbonded joints between three distinct nanostructures; in HYP they are spread over the whole surface, with sin⁡α=(c/r)sinh⁡(u/r)\sin\alpha= (c/r)\sinh(u/r) interpolating continuously between the nanotube-like waist and the flat sheets.

This is why there are clear indications [24] that, if one realizes carbon-made HYPs with the appropriate rr and cc parameters, one solves two problems at once: the HYP structure is unique, not made of three pieces, so that no junctions of the CNCs with the CNT are necessary; and the two ends of the carbon HYP are flat along the meridian direction, so that they fit smoothly with the planar graphene, see Figure 20(b) and Figure 1. Such a structure, a carbon-made HYP joining two graphene sheets, was realized in silico, for r=cr = c, in Ref. [17] (§ 8.3 and Fig. 18), and proved mechanically stable. Moreover, being able to tune the control parameters, rr and cc, according to needs and requirements (the radius of the neck, cc, and the ratio between mouth and neck, 1+r2/c2\sqrt{1+r^2/c^2}), is potentially interesting for creating prototypes of desalination or purification filters. A caveat is in order. The ordered, shell-like (“plug”) flow of water in nanotubes relies on a constant radius, while in HYP the radius is only approximately constant around the waist (R≤1.1cR \le1.1c over a length ≃0.9r\simeq0.9r). Whether such a neck suffices, and for which values of r/cr/c, has to be settled by dedicated simulations of water transport. The final structure should be a matrix of HYPs between planar graphene sheets, as in Figure 20(c), see Ref. [24].

A route to an experimental realization of carbon structures with this geometry is offered by zeolite-templated carbons (ZTCs), obtained by chemical vapor deposition of a carbon precursor inside the channels of a zeolite, followed by the removal of the template. The carbon inherits the three-dimensional topology of the zeolite, and in frameworks such as FAU (faujasite) and BEA (zeolite beta) the intersections of the channels produce hourglass-shaped regions of negative Gaussian curvature, with very narrow distributions of micropores, that can be regarded as experimentally accessible approximants of carbon HYPs. This is the strategy of a research project on the purification and desalination of water, also for lunar applications, that will interpret ZTCs as realizations of carbon pseudospheres [77].

Conclusions and future prospects

We have reviewed and extended the theory and practice of carbon-made Lobachevsky surfaces, organizing them around the balanced Gauss–Bonnet theorem: the total curvature fixed by the topology is shared among disclination defects, elastically deformed hexagons, and boundaries.

On the side of the defect-free curvature, we have shown that elastic deformations can carry intrinsic curvature, but cannot change the topology: a sphere made of hexagons only is a punctured sphere, as shown explicitly by the inverse stereographic image of the honeycomb, whose deficit angles add up to 4π4\pi only in the limit, at the missing point. Defects take over above a threshold set essentially by the integrated curvature, one sixth of the sphere for a pentagon.

On the side of the defects, we have shown that the discrete symmetries of the defects on the three pseudospheres stem from the elliptic (ELL), parabolic (BEL) and hyperbolic (HYP) one-parameter subgroups of SL(2,R)SL(2,\mathbb{R}), and that the defects sit where the surface has accumulated an integer number of units of curvature, π/3\pi/3. We have provided the explicit arrangement for the whole family of ELLs, recovered the group FYF_Y of BEL as their limit, and proposed the corresponding arrangement for HYP.

On the side of the actual carbon structures, we have confronted ideal HYPs by design with HYPs obtained by simulated extrusion and annealing: the total curvature charge, −4π-4\pi, is the same; its distribution is completely different, and the shape is recovered only when the boundary term is controlled.

Two lines of research are open. The first is quantitative: as seen with large fullerenes, and with the large ideal HYPs of § 8.1, the larger the carbon structure with isolated defects, the closer it is to a triangulated polyhedron, the hexagonal regions between the defects tending to be flat. Keeping the curvature between the defects requires deformed hexagons, i.e., stress, that can be imposed by a substrate or by the boundaries, as for the HYP embedded in flat sheets. A theory of how the non-topological curvature of the deformed hexagons and the topological curvature of the defects share the total curvature, for arbitrary configurations, is the main open problem that this work leaves us with.

The second is structural, and it is the one that motivated this program from the start. The finite subgroups of SO(3)SO(3) are few, and they are tied to the five Platonic solids. For the Lobachevsky geometry, this finiteness breaks down: SO(2,1)SO(2,1) has infinitely many discrete subgroups, the Fuchsian groups, among which are the infinitely many hyperbolic triangle groups, symmetry groups of infinitely many “Lobachevsky Platonic solids”. At the same time, there are infinitely many Lobachevsky surfaces in R3\mathbb{R}^{3}, each realizing a portion of the Lobachevsky plane. We have seen that the three pseudospheres, the simplest of them, correspond to the three classes of one-parameter subgroups of SL(2,R)SL(2,\mathbb{R}), with the discrete structures FCk−1FC_{k-1}, FYFY and FD6FD_{6}. It is tempting to conjecture a correspondence between the two infinities, discrete groups on one side and surfaces on the other, of which the pseudospheres are the first instance. To our knowledge, the crystallography of the Lobachevsky plane has been developed for hyperbolic lattices [49, 50] and for periodic minimal surfaces [52, 53], but not in connection with the constant-curvature surfaces embedded in R3\mathbb{R}^{3}. The next steps are then the construction of the full discrete groups for HYP and for the other Lobachevsky surfaces, such as the Dini surface, and the laboratory realization of these structures.

Acknowledgements

P.C., R.G., A.I. and P.P. gladly acknowledge support from Charles University Research Center (UNCE 24/SCI/016). T.P.C.K. and D.L. acknowledge support by the Ministry of Education, Youth and Sports of the Czech Republic through the e-INFRA CZ (ID:90254). O.D.L. and R.G.A. acknowledge the STAR funding: Progetto STAR X – 2583_QIIR111_00544, CUP H23C26000120007, PN RIC 2021-2027, Italian Ministry of University and Research. A.I. thanks the VIS programme of the Regione Calabria for a Visiting Professorship at the University of Calabria (UNICAL), where large parts of this work were carried out.

Declarations The authors declare no competing interests and no conflicts.

Data availability statement All data that support the findings of this study are included within the article. The scripts used for Section 3.3 are available as supplementary material; the input and output files of the simulations of Ref. [17] are available with that work; other data are available from the authors upon reasonable request.

Appendices

Deficit angle and Gaussian curvature of discrete structures

The Gaussian curvature K(u,v)K(u,v) is an intrinsic property of a surface, given at any point by the product of the two principal curvatures at that point. It opposes to extrinsic, or mean, curvature, MM, that is (half) the sum of the principal curvatures. We shall give other formulae for KK later here. For the moment see Ref. [32] and Ref. [33].

With this, surfaces with any (smooth enough) shape in one direction, and flat in the orthogonal direction, are intrinsically flat, K=0K=0, even though M≠0M\ne0 everywhere.

We shall be dealing with discrete structures, mostly hexagonal lattices. We shall also allow for deformations of the polygons. We then allow the edges to have different lengths, but not to bent, for the moment. One can think of those edges as elastic, but in the plane of the polygon.

For such structures the Gaussian/intrinsic curvature is given by the sum of the deficit (δ>0\delta>0) or excess (δ<0\delta<0) angles at every vertex of the structure

δ=2π−∑i=1qθi(44)\delta= 2\pi-\sum_{i=1}^{q}\theta_i \tag*{(44)}

with θi\theta_i the angles at the vertex of the qq, in general irregular, polygons, meeting there. See Figure 21(a). The case of tessellations of regular polygons is clearly included in this general definition.

Figure 21. (a) The angles θi\theta_i of the qq polygons meeting at a vertex and the deficit angle δV\delta_V; (b) square and (c) hexagonal tilings of the plane: zero deficit angle at every vertex.

Locally the curvature is then given, vertex by vertex, by KV=δVK_V=\delta_V, while, for a compact structure (no boundaries)

Ktot=∑all verticesKV(45)K_{\mathrm{tot}}=\sum_{\text{all vertices}}K_V \tag*{(45)}

When there are boundaries, the vertices on the boundary carry the discrete counterpart of the geodesic curvature, that is the exterior angle π−∑iθi\pi-\sum_i\theta_i, with θi\theta_i the angles of the polygons meeting there, cf. Eq. (7), and the Gauss–Bonnet theorem reads

∑V∈bulk(2π−∑iθi)+∑V∈boundary(π−∑iθi)=2πχ(46)\sum_{V\in\mathrm{bulk}}\left(2\pi-\sum_i\theta_i\right)+\sum_{V\in\mathrm{boundary}}\left(\pi-\sum_i\theta_i\right)=2\pi\chi \tag*{(46)}

The first sum is KtotK_{tot}, the second is the boundary term: the boundaries contribute to χ\chi, but not to the total Gaussian curvature. This is important when counting the number of necessary defects for the pseudospheres, that do have boundaries, see Refs. [1, 26] and Table 1.

Two elementary checks. In the square, hexagonal and triangular tilings of the plane (Figure 21(b),(Figure 21c) and Figure 2) δ=2π−4⋅π2=2π−3⋅2π3=2π−6⋅π3=0\delta= 2\pi- 4 \cdot\frac{\pi}{2} = 2\pi- 3 \cdot\frac{2\pi}{3} = 2\pi- 6 \cdot\frac{\pi}{3} = 0 at every vertex, so that KV=0K_V = 0 and Ktot=0K_{tot} = 0. For the tetrahedron, δ=2π−3⋅π3=π\delta= 2\pi- 3 \cdot\frac{\pi}{3} = \pi at each of its 4 vertices, and for the cube δ=2π−3⋅π2=π2\delta= 2\pi- 3 \cdot\frac{\pi}{2} = \frac{\pi}{2} at each of its 8 vertices: in both cases Ktot=4π=2π⋅2K_{tot} = 4\pi= 2\pi\cdot2, the correct total curvature for χ=2\chi= 2.

An illustrative counting

We report here the analysis of the output of the simulation of the carbon HYP obtained by nano-extrusion and annealing (§ 8.3). We do it to show the details of the counting of the intrinsic curvature associated with the discrete structure, that, although elementary, can be easily overlooked. The structure analyzed is one of the HYPs embedded between the two periodic graphene sheets, and the scars that extend across the Hilbert horizons into the sheets are counted in full. In the units used here the total curvature of the sphere is 1, so that −4π-4\pi corresponds to −1-1, i.e., twelve heptagons of −1/12-1/12 each.

The output file was analyzed to characterize morphology and arrangement of the defects. A defect was defined as a non-hexagonal ring, or a set of adjacent ones, and/or non-sp2sp^{2} carbons.

Non-hexagonal rings were mostly 7- and 5-sided. They formed thread-like structures, often one ring thick or, less commonly, two. These structures, also known as scars, are in most cases linear but they can bifurcate, typically branching at 7- and 8-sided polygons. One atom was found to be spsp-hybridized (2 bonds) and three others were sp3sp^{3} (4 bonds).

Heptagonal rings introduce −1/12-1/12, octagonal rings −1/6-1/6 and so on. In general, n-gons bring in (6−n)/12(6-n)/12 of curvature. spsp-hybridized carbons bring in 1/61/6 while sp3sp^{3} −1/6-1/6.

Visual inspection identified the entire expected total curvature. The fifteen defects found are listed in Table 7, and six representative ones are shown in Figure 22. All pictures have been generated with OVITO (Open Visualization Tool) [78], Basic 3.11.0, whose structure identification tool can match various templates, including graphene. For larger structures, an automated method, such as those implemented in polymake [79, 80] or rdkit, would be recommended.

halfdefectnotescurvature
topscar 1triangular ring with an sp3sp^{3} carbon−1/12-1/12
topscars 2, 3−1/12-1/12 each
topscar 4 (above)neutral00
topscar 4 (below)−1/12-1/12
topscar 5−1/6-1/6
topscars 6, 7−1/12-1/12 each
bottomscar 8−1/12-1/12
bottomscar 9contains an sp3sp^{3} carbon+1/12+1/12
bottomscar 10contains an sp3sp^{3} carbon−1/12-1/12
bottomscar 11−1/12-1/12
bottomscar 12contains an spsp carbon, neutral00
bottomscar 13−1/12-1/12
bottomisolated heptagon−1/12-1/12
total−1-1, i.e., −4π-4\pi

Table 7. Charged and most extensive neutral defects of the carbon HYP obtained by nano-extrusion and annealing [17], with their curvature in units of 4π4\pi (one heptagon gives −1/12-1/12). Scar 4 is split in two parts, one above and one below. The charge of a scar is assigned, conventionally, to one of its ends, and some scars cross from one half to the other: only the total is significant, not the partial sums of the two halves.

Figure 22. (a) isolated heptagon, −1/12-1/12; (b) scar 2, −1/12-1/12; (c) scar 5, −1/6-1/6; (d) scar 9, +1/12+1/12; (e) scar 12, 00; (f) scar 1, −1/12-1/12. Representative defects of the carbon HYP of Ref. [17] (see Table 7), with their curvature in units of 4π4\pi. Scar 9 (d) contains an sp3sp^3 carbon, scar 12 (e) an spsp carbon, and scar 1 (f), on top of two extra heptagons (−1/6-1/6), a triangular ring (+1/4+1/4) with an sp3sp^3 carbon (−1/6-1/6), so that it carries −1/12-1/12 overall.

References

  1. [1]Alfredo Iorio. Using Weyl symmetry to make graphene a real lab for fundamental physics. Eur. Phys. J. Plus, 127:156, 2012.DOI
  2. [2]Alfredo Iorio. Curved spacetimes and curved graphene: A status report of the weyl symmetry approach. Int. J. Mod. Phys. D, 24(05):1530013, 2015.DOI
  3. [3]Giovanni Acquaviva, Alfredo Iorio, Pablo Pais, and Luca Smaldone. Hunting quantum gravity with analogs: The case of graphene. Universe, 8(9):455, 2022.DOI
  4. [4]Alfredo Iorio. Weyl-gauge symmetry of graphene. Ann. Phys., 326(5):1334 – 1353, 2011.DOI
  5. [5]Alfredo Iorio and Gaetano Lambiase. The hawking–unruh phenomenon on graphene. Phys. Lett. B, 716(2):334–337, 2012.DOI
  6. [6]Alfredo Iorio and Gaetano Lambiase. Quantum field theory in curved graphene spacetimes, lobachevsky geometry, weyl symmetry, hawking effect, and all that. Phys. Rev. D, 90:025006, 2014.DOI
  7. [7]A Iorio, P Pais, IA Elmashad, AF Ali, Mir Faizal, and LI Abou-Salem. Generalized dirac structure beyond the linear regime in graphene. Int. J. Mod. Phys. D, 27(08):1850080, 2018.DOI
  8. [8]Alfredo Iorio and Pablo Pais. (anti-) de sitter, poincaré, super symmetries, and the two dirac points of graphene. Ann. Phys., 398:265–286, 2018.DOI
  9. [9]Alfredo Iorio and Pablo Pais. Generalized uncertainty principle in graphene. J. Phys.: Conf. Ser., 1275(1):012061, 2019.DOI
  10. [10]Alfredo Iorio, Gaetano Lambiase, Pablo Pais, and Fabio Scardigli. Generalized uncertainty principle in three-dimensional gravity and the btz black hole. Phys. Rev. D, 101(10):105002, 2020.DOI
  11. [11]Marcelo F Ciappina, Alfredo Iorio, Pablo Pais, and Adamantia Zampeli. Torsion in quantum field theory through time-loops on dirac materials. Phys. Rev. D, 101(3):036021, 2020.DOI
  12. [12]A Iorio, B Ivetić, S Mignemi, and P Pais. Three “layers” of graphene monolayer and their analog generalized uncertainty principles. Phys. Rev. D, 106(11):116011, 2022.DOI
  13. [13]Alfredo Iorio, Boris Ivetić, and Pablo Pais. Turning graphene into a lab for noncommutativity. Phys. Lett. B, 852:138630, 2024.DOI
  14. [14]A. H. Castro Neto, F. Guinea, N. M. R. Peres, K. S. Novoselov, and A. K. Geim. The electronic properties of graphene. Rev. Mod. Phys., 81:109–162, 2009.DOI
  15. [15]Valeriia Tynianska and Alfredo Iorio. Classical traversability of a carbon BTZ wormhole. arXiv:2610.37704 [gr-qc], 2026.arxiv.org/abs/2610.37704
  16. [16]Máximo Bañados, Claudio Teitelboim, and Jorge Zanelli. Black hole in three-dimensional spacetime. Phys. Rev. Lett., 69:1849–1851, 1992.pubmed.ncbi.nlm.nih.gov/10046331
  17. [17]T. P. C. Klaver, R. Gabbrielli, V. Tynianska, A. Iorio, and D. Legut. Stability of graphene hyperbolic pseudospheres under harsh conditions. Appl. Phys. A, 131(11):937, 2025.DOI
  18. [18]A. Ovchinnikov. Gallery of pseudospherical surfaces. In D Wójcik and J Cieśliński, editors, Nonlinearity and Geometry, pages 41–60. Polish Scientific Publishers PWN, Warsaw, 1998.
  19. [19]R. McLachlan. A gallery of constant negative curvature surfaces. Math. Intell., 16:31, 1994.DOI
  20. [20]R. K. Joshi, P. Carbone, F. C. Wang, V. G. Kravets, Y. Su, I. V. Grigorieva, H. A. Wu, A. K. Geim, and R. R. Nair. Precise and ultrafast molecular sieving through graphene oxide membranes. Science, 343(6172):752–754, 2014.
  21. [21]Stefania Castelletto and Alberto Boretti. Advantages, limitations, and future suggestions in studying graphene-based desalination membranes. RSC Adv., 11:7981–8002, 2021.
  22. [22]Vishnu Prasad Kurupath, Sridhar Kumar Kannam, Remco Hartkamp, and Sarith P. Sathian. Highly efficient water desalination through hourglass shaped carbon nanopores. Desalination, 505:114978, 2021.
  23. [23]Sidi Abdelmajid Ait Abdelkader, Ismail Benabdallah, Mohammed Amlieh, and Abdelouahad El Fatimy. Nanopore creation in graphene at the nanoscale for water desalination. ACS Omega, 10(9):9113–9119, 2025.
  24. [24]Alfredo Iorio and Paolo Castorina. Desalination with carbon pseudos­pheres. Research project italy–tunisia, Democritos Technologies, 2024.
  25. [25]Tohru Eguchi, Peter B. Gilkey, and Andrew J. Hanson. Gravitation, Gauge Theories and Differential Geometry. Phys. Rep., 66:213–393, 1980.DOI
  26. [26]Simone Taioli, Ruggero Gabbrielli, Stefano Simonucci, Nicola Maria Pugno, and Alfredo Iorio. Lobachevsky crystallography made real through carbon pseudos­pheres. J. Phys.: Condens. Matter, 28(13):13LT01, 2016.
  27. [27]H. S. Seung and D. R. Nelson. Defects in flexible membranes with crystalline order. Phys. Rev. A, 38:1005, 1988.pubmed.ncbi.nlm.nih.gov/9900464
  28. [28]H Kleinert. Gauge Fields in Condensed Matter. World Scientific, 1989.
  29. [29]M. O. Katanaev and I. V. Volovich. Theory of defects in solids and three-dimensional gravity. Ann. Phys., 216:1, 1992.DOI
  30. [30]Mark J. Bowick, David R. Nelson, and Alex Travesset. Interacting topological defects on frozen topographies. Phys. Rev. B, 62(13):8738–8751, 2000.DOI
  31. [31]A. R. Bausch, M. J. Bowick, A. Cacciuto, A. D. Dinsmore, M. F. Hsu, D. R. Nelson, M. G. Nikolaides, A. Travesset, and D. A. Weitz. Grain boundary scars and spherical crystallography. Science, 299:1716, 2003.
  32. [32]Carl Friedrich Gauss. Disquisitiones generales circa superficies curvas. Commentationes Societatis Regiae Scientiarum Gottingensis Recentiores, 6:99–146, 1828. Presented to the Royal Society of Göttingen on 8 October 1827.
  33. [33]Michael Spivak. A Comprehensive Introduction to Differential Geometry, volume I–V. Publish or Perish, Berkeley, 2nd edition, 1979.
  34. [34]Mark J. Bowick and Luca Giomi. Two-dimensional matter: order, curvature and defects. Adv. Phys., 58(5):449–563, 2009.DOI
  35. [35]Vincenzo Vitelli and David R. Nelson. Defect generation and deconfinement on corrugated topographies. Phys. Rev. E, 70:051105, 2004.DOI
  36. [36]Vincenzo Vitelli, Julius B. Lucks, and David R. Nelson. Crystallography on curved surfaces. Proc. Natl. Acad. Sci. USA, 103(33):12323–12328, 2006.DOI
  37. [37]William T. M. Irvine, Vincenzo Vitelli, and Paul M. Chaikin. Pleats in crystals on curved surfaces. Nature, 468:947, 2010.DOI
  38. [38]Amir Azadi and Gregory M. Grason. Emergent structure of multidislocation ground states in curved crystals. Phys. Rev. Lett., 112:225502, 2014.pubmed.ncbi.nlm.nih.gov/24949777
  39. [39]Amir Azadi and Gregory M. Grason. Neutral versus charged defect patterns in curved crystals. Phys. Rev. E, 94:013003, 2016.DOI
  40. [40]Matias Vincenzo Iorio. Using mathematics to prove why a traditional 32-panel football must consist of exactly 12 regular pentagons and 20 regular hexagons, and exploring the possibilities of constructing a similar spherical object using only hexagonal panels. IB Mathematics: Analysis and Approaches HL exploration, Riverside School, 2025.
  41. [41]W. Helfrich. Elastic properties of lipid bilayers: theory and possible experiments. Z. Naturforsch. C, 28:693–703, 1973.pubmed.ncbi.nlm.nih.gov/4273690
  42. [42]L. D. Landau and E. M. Lifshitz. Theory of Elasticity. Butterworth–Heinemann, Oxford, 3rd edition, 1986.
  43. [43]N. Levy, S. A. Burke, K. L. Meaker, M. Panlasigui, A. Zettl, F. Guinea, A. H. Castro Neto, and M. F. Crommie. Strain-induced pseudo-magnetic fields greater than 300 tesla in graphene nanobubbles. Science, 329(5991):544–547, 2010.DOI
  44. [44]Shikai Deng and Vikas Berry. Wrinkled, rippled and crumpled graphene: an overview of formation mechanism, electronic properties, and applications. Mater. Today, 19(4):197–212, 2016.DOI
  45. [45]M.A.H. Vozmediano, M.I. Katsnelson, and F. Guinea. Gauge fields in graphene. Phys. Rep., 496(4-5):109–148, 2010.DOI
  46. [46]Svetlana Katok. Fuchsian Groups. University of Chicago Press, Chicago, 1992.
  47. [47]Wilhelm Magnus. Noneuclidean Tessellations and Their Groups. Academic Press, New York, 1974.
  48. [48]Alan F. Beardon. The Geometry of Discrete Groups, volume 91 of Graduate Texts in Mathematics. Springer, New York, 1983.DOI
  49. [49]Joseph Maciejko and Steven Rayan. Hyperbolic band theory. Sci. Adv., 7(36):eabe9170, 2021.DOI
  50. [50]Igor Boettcher, Alexey V. Gorshkov, Alicia J. Kollár, Joseph Maciejko, Steven Rayan, and Ronny Thomale. Crystallography of hyperbolic lattices. Phys. Rev. B, 105:125118, 2022.DOI
  51. [51]Alicia J. Kollár, Mattias Fitzpatrick, and Andrew A. Houck. Hyperbolic lattices in circuit quantum electrodynamics. Nature, 571:45–50, 2019.
  52. [52]J. F. Sadoc and J. Charvolin. The crystallography of the hyperbolic plane and infinite periodic minimal surfaces. J. Phys. Colloques, 51(C7):C7–319–C7–332, 1990.DOI
  53. [53]Jean-François Sadoc and Rémy Mosseri. Geometrical Frustration. Cambridge University Press, Cambridge, 1999.DOI
  54. [54]Luther Pfahler Eisenhart. A treatise on the differential geometry of curves and surfaces. Ginn, 1909.
  55. [55]Koji Yamamoto, Tadashi Harada, Masao Nakazaki, Takuo Naka, Yasushi Kai, Shigeharu Harada, and Nobutami Kasai. Synthesis and characterization of [7]circulene. J. Am. Chem. Soc., 105(24):7171–7172, 1983.DOI
  56. [56]Koji Yamamoto, Yasuhiro Saitho, Daisuke Iwaki, and Takayuki Ooka. [7.7]Circulene, a molecule shaped like a figure of eight. Angew. Chem. Int. Ed. Engl., 30(9):1173–1174, 1991.
  57. [57]Ka Man Cheung, Yongming Xiong, Sai Ho Pun, Xingjian Zhuo, Qi Gong, Xingwei Zeng, Shilong Su, and Qian Miao. Negatively curved molecular nanocarbons containing multiple heptagons are enabled by the Scholl reactions of macrocyclic precursors. Chem, 9(10):2855–2868, 2023.DOI
  58. [58]Jonas Spengler, Yvonne Wagenhäuser, Marit Kuhstrebe, and Frank Würthner. Flexibility induced by four heptagons: A negatively curved polycyclic aromatic hydrocarbon cut-out of Schwarzite P216. J. Am. Chem. Soc., 2026.
  59. [59]T. P. C. Klaver and D. Legut. Magnetic moments in defect-rich graphene and graphene. Carbon, 262:122267, 2027.DOI
  60. [60]Patrice Simon and Yury Gogotsi. Materials for electrochemical capacitors. Nat. Mater., 7:845–854, 2008.
  61. [61]C. Rogers and W. K. Schief. Bäcklund and Darboux Transformations: Geometry and Modern Applications in Soliton Theory. Cambridge University Press, Cambridge, 2002.DOI
  62. [62]Máximo Bañados, Marc Henneaux, Claudio Teitelboim, and Jorge Zanelli. Geometry of the 2+1 black hole. Phys. Rev. D, 48:1506–1525, 1993.DOI
  63. [63]Marc Casals, Alessandro Fabbri, Cristián Martínez, and Jorge Zanelli. Quantum-corrected rotating black holes and naked singularities in (2 + 1) dimensions. Phys. Rev. D, 99:104023, 2019.DOI
  64. [64]Dieter Brill. 2+ 1-dimensional black holes with momentum and angular momentum. Ann. Phys. (Berlin), 512(3-5):217–226, 2000.arxiv.org/abs/gr-qc/9912079
  65. [65]Dieter Brill. Black Holes and Wormholes in 2+1 Dimensions. In Spiros Cotsakis and Gary W. Gibbons, editors, Mathematical and Quantum Aspects of Relativity and Cosmology, pages 143–179, Berlin, Heidelberg, 2000. Springer Berlin Heidelberg.
  66. [66]Dieter Brill. Geometry of black holes and Multi-Black-Holes in 2+ 1 dimensions. arXiv:gr-qc/9607026, 1996.arxiv.org/abs/gr-qc/9607026
  67. [67]Gorkem Memisoglu, Raghavan Chinnambedu Murugesan, Joseba Zubia, and Aleksey G. Rozhin. Graphene nanocomposite membranes: Fabrication and water treatment applications. Membranes, 13:145, 2023.DOI
  68. [68]Nikita S. Lazarenko, Valerii V. Golovakihin, Artem A. Shestakov, Nikita I. Lapekin, and Alexander G. Bannov. Recent advances on membranes for water purification based on carbon nanomaterials. Membranes, 12:915, 2022.DOI
  69. [69]Dong Han Seo, Shafque Pineda, Yun Chul Woo, Ming Xie, Adrian T. Murdock, Elisa Y. M. Ang, Yalong Jiao, Myoung Jun Park, Sung Il Lim, Malcolm Lawn, Fabricio Frizera Borghi, Zhao Jun Han, Stephen Gray, Graeme Millar, Aijun Du, Ho Kyong Shon, Teng Yong Ng, and Kostya Ostrikov. Anti-fouling graphene-based membranes for effective water desalination. Nat. Commun., 9:683, 2018.DOI
  70. [70]Albert Boretti, Sarim Al-Zubaidy, Miroslava Vaclavikova, Mohammed Al-Abri, Stefania Castelletto, and Sergey Mikhaiovsky. Outlook for graphene-based desalination membranes. npj Clean Water, 1:5, 2018.DOI
  71. [71]Gerhard Hummer, Joseph C. Rasaiah, and John P. Noworyta. Water conduction through the hydrophobic channel of a carbon nanotube. Nature, 414:188–190, 2001.pubmed.ncbi.nlm.nih.gov/11700553
  72. [72]Kenichiro Koga, G. T. Gao, Hideki Tanaka, and Xiao Cheng Zeng. Formation of ordered ice inside carbon nanotubes. Nature, 412:802–805, 2001.
  73. [73]Ruiqian Wang, J. Chen, L. Chen, Z. Ye, C. Wu, W. Gao, X. Li, Y. Zhang, and Y. Ying. Ultrathin and ultradense aligned carbon nanotube membranes for water purification with enhanced rejection performance. Desalination, 494:114671, 2020.DOI
  74. [74]Rasel Das, Md. Eaqub Ali, Sharifah Bee Abd Hamid, Seeram Ramakrishna, and Zaira Zaman Chowdhury. Carbon nanotube membranes for water purification: A bright future in water desalination. Desalination, 336:97–109, 2014.DOI
  75. [75]Chris Buelke, Ali Alshami, James Casler, Jeremy Lewis, Maram Al-Sayaghi, and Michael A. Hickner. Graphene oxide membranes for enhancing water purification in terrestrial and space-born applications: State of the art. Desalination, 448:113–132, 2018.DOI
  76. [76]Simon Gravelle, Laurent Joly, François Detcheverry, Christophe Ybert, Lyderic Bocquet, and Elisabeth Charlaix. Optimizing water permeability through the hourglass shape of aquaporins. Proc. Natl. Acad. Sci. USA, 110:16367–16372, 2013.DOI
  77. [77]R. G. Agostino. ZeoLunar: bio-inspired carbon structures and zeolites for the purification and desalination of lunar water — from the carbon pseudosphere to a proof of concept. PhD research project, University of Calabria and UOP S.r.l. (Honeywell), 2026.
  78. [78]Alexander Stukowski. Visualization and analysis of atomistic simulation data with ovito–the open visualization tool. Model. Simul. Mater. Sci. Eng., 18(1):015012, 2010.DOI
  79. [79]Ewgenij Gawrilow and Michael Joswig. polymake: a framework for analyzing convex polytopes. In Gil Kalai and Günter M. Ziegler, editors, Polytopes—Combinatorics and Computation, pages 43–73. Birkhäuser Basel, Basel, 2000.DOI
  80. [80]Benjamin Assarf, Ewgenij Gawrilow, Katrin Herr, Michael Joswig, Benjamin Lorenz, Andreas Paffenholz, and Thomas Rehn. Computing convex hulls and counting integer points with polymake. Math. Program. Comput., 9(1):1–38, 2017.arxiv.org/abs/1408.4653

Paper details

vlnc:W_uwn9e3tz
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 5

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents