betaFeedback

WorksMathematics

Extrinsic Representation of the Partial Ricci Flow

Abstract

The partial Ricci flow (PRF) modifies the Riemannian metric on a foliation on a distribution orthogonal to the leaves. In this work we develop extrinsic representations of PRF on geodesic foliations (twisted/warped products, weak K-contact and weak \b{eta}-Kenmotsu manifolds) and analyze model cases where PRF admits closed-form solutions. Using the convergence of PRF on model cases, we construct its extrinsic representations initialized by rotational hypersurfaces in Euclidean space as well as parallel curved surfaces and ellipsoids. The extrinsic viewpoint for PRF provides new tools for studying curvature-driven dynamics on foliations, with potential applications to calculus of moving surfaces and mathematical physics.

Introduction

By geometric flow we mean the evolution of a metric gg on a manifold under a differential equation, ∂tg=B(g)\partial_t g = B(g), where the tensor BB is associated with curvature. The Ricci flow, and the mean curvature flow are most popular geometric flows. In [4, 8, 6] geometric flows on a foliated manifold are studied, the metric gg is modified along the normal to the leaves distribution, and the evolution operator is elliptic along the leaves. The partial Ricci flow (PRF), on a foliated manifold (M,F)(M,\mathcal{F})

∂tg=−2Ric⁡g⊥+2rg⊥,r∈R(1)\partial_t g=-2\operatorname{Ric}^{\perp}_{g}+2r g^{\perp},\qquad r\in\mathbb{R} \tag*{(1)}

where Ric⁡g⊥\operatorname{Ric}^{\perp}_{g} is the partial Ricci curvature, and g⊥g^{\perp} is the restriction of gg on the distribution D\mathcal{D} orthogonal to F\mathcal{F}, is the tool to prescribe the mixed sectional curvature, KmixK_{\mathrm{mix}}. The PRF preserves D\mathcal{D} and such geometric properties of the leaves as totally umbilical, geodesic or harmonic (minimal) submanifolds. The symmetric (0,2)(0,2)-tensor Ric⁡g⊥(X,Y)=∑ag(R(X⊥,ea)Y⊥,ea)\operatorname{Ric}^{\perp}_{g}(X,Y)=\sum_a g(R(X^{\perp},e_a)Y^{\perp},e_a), where {ea}\{e_a\} is any local orthonormal basis of D\mathcal{D}, is called the partial Ricci curvature of a foliation F\mathcal{F} of (M,g)(M,g). For a one-dimensional foliation F\mathcal{F} and a unit vector ξ∈TF\xi\in T\mathcal{F}, we have Ric⁡g⊥(X,Y)=g(RXξ,ξ,Y)=g(RξX,Y)\operatorname{Ric}^{\perp}_{g}(X,Y)=g(R_X\xi,\xi,Y)=g(R_{\xi}X,Y), where RξR_{\xi} is the Jacobi operator for ξ\xi. For a foliated surface (M2,F,g)(M^2,\mathcal{F},g), we have Ric⁡g⊥=Kg⊥\operatorname{Ric}^{\perp}_{g}=K g^{\perp} and Rξ=KIDR_{\xi}=K I_{\mathcal{D}}, where KK is the gaussian curvature, hence, (1.1) reduces to the D\mathcal{D}-conformal flow

∂tg=−2(K−r)g⊥.\partial_t g=-2(K-r)g^{\perp}.

For better understanding and visualization the geometric flows, we can represent them extrinsically by depicting a submanifold deforming in Euclidean space, when the metric of the flow is the metric induced by the manifold’s embedding. In this paper, we adapt to PRF the following concept, studied in [2] for the Ricci flow.

An extrinsic representation ensures that, given a geometric flow gtg_t on a differentiable manifold MM, there is a smooth family of immersions ψt:M→St\psi_t:M\to S_t (t∈Rt\in\mathbb{R}) in a fixed Euclidean space, such that the metric extrinsically induced on StS_t by the ambient Euclidean metric is precisely gtg_t.

This approach is motivated by Nash’s embedding theorems, which state that any Riemannian manifold can be isometrically embedded into some Euclidean space. Moreover, for a short time, one can construct a smooth one-parameter family of isometric embeddings of any smooth parametrized family of Riemannian metrics on a smooth closed manifold into Euclidean space, see [10], in particular, the Ricci flow on MnM^n can be represented in Rm\mathbb{R}^m, where m=max⁡{12n(n+5),12(n+3)+5}m=\max\{\frac{1}{2}n(n+5),\frac{1}{2}(n+3)+5\}.

The PRF exists for a short time, see [6], so what is the relation between mm and nn for the extrinsic representations of the PRF on a smooth foliated manifold MnM^n in Rm\mathbb{R}^m?

Some authors studied extrinsic representations of the Ricci flow initiated by rotational hypersurfaces in space forms, e.g., [2, 9], whose metrics are rotational symmetric – a special class of warped products. Meridian curves of a rotational hypersurface form a geodesic foliation. Such foliations arise naturally on twisted and warped products B×ρMˉB\times_{\rho}\bar{M}, on weak β\beta-Kenmotsu manifolds and weak contact metric manifolds M(f,ξ,η,g)M(f,\xi,\eta,g), see [7, 5, 8], obtained by replacing the complex structure of the distribution ker η\eta with a non-singular skew-symmetric tensor. The main achievement of the PRF for geodesic foliations is the dynamical characterizations of KK-contact manifolds and warped products of constant mixed curvature, KmixK_{\mathrm{mix}}

– PRF evolves a weak KK-contact structure (with Kmix>0K_{\mathrm{mix}}>0), exponentially fast towards a KK-contact structure (with Kmix≡1K_{\mathrm{mix}}\equiv1)

– PRF evolves warped products toward warped products with Kmix=constK_{\mathrm{mix}}=\mathrm{const}. This has practical implications because a simply connected weak Kenmotsu β\beta-manifold with a complete Reeb field ξ\xi such that ∇β\nabla\beta is collinear with ξ\xi is a warped product.

These advances raise the following problem: to study extrinsic representations of the PRF on geodesic foliations initialized by rotational hypersurfaces in Euclidean space, or hypersurfaces from other distinguished classes.

The paper focuses on extrinsic representations of PRF on geodesic foliations, when gtg_t correspond to a warped product, β\beta-Kenmotsu or KK-contact structures, and StS_t are rotational hypersurfaces or ellipsoids, respectively. We distinguish 3 assertions: existence of the PRF, realization by a certain isometric immersion, and smooth extension of such immersion across a pole. The first doesn’t imply the other two.

In Section 2 we show the convergence of the PRF for some warped and twisted products, and give examples. In Section 3 we study the PRF initiated by rotational hypersurfaces in Euclidean space and by parallel curved surfaces. In Section 4 we recall the notion of weak almost contact metric manifolds and study extrinsic representations of the PRF on weak KK-contact manifolds initiated by ellipsoids, and on weak β\beta-Kenmotsu manifolds initiated by rotational hypersurfaces.

The PRF on Twisted and Warped Products

The notion of a warped and twisted product is popular in differential geometry as well as in general relativity (some spacetime models are warped products).

Let (Bs,gB)(B^{s},g_{B}) be a compact Riemannian manifold with boundary, and (M‾,g‾)(\overline{M},\overline{g}) be an nn-dimensional Riemannian manifold. The twisted product B×ρM‾B\times_{\rho}\overline{M} is the manifold M=B×M‾M=B\times\overline{M} with the metric g=gB+ρ2g‾g=g_{B}+\rho^{2}\overline{g}, where ρ∈C∞(M)\rho\in C^{\infty}(M) is a positive function. When ρ\rho depends on BB only, the twisted product is a warped product. The special case of a warped product are rotational symmetric metrics, i.e., (M‾,g‾)(\overline{M},\overline{g}) is a unit nn-sphere, such metrics appear on rotational hypersurfaces in space forms.

The leaves B×{v}B\times\{v\} of a twisted product are totally geodesic, and the fibers {u}×M‾\{u\}\times\overline{M} are totally umbilical with the second fundamental form h=−∇F(log⁡ρ)g‾h=-\nabla^{F}(\log\rho)\overline{g}. For twisted products Bs×ρM‾nB^{s}\times_{\rho}\overline{M}^{n}, we have Ric⁡g⊥=(Smixg/n)g⊥\operatorname{Ric}^{\perp}_{g}=(S^{g}_{\mathrm{mix}}/n)g^{\perp}, where Smixg=tr⁡Ric⁡g⊥S^{g}_{\mathrm{mix}}=\operatorname{tr}\operatorname{Ric}^{\perp}_{g}, and the PRF becomes a D\mathcal{D}-conformal flow

∂tg=−2(Smixg−r)g⊥.\partial_{t}g=-2\left(S^{g}_{\mathrm{mix}}-r\right)g^{\perp}.

Since R(X,ξ,Y,ξ)=−ρ−1ξ(ξ(ρ))g‾(X,Y)R(X,\xi,Y,\xi)=-\rho^{-1}\xi(\xi(\rho))\overline{g}(X,Y) for unit ξ∈TF\xi\in TF, the mixed sectional curvature KmixK_{\mathrm{mix}} depends on a point of BB only, and Ric⁡g⊥=−(ΔFρ/ρ)g⊥\operatorname{Ric}^{\perp}_{g}=-(\Delta_{F}\rho/\rho)g^{\perp}. Here, ∇F\nabla^{F} and ΔF\Delta_{F} denote, respectively, the gradient and the laplacian on the leaves.

If twisted product metrics gt=gB+ρ2(t,u,v)g‾g_{t}=g_{B}+\rho^{2}(t,u,v)\overline{g} solve the PRF on B×M‾B\times\overline{M}, then ∂t(ρ2)g‾=2(ΔFρ/ρ)ρ2g‾+2rρ2g‾\partial_{t}(\rho^{2})\overline{g}=2(\Delta_{F}\rho/\rho)\rho^{2}\overline{g}+2r\rho^{2}\overline{g}. This yields the leaf-wise parabolic PDE

∂tρ=ΔFρ+rρ.\partial_{t}\rho=\Delta_{F}\rho+r\rho.

Let μ(t,u)=ρ(t,u)∣∂B\mu(t,u)=\rho(t,u)|_{\partial B} be twice continuously differentiable in tt, and

lim⁡t→∞μ(t,u)=μ~(u),lim⁡t→∞∂tμ(t,u)=0,lim⁡t→∞∂t2μ(t,⋅)=0(2)\lim_{t\to\infty}\mu(t,u)=\widetilde{\mu}(u),\qquad\lim_{t\to\infty}\partial_{t}\mu(t,u)=0,\qquad\lim_{t\to\infty}\partial_{t}^{2}\mu(t,\cdot)=0 \tag*{(2)}

for a smooth nonnegative function μ~:∂B→R\widetilde{\mu}:\partial B\to\mathbb{R} uniformly for u∈∂Bu\in\partial B.

Define

δ(t,⋅)=μ(t,⋅)−μ~,\delta(t,\cdot)=\mu(t,\cdot)-\widetilde{\mu},
ν(t)=max⁡{∥δ(t,⋅)∥C0(∂B),∥∂tδ(t,⋅)∥C0(∂B)},\nu(t)=\max\left\{\|\delta(t,\cdot)\|_{C^{0}(\partial B)},\|\partial_{t}\delta(t,\cdot)\|_{C^{0}(\partial B)}\right\},
ν~(t)=max⁡{∥δ(t,⋅)∥C0(∂B),∥∂tδ(t,⋅)∥C0(∂B),∥∂t,t2δ(t,⋅)∥C0(∂B)}.\widetilde{\nu}(t)=\max\left\{\|\delta(t,\cdot)\|_{C^{0}(\partial B)},\|\partial_{t}\delta(t,\cdot)\|_{C^{0}(\partial B)},\|\partial_{t,t}^{2}\delta(t,\cdot)\|_{C^{0}(\partial B)}\right\}.

We first record the fact identifying the PRF with the profile heat equation, it is the hinge on which the extrinsic representation turns.

Theorem 2.1 (Rovenski-Zelenko). Let the warped product metrics gt=gB+ρ2(t,u)g‾g_{t}=g_{B}+\rho^{2}(t,u)\overline{g} on Bp×ρM‾nB^{p}\times_{\rho}\overline{M}^{n} solve the PRF (1.1), and any of conditions i)–iii) are satisfied

i) r<0r<0 and (2.1)1,2_{1,2}

ii) 0≤r<λ10\leq r<\lambda_{1}, p<4p<4, and (2.1)

iii) r=λ1r=\lambda_{1}, p<4p<4, (2.1) and μ~=0\widetilde{\mu}=0, ∫0∞ν(τ) dτ<∞\int_{0}^{\infty}\nu(\tau)\,d\tau<\infty.

Then gtg_{t} exist for all t≥0t\geq0, and converge, as t→∞t\to\infty, uniformly on Bp×MnB^{p}\times M^{n} in C0C^{0}-norm to the limit warped product metric g∞=gB+ρ∞2(u)g‾g_{\infty}=g_{B}+\rho_{\infty}^{2}(u)\overline{g} with Ric⁡g∞⊥=rg‾\operatorname{Ric}^{\perp}_{g_{\infty}}=r\overline{g}. Moreover, in cases i) and ii), the PRF has a global single point attractor, while in case iii) the limit metric g∞g_{\infty} depends on initial and boundary conditions.

Proof. Let 0<λ1≤…≤λi≤…0 < \lambda_1 \le\ldots\le\lambda_i \le\ldots be eigenvalues of the Schrödinger operator −ΔF−r-\Delta_{\mathcal F}-r on the leaves B×{y}B \times\{y\} counting their multiplicities, lim⁡j→∞λj=∞\lim_{j\to\infty}\lambda_j=\infty, and {ei}1≤i<∞\{e_i\}_{1\le i<\infty} corresponding L2L_2-orthonormal basis of eigenfunctions with ei=0e_i=0 on ∂B\partial B of the eigenvalue problem: −ΔFei=λiei-\Delta_{\mathcal F}e_i=\lambda_i e_i. Then, λ1\lambda_1 has multiplicity 1 (hence, λ1<λ2\lambda_1<\lambda_2) and e1>0e_1>0 on the interior of BB. The leaf-wise parabolic Cauchy’s problem

∂tρ=ΔFρ+rρ,ρ(0,⋅)=ρ0,ρ(t,⋅)∣∂B=μ(t,⋅)(3)\partial_t \rho= \Delta_{\mathcal F}\rho+ r\rho,\qquad\rho(0,\cdot)=\rho_0,\qquad\rho(t,\cdot)|_{\partial B}=\mu(t,\cdot) \tag*{(3)}

has a unique solution ρ:[0,∞)×B→R\rho:[0,\infty)\times B\to\mathbb{R}. One can prove convergence of ρ\rho, as t→∞t\to\infty, to a stationary state – a solution ρ∞:B→R\rho_\infty:B\to\mathbb{R} of the problem

−ΔFρ∞=rρ∞,ρ∞∣∂B=μ~.(4)-\Delta_{\mathcal F}\rho_\infty=r\rho_\infty,\qquad\rho_\infty|_{\partial B}=\widetilde{\mu}. \tag*{(4)}

The rest of proof is similar to the proof of [6 Theorem 5.98].

For p=1p=1 and B=[0,1]B=[0,1], the problems (3) and (4) have the following view

ρt′=ρu,u′′+rρ,ρ(0,⋅)=ρ0,ρ(t,0)=μ0(t),ρ(t,1)=μ1(t)(5)\rho'_t=\rho''_{u,u}+r\rho,\qquad\rho(0,\cdot)=\rho_0,\qquad\rho(t,0)=\mu_0(t),\qquad\rho(t,1)=\mu_1(t) \tag*{(5)}
ρ∞′′+rρ∞=0 on (0,1),ρ∞(0)=α1∈R,ρ∞(1)=α2∈R.(6)\rho''_\infty+r\rho_\infty=0\ \text{on }(0,1),\qquad\rho_\infty(0)=\alpha_1\in\mathbb{R},\qquad\rho_\infty(1)=\alpha_2\in\mathbb{R}. \tag*{(6)}

To describe stationary profiles for r∈Rr\in\mathbb{R}, we use the following pair of solutions of y′′+ry=0y''+ry=0 normalized by y(0)=0y(0)=0, y′(0)=1y'(0)=1 and y(0)=1y(0)=1, y′(0)=0y'(0)=0, respectively,

Sn⁡r(u)={sin⁡(ru)/rr>0ur=0sinh⁡(−ru)/−rr<0Cn⁡r(u)={cos⁡(ru)r>01r=0cosh⁡(−ru)r<0.\operatorname{Sn}_r(u)= \begin{cases} \sin(\sqrt{r}u)/\sqrt{r} & r>0\\ u & r=0\\ \sinh(\sqrt{-r}u)/\sqrt{-r} & r<0 \end{cases} \qquad \operatorname{Cn}_r(u)= \begin{cases} \cos(\sqrt{r}u) & r>0\\ 1 & r=0\\ \cosh(\sqrt{-r}u) & r<0. \end{cases}

The functions Sn⁡r\operatorname{Sn}_r, Cn⁡r\operatorname{Cn}_r are jointly smooth in (u,r)(u,r) and satisfy the following relations: Sn⁡r′=Cn⁡r\operatorname{Sn}'_r=\operatorname{Cn}_r, Cn⁡r′=−rSn⁡r\operatorname{Cn}'_r=-r\operatorname{Sn}_r, Cn⁡r2+rSn⁡r2=1\operatorname{Cn}_r^2+r\operatorname{Sn}_r^2=1.

Corollary 2.2 (Stationary profiles). Let r<π2r<\pi^2 and α1,α2∈R\alpha_1,\alpha_2\in\mathbb{R}. The problem (6) has a unique solution ρ∞(u,r;α1,α2)=α2Sn⁡r(u)+α1Sn⁡r(1−u)Sn⁡r(1)\rho_\infty(u,r;\alpha_1,\alpha_2)=\frac{\alpha_2\operatorname{Sn}_r(u)+\alpha_1\operatorname{Sn}_r(1-u)}{\operatorname{Sn}_r(1)} linear in α1,α2\alpha_1,\alpha_2. For r=π2r=\pi^2 and α1,α2≥0\alpha_1,\alpha_2\ge0 with α12+α22>0\alpha_1^2+\alpha_2^2>0, (6) is solvable if and only if α2=−α1\alpha_2=-\alpha_1, in which case solutions are ρ∞(u,r;α1,α2)=Csin⁡(πu)\rho_\infty(u,r;\alpha_1,\alpha_2)=C\sin(\pi u) for C∈RC\in\mathbb{R}.

Example 2.3. i) For M=[0,1]×S1M=[0,1]\times S^1 the twisted product (semi-geodesic) metrics are

g=du2+ρ2(u,v) dv2,ρ>0.(7)g=du^2+\rho^2(u,v)\,dv^2,\qquad\rho>0. \tag*{(7)}

For the metric (7), each uu-curve (i.e., v=constv=\mathrm{const}) is a unit-speed geodesic, and the Gaussian curvature is K=−ρu,u′′/ρK=-\rho''_{u,u}/\rho (involves no vv-derivatives of ρ\rho).

ii) Let gt=du2+ρ2(t,u)gSn−1g_t=du^2+\rho^2(t,u)g_{S^{n-1}}, ρ(t,⋅)>0\rho(t,\cdot)>0 on (0,1)(0,1), be rotationally symmetric metrics on [0,1]×Sn−1[0,1]\times S^{n-1}. Then gtg_t solve the PRF (1.1) if and only if (5) holds.

Indeed, the leaves {ω=const}\{\omega=\mathrm{const}\} compose a geodesic foliation F\mathcal F with unit tangent N=∂uN=\partial_u and leaf-wise Laplacian ΔF=∂u,u2\Delta_{\mathcal F}=\partial^2_{u,u}, and D=TSn−1D=TS^{n-1}, this is the warped product [0,1]×ρSn−1[0,1]\times_\rho S^{n-1}. We obtain the formula RN(t)=−(ρu,u′′/ρ)gtR_N(t)=-(\rho''_{u,u}/\rho)g_t, while gt=ρ2gSn−1g_t=\rho^2g_{S^{n-1}} and ∂tgt=(ρ2)t′gSn−1=2ρρt′gSn−1\partial_tg_t=(\rho^2)'_t g_{S^{n-1}}=2\rho\rho'_t g_{S^{n-1}}. Thus, the PRF (1.1) reads 2ρρt′=2(ρu,u′′/ρ)ρ2+2rρ22\rho\rho'_t=2(\rho''_{u,u}/\rho)\rho^2+2r\rho^2, that reduces to (5).

Proposition 2.4. Consider metric g0=du2+ρ02(u,v) dv2g_0=du^2+\rho_0^2(u,v)\,dv^2 on [0,1]×S1[0,1]\times S^1 with smooth ρ0>0\rho_0>0 on (0,1)×S1(0,1)\times S^1, with boundary values α1(v):=ρ0(0,v)≥0\alpha_1(v):=\rho_0(0,v)\ge0 and α2(v):=ρ0(1,v)≥0\alpha_2(v):=\rho_0(1,v)\ge0 fixed in time. Then the following holds

i) the PRF with the boundary conditions ρ(t,0,v)=α1(v)\rho(t,0,v)=\alpha_1(v) and ρ(t,1,v)=α2(v)\rho(t,1,v)=\alpha_2(v) has a unique solution gt=du2+ρ2(t,u,v) dv2g_t=du^2+\rho^2(t,u,v)\,dv^2 for all t≥0t\geq0, it is given for r<λ1r<\lambda_1 by

ρ(t,u,v)=ρ∞(u,r,α1(v),α2(v))+∑n=1∞An(v)sin⁡(nπu)exp⁡((r−λn)t)\rho(t,u,v)=\rho_\infty(u,r,\alpha_1(v),\alpha_2(v))+\sum_{n=1}^{\infty}A_n(v)\sin(n\pi u)\exp((r-\lambda_n)t)

where An(v)=2∫01(ρ0(u,v)−ρ∞(u,r,α1(v),α2(v)))sin⁡(nπu) duA_n(v)=2\int_0^1(\rho_0(u,v)-\rho_\infty(u,r,\alpha_1(v),\alpha_2(v)))\sin(n\pi u)\,du. The function ρ>0\rho>0 on (0,1)×S1×[0,∞)(0,1)\times S^1\times[0,\infty), so gtg_t is a Riemannian metric on the interior for all tt, degenerating at a boundary circle when the boundary value vanishes.

ii) if r<λ1r<\lambda_1 then, ∥ρ(t,⋅,v)−ρ∞(⋅,r,α1(v),α2(v))∥Ck([0,1])≤Ckexp⁡((r−λ1)t)\|\rho(t,\cdot,v)-\rho_\infty(\cdot,r,\alpha_1(v),\alpha_2(v))\|_{C^k([0,1])}\leq C_k\exp((r-\lambda_1)t) (t≥1t\geq1), uniformly in vv for every k≥0k\geq0, consequently, gt→g∞g_t\to g_\infty in every CkC^k norm, where g∞=du2+ρ∞2(u,r,α1(v),α2(v)) dv2g_\infty=du^2+\rho_\infty^2(u,r,\alpha_1(v),\alpha_2(v))\,dv^2 has constant Gaussian curvature K≡rK\equiv r.

iii) if r=λ1r=\lambda_1, a limit exists only if α1≡α2≡0\alpha_1\equiv\alpha_2\equiv0, in that case

lim⁡t→∞ρ(t,u,v)=B1(v)sin⁡(πu),B1(v)=2∫01ρ0(u,v)sin⁡(πu) du>0\lim_{t\to\infty}\rho(t,u,v)=B_1(v)\sin(\pi u),\qquad B_1(v)=2\int_0^1\rho_0(u,v)\sin(\pi u)\,du>0

in any CkC^k norm, and the limit metric has constant curvature rr on the interior.

iv) if r>λ1r>\lambda_1, then lim⁡t→∞gt\lim_{t\to\infty}g_t exists if and only if for any vv and nn with λn≤r\lambda_n\leq r, the coefficient An(v)A_n(v) of the solution vanishes (with ρ∞\rho_\infty replaced by any particular stationary solution when one exists), for generic data no limit exists.

Proof. Fourier series are used here in the standard way.

The PRF on Rotational and Parallel Curved Surfaces

In this section, we construct an extrinsic representation of the PRF (1.1) initialized by rotational hypersurfaces and parallel curved surfaces of Euclidean space.

Rotational Hypersurfaces

Rotational hypersurfaces with a one-dimensional axis in Rn+1\mathbb{R}^{n+1} are foliated into (n−1)(n-1)-spheres (parallels) and equipped with rotationally symmetric metrics g=du2+ρ2(u) dsn−12g=du^2+\rho^2(u)\,ds_{n-1}^2 — a special case of a warped product. Such a hypersurface (and its profile) can be given as a graph ρ=F(xn+1)\rho=F(x_{n+1}), where F∈C2F\in C^2, ρ=x12+⋯+xn2\rho=\sqrt{x_1^2+\cdots+x_n^2} and (xi)(x_i) are Cartesian coordinates in Rn+1\mathbb{R}^{n+1}. We assume F′′≢0F''\not\equiv0 to exclude the cases when the profile curve is a straight, and MM is a cone, a cylinder, or a hyperplane.

A rotational hypersurface in Euclidean space Rn+1\mathbb{R}^{n+1} with a multi-dimensional axis is a hypersurface obtained by rotating a lower-dimensional profile around a linear subspace (the “axis”) of dimension >1>1 (not a line). So, instead of the rotational hypersurface with axis = line, we get a foliated hypersurface invariant under the group O(n+1−s)O(n+1-s) acting on the orthogonal complement of the ss-dimensional axis: for each y∈Rsy\in\mathbb{R}^s, the fiber in xx is a round (n−s)(n-s)-sphere of radius ρ=F(y)\rho=F(y) in Rn+1−s\mathbb{R}^{n+1-s}.

Proposition 3.1. Let Sn⊂Rn+1S^n\subset\mathbb{R}^{n+1} be a rotational hypersurface with a profile curve (ρ,h)(\rho,h) parameterized by arclength, where ρ,h:[0,1]→R\rho,h:[0,1]\to\mathbb{R} are smooth and ρ>0\rho>0 on (0,1)(0,1). If SnS^n has Kmax⁡=r=const>0K_{\max}=r=\mathrm{const}>0, then r≤π2r\leq\pi^2, and for r=π2r=\pi^2, we have ρ(0)=ρ(1)=0\rho(0)=\rho(1)=0, and for any C∈(0,1/π]C\in(0,1/\pi], the functions (ρ,h)(\rho,h) are given by

ρ(u)=Csin⁡(πu),h(u)=∫0u1−(πC)2cos⁡2(πs) ds.(8)\rho(u)=C\sin(\pi u),\qquad h(u)=\int_0^u\sqrt{1-(\pi C)^2\cos^2(\pi s)}\,ds. \tag*{(8)}

:::`

Proof. This is standard.

Example 3.2 (see Example 5.101 in [6]). Consider rotational surfaces St⊂R3S_t \subset\mathbb{R}^3

r=(ρ(t,u)cos⁡v,ρ(t,u)sin⁡v,h(t,u)),0≤u≤1,∣v∣≤π,ρ≥0\mathbf{r}=(\rho(t,u)\cos v,\rho(t,u)\sin v,h(t,u)),\qquad0\leq u\leq1,\qquad|v|\leq\pi,\qquad\rho\geq0

with (ρu′)2+(hu′)2=1(\rho'_u)^2+(h'_u)^2=1. Profile curves v=constv=\mathrm{const} of StS_t are unit speed geodesics. The vv-curves of StS_t are circles (in R3\mathbb{R}^3) of geodesic curvature k=−(log⁡ρ)u′k=-(\log\rho)'_u. The metric gt=du2+ρ2(t,u) dv2g_t=du^2+\rho^2(t,u)\,dv^2 is rotational symmetric, and its Gaussian curvature is K=−ρu,u′′/ρK=-\rho''_{u,u}/\rho. Let gtg_t solve the PRF, then ρ(t,u)\rho(t,u) solves the problem (5).

Assume that μ0,μ1∈C1[0,∞)\mu_0,\mu_1\in C^1[0,\infty), then there exist limits lim⁡t→∞μj(t)=μ~j∈[0,∞)\lim_{t\to\infty}\mu_j(t)=\widetilde{\mu}_j\in[0,\infty).

The solution of stationary problem (6) with B=[0,1]B=[0,1] and λj=(πj)2\lambda_j=(\pi j)^2 is

ρ∞(u)={μ~1sin⁡(r u)+μ~0sin⁡(r(1−u))sin⁡(r)if 0<r<λ1,μ~0+(μ~1−μ~0)(u)if r=0,μ~1sinh⁡(−r u)+μ~0sinh⁡(−r(1−u))sinh⁡(−r)if r<0.(9)\rho_\infty(u)=\begin{cases} \dfrac{\widetilde{\mu}_1\sin(\sqrt{r}\,u)+\widetilde{\mu}_0\sin(\sqrt{r}(1-u))}{\sin(\sqrt{r})} & \text{if }0<r<\lambda_1,\\ \widetilde{\mu}_0+(\widetilde{\mu}_1-\widetilde{\mu}_0)(u) & \text{if }r=0,\\ \dfrac{\widetilde{\mu}_1\sinh(\sqrt{-r}\,u)+\widetilde{\mu}_0\sinh(\sqrt{-r}(1-u))}{\sinh(\sqrt{-r})} & \text{if }r<0. \end{cases} \tag*{(9)}

For the resonance case, r=λ1=π2r=\lambda_1=\pi^2, the stationary problem is solvable if and only if μ~0=μ~1=0\widetilde{\mu}_0=\widetilde{\mu}_1=0, and the solution is ρ∞(u)=Csin⁡(πu)\rho_\infty(u)=C\sin(\pi u), where C=const>0C=\mathrm{const}>0.

If r>π2r>\pi^2, then gtg_t diverge as t→∞t\to\infty, and if r≤π2r\leq\pi^2, then gtg_t converge to the limit metric g∞=du2+ρ∞2(u) dv2g_\infty=du^2+\rho_\infty^2(u)\,dv^2 with K(g∞)=rK(g_\infty)=r, see Theorem 2.1.

With known solution ρ(t,u)\rho(t,u) (t≥0t\geq0) of (5) and ∣ρu′∣≤1|\rho'_u|\leq1, we get h=h(t,0)+∫0u1−(ρu′)2 duh=h(t,0)+\int_0^u\sqrt{1-(\rho'_u)^2}\,du, that correspond to rotational surfaces in R3\mathbb{R}^3 of positive constant Gaussian curvature. Assume for simplicity that μj(t)≡μ~j\mu_j(t)\equiv\widetilde{\mu}_j. Then

μ~j=0,∫0∞∣μj(τ)−μ~j∣ dτ<∞,∫0∞∣μj′(τ)∣ dτ<∞,j=0,1.\widetilde{\mu}_j=0,\qquad\int_0^\infty|\mu_j(\tau)-\widetilde{\mu}_j|\,d\tau<\infty,\qquad\int_0^\infty|\mu'_j(\tau)|\,d\tau<\infty,\qquad j=0,1.

For r=π2r=\pi^2, we get ρ∞=Csin⁡(πu)\rho_\infty=C\sin(\pi u), where C=v10=∫01v0(s) dsC=v_1^0=\int_0^1v_0(s)\,ds, and then

h∞=1πEllipticE⁡(cos⁡(πu),Cπ).h_\infty=\frac{1}{\pi}\operatorname{EllipticE}(\cos(\pi u),C\pi).

Here,

EllipticE⁡(z,k)=∫0z1−k2s21−s2 ds\operatorname{EllipticE}(z,k)=\int_0^z\sqrt{\frac{1-k^2s^2}{1-s^2}}\,ds

is the incomplete elliptic integral. To provide a numerical example for r=π2r=\pi^2, let u∈[0,π]u\in[0,\pi], C=1C=1 and μj=0\mu_j=0. Then, the limit profile curve is a semi-circle [sin⁡u,cos⁡u][\sin u,\cos u] for 0≤u≤π0\leq u\leq\pi, and the limit rotational surface in R3\mathbb{R}^3 is a round sphere.

Next, we prove the extrinsic realization theorem: the rotational hypersurfaces built from the normalized solution realize the PRF extrinsically for all t≥0t\geq0, and converge, as t→∞t\to\infty, to the rotational hypersurface of Kmix=rK_{\mathrm{mix}}=r when r≤λ1r\leq\lambda_1.

The key analytic fact is that the total slope δ:=sup⁡(t,u)∈[0,1]×[0,∞)∣ρu′(t,u)∣\delta:=\sup_{(t,u)\in[0,1]\times[0,\infty)}|\rho'_u(t,u)| is finite whenever r≤λ1r\leq\lambda_1, and one can find an explicit bound of δ\delta.

Definition 3.3. Let r≤λ1r\leq\lambda_1, and ρ0:[0,1]→R\rho_0:[0,1]\to\mathbb{R} be smooth and positive on (0,1)(0,1) with ρ0(0)=μ0≥0\rho_0(0)=\mu_0\geq0, ρ0(1)=μ1≥0\rho_0(1)=\mu_1\geq0 (and μ0=μ1=0\mu_0=\mu_1=0 if r=λ1r=\lambda_1), and δ<∞\delta<\infty be the total slope of the solution ρ\rho of (6). We assume δ>0\delta>0: the degenerate case δ=0\delta=0 occurs exactly when ρ≡μ0=μ1>0\rho\equiv\mu_0=\mu_1>0 is constant, and

then we get a trivial representation: every MtM_t below is a fixed round cylinder. The normalized PRF solution is ρ^:=ρ/δ\hat{\rho}:=\rho/\delta, which by linearity of (2.5) is the solution with data ρ0/δ\rho_0/\delta, μ0/δ\mu_0/\delta, μ1/δ\mu_1/\delta, and satisfies

sup⁡(t,u)∣ρ^u′(t,u)∣=1.(10)\sup_{(t,u)}\lvert\hat{\rho}'_u(t,u)\rvert=1. \tag*{(10)}

We represent the PRF by rotational hypersurfaces in Rn+1\mathbb{R}^{n+1}. Given the normalized PRF solution ρ^(t,u)\hat{\rho}(t,u), define the map ψt:[0,1]×Sn−1→Rn+1\psi_t:[0,1]\times S^{n-1}\to\mathbb{R}^{n+1} for t≥0t\geq0

ψt(u,ω)=(ρ^(t,u) ω, h(t,u)),h(t,u)=∫0u1−(ρ^s′(t,s))2 ds(11)\psi_t(u,\omega)=\left(\hat{\rho}(t,u)\,\omega,\ h(t,u)\right),\qquad h(t,u)=\int_0^u\sqrt{1-(\hat{\rho}'_s(t,s))^2}\,\mathrm{d}s \tag*{(11)}

where ω∈Sn−1(1)\omega\in S^{n-1}(1), the square root in the formula for h(t,u)h(t,u) is real due to (3.3), and the choice h(t,0)=0h(t,0)=0, hu′≥0h'_u\geq0 pins down the vertical translation and reflection (an extrinsic representation is defined up to rigid motions of Rn+1\mathbb{R}^{n+1}).

Lemma 3.4. For each t≥0t\geq0 the map (3.4) parameterizes an embedded rotational hypersurface St⊂Rn+1S_t\subset\mathbb{R}^{n+1} (immersed open cylinder (0,1)×Sn−1(0,1)\times S^{n-1}, closing up at a pole u∈{0,1}u\in\{0,1\}, where a boundary radius ρ^\hat{\rho} vanishes, the closing is smooth when ∣ρ^u′∣=1\lvert\hat{\rho}'_u\rvert=1 there, and is conical otherwise), with the following properties

i) The induced metric is gt=du2+ρ^2(t,u)gSn−1g_t=\mathrm{d}u^2+\hat{\rho}^2(t,u)g_{S^{n-1}} and satisfies the PRF (1.1)

ii) For each ω∈Sn−1\omega\in S^{n-1}, ψt(u,ω)\psi_t(u,\omega) in (3.4) is a unit-speed geodesic of StS_t of length 1. Moreover, ψt\psi_t is a C1C^1-regular immersion for every tt, smooth away from the closed set ∣ρ^u′∣=1\lvert\hat{\rho}'_u\rvert=1, the metric gtg_t is smooth in (t,u)(t,u) for t>0t>0 and continuous at t=0t=0.

Proof. This is standard.

Let StS_t (t≥0t\geq0) be rotational hypersurfaces with a common axis and profile curves γt\gamma_t parameterized by arclength on the interval [0,1][0,1]. Then StS_t converge to the limiting hypersurface with a profile curve γ∞\gamma_\infty, if and only if γt→γ∞\gamma_t\to\gamma_\infty uniformly on [0,1][0,1].

Theorem 3.5 (Extrinsic realization). Let r≤λ1r\leq\lambda_1, with μ0=μ1=0\mu_0=\mu_1=0 if r=λ1r=\lambda_1, ρ0\rho_0 be smooth and positive on (0,1)(0,1) with ρ0(0)=μ0≥0\rho_0(0)=\mu_0\geq0, ρ0(1)=μ1≥0\rho_0(1)=\mu_1\geq0, δ∈(0,∞)\delta\in(0,\infty) be the total slope. Suppose that St⊂Rn+1S_t\subset\mathbb{R}^{n+1} are rotational hypersurfaces (3.4) built from the normalized solution ρ^=ρ/δ\hat{\rho}=\rho/\delta. Then {St}t≥0\{S_t\}_{t\geq0} realize the PRF (1.1) extrinsically: the metric induced on StS_t is gt=du2+ρ^2(t,u)gSn−1g_t=\mathrm{d}u^2+\hat{\rho}^2(t,u)g_{S^{n-1}}, and

i) if r<λ1r<\lambda_1, the limiting, as t→∞t\to\infty, hypersurface has Kmix≡rK_{\mathrm{mix}}\equiv r and its profile is

γ∞(u)=(ρ∞(u,r,μ0/δ,μ1/δ), ∫0u(1−ρ∞′(s,r;μ0/δ,μ1/δ)2)1/2 ds)(12)\gamma_\infty(u)=\left(\rho_\infty(u,r,\mu_0/\delta,\mu_1/\delta),\ \int_0^u\left(1-\rho'_\infty(s,r;\mu_0/\delta,\mu_1/\delta)^2\right)^{1/2}\,\mathrm{d}s\right) \tag*{(12)}

ii) if r=λ1r=\lambda_1 (so μ0=μ1=0\mu_0=\mu_1=0), the limiting, as t→∞t\to\infty, hypersurface exists and is given by (3.1) with B=A1/δ∈(0,1/π]B=A_1/\delta\in(0,1/\pi], where A1=2∫01ρ0(u)sin⁡(πu) du>0A_1=2\int_0^1\rho_0(u)\sin(\pi u)\,\mathrm{d}u>0

iii) if r>λ1r>\lambda_1, then for generic data, δ=∞\delta=\infty, the normalization is impossible, and no limiting hypersurface exists. Indeed, the intrinsic limit fails by Proposition 2.4iv) unless all coefficients A~i\tilde{A}_i with λi≤r\lambda_i\leq r vanish.

Proof. The induced metrics are gt=du2+ρ^2gSn−1g_t=\mathrm{d}u^2+\hat{\rho}^2g_{S^{n-1}}, they solve the PRF (1.1), and the meridians are geodesics of length 1 for every tt, see Lemma 3.4i),ii) and Example 2.3ii). Note that ρ^\hat{\rho} is the solution of (2.5) with data (ρ0/δ,μ0/δ,μ1/δ)(\rho_0/\delta,\mu_0/\delta,\mu_1/\delta) by linearity, unique and smooth for t>0t>0 by [6 Section 5.4.4] and (3.3) makes (3.4) well defined for all tt. Thus, ψt\psi_t are the required immersions, smooth for $t>0.

i) By Proposition 2.4 ii) with the normalized data, ρ^(t,⋅)→ρ∞(⋅,r,μ0/δ,μ1/δ)\hat{\rho}(t,\cdot)\to\rho_{\infty}(\cdot,r,\mu_{0}/\delta,\mu_{1}/\delta) in C1([0,1])C^{1}([0,1]), as t→∞t\to\infty, using ρ∞(⋅,μ0/δ,μ1/δ)=1δρ∞(⋅,μ0,μ1)\rho_{\infty}(\cdot,\mu_{0}/\delta,\mu_{1}/\delta)=\frac{1}{\delta}\rho_{\infty}(\cdot,\mu_{0},\mu_{1}). Passing to the limit in ∣ρ^u′∣≤1|\hat{\rho}_{u}^{\prime}|\leq1 gives the derivative of ρ∞\rho_{\infty}, see (13) and also Corollary 2.2. Then

∣h(t,u)−h∞(u)∣≤∫01∣1−(ρ^s′)2(t,s)−1−(ρ∞,u′)2(s)∣ ds→0|h(t,u)-h_{\infty}(u)|\leq\int_{0}^{1}\left|\sqrt{1-(\hat{\rho}_{s}^{\prime})^{2}(t,s)}-\sqrt{1-(\rho_{\infty,u}^{\prime})^{2}(s)}\right|\,\mathrm{d}s\to0

uniformly in uu, since ρ^u′→ρ∞′\hat{\rho}_{u}^{\prime}\to\rho_{\infty}^{\prime} uniformly with values in [−1,1][-1,1]. Hence, the profile curves (ρ^,h)(t,⋅)(\hat{\rho},h)(t,\cdot) converge uniformly to (3.5), which is the limiting hypersurface. A limiting metric is g∞=du2+ρ∞2(u)gSn−1g_{\infty}=\mathrm{d}u^{2}+\rho_{\infty}^{2}(u)g_{\mathbb{S}^{n-1}}, and any plane containing ∂u\partial_{u} has sectional curvature Kmix=−ρ∞′′/ρ∞=rK_{\mathrm{mix}}=-\rho_{\infty}^{\prime\prime}/\rho_{\infty}=r by (2.5), for n=2n=2 this is the Gaussian curvature.

ii) Proposition 2.4 iii) with μ0=μ1=0\mu_{0}=\mu_{1}=0 and r=λ1r=\lambda_{1} gives ρ^(t,⋅)→(A1/δ)sin⁡(πu)\hat{\rho}(t,\cdot)\to(A_{1}/\delta)\sin(\pi u) in C1C^{1}, and A1>0A_{1}>0 since ρ0>0\rho_{0}>0 on (0,1)(0,1) makes the integrand of A1A_{1} positive there. Similar argument as in (1) yields uniform convergence of the generating curves to the curve (3.1) with B=A1/δ>0B=A_{1}/\delta>0, and passing to the limit in ∣ρ^u′∣≤1|\hat{\rho}_{u}^{\prime}|\leq1 gives Bπ≤1B\pi\leq1, i.e., B≤1/πB\leq1/\pi, so the limit lies in the extremal family of Proposition 3.1, for B=1/πB=1/\pi the limiting hypersurface closes smoothly at both poles (∣ρ^u′∣=1|\hat{\rho}_{u}^{\prime}|=1 there), and for B<1/πB<1/\pi it closes with two conical points. By the above, δ≤1\delta\leq1, so (3.4) applies to ρ\rho (in place of ρ^\hat{\rho}), and the proof is similar as for δ=1\delta=1: we have

max⁡u∈[0,1]∣ρ∞′∣=max⁡u∈[0,1]∣(μ1/δ)Cn⁡r(u)−(μ0/δ)Cn⁡r(1−u)Sn⁡r(1)∣≤1(13)\max_{u\in[0,1]}|\rho_{\infty}^{\prime}|=\max_{u\in[0,1]}\left|\frac{(\mu_{1}/\delta)\operatorname{Cn}_{r}(u)-(\mu_{0}/\delta)\operatorname{Cn}_{r}(1-u)}{\operatorname{Sn}_{r}(1)}\right|\leq1 \tag*{(13)}

for the limiting profile curve, (3.5) with (μ0,μ1)(\mu_{0},\mu_{1}), and the slope condition.

Parallel Curved Surfaces

A parallel curved (PC) surface is a surface M2⊂R3M^{2}\subset\mathbb{R}^{3} and a fixed plane EE such that at each point of M2M^{2} at least one principal direction is parallel to EE, see [1]. PC surfaces carry twisted product metrics, see (2.6), with (positive by the focal condition)

ρ0(u,v)=1−ρ0(u)k(v),respectively,ρ0(u,v)=1−a0(u,v)k(v).(14)\rho_{0}(u,v)=1-\rho_{0}(u)k(v),\qquad\text{respectively,}\qquad\rho_{0}(u,v)=1-a_{0}(u,v)k(v). \tag*{(14)}

Let γ:I→R2⊂R3\gamma:I\to\mathbb{R}^{2}\subset\mathbb{R}^{3} be a simple closed convex curve, parameterized by arclength, with curvature k=kγ(v)>0k=k_{\gamma}(v)>0 and inward unit normal N(v)N(v), rfoc:=min⁡v1/k(v)>0r_{\mathrm{foc}}:=\min_{v}1/k(v)>0 its focal radius, and e3e_{3} the vertical unit vector in R3\mathbb{R}^{3}. Let ρ,h:[0,1]→R\rho,h:[0,1]\to\mathbb{R} be smooth arclength parameterization and ∣ρ(u)∣<rfoc|\rho(u)|<r_{\mathrm{foc}} for all uu. Then r=γ(v)+ρ(u)N(v)+h(u)e3\mathbf{r}=\gamma(v)+\rho(u)N(v)+h(u)e_{3} parameterizes a PC surface, called a canonical PC surface, and

i) the induced metric ds2=du2+(1−ρ(u)k(v))2dv2\mathrm{d}s^{2}=\mathrm{d}u^{2}+(1-\rho(u)k(v))^{2}\mathrm{d}v^{2} belongs to twisted products

ii) uu-curves are unit-speed geodesics of length 1

iii) the Gaussian curvature is K=ρu,u′′k/(1−ρk)K=\rho_{u,u}^{\prime\prime}k/(1-\rho k)

iv) with unit normal ν=−hu′N+ρu′e3\nu=-h_{u}^{\prime}N+\rho_{u}^{\prime}e_{3}, the second fundamental form is diagonal with e=hu,u′′/ρu′−hu′′ρu,u′/ρu′e=h_{u,u}^{\prime\prime}/\rho_{u}^{\prime}-h_{u}^{\prime\prime}\rho_{u,u}^{\prime}/\rho_{u}^{\prime}, g=−hu′(u)k(v)(1−ρ(u)k(v))g=-h_{u}^{\prime}(u)k(v)(1-\rho(u)k(v)), so the coordinate directions are principal, r,v′r_{,v}^{\prime} being parallel to the xyxy-plane.

Proposition 3.6 (The PRF on PC surfaces). Let g0g_{0} be induced metric of a canonical PC surface, with ρ0\rho_{0} as in (3.7) and fixed boundary values μ0(v)=ρ0(0,v)\mu_{0}(v)=\rho_{0}(0,v) and μ1(v)=ρ0(1,v)\mu_{1}(v)=\rho_{0}(1,v). Then the PRF has a unique global solution, the foliation by uu-curves remains geodesic with leaves of length 1, and for r<λ1r<\lambda_{1} the metrics converge in every CkC^{k}-norm to metric g∞=du2+ρ∞2(u,r,μ0(v),μ1(v))2dv2g_{\infty}=\mathrm{d}u^{2}+\rho_{\infty}^{2}(u,r,\mu_{0}(v),\mu_{1}(v))^{2}\mathrm{d}v^{2} of constant Gaussian curvature rr, the cases r≥λ1r\geq\lambda_{1} are as in Proposition 2.4.

Proof. This is standard.

Theorem 3.7 (Realization of PRF by PC surfaces). Consider canonical PC surfaces over the fixed base γ\gamma with evolving profile

r(u,v,t)=γ(v)+ρ(t,u)N(v)+h(t,u)e3,h(t,u)=∫0u1−ρs′(t,s)2 ds.(15)\mathbf{r}(u,v,t)=\gamma(v)+\rho(t,u)N(v)+h(t,u)e_{3},\qquad h(t,u)=\int_{0}^{u}\sqrt{1-\rho_{s}^{\prime}(t,s)^{2}}\,\mathrm{d}s. \tag*{(15)}
  1. (Rigidity). If the induced metrics of a family (3.8) with ρt′≢0\rho_{t}^{\prime}\not\equiv0 satisfy the PRF, then either k=constk=\mathrm{const}, i.e., γ\gamma is a circle and the surfaces are rotational, or r=0r=0.

  1. (Existence for r=0r=0). Let ρ(t,u)\rho(t,u) solve the heat equation ρt′=ρu,u′′\rho_{t}^{\prime}=\rho_{u,u}^{\prime\prime} with data ρ(⋅,0)=ϕ\rho(\cdot,0)=\phi, ρ(t,0)=α1\rho(t,0)=\alpha_{1}, ρ(t,1)=α2\rho(t,1)=\alpha_{2}, where ϕ\phi is smooth, ∣α1∣,∣α2∣,sup⁡∣ϕ∣<rfoc|\alpha_{1}|,|\alpha_{2}|,\sup|\phi|<r_{\mathrm{foc}}, and sup⁡t,u∣ρu′∣≤1\sup_{t,u}|\rho_{u}^{\prime}|\le1 (achievable by the normalization of Definition 3.3, which scales α2\alpha_{2} toward 0 and so preserves the focal condition). Then

2a) the focal condition is preserved: by the maximum principle, for all tt we have

sup⁡u∣ρ(t,u)∣≤max⁡(sup⁡∣ϕ∣,∣α1∣,∣α2∣)<rfoc\sup_{u}|\rho(t,u)|\le\max\left(\sup|\phi|,|\alpha_{1}|,|\alpha_{2}|\right)<r_{\mathrm{foc}}

2b) a family (3.8) consists of canonical PC surfaces realizing the PRF with r=0r=0.

2c) the limit is a flat canonical PC surface with profile ρ∞(u)=α1+(α2−α1)u\rho_{\infty}(u)=\alpha_{1}+(\alpha_{2}-\alpha_{1})u.

Proof. 1) The induced metric of (3.8) is du2+b2dv2\mathrm{d}u^{2}+b^{2}\mathrm{d}v^{2} with b=1−ρ(t,u)k(v)b=1-\rho(t,u)k(v). The PRF is equivalent to ρt′=ρu,u′′+rb\rho_{t}^{\prime}=\rho_{u,u}^{\prime\prime}+rb, thus, (using k(v)>0k(v)>0)

ρt′−ρu,u′′−rb=−r/k(v).(16)\rho_{t}^{\prime}-\rho_{u,u}^{\prime\prime}-rb=-r/k(v). \tag*{(16)}

The LHS is independent of vv, hence so is the RHS, which forces r=0r=0 or k=constk=\mathrm{const}.

  1. For r=0r=0, (3.9) reduces to ρt′=ρu,u′′\rho_{t}^{\prime}=\rho_{u,u}^{\prime\prime}, so the induced metrics satisfy the PRF equation with r=0r=0, this proves the realization claim in 2b) once the surfaces are shown to be well defined. 2a) is the maximum principle for the heat equation: both max and min of ρ\rho over [0,1]×[0,T][0,1]\times[0,T] are attained on the parabolic boundary, where ∣ρ∣≤max⁡(sup⁡∣ϕ∣,∣α1∣,∣α2∣)<rfoc|\rho|\le\max(\sup|\phi|,|\alpha_{1}|,|\alpha_{2}|)<r_{\mathrm{foc}}, note that for r=0r=0 there is no zeroth-order term and the principle applies directly. Together with sup⁡∣ρu′∣≤1\sup|\rho_{u}^{\prime}|\le1, the map (3.8) is well defined, and 1−bk≥1−sup⁡∣b∣max⁡k>01-bk\ge1-\sup|b|\max k>0, so it is an immersion with the stated first fundamental form. For 2c), Theorem 2.1 with r=0<λ1r=0<\lambda_{1} gives ρ(t,⋅)→ρ∞(⋅,0,α1,α2)\rho(t,\cdot)\to\rho_{\infty}(\cdot,0,\alpha_{1},\alpha_{2}) in C1C^{1}, and the height functions converge uniformly, the limit profile is linear, so the limiting surface has K=k(ρ∞)u,u′′/(1−kρ∞)≡0K=k(\rho_{\infty})_{u,u}^{\prime\prime}/(1-k\rho_{\infty})\equiv0.

The PRF on Weak Almost Contact Metric Manifolds

Contact Riemannian geometry is a significant field in both mathematics and physics. In 2022, Rovenski and Wolak introduced metric structures on a smooth manifold that generalize the almost contact, KK-contact, Sasakian, cosymplectic, etc. metric structures. Such so-called “weak” structures allow us to take a fresh look at the theory of classical structures and find new applications, see [7, 8, 5].

A Riemannian manifold (M2n,g)(M^{2n},g) equipped with a non-degenerate skew-symmetric (1,1)(1,1)-tensor ff is called a weak Hermitian manifold. This manifold is called weak Kählerian if ∇f=0\nabla f=0, where ∇\nabla is the Levi-Civita connection.

A weak almost contact metric structure on a manifold M2n+1M^{2n+1} is a set (f,Q,ξ,η,g)(f,Q,\xi,\eta,g), where gg is a Riemannian metric on MM, ff is a skew-symmetric (1,1)(1,1)-tensor of rank 2n2n, ξ\xi is a unit (Reeb) vector field, η\eta is a 1-form such that η(ξ)=1\eta(\xi)=1, satisfying

g(fX,fY)=g(X,QY)−η(X)η(Y)g(fX,fY)=g(X,QY)-\eta(X)\eta(Y)

and Q:=−f2+η⊗ξQ := -f^{2} + \eta\otimes\xi is a self-adjoint (1,1)(1,1)-tensor field (for almost contact metric manifolds, Q=IQ = I is true). The self-adjoint (1,1)(1,1)-tensor Q~:=Q−I\widetilde{Q} := Q - I measures the difference between weak and classical metric structures. A weak contact metric structure is a weak almost contact metric structure satisfying the condition dη=Fd\eta= F, where F(X,Y)=g(X,fY)F(X,Y) = g(X,fY), in this case, the Reeb vector field ξ\xi is geodesic.

Weak KK-Contact Manifolds

A weak contact metric structure is called a weak KK-contact structure if ξ\xi is Killing (an infinitesimal isometry), i.e., the Lie derivative of metric vanishes: Lξg=0\mathcal{L}_{\xi}g = 0. We generalize the fact that a Riemannian manifold with a unit Killing vector field ξ\xi such that RξX=XR_{\xi}X = X for X∈ker⁡ηX \in\ker\eta is a KK-contact manifold.

Theorem 4.1 ([8]). A Riemannian manifold (M2n+1,g)(M^{2n+1},g) admitting a unit Killing vector field ξ\xi such that the (1,1)(1,1)-tensor ∇ξ\nabla\xi has constant rank 2n2n and ξ\xi-sectional curvature is positive is a weak KK-contact manifold M(f,Q,ξ,η,g)M(f,Q,\xi,\eta,g) with η=g(⋅,ξ)\eta= g(\cdot,\xi) and the following structural tensors: f=−∇ξf = -\nabla\xi and QX=RξXQX = R_{\xi}X for X∈ker⁡ηX \in\ker\eta.

The PRF prescribes KmixK_{\mathrm{mix}} of a foliation and has recently been used to obtain dynamical characterizations of KK-contact structures, see [8]. For KK-contact structures, Ric⁡g⊥=1\operatorname{Ric}^{\perp}_{g} = 1 and the metric is a fixed point of the normalized PRF with r=1r = 1.

Theorem 4.2 ([8]). Let M2n+1(f0,Q0,ξ,η,g0)M^{2n+1}(f_{0},Q_{0},\xi,\eta,g_{0}) be a weak KK-contact manifold. Then, there exist D\mathcal{D}-adapted weak KK-contact structures (ft,Qt,ξ,η,gt)(f_{t},Q_{t},\xi,\eta,g_{t}) on MM such that gtg_{t} – a solution of the PRF – converges exponentially fast, as t→−∞t \to-\infty, to a limit metric that provides a KK-contact structure.

Example 4.3. Let M2n+1(f0,Q0,ξ,η,g0)M^{2n+1}(f_{0},Q_{0},\xi,\eta,g_{0}) be a weak KK-contact manifold with Q∣D=λIDQ|_{\mathcal{D}} = \lambda I_{\mathcal{D}} for some differentiable function 0<λ≢10 < \lambda\not\equiv1 on MM. From Theorem 4.2, applying the theorem of uniqueness of a solution to ODE

∂tRξ=4Rξ(Rξ−ID)(17)\partial_{t}R_{\xi} = 4R_{\xi}\left(R_{\xi} - I_{\mathcal{D}}\right) \tag*{(17)}

we have a D\mathcal{D}-conformal solution Rξ(t)=μ(t)IDR_{\xi}(t) = \mu(t) I_{\mathcal{D}}. Since ∂tID=0\partial_{t}I_{\mathcal{D}} = 0, we have also that ODE dμ/dt=4μ(μ−1)d\mu/dt = 4\mu(\mu- 1) with μ(0)=λ>0\mu(0) = \lambda> 0. Thus,

μ(t)=11−Cexp⁡(4t)\mu(t) = \frac{1}{1-C\exp(4t)}

(t∈R)(t \in\mathbb{R}), where C=1−λ−1≠0C = 1 - \lambda^{-1} \ne0. Since our PRF reduces to ODE, by the uniqueness of a solution, we have ∂tgt⊥=0\partial_{t}g_{t}^{\perp} = 0 and gt=g0⊥+ρ(t)g0⊤g_{t} = g_{0}^{\perp} + \rho(t)g_{0}^{\top}, where ρ(0)=1\rho(0) = 1 and

∂tgt⊤=2(1−μ(t))gt⊤⟹ddtlog⁡ρ(t)=−2Cexp⁡(4t)1−Cexp⁡(4t)⟹ρ(t)=(1−Cexp⁡(4t)1−C)1/2.\partial_{t}g_{t}^{\top} = 2(1-\mu(t))g_{t}^{\top} \Longrightarrow\frac{d}{dt}\log\rho(t) = -\frac{2C\exp(4t)}{1-C\exp(4t)} \Longrightarrow\rho(t) = \left(\frac{1-C\exp(4t)}{1-C}\right)^{1/2}.

Therefore, ft=f0/ρ(t)f_{t} = f_{0}/\rho(t) and Qt∣D=11−Cexp⁡(4t)IDQ_{t}|_{\mathcal{D}} = \frac{1}{1-C\exp(4t)}I_{\mathcal{D}}. Note that lim⁡t→−∞Qt=ID\lim_{t\to-\infty}Q_{t} = I_{\mathcal{D}}. By Theorem 4.2, (ft,Qt,ξ,η,gt)(f_{t},Q_{t},\xi,\eta,g_{t}) (t∈R)(t \in\mathbb{R}) is a smooth family of weak KK-contact structures on MM – a solution of the PRF – that converges exponentially fast, as t→−∞t \to-\infty, to a KK-contact structure (f0/λ,ξ,η,g0⊥+λg0⊤)(f_{0}/\sqrt{\lambda},\xi,\eta,g_{0}^{\perp}+\sqrt{\lambda}g_{0}^{\top}).

We obtain an extrinsic realization of the PRF on an ellipsoid of R4\mathbb{R}^{4} with the weak KK-contact structure, its limit is a unit three-sphere with the KK-contact structure.

Example 4.4. Let M⊂R2n+2M \subset\mathbb{R}^{2n+2} with n≥1n \ge1 be an ellipsoid with induced metric gg

M={u=(u1,…,u2n+2)∈R2n+2;∑i=1n+1ui2+a∑i=n+22n+1ui2=1}M = \left\{u=(u_{1},\ldots,u_{2n+2})\in\mathbb{R}^{2n+2};\quad\sum_{i=1}^{n+1}u_{i}^{2}+a\sum_{i=n+2}^{2n+1}u_{i}^{2}=1\right\}

where 0<a=const≠10<a=\mathrm{const}\ne1. The sectional curvature of (M,g)(M,g) is positive, and

ξ(u)=(−u2,u1,…,−un+1,un,−aun+3,aun+2,…,−au2n+2,au2n+1)\xi(u)=(-u_{2},u_{1},\ldots,-u_{n+1},u_{n},-\sqrt{a}u_{n+3},\sqrt{a}u_{n+2},\ldots,-\sqrt{a}u_{2n+2},\sqrt{a}u_{2n+1})

is a Killing vector field on R2n+2\mathbb{R}^{2n+2}. (Indeed, there is a constant block-diagonal with 2×22\times2 skew-symmetric blocks matrix JJ such that ξ(u)=Ju\xi(u)=Ju. The integral curves of ξ(u)\xi(u) solve (d/dt)u=Ju(\mathrm{d}/\mathrm{d}t)u=Ju, hence, u(t)=exp⁡(tJ)u(0)u(t)=\exp(tJ)u(0), where exp⁡(tJ)\exp(tJ) is an orthogonal matrix. Thus, the flow Ψt(u)=exp⁡(tJ)u\Psi_t(u)=\exp(tJ)u consists of rotations, so each Ψt\Psi_t is an isometry). Note that ξ\xi is tangent to MM, and ξ\xi is a unit Killing vector field on (M,g)(M,g), which defines a weak KK-contact structure, see Theorem 4.1.

The following result (for n=1n=1) can easily be extended for n>1n>1.

Theorem 4.5. Let (f,Q,ξ,η,g)(f,Q,\xi,\eta,g) be a weak KK-contact structure on an ellipsoid S3={u12+u22+a0(u32+u42)=1}⊂R4S^{3}=\{u_{1}^{2}+u_{2}^{2}+a_{0}(u_{3}^{2}+u_{4}^{2})=1\}\subset\mathbb{R}^{4} with a unit Killing vector field ξ=(−u2,u1,−a0u4,a0u3)\xi=(-u_{2},u_{1},-\sqrt{a_{0}}u_{4},\sqrt{a_{0}}u_{3}) and Q∣ker⁡η=λIker⁡ηQ|_{\ker\eta}=\lambda I_{\ker\eta} for a smooth function λ>0\lambda>0 on S3S^{3}. Then we get an extrinsic realization of the PRF on ellipsoids St3={u12+u22+at(u32+u42)=1}⊂R4S_{t}^{3}=\{u_{1}^{2}+u_{2}^{2}+a_{t}(u_{3}^{2}+u_{4}^{2})=1\}\subset\mathbb{R}^{4} with at=1−(1−a0)exp⁡(4t)a_{t}=1-(1-a_{0})\exp(4t) and t≥0t\geq0, and the limit, as t→−∞t\to-\infty, is a unit sphere S3⊂R4S^{3}\subset\mathbb{R}^{4} equipped with a KK-contact structure (f^,ξ^,η^,g^)(\hat{f},\hat{\xi},\hat{\eta},\hat{g}) and ξ^=(−u2,u1,−u4,u3)\hat{\xi}=(-u_{2},u_{1},-u_{4},u_{3}).

Proof. Using Hopf parametrization u1=cos⁡θcos⁡ϕu_{1}=\cos\theta\cos\phi, u2=cos⁡θsin⁡ϕu_{2}=\cos\theta\sin\phi, u3=a0−1/2sin⁡θcos⁡ψu_{3}=a_{0}^{-1/2}\sin\theta\cos\psi, u4=a0−1/2sin⁡θsin⁡ψ)u_{4}=a_{0}^{-1/2}\sin\theta\sin\psi) of this ellipsoid, we find its metric tensor

ds2=(sin⁡2θ+a0−1cos⁡2θ)dθ2+cos⁡2θ dϕ2+a0−1sin⁡2θ dψ2.\mathrm{d}s^{2}=(\sin^{2}\theta+a_{0}^{-1}\cos^{2}\theta)\mathrm{d}\theta^{2}+\cos^{2}\theta\,\mathrm{d}\phi^{2}+a_{0}^{-1}\sin^{2}\theta\,\mathrm{d}\psi^{2}.

For any tangent vector XX such that g(X,ξ)=0g(X,\xi)=0, the Gauss equation, and (4.1), give RξX=g(Aξ,ξ)AX−g(AX,ξ)AξR_{\xi}X=g(A\xi,\xi)AX-g(AX,\xi)A\xi, where AA is the Weingarten operator. In fact, for this ellipsoid of revolution, XX and ξ\xi are principal directions. By the above, RξX=∥u∥−2X=(1+(1−a0)(u32+u42))−1XR_{\xi}X=\lVert u\rVert^{-2}X=(1+(1-a_{0})(u_{3}^{2}+u_{4}^{2}))^{-1}X, where the factor varies with both a0a_{0} and the point (ui)(u_{i}) on the ellipsoid. Using QX=RξXQX=R_{\xi}X (Theorem 4.1) where X∈ker⁡ηX\in\ker\eta and Example 4.3, we get 1+(1−at)(u32+u42)=1−(1−λ−1)exp⁡(4t)1+(1-a_{t})(u_{3}^{2}+u_{4}^{2})=1-(1-\lambda^{-1})\exp(4t). For at=1−(1−a0)exp⁡(4t)a_{t}=1-(1-a_{0})\exp(4t) and t∈Rt\in\mathbb{R}, we obtain λ=(u32+u42+1)−1\lambda=(u_{3}^{2}+u_{4}^{2}+1)^{-1} and QX=(1+exp⁡(4t)(u32+u42))−1XQX=(1+\exp(4t)(u_{3}^{2}+u_{4}^{2}))^{-1}X (X∈ker⁡ηX\in\ker\eta). We get an extrinsic realization of the PRF on the weak KK-contact ellipsoid. By Theorem 4.2, the limit, as t→−∞t\to-\infty, is a unit sphere S3S^{3} with a KK-contact structure.

Weak β\beta-Kenmotsu Manifolds

In 1972, Kenmotsu introduced and studied a special class of almost contact metric manifolds, which is characterized in terms of warped products of the real line and a Kähler manifold with the warping function ρ=Cexp⁡(u)\rho=C\exp(u). In 1991, Olszak characterized in terms of warped products (with ρ=Cexp⁡(βu)\rho=C\exp(\beta u)) a class of almost contact metric manifolds, known as β\beta-Kenmotsu manifolds for β=const>0\beta=\mathrm{const}>0, given by

(∇Xf)Y=β(g(fX,Y)ξ−η(Y)fX).(18)(\nabla_{X}f)Y=\beta\bigl(g(fX,Y)\xi-\eta(Y)fX\bigr). \tag*{(18)}

Weak β\beta-Kenmotsu manifolds were introduced, see [5], as weak almost contact metric manifolds satisfying (4.2), where β≠0\beta\ne0 is a smooth function on MM, and if n=1n=1, then dβ∧η=0\mathrm{d}\beta\wedge\eta=0 is required. If β≡1\beta\equiv1, then M2n+1(f,Q,ξ,η,g)M^{2n+1}(f,Q,\xi,\eta,g) is called a weak Kenmotsu manifold, and for Q=IQ=I we get a Kenmotsu manifold. By (4.2), ∇Xξ=β(X−η(X)ξ)\nabla_{X}\xi=\beta\bigl(X-\eta(X)\xi\bigr) thus, for weak β\beta-Kenmotsu manifolds, ξ\xi is geodesic (∇ξξ=0\nabla_{\xi}\xi=0).

Theorem 4.6 (see [5]). *Every point of a weak β\beta-Kenmotsu manifold M(f,Q,ξ,η,g)M(f,Q,\xi,\eta,g) has a neighbourhood UU that is isometric to a twisted product (−ε,ε)×ρMˉ(-\varepsilon,\varepsilon)\times_{\rho}\bar{M} (a warped product if ∇β\nabla\beta is parallel to

ξ\xi with the leaves tangent to ξ\xi, where M‾\overline{M} has a weak Hermitian (weak Kähler, respectively) structure and the following is true

ρu′ρ=β.(19)\frac{\rho'_u}{\rho}=\beta. \tag*{(19)}

For a simply connected MM and geodesically complete ξ\xi, the isometry is global.

Consider a (weak) β\beta-Kenmotsu structure on a warped product [0,1]×ρS2(1)[0,1]\times_{\rho} S^2(1) with the (weak) Kähler structure on S2(1)S^2(1), and realize it on a rotational hypersurface in R4\mathbb{R}^4 with profile curve (ρ(u),h(u))(\rho(u),h(u)) (u∈[0,1])(u\in[0,1]), where (ρu′)2+(hu′)2=1(\rho'_u)^2+(h'_u)^2=1. By Theorem 4.6, ∇β\nabla\beta is parallel to the Reeb field ξ\xi, i.e., β\beta is a function of uu. By (19), we have

ρ(u)=ρ0exp⁡(∫0uβ(s) ds),h(u)=∫0u1−β2(s)ρ02exp⁡(2∫0sβ(τ) dτ) ds.\begin{aligned} \rho(u)&=\rho_0\exp\left(\int_0^u\beta(s)\,\mathrm{d}s\right),\\ h(u)&=\int_0^u\sqrt{1-\beta^2(s)\rho_0^2\exp\left(2\int_0^s\beta(\tau)\,\mathrm{d}\tau\right)}\,\mathrm{d}s. \end{aligned}

For β≡1\beta\equiv1, the profile curve is (ρ(u),h(u))=(ρ0exp⁡(u),∫0u1−ρ02exp⁡(2s) ds)(\rho(u),h(u))=\left(\rho_0\exp(u),\int_0^u\sqrt{1-\rho_0^2\exp(2s)}\,\mathrm{d}s\right).

Let the family of warped products Mt3=[0,1]×ρ(t,u)S2M^3_t=[0,1]\times_{\rho(t,u)}S^2, t≥0t\geq0, with ρ(0,u)\rho(0,u) given in i), see Theorem 2.1, carries a (weak) β\beta-Kenmotsu structure with ξ=∂u\xi=\partial_u and β\beta depending on (t,u)(t,u), and g(t)g(t) modified by the PRF (1). Let us realize Mt3M^3_t as hypersurfaces of revolution St3S^3_t in R4\mathbb{R}^4 with profile curves (ρ(t,u),h(t,u))(\rho(t,u),h(t,u)), where the height function h(t,u)h(t,u) is determined (up to an additive constant in uu) by the arc-length condition, i.e., hu′(t,u)=1−(ρu′)2(t,u)h'_u(t,u)=\sqrt{1-(\rho'_u)^2(t,u)}. The rotational hypersurface closes smoothly at both ends u=0u=0 and u=1u=1 (i.e., is diffeomorphic to S3S^3) if and only if

ρ(t,0)=0,ρ(t,1)=0,ρu′(t,0)=1,ρu′(t,1)=−1,hu′(t,0)=0,hu′(t,1)=0.\rho(t,0)=0,\quad\rho(t,1)=0,\quad\rho'_u(t,0)=1,\quad\rho'_u(t,1)=-1,\quad h'_u(t,0)=0,\quad h'_u(t,1)=0.

By (19), β=ρu′/ρ\beta=\rho'_u/\rho is a function of (t,u)(t,u) that defines a β\beta-Kenmotsu structure on Mt3M^3_t.

The PRF modifies g(t)g(t), as t→∞t\to\infty, to a limit metric g∞=du2+ρ∞2(u)gS2g_\infty=\mathrm{d}u^2+\rho_\infty^2(u)g_{S^2} with Kmix≡rK_{\mathrm{mix}}\equiv r. The function ρ∞(u)\rho_\infty(u) satisfies the ODE and is given by (9). The profile curve of the limiting rotational hypersurface S∞3S^3_\infty in R4\mathbb{R}^4 is given by (ρ∞(u),h∞(u))(\rho_\infty(u),h_\infty(u)), where h∞(u)=∫0u1−(ρ∞′)2(s) dsh_\infty(u)=\int_0^u\sqrt{1-(\rho'_\infty)^2(s)}\,\mathrm{d}s. The limiting (weak) β\beta-Kenmotsu structure has β∞=(ρ∞)u/ρ∞\beta_\infty=(\rho_\infty)_u/\rho_\infty. For r=π2r=\pi^2, we get ρ∞=Csin⁡(πu)\rho_\infty=C\sin(\pi u), see Example 3.2, then using (19) we find β∞(u)=πcot⁡(πu)\beta_\infty(u)=\pi\cot(\pi u).

Conclusion

We studied extrinsic representations of the partial Ricci flow (PRF) on manifolds with geodesic foliations. The constructions encompass compact rotational hypersurfaces in Euclidean space, including weak β\beta-Kenmotsu examples, parallel curved surfaces, and ellipsoidal models carrying weak KK-contact structures.

The study of extrinsic representations of the PRF presented in this paper can be extended for totally geodesic foliations, e.g., twisted products and weak metric ff-contact manifolds (as well as to spherical foliations such as in [3]) with applications to the calculus of moving surfaces and mathematical physics.

Acknowledgments

The authors thank Professor D. Blanc for helpful comments on the paper.

References

  1. [1]Ando N., Parallel Curved Surfaces, Tsukuba J. Math. 28 (2004) 223–243.DOI
  2. [2]Coll V., Dodd J. and Johnson D., Ricci Flow on Surfaces of Revolution: An Extrinsic View. Geom. Dedicata 207 (2020) 81–94.DOI
  3. [3]Güler E. and Toda M., A Study of Special Weingarten Hypersurfaces in R^4 with Spherical Foliations, J. Geom. Symmetry Phys. 75 (2026) 39–49.DOI
  4. [4]Rovenski V., The Partial Ricci Flow for Foliations, In: Rovenski, V. et al. (Eds) Geometry and its Applications, Springer Proc. Math. & Statistics 72, Springer, Cham 2014, pp 125–155.DOI
  5. [5]Rovenski V. and Patra D., Weak β-Kenmotsu Manifolds and η-Ricci Solitons. In: Rovenski, V. et al. (Eds) Differential Geometric Structures and Appl., Springer Proc. Math. and Statistics 440, Springer, Cham 2024, pp 53–72.
  6. [6]Rovenski V. and Walczak P., Extrinsic Geometry of Foliations, Progress in Mathematics 339, Birkhäuser, Cham 2021.DOI
  7. [7]Rovenski V. and Wolak R., New Metric Structures on g-Foliations, Indag. Math. 33 (2022) 518–532.DOI
  8. [8]Rovenski V., Generalized Ricci Solitons and Einstein Metrics on Weak K-Contact Manifolds. CAM, 15 (2023) 177–188.arxiv.org/abs/2301.03500
  9. [9]Suda N., Ricci Flow of Discrete Surfaces of Revolution, and Relation to Constant Gaussian Curvature, Diff. Geometry and its Appl. 98 (2025) 102221.arxiv.org/abs/2312.08113
  10. [10]Zergänge N., Stability Analysis of Isometric Embeddings, Preprint 2017, arXiv:1712.02766 [math.DG].

Paper details

vlnc:W_cc7pkhdn
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 6

downloads · 0

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

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents