OctoInformBeta

OpenAI preprint, September 2026 · §2. Volume-normalized Berger links

Lemma 2.1: Cartesian tensor identities and the Laplacian of a homogeneous function with logarithmic radial profile

Volume and curvature of radial Berger metrics.

Hover over a blue or green term in the text, or a node in the graph, to find it in the other. Select it to read its declaration both in Lean and in words.

The Setting

A radial Berger metric distinguishes the radial direction from the Hopf direction and its orthogonal complement in each sphere. Its horizontal scale and Hopf stretching may vary with logarithmic radius.

Work in Euclidean space ℝn\mathbb R^n, and put m=n−1m=n-1 for the link dimension. Write Mn(ℝ)M_n(\mathbb R) for the real matrix space and II for its identity matrix. An orthogonal complex structure J∈Mn(ℝ)J\in M_n(\mathbb R) is a matrix satisfying J2=−IJ^2=-I and JTJ=IJ^T J=I.✦ For x≠0x\ne0, the squared radius SS, radius rr, logarithmic radius tt, unit radial vector ere_r and unit Hopf vector eHe_H are S=x⋅x,r=S,t=12log⁡S=log⁡r,er=xr,eH=Jxr.S=x\cdot x,\qquad r=\sqrt S,\qquad t=\tfrac12\log S=\log r, \qquad e_r=\frac{x}{r},\qquad e_H=\frac{Jx}{r}.(1)

For vectors η,ξ∈ℝn\eta,\xi\in\mathbb R^n and real scalars c,qc,q, the Berger tensor T(η,ξ;c,q)T(\eta,\xi;c,q) is the matrix in (2); for real A,DA,D and x≠0x\ne0, the polar tensor 𝒫J(x;A,D)\mathcal P_J(x;A,D) is the matrix in the same display, formed from the radial vectors of (1). Here z⊗wz\otimes w denotes the matrix with entries ziwjz_iw_j. T(η,ξ;c,q)=cI+(1−c)η⊗η+c(q−1)ξ⊗ξ,𝒫J(x;A,D)=AI+1−ASx⊗x+D−AS(Jx)⊗(Jx).\begin{aligned} T(\eta,\xi;c,q) &=cI+(1-c)\eta\otimes\eta+c(q-1)\xi\otimes\xi,\\ \mathcal P_J(x;A,D) &=AI+\frac{1-A}{S}x\otimes x+\frac{D-A}{S}(Jx)\otimes(Jx). \end{aligned}(2) For m≥1m\ge1 and a,q>0a,q>0, the normalized horizontal factor cc is the positive real scalar given by (3): c=a2q−1/m.c=a^2q^{-1/m}.(3) The Berger link metric GSG_S is the restriction to the unit sphere of T(er,eH;c,q)T(e_r,e_H;c,q) from (2), with cc as in (3); g0g_0 denotes the round metric.

Cartesian Differential Operators

For a differentiable real function FF on ℝn\mathbb R^n, its coordinate derivative ∂iF(x)\partial_iF(x) is DF(x)[ei]DF(x)[e_i], where eie_i is the iith coordinate vector.✦ For a real function FF on ℝn\mathbb R^n, the Cartesian gradient ∇F\nabla F, Hessian D2FD^2F, Euclidean Laplacian Δ0F\Delta_0F, Euler derivative EFEF and Hopf derivative DJFD_JF are respectively the vector field, matrix field and scalar functions given by (4). ∇F=(∂iF)i,D2F=(∂i∂jF)ij,Δ0F=tr⁡(D2F),EF=x⋅∇F,DJF=(Jx)⋅∇F.\begin{gathered} \nabla F=(\partial_iF)_i,\qquad D^2F=(\partial_i\partial_jF)_{ij},\qquad \Delta_0F=\operatorname{tr}(D^2F),\\ EF=x\cdot\nabla F,\qquad D_JF=(Jx)\cdot\nabla F. \end{gathered}(4) For an integer ℓ≥0\ell\ge0, VℓV_\ell is the space of homogeneous Euclidean harmonic polynomials of degree ℓ\ell, and Bℓ=ℓ(ℓ+m−1)B_\ell=\ell(\ell+m-1).

A smooth Riemannian metric gg on ℝn\mathbb R^n is a smooth matrix function G:ℝn→Mn(ℝ)G:\mathbb R^n\to M_n(\mathbb R) whose value is positive definite at every point.✦ Write GijG^{ij} for the entries of G−1G^{-1}. For a smooth Riemannian metric gg on ℝn\mathbb R^n with matrix GG, the Christoffel symbols Γijk\Gamma^k_{ij} are its connection coefficients, the Ricci tensor Ric⁡g\operatorname{Ric}_g is its curvature contraction, and the Laplace-Beltrami operator Δg\Delta_g is its scalar differential operator given by (5), with the curvature convention for which the round sphere has positive Ricci curvature. Γijk=12∑bGkb(∂iGjb+∂jGib−∂bGij),(Ric⁡g)ij=∑k(∂kΓijk−∂jΓikk)+∑k,b(ΓijkΓkbb−ΓikbΓjbk),ΔgF=∑i,jGij(∂i∂jF−∑kΓijk∂kF).\begin{aligned} \Gamma^k_{ij} &=\tfrac12\sum_bG^{kb} (\partial_iG_{jb}+\partial_jG_{ib}-\partial_bG_{ij}),\\ (\operatorname{Ric}_g)_{ij} &=\sum_k(\partial_k\Gamma^k_{ij}-\partial_j\Gamma^k_{ik}) +\sum_{k,b}(\Gamma^k_{ij}\Gamma^b_{kb}-\Gamma^b_{ik}\Gamma^k_{jb}),\\ \Delta_gF &=\sum_{i,j}G^{ij} \left(\partial_i\partial_jF-\sum_k\Gamma^k_{ij}\partial_kF\right). \end{aligned}(5)

Connection and Curvature Coefficients

For real coefficients α,β,γ,δ,ε,ζ\alpha,\beta,\gamma,\delta,\varepsilon,\zeta, the six-term connection expression 𝒦ijk\mathcal K^k_{ij} is defined by (6), where w=Jxw=Jx and δij\delta_{ij} is the Kronecker symbol. 𝒦ijk=α(xiδkj+xjδki)+βxkδij+γxkxixj+δwk(xiwj+xjwi)+εxkwiwj+ζ(wjJki+wiJkj).\begin{aligned} \mathcal K^k_{ij} ={}&\alpha(x_i\delta_{kj}+x_j\delta_{ki})+\beta x_k\delta_{ij} +\gamma x_kx_ix_j\\ &+\delta w_k(x_iw_j+x_jw_i)+\varepsilon x_kw_iw_j +\zeta(w_jJ_{ki}+w_iJ_{kj}). \end{aligned}(6)✦ For real parameters A,q,u,vA,q,u,v, the normalized connection expression 𝒞ijk(J,x;A,q,u,v)\mathcal C^k_{ij}(J,x;A,q,u,v) is (6) with the coefficients in (7). The normalized connection coefficients α,β,γ,δ,ε,ζ\alpha,\beta,\gamma,\delta,\varepsilon,\zeta are the scalar functions of squared radius SS, horizontal scale AA, anisotropy qq and logarithmic rates u,vu,v given in (7). α=uS,β=1−A(1+u)S,γ=A(1+u)−1−2uS2,δ=1−q+vS2,ε=A(1+u)−Aq(1+u+v)+2(q−1)S2,ζ=q−1S.\begin{aligned} \alpha&=\frac{u}{S},& \beta&=\frac{1-A(1+u)}{S},& \gamma&=\frac{A(1+u)-1-2u}{S^2},\\ \delta&=\frac{1-q+v}{S^2},& \varepsilon&=\frac{A(1+u)-Aq(1+u+v)+2(q-1)}{S^2},& \zeta&=\frac{q-1}{S}. \end{aligned}(7)

For nonvanishing differentiable profiles A,q:ℝ→ℝA,q:\mathbb R\to\mathbb R, their logarithmic rates u,v:ℝ→ℝu,v:\mathbb R\to\mathbb R satisfy A′=2Au,q′=2qv.A'=2Au,\qquad q'=2qv.(8) The horizontal, radial and Hopf Ricci numerators ℋ\mathcal H, ℛ\mathcal R and 𝒱\mathcal V are the scalar functions of the profile values and logarithmic rates given in (9); dots denote scalar arguments for the derivatives of uu and vv with respect to tt. ℋ(A,q,u,v,u̇)=n−2q−A[u̇+(m−1+mu+v)(1+u)],ℛ(u,v,u̇,v̇)=−[m(u+u̇+u2)+v+v̇+2uv+v2],𝒱(A,q,u,v,u̇,v̇)=(m−1)q2−Aq[u̇+v̇+(m−1+mu+v)(1+u+v)].\begin{aligned} \mathcal H(A,q,u,v,\dot u) &=n-2q-A\bigl[\dot u+(m-1+mu+v)(1+u)\bigr],\\ \mathcal R(u,v,\dot u,\dot v) &=-\bigl[m(u+\dot u+u^2)+v+\dot v+2uv+v^2\bigr],\\ \mathcal V(A,q,u,v,\dot u,\dot v) &=(m-1)q^2 -Aq\bigl[\dot u+\dot v+(m-1+mu+v)(1+u+v)\bigr]. \end{aligned}(9)

The Claim

Theorem 1. Let m≥1m\ge1, let η,ξ∈ℝm+1\eta,\xi\in\mathbb R^{m+1} be orthogonal unit vectors, and let a,q>0a,q>0. With c=a2q−1/mc=a^2q^{-1/m}, the Cartesian volume density is det⁡T(η,ξ;c,q)=am.\sqrt{\det T(\eta,\xi;c,q)}=a^m.(10)✦

Let gg be a smooth Riemannian metric on ℝn\mathbb R^n, let JJ be an orthogonal complex structure, and fix x≠0x\ne0. Suppose A,qA,q are nonvanishing differentiable profiles with rates u,vu,v satisfying (8), that u,vu,v are differentiable at t=log⁡|x|t=\log|x|, and that, throughout a neighborhood of xx, G(y)=T(er(y),eH(y);A(log⁡|y|),q(log⁡|y|))G(y)=T(e_r(y),e_H(y);A(\log|y|),q(\log|y|)). Let H,R,VH,R,V be the numerators in (9) evaluated along A,q,u,vA,q,u,v at tt. Then Ric⁡g(x)=HSI+R−HS2x⊗x+V−HS2(Jx)⊗(Jx).\operatorname{Ric}_g(x) =\frac{H}{S}I+\frac{R-H}{S^2}x\otimes x +\frac{V-H}{S^2}(Jx)\otimes(Jx).(11)✦

Finally, let f:ℝ→ℝf:\mathbb R\to\mathbb R be twice differentiable and let P:ℝn→ℝP:\mathbb R^n\to\mathbb R be smooth, with EP=ℓPEP=\ell P and Δ0P=0\Delta_0P=0 everywhere for some ℓ∈ℝ\ell\in\mathbb R. Fix nonzero real scalars A,qA,q and real u,vu,v. Suppose at x≠0x\ne0 that G(x)−1=𝒫J(x;A−1,(Aq)−1)G(x)^{-1}=\mathcal P_J(x;A^{-1},(Aq)^{-1}) and that every Christoffel symbol equals 𝒞ijk(J,x;A,q,u,v)\mathcal C^k_{ij}(J,x;A,q,u,v). For t=log⁡|x|t=\log|x|, the separated Laplacian satisfies SΔg(f(log⁡|⋅|)P)(x)=(f″(t)+(2ℓ+m−1+mu+v)f′(t)+[ℓ2+(m−1+mu+v)ℓ−A−1ℓ(ℓ+m−1)]f(t))P(x)+((Aq)−1−A−1)f(t)DJ2P(x).\begin{aligned} S\Delta_g\bigl(f(\log|\cdot|)P\bigr)(x) ={}&\Bigl(f''(t)+(2\ell+m-1+mu+v)f'(t)\\ &\quad+\bigl[\ell^2+(m-1+mu+v)\ell -A^{-1}\ell(\ell+m-1)\bigr]f(t)\Bigr)P(x)\\ &+\bigl((Aq)^{-1}-A^{-1}\bigr)f(t)D_J^2P(x). \end{aligned}(12)✦

In the paper, the result reads: under the preceding hypotheses, the volume form and horizontal and vertical Ricci endomorphism eigenvalues ρH,ρV\rho_H,\rho_V of GSG_S satisfy (13), including the strict Ricci inequality criterion, and every VℓV_\ell is preserved by the positive angular Laplacian with restriction Aℓ(a,q,J)A_\ell(a,q,J) as given there: dvolGS=amdvolg0,ρH=c−1[(m−1)−2(q−1)],ρV=c−1(m−1)q,Ric⁡GS>(m−1)GS⇔a2<q1/mmin⁡{1−2(q−1)m−1,q},Aℓ(a,q,J)=c−1[BℓI+(1−q−1)DJ2|Vℓ].\begin{gathered} \mathrm{dvol}_{G_S}=a^m\mathrm{dvol}_{g_0},\qquad \rho_H=c^{-1}\bigl[(m-1)-2(q-1)\bigr],\qquad \rho_V=c^{-1}(m-1)q,\\ \operatorname{Ric}_{G_S}>(m-1)G_S \quad\Longleftrightarrow\quad a^2<q^{1/m}\min\left\{1-\frac{2(q-1)}{m-1},q\right\},\\ A_\ell(a,q,J) =c^{-1}\left[B_\ell I+(1-q^{-1})D_J^2\big|_{V_\ell}\right]. \end{gathered}(13)

What Is Checked

The density identity (10) is stated over orthogonal unit vectors and positive normalization parameters, the Ricci identity (11) over smooth Cartesian metrics with local radial Berger coefficients and differentiable logarithmic rates, and the Laplacian identity (12) over smooth Euler-homogeneous harmonic functions with twice differentiable radial profiles and the specified inverse metric and Christoffel symbols at a nonzero point.