---
title: 'Sphericalization: Methods, Models & Metrics'
url: https://www.emergentmind.com/topics/sphericalization
type: topic
---

# Sphericalization: Methods, Models & Metrics

Sphericalization denotes several distinct constructions whose common effect is to impose spherical structure on an object, a metric, or a formal system. In astrophysics it can mean the rounding of a triaxial dark-matter halo by a central disk; in stellar-halo tomography it can mean angular randomization at fixed radius; in metric geometry it can mean a conformal deformation that adds a point at infinity and yields a bounded sphere-like compactification; in geometry processing it can mean a map from a surface or point-cloud model to \(S^2\); in atomistic simulation it can mean replacing translational periodicity by infinitesimal rotational periodicity; and in higher category theory it can mean forcing the twist and cotwist of an adjunction to become equivalences [1006.0537; 2204.08425; 2009.13706; 1810.09031; 1010.0067; 2605.15037]. This suggests a unifying but deliberately broad usage: sphericalization is not a single method, but a family of operations that either make an object more spherical, represent it on a sphere, or freely adjoin spherical structure.

## 1. Semantic range and recurrent structural pattern

Across the cited literature, sphericalization appears in five recurrent modalities. One modality is **physical reshaping**: a system that is initially triaxial or anisotropic becomes more axisymmetric or spherical under an added influence. A second is **angular erasure at fixed radius**: radial information is retained while angular information is randomized, so that non-spherical shape and substructure are removed. A third is **conformal compactification**: an unbounded metric or quasimetric space is transformed into a bounded space by damping distances and adjoining a point at infinity. A fourth is **sphere-domain parameterization**: a surface, mesh, or data-derived simplicial complex is mapped to \(S^2\), often with explicit control of angle, area, or homotopy class. A fifth is **formal spherical completion** in higher algebra: an adjunction is modified until its twist and cotwist become equivalences [2204.08425; 2204.01920; 1810.09031; 2209.02791; 2605.15037].

The recurrence of the word “sphere” should not obscure the fact that the preserved invariants differ sharply by field. In some settings the key invariant is the radial profile \(N(<r)\); in others it is Ahlfors regularity, doubling, a \(p\)-Poincaré inequality, Besov energy, or the homotopy class of a map into \(S^2\). In the categorical setting the target invariant is instead the invertibility of twist and cotwist endofunctors. The term therefore functions as a family resemblance rather than a single definition.

## 2. Gravitational and stellar-halo uses

In galactic dynamics, sphericalization refers to the halo becoming more axisymmetric or spherical under the gravitational influence of a growing central disk galaxy. The dark halo shape is quantified by fitting ellipsoids to equipotential surfaces rather than to isodensity contours, with principal axes \(a \ge b \ge c\), axis ratios \(b/a\) and \(c/a\), and triaxiality
\[
T=\frac{a^2-b^2}{a^2-c^2}.
\]
This emphasis on potential-axis ratios is deliberate: potential shapes are typically more axisymmetric than density shapes because the potential is a smoothed integral of the mass distribution and is less sensitive to local substructure or transient density inhomogeneities. In the simulation suite, triaxial halos are built via successive mergers of Hernquist models; the final triaxial halo has \(300{,}000\) particles; disks are rigid exponential potentials with \(R_d \in \{2.8, 8.0, 22.3\}\,\mathrm{kpc}\); and the main control parameter is the peak disk fraction
\[
\eta \equiv \left.\frac{V_{\rm disk}(r)}{V_{\rm tot}(r)}\right|_{r=2.2R_d},
\qquad
V_{\rm tot}^2(r)=V_{\rm disk}^2(r)+V_{\rm halo}^2(r).
\]
The transition to noticeable sphericalization occurs around \(\eta \approx 0.5\): exponential disks with \(f_{\rm disk,peak}\lesssim 0.5\) do not noticeably modify the halo potential shape, whereas realistic HSB disks with \(\eta \approx 0.7\) can drive inner 2D potential axis ratios from \(\sim 0.6\) to \(\sim 0.8\), with effects extending to \(\sim 30\)–\(50\,\mathrm{kpc}\). Orientation and assembly timescale are secondary: the effect is maximal when the disk symmetry axis is aligned with the halo major axis, minimal when aligned with the halo minor axis, and nearly insensitive to whether disk growth is instantaneous, linear over \(1\) or \(10\,\mathrm{Gyr}\), or stepped over \(1\,\mathrm{Gyr}\) [1006.0537].

The same paper treats this as an observational discriminator. Dwarfs and LSB systems with \(\eta \lesssim 0.5\) are expected to retain triaxial potential shapes, so gaseous disks in such systems should exhibit noncircular motions, kinematic twists, and bar-like harmonics without an obvious stellar bar. NGC 2976 is presented as a threshold case, with \(R_d \approx 0.7\,\mathrm{kpc}\) and \(\eta \approx 0.5\), whose complex central velocity field is consistent with bar-like kinematics induced by triaxial halo forcing rather than by a visible optical bar [1006.0537].

A different astrophysical usage appears in stellar-halo tomography. There sphericalization is the operation of randomizing the angular coordinates of all stellar particles while preserving their radial distances from the halo center. For each particle at \((r,\theta,\phi)\), one draws
\[
\phi' \sim U(0,2\pi), \qquad \mu'=\cos\theta' \sim U(-1,1),
\]
sets \(\theta'=\arccos \mu'\), and replaces \((\theta,\phi)\) by \((\theta',\phi')\). The procedure keeps the original radial profile \(N(<r)\) intact, but removes anisotropy arising from small-scale substructures and from large-scale non-spherical shape. The residual anisotropy after sphericalization is then the Poisson baseline. In the Bullock–Johnston halos, whole-sky anisotropy increases with radius and plateaus, all halos are very smooth within \(\sim 50\,\mathrm{kpc}\), and comparison with sphericalized realizations shows that the approximately uniform component of anisotropy originates from discreteness noise and non-spherical shape, whereas the fluctuating spikes originate from substructure [2204.08425].

## 3. Metric and quasimetric constructions

In metric geometry, sphericalization classically converts Euclidean space into a compact sphere by equipping the one-point compactification \(\mathbf{R}^n \cup \{\infty\}\) with the chordal metric
\[
q(x,y)=\frac{|x-y|}{\sqrt{(1+|x|^2)(1+|y|^2)}},
\qquad
q(x,\infty)=\frac{1}{\sqrt{1+|x|^2}}.
\]
In general metric spaces \((X,d)\), fixing a basepoint \(a\in X\), one replaces \(d\) by the damped quasimetric
\[
d_a(x,y)=\frac{d(x,y)}{(1+d(x,a))(1+d(y,a))},\qquad
d_a(x,\infty)=\frac{1}{1+d(x,a)},
\]
and then passes to the chain metric
\[
\widehat d_a(x,y)=\inf \sum_{j=0}^{k} d_a(x_j,x_{j+1}),
\]
where the infimum is over finite chains joining \(x\) to \(y\). The result is a bounded space with an added point at infinity, and in Gromov hyperbolic applications it allows bounded-space arguments to be transferred back to unbounded domains [2009.13706].

A boundary-distance variant replaces damping relative to a basepoint by damping relative to distance from the metric boundary. If \((\Omega,d)\) is locally compact, rectifiably path-connected, non-complete, and unbounded, with \(d_\Omega(x)=\operatorname{dist}_d(x,\partial\Omega)\), and if \(\phi:(0,\infty)\to(0,\infty)\) is continuous, monotone decreasing, reverse doubling, equal to \(1\) on \((0,1]\), and integrable at infinity, then one sets
\[
w(x)=\phi(d_\Omega(x)),\qquad
\ell_\phi(\gamma)=\int_\gamma \phi(d_\Omega(\gamma(t)))\,dt,
\qquad
d_\phi(x,y)=\inf_\gamma \ell_\phi(\gamma).
\]
This deformation agrees with \(d\) on the collar \(\Omega_0=\{x\in\Omega:d_\Omega(x)\le 1\}\), adds exactly one new point \(\infty\) to the completion, keeps \(d\) and \(d_\phi\) locally bi-Lipschitz near \(\partial\Omega\), and preserves the uniform domain property when \(\Omega\) is uniform with respect to its completion [2204.01920].

In quasimetric measure spaces the same idea is paired with a dual operation, flattening. For an unbounded \(K\)-quasimetric space \((X,\rho)\) and \(a\in X\), sphericalization is defined on \(X\cup\{\infty\}\) by
\[
\rho_a(x,y)=\frac{\rho(x,y)}{(1+\rho(x,a))(1+\rho(y,a))},\qquad
\rho_a(x,\infty)=\frac{1}{1+\rho(x,a)}.
\]
For a bounded \(K\)-quasimetric space \((X,\rho)\) and \(c\in X\), flattening is defined on \(X\setminus\{c\}\) by
\[
\rho_c(x,y)=\frac{\rho(x,y)}{\rho(x,c)\rho(y,c)}.
\]
These operations preserve Ahlfors \(Q\)-regularity and doubling under the hypotheses stated in the paper, and flattening followed by sphericalization is bilipschitz equivalent to the original space [1911.01760].

A related uniformization program treats sphericalization and inversion as compatible with conformal deformations of Gromov hyperbolic spaces. If \(X\) is proper geodesic \(\delta\)-hyperbolic and \(\rho_{\lambda,b}(x)=e^{-\lambda b(x)}\) for a distance function or Busemann function \(b\), then the conformal deformation
\[
d_{\lambda,b}(x,y)=\inf_\gamma \int_\gamma e^{-\lambda b(\gamma(s))}\,ds
\]
produces uniform metric models \(X_{\lambda,b}\). In this setting, sphericalization and inversion commute with uniformization up to \(\partial\)-biLipschitz and quasimöbius equivalence via the identity, which places bounded and unbounded models in a common formal framework [2008.06806].

## 4. Analytic invariants, sharp conditions, and boundary value problems

One major branch of the literature asks which analytic structures survive sphericalization. For weighted measures of the form
\[
d\widehat\mu(x)=\frac{d\mu(x)}{(1+|x|)^q},
\]
with \(|x|=d(x,a)\), sphericalization preserves \(p\)-harmonic functions and Poincaré inequalities under assumptions far weaker than Ahlfors regularity. In this setting the sphericalized metric has diameter \(1\), the transformed measure is broader than the Ahlfors-regular case treated earlier, and the preservation theory is formulated for doubling measures with quantitative lower-growth control at the basepoint. The paper emphasizes that some consequences are new even for unweighted \(\mathbf{R}^n\), \(n\ge 2\), and \(p\ne 2\) [2508.09795].

A sharp formulation for preserving uniformity, doubling, and support of a \(p\)-Poincaré inequality uses a radial density \(\rho(|x|)\) and the path metric
\[
d_\rho(x,y)=\inf_{\gamma\in\Gamma(x,y)}\int_\gamma \rho(|\cdot|)\,ds,
\qquad
\mu_\rho(A)=\int_A \rho(|x|)^\sigma\,d\mu(x).
\]
The paper isolates three conditions on \(\rho\):
\[
\rho(r)\le C_A\rho(s)\quad\text{whenever }0<r\le 2s+1\text{ and }0<s\le 2r+1,
\]
\[
\int_r^\infty \rho(t)\,dt \le C_B (r+1)\rho(r)\quad\text{for all }r>0,
\]
\[
\int_{X\setminus B(b,r)} \rho(|x|)^\sigma\,d\mu(x)\le C_C \rho(r)^\sigma \mu(B(b,r+1))\quad\text{for all }r>0.
\]
Under these sharp hypotheses, \((X,d_\rho)\) is bounded, uniform, and has \(\partial_{d_\rho}X=\partial_dX\cup\{\infty\}\); \(\mu_\rho\) is doubling; and, if \(\rho\) is lower semicontinuous, \((X,d_\rho,\mu_\rho)\) supports a \(p\)-Poincaré inequality. The same paper proves sharpness by constructing half-plane counterexamples showing that failure of the first two conditions destroys uniformity, while failure of the third destroys doubling [2501.01348].

A parallel development targets fractional smoothness. Let \(Z\) be complete, \(\nu\) doubling, and
\[
\widehat d(x,y):=\inf \sum_{j=1}^n \bigl(\rho(x_j)+\rho(x_{j-1})\bigr)d(x_j,x_{j-1}),
\qquad
d\widehat\nu=\rho\,d\nu,
\]
with
\[
\rho(x)=\rho(|x|)=\frac{1}{\bigl(m(x)\nu(B_{m(x)})\bigr)^{1/\sigma}},
\qquad
m(t)=t+m_0,\quad m_0\in\{0,1\}.
\]
For \(\sigma=p\theta\), sphericalization and flattening preserve the Besov energy
\[
[u]_{\theta,p}
=
\int_Z\int_Z
\frac{|u(x)-u(y)|^p}{d(x,y)^{\theta p}\nu(B(x,d(x,y)))}\,d\nu(y)\,d\nu(x),
\]
and the corresponding transformed measure remains doubling. The construction is designed to apply even to totally disconnected fractal-type sets, where \(\theta>1\) is meaningful [2409.17809].

For nonlinear potential theory, sphericalization is used to transfer Dirichlet problems on unbounded uniform domains with unbounded boundary to bounded uniform domains. The deformation is
\[
d_\rho(x,y)=\inf_{\gamma\in\Gamma(x,y)}\int_\gamma \rho(|\cdot|)\,ds,
\qquad
\mu_\rho(E)=\int_E \rho(|x|)^p\,d\mu(x),
\qquad
\nu_\rho(E)=\int_E \rho(|x|)^{p-\theta}\,d\nu(x),
\]
and the upper gradients satisfy
\[
\widehat g_u=\frac{g_u}{\rho},
\qquad
\int_\Omega g_u^p\,d\mu=\int_\Omega \widehat g_u^{\,p}\,d\mu_\rho.
\]
Together with trace and extension operators on Besov boundary spaces, this yields existence and uniqueness of \(p\)-harmonic solutions for every \(f\in HB^{1-\theta/p}_{p,p}(\partial\Omega,d,\nu)\), and it links uniqueness at infinity to \(p\)-parabolicity and \(p\)-hyperbolicity via the capacity of the added point \(\infty\) [2602.15701].

An earlier Ahlfors-regular version uses the chain metric \(d_s\) on \(X\cup\{\infty\}\) together with
\[
d\widehat\mu(x)=(1+d(x,a))^{-2p}\,d\mu(x).
\]
There the sphericalization preserves \(p\)-modulus, minimal \(p\)-weak upper gradients, and exact \(p\)-energy, so that \(p\)-harmonicity and superharmonicity are identical on \((\Omega,d,\mu)\) and on \((\Omega,d_s,\widehat\mu)\). This allows resolutivity, Kellogg-type theorems, barrier criteria, and local regularity at \(\infty\) to be transferred from bounded to unbounded domains, including situations with several “approach directions” toward infinity encoded by the Mazurkiewicz boundary [1810.01580].

In Gromov hyperbolic geometry, sphericalization also mediates between boundary gauges. One application shows that the doubling property coincides for Bourdon metrics on \(\partial_\infty X\) and Hamenstädt metrics on \(\partial_\infty X\setminus\{\xi\}\), and another characterizes unbounded Gromov hyperbolic domains by the Gehring–Hayman and ball separation conditions after passing to a sphericalized bounded model [2009.13706].

## 5. Sphere-domain parameterization and sphere-valued coordinates

In geometry processing, sphericalization can mean parameterizing a closed genus-\(0\) triangle mesh \(M=(V,E,F)\) onto the unit sphere \(S^2\) by a bijective map \(f:M\to S^2\). The framework in question distinguishes conformal, area-preserving, and balanced sphericalizations. The conformal branch is based on dynamic Yamabe flow and conformal welding, with discrete conformal factors \(u_i\), edge update
\[
l_{ij}(u)=e^{(u_i+u_j)/2}l^0_{ij},
\]
and discrete Gaussian curvature
\[
K_i(u)=
\begin{cases}
2\pi-\sum \theta_i^{jk}(u), & \text{interior vertex},\\[2mm]
\pi-\sum \theta_i^{jk}(u), & \text{boundary vertex}.
\end{cases}
\]
Area-preserving sphericalization is obtained through discrete optimal mass transport after stereographic projection, and balanced sphericalization is produced by polar factorization via the measure interpolation
\[
\mu_t=(1-t)\,\tau_\# dA_h+t\,(\tau\circ\phi)_\# dA_g,\qquad 0\le t\le 1.
\]
The resulting family interpolates between angle-preserving and area-preserving maps on \(S^2\), with the total area normalized to \(4\pi\) [1810.09031].

For point clouds and general data, sphericalization can mean the construction of sphere-valued coordinates from second persistent cohomology. One builds a filtered Vietoris–Rips complex
\[
\mathrm{VR}_\varepsilon(X)=\{\sigma\subseteq X:\operatorname{diam}(\sigma)\le \varepsilon\},
\]
extracts a persistent class \([\alpha_p]\in H^2(\mathrm{VR}_\varepsilon(X);F_p)\), lifts it to an integral cocycle, and constructs a map \(\alpha:\mathrm{sk}_2(\mathrm{VR}_\varepsilon(X))\to S^2\) that extends to \(\mathrm{sk}_3\). The initial map is then smoothed by minimizing a simplicial energy, either harmonic
\[
\mathcal E_H(f)=\sum_{x\in Y^2}\frac12\,\|A(f(x))\|^2,
\]
or spring-type
\[
\mathcal E_S(f)=\sum_{x\in Y^2}\frac12\,\|k(A(f(x))-R)\|^2,
\]
subject to \(f(v)\in S^2\), zero center of mass, and preservation of the homotopy class of the original cocycle-derived map. The minimizer is unique up to rigid motion, and the method is explicitly positioned as a generalization of circular coordinates from \(PH^1\) to spherical coordinates from \(PH^2\) [2209.02791].

A third usage appears in morphological decomposition by spherical harmonics. For a closed, star-shaped surface with vertices \(p=(x,y,z)\), “sphericalization” maps to the unit sphere by radial rescaling,
\[
(x,y,z)\mapsto (x/r,y/r,z/r),\qquad r=\sqrt{x^2+y^2+z^2},
\]
and then expands a scalar field on the sphere as
\[
f(\theta,\phi)=\sum_{l=0}^{L}\sum_{m=-l}^{l} a_{lm}Y_l^m(\theta,\phi).
\]
The paper shows that this SH-based sphericalization degenerates for oblate and prolate particles because the radial map induces nonuniform sampling density and local angle and area distortions; oscillations appear and the spectral descriptor tail fails to decay. The proposed remedy is to replace the unit sphere by a spheroidal domain with per-particle parameters, leading to rSOH, hSOH, and c-rSOH constructions. In that sense, SOH generalizes sphericalization by allowing the analysis domain to be a tuned spheroid rather than \(S^2\) [2407.03350].

## 6. Local spherical models in membrane simulation

In atomistic and mesoscale simulation, sphericalization can mean modeling a locally spherical membrane patch using revised periodic boundary conditions built from small rotations instead of translations. The generators are
\[
S_1=R(\delta\theta_1,\hat u_1),\qquad
S_2=R(\delta\theta_2,\hat u_2),
\]
and for small \(\delta\theta_i\),
\[
S_1^{n_1}S_2^{n_2}\approx R(n_1\delta\theta_1\hat u_1+n_2\delta\theta_2\hat u_2).
\]
A minimal simulation cell is placed near a point on a notional sphere, and periodic images are created by these infinitesimal rotations. Curvature is imposed statistically through the boundary symmetry rather than by a global constraint on the radial coordinate, with \(\delta\theta_i \approx L_i/R\) linking angular steps to a target radius \(R\). Distances are Euclidean chord lengths \(|r_i-S^n r_j|\), whose difference from geodesic arc length is \(O(\theta^3)\), with relative error \(\sim \theta^2/24\) [1010.0067].

The method is accurate in the locally flat regime
\[
R\gg L,\qquad R\gg r_c,\qquad R\gg t,
\]
and admits both classical and quantum implementations. For classical pair potentials,
\[
E_{\rm pair}=\frac12\sum_{i,j}\sum_n U_{ij}(|r_i-S^n r_j|).
\]
For membranes the elastic interpretation is framed by the Helfrich free energy,
\[
F=\int_{\mathcal S}\left[\frac{\kappa}{2}(2H-c_0)^2+\kappa_G K\right]\,dA+\sigma\int_{\mathcal S} dA,
\]
with spherical energy density
\[
g=\frac{2\kappa+\kappa_G}{R^2}.
\]
The paper demonstrates the approach on single- and multilayer graphene, obtaining \(\kappa \approx 1.61\,\mathrm{eV}\) and \(\kappa_G \approx -0.70\,\mathrm{eV}\) for monolayer graphene, while reducing the active system to as little as a 2-atom unit cell [1010.0067].

## 7. Higher-categorical sphericalization and terminological distinctions

In stable \(\infty\)-category theory, sphericalization is neither a compactification nor a geometric rounding. Given an adjunction \(L\dashv R\) in a locally stable \((\infty,2)\)-category, one defines the twist and cotwist by exact triangles
\[
\mathrm{id}_C \to RL \to T_C \to \Sigma\,\mathrm{id}_C,
\qquad
T_D \to LR \to \mathrm{id}_D \to \Sigma T_D.
\]
An adjunction is spherical if both \(T_C\) and \(T_D\) are equivalences. The sphericalization procedure takes any adjunction and freely forces these endofunctors to become equivalences by iterating the associated endomorphism of adjunctions and taking a sequential colimit or, dually, a sequential limit:
\[
S(L,R)=\operatorname*{colim}_{n\in\mathbf N}\Gamma(L,R),
\qquad
L(L,R)=\operatorname*{lim}_{n\in\mathbf N^{\mathrm{op}}}\Gamma^\vee(L,R).
\]
This yields functors
\[
S\dashv \iota \dashv L
\]
from all adjunctions to spherical adjunctions, a walking spherical adjunction that corepresents spherical adjunctions, and a Fourier-transform autoequivalence sending \(L\dashv R\) to \(T_C^{-1}R\dashv L\) [2605.15037].

A common terminological confusion arises with **spherical localization**, which is a different construction. Spherical localization decomposes \(S^n\) into geodesic needles carrying densities of the form
\[
h(t)=C\,\sin(t+t_0)^{\,n-k},
\]
so that global integral inequalities reduce to one-dimensional weighted inequalities along those needles. It is a technique in convex and metric geometry, not a sphericalization in the compactification, parameterization, or adjunction-theoretic senses discussed above [1507.00915].

The multiplicity of meanings is therefore substantive rather than accidental. In some literatures sphericalization removes anisotropy; in others it compactifies infinity, preserves energy under a conformal deformation, builds coordinates on \(S^2\), realizes local spherical symmetry in a periodic simulation, or freely adjoins sphericality to an adjunction. The term is unified less by a single construction than by a stable formal idea: replacing a problem by one in which spherical structure is explicit, enforced, or analytically advantageous.

Source: https://www.emergentmind.com/topics/sphericalization